projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
io_uring: always progress task_work on task cancel
[linux-2.6-block.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);
11215630 69static int ext4_mark_recovery_complete(struct super_block *sb,
2b2d6d01 70 struct ext4_super_block *es);
11215630
JK		 71static int 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:
c1e8d7c6 96 * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
e74031fd
JK		 97 * page lock -> i_data_sem (rw)
98 *
99 * buffered write path:
c1e8d7c6 100 * sb_start_write -> i_mutex -> mmap_lock
e74031fd
JK		 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:
c1e8d7c6 110 * sb_start_write -> i_mutex -> mmap_lock
1d39834f 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
fa491b14 144
145static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
146 bh_end_io_t *end_io)
147{
148 /*
149 * buffer's verified bit is no longer valid after reading from
150 * disk again due to write out error, clear it to make sure we
151 * recheck the buffer contents.
152 */
153 clear_buffer_verified(bh);
154
155 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
156 get_bh(bh);
157 submit_bh(REQ_OP_READ, op_flags, bh);
158}
159
160void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
161 bh_end_io_t *end_io)
162{
163 BUG_ON(!buffer_locked(bh));
164
165 if (ext4_buffer_uptodate(bh)) {
166 unlock_buffer(bh);
167 return;
168 }
169 __ext4_read_bh(bh, op_flags, end_io);
170}
171
172int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
173{
174 BUG_ON(!buffer_locked(bh));
175
176 if (ext4_buffer_uptodate(bh)) {
177 unlock_buffer(bh);
178 return 0;
179 }
180
181 __ext4_read_bh(bh, op_flags, end_io);
182
183 wait_on_buffer(bh);
184 if (buffer_uptodate(bh))
185 return 0;
186 return -EIO;
187}
188
189int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
190{
191 if (trylock_buffer(bh)) {
192 if (wait)
193 return ext4_read_bh(bh, op_flags, NULL);
194 ext4_read_bh_nowait(bh, op_flags, NULL);
195 return 0;
196 }
197 if (wait) {
198 wait_on_buffer(bh);
199 if (buffer_uptodate(bh))
200 return 0;
201 return -EIO;
202 }
203 return 0;
204}
205
fb265c9c 206/*
8394a6ab 207 * This works like __bread_gfp() except it uses ERR_PTR for error
fb265c9c
TT		 208 * returns. Currently with sb_bread it's impossible to distinguish
209 * between ENOMEM and EIO situations (since both result in a NULL
210 * return.
211 */
8394a6ab 212static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
213 sector_t block, int op_flags,
214 gfp_t gfp)
fb265c9c 215{
2d069c08 216 struct buffer_head *bh;
217 int ret;
fb265c9c 218
8394a6ab 219 bh = sb_getblk_gfp(sb, block, gfp);
fb265c9c
TT		 220 if (bh == NULL)
221 return ERR_PTR(-ENOMEM);
cf2834a5 222 if (ext4_buffer_uptodate(bh))
fb265c9c 223 return bh;
2d069c08 224
225 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
226 if (ret) {
227 put_bh(bh);
228 return ERR_PTR(ret);
229 }
230 return bh;
fb265c9c
TT		 231}
232
8394a6ab 233struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
234 int op_flags)
235{
236 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
237}
238
239struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
240 sector_t block)
241{
242 return __ext4_sb_bread_gfp(sb, block, 0, 0);
243}
244
5df1d412 245void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
246{
247 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
248
249 if (likely(bh)) {
250 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
251 brelse(bh);
252 }
fb265c9c
TT		 253}
254
d25425f8
DW		 255static int ext4_verify_csum_type(struct super_block *sb,
256 struct ext4_super_block *es)
257{
e2b911c5 258 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 259 return 1;
260
261 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
262}
263
a9c47317
DW		 264static __le32 ext4_superblock_csum(struct super_block *sb,
265 struct ext4_super_block *es)
266{
267 struct ext4_sb_info *sbi = EXT4_SB(sb);
268 int offset = offsetof(struct ext4_super_block, s_checksum);
269 __u32 csum;
270
271 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
272
273 return cpu_to_le32(csum);
274}
275
c197855e
SH		 276static int ext4_superblock_csum_verify(struct super_block *sb,
277 struct ext4_super_block *es)
a9c47317 278{
9aa5d32b 279 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 280 return 1;
281
282 return es->s_checksum == ext4_superblock_csum(sb, es);
283}
284
06db49e6 285void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 286{
06db49e6
TT		 287 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
288
9aa5d32b 289 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 290 return;
291
292 es->s_checksum = ext4_superblock_csum(sb, es);
293}
294
8fadc143
AR		 295ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
296 struct ext4_group_desc *bg)
bd81d8ee 297{
3a14589c 298 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 299 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 300 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 301}
302
8fadc143
AR		 303ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
304 struct ext4_group_desc *bg)
bd81d8ee 305{
5272f837 306 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 307 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 308 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 309}
310
8fadc143
AR		 311ext4_fsblk_t ext4_inode_table(struct super_block *sb,
312 struct ext4_group_desc *bg)
bd81d8ee 313{
5272f837 314 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 315 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 316 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 317}
318
021b65bb
TT		 319__u32 ext4_free_group_clusters(struct super_block *sb,
320 struct ext4_group_desc *bg)
560671a0
AK		 321{
322 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
323 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 324 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 325}
326
327__u32 ext4_free_inodes_count(struct super_block *sb,
328 struct ext4_group_desc *bg)
329{
330 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
331 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 332 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 333}
334
335__u32 ext4_used_dirs_count(struct super_block *sb,
336 struct ext4_group_desc *bg)
337{
338 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
339 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 340 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 341}
342
343__u32 ext4_itable_unused_count(struct super_block *sb,
344 struct ext4_group_desc *bg)
345{
346 return le16_to_cpu(bg->bg_itable_unused_lo) |
347 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 348 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 349}
350
8fadc143
AR		 351void ext4_block_bitmap_set(struct super_block *sb,
352 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 353{
3a14589c 354 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 355 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
356 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 357}
358
8fadc143
AR		 359void ext4_inode_bitmap_set(struct super_block *sb,
360 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 361{
5272f837 362 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 363 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
364 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 365}
366
8fadc143
AR		 367void ext4_inode_table_set(struct super_block *sb,
368 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 369{
5272f837 370 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 371 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
372 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 373}
374
021b65bb
TT		 375void ext4_free_group_clusters_set(struct super_block *sb,
376 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 377{
378 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
379 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
380 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
381}
382
383void ext4_free_inodes_set(struct super_block *sb,
384 struct ext4_group_desc *bg, __u32 count)
385{
386 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
387 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
388 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
389}
390
391void ext4_used_dirs_set(struct super_block *sb,
392 struct ext4_group_desc *bg, __u32 count)
393{
394 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
395 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
396 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
397}
398
399void ext4_itable_unused_set(struct super_block *sb,
400 struct ext4_group_desc *bg, __u32 count)
401{
402 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
403 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
404 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
405}
406
6a0678a7
AB		 407static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
408{
409 time64_t now = ktime_get_real_seconds();
410
411 now = clamp_val(now, 0, (1ull << 40) - 1);
412
413 *lo = cpu_to_le32(lower_32_bits(now));
414 *hi = upper_32_bits(now);
415}
416
417static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
418{
419 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
420}
421#define ext4_update_tstamp(es, tstamp) \
422 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
423#define ext4_get_tstamp(es, tstamp) \
424 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
d3d1faf6 425
54d3adbc
TT		 426static void __save_error_info(struct super_block *sb, int error,
427 __u32 ino, __u64 block,
428 const char *func, unsigned int line)
1c13d5c0
TT		 429{
430 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
54d3adbc 431 int err;
1c13d5c0
TT		 432
433 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 434 if (bdev_read_only(sb->s_bdev))
435 return;
1c13d5c0 436 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6a0678a7 437 ext4_update_tstamp(es, s_last_error_time);
1c13d5c0
TT		 438 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
439 es->s_last_error_line = cpu_to_le32(line);
54d3adbc
TT		 440 es->s_last_error_ino = cpu_to_le32(ino);
441 es->s_last_error_block = cpu_to_le64(block);
442 switch (error) {
443 case EIO:
444 err = EXT4_ERR_EIO;
445 break;
446 case ENOMEM:
447 err = EXT4_ERR_ENOMEM;
448 break;
449 case EFSBADCRC:
450 err = EXT4_ERR_EFSBADCRC;
451 break;
452 case 0:
453 case EFSCORRUPTED:
454 err = EXT4_ERR_EFSCORRUPTED;
455 break;
456 case ENOSPC:
457 err = EXT4_ERR_ENOSPC;
458 break;
459 case ENOKEY:
460 err = EXT4_ERR_ENOKEY;
461 break;
462 case EROFS:
463 err = EXT4_ERR_EROFS;
464 break;
465 case EFBIG:
466 err = EXT4_ERR_EFBIG;
467 break;
468 case EEXIST:
469 err = EXT4_ERR_EEXIST;
470 break;
471 case ERANGE:
472 err = EXT4_ERR_ERANGE;
473 break;
474 case EOVERFLOW:
475 err = EXT4_ERR_EOVERFLOW;
476 break;
477 case EBUSY:
478 err = EXT4_ERR_EBUSY;
479 break;
480 case ENOTDIR:
481 err = EXT4_ERR_ENOTDIR;
482 break;
483 case ENOTEMPTY:
484 err = EXT4_ERR_ENOTEMPTY;
485 break;
486 case ESHUTDOWN:
487 err = EXT4_ERR_ESHUTDOWN;
488 break;
489 case EFAULT:
490 err = EXT4_ERR_EFAULT;
491 break;
492 default:
493 err = EXT4_ERR_UNKNOWN;
494 }
495 es->s_last_error_errcode = err;
1c13d5c0
TT		 496 if (!es->s_first_error_time) {
497 es->s_first_error_time = es->s_last_error_time;
6a0678a7 498 es->s_first_error_time_hi = es->s_last_error_time_hi;
1c13d5c0
TT		 499 strncpy(es->s_first_error_func, func,
500 sizeof(es->s_first_error_func));
501 es->s_first_error_line = cpu_to_le32(line);
502 es->s_first_error_ino = es->s_last_error_ino;
503 es->s_first_error_block = es->s_last_error_block;
878520ac 504 es->s_first_error_errcode = es->s_last_error_errcode;
1c13d5c0 505 }
66e61a9e
TT		 506 /*
507 * Start the daily error reporting function if it hasn't been
508 * started already
509 */
510 if (!es->s_error_count)
511 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 512 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 513}
514
54d3adbc
TT		 515static void save_error_info(struct super_block *sb, int error,
516 __u32 ino, __u64 block,
517 const char *func, unsigned int line)
1c13d5c0 518{
54d3adbc 519 __save_error_info(sb, error, ino, block, func, line);
c96e2b85
ES		 520 if (!bdev_read_only(sb->s_bdev))
521 ext4_commit_super(sb, 1);
1c13d5c0
TT		 522}
523
bdfe0cbd
TT		 524/*
525 * The del_gendisk() function uninitializes the disk-specific data
526 * structures, including the bdi structure, without telling anyone
527 * else. Once this happens, any attempt to call mark_buffer_dirty()
528 * (for example, by ext4_commit_super), will cause a kernel OOPS.
529 * This is a kludge to prevent these oops until we can put in a proper
530 * hook in del_gendisk() to inform the VFS and file system layers.
531 */
532static int block_device_ejected(struct super_block *sb)
533{
534 struct inode *bd_inode = sb->s_bdev->bd_inode;
535 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
536
537 return bdi->dev == NULL;
538}
539
18aadd47
BJ		 540static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
541{
542 struct super_block *sb = journal->j_private;
543 struct ext4_sb_info *sbi = EXT4_SB(sb);
544 int error = is_journal_aborted(journal);
5d3ee208 545 struct ext4_journal_cb_entry *jce;
18aadd47 546
5d3ee208 547 BUG_ON(txn->t_state == T_FINISHED);
a0154344
DJ		 548
549 ext4_process_freed_data(sb, txn->t_tid);
550
18aadd47 551 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 552 while (!list_empty(&txn->t_private_list)) {
553 jce = list_entry(txn->t_private_list.next,
554 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 555 list_del_init(&jce->jce_list);
556 spin_unlock(&sbi->s_md_lock);
557 jce->jce_func(sb, jce, error);
558 spin_lock(&sbi->s_md_lock);
559 }
560 spin_unlock(&sbi->s_md_lock);
561}
1c13d5c0 562
afb585a9
MFO		 563/*
564 * This writepage callback for write_cache_pages()
565 * takes care of a few cases after page cleaning.
566 *
567 * write_cache_pages() already checks for dirty pages
568 * and calls clear_page_dirty_for_io(), which we want,
569 * to write protect the pages.
570 *
571 * However, we may have to redirty a page (see below.)
572 */
573static int ext4_journalled_writepage_callback(struct page *page,
574 struct writeback_control *wbc,
575 void *data)
576{
577 transaction_t *transaction = (transaction_t *) data;
578 struct buffer_head *bh, *head;
579 struct journal_head *jh;
580
581 bh = head = page_buffers(page);
582 do {
583 /*
584 * We have to redirty a page in these cases:
585 * 1) If buffer is dirty, it means the page was dirty because it
586 * contains a buffer that needs checkpointing. So the dirty bit
587 * needs to be preserved so that checkpointing writes the buffer
588 * properly.
589 * 2) If buffer is not part of the committing transaction
590 * (we may have just accidentally come across this buffer because
591 * inode range tracking is not exact) or if the currently running
592 * transaction already contains this buffer as well, dirty bit
593 * needs to be preserved so that the buffer gets writeprotected
594 * properly on running transaction's commit.
595 */
596 jh = bh2jh(bh);
597 if (buffer_dirty(bh) ||
598 (jh && (jh->b_transaction != transaction ||
599 jh->b_next_transaction))) {
600 redirty_page_for_writepage(wbc, page);
601 goto out;
602 }
603 } while ((bh = bh->b_this_page) != head);
604
605out:
606 return AOP_WRITEPAGE_ACTIVATE;
607}
608
609static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
610{
611 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
612 struct writeback_control wbc = {
613 .sync_mode = WB_SYNC_ALL,
614 .nr_to_write = LONG_MAX,
615 .range_start = jinode->i_dirty_start,
616 .range_end = jinode->i_dirty_end,
617 };
618
619 return write_cache_pages(mapping, &wbc,
620 ext4_journalled_writepage_callback,
621 jinode->i_transaction);
622}
623
624static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
625{
626 int ret;
627
628 if (ext4_should_journal_data(jinode->i_vfs_inode))
629 ret = ext4_journalled_submit_inode_data_buffers(jinode);
630 else
631 ret = jbd2_journal_submit_inode_data_buffers(jinode);
632
633 return ret;
634}
635
636static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
637{
638 int ret = 0;
639
640 if (!ext4_should_journal_data(jinode->i_vfs_inode))
641 ret = jbd2_journal_finish_inode_data_buffers(jinode);
642
643 return ret;
644}
645
1dc1097f
JK		 646static bool system_going_down(void)
647{
648 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
649 || system_state == SYSTEM_RESTART;
650}
651
ac27a0ec
DK		 652/* Deal with the reporting of failure conditions on a filesystem such as
653 * inconsistencies detected or read IO failures.
654 *
655 * On ext2, we can store the error state of the filesystem in the
617ba13b 656 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 657 * write ordering constraints on the superblock which prevent us from
658 * writing it out straight away; and given that the journal is about to
659 * be aborted, we can't rely on the current, or future, transactions to
660 * write out the superblock safely.
661 *
dab291af 662 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 663 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 664 * that error until we've noted it down and cleared it.
665 */
666
617ba13b 667static void ext4_handle_error(struct super_block *sb)
ac27a0ec 668{
327eaf73
TT		 669 if (test_opt(sb, WARN_ON_ERROR))
670 WARN_ON_ONCE(1);
671
bc98a42c 672 if (sb_rdonly(sb))
ac27a0ec
DK		 673 return;
674
2b2d6d01 675 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 676 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 677
9b5f6c9b 678 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
ac27a0ec 679 if (journal)
dab291af 680 jbd2_journal_abort(journal, -EIO);
ac27a0ec 681 }
1dc1097f
JK		 682 /*
683 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
684 * could panic during 'reboot -f' as the underlying device got already
685 * disabled.
686 */
687 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
b31e1552 688 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 689 /*
690 * Make sure updated value of ->s_mount_flags will be visible
691 * before ->s_flags update
692 */
693 smp_wmb();
1751e8a6 694 sb->s_flags |= SB_RDONLY;
1dc1097f 695 } else if (test_opt(sb, ERRORS_PANIC)) {
617ba13b 696 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 697 sb->s_id);
4327ba52 698 }
ac27a0ec
DK		 699}
700
efbed4dc
TT		 701#define ext4_error_ratelimit(sb) \
702 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
703 "EXT4-fs error")
704
12062ddd 705void __ext4_error(struct super_block *sb, const char *function,
54d3adbc
TT		 706 unsigned int line, int error, __u64 block,
707 const char *fmt, ...)
ac27a0ec 708{
0ff2ea7d 709 struct va_format vaf;
ac27a0ec
DK		 710 va_list args;
711
0db1ff22
TT		 712 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
713 return;
714
ccf0f32a 715 trace_ext4_error(sb, function, line);
efbed4dc
TT		 716 if (ext4_error_ratelimit(sb)) {
717 va_start(args, fmt);
718 vaf.fmt = fmt;
719 vaf.va = &args;
720 printk(KERN_CRIT
721 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
722 sb->s_id, function, line, current->comm, &vaf);
723 va_end(args);
724 }
54d3adbc 725 save_error_info(sb, error, 0, block, function, line);
617ba13b 726 ext4_handle_error(sb);
ac27a0ec
DK		 727}
728
e7c96e8e 729void __ext4_error_inode(struct inode *inode, const char *function,
54d3adbc 730 unsigned int line, ext4_fsblk_t block, int error,
e7c96e8e 731 const char *fmt, ...)
273df556
FM		 732{
733 va_list args;
f7c21177 734 struct va_format vaf;
273df556 735
0db1ff22
TT		 736 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
737 return;
738
ccf0f32a 739 trace_ext4_error(inode->i_sb, function, line);
efbed4dc
TT		 740 if (ext4_error_ratelimit(inode->i_sb)) {
741 va_start(args, fmt);
742 vaf.fmt = fmt;
743 vaf.va = &args;
744 if (block)
745 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
746 "inode #%lu: block %llu: comm %s: %pV\n",
747 inode->i_sb->s_id, function, line, inode->i_ino,
748 block, current->comm, &vaf);
749 else
750 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
751 "inode #%lu: comm %s: %pV\n",
752 inode->i_sb->s_id, function, line, inode->i_ino,
753 current->comm, &vaf);
754 va_end(args);
755 }
54d3adbc
TT		 756 save_error_info(inode->i_sb, error, inode->i_ino, block,
757 function, line);
273df556
FM		 758 ext4_handle_error(inode->i_sb);
759}
760
e7c96e8e
JP		 761void __ext4_error_file(struct file *file, const char *function,
762 unsigned int line, ext4_fsblk_t block,
763 const char *fmt, ...)
273df556
FM		 764{
765 va_list args;
f7c21177 766 struct va_format vaf;
496ad9aa 767 struct inode *inode = file_inode(file);
273df556
FM		 768 char pathname[80], *path;
769
0db1ff22
TT		 770 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
771 return;
772
ccf0f32a 773 trace_ext4_error(inode->i_sb, function, line);
efbed4dc 774 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 775 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 776 if (IS_ERR(path))
777 path = "(unknown)";
778 va_start(args, fmt);
779 vaf.fmt = fmt;
780 vaf.va = &args;
781 if (block)
782 printk(KERN_CRIT
783 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
784 "block %llu: comm %s: path %s: %pV\n",
785 inode->i_sb->s_id, function, line, inode->i_ino,
786 block, current->comm, path, &vaf);
787 else
788 printk(KERN_CRIT
789 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
790 "comm %s: path %s: %pV\n",
791 inode->i_sb->s_id, function, line, inode->i_ino,
792 current->comm, path, &vaf);
793 va_end(args);
794 }
54d3adbc
TT		 795 save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block,
796 function, line);
273df556
FM		 797 ext4_handle_error(inode->i_sb);
798}
799
722887dd
TT		 800const char *ext4_decode_error(struct super_block *sb, int errno,
801 char nbuf[16])
ac27a0ec
DK		 802{
803 char *errstr = NULL;
804
805 switch (errno) {
6a797d27
DW		 806 case -EFSCORRUPTED:
807 errstr = "Corrupt filesystem";
808 break;
809 case -EFSBADCRC:
810 errstr = "Filesystem failed CRC";
811 break;
ac27a0ec
DK		 812 case -EIO:
813 errstr = "IO failure";
814 break;
815 case -ENOMEM:
816 errstr = "Out of memory";
817 break;
818 case -EROFS:
78f1ddbb
TT		 819 if (!sb || (EXT4_SB(sb)->s_journal &&
820 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 821 errstr = "Journal has aborted";
822 else
823 errstr = "Readonly filesystem";
824 break;
825 default:
826 /* If the caller passed in an extra buffer for unknown
827 * errors, textualise them now. Else we just return
828 * NULL. */
829 if (nbuf) {
830 /* Check for truncated error codes... */
831 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
832 errstr = nbuf;
833 }
834 break;
835 }
836
837 return errstr;
838}
839
617ba13b 840/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 841 * automatically and invokes the appropriate error response. */
842
c398eda0
TT		 843void __ext4_std_error(struct super_block *sb, const char *function,
844 unsigned int line, int errno)
ac27a0ec
DK		 845{
846 char nbuf[16];
847 const char *errstr;
848
0db1ff22
TT		 849 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
850 return;
851
ac27a0ec
DK		 852 /* Special case: if the error is EROFS, and we're not already
853 * inside a transaction, then there's really no point in logging
854 * an error. */
bc98a42c 855 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
ac27a0ec
DK		 856 return;
857
efbed4dc
TT		 858 if (ext4_error_ratelimit(sb)) {
859 errstr = ext4_decode_error(sb, errno, nbuf);
860 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
861 sb->s_id, function, line, errstr);
862 }
ac27a0ec 863
54d3adbc 864 save_error_info(sb, -errno, 0, 0, function, line);
617ba13b 865 ext4_handle_error(sb);
ac27a0ec
DK		 866}
867
868/*
617ba13b 869 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 870 * abort function may be used to deal with unrecoverable failures such
871 * as journal IO errors or ENOMEM at a critical moment in log management.
872 *
873 * We unconditionally force the filesystem into an ABORT|READONLY state,
874 * unless the error response on the fs has been set to panic in which
875 * case we take the easy way out and panic immediately.
876 */
877
c67d859e 878void __ext4_abort(struct super_block *sb, const char *function,
54d3adbc 879 unsigned int line, int error, const char *fmt, ...)
ac27a0ec 880{
651e1c3b 881 struct va_format vaf;
ac27a0ec
DK		 882 va_list args;
883
0db1ff22
TT		 884 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
885 return;
886
54d3adbc 887 save_error_info(sb, error, 0, 0, function, line);
ac27a0ec 888 va_start(args, fmt);
651e1c3b
JP		 889 vaf.fmt = fmt;
890 vaf.va = &args;
891 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
892 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 893 va_end(args);
894
bc98a42c 895 if (sb_rdonly(sb) == 0) {
9b5f6c9b 896 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
7b97d868 897 if (EXT4_SB(sb)->s_journal)
898 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
899
900 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 901 /*
902 * Make sure updated value of ->s_mount_flags will be visible
903 * before ->s_flags update
904 */
905 smp_wmb();
1751e8a6 906 sb->s_flags |= SB_RDONLY;
1c13d5c0 907 }
7b97d868 908 if (test_opt(sb, ERRORS_PANIC) && !system_going_down())
617ba13b 909 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 910}
911
e7c96e8e
JP		 912void __ext4_msg(struct super_block *sb,
913 const char *prefix, const char *fmt, ...)
b31e1552 914{
0ff2ea7d 915 struct va_format vaf;
b31e1552
ES		 916 va_list args;
917
1cf006ed 918 atomic_inc(&EXT4_SB(sb)->s_msg_count);
efbed4dc
TT		 919 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
920 return;
921
b31e1552 922 va_start(args, fmt);
0ff2ea7d
JP		 923 vaf.fmt = fmt;
924 vaf.va = &args;
925 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 926 va_end(args);
927}
928
1cf006ed
DM		 929static int ext4_warning_ratelimit(struct super_block *sb)
930{
931 atomic_inc(&EXT4_SB(sb)->s_warning_count);
932 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
933 "EXT4-fs warning");
934}
b03a2f7e 935
12062ddd 936void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 937 unsigned int line, const char *fmt, ...)
ac27a0ec 938{
0ff2ea7d 939 struct va_format vaf;
ac27a0ec
DK		 940 va_list args;
941
b03a2f7e 942 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 943 return;
944
ac27a0ec 945 va_start(args, fmt);
0ff2ea7d
JP		 946 vaf.fmt = fmt;
947 vaf.va = &args;
948 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
949 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 950 va_end(args);
951}
952
b03a2f7e
AD		 953void __ext4_warning_inode(const struct inode *inode, const char *function,
954 unsigned int line, const char *fmt, ...)
955{
956 struct va_format vaf;
957 va_list args;
958
959 if (!ext4_warning_ratelimit(inode->i_sb))
960 return;
961
962 va_start(args, fmt);
963 vaf.fmt = fmt;
964 vaf.va = &args;
965 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
966 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
967 function, line, inode->i_ino, current->comm, &vaf);
968 va_end(args);
969}
970
e29136f8
TT		 971void __ext4_grp_locked_error(const char *function, unsigned int line,
972 struct super_block *sb, ext4_group_t grp,
973 unsigned long ino, ext4_fsblk_t block,
974 const char *fmt, ...)
5d1b1b3f
AK		 975__releases(bitlock)
976__acquires(bitlock)
977{
0ff2ea7d 978 struct va_format vaf;
5d1b1b3f 979 va_list args;
5d1b1b3f 980
0db1ff22
TT		 981 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
982 return;
983
ccf0f32a 984 trace_ext4_error(sb, function, line);
54d3adbc 985 __save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
0ff2ea7d 986
efbed4dc
TT		 987 if (ext4_error_ratelimit(sb)) {
988 va_start(args, fmt);
989 vaf.fmt = fmt;
990 vaf.va = &args;
991 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
992 sb->s_id, function, line, grp);
993 if (ino)
994 printk(KERN_CONT "inode %lu: ", ino);
995 if (block)
996 printk(KERN_CONT "block %llu:",
997 (unsigned long long) block);
998 printk(KERN_CONT "%pV\n", &vaf);
999 va_end(args);
1000 }
5d1b1b3f 1001
327eaf73
TT		 1002 if (test_opt(sb, WARN_ON_ERROR))
1003 WARN_ON_ONCE(1);
1004
5d1b1b3f 1005 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 1006 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 1007 return;
1008 }
1c13d5c0 1009
5d1b1b3f 1010 ext4_unlock_group(sb, grp);
06f29cc8 1011 ext4_commit_super(sb, 1);
5d1b1b3f
AK		 1012 ext4_handle_error(sb);
1013 /*
1014 * We only get here in the ERRORS_RO case; relocking the group
1015 * may be dangerous, but nothing bad will happen since the
1016 * filesystem will have already been marked read/only and the
1017 * journal has been aborted. We return 1 as a hint to callers
1018 * who might what to use the return value from
25985edc 1019 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 1020 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1021 * aggressively from the ext4 function in question, with a
1022 * more appropriate error code.
1023 */
1024 ext4_lock_group(sb, grp);
1025 return;
1026}
1027
db79e6d1
WS		 1028void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1029 ext4_group_t group,
1030 unsigned int flags)
1031{
1032 struct ext4_sb_info *sbi = EXT4_SB(sb);
1033 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1034 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
9af0b3d1
WS		 1035 int ret;
1036
1037 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1038 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1039 &grp->bb_state);
1040 if (!ret)
1041 percpu_counter_sub(&sbi->s_freeclusters_counter,
1042 grp->bb_free);
db79e6d1
WS		 1043 }
1044
9af0b3d1
WS		 1045 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1046 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1047 &grp->bb_state);
1048 if (!ret && gdp) {
db79e6d1
WS		 1049 int count;
1050
1051 count = ext4_free_inodes_count(sb, gdp);
1052 percpu_counter_sub(&sbi->s_freeinodes_counter,
1053 count);
1054 }
db79e6d1
WS		 1055 }
1056}
1057
617ba13b 1058void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 1059{
617ba13b 1060 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 1061
617ba13b 1062 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 1063 return;
1064
12062ddd 1065 ext4_warning(sb,
ac27a0ec
DK		 1066 "updating to rev %d because of new feature flag, "
1067 "running e2fsck is recommended",
617ba13b 1068 EXT4_DYNAMIC_REV);
ac27a0ec 1069
617ba13b
MC		 1070 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1071 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1072 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 1073 /* leave es->s_feature_*compat flags alone */
1074 /* es->s_uuid will be set by e2fsck if empty */
1075
1076 /*
1077 * The rest of the superblock fields should be zero, and if not it
1078 * means they are likely already in use, so leave them alone. We
1079 * can leave it up to e2fsck to clean up any inconsistencies there.
1080 */
1081}
1082
1083/*
1084 * Open the external journal device
1085 */
b31e1552 1086static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 1087{
1088 struct block_device *bdev;
ac27a0ec 1089
d4d77629 1090 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 1091 if (IS_ERR(bdev))
1092 goto fail;
1093 return bdev;
1094
1095fail:
ea3edd4d
CH		 1096 ext4_msg(sb, KERN_ERR,
1097 "failed to open journal device unknown-block(%u,%u) %ld",
1098 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
ac27a0ec
DK		 1099 return NULL;
1100}
1101
1102/*
1103 * Release the journal device
1104 */
4385bab1 1105static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 1106{
4385bab1 1107 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 1108}
1109
4385bab1 1110static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 1111{
1112 struct block_device *bdev;
ee7ed3aa 1113 bdev = sbi->s_journal_bdev;
ac27a0ec 1114 if (bdev) {
4385bab1 1115 ext4_blkdev_put(bdev);
ee7ed3aa 1116 sbi->s_journal_bdev = NULL;
ac27a0ec 1117 }
ac27a0ec
DK		 1118}
1119
1120static inline struct inode *orphan_list_entry(struct list_head *l)
1121{
617ba13b 1122 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 1123}
1124
617ba13b 1125static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 1126{
1127 struct list_head *l;
1128
b31e1552
ES		 1129 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1130 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 1131
1132 printk(KERN_ERR "sb_info orphan list:\n");
1133 list_for_each(l, &sbi->s_orphan) {
1134 struct inode *inode = orphan_list_entry(l);
1135 printk(KERN_ERR " "
1136 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1137 inode->i_sb->s_id, inode->i_ino, inode,
1138 inode->i_mode, inode->i_nlink,
1139 NEXT_ORPHAN(inode));
1140 }
1141}
1142
957153fc
JK		 1143#ifdef CONFIG_QUOTA
1144static int ext4_quota_off(struct super_block *sb, int type);
1145
1146static inline void ext4_quota_off_umount(struct super_block *sb)
1147{
1148 int type;
1149
964edf66
JK		 1150 /* Use our quota_off function to clear inode flags etc. */
1151 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1152 ext4_quota_off(sb, type);
957153fc 1153}
33458eab
TT		 1154
1155/*
1156 * This is a helper function which is used in the mount/remount
1157 * codepaths (which holds s_umount) to fetch the quota file name.
1158 */
1159static inline char *get_qf_name(struct super_block *sb,
1160 struct ext4_sb_info *sbi,
1161 int type)
1162{
1163 return rcu_dereference_protected(sbi->s_qf_names[type],
1164 lockdep_is_held(&sb->s_umount));
1165}
957153fc
JK		 1166#else
1167static inline void ext4_quota_off_umount(struct super_block *sb)
1168{
1169}
1170#endif
1171
2b2d6d01 1172static void ext4_put_super(struct super_block *sb)
ac27a0ec 1173{
617ba13b
MC		 1174 struct ext4_sb_info *sbi = EXT4_SB(sb);
1175 struct ext4_super_block *es = sbi->s_es;
1d0c3924 1176 struct buffer_head **group_desc;
7c990728 1177 struct flex_groups **flex_groups;
97abd7d4 1178 int aborted = 0;
ef2cabf7 1179 int i, err;
ac27a0ec 1180
857ac889 1181 ext4_unregister_li_request(sb);
957153fc 1182 ext4_quota_off_umount(sb);
e0ccfd95 1183
2e8fa54e 1184 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 1185
5e47868f
RH		 1186 /*
1187 * Unregister sysfs before destroying jbd2 journal.
1188 * Since we could still access attr_journal_task attribute via sysfs
1189 * path which could have sbi->s_journal->j_task as NULL
1190 */
1191 ext4_unregister_sysfs(sb);
1192
0390131b 1193 if (sbi->s_journal) {
97abd7d4 1194 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 1195 err = jbd2_journal_destroy(sbi->s_journal);
1196 sbi->s_journal = NULL;
878520ac 1197 if ((err < 0) && !aborted) {
54d3adbc 1198 ext4_abort(sb, -err, "Couldn't clean up the journal");
878520ac 1199 }
0390131b 1200 }
d4edac31 1201
d3922a77 1202 ext4_es_unregister_shrinker(sbi);
9105bb14 1203 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 1204 ext4_release_system_zone(sb);
1205 ext4_mb_release(sb);
1206 ext4_ext_release(sb);
d4edac31 1207
bc98a42c 1208 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 1209 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 1210 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 1211 }
bc98a42c 1212 if (!sb_rdonly(sb))
a8e25a83
AB		 1213 ext4_commit_super(sb, 1);
1214
1d0c3924
TT		 1215 rcu_read_lock();
1216 group_desc = rcu_dereference(sbi->s_group_desc);
ac27a0ec 1217 for (i = 0; i < sbi->s_gdb_count; i++)
1d0c3924
TT		 1218 brelse(group_desc[i]);
1219 kvfree(group_desc);
7c990728
SJS		 1220 flex_groups = rcu_dereference(sbi->s_flex_groups);
1221 if (flex_groups) {
1222 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1223 kvfree(flex_groups[i]);
1224 kvfree(flex_groups);
1225 }
1d0c3924 1226 rcu_read_unlock();
57042651 1227 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 1228 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1229 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 1230 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
bbd55937 1231 percpu_free_rwsem(&sbi->s_writepages_rwsem);
ac27a0ec 1232#ifdef CONFIG_QUOTA
a2d4a646 1233 for (i = 0; i < EXT4_MAXQUOTAS; i++)
33458eab 1234 kfree(get_qf_name(sb, sbi, i));
ac27a0ec
DK		 1235#endif
1236
1237 /* Debugging code just in case the in-memory inode orphan list
1238 * isn't empty. The on-disk one can be non-empty if we've
1239 * detected an error and taken the fs readonly, but the
1240 * in-memory list had better be clean by this point. */
1241 if (!list_empty(&sbi->s_orphan))
1242 dump_orphan_list(sb, sbi);
1243 J_ASSERT(list_empty(&sbi->s_orphan));
1244
89d96a6f 1245 sync_blockdev(sb->s_bdev);
f98393a6 1246 invalidate_bdev(sb->s_bdev);
ee7ed3aa 1247 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
ac27a0ec
DK		 1248 /*
1249 * Invalidate the journal device's buffers. We don't want them
1250 * floating about in memory - the physical journal device may
1251 * hotswapped, and it breaks the `ro-after' testing code.
1252 */
ee7ed3aa
CX		 1253 sync_blockdev(sbi->s_journal_bdev);
1254 invalidate_bdev(sbi->s_journal_bdev);
617ba13b 1255 ext4_blkdev_remove(sbi);
ac27a0ec 1256 }
50c15df6
CX		 1257
1258 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1259 sbi->s_ea_inode_cache = NULL;
1260
1261 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1262 sbi->s_ea_block_cache = NULL;
1263
c5e06d10
JL		 1264 if (sbi->s_mmp_tsk)
1265 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 1266 brelse(sbi->s_sbh);
ac27a0ec 1267 sb->s_fs_info = NULL;
3197ebdb
TT		 1268 /*
1269 * Now that we are completely done shutting down the
1270 * superblock, we need to actually destroy the kobject.
1271 */
3197ebdb
TT		 1272 kobject_put(&sbi->s_kobj);
1273 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 1274 if (sbi->s_chksum_driver)
1275 crypto_free_shash(sbi->s_chksum_driver);
705895b6 1276 kfree(sbi->s_blockgroup_lock);
5e405595 1277 fs_put_dax(sbi->s_daxdev);
ac4acb1f 1278 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
c83ad55e 1279#ifdef CONFIG_UNICODE
f8f4acb6 1280 utf8_unload(sb->s_encoding);
c83ad55e 1281#endif
ac27a0ec 1282 kfree(sbi);
ac27a0ec
DK		 1283}
1284
e18b890b 1285static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 1286
1287/*
1288 * Called inside transaction, so use GFP_NOFS
1289 */
617ba13b 1290static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 1291{
617ba13b 1292 struct ext4_inode_info *ei;
ac27a0ec 1293
e6b4f8da 1294 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 1295 if (!ei)
1296 return NULL;
0b8e58a1 1297
ee73f9a5 1298 inode_set_iversion(&ei->vfs_inode, 1);
202ee5df 1299 spin_lock_init(&ei->i_raw_lock);
c9de560d 1300 INIT_LIST_HEAD(&ei->i_prealloc_list);
27bc446e 1301 atomic_set(&ei->i_prealloc_active, 0);
c9de560d 1302 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 1303 ext4_es_init_tree(&ei->i_es_tree);
1304 rwlock_init(&ei->i_es_lock);
edaa53ca 1305 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 1306 ei->i_es_all_nr = 0;
edaa53ca 1307 ei->i_es_shk_nr = 0;
dd475925 1308 ei->i_es_shrink_lblk = 0;
d2a17637 1309 ei->i_reserved_data_blocks = 0;
d2a17637 1310 spin_lock_init(&(ei->i_block_reservation_lock));
1dc0aa46 1311 ext4_init_pending_tree(&ei->i_pending_tree);
a9e7f447
DM		 1312#ifdef CONFIG_QUOTA
1313 ei->i_reserved_quota = 0;
96c7e0d9 1314 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 1315#endif
8aefcd55 1316 ei->jinode = NULL;
2e8fa54e 1317 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 1318 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 1319 ei->i_sync_tid = 0;
1320 ei->i_datasync_tid = 0;
e27f41e1 1321 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 1322 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
aa75f4d3
HS		 1323 ext4_fc_init_inode(&ei->vfs_inode);
1324 mutex_init(&ei->i_fc_lock);
ac27a0ec
DK		 1325 return &ei->vfs_inode;
1326}
1327
7ff9c073
TT		 1328static int ext4_drop_inode(struct inode *inode)
1329{
1330 int drop = generic_drop_inode(inode);
1331
29b3692e
EB		 1332 if (!drop)
1333 drop = fscrypt_drop_inode(inode);
1334
7ff9c073
TT		 1335 trace_ext4_drop_inode(inode, drop);
1336 return drop;
1337}
1338
94053139 1339static void ext4_free_in_core_inode(struct inode *inode)
fa0d7e3d 1340{
2c58d548 1341 fscrypt_free_inode(inode);
aa75f4d3
HS		 1342 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1343 pr_warn("%s: inode %ld still in fc list",
1344 __func__, inode->i_ino);
1345 }
fa0d7e3d
NP		 1346 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1347}
1348
617ba13b 1349static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1350{
9f7dd93d 1351 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1352 ext4_msg(inode->i_sb, KERN_ERR,
1353 "Inode %lu (%p): orphan list check failed!",
1354 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1355 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1356 EXT4_I(inode), sizeof(struct ext4_inode_info),
1357 true);
1358 dump_stack();
1359 }
ac27a0ec
DK		 1360}
1361
51cc5068 1362static void init_once(void *foo)
ac27a0ec 1363{
617ba13b 1364 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1365
a35afb83 1366 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1367 init_rwsem(&ei->xattr_sem);
0e855ac8 1368 init_rwsem(&ei->i_data_sem);
ea3d7209 1369 init_rwsem(&ei->i_mmap_sem);
a35afb83 1370 inode_init_once(&ei->vfs_inode);
aa75f4d3 1371 ext4_fc_init_inode(&ei->vfs_inode);
ac27a0ec
DK		 1372}
1373
e67bc2b3 1374static int __init init_inodecache(void)
ac27a0ec 1375{
f8dd7c70
DW		 1376 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1377 sizeof(struct ext4_inode_info), 0,
1378 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1379 SLAB_ACCOUNT),
1380 offsetof(struct ext4_inode_info, i_data),
1381 sizeof_field(struct ext4_inode_info, i_data),
1382 init_once);
617ba13b 1383 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1384 return -ENOMEM;
1385 return 0;
1386}
1387
1388static void destroy_inodecache(void)
1389{
8c0a8537
KS		 1390 /*
1391 * Make sure all delayed rcu free inodes are flushed before we
1392 * destroy cache.
1393 */
1394 rcu_barrier();
617ba13b 1395 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1396}
1397
0930fcc1 1398void ext4_clear_inode(struct inode *inode)
ac27a0ec 1399{
aa75f4d3 1400 ext4_fc_del(inode);
0930fcc1 1401 invalidate_inode_buffers(inode);
dbd5768f 1402 clear_inode(inode);
27bc446e 1403 ext4_discard_preallocations(inode, 0);
51865fda 1404 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
f4c2d372 1405 dquot_drop(inode);
8aefcd55
TT		 1406 if (EXT4_I(inode)->jinode) {
1407 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1408 EXT4_I(inode)->jinode);
1409 jbd2_free_inode(EXT4_I(inode)->jinode);
1410 EXT4_I(inode)->jinode = NULL;
1411 }
3d204e24 1412 fscrypt_put_encryption_info(inode);
c93d8f88 1413 fsverity_cleanup_inode(inode);
ac27a0ec
DK		 1414}
1415
1b961ac0 1416static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1417 u64 ino, u32 generation)
ac27a0ec 1418{
ac27a0ec 1419 struct inode *inode;
ac27a0ec 1420
8a363970 1421 /*
ac27a0ec
DK		 1422 * Currently we don't know the generation for parent directory, so
1423 * a generation of 0 means "accept any"
1424 */
8a363970 1425 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1d1fe1ee
DH		 1426 if (IS_ERR(inode))
1427 return ERR_CAST(inode);
1428 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1429 iput(inode);
1430 return ERR_PTR(-ESTALE);
1431 }
1b961ac0
CH		 1432
1433 return inode;
1434}
1435
1436static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1437 int fh_len, int fh_type)
1b961ac0
CH		 1438{
1439 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1440 ext4_nfs_get_inode);
1441}
1442
1443static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1444 int fh_len, int fh_type)
1b961ac0
CH		 1445{
1446 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1447 ext4_nfs_get_inode);
ac27a0ec
DK		 1448}
1449
fde87268
TT		 1450static int ext4_nfs_commit_metadata(struct inode *inode)
1451{
1452 struct writeback_control wbc = {
1453 .sync_mode = WB_SYNC_ALL
1454 };
1455
1456 trace_ext4_nfs_commit_metadata(inode);
1457 return ext4_write_inode(inode, &wbc);
1458}
1459
c39a7f84
TO		 1460/*
1461 * Try to release metadata pages (indirect blocks, directories) which are
1462 * mapped via the block device. Since these pages could have journal heads
1463 * which would prevent try_to_free_buffers() from freeing them, we must use
1464 * jbd2 layer's try_to_free_buffers() function to release them.
1465 */
0b8e58a1
AD		 1466static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1467 gfp_t wait)
c39a7f84
TO		 1468{
1469 journal_t *journal = EXT4_SB(sb)->s_journal;
1470
1471 WARN_ON(PageChecked(page));
1472 if (!page_has_buffers(page))
1473 return 0;
1474 if (journal)
529a781e 1475 return jbd2_journal_try_to_free_buffers(journal, page);
1476
c39a7f84
TO		 1477 return try_to_free_buffers(page);
1478}
1479
643fa961 1480#ifdef CONFIG_FS_ENCRYPTION
a7550b30
JK		 1481static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1482{
1483 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1484 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1485}
1486
a7550b30
JK		 1487static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1488 void *fs_data)
1489{
2f8f5e76 1490 handle_t *handle = fs_data;
c1a5d5f6 1491 int res, res2, credits, retries = 0;
2f8f5e76 1492
9ce0151a
EB		 1493 /*
1494 * Encrypting the root directory is not allowed because e2fsck expects
1495 * lost+found to exist and be unencrypted, and encrypting the root
1496 * directory would imply encrypting the lost+found directory as well as
1497 * the filename "lost+found" itself.
1498 */
1499 if (inode->i_ino == EXT4_ROOT_INO)
1500 return -EPERM;
2f8f5e76 1501
7d3e06a8
RZ		 1502 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1503 return -EINVAL;
1504
b383a73f
IW		 1505 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1506 return -EOPNOTSUPP;
1507
94840e3c
EB		 1508 res = ext4_convert_inline_data(inode);
1509 if (res)
1510 return res;
1511
2f8f5e76
EB		 1512 /*
1513 * If a journal handle was specified, then the encryption context is
1514 * being set on a new inode via inheritance and is part of a larger
1515 * transaction to create the inode. Otherwise the encryption context is
1516 * being set on an existing inode in its own transaction. Only in the
1517 * latter case should the "retry on ENOSPC" logic be used.
1518 */
a7550b30 1519
2f8f5e76
EB		 1520 if (handle) {
1521 res = ext4_xattr_set_handle(handle, inode,
1522 EXT4_XATTR_INDEX_ENCRYPTION,
1523 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1524 ctx, len, 0);
a7550b30
JK		 1525 if (!res) {
1526 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1527 ext4_clear_inode_state(inode,
1528 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b 1529 /*
2ee6a576
EB		 1530 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1531 * S_DAX may be disabled
a3caa24b 1532 */
043546e4 1533 ext4_set_inode_flags(inode, false);
a7550b30
JK		 1534 }
1535 return res;
1536 }
1537
b8cb5a54
TE		 1538 res = dquot_initialize(inode);
1539 if (res)
1540 return res;
2f8f5e76 1541retry:
af65207c
TE		 1542 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1543 &credits);
dec214d0
TE		 1544 if (res)
1545 return res;
1546
c1a5d5f6 1547 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
a7550b30
JK		 1548 if (IS_ERR(handle))
1549 return PTR_ERR(handle);
1550
2f8f5e76
EB		 1551 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1552 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1553 ctx, len, 0);
a7550b30
JK		 1554 if (!res) {
1555 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
2ee6a576
EB		 1556 /*
1557 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1558 * S_DAX may be disabled
1559 */
043546e4 1560 ext4_set_inode_flags(inode, false);
a7550b30
JK		 1561 res = ext4_mark_inode_dirty(handle, inode);
1562 if (res)
1563 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1564 }
1565 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1566
1567 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1568 goto retry;
a7550b30
JK		 1569 if (!res)
1570 res = res2;
1571 return res;
1572}
1573
ac4acb1f 1574static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
a7550b30 1575{
ac4acb1f 1576 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
a7550b30
JK		 1577}
1578
b925acb8
EB		 1579static bool ext4_has_stable_inodes(struct super_block *sb)
1580{
1581 return ext4_has_feature_stable_inodes(sb);
1582}
1583
1584static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1585 int *ino_bits_ret, int *lblk_bits_ret)
1586{
1587 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1588 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1589}
1590
6f69f0ed 1591static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1592 .key_prefix = "ext4:",
a7550b30 1593 .get_context = ext4_get_context,
a7550b30 1594 .set_context = ext4_set_context,
ac4acb1f 1595 .get_dummy_policy = ext4_get_dummy_policy,
a7550b30 1596 .empty_dir = ext4_empty_dir,
e12ee683 1597 .max_namelen = EXT4_NAME_LEN,
b925acb8
EB		 1598 .has_stable_inodes = ext4_has_stable_inodes,
1599 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
a7550b30 1600};
a7550b30
JK		 1601#endif
1602
ac27a0ec 1603#ifdef CONFIG_QUOTA
d6006186 1604static const char * const quotatypes[] = INITQFNAMES;
689c958c 1605#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1606
617ba13b
MC		 1607static int ext4_write_dquot(struct dquot *dquot);
1608static int ext4_acquire_dquot(struct dquot *dquot);
1609static int ext4_release_dquot(struct dquot *dquot);
1610static int ext4_mark_dquot_dirty(struct dquot *dquot);
1611static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1612static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1613 const struct path *path);
617ba13b
MC		 1614static int ext4_quota_on_mount(struct super_block *sb, int type);
1615static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1616 size_t len, loff_t off);
617ba13b 1617static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1618 const char *data, size_t len, loff_t off);
7c319d32
AK		 1619static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1620 unsigned int flags);
1621static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1622
96c7e0d9
JK		 1623static struct dquot **ext4_get_dquots(struct inode *inode)
1624{
1625 return EXT4_I(inode)->i_dquot;
1626}
1627
61e225dc 1628static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1629 .get_reserved_space = ext4_get_reserved_space,
1630 .write_dquot = ext4_write_dquot,
1631 .acquire_dquot = ext4_acquire_dquot,
1632 .release_dquot = ext4_release_dquot,
1633 .mark_dirty = ext4_mark_dquot_dirty,
1634 .write_info = ext4_write_info,
1635 .alloc_dquot = dquot_alloc,
1636 .destroy_dquot = dquot_destroy,
1637 .get_projid = ext4_get_projid,
1638 .get_inode_usage = ext4_get_inode_usage,
ebc11f7b 1639 .get_next_id = dquot_get_next_id,
ac27a0ec
DK		 1640};
1641
0d54b217 1642static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1643 .quota_on = ext4_quota_on,
ca0e05e4 1644 .quota_off = ext4_quota_off,
287a8095 1645 .quota_sync = dquot_quota_sync,
0a240339 1646 .get_state = dquot_get_state,
287a8095
CH		 1647 .set_info = dquot_set_dqinfo,
1648 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1649 .set_dqblk = dquot_set_dqblk,
1650 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1651};
1652#endif
1653
ee9b6d61 1654static const struct super_operations ext4_sops = {
617ba13b 1655 .alloc_inode = ext4_alloc_inode,
94053139 1656 .free_inode = ext4_free_in_core_inode,
617ba13b 1657 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1658 .write_inode = ext4_write_inode,
1659 .dirty_inode = ext4_dirty_inode,
7ff9c073 1660 .drop_inode = ext4_drop_inode,
0930fcc1 1661 .evict_inode = ext4_evict_inode,
617ba13b 1662 .put_super = ext4_put_super,
617ba13b 1663 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1664 .freeze_fs = ext4_freeze,
1665 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1666 .statfs = ext4_statfs,
1667 .remount_fs = ext4_remount,
617ba13b 1668 .show_options = ext4_show_options,
ac27a0ec 1669#ifdef CONFIG_QUOTA
617ba13b
MC		 1670 .quota_read = ext4_quota_read,
1671 .quota_write = ext4_quota_write,
96c7e0d9 1672 .get_dquots = ext4_get_dquots,
ac27a0ec 1673#endif
c39a7f84 1674 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1675};
1676
39655164 1677static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1678 .fh_to_dentry = ext4_fh_to_dentry,
1679 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1680 .get_parent = ext4_get_parent,
fde87268 1681 .commit_metadata = ext4_nfs_commit_metadata,
ac27a0ec
DK		 1682};
1683
1684enum {
1685 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1686 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1687 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1688 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1689 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1690 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1691 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1692 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1693 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
4f74d15f 1694 Opt_inlinecrypt,
ac27a0ec 1695 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1696 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1697 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
9cb20f94
IW		 1698 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1699 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
327eaf73
TT		 1700 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1701 Opt_nowarn_on_error, Opt_mblk_io_submit,
670e9875 1702 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1703 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1704 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1705 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1706 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1707 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
99c880de 1708 Opt_prefetch_block_bitmaps,
8016e29f 1709#ifdef CONFIG_EXT4_DEBUG
99c880de 1710 Opt_fc_debug_max_replay, Opt_fc_debug_force
8016e29f 1711#endif
ac27a0ec
DK		 1712};
1713
a447c093 1714static const match_table_t tokens = {
ac27a0ec
DK		 1715 {Opt_bsd_df, "bsddf"},
1716 {Opt_minix_df, "minixdf"},
1717 {Opt_grpid, "grpid"},
1718 {Opt_grpid, "bsdgroups"},
1719 {Opt_nogrpid, "nogrpid"},
1720 {Opt_nogrpid, "sysvgroups"},
1721 {Opt_resgid, "resgid=%u"},
1722 {Opt_resuid, "resuid=%u"},
1723 {Opt_sb, "sb=%u"},
1724 {Opt_err_cont, "errors=continue"},
1725 {Opt_err_panic, "errors=panic"},
1726 {Opt_err_ro, "errors=remount-ro"},
1727 {Opt_nouid32, "nouid32"},
ac27a0ec 1728 {Opt_debug, "debug"},
72578c33
TT		 1729 {Opt_removed, "oldalloc"},
1730 {Opt_removed, "orlov"},
ac27a0ec
DK		 1731 {Opt_user_xattr, "user_xattr"},
1732 {Opt_nouser_xattr, "nouser_xattr"},
1733 {Opt_acl, "acl"},
1734 {Opt_noacl, "noacl"},
e3bb52ae 1735 {Opt_noload, "norecovery"},
5a916be1 1736 {Opt_noload, "noload"},
72578c33
TT		 1737 {Opt_removed, "nobh"},
1738 {Opt_removed, "bh"},
ac27a0ec 1739 {Opt_commit, "commit=%u"},
30773840
TT		 1740 {Opt_min_batch_time, "min_batch_time=%u"},
1741 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1742 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1743 {Opt_journal_path, "journal_path=%s"},
818d276c 1744 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1745 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1746 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1747 {Opt_abort, "abort"},
1748 {Opt_data_journal, "data=journal"},
1749 {Opt_data_ordered, "data=ordered"},
1750 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1751 {Opt_data_err_abort, "data_err=abort"},
1752 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1753 {Opt_offusrjquota, "usrjquota="},
1754 {Opt_usrjquota, "usrjquota=%s"},
1755 {Opt_offgrpjquota, "grpjquota="},
1756 {Opt_grpjquota, "grpjquota=%s"},
1757 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1758 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1759 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1760 {Opt_grpquota, "grpquota"},
1761 {Opt_noquota, "noquota"},
1762 {Opt_quota, "quota"},
1763 {Opt_usrquota, "usrquota"},
49da9392 1764 {Opt_prjquota, "prjquota"},
ac27a0ec 1765 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1766 {Opt_barrier, "barrier"},
1767 {Opt_nobarrier, "nobarrier"},
25ec56b5 1768 {Opt_i_version, "i_version"},
923ae0ff 1769 {Opt_dax, "dax"},
9cb20f94
IW		 1770 {Opt_dax_always, "dax=always"},
1771 {Opt_dax_inode, "dax=inode"},
1772 {Opt_dax_never, "dax=never"},
c9de560d 1773 {Opt_stripe, "stripe=%u"},
64769240 1774 {Opt_delalloc, "delalloc"},
327eaf73
TT		 1775 {Opt_warn_on_error, "warn_on_error"},
1776 {Opt_nowarn_on_error, "nowarn_on_error"},
a26f4992
TT		 1777 {Opt_lazytime, "lazytime"},
1778 {Opt_nolazytime, "nolazytime"},
670e9875 1779 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1780 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1781 {Opt_removed, "mblk_io_submit"},
1782 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1783 {Opt_block_validity, "block_validity"},
1784 {Opt_noblock_validity, "noblock_validity"},
240799cd 1785 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1786 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1787 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1788 {Opt_auto_da_alloc, "auto_da_alloc"},
1789 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc 1790 {Opt_dioread_nolock, "dioread_nolock"},
244adf64 1791 {Opt_dioread_lock, "nodioread_nolock"},
744692dc 1792 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1793 {Opt_discard, "discard"},
1794 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1795 {Opt_init_itable, "init_itable=%u"},
1796 {Opt_init_itable, "init_itable"},
1797 {Opt_noinit_itable, "noinit_itable"},
8016e29f 1798#ifdef CONFIG_EXT4_DEBUG
99c880de 1799 {Opt_fc_debug_force, "fc_debug_force"},
8016e29f
HS		 1800 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1801#endif
df981d03 1802 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
ed318a6c 1803 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
6ddb2447 1804 {Opt_test_dummy_encryption, "test_dummy_encryption"},
4f74d15f 1805 {Opt_inlinecrypt, "inlinecrypt"},
cdb7ee4c
TE		 1806 {Opt_nombcache, "nombcache"},
1807 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
3d392b26 1808 {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"},
c7198b9c
TT		 1809 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1810 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1811 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1812 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1813 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1814 {Opt_err, NULL},
ac27a0ec
DK		 1815};
1816
617ba13b 1817static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1818{
617ba13b 1819 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1820 char *options = (char *) *data;
1821
1822 if (!options || strncmp(options, "sb=", 3) != 0)
1823 return 1; /* Default location */
0b8e58a1 1824
ac27a0ec 1825 options += 3;
0b8e58a1 1826 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1827 sb_block = simple_strtoul(options, &options, 0);
1828 if (*options && *options != ',') {
4776004f 1829 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1830 (char *) *data);
1831 return 1;
1832 }
1833 if (*options == ',')
1834 options++;
1835 *data = (void *) options;
0b8e58a1 1836
ac27a0ec
DK		 1837 return sb_block;
1838}
1839
b3881f74 1840#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1841static const char deprecated_msg[] =
1842 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1843 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1844
56c50f11
DM		 1845#ifdef CONFIG_QUOTA
1846static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1847{
1848 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 1849 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
03dafb5f 1850 int ret = -1;
56c50f11 1851
33458eab 1852 if (sb_any_quota_loaded(sb) && !old_qname) {
56c50f11
DM		 1853 ext4_msg(sb, KERN_ERR,
1854 "Cannot change journaled "
1855 "quota options when quota turned on");
57f73c2c 1856 return -1;
56c50f11 1857 }
e2b911c5 1858 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1859 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1860 "ignored when QUOTA feature is enabled");
1861 return 1;
262b4662 1862 }
56c50f11
DM		 1863 qname = match_strdup(args);
1864 if (!qname) {
1865 ext4_msg(sb, KERN_ERR,
1866 "Not enough memory for storing quotafile name");
57f73c2c 1867 return -1;
56c50f11 1868 }
33458eab
TT		 1869 if (old_qname) {
1870 if (strcmp(old_qname, qname) == 0)
03dafb5f
CG		 1871 ret = 1;
1872 else
1873 ext4_msg(sb, KERN_ERR,
1874 "%s quota file already specified",
1875 QTYPE2NAME(qtype));
1876 goto errout;
56c50f11 1877 }
03dafb5f 1878 if (strchr(qname, '/')) {
56c50f11
DM		 1879 ext4_msg(sb, KERN_ERR,
1880 "quotafile must be on filesystem root");
03dafb5f 1881 goto errout;
56c50f11 1882 }
33458eab 1883 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
fd8c37ec 1884 set_opt(sb, QUOTA);
56c50f11 1885 return 1;
03dafb5f
CG		 1886errout:
1887 kfree(qname);
1888 return ret;
56c50f11
DM		 1889}
1890
1891static int clear_qf_name(struct super_block *sb, int qtype)
1892{
1893
1894 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 1895 char *old_qname = get_qf_name(sb, sbi, qtype);
56c50f11 1896
33458eab 1897 if (sb_any_quota_loaded(sb) && old_qname) {
56c50f11
DM		 1898 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1899 " when quota turned on");
57f73c2c 1900 return -1;
56c50f11 1901 }
33458eab
TT		 1902 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1903 synchronize_rcu();
1904 kfree(old_qname);
56c50f11
DM		 1905 return 1;
1906}
1907#endif
1908
26092bf5
TT		 1909#define MOPT_SET 0x0001
1910#define MOPT_CLEAR 0x0002
1911#define MOPT_NOSUPPORT 0x0004
1912#define MOPT_EXPLICIT 0x0008
1913#define MOPT_CLEAR_ERR 0x0010
1914#define MOPT_GTE0 0x0020
ac27a0ec 1915#ifdef CONFIG_QUOTA
26092bf5
TT		 1916#define MOPT_Q 0
1917#define MOPT_QFMT 0x0040
1918#else
1919#define MOPT_Q MOPT_NOSUPPORT
1920#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1921#endif
26092bf5 1922#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1923#define MOPT_NO_EXT2 0x0100
1924#define MOPT_NO_EXT3 0x0200
1925#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1926#define MOPT_STRING 0x0400
9cb20f94 1927#define MOPT_SKIP 0x0800
995a3ed6 1928#define MOPT_2 0x1000
26092bf5
TT		 1929
1930static const struct mount_opts {
1931 int token;
1932 int mount_opt;
1933 int flags;
1934} ext4_mount_opts[] = {
1935 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1936 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1937 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1938 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1939 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1940 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1941 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1942 MOPT_EXT4_ONLY | MOPT_SET},
1943 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1944 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1945 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1946 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1947 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1948 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1949 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1950 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1951 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1952 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
c6d3d56d
DW		 1953 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1954 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1955 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1956 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1957 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1958 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1959 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1960 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1961 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1962 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1963 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1964 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1965 MOPT_NO_EXT2},
8dc0aa8c 1966 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1967 MOPT_NO_EXT2},
26092bf5
TT		 1968 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1969 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1970 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1971 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1972 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1973 {Opt_commit, 0, MOPT_GTE0},
1974 {Opt_max_batch_time, 0, MOPT_GTE0},
1975 {Opt_min_batch_time, 0, MOPT_GTE0},
1976 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1977 {Opt_init_itable, 0, MOPT_GTE0},
9cb20f94
IW		 1978 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1979 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1980 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1981 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1982 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1983 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1984 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
26092bf5 1985 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1986 {Opt_resuid, 0, MOPT_GTE0},
1987 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1988 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1989 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1990 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1991 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1992 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1993 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1994 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1995 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1996 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1997#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1998 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1999 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 2000#else
26092bf5
TT		 2001 {Opt_acl, 0, MOPT_NOSUPPORT},
2002 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 2003#endif
26092bf5
TT		 2004 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
2005 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 2006 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 2007 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
2008 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
2009 MOPT_SET | MOPT_Q},
2010 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
2011 MOPT_SET | MOPT_Q},
49da9392
JK		 2012 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
2013 MOPT_SET | MOPT_Q},
26092bf5 2014 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 2015 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
2016 MOPT_CLEAR | MOPT_Q},
174fe5ba
KX		 2017 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
2018 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
26092bf5
TT		 2019 {Opt_offusrjquota, 0, MOPT_Q},
2020 {Opt_offgrpjquota, 0, MOPT_Q},
2021 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
2022 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
2023 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 2024 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
ed318a6c 2025 {Opt_test_dummy_encryption, 0, MOPT_STRING},
cdb7ee4c 2026 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
3d392b26
TT		 2027 {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS,
2028 MOPT_SET},
99c880de 2029#ifdef CONFIG_EXT4_DEBUG
0f0672ff
HS		 2030 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2031 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
8016e29f
HS		 2032 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2033#endif
26092bf5
TT		 2034 {Opt_err, 0, 0}
2035};
2036
c83ad55e
GKB		 2037#ifdef CONFIG_UNICODE
2038static const struct ext4_sb_encodings {
2039 __u16 magic;
2040 char *name;
2041 char *version;
2042} ext4_sb_encoding_map[] = {
2043 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2044};
2045
2046static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2047 const struct ext4_sb_encodings **encoding,
2048 __u16 *flags)
2049{
2050 __u16 magic = le16_to_cpu(es->s_encoding);
2051 int i;
2052
2053 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2054 if (magic == ext4_sb_encoding_map[i].magic)
2055 break;
2056
2057 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2058 return -EINVAL;
2059
2060 *encoding = &ext4_sb_encoding_map[i];
2061 *flags = le16_to_cpu(es->s_encoding_flags);
2062
2063 return 0;
2064}
2065#endif
2066
ed318a6c
EB		 2067static int ext4_set_test_dummy_encryption(struct super_block *sb,
2068 const char *opt,
2069 const substring_t *arg,
2070 bool is_remount)
2071{
2072#ifdef CONFIG_FS_ENCRYPTION
2073 struct ext4_sb_info *sbi = EXT4_SB(sb);
2074 int err;
2075
2076 /*
2077 * This mount option is just for testing, and it's not worthwhile to
2078 * implement the extra complexity (e.g. RCU protection) that would be
2079 * needed to allow it to be set or changed during remount. We do allow
2080 * it to be specified during remount, but only if there is no change.
2081 */
ac4acb1f 2082 if (is_remount && !sbi->s_dummy_enc_policy.policy) {
ed318a6c
EB		 2083 ext4_msg(sb, KERN_WARNING,
2084 "Can't set test_dummy_encryption on remount");
2085 return -1;
2086 }
c8c868ab 2087 err = fscrypt_set_test_dummy_encryption(sb, arg->from,
ac4acb1f 2088 &sbi->s_dummy_enc_policy);
ed318a6c
EB		 2089 if (err) {
2090 if (err == -EEXIST)
2091 ext4_msg(sb, KERN_WARNING,
2092 "Can't change test_dummy_encryption on remount");
2093 else if (err == -EINVAL)
2094 ext4_msg(sb, KERN_WARNING,
2095 "Value of option \"%s\" is unrecognized", opt);
2096 else
2097 ext4_msg(sb, KERN_WARNING,
2098 "Error processing option \"%s\" [%d]",
2099 opt, err);
2100 return -1;
2101 }
2102 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2103#else
2104 ext4_msg(sb, KERN_WARNING,
2105 "Test dummy encryption mount option ignored");
2106#endif
2107 return 1;
2108}
2109
26092bf5
TT		 2110static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2111 substring_t *args, unsigned long *journal_devnum,
2112 unsigned int *journal_ioprio, int is_remount)
2113{
2114 struct ext4_sb_info *sbi = EXT4_SB(sb);
2115 const struct mount_opts *m;
08cefc7a
EB		 2116 kuid_t uid;
2117 kgid_t gid;
26092bf5
TT		 2118 int arg = 0;
2119
57f73c2c
TT		 2120#ifdef CONFIG_QUOTA
2121 if (token == Opt_usrjquota)
2122 return set_qf_name(sb, USRQUOTA, &args[0]);
2123 else if (token == Opt_grpjquota)
2124 return set_qf_name(sb, GRPQUOTA, &args[0]);
2125 else if (token == Opt_offusrjquota)
2126 return clear_qf_name(sb, USRQUOTA);
2127 else if (token == Opt_offgrpjquota)
2128 return clear_qf_name(sb, GRPQUOTA);
2129#endif
26092bf5 2130 switch (token) {
f7048605
TT		 2131 case Opt_noacl:
2132 case Opt_nouser_xattr:
2133 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2134 break;
26092bf5
TT		 2135 case Opt_sb:
2136 return 1; /* handled by get_sb_block() */
2137 case Opt_removed:
5f3633e3 2138 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 2139 return 1;
26092bf5 2140 case Opt_abort:
9b5f6c9b 2141 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
26092bf5
TT		 2142 return 1;
2143 case Opt_i_version:
357fdad0 2144 sb->s_flags |= SB_I_VERSION;
26092bf5 2145 return 1;
a26f4992 2146 case Opt_lazytime:
1751e8a6 2147 sb->s_flags |= SB_LAZYTIME;
a26f4992
TT		 2148 return 1;
2149 case Opt_nolazytime:
1751e8a6 2150 sb->s_flags &= ~SB_LAZYTIME;
a26f4992 2151 return 1;
4f74d15f
EB		 2152 case Opt_inlinecrypt:
2153#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2154 sb->s_flags |= SB_INLINECRYPT;
2155#else
2156 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2157#endif
2158 return 1;
26092bf5
TT		 2159 }
2160
5f3633e3
JK		 2161 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2162 if (token == m->token)
2163 break;
2164
2165 if (m->token == Opt_err) {
2166 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2167 "or missing value", opt);
2168 return -1;
2169 }
2170
8dc0aa8c
TT		 2171 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2172 ext4_msg(sb, KERN_ERR,
2173 "Mount option \"%s\" incompatible with ext2", opt);
2174 return -1;
2175 }
2176 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2177 ext4_msg(sb, KERN_ERR,
2178 "Mount option \"%s\" incompatible with ext3", opt);
2179 return -1;
2180 }
2181
ad4eec61 2182 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 2183 return -1;
2184 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2185 return -1;
c93cf2d7
DM		 2186 if (m->flags & MOPT_EXPLICIT) {
2187 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2188 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 2189 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2190 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 2191 } else
2192 return -1;
2193 }
5f3633e3
JK		 2194 if (m->flags & MOPT_CLEAR_ERR)
2195 clear_opt(sb, ERRORS_MASK);
2196 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2197 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2198 "options when quota turned on");
2199 return -1;
2200 }
2201
2202 if (m->flags & MOPT_NOSUPPORT) {
2203 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2204 } else if (token == Opt_commit) {
2205 if (arg == 0)
2206 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
9ba55543 2207 else if (arg > INT_MAX / HZ) {
2208 ext4_msg(sb, KERN_ERR,
2209 "Invalid commit interval %d, "
2210 "must be smaller than %d",
2211 arg, INT_MAX / HZ);
2212 return -1;
2213 }
5f3633e3 2214 sbi->s_commit_interval = HZ * arg;
670e9875 2215 } else if (token == Opt_debug_want_extra_isize) {
9803387c
TT		 2216 if ((arg & 1) ||
2217 (arg < 4) ||
2218 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2219 ext4_msg(sb, KERN_ERR,
2220 "Invalid want_extra_isize %d", arg);
2221 return -1;
2222 }
670e9875 2223 sbi->s_want_extra_isize = arg;
5f3633e3 2224 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 2225 sbi->s_max_batch_time = arg;
2226 } else if (token == Opt_min_batch_time) {
2227 sbi->s_min_batch_time = arg;
2228 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 2229 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2230 ext4_msg(sb, KERN_ERR,
2231 "EXT4-fs: inode_readahead_blks must be "
2232 "0 or a power of 2 smaller than 2^31");
26092bf5 2233 return -1;
5f3633e3
JK		 2234 }
2235 sbi->s_inode_readahead_blks = arg;
2236 } else if (token == Opt_init_itable) {
2237 set_opt(sb, INIT_INODE_TABLE);
2238 if (!args->from)
2239 arg = EXT4_DEF_LI_WAIT_MULT;
2240 sbi->s_li_wait_mult = arg;
2241 } else if (token == Opt_max_dir_size_kb) {
2242 sbi->s_max_dir_size_kb = arg;
8016e29f
HS		 2243#ifdef CONFIG_EXT4_DEBUG
2244 } else if (token == Opt_fc_debug_max_replay) {
2245 sbi->s_fc_debug_max_replay = arg;
2246#endif
5f3633e3
JK		 2247 } else if (token == Opt_stripe) {
2248 sbi->s_stripe = arg;
2249 } else if (token == Opt_resuid) {
2250 uid = make_kuid(current_user_ns(), arg);
2251 if (!uid_valid(uid)) {
2252 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 2253 return -1;
2254 }
5f3633e3
JK		 2255 sbi->s_resuid = uid;
2256 } else if (token == Opt_resgid) {
2257 gid = make_kgid(current_user_ns(), arg);
2258 if (!gid_valid(gid)) {
2259 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2260 return -1;
2261 }
2262 sbi->s_resgid = gid;
2263 } else if (token == Opt_journal_dev) {
2264 if (is_remount) {
2265 ext4_msg(sb, KERN_ERR,
2266 "Cannot specify journal on remount");
2267 return -1;
2268 }
2269 *journal_devnum = arg;
ad4eec61
ES		 2270 } else if (token == Opt_journal_path) {
2271 char *journal_path;
2272 struct inode *journal_inode;
2273 struct path path;
2274 int error;
2275
2276 if (is_remount) {
2277 ext4_msg(sb, KERN_ERR,
2278 "Cannot specify journal on remount");
2279 return -1;
2280 }
2281 journal_path = match_strdup(&args[0]);
2282 if (!journal_path) {
2283 ext4_msg(sb, KERN_ERR, "error: could not dup "
2284 "journal device string");
2285 return -1;
2286 }
2287
2288 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2289 if (error) {
2290 ext4_msg(sb, KERN_ERR, "error: could not find "
2291 "journal device path: error %d", error);
2292 kfree(journal_path);
2293 return -1;
2294 }
2295
2b0143b5 2296 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 2297 if (!S_ISBLK(journal_inode->i_mode)) {
2298 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2299 "is not a block device", journal_path);
2300 path_put(&path);
2301 kfree(journal_path);
2302 return -1;
2303 }
2304
2305 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2306 path_put(&path);
2307 kfree(journal_path);
5f3633e3
JK		 2308 } else if (token == Opt_journal_ioprio) {
2309 if (arg > 7) {
2310 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2311 " (must be 0-7)");
2312 return -1;
2313 }
2314 *journal_ioprio =
2315 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447 2316 } else if (token == Opt_test_dummy_encryption) {
ed318a6c
EB		 2317 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2318 is_remount);
5f3633e3
JK		 2319 } else if (m->flags & MOPT_DATAJ) {
2320 if (is_remount) {
2321 if (!sbi->s_journal)
2322 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2323 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 2324 ext4_msg(sb, KERN_ERR,
26092bf5 2325 "Cannot change data mode on remount");
26092bf5 2326 return -1;
ac27a0ec 2327 }
26092bf5 2328 } else {
5f3633e3
JK		 2329 clear_opt(sb, DATA_FLAGS);
2330 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 2331 }
5f3633e3
JK		 2332#ifdef CONFIG_QUOTA
2333 } else if (m->flags & MOPT_QFMT) {
2334 if (sb_any_quota_loaded(sb) &&
2335 sbi->s_jquota_fmt != m->mount_opt) {
2336 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2337 "quota options when quota turned on");
2338 return -1;
2339 }
e2b911c5 2340 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 2341 ext4_msg(sb, KERN_INFO,
2342 "Quota format mount options ignored "
262b4662 2343 "when QUOTA feature is enabled");
c325a67c 2344 return 1;
262b4662 2345 }
5f3633e3 2346 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 2347#endif
9cb20f94
IW		 2348 } else if (token == Opt_dax || token == Opt_dax_always ||
2349 token == Opt_dax_inode || token == Opt_dax_never) {
ef83b6e8 2350#ifdef CONFIG_FS_DAX
9cb20f94
IW		 2351 switch (token) {
2352 case Opt_dax:
2353 case Opt_dax_always:
829b37b8
TT		 2354 if (is_remount &&
2355 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2356 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2357 fail_dax_change_remount:
2358 ext4_msg(sb, KERN_ERR, "can't change "
2359 "dax mount option while remounting");
2360 return -1;
2361 }
2362 if (is_remount &&
2363 (test_opt(sb, DATA_FLAGS) ==
2364 EXT4_MOUNT_JOURNAL_DATA)) {
2365 ext4_msg(sb, KERN_ERR, "can't mount with "
2366 "both data=journal and dax");
2367 return -1;
2368 }
9cb20f94
IW		 2369 ext4_msg(sb, KERN_WARNING,
2370 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2371 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2372 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2373 break;
2374 case Opt_dax_never:
829b37b8
TT		 2375 if (is_remount &&
2376 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2377 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2378 goto fail_dax_change_remount;
9cb20f94
IW		 2379 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2380 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2381 break;
2382 case Opt_dax_inode:
829b37b8
TT		 2383 if (is_remount &&
2384 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2385 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2386 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2387 goto fail_dax_change_remount;
9cb20f94
IW		 2388 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2389 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2390 /* Strictly for printing options */
2391 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2392 break;
2393 }
ef83b6e8 2394#else
923ae0ff 2395 ext4_msg(sb, KERN_INFO, "dax option not supported");
9cb20f94
IW		 2396 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2397 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
923ae0ff 2398 return -1;
5f3633e3 2399#endif
7915a861
AN		 2400 } else if (token == Opt_data_err_abort) {
2401 sbi->s_mount_opt |= m->mount_opt;
2402 } else if (token == Opt_data_err_ignore) {
2403 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 2404 } else {
2405 if (!args->from)
2406 arg = 1;
2407 if (m->flags & MOPT_CLEAR)
2408 arg = !arg;
2409 else if (unlikely(!(m->flags & MOPT_SET))) {
2410 ext4_msg(sb, KERN_WARNING,
2411 "buggy handling of option %s", opt);
2412 WARN_ON(1);
2413 return -1;
2414 }
995a3ed6
HS		 2415 if (m->flags & MOPT_2) {
2416 if (arg != 0)
2417 sbi->s_mount_opt2 |= m->mount_opt;
2418 else
2419 sbi->s_mount_opt2 &= ~m->mount_opt;
2420 } else {
2421 if (arg != 0)
2422 sbi->s_mount_opt |= m->mount_opt;
2423 else
2424 sbi->s_mount_opt &= ~m->mount_opt;
2425 }
26092bf5 2426 }
5f3633e3 2427 return 1;
26092bf5
TT		 2428}
2429
2430static int parse_options(char *options, struct super_block *sb,
2431 unsigned long *journal_devnum,
2432 unsigned int *journal_ioprio,
2433 int is_remount)
2434{
1e1a76ed 2435 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
33458eab 2436 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
26092bf5
TT		 2437 substring_t args[MAX_OPT_ARGS];
2438 int token;
2439
2440 if (!options)
2441 return 1;
2442
2443 while ((p = strsep(&options, ",")) != NULL) {
2444 if (!*p)
2445 continue;
2446 /*
2447 * Initialize args struct so we know whether arg was
2448 * found; some options take optional arguments.
2449 */
caecd0af 2450 args[0].to = args[0].from = NULL;
26092bf5
TT		 2451 token = match_token(p, tokens, args);
2452 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2453 journal_ioprio, is_remount) < 0)
2454 return 0;
ac27a0ec
DK		 2455 }
2456#ifdef CONFIG_QUOTA
49da9392
JK		 2457 /*
2458 * We do the test below only for project quotas. 'usrquota' and
2459 * 'grpquota' mount options are allowed even without quota feature
2460 * to support legacy quotas in quota files.
2461 */
2462 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2463 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2464 "Cannot enable project quota enforcement.");
2465 return 0;
2466 }
33458eab
TT		 2467 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2468 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2469 if (usr_qf_name || grp_qf_name) {
2470 if (test_opt(sb, USRQUOTA) && usr_qf_name)
fd8c37ec 2471 clear_opt(sb, USRQUOTA);
ac27a0ec 2472
33458eab 2473 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
fd8c37ec 2474 clear_opt(sb, GRPQUOTA);
ac27a0ec 2475
56c50f11 2476 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 2477 ext4_msg(sb, KERN_ERR, "old and new quota "
2478 "format mixing");
ac27a0ec
DK		 2479 return 0;
2480 }
2481
2482 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 2483 ext4_msg(sb, KERN_ERR, "journaled quota format "
2484 "not specified");
ac27a0ec
DK		 2485 return 0;
2486 }
ac27a0ec
DK		 2487 }
2488#endif
626b035b
RH		 2489 if (test_opt(sb, DIOREAD_NOLOCK)) {
2490 int blocksize =
2491 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2492 if (blocksize < PAGE_SIZE)
2493 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2494 "experimental mount option 'dioread_nolock' "
2495 "for blocksize < PAGE_SIZE");
2496 }
ac27a0ec
DK		 2497 return 1;
2498}
2499
2adf6da8
TT		 2500static inline void ext4_show_quota_options(struct seq_file *seq,
2501 struct super_block *sb)
2502{
2503#if defined(CONFIG_QUOTA)
2504 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 2505 char *usr_qf_name, *grp_qf_name;
2adf6da8
TT		 2506
2507 if (sbi->s_jquota_fmt) {
2508 char *fmtname = "";
2509
2510 switch (sbi->s_jquota_fmt) {
2511 case QFMT_VFS_OLD:
2512 fmtname = "vfsold";
2513 break;
2514 case QFMT_VFS_V0:
2515 fmtname = "vfsv0";
2516 break;
2517 case QFMT_VFS_V1:
2518 fmtname = "vfsv1";
2519 break;
2520 }
2521 seq_printf(seq, ",jqfmt=%s", fmtname);
2522 }
2523
33458eab
TT		 2524 rcu_read_lock();
2525 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2526 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2527 if (usr_qf_name)
2528 seq_show_option(seq, "usrjquota", usr_qf_name);
2529 if (grp_qf_name)
2530 seq_show_option(seq, "grpjquota", grp_qf_name);
2531 rcu_read_unlock();
2adf6da8
TT		 2532#endif
2533}
2534
5a916be1
TT		 2535static const char *token2str(int token)
2536{
50df9fd5 2537 const struct match_token *t;
5a916be1
TT		 2538
2539 for (t = tokens; t->token != Opt_err; t++)
2540 if (t->token == token && !strchr(t->pattern, '='))
2541 break;
2542 return t->pattern;
2543}
2544
2adf6da8
TT		 2545/*
2546 * Show an option if
2547 * - it's set to a non-default value OR
2548 * - if the per-sb default is different from the global default
2549 */
66acdcf4
TT		 2550static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2551 int nodefs)
2adf6da8 2552{
2adf6da8
TT		 2553 struct ext4_sb_info *sbi = EXT4_SB(sb);
2554 struct ext4_super_block *es = sbi->s_es;
68afa7e0 2555 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
5a916be1 2556 const struct mount_opts *m;
66acdcf4 2557 char sep = nodefs ? '\n' : ',';
2adf6da8 2558
66acdcf4
TT		 2559#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2560#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2561
2562 if (sbi->s_sb_block != 1)
5a916be1
TT		 2563 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2564
2565 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2566 int want_set = m->flags & MOPT_SET;
2567 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
9cb20f94 2568 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
5a916be1 2569 continue;
68afa7e0 2570 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2571 continue; /* skip if same as the default */
2572 if ((want_set &&
2573 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2574 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2575 continue; /* select Opt_noFoo vs Opt_Foo */
2576 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2577 }
5a916be1 2578
08cefc7a 2579 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2580 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2581 SEQ_OPTS_PRINT("resuid=%u",
2582 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2583 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2584 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2585 SEQ_OPTS_PRINT("resgid=%u",
2586 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2587 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2588 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2589 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2590 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2591 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2592 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2593 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2594 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2595 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2596 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2597 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2598 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2599 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
357fdad0 2600 if (sb->s_flags & SB_I_VERSION)
5a916be1 2601 SEQ_OPTS_PUTS("i_version");
66acdcf4 2602 if (nodefs || sbi->s_stripe)
5a916be1 2603 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0
TN		 2604 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2605 (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2606 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2607 SEQ_OPTS_PUTS("data=journal");
2608 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2609 SEQ_OPTS_PUTS("data=ordered");
2610 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2611 SEQ_OPTS_PUTS("data=writeback");
2612 }
66acdcf4
TT		 2613 if (nodefs ||
2614 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2615 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2616 sbi->s_inode_readahead_blks);
2adf6da8 2617
ceec0376 2618 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 2619 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2620 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2621 if (nodefs || sbi->s_max_dir_size_kb)
2622 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2623 if (test_opt(sb, DATA_ERR_ABORT))
2624 SEQ_OPTS_PUTS("data_err=abort");
ed318a6c
EB		 2625
2626 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2adf6da8 2627
4f74d15f
EB		 2628 if (sb->s_flags & SB_INLINECRYPT)
2629 SEQ_OPTS_PUTS("inlinecrypt");
2630
9cb20f94
IW		 2631 if (test_opt(sb, DAX_ALWAYS)) {
2632 if (IS_EXT2_SB(sb))
2633 SEQ_OPTS_PUTS("dax");
2634 else
2635 SEQ_OPTS_PUTS("dax=always");
2636 } else if (test_opt2(sb, DAX_NEVER)) {
2637 SEQ_OPTS_PUTS("dax=never");
2638 } else if (test_opt2(sb, DAX_INODE)) {
2639 SEQ_OPTS_PUTS("dax=inode");
2640 }
2adf6da8 2641 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2642 return 0;
2643}
2644
66acdcf4
TT		 2645static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2646{
2647 return _ext4_show_options(seq, root->d_sb, 0);
2648}
2649
ebd173be 2650int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2651{
2652 struct super_block *sb = seq->private;
2653 int rc;
2654
bc98a42c 2655 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 2656 rc = _ext4_show_options(seq, sb, 1);
2657 seq_puts(seq, "\n");
2658 return rc;
2659}
2660
617ba13b 2661static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2662 int read_only)
2663{
617ba13b 2664 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 2665 int err = 0;
ac27a0ec 2666
617ba13b 2667 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2668 ext4_msg(sb, KERN_ERR, "revision level too high, "
2669 "forcing read-only mode");
c89128a0 2670 err = -EROFS;
5adaccac 2671 goto done;
ac27a0ec
DK		 2672 }
2673 if (read_only)
281b5995 2674 goto done;
617ba13b 2675 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2676 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2677 "running e2fsck is recommended");
c8b459f4 2678 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2679 ext4_msg(sb, KERN_WARNING,
2680 "warning: mounting fs with errors, "
2681 "running e2fsck is recommended");
ed3ce80a 2682 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2683 le16_to_cpu(es->s_mnt_count) >=
2684 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2685 ext4_msg(sb, KERN_WARNING,
2686 "warning: maximal mount count reached, "
2687 "running e2fsck is recommended");
ac27a0ec 2688 else if (le32_to_cpu(es->s_checkinterval) &&
6a0678a7
AB		 2689 (ext4_get_tstamp(es, s_lastcheck) +
2690 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
b31e1552
ES		 2691 ext4_msg(sb, KERN_WARNING,
2692 "warning: checktime reached, "
2693 "running e2fsck is recommended");
0b8e58a1 2694 if (!sbi->s_journal)
0390131b 2695 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2696 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2697 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2698 le16_add_cpu(&es->s_mnt_count, 1);
6a0678a7 2699 ext4_update_tstamp(es, s_mtime);
0390131b 2700 if (sbi->s_journal)
e2b911c5 2701 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2702
c89128a0 2703 err = ext4_commit_super(sb, 1);
281b5995 2704done:
ac27a0ec 2705 if (test_opt(sb, DEBUG))
a9df9a49 2706 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2707 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2708 sb->s_blocksize,
2709 sbi->s_groups_count,
617ba13b
MC		 2710 EXT4_BLOCKS_PER_GROUP(sb),
2711 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2712 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2713
7abc52c2 2714 cleancache_init_fs(sb);
c89128a0 2715 return err;
ac27a0ec
DK		 2716}
2717
117fff10
TT		 2718int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2719{
2720 struct ext4_sb_info *sbi = EXT4_SB(sb);
7c990728 2721 struct flex_groups **old_groups, **new_groups;
37b0b6b8 2722 int size, i, j;
117fff10
TT		 2723
2724 if (!sbi->s_log_groups_per_flex)
2725 return 0;
2726
2727 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2728 if (size <= sbi->s_flex_groups_allocated)
2729 return 0;
2730
7c990728
SJS		 2731 new_groups = kvzalloc(roundup_pow_of_two(size *
2732 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
117fff10 2733 if (!new_groups) {
7c990728
SJS		 2734 ext4_msg(sb, KERN_ERR,
2735 "not enough memory for %d flex group pointers", size);
117fff10
TT		 2736 return -ENOMEM;
2737 }
7c990728
SJS		 2738 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2739 new_groups[i] = kvzalloc(roundup_pow_of_two(
2740 sizeof(struct flex_groups)),
2741 GFP_KERNEL);
2742 if (!new_groups[i]) {
37b0b6b8
DC		 2743 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2744 kvfree(new_groups[j]);
7c990728
SJS		 2745 kvfree(new_groups);
2746 ext4_msg(sb, KERN_ERR,
2747 "not enough memory for %d flex groups", size);
2748 return -ENOMEM;
2749 }
117fff10 2750 }
7c990728
SJS		 2751 rcu_read_lock();
2752 old_groups = rcu_dereference(sbi->s_flex_groups);
2753 if (old_groups)
2754 memcpy(new_groups, old_groups,
2755 (sbi->s_flex_groups_allocated *
2756 sizeof(struct flex_groups *)));
2757 rcu_read_unlock();
2758 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2759 sbi->s_flex_groups_allocated = size;
2760 if (old_groups)
2761 ext4_kvfree_array_rcu(old_groups);
117fff10
TT		 2762 return 0;
2763}
2764
772cb7c8
JS		 2765static int ext4_fill_flex_info(struct super_block *sb)
2766{
2767 struct ext4_sb_info *sbi = EXT4_SB(sb);
2768 struct ext4_group_desc *gdp = NULL;
7c990728 2769 struct flex_groups *fg;
772cb7c8 2770 ext4_group_t flex_group;
117fff10 2771 int i, err;
772cb7c8 2772
503358ae 2773 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2774 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2775 sbi->s_log_groups_per_flex = 0;
2776 return 1;
2777 }
2778
117fff10
TT		 2779 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2780 if (err)
9933fc0a 2781 goto failed;
772cb7c8 2782
772cb7c8 2783 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2784 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2785
2786 flex_group = ext4_flex_group(sbi, i);
7c990728
SJS		 2787 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2788 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
90ba983f 2789 atomic64_add(ext4_free_group_clusters(sb, gdp),
7c990728
SJS		 2790 &fg->free_clusters);
2791 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
772cb7c8
JS		 2792 }
2793
2794 return 1;
2795failed:
2796 return 0;
2797}
2798
e2b911c5 2799static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2800 struct ext4_group_desc *gdp)
717d50e4 2801{
b47820ed 2802 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2803 __u16 crc = 0;
feb0ab32 2804 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2805 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2806
9aa5d32b 2807 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2808 /* Use new metadata_csum algorithm */
feb0ab32 2809 __u32 csum32;
b47820ed 2810 __u16 dummy_csum = 0;
feb0ab32 2811
feb0ab32
DW		 2812 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2813 sizeof(le_group));
b47820ed
DJ		 2814 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2815 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2816 sizeof(dummy_csum));
2817 offset += sizeof(dummy_csum);
2818 if (offset < sbi->s_desc_size)
2819 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2820 sbi->s_desc_size - offset);
feb0ab32
DW		 2821
2822 crc = csum32 & 0xFFFF;
2823 goto out;
717d50e4
AD		 2824 }
2825
feb0ab32 2826 /* old crc16 code */
e2b911c5 2827 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2828 return 0;
2829
feb0ab32
DW		 2830 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2831 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2832 crc = crc16(crc, (__u8 *)gdp, offset);
2833 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2834 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2835 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2836 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2837 crc = crc16(crc, (__u8 *)gdp + offset,
2838 le16_to_cpu(sbi->s_es->s_desc_size) -
2839 offset);
2840
2841out:
717d50e4
AD		 2842 return cpu_to_le16(crc);
2843}
2844
feb0ab32 2845int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2846 struct ext4_group_desc *gdp)
2847{
feb0ab32 2848 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2849 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2850 return 0;
2851
2852 return 1;
2853}
2854
feb0ab32
DW		 2855void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2856 struct ext4_group_desc *gdp)
2857{
2858 if (!ext4_has_group_desc_csum(sb))
2859 return;
e2b911c5 2860 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2861}
2862
ac27a0ec 2863/* Called at mount-time, super-block is locked */
bfff6873 2864static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2865 ext4_fsblk_t sb_block,
bfff6873 2866 ext4_group_t *first_not_zeroed)
ac27a0ec 2867{
617ba13b
MC		 2868 struct ext4_sb_info *sbi = EXT4_SB(sb);
2869 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2870 ext4_fsblk_t last_block;
44de022c 2871 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
bd81d8ee
LV		 2872 ext4_fsblk_t block_bitmap;
2873 ext4_fsblk_t inode_bitmap;
2874 ext4_fsblk_t inode_table;
ce421581 2875 int flexbg_flag = 0;
bfff6873 2876 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2877
e2b911c5 2878 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2879 flexbg_flag = 1;
2880
af5bc92d 2881 ext4_debug("Checking group descriptors");
ac27a0ec 2882
197cd65a
AM		 2883 for (i = 0; i < sbi->s_groups_count; i++) {
2884 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2885
ce421581 2886 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2887 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2888 else
2889 last_block = first_block +
617ba13b 2890 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2891
bfff6873
LC		 2892 if ((grp == sbi->s_groups_count) &&
2893 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2894 grp = i;
2895
8fadc143 2896 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2897 if (block_bitmap == sb_block) {
2898 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2899 "Block bitmap for group %u overlaps "
2900 "superblock", i);
18db4b4e
TT		 2901 if (!sb_rdonly(sb))
2902 return 0;
829fa70d 2903 }
77260807
TT		 2904 if (block_bitmap >= sb_block + 1 &&
2905 block_bitmap <= last_bg_block) {
2906 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2907 "Block bitmap for group %u overlaps "
2908 "block group descriptors", i);
2909 if (!sb_rdonly(sb))
2910 return 0;
2911 }
2b2d6d01 2912 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2913 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2914 "Block bitmap for group %u not in group "
b31e1552 2915 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2916 return 0;
2917 }
8fadc143 2918 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2919 if (inode_bitmap == sb_block) {
2920 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2921 "Inode bitmap for group %u overlaps "
2922 "superblock", i);
18db4b4e
TT		 2923 if (!sb_rdonly(sb))
2924 return 0;
829fa70d 2925 }
77260807
TT		 2926 if (inode_bitmap >= sb_block + 1 &&
2927 inode_bitmap <= last_bg_block) {
2928 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2929 "Inode bitmap for group %u overlaps "
2930 "block group descriptors", i);
2931 if (!sb_rdonly(sb))
2932 return 0;
2933 }
2b2d6d01 2934 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2935 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2936 "Inode bitmap for group %u not in group "
b31e1552 2937 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2938 return 0;
2939 }
8fadc143 2940 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2941 if (inode_table == sb_block) {
2942 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2943 "Inode table for group %u overlaps "
2944 "superblock", i);
18db4b4e
TT		 2945 if (!sb_rdonly(sb))
2946 return 0;
829fa70d 2947 }
77260807
TT		 2948 if (inode_table >= sb_block + 1 &&
2949 inode_table <= last_bg_block) {
2950 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2951 "Inode table for group %u overlaps "
2952 "block group descriptors", i);
2953 if (!sb_rdonly(sb))
2954 return 0;
2955 }
bd81d8ee 2956 if (inode_table < first_block ||
2b2d6d01 2957 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2958 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2959 "Inode table for group %u not in group "
b31e1552 2960 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2961 return 0;
2962 }
955ce5f5 2963 ext4_lock_group(sb, i);
feb0ab32 2964 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2965 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2966 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2967 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2968 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 2969 if (!sb_rdonly(sb)) {
955ce5f5 2970 ext4_unlock_group(sb, i);
8a266467 2971 return 0;
7ee1ec4c 2972 }
717d50e4 2973 }
955ce5f5 2974 ext4_unlock_group(sb, i);
ce421581
JS		 2975 if (!flexbg_flag)
2976 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2977 }
bfff6873
LC		 2978 if (NULL != first_not_zeroed)
2979 *first_not_zeroed = grp;
ac27a0ec
DK		 2980 return 1;
2981}
2982
617ba13b 2983/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2984 * the superblock) which were deleted from all directories, but held open by
2985 * a process at the time of a crash. We walk the list and try to delete these
2986 * inodes at recovery time (only with a read-write filesystem).
2987 *
2988 * In order to keep the orphan inode chain consistent during traversal (in
2989 * case of crash during recovery), we link each inode into the superblock
2990 * orphan list_head and handle it the same way as an inode deletion during
2991 * normal operation (which journals the operations for us).
2992 *
2993 * We only do an iget() and an iput() on each inode, which is very safe if we
2994 * accidentally point at an in-use or already deleted inode. The worst that
2995 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2996 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2997 * e2fsck was run on this filesystem, and it must have already done the orphan
2998 * inode cleanup for us, so we can safely abort without any further action.
2999 */
2b2d6d01
TT		 3000static void ext4_orphan_cleanup(struct super_block *sb,
3001 struct ext4_super_block *es)
ac27a0ec
DK		 3002{
3003 unsigned int s_flags = sb->s_flags;
2c98eb5e 3004 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec 3005#ifdef CONFIG_QUOTA
95f1fda4 3006 int quota_update = 0;
ac27a0ec
DK		 3007 int i;
3008#endif
3009 if (!es->s_last_orphan) {
3010 jbd_debug(4, "no orphan inodes to clean up\n");
3011 return;
3012 }
3013
a8f48a95 3014 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 3015 ext4_msg(sb, KERN_ERR, "write access "
3016 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 3017 return;
3018 }
3019
d39195c3
AG		 3020 /* Check if feature set would not allow a r/w mount */
3021 if (!ext4_feature_set_ok(sb, 0)) {
3022 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
3023 "unknown ROCOMPAT features");
3024 return;
3025 }
3026
617ba13b 3027 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6 3028 /* don't clear list on RO mount w/ errors */
1751e8a6 3029 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
84474976 3030 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 3031 "clearing orphan list.\n");
c25f9bc6
ES		 3032 es->s_last_orphan = 0;
3033 }
ac27a0ec
DK		 3034 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3035 return;
3036 }
3037
1751e8a6 3038 if (s_flags & SB_RDONLY) {
b31e1552 3039 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1751e8a6 3040 sb->s_flags &= ~SB_RDONLY;
ac27a0ec
DK		 3041 }
3042#ifdef CONFIG_QUOTA
3043 /* Needed for iput() to work correctly and not trash data */
1751e8a6 3044 sb->s_flags |= SB_ACTIVE;
95f1fda4 3045
3046 /*
3047 * Turn on quotas which were not enabled for read-only mounts if
3048 * filesystem has quota feature, so that they are updated correctly.
3049 */
1751e8a6 3050 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
95f1fda4 3051 int ret = ext4_enable_quotas(sb);
3052
3053 if (!ret)
3054 quota_update = 1;
3055 else
3056 ext4_msg(sb, KERN_ERR,
3057 "Cannot turn on quotas: error %d", ret);
3058 }
3059
3060 /* Turn on journaled quotas used for old sytle */
a2d4a646 3061 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 3062 if (EXT4_SB(sb)->s_qf_names[i]) {
3063 int ret = ext4_quota_on_mount(sb, i);
95f1fda4 3064
3065 if (!ret)
3066 quota_update = 1;
3067 else
b31e1552
ES		 3068 ext4_msg(sb, KERN_ERR,
3069 "Cannot turn on journaled "
95f1fda4 3070 "quota: type %d: error %d", i, ret);
ac27a0ec
DK		 3071 }
3072 }
3073#endif
3074
3075 while (es->s_last_orphan) {
3076 struct inode *inode;
3077
c65d5c6c
VN		 3078 /*
3079 * We may have encountered an error during cleanup; if
3080 * so, skip the rest.
3081 */
3082 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3083 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3084 es->s_last_orphan = 0;
3085 break;
3086 }
3087
97bd42b9
JB		 3088 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
3089 if (IS_ERR(inode)) {
ac27a0ec
DK		 3090 es->s_last_orphan = 0;
3091 break;
3092 }
3093
617ba13b 3094 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 3095 dquot_initialize(inode);
ac27a0ec 3096 if (inode->i_nlink) {
566370a2
PT		 3097 if (test_opt(sb, DEBUG))
3098 ext4_msg(sb, KERN_DEBUG,
3099 "%s: truncating inode %lu to %lld bytes",
3100 __func__, inode->i_ino, inode->i_size);
e5f8eab8 3101 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 3102 inode->i_ino, inode->i_size);
5955102c 3103 inode_lock(inode);
55f252c9 3104 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 3105 ret = ext4_truncate(inode);
3106 if (ret)
3107 ext4_std_error(inode->i_sb, ret);
5955102c 3108 inode_unlock(inode);
ac27a0ec
DK		 3109 nr_truncates++;
3110 } else {
566370a2
PT		 3111 if (test_opt(sb, DEBUG))
3112 ext4_msg(sb, KERN_DEBUG,
3113 "%s: deleting unreferenced inode %lu",
3114 __func__, inode->i_ino);
ac27a0ec
DK		 3115 jbd_debug(2, "deleting unreferenced inode %lu\n",
3116 inode->i_ino);
3117 nr_orphans++;
3118 }
3119 iput(inode); /* The delete magic happens here! */
3120 }
3121
2b2d6d01 3122#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 3123
3124 if (nr_orphans)
b31e1552
ES		 3125 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
3126 PLURAL(nr_orphans));
ac27a0ec 3127 if (nr_truncates)
b31e1552
ES		 3128 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3129 PLURAL(nr_truncates));
ac27a0ec 3130#ifdef CONFIG_QUOTA
95f1fda4 3131 /* Turn off quotas if they were enabled for orphan cleanup */
3132 if (quota_update) {
3133 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3134 if (sb_dqopt(sb)->files[i])
3135 dquot_quota_off(sb, i);
3136 }
ac27a0ec
DK		 3137 }
3138#endif
1751e8a6 3139 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
ac27a0ec 3140}
0b8e58a1 3141
cd2291a4
ES		 3142/*
3143 * Maximal extent format file size.
3144 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3145 * extent format containers, within a sector_t, and within i_blocks
3146 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3147 * so that won't be a limiting factor.
3148 *
f17722f9
LC		 3149 * However there is other limiting factor. We do store extents in the form
3150 * of starting block and length, hence the resulting length of the extent
3151 * covering maximum file size must fit into on-disk format containers as
3152 * well. Given that length is always by 1 unit bigger than max unit (because
3153 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3154 *
cd2291a4
ES		 3155 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3156 */
f287a1a5 3157static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 3158{
3159 loff_t res;
3160 loff_t upper_limit = MAX_LFS_FILESIZE;
3161
72deb455
CH		 3162 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3163
3164 if (!has_huge_files) {
cd2291a4
ES		 3165 upper_limit = (1LL << 32) - 1;
3166
3167 /* total blocks in file system block size */
3168 upper_limit >>= (blkbits - 9);
3169 upper_limit <<= blkbits;
3170 }
3171
f17722f9
LC		 3172 /*
3173 * 32-bit extent-start container, ee_block. We lower the maxbytes
3174 * by one fs block, so ee_len can cover the extent of maximum file
3175 * size
3176 */
3177 res = (1LL << 32) - 1;
cd2291a4 3178 res <<= blkbits;
cd2291a4
ES		 3179
3180 /* Sanity check against vm- & vfs- imposed limits */
3181 if (res > upper_limit)
3182 res = upper_limit;
3183
3184 return res;
3185}
ac27a0ec 3186
ac27a0ec 3187/*
cd2291a4 3188 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 3189 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3190 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 3191 */
f287a1a5 3192static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 3193{
617ba13b 3194 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 3195 int meta_blocks;
3196 loff_t upper_limit;
0b8e58a1
AD		 3197 /* This is calculated to be the largest file size for a dense, block
3198 * mapped file such that the file's total number of 512-byte sectors,
3199 * including data and all indirect blocks, does not exceed (2^48 - 1).
3200 *
3201 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3202 * number of 512-byte sectors of the file.
0fc1b451
AK		 3203 */
3204
72deb455 3205 if (!has_huge_files) {
0fc1b451 3206 /*
72deb455
CH		 3207 * !has_huge_files or implies that the inode i_block field
3208 * represents total file blocks in 2^32 512-byte sectors ==
3209 * size of vfs inode i_blocks * 8
0fc1b451
AK		 3210 */
3211 upper_limit = (1LL << 32) - 1;
3212
3213 /* total blocks in file system block size */
3214 upper_limit >>= (bits - 9);
3215
3216 } else {
8180a562
AK		 3217 /*
3218 * We use 48 bit ext4_inode i_blocks
3219 * With EXT4_HUGE_FILE_FL set the i_blocks
3220 * represent total number of blocks in
3221 * file system block size
3222 */
0fc1b451
AK		 3223 upper_limit = (1LL << 48) - 1;
3224
0fc1b451
AK		 3225 }
3226
3227 /* indirect blocks */
3228 meta_blocks = 1;
3229 /* double indirect blocks */
3230 meta_blocks += 1 + (1LL << (bits-2));
3231 /* tripple indirect blocks */
3232 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3233
3234 upper_limit -= meta_blocks;
3235 upper_limit <<= bits;
ac27a0ec
DK		 3236
3237 res += 1LL << (bits-2);
3238 res += 1LL << (2*(bits-2));
3239 res += 1LL << (3*(bits-2));
3240 res <<= bits;
3241 if (res > upper_limit)
3242 res = upper_limit;
0fc1b451
AK		 3243
3244 if (res > MAX_LFS_FILESIZE)
3245 res = MAX_LFS_FILESIZE;
3246
ac27a0ec
DK		 3247 return res;
3248}
3249
617ba13b 3250static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 3251 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 3252{
617ba13b 3253 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 3254 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 3255 int has_super = 0;
3256
3257 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3258
e2b911c5 3259 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 3260 return logical_sb_block + nr + 1;
ac27a0ec 3261 bg = sbi->s_desc_per_block * nr;
617ba13b 3262 if (ext4_bg_has_super(sb, bg))
ac27a0ec 3263 has_super = 1;
0b8e58a1 3264
bd63f6b0
DW		 3265 /*
3266 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3267 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3268 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3269 * compensate.
3270 */
3271 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 3272 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 3273 has_super++;
3274
617ba13b 3275 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 3276}
3277
c9de560d
AT		 3278/**
3279 * ext4_get_stripe_size: Get the stripe size.
3280 * @sbi: In memory super block info
3281 *
3282 * If we have specified it via mount option, then
3283 * use the mount option value. If the value specified at mount time is
3284 * greater than the blocks per group use the super block value.
3285 * If the super block value is greater than blocks per group return 0.
3286 * Allocator needs it be less than blocks per group.
3287 *
3288 */
3289static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3290{
3291 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3292 unsigned long stripe_width =
3293 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 3294 int ret;
c9de560d
AT		 3295
3296 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 3297 ret = sbi->s_stripe;
5469d7c3 3298 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 3299 ret = stripe_width;
5469d7c3 3300 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 3301 ret = stride;
3302 else
3303 ret = 0;
c9de560d 3304
3eb08658
DE		 3305 /*
3306 * If the stripe width is 1, this makes no sense and
3307 * we set it to 0 to turn off stripe handling code.
3308 */
3309 if (ret <= 1)
3310 ret = 0;
c9de560d 3311
3eb08658 3312 return ret;
c9de560d 3313}
ac27a0ec 3314
a13fb1a4
ES		 3315/*
3316 * Check whether this filesystem can be mounted based on
3317 * the features present and the RDONLY/RDWR mount requested.
3318 * Returns 1 if this filesystem can be mounted as requested,
3319 * 0 if it cannot be.
3320 */
3321static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3322{
e2b911c5 3323 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 3324 ext4_msg(sb, KERN_ERR,
3325 "Couldn't mount because of "
3326 "unsupported optional features (%x)",
3327 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3328 ~EXT4_FEATURE_INCOMPAT_SUPP));
3329 return 0;
3330 }
3331
c83ad55e
GKB		 3332#ifndef CONFIG_UNICODE
3333 if (ext4_has_feature_casefold(sb)) {
3334 ext4_msg(sb, KERN_ERR,
3335 "Filesystem with casefold feature cannot be "
3336 "mounted without CONFIG_UNICODE");
3337 return 0;
3338 }
3339#endif
3340
a13fb1a4
ES		 3341 if (readonly)
3342 return 1;
3343
e2b911c5 3344 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 3345 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 3346 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 3347 return 1;
3348 }
3349
a13fb1a4 3350 /* Check that feature set is OK for a read-write mount */
e2b911c5 3351 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 3352 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3353 "unsupported optional features (%x)",
3354 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3355 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3356 return 0;
3357 }
e2b911c5 3358 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 3359 ext4_msg(sb, KERN_ERR,
3360 "Can't support bigalloc feature without "
3361 "extents feature\n");
3362 return 0;
3363 }
7c319d32 3364
9db176bc 3365#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
d65d87a0
TT		 3366 if (!readonly && (ext4_has_feature_quota(sb) ||
3367 ext4_has_feature_project(sb))) {
7c319d32 3368 ext4_msg(sb, KERN_ERR,
d65d87a0 3369 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
689c958c
LX		 3370 return 0;
3371 }
7c319d32 3372#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 3373 return 1;
3374}
3375
66e61a9e
TT		 3376/*
3377 * This function is called once a day if we have errors logged
3378 * on the file system
3379 */
235699a8 3380static void print_daily_error_info(struct timer_list *t)
66e61a9e 3381{
235699a8
KC		 3382 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3383 struct super_block *sb = sbi->s_sb;
3384 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 3385
3386 if (es->s_error_count)
ae0f78de
TT		 3387 /* fsck newer than v1.41.13 is needed to clean this condition. */
3388 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 3389 le32_to_cpu(es->s_error_count));
3390 if (es->s_first_error_time) {
6a0678a7
AB		 3391 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3392 sb->s_id,
3393 ext4_get_tstamp(es, s_first_error_time),
66e61a9e
TT		 3394 (int) sizeof(es->s_first_error_func),
3395 es->s_first_error_func,
3396 le32_to_cpu(es->s_first_error_line));
3397 if (es->s_first_error_ino)
651e1c3b 3398 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3399 le32_to_cpu(es->s_first_error_ino));
3400 if (es->s_first_error_block)
651e1c3b 3401 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3402 le64_to_cpu(es->s_first_error_block));
651e1c3b 3403 printk(KERN_CONT "\n");
66e61a9e
TT		 3404 }
3405 if (es->s_last_error_time) {
6a0678a7
AB		 3406 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3407 sb->s_id,
3408 ext4_get_tstamp(es, s_last_error_time),
66e61a9e
TT		 3409 (int) sizeof(es->s_last_error_func),
3410 es->s_last_error_func,
3411 le32_to_cpu(es->s_last_error_line));
3412 if (es->s_last_error_ino)
651e1c3b 3413 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3414 le32_to_cpu(es->s_last_error_ino));
3415 if (es->s_last_error_block)
651e1c3b 3416 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3417 le64_to_cpu(es->s_last_error_block));
651e1c3b 3418 printk(KERN_CONT "\n");
66e61a9e
TT		 3419 }
3420 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3421}
3422
bfff6873
LC		 3423/* Find next suitable group and run ext4_init_inode_table */
3424static int ext4_run_li_request(struct ext4_li_request *elr)
3425{
3426 struct ext4_group_desc *gdp = NULL;
3d392b26
TT		 3427 struct super_block *sb = elr->lr_super;
3428 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3429 ext4_group_t group = elr->lr_next_group;
bfff6873 3430 unsigned long timeout = 0;
3d392b26 3431 unsigned int prefetch_ios = 0;
bfff6873
LC		 3432 int ret = 0;
3433
3d392b26
TT		 3434 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3435 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3436 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3437 if (prefetch_ios)
3438 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3439 prefetch_ios);
3440 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3441 prefetch_ios);
3442 if (group >= elr->lr_next_group) {
3443 ret = 1;
3444 if (elr->lr_first_not_zeroed != ngroups &&
3445 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3446 elr->lr_next_group = elr->lr_first_not_zeroed;
3447 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3448 ret = 0;
3449 }
3450 }
3451 return ret;
3452 }
bfff6873 3453
3d392b26 3454 for (; group < ngroups; group++) {
bfff6873
LC		 3455 gdp = ext4_get_group_desc(sb, group, NULL);
3456 if (!gdp) {
3457 ret = 1;
3458 break;
3459 }
3460
3461 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3462 break;
3463 }
3464
7f511862 3465 if (group >= ngroups)
bfff6873
LC		 3466 ret = 1;
3467
3468 if (!ret) {
3469 timeout = jiffies;
3470 ret = ext4_init_inode_table(sb, group,
3471 elr->lr_timeout ? 0 : 1);
3d392b26 3472 trace_ext4_lazy_itable_init(sb, group);
bfff6873 3473 if (elr->lr_timeout == 0) {
51ce6511 3474 timeout = (jiffies - timeout) *
3d392b26 3475 EXT4_SB(elr->lr_super)->s_li_wait_mult;
bfff6873
LC		 3476 elr->lr_timeout = timeout;
3477 }
3478 elr->lr_next_sched = jiffies + elr->lr_timeout;
3479 elr->lr_next_group = group + 1;
3480 }
bfff6873
LC		 3481 return ret;
3482}
3483
3484/*
3485 * Remove lr_request from the list_request and free the
4ed5c033 3486 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 3487 */
3488static void ext4_remove_li_request(struct ext4_li_request *elr)
3489{
bfff6873
LC		 3490 if (!elr)
3491 return;
3492
bfff6873 3493 list_del(&elr->lr_request);
3d392b26 3494 EXT4_SB(elr->lr_super)->s_li_request = NULL;
bfff6873
LC		 3495 kfree(elr);
3496}
3497
3498static void ext4_unregister_li_request(struct super_block *sb)
3499{
1bb933fb
LC		 3500 mutex_lock(&ext4_li_mtx);
3501 if (!ext4_li_info) {
3502 mutex_unlock(&ext4_li_mtx);
bfff6873 3503 return;
1bb933fb 3504 }
bfff6873
LC		 3505
3506 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 3507 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 3508 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 3509 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 3510}
3511
8f1f7453
ES		 3512static struct task_struct *ext4_lazyinit_task;
3513
bfff6873
LC		 3514/*
3515 * This is the function where ext4lazyinit thread lives. It walks
3516 * through the request list searching for next scheduled filesystem.
3517 * When such a fs is found, run the lazy initialization request
3518 * (ext4_rn_li_request) and keep track of the time spend in this
3519 * function. Based on that time we compute next schedule time of
3520 * the request. When walking through the list is complete, compute
3521 * next waking time and put itself into sleep.
3522 */
3523static int ext4_lazyinit_thread(void *arg)
3524{
3525 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3526 struct list_head *pos, *n;
3527 struct ext4_li_request *elr;
4ed5c033 3528 unsigned long next_wakeup, cur;
bfff6873
LC		 3529
3530 BUG_ON(NULL == eli);
3531
bfff6873
LC		 3532cont_thread:
3533 while (true) {
3534 next_wakeup = MAX_JIFFY_OFFSET;
3535
3536 mutex_lock(&eli->li_list_mtx);
3537 if (list_empty(&eli->li_request_list)) {
3538 mutex_unlock(&eli->li_list_mtx);
3539 goto exit_thread;
3540 }
bfff6873 3541 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3542 int err = 0;
3543 int progress = 0;
bfff6873
LC		 3544 elr = list_entry(pos, struct ext4_li_request,
3545 lr_request);
3546
e22834f0
DM		 3547 if (time_before(jiffies, elr->lr_next_sched)) {
3548 if (time_before(elr->lr_next_sched, next_wakeup))
3549 next_wakeup = elr->lr_next_sched;
3550 continue;
3551 }
3552 if (down_read_trylock(&elr->lr_super->s_umount)) {
3553 if (sb_start_write_trylock(elr->lr_super)) {
3554 progress = 1;
3555 /*
3556 * We hold sb->s_umount, sb can not
3557 * be removed from the list, it is
3558 * now safe to drop li_list_mtx
3559 */
3560 mutex_unlock(&eli->li_list_mtx);
3561 err = ext4_run_li_request(elr);
3562 sb_end_write(elr->lr_super);
3563 mutex_lock(&eli->li_list_mtx);
3564 n = pos->next;
b2c78cd0 3565 }
e22834f0
DM		 3566 up_read((&elr->lr_super->s_umount));
3567 }
3568 /* error, remove the lazy_init job */
3569 if (err) {
3570 ext4_remove_li_request(elr);
3571 continue;
3572 }
3573 if (!progress) {
3574 elr->lr_next_sched = jiffies +
3575 (prandom_u32()
3576 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 3577 }
bfff6873
LC		 3578 if (time_before(elr->lr_next_sched, next_wakeup))
3579 next_wakeup = elr->lr_next_sched;
3580 }
3581 mutex_unlock(&eli->li_list_mtx);
3582
a0acae0e 3583 try_to_freeze();
bfff6873 3584
4ed5c033
LC		 3585 cur = jiffies;
3586 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3587 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3588 cond_resched();
3589 continue;
3590 }
3591
4ed5c033
LC		 3592 schedule_timeout_interruptible(next_wakeup - cur);
3593
8f1f7453
ES		 3594 if (kthread_should_stop()) {
3595 ext4_clear_request_list();
3596 goto exit_thread;
3597 }
bfff6873
LC		 3598 }
3599
3600exit_thread:
3601 /*
3602 * It looks like the request list is empty, but we need
3603 * to check it under the li_list_mtx lock, to prevent any
3604 * additions into it, and of course we should lock ext4_li_mtx
3605 * to atomically free the list and ext4_li_info, because at
3606 * this point another ext4 filesystem could be registering
3607 * new one.
3608 */
3609 mutex_lock(&ext4_li_mtx);
3610 mutex_lock(&eli->li_list_mtx);
3611 if (!list_empty(&eli->li_request_list)) {
3612 mutex_unlock(&eli->li_list_mtx);
3613 mutex_unlock(&ext4_li_mtx);
3614 goto cont_thread;
3615 }
3616 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3617 kfree(ext4_li_info);
3618 ext4_li_info = NULL;
3619 mutex_unlock(&ext4_li_mtx);
3620
3621 return 0;
3622}
3623
3624static void ext4_clear_request_list(void)
3625{
3626 struct list_head *pos, *n;
3627 struct ext4_li_request *elr;
3628
3629 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3630 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3631 elr = list_entry(pos, struct ext4_li_request,
3632 lr_request);
3633 ext4_remove_li_request(elr);
3634 }
3635 mutex_unlock(&ext4_li_info->li_list_mtx);
3636}
3637
3638static int ext4_run_lazyinit_thread(void)
3639{
8f1f7453
ES		 3640 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3641 ext4_li_info, "ext4lazyinit");
3642 if (IS_ERR(ext4_lazyinit_task)) {
3643 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3644 ext4_clear_request_list();
bfff6873
LC		 3645 kfree(ext4_li_info);
3646 ext4_li_info = NULL;
92b97816 3647 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3648 "initialization thread\n",
3649 err);
3650 return err;
3651 }
3652 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3653 return 0;
3654}
3655
3656/*
3657 * Check whether it make sense to run itable init. thread or not.
3658 * If there is at least one uninitialized inode table, return
3659 * corresponding group number, else the loop goes through all
3660 * groups and return total number of groups.
3661 */
3662static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3663{
3664 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3665 struct ext4_group_desc *gdp = NULL;
3666
8844618d
TT		 3667 if (!ext4_has_group_desc_csum(sb))
3668 return ngroups;
3669
bfff6873
LC		 3670 for (group = 0; group < ngroups; group++) {
3671 gdp = ext4_get_group_desc(sb, group, NULL);
3672 if (!gdp)
3673 continue;
3674
50122847 3675 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
bfff6873
LC		 3676 break;
3677 }
3678
3679 return group;
3680}
3681
3682static int ext4_li_info_new(void)
3683{
3684 struct ext4_lazy_init *eli = NULL;
3685
3686 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3687 if (!eli)
3688 return -ENOMEM;
3689
bfff6873
LC		 3690 INIT_LIST_HEAD(&eli->li_request_list);
3691 mutex_init(&eli->li_list_mtx);
3692
bfff6873
LC		 3693 eli->li_state |= EXT4_LAZYINIT_QUIT;
3694
3695 ext4_li_info = eli;
3696
3697 return 0;
3698}
3699
3700static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3701 ext4_group_t start)
3702{
bfff6873 3703 struct ext4_li_request *elr;
bfff6873
LC		 3704
3705 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3706 if (!elr)
3707 return NULL;
3708
3709 elr->lr_super = sb;
3d392b26
TT		 3710 elr->lr_first_not_zeroed = start;
3711 if (test_opt(sb, PREFETCH_BLOCK_BITMAPS))
3712 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3713 else {
3714 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3715 elr->lr_next_group = start;
3716 }
bfff6873
LC		 3717
3718 /*
3719 * Randomize first schedule time of the request to
3720 * spread the inode table initialization requests
3721 * better.
3722 */
dd1f723b
TT		 3723 elr->lr_next_sched = jiffies + (prandom_u32() %
3724 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3725 return elr;
3726}
3727
7f511862
TT		 3728int ext4_register_li_request(struct super_block *sb,
3729 ext4_group_t first_not_zeroed)
bfff6873
LC		 3730{
3731 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3732 struct ext4_li_request *elr = NULL;
49598e04 3733 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 3734 int ret = 0;
bfff6873 3735
7f511862 3736 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3737 if (sbi->s_li_request != NULL) {
3738 /*
3739 * Reset timeout so it can be computed again, because
3740 * s_li_wait_mult might have changed.
3741 */
3742 sbi->s_li_request->lr_timeout = 0;
7f511862 3743 goto out;
51ce6511 3744 }
bfff6873 3745
3d392b26
TT		 3746 if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) &&
3747 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3748 !test_opt(sb, INIT_INODE_TABLE)))
7f511862 3749 goto out;
bfff6873
LC		 3750
3751 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3752 if (!elr) {
3753 ret = -ENOMEM;
3754 goto out;
3755 }
bfff6873
LC		 3756
3757 if (NULL == ext4_li_info) {
3758 ret = ext4_li_info_new();
3759 if (ret)
3760 goto out;
3761 }
3762
3763 mutex_lock(&ext4_li_info->li_list_mtx);
3764 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3765 mutex_unlock(&ext4_li_info->li_list_mtx);
3766
3767 sbi->s_li_request = elr;
46e4690b
TM		 3768 /*
3769 * set elr to NULL here since it has been inserted to
3770 * the request_list and the removal and free of it is
3771 * handled by ext4_clear_request_list from now on.
3772 */
3773 elr = NULL;
bfff6873
LC		 3774
3775 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3776 ret = ext4_run_lazyinit_thread();
3777 if (ret)
3778 goto out;
3779 }
bfff6873 3780out:
beed5ecb
NK		 3781 mutex_unlock(&ext4_li_mtx);
3782 if (ret)
bfff6873 3783 kfree(elr);
bfff6873
LC		 3784 return ret;
3785}
3786
3787/*
3788 * We do not need to lock anything since this is called on
3789 * module unload.
3790 */
3791static void ext4_destroy_lazyinit_thread(void)
3792{
3793 /*
3794 * If thread exited earlier
3795 * there's nothing to be done.
3796 */
8f1f7453 3797 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3798 return;
3799
8f1f7453 3800 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3801}
3802
25ed6e8a
DW		 3803static int set_journal_csum_feature_set(struct super_block *sb)
3804{
3805 int ret = 1;
3806 int compat, incompat;
3807 struct ext4_sb_info *sbi = EXT4_SB(sb);
3808
9aa5d32b 3809 if (ext4_has_metadata_csum(sb)) {
db9ee220 3810 /* journal checksum v3 */
25ed6e8a 3811 compat = 0;
db9ee220 3812 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3813 } else {
3814 /* journal checksum v1 */
3815 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3816 incompat = 0;
3817 }
3818
feb8c6d3
DW		 3819 jbd2_journal_clear_features(sbi->s_journal,
3820 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3821 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3822 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3823 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3824 ret = jbd2_journal_set_features(sbi->s_journal,
3825 compat, 0,
3826 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3827 incompat);
3828 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3829 ret = jbd2_journal_set_features(sbi->s_journal,
3830 compat, 0,
3831 incompat);
3832 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3833 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3834 } else {
feb8c6d3
DW		 3835 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3836 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3837 }
3838
3839 return ret;
3840}
3841
952fc18e
TT		 3842/*
3843 * Note: calculating the overhead so we can be compatible with
3844 * historical BSD practice is quite difficult in the face of
3845 * clusters/bigalloc. This is because multiple metadata blocks from
3846 * different block group can end up in the same allocation cluster.
3847 * Calculating the exact overhead in the face of clustered allocation
3848 * requires either O(all block bitmaps) in memory or O(number of block
3849 * groups**2) in time. We will still calculate the superblock for
3850 * older file systems --- and if we come across with a bigalloc file
3851 * system with zero in s_overhead_clusters the estimate will be close to
3852 * correct especially for very large cluster sizes --- but for newer
3853 * file systems, it's better to calculate this figure once at mkfs
3854 * time, and store it in the superblock. If the superblock value is
3855 * present (even for non-bigalloc file systems), we will use it.
3856 */
3857static int count_overhead(struct super_block *sb, ext4_group_t grp,
3858 char *buf)
3859{
3860 struct ext4_sb_info *sbi = EXT4_SB(sb);
3861 struct ext4_group_desc *gdp;
3862 ext4_fsblk_t first_block, last_block, b;
3863 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3864 int s, j, count = 0;
3865
e2b911c5 3866 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3867 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3868 sbi->s_itb_per_group + 2);
3869
952fc18e
TT		 3870 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3871 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3872 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3873 for (i = 0; i < ngroups; i++) {
3874 gdp = ext4_get_group_desc(sb, i, NULL);
3875 b = ext4_block_bitmap(sb, gdp);
3876 if (b >= first_block && b <= last_block) {
3877 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3878 count++;
3879 }
3880 b = ext4_inode_bitmap(sb, gdp);
3881 if (b >= first_block && b <= last_block) {
3882 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3883 count++;
3884 }
3885 b = ext4_inode_table(sb, gdp);
3886 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3887 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3888 int c = EXT4_B2C(sbi, b - first_block);
3889 ext4_set_bit(c, buf);
3890 count++;
3891 }
3892 if (i != grp)
3893 continue;
3894 s = 0;
3895 if (ext4_bg_has_super(sb, grp)) {
3896 ext4_set_bit(s++, buf);
3897 count++;
3898 }
c48ae41b
TT		 3899 j = ext4_bg_num_gdb(sb, grp);
3900 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3901 ext4_error(sb, "Invalid number of block group "
3902 "descriptor blocks: %d", j);
3903 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3904 }
c48ae41b
TT		 3905 count += j;
3906 for (; j > 0; j--)
3907 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3908 }
3909 if (!count)
3910 return 0;
3911 return EXT4_CLUSTERS_PER_GROUP(sb) -
3912 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3913}
3914
3915/*
3916 * Compute the overhead and stash it in sbi->s_overhead
3917 */
3918int ext4_calculate_overhead(struct super_block *sb)
3919{
3920 struct ext4_sb_info *sbi = EXT4_SB(sb);
3921 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3922 struct inode *j_inode;
3923 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3924 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3925 ext4_fsblk_t overhead = 0;
4fdb5543 3926 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3927
952fc18e
TT		 3928 if (!buf)
3929 return -ENOMEM;
3930
3931 /*
3932 * Compute the overhead (FS structures). This is constant
3933 * for a given filesystem unless the number of block groups
3934 * changes so we cache the previous value until it does.
3935 */
3936
3937 /*
3938 * All of the blocks before first_data_block are overhead
3939 */
3940 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3941
3942 /*
3943 * Add the overhead found in each block group
3944 */
3945 for (i = 0; i < ngroups; i++) {
3946 int blks;
3947
3948 blks = count_overhead(sb, i, buf);
3949 overhead += blks;
3950 if (blks)
3951 memset(buf, 0, PAGE_SIZE);
3952 cond_resched();
3953 }
3c816ded
EW		 3954
3955 /*
3956 * Add the internal journal blocks whether the journal has been
3957 * loaded or not
3958 */
ee7ed3aa 3959 if (sbi->s_journal && !sbi->s_journal_bdev)
ede7dc7f 3960 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
f1eec3b0
RH		 3961 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3962 /* j_inum for internal journal is non-zero */
3c816ded
EW		 3963 j_inode = ext4_get_journal_inode(sb, j_inum);
3964 if (j_inode) {
3965 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3966 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3967 iput(j_inode);
3968 } else {
3969 ext4_msg(sb, KERN_ERR, "can't get journal size");
3970 }
3971 }
952fc18e
TT		 3972 sbi->s_overhead = overhead;
3973 smp_wmb();
3974 free_page((unsigned long) buf);
3975 return 0;
3976}
3977
b5799018 3978static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3979{
3980 ext4_fsblk_t resv_clusters;
b5799018 3981 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3982
30fac0f7
JK		 3983 /*
3984 * There's no need to reserve anything when we aren't using extents.
3985 * The space estimates are exact, there are no unwritten extents,
3986 * hole punching doesn't need new metadata... This is needed especially
3987 * to keep ext2/3 backward compatibility.
3988 */
e2b911c5 3989 if (!ext4_has_feature_extents(sb))
b5799018 3990 return;
27dd4385
LC		 3991 /*
3992 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3993 * This should cover the situations where we can not afford to run
3994 * out of space like for example punch hole, or converting
556615dc 3995 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3996 * allocation would require 1, or 2 blocks, higher numbers are
3997 * very rare.
3998 */
b5799018
TT		 3999 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4000 sbi->s_cluster_bits);
27dd4385
LC		 4001
4002 do_div(resv_clusters, 50);
4003 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4004
b5799018 4005 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 4006}
4007
2b2d6d01 4008static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 4009{
5e405595 4010 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
d4c402d9 4011 char *orig_data = kstrdup(data, GFP_KERNEL);
1d0c3924 4012 struct buffer_head *bh, **group_desc;
617ba13b 4013 struct ext4_super_block *es = NULL;
5aee0f8a 4014 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
7c990728 4015 struct flex_groups **flex_groups;
617ba13b
MC		 4016 ext4_fsblk_t block;
4017 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 4018 ext4_fsblk_t logical_sb_block;
ac27a0ec 4019 unsigned long offset = 0;
ac27a0ec
DK		 4020 unsigned long journal_devnum = 0;
4021 unsigned long def_mount_opts;
4022 struct inode *root;
0390131b 4023 const char *descr;
dcc7dae3 4024 int ret = -ENOMEM;
281b5995 4025 int blocksize, clustersize;
4ec11028
TT		 4026 unsigned int db_count;
4027 unsigned int i;
ef5fd681 4028 int needs_recovery, has_huge_files;
bd81d8ee 4029 __u64 blocks_count;
07aa2ea1 4030 int err = 0;
b3881f74 4031 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 4032 ext4_group_t first_not_zeroed;
ac27a0ec 4033
5aee0f8a
TT		 4034 if ((data && !orig_data) || !sbi)
4035 goto out_free_base;
705895b6 4036
aed9eb1b 4037 sbi->s_daxdev = dax_dev;
705895b6
PE		 4038 sbi->s_blockgroup_lock =
4039 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 4040 if (!sbi->s_blockgroup_lock)
4041 goto out_free_base;
4042
ac27a0ec 4043 sb->s_fs_info = sbi;
2c0544b2 4044 sbi->s_sb = sb;
240799cd 4045 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 4046 sbi->s_sb_block = sb_block;
8446fe92
CH		 4047 sbi->s_sectors_written_start =
4048 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
ac27a0ec 4049
9f6200bb 4050 /* Cleanup superblock name */
ec3904dc 4051 strreplace(sb->s_id, '/', '!');
9f6200bb 4052
07aa2ea1 4053 /* -EINVAL is default */
dcc7dae3 4054 ret = -EINVAL;
617ba13b 4055 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 4056 if (!blocksize) {
b31e1552 4057 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 4058 goto out_fail;
4059 }
4060
4061 /*
617ba13b 4062 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 4063 * block sizes. We need to calculate the offset from buffer start.
4064 */
617ba13b 4065 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 4066 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4067 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 4068 } else {
70bbb3e0 4069 logical_sb_block = sb_block;
ac27a0ec
DK		 4070 }
4071
8394a6ab 4072 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4073 if (IS_ERR(bh)) {
b31e1552 4074 ext4_msg(sb, KERN_ERR, "unable to read superblock");
8394a6ab 4075 ret = PTR_ERR(bh);
4076 bh = NULL;
ac27a0ec
DK		 4077 goto out_fail;
4078 }
4079 /*
4080 * Note: s_es must be initialized as soon as possible because
617ba13b 4081 * some ext4 macro-instructions depend on its value
ac27a0ec 4082 */
2716b802 4083 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 4084 sbi->s_es = es;
4085 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 4086 if (sb->s_magic != EXT4_SUPER_MAGIC)
4087 goto cantfind_ext4;
afc32f7e 4088 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 4089
feb0ab32 4090 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 4091 if (ext4_has_feature_metadata_csum(sb) &&
4092 ext4_has_feature_gdt_csum(sb))
363307e6 4093 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 4094 "redundant flags; please run fsck.");
4095
d25425f8
DW		 4096 /* Check for a known checksum algorithm */
4097 if (!ext4_verify_csum_type(sb, es)) {
4098 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4099 "unknown checksum algorithm.");
4100 silent = 1;
4101 goto cantfind_ext4;
4102 }
4103
0441984a 4104 /* Load the checksum driver */
a45403b5
TT		 4105 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4106 if (IS_ERR(sbi->s_chksum_driver)) {
4107 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4108 ret = PTR_ERR(sbi->s_chksum_driver);
4109 sbi->s_chksum_driver = NULL;
4110 goto failed_mount;
0441984a
DW		 4111 }
4112
a9c47317
DW		 4113 /* Check superblock checksum */
4114 if (!ext4_superblock_csum_verify(sb, es)) {
4115 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4116 "invalid superblock checksum. Run e2fsck?");
4117 silent = 1;
6a797d27 4118 ret = -EFSBADCRC;
a9c47317
DW		 4119 goto cantfind_ext4;
4120 }
4121
4122 /* Precompute checksum seed for all metadata */
e2b911c5 4123 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 4124 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 4125 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 4126 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4127 sizeof(es->s_uuid));
4128
ac27a0ec
DK		 4129 /* Set defaults before we parse the mount options */
4130 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 4131 set_opt(sb, INIT_INODE_TABLE);
617ba13b 4132 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 4133 set_opt(sb, DEBUG);
87f26807 4134 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 4135 set_opt(sb, GRPID);
617ba13b 4136 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 4137 set_opt(sb, NO_UID32);
ea663336 4138 /* xattr user namespace & acls are now defaulted on */
ea663336 4139 set_opt(sb, XATTR_USER);
03010a33 4140#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 4141 set_opt(sb, POSIX_ACL);
2e7842b8 4142#endif
995a3ed6
HS		 4143 if (ext4_has_feature_fast_commit(sb))
4144 set_opt2(sb, JOURNAL_FAST_COMMIT);
98c1a759
DW		 4145 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4146 if (ext4_has_metadata_csum(sb))
4147 set_opt(sb, JOURNAL_CHECKSUM);
4148
617ba13b 4149 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 4150 set_opt(sb, JOURNAL_DATA);
617ba13b 4151 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 4152 set_opt(sb, ORDERED_DATA);
617ba13b 4153 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 4154 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 4155
4156 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 4157 set_opt(sb, ERRORS_PANIC);
bb4f397a 4158 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 4159 set_opt(sb, ERRORS_CONT);
bb4f397a 4160 else
fd8c37ec 4161 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 4162 /* block_validity enabled by default; disable with noblock_validity */
4163 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 4164 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 4165 set_opt(sb, DISCARD);
ac27a0ec 4166
08cefc7a
EB		 4167 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4168 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 4169 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4170 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4171 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 4172
8b67f04a 4173 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 4174 set_opt(sb, BARRIER);
ac27a0ec 4175
dd919b98
AK		 4176 /*
4177 * enable delayed allocation by default
4178 * Use -o nodelalloc to turn it off
4179 */
bc0b75f7 4180 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 4181 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 4182 set_opt(sb, DELALLOC);
dd919b98 4183
51ce6511
LC		 4184 /*
4185 * set default s_li_wait_mult for lazyinit, for the case there is
4186 * no mount option specified.
4187 */
4188 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4189
4f97a681 4190 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
626b035b
RH		 4191
4192 if (blocksize == PAGE_SIZE)
4193 set_opt(sb, DIOREAD_NOLOCK);
4194
4f97a681
TT		 4195 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
4196 blocksize > EXT4_MAX_BLOCK_SIZE) {
4197 ext4_msg(sb, KERN_ERR,
4198 "Unsupported filesystem blocksize %d (%d log_block_size)",
4199 blocksize, le32_to_cpu(es->s_log_block_size));
4200 goto failed_mount;
4201 }
4202
9803387c
TT		 4203 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4204 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4205 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4206 } else {
4207 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4208 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4209 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4210 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4211 sbi->s_first_ino);
4212 goto failed_mount;
4213 }
4214 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4215 (!is_power_of_2(sbi->s_inode_size)) ||
4216 (sbi->s_inode_size > blocksize)) {
4217 ext4_msg(sb, KERN_ERR,
4218 "unsupported inode size: %d",
4219 sbi->s_inode_size);
4f97a681 4220 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
9803387c
TT		 4221 goto failed_mount;
4222 }
4223 /*
4224 * i_atime_extra is the last extra field available for
4225 * [acm]times in struct ext4_inode. Checking for that
4226 * field should suffice to ensure we have extra space
4227 * for all three.
4228 */
4229 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4230 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4231 sb->s_time_gran = 1;
4232 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4233 } else {
4234 sb->s_time_gran = NSEC_PER_SEC;
4235 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4236 }
4237 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4238 }
4239 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4240 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4241 EXT4_GOOD_OLD_INODE_SIZE;
4242 if (ext4_has_feature_extra_isize(sb)) {
4243 unsigned v, max = (sbi->s_inode_size -
4244 EXT4_GOOD_OLD_INODE_SIZE);
4245
4246 v = le16_to_cpu(es->s_want_extra_isize);
4247 if (v > max) {
4248 ext4_msg(sb, KERN_ERR,
4249 "bad s_want_extra_isize: %d", v);
4250 goto failed_mount;
4251 }
4252 if (sbi->s_want_extra_isize < v)
4253 sbi->s_want_extra_isize = v;
4254
4255 v = le16_to_cpu(es->s_min_extra_isize);
4256 if (v > max) {
4257 ext4_msg(sb, KERN_ERR,
4258 "bad s_min_extra_isize: %d", v);
4259 goto failed_mount;
4260 }
4261 if (sbi->s_want_extra_isize < v)
4262 sbi->s_want_extra_isize = v;
4263 }
4264 }
4265
5aee0f8a
TT		 4266 if (sbi->s_es->s_mount_opts[0]) {
4267 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4268 sizeof(sbi->s_es->s_mount_opts),
4269 GFP_KERNEL);
4270 if (!s_mount_opts)
4271 goto failed_mount;
4272 if (!parse_options(s_mount_opts, sb, &journal_devnum,
4273 &journal_ioprio, 0)) {
4274 ext4_msg(sb, KERN_WARNING,
4275 "failed to parse options in superblock: %s",
4276 s_mount_opts);
4277 }
4278 kfree(s_mount_opts);
8b67f04a 4279 }
5a916be1 4280 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 4281 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 4282 &journal_ioprio, 0))
ac27a0ec
DK		 4283 goto failed_mount;
4284
c83ad55e 4285#ifdef CONFIG_UNICODE
f8f4acb6 4286 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
c83ad55e
GKB		 4287 const struct ext4_sb_encodings *encoding_info;
4288 struct unicode_map *encoding;
4289 __u16 encoding_flags;
4290
4291 if (ext4_has_feature_encrypt(sb)) {
4292 ext4_msg(sb, KERN_ERR,
4293 "Can't mount with encoding and encryption");
4294 goto failed_mount;
4295 }
4296
4297 if (ext4_sb_read_encoding(es, &encoding_info,
4298 &encoding_flags)) {
4299 ext4_msg(sb, KERN_ERR,
4300 "Encoding requested by superblock is unknown");
4301 goto failed_mount;
4302 }
4303
4304 encoding = utf8_load(encoding_info->version);
4305 if (IS_ERR(encoding)) {
4306 ext4_msg(sb, KERN_ERR,
4307 "can't mount with superblock charset: %s-%s "
4308 "not supported by the kernel. flags: 0x%x.",
4309 encoding_info->name, encoding_info->version,
4310 encoding_flags);
4311 goto failed_mount;
4312 }
4313 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4314 "%s-%s with flags 0x%hx", encoding_info->name,
4315 encoding_info->version?:"\b", encoding_flags);
4316
f8f4acb6
DR		 4317 sb->s_encoding = encoding;
4318 sb->s_encoding_flags = encoding_flags;
c83ad55e
GKB		 4319 }
4320#endif
4321
56889787 4322 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
556e0319 4323 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
781c036b 4324 /* can't mount with both data=journal and dioread_nolock. */
244adf64 4325 clear_opt(sb, DIOREAD_NOLOCK);
556e0319 4326 clear_opt2(sb, JOURNAL_FAST_COMMIT);
56889787
TT		 4327 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4328 ext4_msg(sb, KERN_ERR, "can't mount with "
4329 "both data=journal and delalloc");
4330 goto failed_mount;
4331 }
fc626fe3 4332 if (test_opt(sb, DAX_ALWAYS)) {
923ae0ff
RZ		 4333 ext4_msg(sb, KERN_ERR, "can't mount with "
4334 "both data=journal and dax");
4335 goto failed_mount;
4336 }
73b92a2a
SK		 4337 if (ext4_has_feature_encrypt(sb)) {
4338 ext4_msg(sb, KERN_WARNING,
4339 "encrypted files will use data=ordered "
4340 "instead of data journaling mode");
4341 }
56889787
TT		 4342 if (test_opt(sb, DELALLOC))
4343 clear_opt(sb, DELALLOC);
001e4a87
TH		 4344 } else {
4345 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 4346 }
4347
1751e8a6
LT		 4348 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4349 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 4350
617ba13b 4351 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 4352 (ext4_has_compat_features(sb) ||
4353 ext4_has_ro_compat_features(sb) ||
4354 ext4_has_incompat_features(sb)))
b31e1552
ES		 4355 ext4_msg(sb, KERN_WARNING,
4356 "feature flags set on rev 0 fs, "
4357 "running e2fsck is recommended");
469108ff 4358
ed3654eb
TT		 4359 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4360 set_opt2(sb, HURD_COMPAT);
e2b911c5 4361 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 4362 ext4_msg(sb, KERN_ERR,
4363 "The Hurd can't support 64-bit file systems");
4364 goto failed_mount;
4365 }
dec214d0
TE		 4366
4367 /*
4368 * ea_inode feature uses l_i_version field which is not
4369 * available in HURD_COMPAT mode.
4370 */
4371 if (ext4_has_feature_ea_inode(sb)) {
4372 ext4_msg(sb, KERN_ERR,
4373 "ea_inode feature is not supported for Hurd");
4374 goto failed_mount;
4375 }
ed3654eb
TT		 4376 }
4377
2035e776
TT		 4378 if (IS_EXT2_SB(sb)) {
4379 if (ext2_feature_set_ok(sb))
4380 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4381 "using the ext4 subsystem");
4382 else {
0d9366d6
ES		 4383 /*
4384 * If we're probing be silent, if this looks like
4385 * it's actually an ext[34] filesystem.
4386 */
4387 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4388 goto failed_mount;
2035e776
TT		 4389 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4390 "to feature incompatibilities");
4391 goto failed_mount;
4392 }
4393 }
4394
4395 if (IS_EXT3_SB(sb)) {
4396 if (ext3_feature_set_ok(sb))
4397 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4398 "using the ext4 subsystem");
4399 else {
0d9366d6
ES		 4400 /*
4401 * If we're probing be silent, if this looks like
4402 * it's actually an ext4 filesystem.
4403 */
4404 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4405 goto failed_mount;
2035e776
TT		 4406 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4407 "to feature incompatibilities");
4408 goto failed_mount;
4409 }
4410 }
4411
ac27a0ec
DK		 4412 /*
4413 * Check feature flags regardless of the revision level, since we
4414 * previously didn't change the revision level when setting the flags,
4415 * so there is a chance incompat flags are set on a rev 0 filesystem.
4416 */
bc98a42c 4417 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
ac27a0ec 4418 goto failed_mount;
a13fb1a4 4419
8cdf3372
TT		 4420 if (le32_to_cpu(es->s_log_block_size) >
4421 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4422 ext4_msg(sb, KERN_ERR,
4423 "Invalid log block size: %u",
4424 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 4425 goto failed_mount;
4426 }
bfe0a5f4
TT		 4427 if (le32_to_cpu(es->s_log_cluster_size) >
4428 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4429 ext4_msg(sb, KERN_ERR,
4430 "Invalid log cluster size: %u",
4431 le32_to_cpu(es->s_log_cluster_size));
4432 goto failed_mount;
4433 }
ac27a0ec 4434
5b9554dc
TT		 4435 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4436 ext4_msg(sb, KERN_ERR,
4437 "Number of reserved GDT blocks insanely large: %d",
4438 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4439 goto failed_mount;
4440 }
4441
a8ab6d38
IW		 4442 if (bdev_dax_supported(sb->s_bdev, blocksize))
4443 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4444
fc626fe3 4445 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
559db4c6
RZ		 4446 if (ext4_has_feature_inline_data(sb)) {
4447 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4448 " that may contain inline data");
361d24d4 4449 goto failed_mount;
559db4c6 4450 }
a8ab6d38 4451 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
24f3478d 4452 ext4_msg(sb, KERN_ERR,
361d24d4
ES		 4453 "DAX unsupported by block device.");
4454 goto failed_mount;
24f3478d 4455 }
923ae0ff
RZ		 4456 }
4457
e2b911c5 4458 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 4459 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4460 es->s_encryption_level);
4461 goto failed_mount;
4462 }
4463
ac27a0ec 4464 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 4465 /* Validate the filesystem blocksize */
4466 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 4467 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 4468 blocksize);
ac27a0ec
DK		 4469 goto failed_mount;
4470 }
4471
2b2d6d01 4472 brelse(bh);
70bbb3e0
AM		 4473 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4474 offset = do_div(logical_sb_block, blocksize);
8394a6ab 4475 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4476 if (IS_ERR(bh)) {
b31e1552
ES		 4477 ext4_msg(sb, KERN_ERR,
4478 "Can't read superblock on 2nd try");
8394a6ab 4479 ret = PTR_ERR(bh);
4480 bh = NULL;
ac27a0ec
DK		 4481 goto failed_mount;
4482 }
2716b802 4483 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 4484 sbi->s_es = es;
617ba13b 4485 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 4486 ext4_msg(sb, KERN_ERR,
4487 "Magic mismatch, very weird!");
ac27a0ec
DK		 4488 goto failed_mount;
4489 }
4490 }
4491
e2b911c5 4492 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 4493 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4494 has_huge_files);
4495 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 4496
0d1ee42f 4497 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 4498 if (ext4_has_feature_64bit(sb)) {
8fadc143 4499 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 4500 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 4501 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 4502 ext4_msg(sb, KERN_ERR,
4503 "unsupported descriptor size %lu",
0d1ee42f
AR		 4504 sbi->s_desc_size);
4505 goto failed_mount;
4506 }
4507 } else
4508 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 4509
ac27a0ec 4510 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 4511 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 4512
617ba13b 4513 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 4514 if (sbi->s_inodes_per_block == 0)
617ba13b 4515 goto cantfind_ext4;
cd6bb35b
TT		 4516 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4517 sbi->s_inodes_per_group > blocksize * 8) {
4518 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
b9c538da 4519 sbi->s_inodes_per_group);
cd6bb35b
TT		 4520 goto failed_mount;
4521 }
ac27a0ec
DK		 4522 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4523 sbi->s_inodes_per_block;
0d1ee42f 4524 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 4525 sbi->s_sbh = bh;
4526 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 4527 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4528 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 4529
2b2d6d01 4530 for (i = 0; i < 4; i++)
ac27a0ec
DK		 4531 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4532 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 4533 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 4534 i = le32_to_cpu(es->s_flags);
4535 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4536 sbi->s_hash_unsigned = 3;
4537 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 4538#ifdef __CHAR_UNSIGNED__
bc98a42c 4539 if (!sb_rdonly(sb))
23301410
TT		 4540 es->s_flags |=
4541 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4542 sbi->s_hash_unsigned = 3;
f99b2589 4543#else
bc98a42c 4544 if (!sb_rdonly(sb))
23301410
TT		 4545 es->s_flags |=
4546 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 4547#endif
23301410 4548 }
f99b2589 4549 }
ac27a0ec 4550
281b5995
TT		 4551 /* Handle clustersize */
4552 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
ef5fd681 4553 if (ext4_has_feature_bigalloc(sb)) {
281b5995
TT		 4554 if (clustersize < blocksize) {
4555 ext4_msg(sb, KERN_ERR,
4556 "cluster size (%d) smaller than "
4557 "block size (%d)", clustersize, blocksize);
4558 goto failed_mount;
4559 }
4560 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4561 le32_to_cpu(es->s_log_block_size);
4562 sbi->s_clusters_per_group =
4563 le32_to_cpu(es->s_clusters_per_group);
4564 if (sbi->s_clusters_per_group > blocksize * 8) {
4565 ext4_msg(sb, KERN_ERR,
4566 "#clusters per group too big: %lu",
4567 sbi->s_clusters_per_group);
4568 goto failed_mount;
4569 }
4570 if (sbi->s_blocks_per_group !=
4571 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4572 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4573 "clusters per group (%lu) inconsistent",
4574 sbi->s_blocks_per_group,
4575 sbi->s_clusters_per_group);
4576 goto failed_mount;
4577 }
4578 } else {
4579 if (clustersize != blocksize) {
bfe0a5f4
TT		 4580 ext4_msg(sb, KERN_ERR,
4581 "fragment/cluster size (%d) != "
4582 "block size (%d)", clustersize, blocksize);
4583 goto failed_mount;
281b5995
TT		 4584 }
4585 if (sbi->s_blocks_per_group > blocksize * 8) {
4586 ext4_msg(sb, KERN_ERR,
4587 "#blocks per group too big: %lu",
4588 sbi->s_blocks_per_group);
4589 goto failed_mount;
4590 }
4591 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4592 sbi->s_cluster_bits = 0;
ac27a0ec 4593 }
281b5995
TT		 4594 sbi->s_cluster_ratio = clustersize / blocksize;
4595
960fd856
TT		 4596 /* Do we have standard group size of clustersize * 8 blocks ? */
4597 if (sbi->s_blocks_per_group == clustersize << 3)
4598 set_opt2(sb, STD_GROUP_SIZE);
4599
bf43d84b
ES		 4600 /*
4601 * Test whether we have more sectors than will fit in sector_t,
4602 * and whether the max offset is addressable by the page cache.
4603 */
5a9ae68a 4604 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 4605 ext4_blocks_count(es));
5a9ae68a 4606 if (err) {
b31e1552 4607 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 4608 " too large to mount safely on this system");
ac27a0ec
DK		 4609 goto failed_mount;
4610 }
4611
617ba13b
MC		 4612 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4613 goto cantfind_ext4;
e7c95593 4614
0f2ddca6
FTN		 4615 /* check blocks count against device size */
4616 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4617 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 4618 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4619 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 4620 ext4_blocks_count(es), blocks_count);
4621 goto failed_mount;
4622 }
4623
0b8e58a1
AD		 4624 /*
4625 * It makes no sense for the first data block to be beyond the end
4626 * of the filesystem.
4627 */
4628 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 4629 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 4630 "block %u is beyond end of filesystem (%llu)",
4631 le32_to_cpu(es->s_first_data_block),
4632 ext4_blocks_count(es));
e7c95593
ES		 4633 goto failed_mount;
4634 }
bfe0a5f4
TT		 4635 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4636 (sbi->s_cluster_ratio == 1)) {
4637 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4638 "block is 0 with a 1k block and cluster size");
4639 goto failed_mount;
4640 }
4641
bd81d8ee
LV		 4642 blocks_count = (ext4_blocks_count(es) -
4643 le32_to_cpu(es->s_first_data_block) +
4644 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4645 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 4646 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
df41460a 4647 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4ec11028 4648 "(block count %llu, first data block %u, "
df41460a 4649 "blocks per group %lu)", blocks_count,
4ec11028
TT		 4650 ext4_blocks_count(es),
4651 le32_to_cpu(es->s_first_data_block),
4652 EXT4_BLOCKS_PER_GROUP(sb));
4653 goto failed_mount;
4654 }
bd81d8ee 4655 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 4656 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4657 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
9e463084
TT		 4658 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4659 le32_to_cpu(es->s_inodes_count)) {
4660 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4661 le32_to_cpu(es->s_inodes_count),
4662 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4663 ret = -EINVAL;
4664 goto failed_mount;
4665 }
617ba13b
MC		 4666 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4667 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 4668 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 4669 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 4670 ext4_msg(sb, KERN_WARNING,
4671 "first meta block group too large: %u "
4672 "(group descriptor block count %u)",
4673 le32_to_cpu(es->s_first_meta_bg), db_count);
4674 goto failed_mount;
4675 }
4676 }
1d0c3924
TT		 4677 rcu_assign_pointer(sbi->s_group_desc,
4678 kvmalloc_array(db_count,
4679 sizeof(struct buffer_head *),
4680 GFP_KERNEL));
ac27a0ec 4681 if (sbi->s_group_desc == NULL) {
b31e1552 4682 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 4683 ret = -ENOMEM;
ac27a0ec
DK		 4684 goto failed_mount;
4685 }
4686
705895b6 4687 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 4688
85c8f176
AP		 4689 /* Pre-read the descriptors into the buffer cache */
4690 for (i = 0; i < db_count; i++) {
4691 block = descriptor_loc(sb, logical_sb_block, i);
5df1d412 4692 ext4_sb_breadahead_unmovable(sb, block);
85c8f176
AP		 4693 }
4694
ac27a0ec 4695 for (i = 0; i < db_count; i++) {
1d0c3924
TT		 4696 struct buffer_head *bh;
4697
70bbb3e0 4698 block = descriptor_loc(sb, logical_sb_block, i);
8394a6ab 4699 bh = ext4_sb_bread_unmovable(sb, block);
4700 if (IS_ERR(bh)) {
b31e1552
ES		 4701 ext4_msg(sb, KERN_ERR,
4702 "can't read group descriptor %d", i);
ac27a0ec 4703 db_count = i;
8394a6ab 4704 ret = PTR_ERR(bh);
4705 bh = NULL;
ac27a0ec
DK		 4706 goto failed_mount2;
4707 }
1d0c3924
TT		 4708 rcu_read_lock();
4709 rcu_dereference(sbi->s_group_desc)[i] = bh;
4710 rcu_read_unlock();
ac27a0ec 4711 }
44de022c 4712 sbi->s_gdb_count = db_count;
829fa70d 4713 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 4714 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 4715 ret = -EFSCORRUPTED;
f9ae9cf5 4716 goto failed_mount2;
ac27a0ec 4717 }
772cb7c8 4718
235699a8 4719 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
04496411 4720
a75ae78f 4721 /* Register extent status tree shrinker */
eb68d0e2 4722 if (ext4_es_register_shrinker(sbi))
ce7e010a 4723 goto failed_mount3;
ce7e010a 4724
c9de560d 4725 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 4726 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 4727
f9ae9cf5
TT		 4728 /*
4729 * set up enough so that it can read an inode
4730 */
f6e63f90 4731 sb->s_op = &ext4_sops;
617ba13b
MC		 4732 sb->s_export_op = &ext4_export_ops;
4733 sb->s_xattr = ext4_xattr_handlers;
643fa961 4734#ifdef CONFIG_FS_ENCRYPTION
a7550b30 4735 sb->s_cop = &ext4_cryptops;
ffcc4182 4736#endif
c93d8f88
EB		 4737#ifdef CONFIG_FS_VERITY
4738 sb->s_vop = &ext4_verityops;
4739#endif
ac27a0ec 4740#ifdef CONFIG_QUOTA
617ba13b 4741 sb->dq_op = &ext4_quota_operations;
e2b911c5 4742 if (ext4_has_feature_quota(sb))
1fa5efe3 4743 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 4744 else
4745 sb->s_qcop = &ext4_qctl_operations;
689c958c 4746 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 4747#endif
85787090 4748 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 4749
ac27a0ec 4750 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4751 mutex_init(&sbi->s_orphan_lock);
ac27a0ec 4752
aa75f4d3
HS		 4753 /* Initialize fast commit stuff */
4754 atomic_set(&sbi->s_fc_subtid, 0);
4755 atomic_set(&sbi->s_fc_ineligible_updates, 0);
4756 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4757 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4758 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4759 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4760 sbi->s_fc_bytes = 0;
9b5f6c9b
HS		 4761 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4762 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
aa75f4d3
HS		 4763 spin_lock_init(&sbi->s_fc_lock);
4764 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
8016e29f
HS		 4765 sbi->s_fc_replay_state.fc_regions = NULL;
4766 sbi->s_fc_replay_state.fc_regions_size = 0;
4767 sbi->s_fc_replay_state.fc_regions_used = 0;
4768 sbi->s_fc_replay_state.fc_regions_valid = 0;
4769 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4770 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4771 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
aa75f4d3 4772
ac27a0ec
DK		 4773 sb->s_root = NULL;
4774
4775 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 4776 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 4777
bc98a42c 4778 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 4779 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4780 goto failed_mount3a;
c5e06d10 4781
ac27a0ec
DK		 4782 /*
4783 * The first inode we look at is the journal inode. Don't try
4784 * root first: it may be modified in the journal!
4785 */
e2b911c5 4786 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 4787 err = ext4_load_journal(sb, es, journal_devnum);
4788 if (err)
50460fe8 4789 goto failed_mount3a;
bc98a42c 4790 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 4791 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4792 ext4_msg(sb, KERN_ERR, "required journal recovery "
4793 "suppressed and not mounted read-only");
744692dc 4794 goto failed_mount_wq;
ac27a0ec 4795 } else {
1e381f60
DM		 4796 /* Nojournal mode, all journal mount options are illegal */
4797 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4798 ext4_msg(sb, KERN_ERR, "can't mount with "
4799 "journal_checksum, fs mounted w/o journal");
4800 goto failed_mount_wq;
4801 }
4802 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4803 ext4_msg(sb, KERN_ERR, "can't mount with "
4804 "journal_async_commit, fs mounted w/o journal");
4805 goto failed_mount_wq;
4806 }
4807 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4808 ext4_msg(sb, KERN_ERR, "can't mount with "
4809 "commit=%lu, fs mounted w/o journal",
4810 sbi->s_commit_interval / HZ);
4811 goto failed_mount_wq;
4812 }
4813 if (EXT4_MOUNT_DATA_FLAGS &
4814 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4815 ext4_msg(sb, KERN_ERR, "can't mount with "
4816 "data=, fs mounted w/o journal");
4817 goto failed_mount_wq;
4818 }
50b29d8f 4819 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
1e381f60 4820 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4821 clear_opt(sb, DATA_FLAGS);
995a3ed6 4822 clear_opt2(sb, JOURNAL_FAST_COMMIT);
0390131b
FM		 4823 sbi->s_journal = NULL;
4824 needs_recovery = 0;
4825 goto no_journal;
ac27a0ec
DK		 4826 }
4827
e2b911c5 4828 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4829 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4830 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4831 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4832 goto failed_mount_wq;
eb40a09c
JS		 4833 }
4834
25ed6e8a
DW		 4835 if (!set_journal_csum_feature_set(sb)) {
4836 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4837 "feature set");
4838 goto failed_mount_wq;
d4da6c9c 4839 }
818d276c 4840
a1e5e465
HS		 4841 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4842 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4843 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4844 ext4_msg(sb, KERN_ERR,
4845 "Failed to set fast commit journal feature");
4846 goto failed_mount_wq;
4847 }
4848
ac27a0ec
DK		 4849 /* We have now updated the journal if required, so we can
4850 * validate the data journaling mode. */
4851 switch (test_opt(sb, DATA_FLAGS)) {
4852 case 0:
4853 /* No mode set, assume a default based on the journal
63f57933
AM		 4854 * capabilities: ORDERED_DATA if the journal can
4855 * cope, else JOURNAL_DATA
4856 */
dab291af 4857 if (jbd2_journal_check_available_features
27f394a7 4858 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
fd8c37ec 4859 set_opt(sb, ORDERED_DATA);
27f394a7
TN		 4860 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4861 } else {
fd8c37ec 4862 set_opt(sb, JOURNAL_DATA);
27f394a7
TN		 4863 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4864 }
ac27a0ec
DK		 4865 break;
4866
617ba13b
MC		 4867 case EXT4_MOUNT_ORDERED_DATA:
4868 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4869 if (!jbd2_journal_check_available_features
4870 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4871 ext4_msg(sb, KERN_ERR, "Journal does not support "
4872 "requested data journaling mode");
744692dc 4873 goto failed_mount_wq;
ac27a0ec
DK		 4874 }
4875 default:
4876 break;
4877 }
ab04df78
JK		 4878
4879 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4880 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4881 ext4_msg(sb, KERN_ERR, "can't mount with "
4882 "journal_async_commit in data=ordered mode");
4883 goto failed_mount_wq;
4884 }
4885
b3881f74 4886 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4887
18aadd47 4888 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
342af94e 4889 sbi->s_journal->j_submit_inode_data_buffers =
afb585a9 4890 ext4_journal_submit_inode_data_buffers;
342af94e 4891 sbi->s_journal->j_finish_inode_data_buffers =
afb585a9 4892 ext4_journal_finish_inode_data_buffers;
18aadd47 4893
ce7e010a 4894no_journal:
cdb7ee4c
TE		 4895 if (!test_opt(sb, NO_MBCACHE)) {
4896 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4897 if (!sbi->s_ea_block_cache) {
dec214d0 4898 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4899 "Failed to create ea_block_cache");
dec214d0
TE		 4900 goto failed_mount_wq;
4901 }
cdb7ee4c
TE		 4902
4903 if (ext4_has_feature_ea_inode(sb)) {
4904 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4905 if (!sbi->s_ea_inode_cache) {
4906 ext4_msg(sb, KERN_ERR,
4907 "Failed to create ea_inode_cache");
4908 goto failed_mount_wq;
4909 }
4910 }
9c191f70
M		 4911 }
4912
c93d8f88
EB		 4913 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4914 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4915 goto failed_mount_wq;
4916 }
4917
bc98a42c 4918 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
e2b911c5
DW		 4919 !ext4_has_feature_encrypt(sb)) {
4920 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4921 ext4_commit_super(sb, 1);
4922 }
4923
952fc18e
TT		 4924 /*
4925 * Get the # of file system overhead blocks from the
4926 * superblock if present.
4927 */
4928 if (es->s_overhead_clusters)
4929 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4930 else {
07aa2ea1
LC		 4931 err = ext4_calculate_overhead(sb);
4932 if (err)
952fc18e
TT		 4933 goto failed_mount_wq;
4934 }
4935
fd89d5f2
TH		 4936 /*
4937 * The maximum number of concurrent works can be high and
4938 * concurrency isn't really necessary. Limit it to 1.
4939 */
2e8fa54e
JK		 4940 EXT4_SB(sb)->rsv_conversion_wq =
4941 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4942 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4943 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4944 ret = -ENOMEM;
2e8fa54e
JK		 4945 goto failed_mount4;
4946 }
4947
ac27a0ec 4948 /*
dab291af 4949 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4950 * so we can safely mount the rest of the filesystem now.
4951 */
4952
8a363970 4953 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
1d1fe1ee 4954 if (IS_ERR(root)) {
b31e1552 4955 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4956 ret = PTR_ERR(root);
32a9bb57 4957 root = NULL;
ac27a0ec
DK		 4958 goto failed_mount4;
4959 }
4960 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4961 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4962 iput(root);
ac27a0ec
DK		 4963 goto failed_mount4;
4964 }
b886ee3e
GKB		 4965
4966#ifdef CONFIG_UNICODE
f8f4acb6 4967 if (sb->s_encoding)
b886ee3e
GKB		 4968 sb->s_d_op = &ext4_dentry_ops;
4969#endif
4970
48fde701 4971 sb->s_root = d_make_root(root);
1d1fe1ee 4972 if (!sb->s_root) {
b31e1552 4973 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4974 ret = -ENOMEM;
4975 goto failed_mount4;
4976 }
ac27a0ec 4977
c89128a0
JK		 4978 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4979 if (ret == -EROFS) {
1751e8a6 4980 sb->s_flags |= SB_RDONLY;
c89128a0
JK		 4981 ret = 0;
4982 } else if (ret)
4983 goto failed_mount4a;
ef7f3835 4984
b5799018 4985 ext4_set_resv_clusters(sb);
27dd4385 4986
0f5bde1d
JK		 4987 if (test_opt(sb, BLOCK_VALIDITY)) {
4988 err = ext4_setup_system_zone(sb);
4989 if (err) {
4990 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4991 "zone (%d)", err);
4992 goto failed_mount4a;
4993 }
f9ae9cf5 4994 }
8016e29f 4995 ext4_fc_replay_cleanup(sb);
f9ae9cf5
TT		 4996
4997 ext4_ext_init(sb);
4998 err = ext4_mb_init(sb);
4999 if (err) {
5000 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5001 err);
dcf2d804 5002 goto failed_mount5;
c2774d84
AK		 5003 }
5004
d5e03cbb
TT		 5005 block = ext4_count_free_clusters(sb);
5006 ext4_free_blocks_count_set(sbi->s_es,
5007 EXT4_C2B(sbi, block));
4274f516 5008 ext4_superblock_csum_set(sb);
908c7f19
TH		 5009 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5010 GFP_KERNEL);
d5e03cbb
TT		 5011 if (!err) {
5012 unsigned long freei = ext4_count_free_inodes(sb);
5013 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4274f516 5014 ext4_superblock_csum_set(sb);
908c7f19
TH		 5015 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5016 GFP_KERNEL);
d5e03cbb
TT		 5017 }
5018 if (!err)
5019 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 5020 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 5021 if (!err)
908c7f19
TH		 5022 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5023 GFP_KERNEL);
c8585c6f 5024 if (!err)
bbd55937 5025 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
c8585c6f 5026
d5e03cbb
TT		 5027 if (err) {
5028 ext4_msg(sb, KERN_ERR, "insufficient memory");
5029 goto failed_mount6;
5030 }
5031
e2b911c5 5032 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 5033 if (!ext4_fill_flex_info(sb)) {
5034 ext4_msg(sb, KERN_ERR,
5035 "unable to initialize "
5036 "flex_bg meta info!");
5037 goto failed_mount6;
5038 }
5039
bfff6873
LC		 5040 err = ext4_register_li_request(sb, first_not_zeroed);
5041 if (err)
dcf2d804 5042 goto failed_mount6;
bfff6873 5043
b5799018 5044 err = ext4_register_sysfs(sb);
dcf2d804
TM		 5045 if (err)
5046 goto failed_mount7;
3197ebdb 5047
9b2ff357
JK		 5048#ifdef CONFIG_QUOTA
5049 /* Enable quota usage during mount. */
bc98a42c 5050 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 5051 err = ext4_enable_quotas(sb);
5052 if (err)
5053 goto failed_mount8;
5054 }
5055#endif /* CONFIG_QUOTA */
5056
bc71726c 5057 /*
5058 * Save the original bdev mapping's wb_err value which could be
5059 * used to detect the metadata async write error.
5060 */
5061 spin_lock_init(&sbi->s_bdev_wb_lock);
9704a322
ZX		 5062 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5063 &sbi->s_bdev_wb_err);
bc71726c 5064 sb->s_bdev->bd_super = sb;
617ba13b
MC		 5065 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5066 ext4_orphan_cleanup(sb, es);
5067 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 5068 if (needs_recovery) {
b31e1552 5069 ext4_msg(sb, KERN_INFO, "recovery complete");
11215630
JK		 5070 err = ext4_mark_recovery_complete(sb, es);
5071 if (err)
5072 goto failed_mount8;
0390131b
FM		 5073 }
5074 if (EXT4_SB(sb)->s_journal) {
5075 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5076 descr = " journalled data mode";
5077 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5078 descr = " ordered data mode";
5079 else
5080 descr = " writeback data mode";
5081 } else
5082 descr = "out journal";
5083
79add3a3
LC		 5084 if (test_opt(sb, DISCARD)) {
5085 struct request_queue *q = bdev_get_queue(sb->s_bdev);
5086 if (!blk_queue_discard(q))
5087 ext4_msg(sb, KERN_WARNING,
5088 "mounting with \"discard\" option, but "
5089 "the device does not support discard");
5090 }
5091
e294a537
TT		 5092 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5093 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 5094 "Opts: %.*s%s%s", descr,
5095 (int) sizeof(sbi->s_es->s_mount_opts),
5096 sbi->s_es->s_mount_opts,
e294a537 5097 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 5098
66e61a9e
TT		 5099 if (es->s_error_count)
5100 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 5101
efbed4dc
TT		 5102 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5103 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5104 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5105 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
1cf006ed
DM		 5106 atomic_set(&sbi->s_warning_count, 0);
5107 atomic_set(&sbi->s_msg_count, 0);
efbed4dc 5108
d4c402d9 5109 kfree(orig_data);
ac27a0ec
DK		 5110 return 0;
5111
617ba13b 5112cantfind_ext4:
ac27a0ec 5113 if (!silent)
b31e1552 5114 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 5115 goto failed_mount;
5116
72ba7450 5117failed_mount8:
ebd173be 5118 ext4_unregister_sysfs(sb);
cb8d53d2 5119 kobject_put(&sbi->s_kobj);
dcf2d804
TM		 5120failed_mount7:
5121 ext4_unregister_li_request(sb);
5122failed_mount6:
f9ae9cf5 5123 ext4_mb_release(sb);
7c990728
SJS		 5124 rcu_read_lock();
5125 flex_groups = rcu_dereference(sbi->s_flex_groups);
5126 if (flex_groups) {
5127 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5128 kvfree(flex_groups[i]);
5129 kvfree(flex_groups);
5130 }
5131 rcu_read_unlock();
d5e03cbb
TT		 5132 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5133 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5134 percpu_counter_destroy(&sbi->s_dirs_counter);
5135 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
bbd55937 5136 percpu_free_rwsem(&sbi->s_writepages_rwsem);
00764937 5137failed_mount5:
f9ae9cf5
TT		 5138 ext4_ext_release(sb);
5139 ext4_release_system_zone(sb);
5140failed_mount4a:
94bf608a 5141 dput(sb->s_root);
32a9bb57 5142 sb->s_root = NULL;
94bf608a 5143failed_mount4:
b31e1552 5144 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 5145 if (EXT4_SB(sb)->rsv_conversion_wq)
5146 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 5147failed_mount_wq:
50c15df6
CX		 5148 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5149 sbi->s_ea_inode_cache = NULL;
5150
5151 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5152 sbi->s_ea_block_cache = NULL;
5153
0390131b
FM		 5154 if (sbi->s_journal) {
5155 jbd2_journal_destroy(sbi->s_journal);
5156 sbi->s_journal = NULL;
5157 }
50460fe8 5158failed_mount3a:
d3922a77 5159 ext4_es_unregister_shrinker(sbi);
eb68d0e2 5160failed_mount3:
9105bb14 5161 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 5162 if (sbi->s_mmp_tsk)
5163 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 5164failed_mount2:
1d0c3924
TT		 5165 rcu_read_lock();
5166 group_desc = rcu_dereference(sbi->s_group_desc);
ac27a0ec 5167 for (i = 0; i < db_count; i++)
1d0c3924
TT		 5168 brelse(group_desc[i]);
5169 kvfree(group_desc);
5170 rcu_read_unlock();
ac27a0ec 5171failed_mount:
0441984a
DW		 5172 if (sbi->s_chksum_driver)
5173 crypto_free_shash(sbi->s_chksum_driver);
c83ad55e
GKB		 5174
5175#ifdef CONFIG_UNICODE
f8f4acb6 5176 utf8_unload(sb->s_encoding);
c83ad55e
GKB		 5177#endif
5178
ac27a0ec 5179#ifdef CONFIG_QUOTA
a2d4a646 5180 for (i = 0; i < EXT4_MAXQUOTAS; i++)
0ba33fac 5181 kfree(get_qf_name(sb, sbi, i));
ac27a0ec 5182#endif
ac4acb1f 5183 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
617ba13b 5184 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 5185 brelse(bh);
5186out_fail:
5187 sb->s_fs_info = NULL;
f6830165 5188 kfree(sbi->s_blockgroup_lock);
5aee0f8a 5189out_free_base:
ac27a0ec 5190 kfree(sbi);
d4c402d9 5191 kfree(orig_data);
5e405595 5192 fs_put_dax(dax_dev);
07aa2ea1 5193 return err ? err : ret;
ac27a0ec
DK		 5194}
5195
5196/*
5197 * Setup any per-fs journal parameters now. We'll do this both on
5198 * initial mount, once the journal has been initialised but before we've
5199 * done any recovery; and again on any subsequent remount.
5200 */
617ba13b 5201static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 5202{
617ba13b 5203 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5204
30773840
TT		 5205 journal->j_commit_interval = sbi->s_commit_interval;
5206 journal->j_min_batch_time = sbi->s_min_batch_time;
5207 journal->j_max_batch_time = sbi->s_max_batch_time;
6866d7b3 5208 ext4_fc_init(sb, journal);
ac27a0ec 5209
a931da6a 5210 write_lock(&journal->j_state_lock);
ac27a0ec 5211 if (test_opt(sb, BARRIER))
dab291af 5212 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 5213 else
dab291af 5214 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 5215 if (test_opt(sb, DATA_ERR_ABORT))
5216 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5217 else
5218 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 5219 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 5220}
5221
c6cb7e77
EW		 5222static struct inode *ext4_get_journal_inode(struct super_block *sb,
5223 unsigned int journal_inum)
ac27a0ec
DK		 5224{
5225 struct inode *journal_inode;
ac27a0ec 5226
c6cb7e77
EW		 5227 /*
5228 * Test for the existence of a valid inode on disk. Bad things
5229 * happen if we iget() an unused inode, as the subsequent iput()
5230 * will try to delete it.
5231 */
8a363970 5232 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
1d1fe1ee 5233 if (IS_ERR(journal_inode)) {
b31e1552 5234 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 5235 return NULL;
5236 }
5237 if (!journal_inode->i_nlink) {
5238 make_bad_inode(journal_inode);
5239 iput(journal_inode);
b31e1552 5240 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 5241 return NULL;
5242 }
5243
e5f8eab8 5244 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 5245 journal_inode, journal_inode->i_size);
1d1fe1ee 5246 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 5247 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 5248 iput(journal_inode);
5249 return NULL;
5250 }
c6cb7e77
EW		 5251 return journal_inode;
5252}
5253
5254static journal_t *ext4_get_journal(struct super_block *sb,
5255 unsigned int journal_inum)
5256{
5257 struct inode *journal_inode;
5258 journal_t *journal;
5259
11215630
JK		 5260 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5261 return NULL;
c6cb7e77
EW		 5262
5263 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5264 if (!journal_inode)
5265 return NULL;
ac27a0ec 5266
dab291af 5267 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 5268 if (!journal) {
b31e1552 5269 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 5270 iput(journal_inode);
5271 return NULL;
5272 }
5273 journal->j_private = sb;
617ba13b 5274 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 5275 return journal;
5276}
5277
617ba13b 5278static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 5279 dev_t j_dev)
5280{
2b2d6d01 5281 struct buffer_head *bh;
ac27a0ec 5282 journal_t *journal;
617ba13b
MC		 5283 ext4_fsblk_t start;
5284 ext4_fsblk_t len;
ac27a0ec 5285 int hblock, blocksize;
617ba13b 5286 ext4_fsblk_t sb_block;
ac27a0ec 5287 unsigned long offset;
2b2d6d01 5288 struct ext4_super_block *es;
ac27a0ec
DK		 5289 struct block_device *bdev;
5290
11215630
JK		 5291 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5292 return NULL;
0390131b 5293
b31e1552 5294 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 5295 if (bdev == NULL)
5296 return NULL;
5297
ac27a0ec 5298 blocksize = sb->s_blocksize;
e1defc4f 5299 hblock = bdev_logical_block_size(bdev);
ac27a0ec 5300 if (blocksize < hblock) {
b31e1552
ES		 5301 ext4_msg(sb, KERN_ERR,
5302 "blocksize too small for journal device");
ac27a0ec
DK		 5303 goto out_bdev;
5304 }
5305
617ba13b
MC		 5306 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5307 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 5308 set_blocksize(bdev, blocksize);
5309 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 5310 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5311 "external journal");
ac27a0ec
DK		 5312 goto out_bdev;
5313 }
5314
2716b802 5315 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 5316 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 5317 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 5318 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 5319 ext4_msg(sb, KERN_ERR, "external journal has "
5320 "bad superblock");
ac27a0ec
DK		 5321 brelse(bh);
5322 goto out_bdev;
5323 }
5324
df4763be
DW		 5325 if ((le32_to_cpu(es->s_feature_ro_compat) &
5326 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5327 es->s_checksum != ext4_superblock_csum(sb, es)) {
5328 ext4_msg(sb, KERN_ERR, "external journal has "
5329 "corrupt superblock");
5330 brelse(bh);
5331 goto out_bdev;
5332 }
5333
617ba13b 5334 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 5335 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 5336 brelse(bh);
5337 goto out_bdev;
5338 }
5339
bd81d8ee 5340 len = ext4_blocks_count(es);
ac27a0ec
DK		 5341 start = sb_block + 1;
5342 brelse(bh); /* we're done with the superblock */
5343
dab291af 5344 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 5345 start, len, blocksize);
5346 if (!journal) {
b31e1552 5347 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 5348 goto out_bdev;
5349 }
5350 journal->j_private = sb;
2d069c08 5351 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
b31e1552 5352 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 5353 goto out_journal;
5354 }
5355 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 5356 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5357 "user (unsupported) - %d",
ac27a0ec
DK		 5358 be32_to_cpu(journal->j_superblock->s_nr_users));
5359 goto out_journal;
5360 }
ee7ed3aa 5361 EXT4_SB(sb)->s_journal_bdev = bdev;
617ba13b 5362 ext4_init_journal_params(sb, journal);
ac27a0ec 5363 return journal;
0b8e58a1 5364
ac27a0ec 5365out_journal:
dab291af 5366 jbd2_journal_destroy(journal);
ac27a0ec 5367out_bdev:
617ba13b 5368 ext4_blkdev_put(bdev);
ac27a0ec
DK		 5369 return NULL;
5370}
5371
617ba13b
MC		 5372static int ext4_load_journal(struct super_block *sb,
5373 struct ext4_super_block *es,
ac27a0ec
DK		 5374 unsigned long journal_devnum)
5375{
5376 journal_t *journal;
5377 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5378 dev_t journal_dev;
5379 int err = 0;
5380 int really_read_only;
273108fa 5381 int journal_dev_ro;
ac27a0ec 5382
11215630
JK		 5383 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5384 return -EFSCORRUPTED;
0390131b 5385
ac27a0ec
DK		 5386 if (journal_devnum &&
5387 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 5388 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5389 "numbers have changed");
ac27a0ec
DK		 5390 journal_dev = new_decode_dev(journal_devnum);
5391 } else
5392 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5393
273108fa
LC		 5394 if (journal_inum && journal_dev) {
5395 ext4_msg(sb, KERN_ERR,
5396 "filesystem has both journal inode and journal device!");
5397 return -EINVAL;
5398 }
5399
5400 if (journal_inum) {
5401 journal = ext4_get_journal(sb, journal_inum);
5402 if (!journal)
5403 return -EINVAL;
5404 } else {
5405 journal = ext4_get_dev_journal(sb, journal_dev);
5406 if (!journal)
5407 return -EINVAL;
5408 }
5409
5410 journal_dev_ro = bdev_read_only(journal->j_dev);
5411 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5412
5413 if (journal_dev_ro && !sb_rdonly(sb)) {
5414 ext4_msg(sb, KERN_ERR,
5415 "journal device read-only, try mounting with '-o ro'");
5416 err = -EROFS;
5417 goto err_out;
5418 }
ac27a0ec
DK		 5419
5420 /*
5421 * Are we loading a blank journal or performing recovery after a
5422 * crash? For recovery, we need to check in advance whether we
5423 * can get read-write access to the device.
5424 */
e2b911c5 5425 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 5426 if (sb_rdonly(sb)) {
b31e1552
ES		 5427 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5428 "required on readonly filesystem");
ac27a0ec 5429 if (really_read_only) {
b31e1552 5430 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 5431 "unavailable, cannot proceed "
5432 "(try mounting with noload)");
273108fa
LC		 5433 err = -EROFS;
5434 goto err_out;
ac27a0ec 5435 }
b31e1552
ES		 5436 ext4_msg(sb, KERN_INFO, "write access will "
5437 "be enabled during recovery");
ac27a0ec
DK		 5438 }
5439 }
5440
90576c0b 5441 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 5442 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 5443
e2b911c5 5444 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 5445 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 5446 if (!err) {
5447 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5448 if (save)
5449 memcpy(save, ((char *) es) +
5450 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 5451 err = jbd2_journal_load(journal);
1c13d5c0
TT		 5452 if (save)
5453 memcpy(((char *) es) + EXT4_S_ERR_START,
5454 save, EXT4_S_ERR_LEN);
5455 kfree(save);
5456 }
ac27a0ec
DK		 5457
5458 if (err) {
b31e1552 5459 ext4_msg(sb, KERN_ERR, "error loading journal");
273108fa 5460 goto err_out;
ac27a0ec
DK		 5461 }
5462
617ba13b 5463 EXT4_SB(sb)->s_journal = journal;
11215630
JK		 5464 err = ext4_clear_journal_err(sb, es);
5465 if (err) {
5466 EXT4_SB(sb)->s_journal = NULL;
5467 jbd2_journal_destroy(journal);
5468 return err;
5469 }
ac27a0ec 5470
c41303ce 5471 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 5472 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5473 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 5474
5475 /* Make sure we flush the recovery flag to disk. */
e2d67052 5476 ext4_commit_super(sb, 1);
ac27a0ec
DK		 5477 }
5478
5479 return 0;
273108fa
LC		 5480
5481err_out:
5482 jbd2_journal_destroy(journal);
5483 return err;
ac27a0ec
DK		 5484}
5485
e2d67052 5486static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 5487{
e2d67052 5488 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 5489 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 5490 int error = 0;
ac27a0ec 5491
bdfe0cbd 5492 if (!sbh || block_device_ejected(sb))
c4be0c1d 5493 return error;
a17712c8 5494
71290b36
TT		 5495 /*
5496 * If the file system is mounted read-only, don't update the
5497 * superblock write time. This avoids updating the superblock
5498 * write time when we are mounting the root file system
5499 * read/only but we need to replay the journal; at that point,
5500 * for people who are east of GMT and who make their clock
5501 * tick in localtime for Windows bug-for-bug compatibility,
5502 * the clock is set in the future, and this will cause e2fsck
5503 * to complain and force a full file system check.
5504 */
1751e8a6 5505 if (!(sb->s_flags & SB_RDONLY))
6a0678a7 5506 ext4_update_tstamp(es, s_wtime);
8446fe92
CH		 5507 es->s_kbytes_written =
5508 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5509 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
5510 EXT4_SB(sb)->s_sectors_written_start) >> 1));
d5e03cbb
TT		 5511 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5512 ext4_free_blocks_count_set(es,
57042651
TT		 5513 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5514 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 5515 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5516 es->s_free_inodes_count =
5517 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 5518 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 5519 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 5520 ext4_superblock_csum_set(sb);
1566a48a
TT		 5521 if (sync)
5522 lock_buffer(sbh);
e8680786 5523 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4743f839
PKS		 5524 /*
5525 * Oh, dear. A previous attempt to write the
5526 * superblock failed. This could happen because the
5527 * USB device was yanked out. Or it could happen to
5528 * be a transient write error and maybe the block will
5529 * be remapped. Nothing we can do but to retry the
5530 * write and hope for the best.
5531 */
5532 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5533 "superblock detected");
5534 clear_buffer_write_io_error(sbh);
5535 set_buffer_uptodate(sbh);
5536 }
ac27a0ec 5537 mark_buffer_dirty(sbh);
914258bf 5538 if (sync) {
1566a48a 5539 unlock_buffer(sbh);
564bc402 5540 error = __sync_dirty_buffer(sbh,
00473374 5541 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
c89128a0 5542 if (buffer_write_io_error(sbh)) {
b31e1552
ES		 5543 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5544 "superblock");
914258bf
TT		 5545 clear_buffer_write_io_error(sbh);
5546 set_buffer_uptodate(sbh);
5547 }
5548 }
c4be0c1d 5549 return error;
ac27a0ec
DK		 5550}
5551
ac27a0ec
DK		 5552/*
5553 * Have we just finished recovery? If so, and if we are mounting (or
5554 * remounting) the filesystem readonly, then we will end up with a
5555 * consistent fs on disk. Record that fact.
5556 */
11215630
JK		 5557static int ext4_mark_recovery_complete(struct super_block *sb,
5558 struct ext4_super_block *es)
ac27a0ec 5559{
11215630 5560 int err;
617ba13b 5561 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 5562
e2b911c5 5563 if (!ext4_has_feature_journal(sb)) {
11215630
JK		 5564 if (journal != NULL) {
5565 ext4_error(sb, "Journal got removed while the fs was "
5566 "mounted!");
5567 return -EFSCORRUPTED;
5568 }
5569 return 0;
0390131b 5570 }
dab291af 5571 jbd2_journal_lock_updates(journal);
11215630
JK		 5572 err = jbd2_journal_flush(journal);
5573 if (err < 0)
7ffe1ea8
HK		 5574 goto out;
5575
bc98a42c 5576 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
e2b911c5 5577 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 5578 ext4_commit_super(sb, 1);
ac27a0ec 5579 }
7ffe1ea8 5580out:
dab291af 5581 jbd2_journal_unlock_updates(journal);
11215630 5582 return err;
ac27a0ec
DK		 5583}
5584
5585/*
5586 * If we are mounting (or read-write remounting) a filesystem whose journal
5587 * has recorded an error from a previous lifetime, move that error to the
5588 * main filesystem now.
5589 */
11215630 5590static int ext4_clear_journal_err(struct super_block *sb,
2b2d6d01 5591 struct ext4_super_block *es)
ac27a0ec
DK		 5592{
5593 journal_t *journal;
5594 int j_errno;
5595 const char *errstr;
5596
11215630
JK		 5597 if (!ext4_has_feature_journal(sb)) {
5598 ext4_error(sb, "Journal got removed while the fs was mounted!");
5599 return -EFSCORRUPTED;
5600 }
0390131b 5601
617ba13b 5602 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 5603
5604 /*
5605 * Now check for any error status which may have been recorded in the
617ba13b 5606 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 5607 */
5608
dab291af 5609 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 5610 if (j_errno) {
5611 char nbuf[16];
5612
617ba13b 5613 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 5614 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 5615 "from previous mount: %s", errstr);
12062ddd 5616 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 5617
617ba13b
MC		 5618 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5619 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 5620 ext4_commit_super(sb, 1);
ac27a0ec 5621
dab291af 5622 jbd2_journal_clear_err(journal);
d796c52e 5623 jbd2_journal_update_sb_errno(journal);
ac27a0ec 5624 }
11215630 5625 return 0;
ac27a0ec
DK		 5626}
5627
5628/*
5629 * Force the running and committing transactions to commit,
5630 * and wait on the commit.
5631 */
617ba13b 5632int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 5633{
5634 journal_t *journal;
ac27a0ec 5635
bc98a42c 5636 if (sb_rdonly(sb))
ac27a0ec
DK		 5637 return 0;
5638
617ba13b 5639 journal = EXT4_SB(sb)->s_journal;
b1deefc9 5640 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 5641}
5642
617ba13b 5643static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 5644{
14ce0cb4 5645 int ret = 0;
9eddacf9 5646 tid_t target;
06a407f1 5647 bool needs_barrier = false;
8d5d02e6 5648 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5649
49598e04 5650 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 5651 return 0;
5652
9bffad1e 5653 trace_ext4_sync_fs(sb, wait);
2e8fa54e 5654 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 5655 /*
5656 * Writeback quota in non-journalled quota case - journalled quota has
5657 * no dirty dquots
5658 */
5659 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 5660 /*
5661 * Data writeback is possible w/o journal transaction, so barrier must
5662 * being sent at the end of the function. But we can skip it if
5663 * transaction_commit will do it for us.
5664 */
bda32530
TT		 5665 if (sbi->s_journal) {
5666 target = jbd2_get_latest_transaction(sbi->s_journal);
5667 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5668 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5669 needs_barrier = true;
5670
5671 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5672 if (wait)
5673 ret = jbd2_log_wait_commit(sbi->s_journal,
5674 target);
5675 }
5676 } else if (wait && test_opt(sb, BARRIER))
06a407f1 5677 needs_barrier = true;
06a407f1
DM		 5678 if (needs_barrier) {
5679 int err;
9398554f 5680 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL);
06a407f1
DM		 5681 if (!ret)
5682 ret = err;
0390131b 5683 }
06a407f1
DM		 5684
5685 return ret;
5686}
5687
ac27a0ec
DK		 5688/*
5689 * LVM calls this function before a (read-only) snapshot is created. This
5690 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 5691 *
5692 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 5693 * state independently. It relies on upper layer to stop all data & metadata
5694 * modifications.
ac27a0ec 5695 */
c4be0c1d 5696static int ext4_freeze(struct super_block *sb)
ac27a0ec 5697{
c4be0c1d
TS		 5698 int error = 0;
5699 journal_t *journal;
ac27a0ec 5700
bc98a42c 5701 if (sb_rdonly(sb))
9ca92389 5702 return 0;
ac27a0ec 5703
9ca92389 5704 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 5705
bb044576
TT		 5706 if (journal) {
5707 /* Now we set up the journal barrier. */
5708 jbd2_journal_lock_updates(journal);
ac27a0ec 5709
bb044576
TT		 5710 /*
5711 * Don't clear the needs_recovery flag if we failed to
5712 * flush the journal.
5713 */
5714 error = jbd2_journal_flush(journal);
5715 if (error < 0)
5716 goto out;
c642dc9e
ES		 5717
5718 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 5719 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 5720 }
9ca92389 5721
9ca92389 5722 error = ext4_commit_super(sb, 1);
6b0310fb 5723out:
bb044576
TT		 5724 if (journal)
5725 /* we rely on upper layer to stop further updates */
5726 jbd2_journal_unlock_updates(journal);
6b0310fb 5727 return error;
ac27a0ec
DK		 5728}
5729
5730/*
5731 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5732 * flag here, even though the filesystem is not technically dirty yet.
5733 */
c4be0c1d 5734static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 5735{
bc98a42c 5736 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 5737 return 0;
5738
c642dc9e
ES		 5739 if (EXT4_SB(sb)->s_journal) {
5740 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 5741 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 5742 }
5743
9ca92389 5744 ext4_commit_super(sb, 1);
c4be0c1d 5745 return 0;
ac27a0ec
DK		 5746}
5747
673c6100
TT		 5748/*
5749 * Structure to save mount options for ext4_remount's benefit
5750 */
5751struct ext4_mount_options {
5752 unsigned long s_mount_opt;
a2595b8a 5753 unsigned long s_mount_opt2;
08cefc7a
EB		 5754 kuid_t s_resuid;
5755 kgid_t s_resgid;
673c6100
TT		 5756 unsigned long s_commit_interval;
5757 u32 s_min_batch_time, s_max_batch_time;
5758#ifdef CONFIG_QUOTA
5759 int s_jquota_fmt;
a2d4a646 5760 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 5761#endif
5762};
5763
2b2d6d01 5764static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 5765{
2b2d6d01 5766 struct ext4_super_block *es;
617ba13b 5767 struct ext4_sb_info *sbi = EXT4_SB(sb);
f25391eb 5768 unsigned long old_sb_flags, vfs_flags;
617ba13b 5769 struct ext4_mount_options old_opts;
c79d967d 5770 int enable_quota = 0;
8a266467 5771 ext4_group_t g;
b3881f74 5772 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 5773 int err = 0;
ac27a0ec 5774#ifdef CONFIG_QUOTA
03dafb5f 5775 int i, j;
33458eab 5776 char *to_free[EXT4_MAXQUOTAS];
ac27a0ec 5777#endif
d4c402d9 5778 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec 5779
21ac738e
CX		 5780 if (data && !orig_data)
5781 return -ENOMEM;
5782
ac27a0ec
DK		 5783 /* Store the original options */
5784 old_sb_flags = sb->s_flags;
5785 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 5786 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 5787 old_opts.s_resuid = sbi->s_resuid;
5788 old_opts.s_resgid = sbi->s_resgid;
5789 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 5790 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5791 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 5792#ifdef CONFIG_QUOTA
5793 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 5794 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5795 if (sbi->s_qf_names[i]) {
33458eab
TT		 5796 char *qf_name = get_qf_name(sb, sbi, i);
5797
5798 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
03dafb5f
CG		 5799 if (!old_opts.s_qf_names[i]) {
5800 for (j = 0; j < i; j++)
5801 kfree(old_opts.s_qf_names[j]);
3e36a163 5802 kfree(orig_data);
03dafb5f
CG		 5803 return -ENOMEM;
5804 }
5805 } else
5806 old_opts.s_qf_names[i] = NULL;
ac27a0ec 5807#endif
b3881f74
TT		 5808 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5809 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 5810
f25391eb
LC		 5811 /*
5812 * Some options can be enabled by ext4 and/or by VFS mount flag
5813 * either way we need to make sure it matches in both *flags and
5814 * s_flags. Copy those selected flags from *flags to s_flags
5815 */
5816 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5817 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5818
661aa520 5819 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 5820 err = -EINVAL;
5821 goto restore_opts;
5822 }
5823
6b992ff2 5824 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 5825 test_opt(sb, JOURNAL_CHECKSUM)) {
5826 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 5827 "during remount not supported; ignoring");
5828 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 5829 }
5830
6ae6514b
PS		 5831 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5832 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5833 ext4_msg(sb, KERN_ERR, "can't mount with "
5834 "both data=journal and delalloc");
5835 err = -EINVAL;
5836 goto restore_opts;
5837 }
5838 if (test_opt(sb, DIOREAD_NOLOCK)) {
5839 ext4_msg(sb, KERN_ERR, "can't mount with "
5840 "both data=journal and dioread_nolock");
5841 err = -EINVAL;
5842 goto restore_opts;
5843 }
ab04df78
JK		 5844 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5845 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5846 ext4_msg(sb, KERN_ERR, "can't mount with "
5847 "journal_async_commit in data=ordered mode");
5848 err = -EINVAL;
5849 goto restore_opts;
5850 }
923ae0ff
RZ		 5851 }
5852
cdb7ee4c
TE		 5853 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5854 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5855 err = -EINVAL;
5856 goto restore_opts;
5857 }
5858
9b5f6c9b 5859 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
54d3adbc 5860 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
ac27a0ec 5861
1751e8a6
LT		 5862 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5863 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 5864
5865 es = sbi->s_es;
5866
b3881f74 5867 if (sbi->s_journal) {
0390131b 5868 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5869 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5870 }
ac27a0ec 5871
1751e8a6 5872 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
9b5f6c9b 5873 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
ac27a0ec
DK		 5874 err = -EROFS;
5875 goto restore_opts;
5876 }
5877
1751e8a6 5878 if (*flags & SB_RDONLY) {
38c03b34
TT		 5879 err = sync_filesystem(sb);
5880 if (err < 0)
5881 goto restore_opts;
0f0dd62f
CH		 5882 err = dquot_suspend(sb, -1);
5883 if (err < 0)
c79d967d 5884 goto restore_opts;
c79d967d 5885
ac27a0ec
DK		 5886 /*
5887 * First of all, the unconditional stuff we have to do
5888 * to disable replay of the journal when we next remount
5889 */
1751e8a6 5890 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 5891
5892 /*
5893 * OK, test if we are remounting a valid rw partition
5894 * readonly, and if so set the rdonly flag and then
5895 * mark the partition as valid again.
5896 */
617ba13b
MC		 5897 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5898 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5899 es->s_state = cpu_to_le16(sbi->s_mount_state);
5900
11215630
JK		 5901 if (sbi->s_journal) {
5902 /*
5903 * We let remount-ro finish even if marking fs
5904 * as clean failed...
5905 */
0390131b 5906 ext4_mark_recovery_complete(sb, es);
11215630 5907 }
2dca60d9
TT		 5908 if (sbi->s_mmp_tsk)
5909 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 5910 } else {
a13fb1a4 5911 /* Make sure we can mount this feature set readwrite */
e2b911c5 5912 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5913 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5914 err = -EROFS;
5915 goto restore_opts;
5916 }
8a266467
TT		 5917 /*
5918 * Make sure the group descriptor checksums
0b8e58a1 5919 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5920 */
5921 for (g = 0; g < sbi->s_groups_count; g++) {
5922 struct ext4_group_desc *gdp =
5923 ext4_get_group_desc(sb, g, NULL);
5924
feb0ab32 5925 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5926 ext4_msg(sb, KERN_ERR,
5927 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5928 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5929 le16_to_cpu(gdp->bg_checksum));
6a797d27 5930 err = -EFSBADCRC;
8a266467
TT		 5931 goto restore_opts;
5932 }
5933 }
5934
ead6596b
ES		 5935 /*
5936 * If we have an unprocessed orphan list hanging
5937 * around from a previously readonly bdev mount,
5938 * require a full umount/remount for now.
5939 */
5940 if (es->s_last_orphan) {
b31e1552 5941 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5942 "remount RDWR because of unprocessed "
5943 "orphan inode list. Please "
b31e1552 5944 "umount/remount instead");
ead6596b
ES		 5945 err = -EINVAL;
5946 goto restore_opts;
5947 }
5948
ac27a0ec
DK		 5949 /*
5950 * Mounting a RDONLY partition read-write, so reread
5951 * and store the current valid flag. (It may have
5952 * been changed by e2fsck since we originally mounted
5953 * the partition.)
5954 */
11215630
JK		 5955 if (sbi->s_journal) {
5956 err = ext4_clear_journal_err(sb, es);
5957 if (err)
5958 goto restore_opts;
5959 }
ac27a0ec 5960 sbi->s_mount_state = le16_to_cpu(es->s_state);
c89128a0
JK		 5961
5962 err = ext4_setup_super(sb, es, 0);
5963 if (err)
5964 goto restore_opts;
5965
5966 sb->s_flags &= ~SB_RDONLY;
e2b911c5 5967 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5968 if (ext4_multi_mount_protect(sb,
5969 le64_to_cpu(es->s_mmp_block))) {
5970 err = -EROFS;
5971 goto restore_opts;
5972 }
c79d967d 5973 enable_quota = 1;
ac27a0ec
DK		 5974 }
5975 }
bfff6873
LC		 5976
5977 /*
5978 * Reinitialize lazy itable initialization thread based on
5979 * current settings
5980 */
bc98a42c 5981 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 5982 ext4_unregister_li_request(sb);
5983 else {
5984 ext4_group_t first_not_zeroed;
5985 first_not_zeroed = ext4_has_uninit_itable(sb);
5986 ext4_register_li_request(sb, first_not_zeroed);
5987 }
5988
0f5bde1d
JK		 5989 /*
5990 * Handle creation of system zone data early because it can fail.
5991 * Releasing of existing data is done when we are sure remount will
5992 * succeed.
5993 */
dd0db94f 5994 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
0f5bde1d
JK		 5995 err = ext4_setup_system_zone(sb);
5996 if (err)
5997 goto restore_opts;
5998 }
d176b1f6 5999
c89128a0
JK		 6000 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6001 err = ext4_commit_super(sb, 1);
6002 if (err)
6003 goto restore_opts;
6004 }
0390131b 6005
ac27a0ec
DK		 6006#ifdef CONFIG_QUOTA
6007 /* Release old quota file names */
a2d4a646 6008 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 6009 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 6010 if (enable_quota) {
6011 if (sb_any_quota_suspended(sb))
6012 dquot_resume(sb, -1);
e2b911c5 6013 else if (ext4_has_feature_quota(sb)) {
7c319d32 6014 err = ext4_enable_quotas(sb);
07724f98 6015 if (err)
7c319d32 6016 goto restore_opts;
7c319d32
AK		 6017 }
6018 }
ac27a0ec 6019#endif
dd0db94f 6020 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
0f5bde1d 6021 ext4_release_system_zone(sb);
d4c402d9 6022
f25391eb
LC		 6023 /*
6024 * Some options can be enabled by ext4 and/or by VFS mount flag
6025 * either way we need to make sure it matches in both *flags and
6026 * s_flags. Copy those selected flags from s_flags to *flags
6027 */
6028 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
d4c402d9
CW		 6029
6030 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
6031 kfree(orig_data);
ac27a0ec 6032 return 0;
0b8e58a1 6033
ac27a0ec
DK		 6034restore_opts:
6035 sb->s_flags = old_sb_flags;
6036 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 6037 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 6038 sbi->s_resuid = old_opts.s_resuid;
6039 sbi->s_resgid = old_opts.s_resgid;
6040 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 6041 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6042 sbi->s_max_batch_time = old_opts.s_max_batch_time;
dd0db94f 6043 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
0f5bde1d 6044 ext4_release_system_zone(sb);
ac27a0ec
DK		 6045#ifdef CONFIG_QUOTA
6046 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 6047 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
33458eab
TT		 6048 to_free[i] = get_qf_name(sb, sbi, i);
6049 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
ac27a0ec 6050 }
33458eab
TT		 6051 synchronize_rcu();
6052 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6053 kfree(to_free[i]);
ac27a0ec 6054#endif
d4c402d9 6055 kfree(orig_data);
ac27a0ec
DK		 6056 return err;
6057}
6058
689c958c
LX		 6059#ifdef CONFIG_QUOTA
6060static int ext4_statfs_project(struct super_block *sb,
6061 kprojid_t projid, struct kstatfs *buf)
6062{
6063 struct kqid qid;
6064 struct dquot *dquot;
6065 u64 limit;
6066 u64 curblock;
6067
6068 qid = make_kqid_projid(projid);
6069 dquot = dqget(sb, qid);
6070 if (IS_ERR(dquot))
6071 return PTR_ERR(dquot);
7b9ca4c6 6072 spin_lock(&dquot->dq_dqb_lock);
689c958c 6073
a08fe66e
CX		 6074 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6075 dquot->dq_dqb.dqb_bhardlimit);
57c32ea4
CX		 6076 limit >>= sb->s_blocksize_bits;
6077
689c958c 6078 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 6079 curblock = (dquot->dq_dqb.dqb_curspace +
6080 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 6081 buf->f_blocks = limit;
6082 buf->f_bfree = buf->f_bavail =
6083 (buf->f_blocks > curblock) ?
6084 (buf->f_blocks - curblock) : 0;
6085 }
6086
a08fe66e
CX		 6087 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6088 dquot->dq_dqb.dqb_ihardlimit);
689c958c
LX		 6089 if (limit && buf->f_files > limit) {
6090 buf->f_files = limit;
6091 buf->f_ffree =
6092 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6093 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6094 }
6095
7b9ca4c6 6096 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 6097 dqput(dquot);
6098 return 0;
6099}
6100#endif
6101
2b2d6d01 6102static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 6103{
6104 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 6105 struct ext4_sb_info *sbi = EXT4_SB(sb);
6106 struct ext4_super_block *es = sbi->s_es;
27dd4385 6107 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 6108 u64 fsid;
d02a9391 6109 s64 bfree;
27dd4385 6110 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 6111
952fc18e
TT		 6112 if (!test_opt(sb, MINIX_DF))
6113 overhead = sbi->s_overhead;
ac27a0ec 6114
617ba13b 6115 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 6116 buf->f_bsize = sb->s_blocksize;
b72f78cb 6117 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 6118 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6119 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 6120 /* prevent underflow in case that few free space is available */
57042651 6121 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 6122 buf->f_bavail = buf->f_bfree -
6123 (ext4_r_blocks_count(es) + resv_blocks);
6124 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 6125 buf->f_bavail = 0;
6126 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 6127 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 6128 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 6129 fsid = le64_to_cpup((void *)es->s_uuid) ^
6130 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
6d1349c7 6131 buf->f_fsid = u64_to_fsid(fsid);
0b8e58a1 6132
689c958c
LX		 6133#ifdef CONFIG_QUOTA
6134 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6135 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6136 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6137#endif
ac27a0ec
DK		 6138 return 0;
6139}
6140
ac27a0ec
DK		 6141
6142#ifdef CONFIG_QUOTA
6143
bc8230ee
JK		 6144/*
6145 * Helper functions so that transaction is started before we acquire dqio_sem
6146 * to keep correct lock ordering of transaction > dqio_sem
6147 */
ac27a0ec
DK		 6148static inline struct inode *dquot_to_inode(struct dquot *dquot)
6149{
4c376dca 6150 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 6151}
6152
617ba13b 6153static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 6154{
6155 int ret, err;
6156 handle_t *handle;
6157 struct inode *inode;
6158
6159 inode = dquot_to_inode(dquot);
9924a92a 6160 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 6161 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 6162 if (IS_ERR(handle))
6163 return PTR_ERR(handle);
6164 ret = dquot_commit(dquot);
617ba13b 6165 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6166 if (!ret)
6167 ret = err;
6168 return ret;
6169}
6170
617ba13b 6171static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 6172{
6173 int ret, err;
6174 handle_t *handle;
6175
9924a92a 6176 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 6177 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 6178 if (IS_ERR(handle))
6179 return PTR_ERR(handle);
6180 ret = dquot_acquire(dquot);
617ba13b 6181 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6182 if (!ret)
6183 ret = err;
6184 return ret;
6185}
6186
617ba13b 6187static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 6188{
6189 int ret, err;
6190 handle_t *handle;
6191
9924a92a 6192 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 6193 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 6194 if (IS_ERR(handle)) {
6195 /* Release dquot anyway to avoid endless cycle in dqput() */
6196 dquot_release(dquot);
ac27a0ec 6197 return PTR_ERR(handle);
9c3013e9 6198 }
ac27a0ec 6199 ret = dquot_release(dquot);
617ba13b 6200 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6201 if (!ret)
6202 ret = err;
6203 return ret;
6204}
6205
617ba13b 6206static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 6207{
262b4662
JK		 6208 struct super_block *sb = dquot->dq_sb;
6209 struct ext4_sb_info *sbi = EXT4_SB(sb);
6210
2c8be6b2 6211 /* Are we journaling quotas? */
e2b911c5 6212 if (ext4_has_feature_quota(sb) ||
262b4662 6213 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 6214 dquot_mark_dquot_dirty(dquot);
617ba13b 6215 return ext4_write_dquot(dquot);
ac27a0ec
DK		 6216 } else {
6217 return dquot_mark_dquot_dirty(dquot);
6218 }
6219}
6220
617ba13b 6221static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 6222{
6223 int ret, err;
6224 handle_t *handle;
6225
6226 /* Data block + inode block */
2b0143b5 6227 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 6228 if (IS_ERR(handle))
6229 return PTR_ERR(handle);
6230 ret = dquot_commit_info(sb, type);
617ba13b 6231 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6232 if (!ret)
6233 ret = err;
6234 return ret;
6235}
6236
6237/*
6238 * Turn on quotas during mount time - we need to find
6239 * the quota file and such...
6240 */
617ba13b 6241static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 6242{
33458eab 6243 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
287a8095 6244 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 6245}
6246
daf647d2
TT		 6247static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6248{
6249 struct ext4_inode_info *ei = EXT4_I(inode);
6250
6251 /* The first argument of lockdep_set_subclass has to be
6252 * *exactly* the same as the argument to init_rwsem() --- in
6253 * this case, in init_once() --- or lockdep gets unhappy
6254 * because the name of the lock is set using the
6255 * stringification of the argument to init_rwsem().
6256 */
6257 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6258 lockdep_set_subclass(&ei->i_data_sem, subclass);
6259}
6260
ac27a0ec
DK		 6261/*
6262 * Standard function to be called on quota_on
6263 */
617ba13b 6264static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 6265 const struct path *path)
ac27a0ec
DK		 6266{
6267 int err;
ac27a0ec
DK		 6268
6269 if (!test_opt(sb, QUOTA))
6270 return -EINVAL;
0623543b 6271
ac27a0ec 6272 /* Quotafile not on the same filesystem? */
d8c9584e 6273 if (path->dentry->d_sb != sb)
ac27a0ec 6274 return -EXDEV;
e0770e91
JK		 6275
6276 /* Quota already enabled for this file? */
6277 if (IS_NOQUOTA(d_inode(path->dentry)))
6278 return -EBUSY;
6279
0623543b
JK		 6280 /* Journaling quota? */
6281 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 6282 /* Quotafile not in fs root? */
f00c9e44 6283 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 6284 ext4_msg(sb, KERN_WARNING,
6285 "Quota file not on filesystem root. "
6286 "Journaled quota will not work");
91389240
JK		 6287 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6288 } else {
6289 /*
6290 * Clear the flag just in case mount options changed since
6291 * last time.
6292 */
6293 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 6294 }
0623543b
JK		 6295
6296 /*
6297 * When we journal data on quota file, we have to flush journal to see
6298 * all updates to the file when we bypass pagecache...
6299 */
0390131b 6300 if (EXT4_SB(sb)->s_journal &&
2b0143b5 6301 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 6302 /*
6303 * We don't need to lock updates but journal_flush() could
6304 * otherwise be livelocked...
6305 */
6306 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 6307 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 6308 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 6309 if (err)
7ffe1ea8 6310 return err;
0623543b 6311 }
957153fc 6312
daf647d2
TT		 6313 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6314 err = dquot_quota_on(sb, type, format_id, path);
957153fc 6315 if (err) {
daf647d2
TT		 6316 lockdep_set_quota_inode(path->dentry->d_inode,
6317 I_DATA_SEM_NORMAL);
957153fc
JK		 6318 } else {
6319 struct inode *inode = d_inode(path->dentry);
6320 handle_t *handle;
6321
61a92987
JK		 6322 /*
6323 * Set inode flags to prevent userspace from messing with quota
6324 * files. If this fails, we return success anyway since quotas
6325 * are already enabled and this is not a hard failure.
6326 */
957153fc
JK		 6327 inode_lock(inode);
6328 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6329 if (IS_ERR(handle))
6330 goto unlock_inode;
6331 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6332 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6333 S_NOATIME | S_IMMUTABLE);
4209ae12 6334 err = ext4_mark_inode_dirty(handle, inode);
957153fc
JK		 6335 ext4_journal_stop(handle);
6336 unlock_inode:
6337 inode_unlock(inode);
6338 }
daf647d2 6339 return err;
ac27a0ec
DK		 6340}
6341
7c319d32
AK		 6342static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6343 unsigned int flags)
6344{
6345 int err;
6346 struct inode *qf_inode;
a2d4a646 6347 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 6348 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 6349 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6350 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 6351 };
6352
e2b911c5 6353 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 6354
6355 if (!qf_inums[type])
6356 return -EPERM;
6357
8a363970 6358 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7c319d32
AK		 6359 if (IS_ERR(qf_inode)) {
6360 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6361 return PTR_ERR(qf_inode);
6362 }
6363
bcb13850
JK		 6364 /* Don't account quota for quota files to avoid recursion */
6365 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 6366 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7212b95e 6367 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
daf647d2
TT		 6368 if (err)
6369 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
61157b24 6370 iput(qf_inode);
7c319d32
AK		 6371
6372 return err;
6373}
6374
6375/* Enable usage tracking for all quota types. */
6376static int ext4_enable_quotas(struct super_block *sb)
6377{
6378 int type, err = 0;
a2d4a646 6379 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 6380 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 6381 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6382 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 6383 };
49da9392
JK		 6384 bool quota_mopt[EXT4_MAXQUOTAS] = {
6385 test_opt(sb, USRQUOTA),
6386 test_opt(sb, GRPQUOTA),
6387 test_opt(sb, PRJQUOTA),
6388 };
7c319d32 6389
91389240 6390 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 6391 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 6392 if (qf_inums[type]) {
6393 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 6394 DQUOT_USAGE_ENABLED |
6395 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 6396 if (err) {
6397 ext4_warning(sb,
72ba7450
TT		 6398 "Failed to enable quota tracking "
6399 "(type=%d, err=%d). Please run "
6400 "e2fsck to fix.", type, err);
7f144fd0
JU		 6401 for (type--; type >= 0; type--)
6402 dquot_quota_off(sb, type);
6403
7c319d32
AK		 6404 return err;
6405 }
6406 }
6407 }
6408 return 0;
6409}
6410
ca0e05e4
DM		 6411static int ext4_quota_off(struct super_block *sb, int type)
6412{
21f97697
JK		 6413 struct inode *inode = sb_dqopt(sb)->files[type];
6414 handle_t *handle;
957153fc 6415 int err;
21f97697 6416
87009d86
DM		 6417 /* Force all delayed allocation blocks to be allocated.
6418 * Caller already holds s_umount sem */
6419 if (test_opt(sb, DELALLOC))
ca0e05e4 6420 sync_filesystem(sb);
ca0e05e4 6421
957153fc 6422 if (!inode || !igrab(inode))
0b268590
AG		 6423 goto out;
6424
957153fc 6425 err = dquot_quota_off(sb, type);
964edf66 6426 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 6427 goto out_put;
6428
6429 inode_lock(inode);
61a92987
JK		 6430 /*
6431 * Update modification times of quota files when userspace can
6432 * start looking at them. If we fail, we return success anyway since
6433 * this is not a hard failure and quotas are already disabled.
6434 */
9924a92a 6435 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
4209ae12
HS		 6436 if (IS_ERR(handle)) {
6437 err = PTR_ERR(handle);
957153fc 6438 goto out_unlock;
4209ae12 6439 }
957153fc
JK		 6440 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6441 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 6442 inode->i_mtime = inode->i_ctime = current_time(inode);
4209ae12 6443 err = ext4_mark_inode_dirty(handle, inode);
21f97697 6444 ext4_journal_stop(handle);
957153fc
JK		 6445out_unlock:
6446 inode_unlock(inode);
6447out_put:
964edf66 6448 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 6449 iput(inode);
6450 return err;
21f97697 6451out:
ca0e05e4
DM		 6452 return dquot_quota_off(sb, type);
6453}
6454
ac27a0ec
DK		 6455/* Read data from quotafile - avoid pagecache and such because we cannot afford
6456 * acquiring the locks... As quota files are never truncated and quota code
25985edc 6457 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 6458 * we don't have to be afraid of races */
617ba13b 6459static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 6460 size_t len, loff_t off)
6461{
6462 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 6463 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 6464 int offset = off & (sb->s_blocksize - 1);
6465 int tocopy;
6466 size_t toread;
6467 struct buffer_head *bh;
6468 loff_t i_size = i_size_read(inode);
6469
6470 if (off > i_size)
6471 return 0;
6472 if (off+len > i_size)
6473 len = i_size-off;
6474 toread = len;
6475 while (toread > 0) {
6476 tocopy = sb->s_blocksize - offset < toread ?
6477 sb->s_blocksize - offset : toread;
1c215028
TT		 6478 bh = ext4_bread(NULL, inode, blk, 0);
6479 if (IS_ERR(bh))
6480 return PTR_ERR(bh);
ac27a0ec
DK		 6481 if (!bh) /* A hole? */
6482 memset(data, 0, tocopy);
6483 else
6484 memcpy(data, bh->b_data+offset, tocopy);
6485 brelse(bh);
6486 offset = 0;
6487 toread -= tocopy;
6488 data += tocopy;
6489 blk++;
6490 }
6491 return len;
6492}
6493
6494/* Write to quotafile (we know the transaction is already started and has
6495 * enough credits) */
617ba13b 6496static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 6497 const char *data, size_t len, loff_t off)
6498{
6499 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 6500 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4209ae12 6501 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
c5e298ae 6502 int retries = 0;
ac27a0ec
DK		 6503 struct buffer_head *bh;
6504 handle_t *handle = journal_current_handle();
6505
0390131b 6506 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 6507 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6508 " cancelled because transaction is not started",
9c3013e9
JK		 6509 (unsigned long long)off, (unsigned long long)len);
6510 return -EIO;
6511 }
67eeb568
DM		 6512 /*
6513 * Since we account only one data block in transaction credits,
6514 * then it is impossible to cross a block boundary.
6515 */
6516 if (sb->s_blocksize - offset < len) {
6517 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6518 " cancelled because not block aligned",
6519 (unsigned long long)off, (unsigned long long)len);
6520 return -EIO;
6521 }
6522
c5e298ae
TT		 6523 do {
6524 bh = ext4_bread(handle, inode, blk,
6525 EXT4_GET_BLOCKS_CREATE |
6526 EXT4_GET_BLOCKS_METADATA_NOFAIL);
45586c70 6527 } while (PTR_ERR(bh) == -ENOSPC &&
c5e298ae 6528 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 6529 if (IS_ERR(bh))
6530 return PTR_ERR(bh);
67eeb568
DM		 6531 if (!bh)
6532 goto out;
5d601255 6533 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 6534 err = ext4_journal_get_write_access(handle, bh);
6535 if (err) {
6536 brelse(bh);
1c215028 6537 return err;
ac27a0ec 6538 }
67eeb568
DM		 6539 lock_buffer(bh);
6540 memcpy(bh->b_data+offset, data, len);
6541 flush_dcache_page(bh->b_page);
6542 unlock_buffer(bh);
62d2b5f2 6543 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 6544 brelse(bh);
ac27a0ec 6545out:
67eeb568
DM		 6546 if (inode->i_size < off + len) {
6547 i_size_write(inode, off + len);
617ba13b 6548 EXT4_I(inode)->i_disksize = inode->i_size;
4209ae12
HS		 6549 err2 = ext4_mark_inode_dirty(handle, inode);
6550 if (unlikely(err2 && !err))
6551 err = err2;
ac27a0ec 6552 }
4209ae12 6553 return err ? err : len;
ac27a0ec 6554}
ac27a0ec
DK		 6555#endif
6556
152a0836
AV		 6557static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6558 const char *dev_name, void *data)
ac27a0ec 6559{
152a0836 6560 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 6561}
6562
c290ea01 6563#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 6564static inline void register_as_ext2(void)
6565{
6566 int err = register_filesystem(&ext2_fs_type);
6567 if (err)
6568 printk(KERN_WARNING
6569 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6570}
6571
6572static inline void unregister_as_ext2(void)
6573{
6574 unregister_filesystem(&ext2_fs_type);
6575}
2035e776
TT		 6576
6577static inline int ext2_feature_set_ok(struct super_block *sb)
6578{
e2b911c5 6579 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 6580 return 0;
bc98a42c 6581 if (sb_rdonly(sb))
2035e776 6582 return 1;
e2b911c5 6583 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 6584 return 0;
6585 return 1;
6586}
24b58424
TT		 6587#else
6588static inline void register_as_ext2(void) { }
6589static inline void unregister_as_ext2(void) { }
2035e776 6590static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 6591#endif
6592
24b58424
TT		 6593static inline void register_as_ext3(void)
6594{
6595 int err = register_filesystem(&ext3_fs_type);
6596 if (err)
6597 printk(KERN_WARNING
6598 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6599}
6600
6601static inline void unregister_as_ext3(void)
6602{
6603 unregister_filesystem(&ext3_fs_type);
6604}
2035e776
TT		 6605
6606static inline int ext3_feature_set_ok(struct super_block *sb)
6607{
e2b911c5 6608 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 6609 return 0;
e2b911c5 6610 if (!ext4_has_feature_journal(sb))
2035e776 6611 return 0;
bc98a42c 6612 if (sb_rdonly(sb))
2035e776 6613 return 1;
e2b911c5 6614 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 6615 return 0;
6616 return 1;
6617}
24b58424 6618
03010a33
TT		 6619static struct file_system_type ext4_fs_type = {
6620 .owner = THIS_MODULE,
6621 .name = "ext4",
152a0836 6622 .mount = ext4_mount,
03010a33
TT		 6623 .kill_sb = kill_block_super,
6624 .fs_flags = FS_REQUIRES_DEV,
6625};
7f78e035 6626MODULE_ALIAS_FS("ext4");
03010a33 6627
e9e3bcec
ES		 6628/* Shared across all ext4 file systems */
6629wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 6630
5dabfc78 6631static int __init ext4_init_fs(void)
ac27a0ec 6632{
e9e3bcec 6633 int i, err;
c9de560d 6634
e294a537 6635 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 6636 ext4_li_info = NULL;
6637 mutex_init(&ext4_li_mtx);
6638
9a4c8019 6639 /* Build-time check for flags consistency */
12e9b892 6640 ext4_check_flag_values();
e9e3bcec 6641
e142d052 6642 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 6643 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 6644
51865fda 6645 err = ext4_init_es();
6fd058f7
TT		 6646 if (err)
6647 return err;
51865fda 6648
1dc0aa46 6649 err = ext4_init_pending();
22cfe4b4
EB		 6650 if (err)
6651 goto out7;
6652
6653 err = ext4_init_post_read_processing();
1dc0aa46
EW		 6654 if (err)
6655 goto out6;
6656
51865fda
ZL		 6657 err = ext4_init_pageio();
6658 if (err)
b5799018 6659 goto out5;
51865fda 6660
5dabfc78 6661 err = ext4_init_system_zone();
bd2d0210 6662 if (err)
b5799018 6663 goto out4;
857ac889 6664
b5799018 6665 err = ext4_init_sysfs();
dd68314c 6666 if (err)
b5799018 6667 goto out3;
857ac889 6668
5dabfc78 6669 err = ext4_init_mballoc();
c9de560d
AT		 6670 if (err)
6671 goto out2;
ac27a0ec
DK		 6672 err = init_inodecache();
6673 if (err)
6674 goto out1;
aa75f4d3
HS		 6675
6676 err = ext4_fc_init_dentry_cache();
6677 if (err)
6678 goto out05;
6679
24b58424 6680 register_as_ext3();
2035e776 6681 register_as_ext2();
03010a33 6682 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 6683 if (err)
6684 goto out;
bfff6873 6685
ac27a0ec
DK		 6686 return 0;
6687out:
24b58424
TT		 6688 unregister_as_ext2();
6689 unregister_as_ext3();
aa75f4d3 6690out05:
ac27a0ec
DK		 6691 destroy_inodecache();
6692out1:
5dabfc78 6693 ext4_exit_mballoc();
9c191f70 6694out2:
b5799018
TT		 6695 ext4_exit_sysfs();
6696out3:
5dabfc78 6697 ext4_exit_system_zone();
b5799018 6698out4:
5dabfc78 6699 ext4_exit_pageio();
b5799018 6700out5:
22cfe4b4 6701 ext4_exit_post_read_processing();
1dc0aa46 6702out6:
22cfe4b4
EB		 6703 ext4_exit_pending();
6704out7:
51865fda
ZL		 6705 ext4_exit_es();
6706
ac27a0ec
DK		 6707 return err;
6708}
6709
5dabfc78 6710static void __exit ext4_exit_fs(void)
ac27a0ec 6711{
bfff6873 6712 ext4_destroy_lazyinit_thread();
24b58424
TT		 6713 unregister_as_ext2();
6714 unregister_as_ext3();
03010a33 6715 unregister_filesystem(&ext4_fs_type);
ac27a0ec 6716 destroy_inodecache();
5dabfc78 6717 ext4_exit_mballoc();
b5799018 6718 ext4_exit_sysfs();
5dabfc78
TT		 6719 ext4_exit_system_zone();
6720 ext4_exit_pageio();
22cfe4b4 6721 ext4_exit_post_read_processing();
dd12ed14 6722 ext4_exit_es();
1dc0aa46 6723 ext4_exit_pending();
ac27a0ec
DK		 6724}
6725
6726MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 6727MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 6728MODULE_LICENSE("GPL");
7ef79ad5 6729MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 6730module_init(ext4_init_fs)
6731module_exit(ext4_exit_fs)
Linux 6.x block layer and io_uring tree(s)