]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
block: remove CONFIG_LBDAF
[thirdparty/kernel/linux.git] / fs / ext4 / super.c
CommitLineData
f5166768 1// SPDX-License-Identifier: GPL-2.0
ac27a0ec 2/*
617ba13b 3 * linux/fs/ext4/super.c
ac27a0ec
DK		 4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/fs/minix/inode.c
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 *
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
18 */
19
20#include <linux/module.h>
21#include <linux/string.h>
22#include <linux/fs.h>
23#include <linux/time.h>
c5ca7c76 24#include <linux/vmalloc.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
66114cad 28#include <linux/backing-dev.h>
ac27a0ec 29#include <linux/parser.h>
ac27a0ec 30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ef510424 41#include <linux/dax.h>
7abc52c2 42#include <linux/cleancache.h>
7c0f6ba6 43#include <linux/uaccess.h>
ee73f9a5 44#include <linux/iversion.h>
ac27a0ec 45
bfff6873
LC		 46#include <linux/kthread.h>
47#include <linux/freezer.h>
48
3dcf5451 49#include "ext4.h"
4a092d73 50#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 51#include "ext4_jbd2.h"
ac27a0ec
DK		 52#include "xattr.h"
53#include "acl.h"
3661d286 54#include "mballoc.h"
0c9ec4be 55#include "fsmap.h"
ac27a0ec 56
9bffad1e
TT		 57#define CREATE_TRACE_POINTS
58#include <trace/events/ext4.h>
59
0b75a840
LC		 60static struct ext4_lazy_init *ext4_li_info;
61static struct mutex ext4_li_mtx;
e294a537 62static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 63
617ba13b 64static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 65 unsigned long journal_devnum);
2adf6da8 66static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 67static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 68static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
617ba13b 72static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 73static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 75static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 76static int ext4_freeze(struct super_block *sb);
152a0836
AV		 77static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
2035e776
TT		 79static inline int ext2_feature_set_ok(struct super_block *sb);
80static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 81static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 82static void ext4_destroy_lazyinit_thread(void);
83static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 84static void ext4_clear_request_list(void);
c6cb7e77
EW		 85static struct inode *ext4_get_journal_inode(struct super_block *sb,
86 unsigned int journal_inum);
ac27a0ec 87
e74031fd
JK		 88/*
89 * Lock ordering
90 *
91 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
92 * i_mmap_rwsem (inode->i_mmap_rwsem)!
93 *
94 * page fault path:
95 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
96 * page lock -> i_data_sem (rw)
97 *
98 * buffered write path:
99 * sb_start_write -> i_mutex -> mmap_sem
100 * sb_start_write -> i_mutex -> transaction start -> page lock ->
101 * i_data_sem (rw)
102 *
103 * truncate:
1d39834f
NB		 104 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
106 * i_data_sem (rw)
e74031fd
JK		 107 *
108 * direct IO:
1d39834f
NB		 109 * sb_start_write -> i_mutex -> mmap_sem
110 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
e74031fd
JK		 111 *
112 * writepages:
113 * transaction start -> page lock(s) -> i_data_sem (rw)
114 */
115
c290ea01 116#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 117static struct file_system_type ext2_fs_type = {
118 .owner = THIS_MODULE,
119 .name = "ext2",
120 .mount = ext4_mount,
121 .kill_sb = kill_block_super,
122 .fs_flags = FS_REQUIRES_DEV,
123};
7f78e035 124MODULE_ALIAS_FS("ext2");
fa7614dd 125MODULE_ALIAS("ext2");
2035e776
TT		 126#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
127#else
128#define IS_EXT2_SB(sb) (0)
129#endif
130
131
ba69f9ab
JK		 132static struct file_system_type ext3_fs_type = {
133 .owner = THIS_MODULE,
134 .name = "ext3",
152a0836 135 .mount = ext4_mount,
ba69f9ab
JK		 136 .kill_sb = kill_block_super,
137 .fs_flags = FS_REQUIRES_DEV,
138};
7f78e035 139MODULE_ALIAS_FS("ext3");
fa7614dd 140MODULE_ALIAS("ext3");
ba69f9ab 141#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 142
fb265c9c
TT		 143/*
144 * This works like sb_bread() except it uses ERR_PTR for error
145 * returns. Currently with sb_bread it's impossible to distinguish
146 * between ENOMEM and EIO situations (since both result in a NULL
147 * return.
148 */
149struct buffer_head *
150ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags)
151{
152 struct buffer_head *bh = sb_getblk(sb, block);
153
154 if (bh == NULL)
155 return ERR_PTR(-ENOMEM);
156 if (buffer_uptodate(bh))
157 return bh;
158 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh);
159 wait_on_buffer(bh);
160 if (buffer_uptodate(bh))
161 return bh;
162 put_bh(bh);
163 return ERR_PTR(-EIO);
164}
165
d25425f8
DW		 166static int ext4_verify_csum_type(struct super_block *sb,
167 struct ext4_super_block *es)
168{
e2b911c5 169 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 170 return 1;
171
172 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
173}
174
a9c47317
DW		 175static __le32 ext4_superblock_csum(struct super_block *sb,
176 struct ext4_super_block *es)
177{
178 struct ext4_sb_info *sbi = EXT4_SB(sb);
179 int offset = offsetof(struct ext4_super_block, s_checksum);
180 __u32 csum;
181
182 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
183
184 return cpu_to_le32(csum);
185}
186
c197855e
SH		 187static int ext4_superblock_csum_verify(struct super_block *sb,
188 struct ext4_super_block *es)
a9c47317 189{
9aa5d32b 190 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 191 return 1;
192
193 return es->s_checksum == ext4_superblock_csum(sb, es);
194}
195
06db49e6 196void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 197{
06db49e6
TT		 198 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
199
9aa5d32b 200 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 201 return;
202
203 es->s_checksum = ext4_superblock_csum(sb, es);
204}
205
9933fc0a
TT		 206void *ext4_kvmalloc(size_t size, gfp_t flags)
207{
208 void *ret;
209
8be04b93 210 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 211 if (!ret)
212 ret = __vmalloc(size, flags, PAGE_KERNEL);
213 return ret;
214}
215
216void *ext4_kvzalloc(size_t size, gfp_t flags)
217{
218 void *ret;
219
8be04b93 220 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 221 if (!ret)
222 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
223 return ret;
224}
225
8fadc143
AR		 226ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
227 struct ext4_group_desc *bg)
bd81d8ee 228{
3a14589c 229 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 231 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 232}
233
8fadc143
AR		 234ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
235 struct ext4_group_desc *bg)
bd81d8ee 236{
5272f837 237 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 239 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 240}
241
8fadc143
AR		 242ext4_fsblk_t ext4_inode_table(struct super_block *sb,
243 struct ext4_group_desc *bg)
bd81d8ee 244{
5272f837 245 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 246 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 247 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 248}
249
021b65bb
TT		 250__u32 ext4_free_group_clusters(struct super_block *sb,
251 struct ext4_group_desc *bg)
560671a0
AK		 252{
253 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
254 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 255 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 256}
257
258__u32 ext4_free_inodes_count(struct super_block *sb,
259 struct ext4_group_desc *bg)
260{
261 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
262 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 263 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 264}
265
266__u32 ext4_used_dirs_count(struct super_block *sb,
267 struct ext4_group_desc *bg)
268{
269 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
270 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 271 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 272}
273
274__u32 ext4_itable_unused_count(struct super_block *sb,
275 struct ext4_group_desc *bg)
276{
277 return le16_to_cpu(bg->bg_itable_unused_lo) |
278 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 279 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 280}
281
8fadc143
AR		 282void ext4_block_bitmap_set(struct super_block *sb,
283 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 284{
3a14589c 285 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 288}
289
8fadc143
AR		 290void ext4_inode_bitmap_set(struct super_block *sb,
291 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 292{
5272f837 293 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 296}
297
8fadc143
AR		 298void ext4_inode_table_set(struct super_block *sb,
299 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 300{
5272f837 301 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 302 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
303 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 304}
305
021b65bb
TT		 306void ext4_free_group_clusters_set(struct super_block *sb,
307 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 308{
309 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
310 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
311 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
312}
313
314void ext4_free_inodes_set(struct super_block *sb,
315 struct ext4_group_desc *bg, __u32 count)
316{
317 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
318 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
319 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
320}
321
322void ext4_used_dirs_set(struct super_block *sb,
323 struct ext4_group_desc *bg, __u32 count)
324{
325 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
326 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
327 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
328}
329
330void ext4_itable_unused_set(struct super_block *sb,
331 struct ext4_group_desc *bg, __u32 count)
332{
333 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
334 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
335 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
336}
337
6a0678a7
AB		 338static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
339{
340 time64_t now = ktime_get_real_seconds();
341
342 now = clamp_val(now, 0, (1ull << 40) - 1);
343
344 *lo = cpu_to_le32(lower_32_bits(now));
345 *hi = upper_32_bits(now);
346}
347
348static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
349{
350 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
351}
352#define ext4_update_tstamp(es, tstamp) \
353 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
354#define ext4_get_tstamp(es, tstamp) \
355 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
d3d1faf6 356
1c13d5c0
TT		 357static void __save_error_info(struct super_block *sb, const char *func,
358 unsigned int line)
359{
360 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
361
362 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 363 if (bdev_read_only(sb->s_bdev))
364 return;
1c13d5c0 365 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6a0678a7 366 ext4_update_tstamp(es, s_last_error_time);
1c13d5c0
TT		 367 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
368 es->s_last_error_line = cpu_to_le32(line);
369 if (!es->s_first_error_time) {
370 es->s_first_error_time = es->s_last_error_time;
6a0678a7 371 es->s_first_error_time_hi = es->s_last_error_time_hi;
1c13d5c0
TT		 372 strncpy(es->s_first_error_func, func,
373 sizeof(es->s_first_error_func));
374 es->s_first_error_line = cpu_to_le32(line);
375 es->s_first_error_ino = es->s_last_error_ino;
376 es->s_first_error_block = es->s_last_error_block;
377 }
66e61a9e
TT		 378 /*
379 * Start the daily error reporting function if it hasn't been
380 * started already
381 */
382 if (!es->s_error_count)
383 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 384 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 385}
386
387static void save_error_info(struct super_block *sb, const char *func,
388 unsigned int line)
389{
390 __save_error_info(sb, func, line);
391 ext4_commit_super(sb, 1);
392}
393
bdfe0cbd
TT		 394/*
395 * The del_gendisk() function uninitializes the disk-specific data
396 * structures, including the bdi structure, without telling anyone
397 * else. Once this happens, any attempt to call mark_buffer_dirty()
398 * (for example, by ext4_commit_super), will cause a kernel OOPS.
399 * This is a kludge to prevent these oops until we can put in a proper
400 * hook in del_gendisk() to inform the VFS and file system layers.
401 */
402static int block_device_ejected(struct super_block *sb)
403{
404 struct inode *bd_inode = sb->s_bdev->bd_inode;
405 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
406
407 return bdi->dev == NULL;
408}
409
18aadd47
BJ		 410static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
411{
412 struct super_block *sb = journal->j_private;
413 struct ext4_sb_info *sbi = EXT4_SB(sb);
414 int error = is_journal_aborted(journal);
5d3ee208 415 struct ext4_journal_cb_entry *jce;
18aadd47 416
5d3ee208 417 BUG_ON(txn->t_state == T_FINISHED);
a0154344
DJ		 418
419 ext4_process_freed_data(sb, txn->t_tid);
420
18aadd47 421 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 422 while (!list_empty(&txn->t_private_list)) {
423 jce = list_entry(txn->t_private_list.next,
424 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 425 list_del_init(&jce->jce_list);
426 spin_unlock(&sbi->s_md_lock);
427 jce->jce_func(sb, jce, error);
428 spin_lock(&sbi->s_md_lock);
429 }
430 spin_unlock(&sbi->s_md_lock);
431}
1c13d5c0 432
1dc1097f
JK		 433static bool system_going_down(void)
434{
435 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
436 || system_state == SYSTEM_RESTART;
437}
438
ac27a0ec
DK		 439/* Deal with the reporting of failure conditions on a filesystem such as
440 * inconsistencies detected or read IO failures.
441 *
442 * On ext2, we can store the error state of the filesystem in the
617ba13b 443 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 444 * write ordering constraints on the superblock which prevent us from
445 * writing it out straight away; and given that the journal is about to
446 * be aborted, we can't rely on the current, or future, transactions to
447 * write out the superblock safely.
448 *
dab291af 449 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 450 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 451 * that error until we've noted it down and cleared it.
452 */
453
617ba13b 454static void ext4_handle_error(struct super_block *sb)
ac27a0ec 455{
327eaf73
TT		 456 if (test_opt(sb, WARN_ON_ERROR))
457 WARN_ON_ONCE(1);
458
bc98a42c 459 if (sb_rdonly(sb))
ac27a0ec
DK		 460 return;
461
2b2d6d01 462 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 463 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 464
4ab2f15b 465 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 466 if (journal)
dab291af 467 jbd2_journal_abort(journal, -EIO);
ac27a0ec 468 }
1dc1097f
JK		 469 /*
470 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
471 * could panic during 'reboot -f' as the underlying device got already
472 * disabled.
473 */
474 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
b31e1552 475 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 476 /*
477 * Make sure updated value of ->s_mount_flags will be visible
478 * before ->s_flags update
479 */
480 smp_wmb();
1751e8a6 481 sb->s_flags |= SB_RDONLY;
1dc1097f 482 } else if (test_opt(sb, ERRORS_PANIC)) {
4327ba52
DJ		 483 if (EXT4_SB(sb)->s_journal &&
484 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
485 return;
617ba13b 486 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 487 sb->s_id);
4327ba52 488 }
ac27a0ec
DK		 489}
490
efbed4dc
TT		 491#define ext4_error_ratelimit(sb) \
492 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
493 "EXT4-fs error")
494
12062ddd 495void __ext4_error(struct super_block *sb, const char *function,
c398eda0 496 unsigned int line, const char *fmt, ...)
ac27a0ec 497{
0ff2ea7d 498 struct va_format vaf;
ac27a0ec
DK		 499 va_list args;
500
0db1ff22
TT		 501 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
502 return;
503
ccf0f32a 504 trace_ext4_error(sb, function, line);
efbed4dc
TT		 505 if (ext4_error_ratelimit(sb)) {
506 va_start(args, fmt);
507 vaf.fmt = fmt;
508 vaf.va = &args;
509 printk(KERN_CRIT
510 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
511 sb->s_id, function, line, current->comm, &vaf);
512 va_end(args);
513 }
f3fc0210 514 save_error_info(sb, function, line);
617ba13b 515 ext4_handle_error(sb);
ac27a0ec
DK		 516}
517
e7c96e8e
JP		 518void __ext4_error_inode(struct inode *inode, const char *function,
519 unsigned int line, ext4_fsblk_t block,
520 const char *fmt, ...)
273df556
FM		 521{
522 va_list args;
f7c21177 523 struct va_format vaf;
1c13d5c0 524 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 525
0db1ff22
TT		 526 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
527 return;
528
ccf0f32a 529 trace_ext4_error(inode->i_sb, function, line);
1c13d5c0
TT		 530 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
531 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 532 if (ext4_error_ratelimit(inode->i_sb)) {
533 va_start(args, fmt);
534 vaf.fmt = fmt;
535 vaf.va = &args;
536 if (block)
537 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
538 "inode #%lu: block %llu: comm %s: %pV\n",
539 inode->i_sb->s_id, function, line, inode->i_ino,
540 block, current->comm, &vaf);
541 else
542 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
543 "inode #%lu: comm %s: %pV\n",
544 inode->i_sb->s_id, function, line, inode->i_ino,
545 current->comm, &vaf);
546 va_end(args);
547 }
1c13d5c0 548 save_error_info(inode->i_sb, function, line);
273df556
FM		 549 ext4_handle_error(inode->i_sb);
550}
551
e7c96e8e
JP		 552void __ext4_error_file(struct file *file, const char *function,
553 unsigned int line, ext4_fsblk_t block,
554 const char *fmt, ...)
273df556
FM		 555{
556 va_list args;
f7c21177 557 struct va_format vaf;
1c13d5c0 558 struct ext4_super_block *es;
496ad9aa 559 struct inode *inode = file_inode(file);
273df556
FM		 560 char pathname[80], *path;
561
0db1ff22
TT		 562 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
563 return;
564
ccf0f32a 565 trace_ext4_error(inode->i_sb, function, line);
1c13d5c0
TT		 566 es = EXT4_SB(inode->i_sb)->s_es;
567 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 568 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 569 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 570 if (IS_ERR(path))
571 path = "(unknown)";
572 va_start(args, fmt);
573 vaf.fmt = fmt;
574 vaf.va = &args;
575 if (block)
576 printk(KERN_CRIT
577 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
578 "block %llu: comm %s: path %s: %pV\n",
579 inode->i_sb->s_id, function, line, inode->i_ino,
580 block, current->comm, path, &vaf);
581 else
582 printk(KERN_CRIT
583 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
584 "comm %s: path %s: %pV\n",
585 inode->i_sb->s_id, function, line, inode->i_ino,
586 current->comm, path, &vaf);
587 va_end(args);
588 }
1c13d5c0 589 save_error_info(inode->i_sb, function, line);
273df556
FM		 590 ext4_handle_error(inode->i_sb);
591}
592
722887dd
TT		 593const char *ext4_decode_error(struct super_block *sb, int errno,
594 char nbuf[16])
ac27a0ec
DK		 595{
596 char *errstr = NULL;
597
598 switch (errno) {
6a797d27
DW		 599 case -EFSCORRUPTED:
600 errstr = "Corrupt filesystem";
601 break;
602 case -EFSBADCRC:
603 errstr = "Filesystem failed CRC";
604 break;
ac27a0ec
DK		 605 case -EIO:
606 errstr = "IO failure";
607 break;
608 case -ENOMEM:
609 errstr = "Out of memory";
610 break;
611 case -EROFS:
78f1ddbb
TT		 612 if (!sb || (EXT4_SB(sb)->s_journal &&
613 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 614 errstr = "Journal has aborted";
615 else
616 errstr = "Readonly filesystem";
617 break;
618 default:
619 /* If the caller passed in an extra buffer for unknown
620 * errors, textualise them now. Else we just return
621 * NULL. */
622 if (nbuf) {
623 /* Check for truncated error codes... */
624 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
625 errstr = nbuf;
626 }
627 break;
628 }
629
630 return errstr;
631}
632
617ba13b 633/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 634 * automatically and invokes the appropriate error response. */
635
c398eda0
TT		 636void __ext4_std_error(struct super_block *sb, const char *function,
637 unsigned int line, int errno)
ac27a0ec
DK		 638{
639 char nbuf[16];
640 const char *errstr;
641
0db1ff22
TT		 642 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
643 return;
644
ac27a0ec
DK		 645 /* Special case: if the error is EROFS, and we're not already
646 * inside a transaction, then there's really no point in logging
647 * an error. */
bc98a42c 648 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
ac27a0ec
DK		 649 return;
650
efbed4dc
TT		 651 if (ext4_error_ratelimit(sb)) {
652 errstr = ext4_decode_error(sb, errno, nbuf);
653 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
654 sb->s_id, function, line, errstr);
655 }
ac27a0ec 656
efbed4dc 657 save_error_info(sb, function, line);
617ba13b 658 ext4_handle_error(sb);
ac27a0ec
DK		 659}
660
661/*
617ba13b 662 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 663 * abort function may be used to deal with unrecoverable failures such
664 * as journal IO errors or ENOMEM at a critical moment in log management.
665 *
666 * We unconditionally force the filesystem into an ABORT|READONLY state,
667 * unless the error response on the fs has been set to panic in which
668 * case we take the easy way out and panic immediately.
669 */
670
c67d859e 671void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 672 unsigned int line, const char *fmt, ...)
ac27a0ec 673{
651e1c3b 674 struct va_format vaf;
ac27a0ec
DK		 675 va_list args;
676
0db1ff22
TT		 677 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
678 return;
679
1c13d5c0 680 save_error_info(sb, function, line);
ac27a0ec 681 va_start(args, fmt);
651e1c3b
JP		 682 vaf.fmt = fmt;
683 vaf.va = &args;
684 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
685 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 686 va_end(args);
687
bc98a42c 688 if (sb_rdonly(sb) == 0) {
1c13d5c0 689 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 690 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 691 /*
692 * Make sure updated value of ->s_mount_flags will be visible
693 * before ->s_flags update
694 */
695 smp_wmb();
1751e8a6 696 sb->s_flags |= SB_RDONLY;
1c13d5c0
TT		 697 if (EXT4_SB(sb)->s_journal)
698 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
699 save_error_info(sb, function, line);
700 }
4327ba52
DJ		 701 if (test_opt(sb, ERRORS_PANIC)) {
702 if (EXT4_SB(sb)->s_journal &&
703 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
704 return;
617ba13b 705 panic("EXT4-fs panic from previous error\n");
4327ba52 706 }
ac27a0ec
DK		 707}
708
e7c96e8e
JP		 709void __ext4_msg(struct super_block *sb,
710 const char *prefix, const char *fmt, ...)
b31e1552 711{
0ff2ea7d 712 struct va_format vaf;
b31e1552
ES		 713 va_list args;
714
efbed4dc
TT		 715 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
716 return;
717
b31e1552 718 va_start(args, fmt);
0ff2ea7d
JP		 719 vaf.fmt = fmt;
720 vaf.va = &args;
721 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 722 va_end(args);
723}
724
b03a2f7e
AD		 725#define ext4_warning_ratelimit(sb) \
726 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
727 "EXT4-fs warning")
728
12062ddd 729void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 730 unsigned int line, const char *fmt, ...)
ac27a0ec 731{
0ff2ea7d 732 struct va_format vaf;
ac27a0ec
DK		 733 va_list args;
734
b03a2f7e 735 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 736 return;
737
ac27a0ec 738 va_start(args, fmt);
0ff2ea7d
JP		 739 vaf.fmt = fmt;
740 vaf.va = &args;
741 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
742 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 743 va_end(args);
744}
745
b03a2f7e
AD		 746void __ext4_warning_inode(const struct inode *inode, const char *function,
747 unsigned int line, const char *fmt, ...)
748{
749 struct va_format vaf;
750 va_list args;
751
752 if (!ext4_warning_ratelimit(inode->i_sb))
753 return;
754
755 va_start(args, fmt);
756 vaf.fmt = fmt;
757 vaf.va = &args;
758 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
759 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
760 function, line, inode->i_ino, current->comm, &vaf);
761 va_end(args);
762}
763
e29136f8
TT		 764void __ext4_grp_locked_error(const char *function, unsigned int line,
765 struct super_block *sb, ext4_group_t grp,
766 unsigned long ino, ext4_fsblk_t block,
767 const char *fmt, ...)
5d1b1b3f
AK		 768__releases(bitlock)
769__acquires(bitlock)
770{
0ff2ea7d 771 struct va_format vaf;
5d1b1b3f
AK		 772 va_list args;
773 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
774
0db1ff22
TT		 775 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
776 return;
777
ccf0f32a 778 trace_ext4_error(sb, function, line);
1c13d5c0
TT		 779 es->s_last_error_ino = cpu_to_le32(ino);
780 es->s_last_error_block = cpu_to_le64(block);
781 __save_error_info(sb, function, line);
0ff2ea7d 782
efbed4dc
TT		 783 if (ext4_error_ratelimit(sb)) {
784 va_start(args, fmt);
785 vaf.fmt = fmt;
786 vaf.va = &args;
787 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
788 sb->s_id, function, line, grp);
789 if (ino)
790 printk(KERN_CONT "inode %lu: ", ino);
791 if (block)
792 printk(KERN_CONT "block %llu:",
793 (unsigned long long) block);
794 printk(KERN_CONT "%pV\n", &vaf);
795 va_end(args);
796 }
5d1b1b3f 797
327eaf73
TT		 798 if (test_opt(sb, WARN_ON_ERROR))
799 WARN_ON_ONCE(1);
800
5d1b1b3f 801 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 802 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 803 return;
804 }
1c13d5c0 805
5d1b1b3f 806 ext4_unlock_group(sb, grp);
06f29cc8 807 ext4_commit_super(sb, 1);
5d1b1b3f
AK		 808 ext4_handle_error(sb);
809 /*
810 * We only get here in the ERRORS_RO case; relocking the group
811 * may be dangerous, but nothing bad will happen since the
812 * filesystem will have already been marked read/only and the
813 * journal has been aborted. We return 1 as a hint to callers
814 * who might what to use the return value from
25985edc 815 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 816 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
817 * aggressively from the ext4 function in question, with a
818 * more appropriate error code.
819 */
820 ext4_lock_group(sb, grp);
821 return;
822}
823
db79e6d1
WS		 824void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
825 ext4_group_t group,
826 unsigned int flags)
827{
828 struct ext4_sb_info *sbi = EXT4_SB(sb);
829 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
830 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
9af0b3d1
WS		 831 int ret;
832
833 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
834 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
835 &grp->bb_state);
836 if (!ret)
837 percpu_counter_sub(&sbi->s_freeclusters_counter,
838 grp->bb_free);
db79e6d1
WS		 839 }
840
9af0b3d1
WS		 841 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
842 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
843 &grp->bb_state);
844 if (!ret && gdp) {
db79e6d1
WS		 845 int count;
846
847 count = ext4_free_inodes_count(sb, gdp);
848 percpu_counter_sub(&sbi->s_freeinodes_counter,
849 count);
850 }
db79e6d1
WS		 851 }
852}
853
617ba13b 854void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 855{
617ba13b 856 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 857
617ba13b 858 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 859 return;
860
12062ddd 861 ext4_warning(sb,
ac27a0ec
DK		 862 "updating to rev %d because of new feature flag, "
863 "running e2fsck is recommended",
617ba13b 864 EXT4_DYNAMIC_REV);
ac27a0ec 865
617ba13b
MC		 866 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
867 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
868 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 869 /* leave es->s_feature_*compat flags alone */
870 /* es->s_uuid will be set by e2fsck if empty */
871
872 /*
873 * The rest of the superblock fields should be zero, and if not it
874 * means they are likely already in use, so leave them alone. We
875 * can leave it up to e2fsck to clean up any inconsistencies there.
876 */
877}
878
879/*
880 * Open the external journal device
881 */
b31e1552 882static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 883{
884 struct block_device *bdev;
885 char b[BDEVNAME_SIZE];
886
d4d77629 887 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 888 if (IS_ERR(bdev))
889 goto fail;
890 return bdev;
891
892fail:
b31e1552 893 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 894 __bdevname(dev, b), PTR_ERR(bdev));
895 return NULL;
896}
897
898/*
899 * Release the journal device
900 */
4385bab1 901static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 902{
4385bab1 903 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 904}
905
4385bab1 906static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 907{
908 struct block_device *bdev;
ac27a0ec
DK		 909 bdev = sbi->journal_bdev;
910 if (bdev) {
4385bab1 911 ext4_blkdev_put(bdev);
ac27a0ec
DK		 912 sbi->journal_bdev = NULL;
913 }
ac27a0ec
DK		 914}
915
916static inline struct inode *orphan_list_entry(struct list_head *l)
917{
617ba13b 918 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 919}
920
617ba13b 921static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 922{
923 struct list_head *l;
924
b31e1552
ES		 925 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
926 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 927
928 printk(KERN_ERR "sb_info orphan list:\n");
929 list_for_each(l, &sbi->s_orphan) {
930 struct inode *inode = orphan_list_entry(l);
931 printk(KERN_ERR " "
932 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
933 inode->i_sb->s_id, inode->i_ino, inode,
934 inode->i_mode, inode->i_nlink,
935 NEXT_ORPHAN(inode));
936 }
937}
938
957153fc
JK		 939#ifdef CONFIG_QUOTA
940static int ext4_quota_off(struct super_block *sb, int type);
941
942static inline void ext4_quota_off_umount(struct super_block *sb)
943{
944 int type;
945
964edf66
JK		 946 /* Use our quota_off function to clear inode flags etc. */
947 for (type = 0; type < EXT4_MAXQUOTAS; type++)
948 ext4_quota_off(sb, type);
957153fc 949}
33458eab
TT		 950
951/*
952 * This is a helper function which is used in the mount/remount
953 * codepaths (which holds s_umount) to fetch the quota file name.
954 */
955static inline char *get_qf_name(struct super_block *sb,
956 struct ext4_sb_info *sbi,
957 int type)
958{
959 return rcu_dereference_protected(sbi->s_qf_names[type],
960 lockdep_is_held(&sb->s_umount));
961}
957153fc
JK		 962#else
963static inline void ext4_quota_off_umount(struct super_block *sb)
964{
965}
966#endif
967
2b2d6d01 968static void ext4_put_super(struct super_block *sb)
ac27a0ec 969{
617ba13b
MC		 970 struct ext4_sb_info *sbi = EXT4_SB(sb);
971 struct ext4_super_block *es = sbi->s_es;
97abd7d4 972 int aborted = 0;
ef2cabf7 973 int i, err;
ac27a0ec 974
857ac889 975 ext4_unregister_li_request(sb);
957153fc 976 ext4_quota_off_umount(sb);
e0ccfd95 977
2e8fa54e 978 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 979
0390131b 980 if (sbi->s_journal) {
97abd7d4 981 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 982 err = jbd2_journal_destroy(sbi->s_journal);
983 sbi->s_journal = NULL;
97abd7d4 984 if ((err < 0) && !aborted)
c67d859e 985 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 986 }
d4edac31 987
ebd173be 988 ext4_unregister_sysfs(sb);
d3922a77 989 ext4_es_unregister_shrinker(sbi);
9105bb14 990 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 991 ext4_release_system_zone(sb);
992 ext4_mb_release(sb);
993 ext4_ext_release(sb);
d4edac31 994
bc98a42c 995 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 996 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 997 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 998 }
bc98a42c 999 if (!sb_rdonly(sb))
a8e25a83
AB		 1000 ext4_commit_super(sb, 1);
1001
ac27a0ec
DK		 1002 for (i = 0; i < sbi->s_gdb_count; i++)
1003 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 1004 kvfree(sbi->s_group_desc);
1005 kvfree(sbi->s_flex_groups);
57042651 1006 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 1007 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1008 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 1009 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c8585c6f 1010 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec 1011#ifdef CONFIG_QUOTA
a2d4a646 1012 for (i = 0; i < EXT4_MAXQUOTAS; i++)
33458eab 1013 kfree(get_qf_name(sb, sbi, i));
ac27a0ec
DK		 1014#endif
1015
1016 /* Debugging code just in case the in-memory inode orphan list
1017 * isn't empty. The on-disk one can be non-empty if we've
1018 * detected an error and taken the fs readonly, but the
1019 * in-memory list had better be clean by this point. */
1020 if (!list_empty(&sbi->s_orphan))
1021 dump_orphan_list(sb, sbi);
1022 J_ASSERT(list_empty(&sbi->s_orphan));
1023
89d96a6f 1024 sync_blockdev(sb->s_bdev);
f98393a6 1025 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 1026 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
1027 /*
1028 * Invalidate the journal device's buffers. We don't want them
1029 * floating about in memory - the physical journal device may
1030 * hotswapped, and it breaks the `ro-after' testing code.
1031 */
1032 sync_blockdev(sbi->journal_bdev);
f98393a6 1033 invalidate_bdev(sbi->journal_bdev);
617ba13b 1034 ext4_blkdev_remove(sbi);
ac27a0ec 1035 }
50c15df6
CX		 1036
1037 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1038 sbi->s_ea_inode_cache = NULL;
1039
1040 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1041 sbi->s_ea_block_cache = NULL;
1042
c5e06d10
JL		 1043 if (sbi->s_mmp_tsk)
1044 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 1045 brelse(sbi->s_sbh);
ac27a0ec 1046 sb->s_fs_info = NULL;
3197ebdb
TT		 1047 /*
1048 * Now that we are completely done shutting down the
1049 * superblock, we need to actually destroy the kobject.
1050 */
3197ebdb
TT		 1051 kobject_put(&sbi->s_kobj);
1052 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 1053 if (sbi->s_chksum_driver)
1054 crypto_free_shash(sbi->s_chksum_driver);
705895b6 1055 kfree(sbi->s_blockgroup_lock);
5e405595 1056 fs_put_dax(sbi->s_daxdev);
ac27a0ec 1057 kfree(sbi);
ac27a0ec
DK		 1058}
1059
e18b890b 1060static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 1061
1062/*
1063 * Called inside transaction, so use GFP_NOFS
1064 */
617ba13b 1065static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 1066{
617ba13b 1067 struct ext4_inode_info *ei;
ac27a0ec 1068
e6b4f8da 1069 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 1070 if (!ei)
1071 return NULL;
0b8e58a1 1072
ee73f9a5 1073 inode_set_iversion(&ei->vfs_inode, 1);
202ee5df 1074 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 1075 INIT_LIST_HEAD(&ei->i_prealloc_list);
1076 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 1077 ext4_es_init_tree(&ei->i_es_tree);
1078 rwlock_init(&ei->i_es_lock);
edaa53ca 1079 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 1080 ei->i_es_all_nr = 0;
edaa53ca 1081 ei->i_es_shk_nr = 0;
dd475925 1082 ei->i_es_shrink_lblk = 0;
d2a17637 1083 ei->i_reserved_data_blocks = 0;
9d0be502 1084 ei->i_da_metadata_calc_len = 0;
7e731bc9 1085 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 1086 spin_lock_init(&(ei->i_block_reservation_lock));
1dc0aa46 1087 ext4_init_pending_tree(&ei->i_pending_tree);
a9e7f447
DM		 1088#ifdef CONFIG_QUOTA
1089 ei->i_reserved_quota = 0;
96c7e0d9 1090 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 1091#endif
8aefcd55 1092 ei->jinode = NULL;
2e8fa54e 1093 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 1094 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 1095 ei->i_sync_tid = 0;
1096 ei->i_datasync_tid = 0;
e27f41e1 1097 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 1098 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 1099 return &ei->vfs_inode;
1100}
1101
7ff9c073
TT		 1102static int ext4_drop_inode(struct inode *inode)
1103{
1104 int drop = generic_drop_inode(inode);
1105
1106 trace_ext4_drop_inode(inode, drop);
1107 return drop;
1108}
1109
fa0d7e3d
NP		 1110static void ext4_i_callback(struct rcu_head *head)
1111{
1112 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 1113 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1114}
1115
617ba13b 1116static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1117{
9f7dd93d 1118 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1119 ext4_msg(inode->i_sb, KERN_ERR,
1120 "Inode %lu (%p): orphan list check failed!",
1121 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1122 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1123 EXT4_I(inode), sizeof(struct ext4_inode_info),
1124 true);
1125 dump_stack();
1126 }
fa0d7e3d 1127 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1128}
1129
51cc5068 1130static void init_once(void *foo)
ac27a0ec 1131{
617ba13b 1132 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1133
a35afb83 1134 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1135 init_rwsem(&ei->xattr_sem);
0e855ac8 1136 init_rwsem(&ei->i_data_sem);
ea3d7209 1137 init_rwsem(&ei->i_mmap_sem);
a35afb83 1138 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1139}
1140
e67bc2b3 1141static int __init init_inodecache(void)
ac27a0ec 1142{
f8dd7c70
DW		 1143 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1144 sizeof(struct ext4_inode_info), 0,
1145 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1146 SLAB_ACCOUNT),
1147 offsetof(struct ext4_inode_info, i_data),
1148 sizeof_field(struct ext4_inode_info, i_data),
1149 init_once);
617ba13b 1150 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1151 return -ENOMEM;
1152 return 0;
1153}
1154
1155static void destroy_inodecache(void)
1156{
8c0a8537
KS		 1157 /*
1158 * Make sure all delayed rcu free inodes are flushed before we
1159 * destroy cache.
1160 */
1161 rcu_barrier();
617ba13b 1162 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1163}
1164
0930fcc1 1165void ext4_clear_inode(struct inode *inode)
ac27a0ec 1166{
0930fcc1 1167 invalidate_inode_buffers(inode);
dbd5768f 1168 clear_inode(inode);
9f754758 1169 dquot_drop(inode);
c2ea3fde 1170 ext4_discard_preallocations(inode);
51865fda 1171 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1172 if (EXT4_I(inode)->jinode) {
1173 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1174 EXT4_I(inode)->jinode);
1175 jbd2_free_inode(EXT4_I(inode)->jinode);
1176 EXT4_I(inode)->jinode = NULL;
1177 }
3d204e24 1178 fscrypt_put_encryption_info(inode);
ac27a0ec
DK		 1179}
1180
1b961ac0 1181static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1182 u64 ino, u32 generation)
ac27a0ec 1183{
ac27a0ec 1184 struct inode *inode;
ac27a0ec 1185
8a363970 1186 /*
ac27a0ec
DK		 1187 * Currently we don't know the generation for parent directory, so
1188 * a generation of 0 means "accept any"
1189 */
8a363970 1190 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1d1fe1ee
DH		 1191 if (IS_ERR(inode))
1192 return ERR_CAST(inode);
1193 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1194 iput(inode);
1195 return ERR_PTR(-ESTALE);
1196 }
1b961ac0
CH		 1197
1198 return inode;
1199}
1200
1201static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1202 int fh_len, int fh_type)
1b961ac0
CH		 1203{
1204 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1205 ext4_nfs_get_inode);
1206}
1207
1208static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1209 int fh_len, int fh_type)
1b961ac0
CH		 1210{
1211 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1212 ext4_nfs_get_inode);
ac27a0ec
DK		 1213}
1214
fde87268
TT		 1215static int ext4_nfs_commit_metadata(struct inode *inode)
1216{
1217 struct writeback_control wbc = {
1218 .sync_mode = WB_SYNC_ALL
1219 };
1220
1221 trace_ext4_nfs_commit_metadata(inode);
1222 return ext4_write_inode(inode, &wbc);
1223}
1224
c39a7f84
TO		 1225/*
1226 * Try to release metadata pages (indirect blocks, directories) which are
1227 * mapped via the block device. Since these pages could have journal heads
1228 * which would prevent try_to_free_buffers() from freeing them, we must use
1229 * jbd2 layer's try_to_free_buffers() function to release them.
1230 */
0b8e58a1
AD		 1231static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1232 gfp_t wait)
c39a7f84
TO		 1233{
1234 journal_t *journal = EXT4_SB(sb)->s_journal;
1235
1236 WARN_ON(PageChecked(page));
1237 if (!page_has_buffers(page))
1238 return 0;
1239 if (journal)
1240 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1241 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1242 return try_to_free_buffers(page);
1243}
1244
643fa961 1245#ifdef CONFIG_FS_ENCRYPTION
a7550b30
JK		 1246static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1247{
1248 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1249 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1250}
1251
a7550b30
JK		 1252static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1253 void *fs_data)
1254{
2f8f5e76 1255 handle_t *handle = fs_data;
c1a5d5f6 1256 int res, res2, credits, retries = 0;
2f8f5e76 1257
9ce0151a
EB		 1258 /*
1259 * Encrypting the root directory is not allowed because e2fsck expects
1260 * lost+found to exist and be unencrypted, and encrypting the root
1261 * directory would imply encrypting the lost+found directory as well as
1262 * the filename "lost+found" itself.
1263 */
1264 if (inode->i_ino == EXT4_ROOT_INO)
1265 return -EPERM;
2f8f5e76 1266
7d3e06a8
RZ		 1267 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1268 return -EINVAL;
1269
94840e3c
EB		 1270 res = ext4_convert_inline_data(inode);
1271 if (res)
1272 return res;
1273
2f8f5e76
EB		 1274 /*
1275 * If a journal handle was specified, then the encryption context is
1276 * being set on a new inode via inheritance and is part of a larger
1277 * transaction to create the inode. Otherwise the encryption context is
1278 * being set on an existing inode in its own transaction. Only in the
1279 * latter case should the "retry on ENOSPC" logic be used.
1280 */
a7550b30 1281
2f8f5e76
EB		 1282 if (handle) {
1283 res = ext4_xattr_set_handle(handle, inode,
1284 EXT4_XATTR_INDEX_ENCRYPTION,
1285 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1286 ctx, len, 0);
a7550b30
JK		 1287 if (!res) {
1288 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1289 ext4_clear_inode_state(inode,
1290 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b 1291 /*
2ee6a576
EB		 1292 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1293 * S_DAX may be disabled
a3caa24b
JK		 1294 */
1295 ext4_set_inode_flags(inode);
a7550b30
JK		 1296 }
1297 return res;
1298 }
1299
b8cb5a54
TE		 1300 res = dquot_initialize(inode);
1301 if (res)
1302 return res;
2f8f5e76 1303retry:
af65207c
TE		 1304 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1305 &credits);
dec214d0
TE		 1306 if (res)
1307 return res;
1308
c1a5d5f6 1309 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
a7550b30
JK		 1310 if (IS_ERR(handle))
1311 return PTR_ERR(handle);
1312
2f8f5e76
EB		 1313 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1314 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1315 ctx, len, 0);
a7550b30
JK		 1316 if (!res) {
1317 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
2ee6a576
EB		 1318 /*
1319 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1320 * S_DAX may be disabled
1321 */
a3caa24b 1322 ext4_set_inode_flags(inode);
a7550b30
JK		 1323 res = ext4_mark_inode_dirty(handle, inode);
1324 if (res)
1325 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1326 }
1327 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1328
1329 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1330 goto retry;
a7550b30
JK		 1331 if (!res)
1332 res = res2;
1333 return res;
1334}
1335
c250b7dd 1336static bool ext4_dummy_context(struct inode *inode)
a7550b30
JK		 1337{
1338 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1339}
1340
6f69f0ed 1341static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1342 .key_prefix = "ext4:",
a7550b30 1343 .get_context = ext4_get_context,
a7550b30
JK		 1344 .set_context = ext4_set_context,
1345 .dummy_context = ext4_dummy_context,
a7550b30 1346 .empty_dir = ext4_empty_dir,
e12ee683 1347 .max_namelen = EXT4_NAME_LEN,
a7550b30 1348};
a7550b30
JK		 1349#endif
1350
ac27a0ec 1351#ifdef CONFIG_QUOTA
d6006186 1352static const char * const quotatypes[] = INITQFNAMES;
689c958c 1353#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1354
617ba13b
MC		 1355static int ext4_write_dquot(struct dquot *dquot);
1356static int ext4_acquire_dquot(struct dquot *dquot);
1357static int ext4_release_dquot(struct dquot *dquot);
1358static int ext4_mark_dquot_dirty(struct dquot *dquot);
1359static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1360static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1361 const struct path *path);
617ba13b
MC		 1362static int ext4_quota_on_mount(struct super_block *sb, int type);
1363static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1364 size_t len, loff_t off);
617ba13b 1365static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1366 const char *data, size_t len, loff_t off);
7c319d32
AK		 1367static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1368 unsigned int flags);
1369static int ext4_enable_quotas(struct super_block *sb);
8f0e8746 1370static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1371
96c7e0d9
JK		 1372static struct dquot **ext4_get_dquots(struct inode *inode)
1373{
1374 return EXT4_I(inode)->i_dquot;
1375}
1376
61e225dc 1377static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1378 .get_reserved_space = ext4_get_reserved_space,
1379 .write_dquot = ext4_write_dquot,
1380 .acquire_dquot = ext4_acquire_dquot,
1381 .release_dquot = ext4_release_dquot,
1382 .mark_dirty = ext4_mark_dquot_dirty,
1383 .write_info = ext4_write_info,
1384 .alloc_dquot = dquot_alloc,
1385 .destroy_dquot = dquot_destroy,
1386 .get_projid = ext4_get_projid,
1387 .get_inode_usage = ext4_get_inode_usage,
1388 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1389};
1390
0d54b217 1391static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1392 .quota_on = ext4_quota_on,
ca0e05e4 1393 .quota_off = ext4_quota_off,
287a8095 1394 .quota_sync = dquot_quota_sync,
0a240339 1395 .get_state = dquot_get_state,
287a8095
CH		 1396 .set_info = dquot_set_dqinfo,
1397 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1398 .set_dqblk = dquot_set_dqblk,
1399 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1400};
1401#endif
1402
ee9b6d61 1403static const struct super_operations ext4_sops = {
617ba13b
MC		 1404 .alloc_inode = ext4_alloc_inode,
1405 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1406 .write_inode = ext4_write_inode,
1407 .dirty_inode = ext4_dirty_inode,
7ff9c073 1408 .drop_inode = ext4_drop_inode,
0930fcc1 1409 .evict_inode = ext4_evict_inode,
617ba13b 1410 .put_super = ext4_put_super,
617ba13b 1411 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1412 .freeze_fs = ext4_freeze,
1413 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1414 .statfs = ext4_statfs,
1415 .remount_fs = ext4_remount,
617ba13b 1416 .show_options = ext4_show_options,
ac27a0ec 1417#ifdef CONFIG_QUOTA
617ba13b
MC		 1418 .quota_read = ext4_quota_read,
1419 .quota_write = ext4_quota_write,
96c7e0d9 1420 .get_dquots = ext4_get_dquots,
ac27a0ec 1421#endif
c39a7f84 1422 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1423};
1424
39655164 1425static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1426 .fh_to_dentry = ext4_fh_to_dentry,
1427 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1428 .get_parent = ext4_get_parent,
fde87268 1429 .commit_metadata = ext4_nfs_commit_metadata,
ac27a0ec
DK		 1430};
1431
1432enum {
1433 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1434 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1435 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1436 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1437 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1438 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1439 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1440 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1441 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1442 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1443 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1444 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1445 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
327eaf73
TT		 1446 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1447 Opt_nowarn_on_error, Opt_mblk_io_submit,
670e9875 1448 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1449 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1450 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1451 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1452 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1453 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
ac27a0ec
DK		 1454};
1455
a447c093 1456static const match_table_t tokens = {
ac27a0ec
DK		 1457 {Opt_bsd_df, "bsddf"},
1458 {Opt_minix_df, "minixdf"},
1459 {Opt_grpid, "grpid"},
1460 {Opt_grpid, "bsdgroups"},
1461 {Opt_nogrpid, "nogrpid"},
1462 {Opt_nogrpid, "sysvgroups"},
1463 {Opt_resgid, "resgid=%u"},
1464 {Opt_resuid, "resuid=%u"},
1465 {Opt_sb, "sb=%u"},
1466 {Opt_err_cont, "errors=continue"},
1467 {Opt_err_panic, "errors=panic"},
1468 {Opt_err_ro, "errors=remount-ro"},
1469 {Opt_nouid32, "nouid32"},
ac27a0ec 1470 {Opt_debug, "debug"},
72578c33
TT		 1471 {Opt_removed, "oldalloc"},
1472 {Opt_removed, "orlov"},
ac27a0ec
DK		 1473 {Opt_user_xattr, "user_xattr"},
1474 {Opt_nouser_xattr, "nouser_xattr"},
1475 {Opt_acl, "acl"},
1476 {Opt_noacl, "noacl"},
e3bb52ae 1477 {Opt_noload, "norecovery"},
5a916be1 1478 {Opt_noload, "noload"},
72578c33
TT		 1479 {Opt_removed, "nobh"},
1480 {Opt_removed, "bh"},
ac27a0ec 1481 {Opt_commit, "commit=%u"},
30773840
TT		 1482 {Opt_min_batch_time, "min_batch_time=%u"},
1483 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1484 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1485 {Opt_journal_path, "journal_path=%s"},
818d276c 1486 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1487 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1488 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1489 {Opt_abort, "abort"},
1490 {Opt_data_journal, "data=journal"},
1491 {Opt_data_ordered, "data=ordered"},
1492 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1493 {Opt_data_err_abort, "data_err=abort"},
1494 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1495 {Opt_offusrjquota, "usrjquota="},
1496 {Opt_usrjquota, "usrjquota=%s"},
1497 {Opt_offgrpjquota, "grpjquota="},
1498 {Opt_grpjquota, "grpjquota=%s"},
1499 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1500 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1501 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1502 {Opt_grpquota, "grpquota"},
1503 {Opt_noquota, "noquota"},
1504 {Opt_quota, "quota"},
1505 {Opt_usrquota, "usrquota"},
49da9392 1506 {Opt_prjquota, "prjquota"},
ac27a0ec 1507 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1508 {Opt_barrier, "barrier"},
1509 {Opt_nobarrier, "nobarrier"},
25ec56b5 1510 {Opt_i_version, "i_version"},
923ae0ff 1511 {Opt_dax, "dax"},
c9de560d 1512 {Opt_stripe, "stripe=%u"},
64769240 1513 {Opt_delalloc, "delalloc"},
327eaf73
TT		 1514 {Opt_warn_on_error, "warn_on_error"},
1515 {Opt_nowarn_on_error, "nowarn_on_error"},
a26f4992
TT		 1516 {Opt_lazytime, "lazytime"},
1517 {Opt_nolazytime, "nolazytime"},
670e9875 1518 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1519 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1520 {Opt_removed, "mblk_io_submit"},
1521 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1522 {Opt_block_validity, "block_validity"},
1523 {Opt_noblock_validity, "noblock_validity"},
240799cd 1524 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1525 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1526 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1527 {Opt_auto_da_alloc, "auto_da_alloc"},
1528 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1529 {Opt_dioread_nolock, "dioread_nolock"},
1530 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1531 {Opt_discard, "discard"},
1532 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1533 {Opt_init_itable, "init_itable=%u"},
1534 {Opt_init_itable, "init_itable"},
1535 {Opt_noinit_itable, "noinit_itable"},
df981d03 1536 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1537 {Opt_test_dummy_encryption, "test_dummy_encryption"},
cdb7ee4c
TE		 1538 {Opt_nombcache, "nombcache"},
1539 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
c7198b9c
TT		 1540 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1541 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1542 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1543 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1544 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1545 {Opt_err, NULL},
ac27a0ec
DK		 1546};
1547
617ba13b 1548static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1549{
617ba13b 1550 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1551 char *options = (char *) *data;
1552
1553 if (!options || strncmp(options, "sb=", 3) != 0)
1554 return 1; /* Default location */
0b8e58a1 1555
ac27a0ec 1556 options += 3;
0b8e58a1 1557 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1558 sb_block = simple_strtoul(options, &options, 0);
1559 if (*options && *options != ',') {
4776004f 1560 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1561 (char *) *data);
1562 return 1;
1563 }
1564 if (*options == ',')
1565 options++;
1566 *data = (void *) options;
0b8e58a1 1567
ac27a0ec
DK		 1568 return sb_block;
1569}
1570
b3881f74 1571#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1572static const char deprecated_msg[] =
1573 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1574 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1575
56c50f11
DM		 1576#ifdef CONFIG_QUOTA
1577static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1578{
1579 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 1580 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
03dafb5f 1581 int ret = -1;
56c50f11 1582
33458eab 1583 if (sb_any_quota_loaded(sb) && !old_qname) {
56c50f11
DM		 1584 ext4_msg(sb, KERN_ERR,
1585 "Cannot change journaled "
1586 "quota options when quota turned on");
57f73c2c 1587 return -1;
56c50f11 1588 }
e2b911c5 1589 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1590 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1591 "ignored when QUOTA feature is enabled");
1592 return 1;
262b4662 1593 }
56c50f11
DM		 1594 qname = match_strdup(args);
1595 if (!qname) {
1596 ext4_msg(sb, KERN_ERR,
1597 "Not enough memory for storing quotafile name");
57f73c2c 1598 return -1;
56c50f11 1599 }
33458eab
TT		 1600 if (old_qname) {
1601 if (strcmp(old_qname, qname) == 0)
03dafb5f
CG		 1602 ret = 1;
1603 else
1604 ext4_msg(sb, KERN_ERR,
1605 "%s quota file already specified",
1606 QTYPE2NAME(qtype));
1607 goto errout;
56c50f11 1608 }
03dafb5f 1609 if (strchr(qname, '/')) {
56c50f11
DM		 1610 ext4_msg(sb, KERN_ERR,
1611 "quotafile must be on filesystem root");
03dafb5f 1612 goto errout;
56c50f11 1613 }
33458eab 1614 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
fd8c37ec 1615 set_opt(sb, QUOTA);
56c50f11 1616 return 1;
03dafb5f
CG		 1617errout:
1618 kfree(qname);
1619 return ret;
56c50f11
DM		 1620}
1621
1622static int clear_qf_name(struct super_block *sb, int qtype)
1623{
1624
1625 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 1626 char *old_qname = get_qf_name(sb, sbi, qtype);
56c50f11 1627
33458eab 1628 if (sb_any_quota_loaded(sb) && old_qname) {
56c50f11
DM		 1629 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1630 " when quota turned on");
57f73c2c 1631 return -1;
56c50f11 1632 }
33458eab
TT		 1633 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1634 synchronize_rcu();
1635 kfree(old_qname);
56c50f11
DM		 1636 return 1;
1637}
1638#endif
1639
26092bf5
TT		 1640#define MOPT_SET 0x0001
1641#define MOPT_CLEAR 0x0002
1642#define MOPT_NOSUPPORT 0x0004
1643#define MOPT_EXPLICIT 0x0008
1644#define MOPT_CLEAR_ERR 0x0010
1645#define MOPT_GTE0 0x0020
ac27a0ec 1646#ifdef CONFIG_QUOTA
26092bf5
TT		 1647#define MOPT_Q 0
1648#define MOPT_QFMT 0x0040
1649#else
1650#define MOPT_Q MOPT_NOSUPPORT
1651#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1652#endif
26092bf5 1653#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1654#define MOPT_NO_EXT2 0x0100
1655#define MOPT_NO_EXT3 0x0200
1656#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1657#define MOPT_STRING 0x0400
26092bf5
TT		 1658
1659static const struct mount_opts {
1660 int token;
1661 int mount_opt;
1662 int flags;
1663} ext4_mount_opts[] = {
1664 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1665 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1666 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1667 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1668 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1669 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1670 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1671 MOPT_EXT4_ONLY | MOPT_SET},
1672 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1673 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1674 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1675 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1676 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1677 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1678 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1679 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1680 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1681 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
c6d3d56d
DW		 1682 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1683 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1684 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1685 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1686 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1687 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1688 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1689 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1690 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1691 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1692 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1693 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1694 MOPT_NO_EXT2},
8dc0aa8c 1695 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1696 MOPT_NO_EXT2},
26092bf5
TT		 1697 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1698 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1699 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1700 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1701 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1702 {Opt_commit, 0, MOPT_GTE0},
1703 {Opt_max_batch_time, 0, MOPT_GTE0},
1704 {Opt_min_batch_time, 0, MOPT_GTE0},
1705 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1706 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1707 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1708 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1709 {Opt_resuid, 0, MOPT_GTE0},
1710 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1711 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1712 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1713 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1714 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1715 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1716 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1717 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1718 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1719 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1720#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1721 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1722 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1723#else
26092bf5
TT		 1724 {Opt_acl, 0, MOPT_NOSUPPORT},
1725 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1726#endif
26092bf5
TT		 1727 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1728 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1729 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1730 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1731 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1732 MOPT_SET | MOPT_Q},
1733 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1734 MOPT_SET | MOPT_Q},
49da9392
JK		 1735 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1736 MOPT_SET | MOPT_Q},
26092bf5 1737 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1738 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1739 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1740 {Opt_usrjquota, 0, MOPT_Q},
1741 {Opt_grpjquota, 0, MOPT_Q},
1742 {Opt_offusrjquota, 0, MOPT_Q},
1743 {Opt_offgrpjquota, 0, MOPT_Q},
1744 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1745 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1746 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1747 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1748 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
cdb7ee4c 1749 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
26092bf5
TT		 1750 {Opt_err, 0, 0}
1751};
1752
1753static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1754 substring_t *args, unsigned long *journal_devnum,
1755 unsigned int *journal_ioprio, int is_remount)
1756{
1757 struct ext4_sb_info *sbi = EXT4_SB(sb);
1758 const struct mount_opts *m;
08cefc7a
EB		 1759 kuid_t uid;
1760 kgid_t gid;
26092bf5
TT		 1761 int arg = 0;
1762
57f73c2c
TT		 1763#ifdef CONFIG_QUOTA
1764 if (token == Opt_usrjquota)
1765 return set_qf_name(sb, USRQUOTA, &args[0]);
1766 else if (token == Opt_grpjquota)
1767 return set_qf_name(sb, GRPQUOTA, &args[0]);
1768 else if (token == Opt_offusrjquota)
1769 return clear_qf_name(sb, USRQUOTA);
1770 else if (token == Opt_offgrpjquota)
1771 return clear_qf_name(sb, GRPQUOTA);
1772#endif
26092bf5 1773 switch (token) {
f7048605
TT		 1774 case Opt_noacl:
1775 case Opt_nouser_xattr:
1776 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1777 break;
26092bf5
TT		 1778 case Opt_sb:
1779 return 1; /* handled by get_sb_block() */
1780 case Opt_removed:
5f3633e3 1781 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1782 return 1;
26092bf5
TT		 1783 case Opt_abort:
1784 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1785 return 1;
1786 case Opt_i_version:
357fdad0 1787 sb->s_flags |= SB_I_VERSION;
26092bf5 1788 return 1;
a26f4992 1789 case Opt_lazytime:
1751e8a6 1790 sb->s_flags |= SB_LAZYTIME;
a26f4992
TT		 1791 return 1;
1792 case Opt_nolazytime:
1751e8a6 1793 sb->s_flags &= ~SB_LAZYTIME;
a26f4992 1794 return 1;
26092bf5
TT		 1795 }
1796
5f3633e3
JK		 1797 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1798 if (token == m->token)
1799 break;
1800
1801 if (m->token == Opt_err) {
1802 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1803 "or missing value", opt);
1804 return -1;
1805 }
1806
8dc0aa8c
TT		 1807 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1808 ext4_msg(sb, KERN_ERR,
1809 "Mount option \"%s\" incompatible with ext2", opt);
1810 return -1;
1811 }
1812 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1813 ext4_msg(sb, KERN_ERR,
1814 "Mount option \"%s\" incompatible with ext3", opt);
1815 return -1;
1816 }
1817
ad4eec61 1818 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1819 return -1;
1820 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1821 return -1;
c93cf2d7
DM		 1822 if (m->flags & MOPT_EXPLICIT) {
1823 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1824 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1825 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1826 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1827 } else
1828 return -1;
1829 }
5f3633e3
JK		 1830 if (m->flags & MOPT_CLEAR_ERR)
1831 clear_opt(sb, ERRORS_MASK);
1832 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1833 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1834 "options when quota turned on");
1835 return -1;
1836 }
1837
1838 if (m->flags & MOPT_NOSUPPORT) {
1839 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1840 } else if (token == Opt_commit) {
1841 if (arg == 0)
1842 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1843 sbi->s_commit_interval = HZ * arg;
670e9875
TT		 1844 } else if (token == Opt_debug_want_extra_isize) {
1845 sbi->s_want_extra_isize = arg;
5f3633e3 1846 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1847 sbi->s_max_batch_time = arg;
1848 } else if (token == Opt_min_batch_time) {
1849 sbi->s_min_batch_time = arg;
1850 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1851 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1852 ext4_msg(sb, KERN_ERR,
1853 "EXT4-fs: inode_readahead_blks must be "
1854 "0 or a power of 2 smaller than 2^31");
26092bf5 1855 return -1;
5f3633e3
JK		 1856 }
1857 sbi->s_inode_readahead_blks = arg;
1858 } else if (token == Opt_init_itable) {
1859 set_opt(sb, INIT_INODE_TABLE);
1860 if (!args->from)
1861 arg = EXT4_DEF_LI_WAIT_MULT;
1862 sbi->s_li_wait_mult = arg;
1863 } else if (token == Opt_max_dir_size_kb) {
1864 sbi->s_max_dir_size_kb = arg;
1865 } else if (token == Opt_stripe) {
1866 sbi->s_stripe = arg;
1867 } else if (token == Opt_resuid) {
1868 uid = make_kuid(current_user_ns(), arg);
1869 if (!uid_valid(uid)) {
1870 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1871 return -1;
1872 }
5f3633e3
JK		 1873 sbi->s_resuid = uid;
1874 } else if (token == Opt_resgid) {
1875 gid = make_kgid(current_user_ns(), arg);
1876 if (!gid_valid(gid)) {
1877 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1878 return -1;
1879 }
1880 sbi->s_resgid = gid;
1881 } else if (token == Opt_journal_dev) {
1882 if (is_remount) {
1883 ext4_msg(sb, KERN_ERR,
1884 "Cannot specify journal on remount");
1885 return -1;
1886 }
1887 *journal_devnum = arg;
ad4eec61
ES		 1888 } else if (token == Opt_journal_path) {
1889 char *journal_path;
1890 struct inode *journal_inode;
1891 struct path path;
1892 int error;
1893
1894 if (is_remount) {
1895 ext4_msg(sb, KERN_ERR,
1896 "Cannot specify journal on remount");
1897 return -1;
1898 }
1899 journal_path = match_strdup(&args[0]);
1900 if (!journal_path) {
1901 ext4_msg(sb, KERN_ERR, "error: could not dup "
1902 "journal device string");
1903 return -1;
1904 }
1905
1906 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1907 if (error) {
1908 ext4_msg(sb, KERN_ERR, "error: could not find "
1909 "journal device path: error %d", error);
1910 kfree(journal_path);
1911 return -1;
1912 }
1913
2b0143b5 1914 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1915 if (!S_ISBLK(journal_inode->i_mode)) {
1916 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1917 "is not a block device", journal_path);
1918 path_put(&path);
1919 kfree(journal_path);
1920 return -1;
1921 }
1922
1923 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1924 path_put(&path);
1925 kfree(journal_path);
5f3633e3
JK		 1926 } else if (token == Opt_journal_ioprio) {
1927 if (arg > 7) {
1928 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1929 " (must be 0-7)");
1930 return -1;
1931 }
1932 *journal_ioprio =
1933 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447 1934 } else if (token == Opt_test_dummy_encryption) {
643fa961 1935#ifdef CONFIG_FS_ENCRYPTION
6ddb2447
TT		 1936 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1937 ext4_msg(sb, KERN_WARNING,
1938 "Test dummy encryption mode enabled");
1939#else
1940 ext4_msg(sb, KERN_WARNING,
1941 "Test dummy encryption mount option ignored");
1942#endif
5f3633e3
JK		 1943 } else if (m->flags & MOPT_DATAJ) {
1944 if (is_remount) {
1945 if (!sbi->s_journal)
1946 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1947 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1948 ext4_msg(sb, KERN_ERR,
26092bf5 1949 "Cannot change data mode on remount");
26092bf5 1950 return -1;
ac27a0ec 1951 }
26092bf5 1952 } else {
5f3633e3
JK		 1953 clear_opt(sb, DATA_FLAGS);
1954 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1955 }
5f3633e3
JK		 1956#ifdef CONFIG_QUOTA
1957 } else if (m->flags & MOPT_QFMT) {
1958 if (sb_any_quota_loaded(sb) &&
1959 sbi->s_jquota_fmt != m->mount_opt) {
1960 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1961 "quota options when quota turned on");
1962 return -1;
1963 }
e2b911c5 1964 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1965 ext4_msg(sb, KERN_INFO,
1966 "Quota format mount options ignored "
262b4662 1967 "when QUOTA feature is enabled");
c325a67c 1968 return 1;
262b4662 1969 }
5f3633e3 1970 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1971#endif
923ae0ff 1972 } else if (token == Opt_dax) {
ef83b6e8
DW		 1973#ifdef CONFIG_FS_DAX
1974 ext4_msg(sb, KERN_WARNING,
1975 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
a92abd73 1976 sbi->s_mount_opt |= m->mount_opt;
ef83b6e8 1977#else
923ae0ff
RZ		 1978 ext4_msg(sb, KERN_INFO, "dax option not supported");
1979 return -1;
5f3633e3 1980#endif
7915a861
AN		 1981 } else if (token == Opt_data_err_abort) {
1982 sbi->s_mount_opt |= m->mount_opt;
1983 } else if (token == Opt_data_err_ignore) {
1984 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1985 } else {
1986 if (!args->from)
1987 arg = 1;
1988 if (m->flags & MOPT_CLEAR)
1989 arg = !arg;
1990 else if (unlikely(!(m->flags & MOPT_SET))) {
1991 ext4_msg(sb, KERN_WARNING,
1992 "buggy handling of option %s", opt);
1993 WARN_ON(1);
1994 return -1;
1995 }
1996 if (arg != 0)
1997 sbi->s_mount_opt |= m->mount_opt;
1998 else
1999 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 2000 }
5f3633e3 2001 return 1;
26092bf5
TT		 2002}
2003
2004static int parse_options(char *options, struct super_block *sb,
2005 unsigned long *journal_devnum,
2006 unsigned int *journal_ioprio,
2007 int is_remount)
2008{
2009 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 2010 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
26092bf5
TT		 2011 substring_t args[MAX_OPT_ARGS];
2012 int token;
2013
2014 if (!options)
2015 return 1;
2016
2017 while ((p = strsep(&options, ",")) != NULL) {
2018 if (!*p)
2019 continue;
2020 /*
2021 * Initialize args struct so we know whether arg was
2022 * found; some options take optional arguments.
2023 */
caecd0af 2024 args[0].to = args[0].from = NULL;
26092bf5
TT		 2025 token = match_token(p, tokens, args);
2026 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2027 journal_ioprio, is_remount) < 0)
2028 return 0;
ac27a0ec
DK		 2029 }
2030#ifdef CONFIG_QUOTA
49da9392
JK		 2031 /*
2032 * We do the test below only for project quotas. 'usrquota' and
2033 * 'grpquota' mount options are allowed even without quota feature
2034 * to support legacy quotas in quota files.
2035 */
2036 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2037 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2038 "Cannot enable project quota enforcement.");
2039 return 0;
2040 }
33458eab
TT		 2041 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2042 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2043 if (usr_qf_name || grp_qf_name) {
2044 if (test_opt(sb, USRQUOTA) && usr_qf_name)
fd8c37ec 2045 clear_opt(sb, USRQUOTA);
ac27a0ec 2046
33458eab 2047 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
fd8c37ec 2048 clear_opt(sb, GRPQUOTA);
ac27a0ec 2049
56c50f11 2050 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 2051 ext4_msg(sb, KERN_ERR, "old and new quota "
2052 "format mixing");
ac27a0ec
DK		 2053 return 0;
2054 }
2055
2056 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 2057 ext4_msg(sb, KERN_ERR, "journaled quota format "
2058 "not specified");
ac27a0ec
DK		 2059 return 0;
2060 }
ac27a0ec
DK		 2061 }
2062#endif
261cb20c
JK		 2063 if (test_opt(sb, DIOREAD_NOLOCK)) {
2064 int blocksize =
2065 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2066
09cbfeaf 2067 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 2068 ext4_msg(sb, KERN_ERR, "can't mount with "
2069 "dioread_nolock if block size != PAGE_SIZE");
2070 return 0;
2071 }
2072 }
ac27a0ec
DK		 2073 return 1;
2074}
2075
2adf6da8
TT		 2076static inline void ext4_show_quota_options(struct seq_file *seq,
2077 struct super_block *sb)
2078{
2079#if defined(CONFIG_QUOTA)
2080 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 2081 char *usr_qf_name, *grp_qf_name;
2adf6da8
TT		 2082
2083 if (sbi->s_jquota_fmt) {
2084 char *fmtname = "";
2085
2086 switch (sbi->s_jquota_fmt) {
2087 case QFMT_VFS_OLD:
2088 fmtname = "vfsold";
2089 break;
2090 case QFMT_VFS_V0:
2091 fmtname = "vfsv0";
2092 break;
2093 case QFMT_VFS_V1:
2094 fmtname = "vfsv1";
2095 break;
2096 }
2097 seq_printf(seq, ",jqfmt=%s", fmtname);
2098 }
2099
33458eab
TT		 2100 rcu_read_lock();
2101 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2102 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2103 if (usr_qf_name)
2104 seq_show_option(seq, "usrjquota", usr_qf_name);
2105 if (grp_qf_name)
2106 seq_show_option(seq, "grpjquota", grp_qf_name);
2107 rcu_read_unlock();
2adf6da8
TT		 2108#endif
2109}
2110
5a916be1
TT		 2111static const char *token2str(int token)
2112{
50df9fd5 2113 const struct match_token *t;
5a916be1
TT		 2114
2115 for (t = tokens; t->token != Opt_err; t++)
2116 if (t->token == token && !strchr(t->pattern, '='))
2117 break;
2118 return t->pattern;
2119}
2120
2adf6da8
TT		 2121/*
2122 * Show an option if
2123 * - it's set to a non-default value OR
2124 * - if the per-sb default is different from the global default
2125 */
66acdcf4
TT		 2126static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2127 int nodefs)
2adf6da8 2128{
2adf6da8
TT		 2129 struct ext4_sb_info *sbi = EXT4_SB(sb);
2130 struct ext4_super_block *es = sbi->s_es;
68afa7e0 2131 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
5a916be1 2132 const struct mount_opts *m;
66acdcf4 2133 char sep = nodefs ? '\n' : ',';
2adf6da8 2134
66acdcf4
TT		 2135#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2136#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2137
2138 if (sbi->s_sb_block != 1)
5a916be1
TT		 2139 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2140
2141 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2142 int want_set = m->flags & MOPT_SET;
2143 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2144 (m->flags & MOPT_CLEAR_ERR))
2145 continue;
68afa7e0 2146 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2147 continue; /* skip if same as the default */
2148 if ((want_set &&
2149 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2150 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2151 continue; /* select Opt_noFoo vs Opt_Foo */
2152 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2153 }
5a916be1 2154
08cefc7a 2155 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2156 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2157 SEQ_OPTS_PRINT("resuid=%u",
2158 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2159 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2160 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2161 SEQ_OPTS_PRINT("resgid=%u",
2162 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2163 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2164 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2165 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2166 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2167 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2168 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2169 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2170 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2171 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2172 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2173 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2174 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2175 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
357fdad0 2176 if (sb->s_flags & SB_I_VERSION)
5a916be1 2177 SEQ_OPTS_PUTS("i_version");
66acdcf4 2178 if (nodefs || sbi->s_stripe)
5a916be1 2179 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0
TN		 2180 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2181 (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2182 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2183 SEQ_OPTS_PUTS("data=journal");
2184 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2185 SEQ_OPTS_PUTS("data=ordered");
2186 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2187 SEQ_OPTS_PUTS("data=writeback");
2188 }
66acdcf4
TT		 2189 if (nodefs ||
2190 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2191 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2192 sbi->s_inode_readahead_blks);
2adf6da8 2193
ceec0376 2194 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 2195 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2196 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2197 if (nodefs || sbi->s_max_dir_size_kb)
2198 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2199 if (test_opt(sb, DATA_ERR_ABORT))
2200 SEQ_OPTS_PUTS("data_err=abort");
338affb5
EB		 2201 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2202 SEQ_OPTS_PUTS("test_dummy_encryption");
2adf6da8
TT		 2203
2204 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2205 return 0;
2206}
2207
66acdcf4
TT		 2208static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2209{
2210 return _ext4_show_options(seq, root->d_sb, 0);
2211}
2212
ebd173be 2213int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2214{
2215 struct super_block *sb = seq->private;
2216 int rc;
2217
bc98a42c 2218 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 2219 rc = _ext4_show_options(seq, sb, 1);
2220 seq_puts(seq, "\n");
2221 return rc;
2222}
2223
617ba13b 2224static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2225 int read_only)
2226{
617ba13b 2227 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 2228 int err = 0;
ac27a0ec 2229
617ba13b 2230 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2231 ext4_msg(sb, KERN_ERR, "revision level too high, "
2232 "forcing read-only mode");
c89128a0 2233 err = -EROFS;
ac27a0ec
DK		 2234 }
2235 if (read_only)
281b5995 2236 goto done;
617ba13b 2237 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2238 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2239 "running e2fsck is recommended");
c8b459f4 2240 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2241 ext4_msg(sb, KERN_WARNING,
2242 "warning: mounting fs with errors, "
2243 "running e2fsck is recommended");
ed3ce80a 2244 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2245 le16_to_cpu(es->s_mnt_count) >=
2246 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2247 ext4_msg(sb, KERN_WARNING,
2248 "warning: maximal mount count reached, "
2249 "running e2fsck is recommended");
ac27a0ec 2250 else if (le32_to_cpu(es->s_checkinterval) &&
6a0678a7
AB		 2251 (ext4_get_tstamp(es, s_lastcheck) +
2252 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
b31e1552
ES		 2253 ext4_msg(sb, KERN_WARNING,
2254 "warning: checktime reached, "
2255 "running e2fsck is recommended");
0b8e58a1 2256 if (!sbi->s_journal)
0390131b 2257 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2258 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2259 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2260 le16_add_cpu(&es->s_mnt_count, 1);
6a0678a7 2261 ext4_update_tstamp(es, s_mtime);
0390131b 2262 if (sbi->s_journal)
e2b911c5 2263 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2264
c89128a0 2265 err = ext4_commit_super(sb, 1);
281b5995 2266done:
ac27a0ec 2267 if (test_opt(sb, DEBUG))
a9df9a49 2268 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2269 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2270 sb->s_blocksize,
2271 sbi->s_groups_count,
617ba13b
MC		 2272 EXT4_BLOCKS_PER_GROUP(sb),
2273 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2274 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2275
7abc52c2 2276 cleancache_init_fs(sb);
c89128a0 2277 return err;
ac27a0ec
DK		 2278}
2279
117fff10
TT		 2280int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2281{
2282 struct ext4_sb_info *sbi = EXT4_SB(sb);
2283 struct flex_groups *new_groups;
2284 int size;
2285
2286 if (!sbi->s_log_groups_per_flex)
2287 return 0;
2288
2289 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2290 if (size <= sbi->s_flex_groups_allocated)
2291 return 0;
2292
2293 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
a7c3e901 2294 new_groups = kvzalloc(size, GFP_KERNEL);
117fff10
TT		 2295 if (!new_groups) {
2296 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2297 size / (int) sizeof(struct flex_groups));
2298 return -ENOMEM;
2299 }
2300
2301 if (sbi->s_flex_groups) {
2302 memcpy(new_groups, sbi->s_flex_groups,
2303 (sbi->s_flex_groups_allocated *
2304 sizeof(struct flex_groups)));
b93b41d4 2305 kvfree(sbi->s_flex_groups);
117fff10
TT		 2306 }
2307 sbi->s_flex_groups = new_groups;
2308 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2309 return 0;
2310}
2311
772cb7c8
JS		 2312static int ext4_fill_flex_info(struct super_block *sb)
2313{
2314 struct ext4_sb_info *sbi = EXT4_SB(sb);
2315 struct ext4_group_desc *gdp = NULL;
772cb7c8 2316 ext4_group_t flex_group;
117fff10 2317 int i, err;
772cb7c8 2318
503358ae 2319 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2320 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2321 sbi->s_log_groups_per_flex = 0;
2322 return 1;
2323 }
2324
117fff10
TT		 2325 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2326 if (err)
9933fc0a 2327 goto failed;
772cb7c8 2328
772cb7c8 2329 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2330 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2331
2332 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2333 atomic_add(ext4_free_inodes_count(sb, gdp),
2334 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2335 atomic64_add(ext4_free_group_clusters(sb, gdp),
2336 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2337 atomic_add(ext4_used_dirs_count(sb, gdp),
2338 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2339 }
2340
2341 return 1;
2342failed:
2343 return 0;
2344}
2345
e2b911c5 2346static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2347 struct ext4_group_desc *gdp)
717d50e4 2348{
b47820ed 2349 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2350 __u16 crc = 0;
feb0ab32 2351 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2352 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2353
9aa5d32b 2354 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2355 /* Use new metadata_csum algorithm */
feb0ab32 2356 __u32 csum32;
b47820ed 2357 __u16 dummy_csum = 0;
feb0ab32 2358
feb0ab32
DW		 2359 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2360 sizeof(le_group));
b47820ed
DJ		 2361 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2362 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2363 sizeof(dummy_csum));
2364 offset += sizeof(dummy_csum);
2365 if (offset < sbi->s_desc_size)
2366 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2367 sbi->s_desc_size - offset);
feb0ab32
DW		 2368
2369 crc = csum32 & 0xFFFF;
2370 goto out;
717d50e4
AD		 2371 }
2372
feb0ab32 2373 /* old crc16 code */
e2b911c5 2374 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2375 return 0;
2376
feb0ab32
DW		 2377 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2378 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2379 crc = crc16(crc, (__u8 *)gdp, offset);
2380 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2381 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2382 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2383 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2384 crc = crc16(crc, (__u8 *)gdp + offset,
2385 le16_to_cpu(sbi->s_es->s_desc_size) -
2386 offset);
2387
2388out:
717d50e4
AD		 2389 return cpu_to_le16(crc);
2390}
2391
feb0ab32 2392int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2393 struct ext4_group_desc *gdp)
2394{
feb0ab32 2395 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2396 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2397 return 0;
2398
2399 return 1;
2400}
2401
feb0ab32
DW		 2402void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2403 struct ext4_group_desc *gdp)
2404{
2405 if (!ext4_has_group_desc_csum(sb))
2406 return;
e2b911c5 2407 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2408}
2409
ac27a0ec 2410/* Called at mount-time, super-block is locked */
bfff6873 2411static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2412 ext4_fsblk_t sb_block,
bfff6873 2413 ext4_group_t *first_not_zeroed)
ac27a0ec 2414{
617ba13b
MC		 2415 struct ext4_sb_info *sbi = EXT4_SB(sb);
2416 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2417 ext4_fsblk_t last_block;
44de022c 2418 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
bd81d8ee
LV		 2419 ext4_fsblk_t block_bitmap;
2420 ext4_fsblk_t inode_bitmap;
2421 ext4_fsblk_t inode_table;
ce421581 2422 int flexbg_flag = 0;
bfff6873 2423 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2424
e2b911c5 2425 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2426 flexbg_flag = 1;
2427
af5bc92d 2428 ext4_debug("Checking group descriptors");
ac27a0ec 2429
197cd65a
AM		 2430 for (i = 0; i < sbi->s_groups_count; i++) {
2431 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2432
ce421581 2433 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2434 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2435 else
2436 last_block = first_block +
617ba13b 2437 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2438
bfff6873
LC		 2439 if ((grp == sbi->s_groups_count) &&
2440 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2441 grp = i;
2442
8fadc143 2443 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2444 if (block_bitmap == sb_block) {
2445 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2446 "Block bitmap for group %u overlaps "
2447 "superblock", i);
18db4b4e
TT		 2448 if (!sb_rdonly(sb))
2449 return 0;
829fa70d 2450 }
77260807
TT		 2451 if (block_bitmap >= sb_block + 1 &&
2452 block_bitmap <= last_bg_block) {
2453 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2454 "Block bitmap for group %u overlaps "
2455 "block group descriptors", i);
2456 if (!sb_rdonly(sb))
2457 return 0;
2458 }
2b2d6d01 2459 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2460 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2461 "Block bitmap for group %u not in group "
b31e1552 2462 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2463 return 0;
2464 }
8fadc143 2465 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2466 if (inode_bitmap == sb_block) {
2467 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2468 "Inode bitmap for group %u overlaps "
2469 "superblock", i);
18db4b4e
TT		 2470 if (!sb_rdonly(sb))
2471 return 0;
829fa70d 2472 }
77260807
TT		 2473 if (inode_bitmap >= sb_block + 1 &&
2474 inode_bitmap <= last_bg_block) {
2475 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2476 "Inode bitmap for group %u overlaps "
2477 "block group descriptors", i);
2478 if (!sb_rdonly(sb))
2479 return 0;
2480 }
2b2d6d01 2481 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2482 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2483 "Inode bitmap for group %u not in group "
b31e1552 2484 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2485 return 0;
2486 }
8fadc143 2487 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2488 if (inode_table == sb_block) {
2489 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2490 "Inode table for group %u overlaps "
2491 "superblock", i);
18db4b4e
TT		 2492 if (!sb_rdonly(sb))
2493 return 0;
829fa70d 2494 }
77260807
TT		 2495 if (inode_table >= sb_block + 1 &&
2496 inode_table <= last_bg_block) {
2497 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2498 "Inode table for group %u overlaps "
2499 "block group descriptors", i);
2500 if (!sb_rdonly(sb))
2501 return 0;
2502 }
bd81d8ee 2503 if (inode_table < first_block ||
2b2d6d01 2504 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2505 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2506 "Inode table for group %u not in group "
b31e1552 2507 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2508 return 0;
2509 }
955ce5f5 2510 ext4_lock_group(sb, i);
feb0ab32 2511 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2512 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2513 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2514 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2515 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 2516 if (!sb_rdonly(sb)) {
955ce5f5 2517 ext4_unlock_group(sb, i);
8a266467 2518 return 0;
7ee1ec4c 2519 }
717d50e4 2520 }
955ce5f5 2521 ext4_unlock_group(sb, i);
ce421581
JS		 2522 if (!flexbg_flag)
2523 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2524 }
bfff6873
LC		 2525 if (NULL != first_not_zeroed)
2526 *first_not_zeroed = grp;
ac27a0ec
DK		 2527 return 1;
2528}
2529
617ba13b 2530/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2531 * the superblock) which were deleted from all directories, but held open by
2532 * a process at the time of a crash. We walk the list and try to delete these
2533 * inodes at recovery time (only with a read-write filesystem).
2534 *
2535 * In order to keep the orphan inode chain consistent during traversal (in
2536 * case of crash during recovery), we link each inode into the superblock
2537 * orphan list_head and handle it the same way as an inode deletion during
2538 * normal operation (which journals the operations for us).
2539 *
2540 * We only do an iget() and an iput() on each inode, which is very safe if we
2541 * accidentally point at an in-use or already deleted inode. The worst that
2542 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2543 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2544 * e2fsck was run on this filesystem, and it must have already done the orphan
2545 * inode cleanup for us, so we can safely abort without any further action.
2546 */
2b2d6d01
TT		 2547static void ext4_orphan_cleanup(struct super_block *sb,
2548 struct ext4_super_block *es)
ac27a0ec
DK		 2549{
2550 unsigned int s_flags = sb->s_flags;
2c98eb5e 2551 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec 2552#ifdef CONFIG_QUOTA
95f1fda4 2553 int quota_update = 0;
ac27a0ec
DK		 2554 int i;
2555#endif
2556 if (!es->s_last_orphan) {
2557 jbd_debug(4, "no orphan inodes to clean up\n");
2558 return;
2559 }
2560
a8f48a95 2561 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2562 ext4_msg(sb, KERN_ERR, "write access "
2563 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2564 return;
2565 }
2566
d39195c3
AG		 2567 /* Check if feature set would not allow a r/w mount */
2568 if (!ext4_feature_set_ok(sb, 0)) {
2569 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2570 "unknown ROCOMPAT features");
2571 return;
2572 }
2573
617ba13b 2574 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6 2575 /* don't clear list on RO mount w/ errors */
1751e8a6 2576 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
84474976 2577 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2578 "clearing orphan list.\n");
c25f9bc6
ES		 2579 es->s_last_orphan = 0;
2580 }
ac27a0ec
DK		 2581 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2582 return;
2583 }
2584
1751e8a6 2585 if (s_flags & SB_RDONLY) {
b31e1552 2586 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1751e8a6 2587 sb->s_flags &= ~SB_RDONLY;
ac27a0ec
DK		 2588 }
2589#ifdef CONFIG_QUOTA
2590 /* Needed for iput() to work correctly and not trash data */
1751e8a6 2591 sb->s_flags |= SB_ACTIVE;
95f1fda4 2592
2593 /*
2594 * Turn on quotas which were not enabled for read-only mounts if
2595 * filesystem has quota feature, so that they are updated correctly.
2596 */
1751e8a6 2597 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
95f1fda4 2598 int ret = ext4_enable_quotas(sb);
2599
2600 if (!ret)
2601 quota_update = 1;
2602 else
2603 ext4_msg(sb, KERN_ERR,
2604 "Cannot turn on quotas: error %d", ret);
2605 }
2606
2607 /* Turn on journaled quotas used for old sytle */
a2d4a646 2608 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2609 if (EXT4_SB(sb)->s_qf_names[i]) {
2610 int ret = ext4_quota_on_mount(sb, i);
95f1fda4 2611
2612 if (!ret)
2613 quota_update = 1;
2614 else
b31e1552
ES		 2615 ext4_msg(sb, KERN_ERR,
2616 "Cannot turn on journaled "
95f1fda4 2617 "quota: type %d: error %d", i, ret);
ac27a0ec
DK		 2618 }
2619 }
2620#endif
2621
2622 while (es->s_last_orphan) {
2623 struct inode *inode;
2624
c65d5c6c
VN		 2625 /*
2626 * We may have encountered an error during cleanup; if
2627 * so, skip the rest.
2628 */
2629 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2630 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2631 es->s_last_orphan = 0;
2632 break;
2633 }
2634
97bd42b9
JB		 2635 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2636 if (IS_ERR(inode)) {
ac27a0ec
DK		 2637 es->s_last_orphan = 0;
2638 break;
2639 }
2640
617ba13b 2641 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2642 dquot_initialize(inode);
ac27a0ec 2643 if (inode->i_nlink) {
566370a2
PT		 2644 if (test_opt(sb, DEBUG))
2645 ext4_msg(sb, KERN_DEBUG,
2646 "%s: truncating inode %lu to %lld bytes",
2647 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2648 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2649 inode->i_ino, inode->i_size);
5955102c 2650 inode_lock(inode);
55f252c9 2651 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2652 ret = ext4_truncate(inode);
2653 if (ret)
2654 ext4_std_error(inode->i_sb, ret);
5955102c 2655 inode_unlock(inode);
ac27a0ec
DK		 2656 nr_truncates++;
2657 } else {
566370a2
PT		 2658 if (test_opt(sb, DEBUG))
2659 ext4_msg(sb, KERN_DEBUG,
2660 "%s: deleting unreferenced inode %lu",
2661 __func__, inode->i_ino);
ac27a0ec
DK		 2662 jbd_debug(2, "deleting unreferenced inode %lu\n",
2663 inode->i_ino);
2664 nr_orphans++;
2665 }
2666 iput(inode); /* The delete magic happens here! */
2667 }
2668
2b2d6d01 2669#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2670
2671 if (nr_orphans)
b31e1552
ES		 2672 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2673 PLURAL(nr_orphans));
ac27a0ec 2674 if (nr_truncates)
b31e1552
ES		 2675 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2676 PLURAL(nr_truncates));
ac27a0ec 2677#ifdef CONFIG_QUOTA
95f1fda4 2678 /* Turn off quotas if they were enabled for orphan cleanup */
2679 if (quota_update) {
2680 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2681 if (sb_dqopt(sb)->files[i])
2682 dquot_quota_off(sb, i);
2683 }
ac27a0ec
DK		 2684 }
2685#endif
1751e8a6 2686 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
ac27a0ec 2687}
0b8e58a1 2688
cd2291a4
ES		 2689/*
2690 * Maximal extent format file size.
2691 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2692 * extent format containers, within a sector_t, and within i_blocks
2693 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2694 * so that won't be a limiting factor.
2695 *
f17722f9
LC		 2696 * However there is other limiting factor. We do store extents in the form
2697 * of starting block and length, hence the resulting length of the extent
2698 * covering maximum file size must fit into on-disk format containers as
2699 * well. Given that length is always by 1 unit bigger than max unit (because
2700 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2701 *
cd2291a4
ES		 2702 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2703 */
f287a1a5 2704static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2705{
2706 loff_t res;
2707 loff_t upper_limit = MAX_LFS_FILESIZE;
2708
72deb455
CH		 2709 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2710
2711 if (!has_huge_files) {
cd2291a4
ES		 2712 upper_limit = (1LL << 32) - 1;
2713
2714 /* total blocks in file system block size */
2715 upper_limit >>= (blkbits - 9);
2716 upper_limit <<= blkbits;
2717 }
2718
f17722f9
LC		 2719 /*
2720 * 32-bit extent-start container, ee_block. We lower the maxbytes
2721 * by one fs block, so ee_len can cover the extent of maximum file
2722 * size
2723 */
2724 res = (1LL << 32) - 1;
cd2291a4 2725 res <<= blkbits;
cd2291a4
ES		 2726
2727 /* Sanity check against vm- & vfs- imposed limits */
2728 if (res > upper_limit)
2729 res = upper_limit;
2730
2731 return res;
2732}
ac27a0ec 2733
ac27a0ec 2734/*
cd2291a4 2735 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2736 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2737 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2738 */
f287a1a5 2739static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2740{
617ba13b 2741 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2742 int meta_blocks;
2743 loff_t upper_limit;
0b8e58a1
AD		 2744 /* This is calculated to be the largest file size for a dense, block
2745 * mapped file such that the file's total number of 512-byte sectors,
2746 * including data and all indirect blocks, does not exceed (2^48 - 1).
2747 *
2748 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2749 * number of 512-byte sectors of the file.
0fc1b451
AK		 2750 */
2751
72deb455 2752 if (!has_huge_files) {
0fc1b451 2753 /*
72deb455
CH		 2754 * !has_huge_files or implies that the inode i_block field
2755 * represents total file blocks in 2^32 512-byte sectors ==
2756 * size of vfs inode i_blocks * 8
0fc1b451
AK		 2757 */
2758 upper_limit = (1LL << 32) - 1;
2759
2760 /* total blocks in file system block size */
2761 upper_limit >>= (bits - 9);
2762
2763 } else {
8180a562
AK		 2764 /*
2765 * We use 48 bit ext4_inode i_blocks
2766 * With EXT4_HUGE_FILE_FL set the i_blocks
2767 * represent total number of blocks in
2768 * file system block size
2769 */
0fc1b451
AK		 2770 upper_limit = (1LL << 48) - 1;
2771
0fc1b451
AK		 2772 }
2773
2774 /* indirect blocks */
2775 meta_blocks = 1;
2776 /* double indirect blocks */
2777 meta_blocks += 1 + (1LL << (bits-2));
2778 /* tripple indirect blocks */
2779 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2780
2781 upper_limit -= meta_blocks;
2782 upper_limit <<= bits;
ac27a0ec
DK		 2783
2784 res += 1LL << (bits-2);
2785 res += 1LL << (2*(bits-2));
2786 res += 1LL << (3*(bits-2));
2787 res <<= bits;
2788 if (res > upper_limit)
2789 res = upper_limit;
0fc1b451
AK		 2790
2791 if (res > MAX_LFS_FILESIZE)
2792 res = MAX_LFS_FILESIZE;
2793
ac27a0ec
DK		 2794 return res;
2795}
2796
617ba13b 2797static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2798 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2799{
617ba13b 2800 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2801 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2802 int has_super = 0;
2803
2804 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2805
e2b911c5 2806 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2807 return logical_sb_block + nr + 1;
ac27a0ec 2808 bg = sbi->s_desc_per_block * nr;
617ba13b 2809 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2810 has_super = 1;
0b8e58a1 2811
bd63f6b0
DW		 2812 /*
2813 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2814 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2815 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2816 * compensate.
2817 */
2818 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 2819 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 2820 has_super++;
2821
617ba13b 2822 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2823}
2824
c9de560d
AT		 2825/**
2826 * ext4_get_stripe_size: Get the stripe size.
2827 * @sbi: In memory super block info
2828 *
2829 * If we have specified it via mount option, then
2830 * use the mount option value. If the value specified at mount time is
2831 * greater than the blocks per group use the super block value.
2832 * If the super block value is greater than blocks per group return 0.
2833 * Allocator needs it be less than blocks per group.
2834 *
2835 */
2836static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2837{
2838 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2839 unsigned long stripe_width =
2840 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2841 int ret;
c9de560d
AT		 2842
2843 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 2844 ret = sbi->s_stripe;
5469d7c3 2845 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 2846 ret = stripe_width;
5469d7c3 2847 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 2848 ret = stride;
2849 else
2850 ret = 0;
c9de560d 2851
3eb08658
DE		 2852 /*
2853 * If the stripe width is 1, this makes no sense and
2854 * we set it to 0 to turn off stripe handling code.
2855 */
2856 if (ret <= 1)
2857 ret = 0;
c9de560d 2858
3eb08658 2859 return ret;
c9de560d 2860}
ac27a0ec 2861
a13fb1a4
ES		 2862/*
2863 * Check whether this filesystem can be mounted based on
2864 * the features present and the RDONLY/RDWR mount requested.
2865 * Returns 1 if this filesystem can be mounted as requested,
2866 * 0 if it cannot be.
2867 */
2868static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2869{
e2b911c5 2870 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2871 ext4_msg(sb, KERN_ERR,
2872 "Couldn't mount because of "
2873 "unsupported optional features (%x)",
2874 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2875 ~EXT4_FEATURE_INCOMPAT_SUPP));
2876 return 0;
2877 }
2878
2879 if (readonly)
2880 return 1;
2881
e2b911c5 2882 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 2883 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 2884 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 2885 return 1;
2886 }
2887
a13fb1a4 2888 /* Check that feature set is OK for a read-write mount */
e2b911c5 2889 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2890 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2891 "unsupported optional features (%x)",
2892 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2893 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2894 return 0;
2895 }
e2b911c5 2896 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2897 ext4_msg(sb, KERN_ERR,
2898 "Can't support bigalloc feature without "
2899 "extents feature\n");
2900 return 0;
2901 }
7c319d32
AK		 2902
2903#ifndef CONFIG_QUOTA
e2b911c5 2904 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2905 ext4_msg(sb, KERN_ERR,
2906 "Filesystem with quota feature cannot be mounted RDWR "
2907 "without CONFIG_QUOTA");
2908 return 0;
2909 }
689c958c
LX		 2910 if (ext4_has_feature_project(sb) && !readonly) {
2911 ext4_msg(sb, KERN_ERR,
2912 "Filesystem with project quota feature cannot be mounted RDWR "
2913 "without CONFIG_QUOTA");
2914 return 0;
2915 }
7c319d32 2916#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2917 return 1;
2918}
2919
66e61a9e
TT		 2920/*
2921 * This function is called once a day if we have errors logged
2922 * on the file system
2923 */
235699a8 2924static void print_daily_error_info(struct timer_list *t)
66e61a9e 2925{
235699a8
KC		 2926 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2927 struct super_block *sb = sbi->s_sb;
2928 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 2929
2930 if (es->s_error_count)
ae0f78de
TT		 2931 /* fsck newer than v1.41.13 is needed to clean this condition. */
2932 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2933 le32_to_cpu(es->s_error_count));
2934 if (es->s_first_error_time) {
6a0678a7
AB		 2935 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
2936 sb->s_id,
2937 ext4_get_tstamp(es, s_first_error_time),
66e61a9e
TT		 2938 (int) sizeof(es->s_first_error_func),
2939 es->s_first_error_func,
2940 le32_to_cpu(es->s_first_error_line));
2941 if (es->s_first_error_ino)
651e1c3b 2942 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2943 le32_to_cpu(es->s_first_error_ino));
2944 if (es->s_first_error_block)
651e1c3b 2945 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2946 le64_to_cpu(es->s_first_error_block));
651e1c3b 2947 printk(KERN_CONT "\n");
66e61a9e
TT		 2948 }
2949 if (es->s_last_error_time) {
6a0678a7
AB		 2950 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
2951 sb->s_id,
2952 ext4_get_tstamp(es, s_last_error_time),
66e61a9e
TT		 2953 (int) sizeof(es->s_last_error_func),
2954 es->s_last_error_func,
2955 le32_to_cpu(es->s_last_error_line));
2956 if (es->s_last_error_ino)
651e1c3b 2957 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2958 le32_to_cpu(es->s_last_error_ino));
2959 if (es->s_last_error_block)
651e1c3b 2960 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2961 le64_to_cpu(es->s_last_error_block));
651e1c3b 2962 printk(KERN_CONT "\n");
66e61a9e
TT		 2963 }
2964 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2965}
2966
bfff6873
LC		 2967/* Find next suitable group and run ext4_init_inode_table */
2968static int ext4_run_li_request(struct ext4_li_request *elr)
2969{
2970 struct ext4_group_desc *gdp = NULL;
2971 ext4_group_t group, ngroups;
2972 struct super_block *sb;
2973 unsigned long timeout = 0;
2974 int ret = 0;
2975
2976 sb = elr->lr_super;
2977 ngroups = EXT4_SB(sb)->s_groups_count;
2978
2979 for (group = elr->lr_next_group; group < ngroups; group++) {
2980 gdp = ext4_get_group_desc(sb, group, NULL);
2981 if (!gdp) {
2982 ret = 1;
2983 break;
2984 }
2985
2986 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2987 break;
2988 }
2989
7f511862 2990 if (group >= ngroups)
bfff6873
LC		 2991 ret = 1;
2992
2993 if (!ret) {
2994 timeout = jiffies;
2995 ret = ext4_init_inode_table(sb, group,
2996 elr->lr_timeout ? 0 : 1);
2997 if (elr->lr_timeout == 0) {
51ce6511
LC		 2998 timeout = (jiffies - timeout) *
2999 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 3000 elr->lr_timeout = timeout;
3001 }
3002 elr->lr_next_sched = jiffies + elr->lr_timeout;
3003 elr->lr_next_group = group + 1;
3004 }
bfff6873
LC		 3005 return ret;
3006}
3007
3008/*
3009 * Remove lr_request from the list_request and free the
4ed5c033 3010 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 3011 */
3012static void ext4_remove_li_request(struct ext4_li_request *elr)
3013{
3014 struct ext4_sb_info *sbi;
3015
3016 if (!elr)
3017 return;
3018
3019 sbi = elr->lr_sbi;
3020
3021 list_del(&elr->lr_request);
3022 sbi->s_li_request = NULL;
3023 kfree(elr);
3024}
3025
3026static void ext4_unregister_li_request(struct super_block *sb)
3027{
1bb933fb
LC		 3028 mutex_lock(&ext4_li_mtx);
3029 if (!ext4_li_info) {
3030 mutex_unlock(&ext4_li_mtx);
bfff6873 3031 return;
1bb933fb 3032 }
bfff6873
LC		 3033
3034 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 3035 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 3036 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 3037 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 3038}
3039
8f1f7453
ES		 3040static struct task_struct *ext4_lazyinit_task;
3041
bfff6873
LC		 3042/*
3043 * This is the function where ext4lazyinit thread lives. It walks
3044 * through the request list searching for next scheduled filesystem.
3045 * When such a fs is found, run the lazy initialization request
3046 * (ext4_rn_li_request) and keep track of the time spend in this
3047 * function. Based on that time we compute next schedule time of
3048 * the request. When walking through the list is complete, compute
3049 * next waking time and put itself into sleep.
3050 */
3051static int ext4_lazyinit_thread(void *arg)
3052{
3053 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3054 struct list_head *pos, *n;
3055 struct ext4_li_request *elr;
4ed5c033 3056 unsigned long next_wakeup, cur;
bfff6873
LC		 3057
3058 BUG_ON(NULL == eli);
3059
bfff6873
LC		 3060cont_thread:
3061 while (true) {
3062 next_wakeup = MAX_JIFFY_OFFSET;
3063
3064 mutex_lock(&eli->li_list_mtx);
3065 if (list_empty(&eli->li_request_list)) {
3066 mutex_unlock(&eli->li_list_mtx);
3067 goto exit_thread;
3068 }
bfff6873 3069 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3070 int err = 0;
3071 int progress = 0;
bfff6873
LC		 3072 elr = list_entry(pos, struct ext4_li_request,
3073 lr_request);
3074
e22834f0
DM		 3075 if (time_before(jiffies, elr->lr_next_sched)) {
3076 if (time_before(elr->lr_next_sched, next_wakeup))
3077 next_wakeup = elr->lr_next_sched;
3078 continue;
3079 }
3080 if (down_read_trylock(&elr->lr_super->s_umount)) {
3081 if (sb_start_write_trylock(elr->lr_super)) {
3082 progress = 1;
3083 /*
3084 * We hold sb->s_umount, sb can not
3085 * be removed from the list, it is
3086 * now safe to drop li_list_mtx
3087 */
3088 mutex_unlock(&eli->li_list_mtx);
3089 err = ext4_run_li_request(elr);
3090 sb_end_write(elr->lr_super);
3091 mutex_lock(&eli->li_list_mtx);
3092 n = pos->next;
b2c78cd0 3093 }
e22834f0
DM		 3094 up_read((&elr->lr_super->s_umount));
3095 }
3096 /* error, remove the lazy_init job */
3097 if (err) {
3098 ext4_remove_li_request(elr);
3099 continue;
3100 }
3101 if (!progress) {
3102 elr->lr_next_sched = jiffies +
3103 (prandom_u32()
3104 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 3105 }
bfff6873
LC		 3106 if (time_before(elr->lr_next_sched, next_wakeup))
3107 next_wakeup = elr->lr_next_sched;
3108 }
3109 mutex_unlock(&eli->li_list_mtx);
3110
a0acae0e 3111 try_to_freeze();
bfff6873 3112
4ed5c033
LC		 3113 cur = jiffies;
3114 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3115 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3116 cond_resched();
3117 continue;
3118 }
3119
4ed5c033
LC		 3120 schedule_timeout_interruptible(next_wakeup - cur);
3121
8f1f7453
ES		 3122 if (kthread_should_stop()) {
3123 ext4_clear_request_list();
3124 goto exit_thread;
3125 }
bfff6873
LC		 3126 }
3127
3128exit_thread:
3129 /*
3130 * It looks like the request list is empty, but we need
3131 * to check it under the li_list_mtx lock, to prevent any
3132 * additions into it, and of course we should lock ext4_li_mtx
3133 * to atomically free the list and ext4_li_info, because at
3134 * this point another ext4 filesystem could be registering
3135 * new one.
3136 */
3137 mutex_lock(&ext4_li_mtx);
3138 mutex_lock(&eli->li_list_mtx);
3139 if (!list_empty(&eli->li_request_list)) {
3140 mutex_unlock(&eli->li_list_mtx);
3141 mutex_unlock(&ext4_li_mtx);
3142 goto cont_thread;
3143 }
3144 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3145 kfree(ext4_li_info);
3146 ext4_li_info = NULL;
3147 mutex_unlock(&ext4_li_mtx);
3148
3149 return 0;
3150}
3151
3152static void ext4_clear_request_list(void)
3153{
3154 struct list_head *pos, *n;
3155 struct ext4_li_request *elr;
3156
3157 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3158 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3159 elr = list_entry(pos, struct ext4_li_request,
3160 lr_request);
3161 ext4_remove_li_request(elr);
3162 }
3163 mutex_unlock(&ext4_li_info->li_list_mtx);
3164}
3165
3166static int ext4_run_lazyinit_thread(void)
3167{
8f1f7453
ES		 3168 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3169 ext4_li_info, "ext4lazyinit");
3170 if (IS_ERR(ext4_lazyinit_task)) {
3171 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3172 ext4_clear_request_list();
bfff6873
LC		 3173 kfree(ext4_li_info);
3174 ext4_li_info = NULL;
92b97816 3175 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3176 "initialization thread\n",
3177 err);
3178 return err;
3179 }
3180 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3181 return 0;
3182}
3183
3184/*
3185 * Check whether it make sense to run itable init. thread or not.
3186 * If there is at least one uninitialized inode table, return
3187 * corresponding group number, else the loop goes through all
3188 * groups and return total number of groups.
3189 */
3190static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3191{
3192 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3193 struct ext4_group_desc *gdp = NULL;
3194
8844618d
TT		 3195 if (!ext4_has_group_desc_csum(sb))
3196 return ngroups;
3197
bfff6873
LC		 3198 for (group = 0; group < ngroups; group++) {
3199 gdp = ext4_get_group_desc(sb, group, NULL);
3200 if (!gdp)
3201 continue;
3202
50122847 3203 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
bfff6873
LC		 3204 break;
3205 }
3206
3207 return group;
3208}
3209
3210static int ext4_li_info_new(void)
3211{
3212 struct ext4_lazy_init *eli = NULL;
3213
3214 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3215 if (!eli)
3216 return -ENOMEM;
3217
bfff6873
LC		 3218 INIT_LIST_HEAD(&eli->li_request_list);
3219 mutex_init(&eli->li_list_mtx);
3220
bfff6873
LC		 3221 eli->li_state |= EXT4_LAZYINIT_QUIT;
3222
3223 ext4_li_info = eli;
3224
3225 return 0;
3226}
3227
3228static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3229 ext4_group_t start)
3230{
3231 struct ext4_sb_info *sbi = EXT4_SB(sb);
3232 struct ext4_li_request *elr;
bfff6873
LC		 3233
3234 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3235 if (!elr)
3236 return NULL;
3237
3238 elr->lr_super = sb;
3239 elr->lr_sbi = sbi;
3240 elr->lr_next_group = start;
3241
3242 /*
3243 * Randomize first schedule time of the request to
3244 * spread the inode table initialization requests
3245 * better.
3246 */
dd1f723b
TT		 3247 elr->lr_next_sched = jiffies + (prandom_u32() %
3248 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3249 return elr;
3250}
3251
7f511862
TT		 3252int ext4_register_li_request(struct super_block *sb,
3253 ext4_group_t first_not_zeroed)
bfff6873
LC		 3254{
3255 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3256 struct ext4_li_request *elr = NULL;
49598e04 3257 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 3258 int ret = 0;
bfff6873 3259
7f511862 3260 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3261 if (sbi->s_li_request != NULL) {
3262 /*
3263 * Reset timeout so it can be computed again, because
3264 * s_li_wait_mult might have changed.
3265 */
3266 sbi->s_li_request->lr_timeout = 0;
7f511862 3267 goto out;
51ce6511 3268 }
bfff6873 3269
bc98a42c 3270 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
55ff3840 3271 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3272 goto out;
bfff6873
LC		 3273
3274 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3275 if (!elr) {
3276 ret = -ENOMEM;
3277 goto out;
3278 }
bfff6873
LC		 3279
3280 if (NULL == ext4_li_info) {
3281 ret = ext4_li_info_new();
3282 if (ret)
3283 goto out;
3284 }
3285
3286 mutex_lock(&ext4_li_info->li_list_mtx);
3287 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3288 mutex_unlock(&ext4_li_info->li_list_mtx);
3289
3290 sbi->s_li_request = elr;
46e4690b
TM		 3291 /*
3292 * set elr to NULL here since it has been inserted to
3293 * the request_list and the removal and free of it is
3294 * handled by ext4_clear_request_list from now on.
3295 */
3296 elr = NULL;
bfff6873
LC		 3297
3298 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3299 ret = ext4_run_lazyinit_thread();
3300 if (ret)
3301 goto out;
3302 }
bfff6873 3303out:
beed5ecb
NK		 3304 mutex_unlock(&ext4_li_mtx);
3305 if (ret)
bfff6873 3306 kfree(elr);
bfff6873
LC		 3307 return ret;
3308}
3309
3310/*
3311 * We do not need to lock anything since this is called on
3312 * module unload.
3313 */
3314static void ext4_destroy_lazyinit_thread(void)
3315{
3316 /*
3317 * If thread exited earlier
3318 * there's nothing to be done.
3319 */
8f1f7453 3320 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3321 return;
3322
8f1f7453 3323 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3324}
3325
25ed6e8a
DW		 3326static int set_journal_csum_feature_set(struct super_block *sb)
3327{
3328 int ret = 1;
3329 int compat, incompat;
3330 struct ext4_sb_info *sbi = EXT4_SB(sb);
3331
9aa5d32b 3332 if (ext4_has_metadata_csum(sb)) {
db9ee220 3333 /* journal checksum v3 */
25ed6e8a 3334 compat = 0;
db9ee220 3335 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3336 } else {
3337 /* journal checksum v1 */
3338 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3339 incompat = 0;
3340 }
3341
feb8c6d3
DW		 3342 jbd2_journal_clear_features(sbi->s_journal,
3343 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3344 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3345 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3346 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3347 ret = jbd2_journal_set_features(sbi->s_journal,
3348 compat, 0,
3349 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3350 incompat);
3351 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3352 ret = jbd2_journal_set_features(sbi->s_journal,
3353 compat, 0,
3354 incompat);
3355 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3356 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3357 } else {
feb8c6d3
DW		 3358 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3359 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3360 }
3361
3362 return ret;
3363}
3364
952fc18e
TT		 3365/*
3366 * Note: calculating the overhead so we can be compatible with
3367 * historical BSD practice is quite difficult in the face of
3368 * clusters/bigalloc. This is because multiple metadata blocks from
3369 * different block group can end up in the same allocation cluster.
3370 * Calculating the exact overhead in the face of clustered allocation
3371 * requires either O(all block bitmaps) in memory or O(number of block
3372 * groups**2) in time. We will still calculate the superblock for
3373 * older file systems --- and if we come across with a bigalloc file
3374 * system with zero in s_overhead_clusters the estimate will be close to
3375 * correct especially for very large cluster sizes --- but for newer
3376 * file systems, it's better to calculate this figure once at mkfs
3377 * time, and store it in the superblock. If the superblock value is
3378 * present (even for non-bigalloc file systems), we will use it.
3379 */
3380static int count_overhead(struct super_block *sb, ext4_group_t grp,
3381 char *buf)
3382{
3383 struct ext4_sb_info *sbi = EXT4_SB(sb);
3384 struct ext4_group_desc *gdp;
3385 ext4_fsblk_t first_block, last_block, b;
3386 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3387 int s, j, count = 0;
3388
e2b911c5 3389 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3390 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3391 sbi->s_itb_per_group + 2);
3392
952fc18e
TT		 3393 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3394 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3395 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3396 for (i = 0; i < ngroups; i++) {
3397 gdp = ext4_get_group_desc(sb, i, NULL);
3398 b = ext4_block_bitmap(sb, gdp);
3399 if (b >= first_block && b <= last_block) {
3400 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3401 count++;
3402 }
3403 b = ext4_inode_bitmap(sb, gdp);
3404 if (b >= first_block && b <= last_block) {
3405 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3406 count++;
3407 }
3408 b = ext4_inode_table(sb, gdp);
3409 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3410 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3411 int c = EXT4_B2C(sbi, b - first_block);
3412 ext4_set_bit(c, buf);
3413 count++;
3414 }
3415 if (i != grp)
3416 continue;
3417 s = 0;
3418 if (ext4_bg_has_super(sb, grp)) {
3419 ext4_set_bit(s++, buf);
3420 count++;
3421 }
c48ae41b
TT		 3422 j = ext4_bg_num_gdb(sb, grp);
3423 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3424 ext4_error(sb, "Invalid number of block group "
3425 "descriptor blocks: %d", j);
3426 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3427 }
c48ae41b
TT		 3428 count += j;
3429 for (; j > 0; j--)
3430 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3431 }
3432 if (!count)
3433 return 0;
3434 return EXT4_CLUSTERS_PER_GROUP(sb) -
3435 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3436}
3437
3438/*
3439 * Compute the overhead and stash it in sbi->s_overhead
3440 */
3441int ext4_calculate_overhead(struct super_block *sb)
3442{
3443 struct ext4_sb_info *sbi = EXT4_SB(sb);
3444 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3445 struct inode *j_inode;
3446 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3447 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3448 ext4_fsblk_t overhead = 0;
4fdb5543 3449 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3450
952fc18e
TT		 3451 if (!buf)
3452 return -ENOMEM;
3453
3454 /*
3455 * Compute the overhead (FS structures). This is constant
3456 * for a given filesystem unless the number of block groups
3457 * changes so we cache the previous value until it does.
3458 */
3459
3460 /*
3461 * All of the blocks before first_data_block are overhead
3462 */
3463 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3464
3465 /*
3466 * Add the overhead found in each block group
3467 */
3468 for (i = 0; i < ngroups; i++) {
3469 int blks;
3470
3471 blks = count_overhead(sb, i, buf);
3472 overhead += blks;
3473 if (blks)
3474 memset(buf, 0, PAGE_SIZE);
3475 cond_resched();
3476 }
3c816ded
EW		 3477
3478 /*
3479 * Add the internal journal blocks whether the journal has been
3480 * loaded or not
3481 */
b003b524 3482 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3483 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3484 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3485 j_inode = ext4_get_journal_inode(sb, j_inum);
3486 if (j_inode) {
3487 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3488 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3489 iput(j_inode);
3490 } else {
3491 ext4_msg(sb, KERN_ERR, "can't get journal size");
3492 }
3493 }
952fc18e
TT		 3494 sbi->s_overhead = overhead;
3495 smp_wmb();
3496 free_page((unsigned long) buf);
3497 return 0;
3498}
3499
b5799018 3500static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3501{
3502 ext4_fsblk_t resv_clusters;
b5799018 3503 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3504
30fac0f7
JK		 3505 /*
3506 * There's no need to reserve anything when we aren't using extents.
3507 * The space estimates are exact, there are no unwritten extents,
3508 * hole punching doesn't need new metadata... This is needed especially
3509 * to keep ext2/3 backward compatibility.
3510 */
e2b911c5 3511 if (!ext4_has_feature_extents(sb))
b5799018 3512 return;
27dd4385
LC		 3513 /*
3514 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3515 * This should cover the situations where we can not afford to run
3516 * out of space like for example punch hole, or converting
556615dc 3517 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3518 * allocation would require 1, or 2 blocks, higher numbers are
3519 * very rare.
3520 */
b5799018
TT		 3521 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3522 sbi->s_cluster_bits);
27dd4385
LC		 3523
3524 do_div(resv_clusters, 50);
3525 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3526
b5799018 3527 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3528}
3529
2b2d6d01 3530static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3531{
5e405595 3532 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
d4c402d9 3533 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3534 struct buffer_head *bh;
617ba13b 3535 struct ext4_super_block *es = NULL;
5aee0f8a 3536 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3537 ext4_fsblk_t block;
3538 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3539 ext4_fsblk_t logical_sb_block;
ac27a0ec 3540 unsigned long offset = 0;
ac27a0ec
DK		 3541 unsigned long journal_devnum = 0;
3542 unsigned long def_mount_opts;
3543 struct inode *root;
0390131b 3544 const char *descr;
dcc7dae3 3545 int ret = -ENOMEM;
281b5995 3546 int blocksize, clustersize;
4ec11028
TT		 3547 unsigned int db_count;
3548 unsigned int i;
281b5995 3549 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3550 __u64 blocks_count;
07aa2ea1 3551 int err = 0;
b3881f74 3552 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3553 ext4_group_t first_not_zeroed;
ac27a0ec 3554
5aee0f8a
TT		 3555 if ((data && !orig_data) || !sbi)
3556 goto out_free_base;
705895b6 3557
aed9eb1b 3558 sbi->s_daxdev = dax_dev;
705895b6
PE		 3559 sbi->s_blockgroup_lock =
3560 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3561 if (!sbi->s_blockgroup_lock)
3562 goto out_free_base;
3563
ac27a0ec 3564 sb->s_fs_info = sbi;
2c0544b2 3565 sbi->s_sb = sb;
240799cd 3566 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3567 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3568 if (sb->s_bdev->bd_part)
3569 sbi->s_sectors_written_start =
dbae2c55 3570 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
ac27a0ec 3571
9f6200bb 3572 /* Cleanup superblock name */
ec3904dc 3573 strreplace(sb->s_id, '/', '!');
9f6200bb 3574
07aa2ea1 3575 /* -EINVAL is default */
dcc7dae3 3576 ret = -EINVAL;
617ba13b 3577 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3578 if (!blocksize) {
b31e1552 3579 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3580 goto out_fail;
3581 }
3582
3583 /*
617ba13b 3584 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3585 * block sizes. We need to calculate the offset from buffer start.
3586 */
617ba13b 3587 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3588 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3589 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3590 } else {
70bbb3e0 3591 logical_sb_block = sb_block;
ac27a0ec
DK		 3592 }
3593
a8ac900b 3594 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3595 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3596 goto out_fail;
3597 }
3598 /*
3599 * Note: s_es must be initialized as soon as possible because
617ba13b 3600 * some ext4 macro-instructions depend on its value
ac27a0ec 3601 */
2716b802 3602 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3603 sbi->s_es = es;
3604 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3605 if (sb->s_magic != EXT4_SUPER_MAGIC)
3606 goto cantfind_ext4;
afc32f7e 3607 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3608
feb0ab32 3609 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3610 if (ext4_has_feature_metadata_csum(sb) &&
3611 ext4_has_feature_gdt_csum(sb))
363307e6 3612 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3613 "redundant flags; please run fsck.");
3614
d25425f8
DW		 3615 /* Check for a known checksum algorithm */
3616 if (!ext4_verify_csum_type(sb, es)) {
3617 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3618 "unknown checksum algorithm.");
3619 silent = 1;
3620 goto cantfind_ext4;
3621 }
3622
0441984a 3623 /* Load the checksum driver */
a45403b5
TT		 3624 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3625 if (IS_ERR(sbi->s_chksum_driver)) {
3626 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3627 ret = PTR_ERR(sbi->s_chksum_driver);
3628 sbi->s_chksum_driver = NULL;
3629 goto failed_mount;
0441984a
DW		 3630 }
3631
a9c47317
DW		 3632 /* Check superblock checksum */
3633 if (!ext4_superblock_csum_verify(sb, es)) {
3634 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3635 "invalid superblock checksum. Run e2fsck?");
3636 silent = 1;
6a797d27 3637 ret = -EFSBADCRC;
a9c47317
DW		 3638 goto cantfind_ext4;
3639 }
3640
3641 /* Precompute checksum seed for all metadata */
e2b911c5 3642 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 3643 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 3644 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 3645 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3646 sizeof(es->s_uuid));
3647
ac27a0ec
DK		 3648 /* Set defaults before we parse the mount options */
3649 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3650 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3651 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3652 set_opt(sb, DEBUG);
87f26807 3653 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3654 set_opt(sb, GRPID);
617ba13b 3655 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3656 set_opt(sb, NO_UID32);
ea663336 3657 /* xattr user namespace & acls are now defaulted on */
ea663336 3658 set_opt(sb, XATTR_USER);
03010a33 3659#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3660 set_opt(sb, POSIX_ACL);
2e7842b8 3661#endif
98c1a759
DW		 3662 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3663 if (ext4_has_metadata_csum(sb))
3664 set_opt(sb, JOURNAL_CHECKSUM);
3665
617ba13b 3666 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3667 set_opt(sb, JOURNAL_DATA);
617ba13b 3668 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3669 set_opt(sb, ORDERED_DATA);
617ba13b 3670 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3671 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3672
3673 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3674 set_opt(sb, ERRORS_PANIC);
bb4f397a 3675 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3676 set_opt(sb, ERRORS_CONT);
bb4f397a 3677 else
fd8c37ec 3678 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3679 /* block_validity enabled by default; disable with noblock_validity */
3680 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3681 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3682 set_opt(sb, DISCARD);
ac27a0ec 3683
08cefc7a
EB		 3684 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3685 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3686 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3687 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3688 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3689
8b67f04a 3690 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3691 set_opt(sb, BARRIER);
ac27a0ec 3692
dd919b98
AK		 3693 /*
3694 * enable delayed allocation by default
3695 * Use -o nodelalloc to turn it off
3696 */
bc0b75f7 3697 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3698 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3699 set_opt(sb, DELALLOC);
dd919b98 3700
51ce6511
LC		 3701 /*
3702 * set default s_li_wait_mult for lazyinit, for the case there is
3703 * no mount option specified.
3704 */
3705 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3706
5aee0f8a
TT		 3707 if (sbi->s_es->s_mount_opts[0]) {
3708 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3709 sizeof(sbi->s_es->s_mount_opts),
3710 GFP_KERNEL);
3711 if (!s_mount_opts)
3712 goto failed_mount;
3713 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3714 &journal_ioprio, 0)) {
3715 ext4_msg(sb, KERN_WARNING,
3716 "failed to parse options in superblock: %s",
3717 s_mount_opts);
3718 }
3719 kfree(s_mount_opts);
8b67f04a 3720 }
5a916be1 3721 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3722 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3723 &journal_ioprio, 0))
ac27a0ec
DK		 3724 goto failed_mount;
3725
56889787
TT		 3726 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3727 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3728 "with data=journal disables delayed "
3729 "allocation and O_DIRECT support!\n");
3730 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3731 ext4_msg(sb, KERN_ERR, "can't mount with "
3732 "both data=journal and delalloc");
3733 goto failed_mount;
3734 }
3735 if (test_opt(sb, DIOREAD_NOLOCK)) {
3736 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3737 "both data=journal and dioread_nolock");
56889787
TT		 3738 goto failed_mount;
3739 }
923ae0ff
RZ		 3740 if (test_opt(sb, DAX)) {
3741 ext4_msg(sb, KERN_ERR, "can't mount with "
3742 "both data=journal and dax");
3743 goto failed_mount;
3744 }
73b92a2a
SK		 3745 if (ext4_has_feature_encrypt(sb)) {
3746 ext4_msg(sb, KERN_WARNING,
3747 "encrypted files will use data=ordered "
3748 "instead of data journaling mode");
3749 }
56889787
TT		 3750 if (test_opt(sb, DELALLOC))
3751 clear_opt(sb, DELALLOC);
001e4a87
TH		 3752 } else {
3753 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3754 }
3755
1751e8a6
LT		 3756 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3757 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 3758
617ba13b 3759 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3760 (ext4_has_compat_features(sb) ||
3761 ext4_has_ro_compat_features(sb) ||
3762 ext4_has_incompat_features(sb)))
b31e1552
ES		 3763 ext4_msg(sb, KERN_WARNING,
3764 "feature flags set on rev 0 fs, "
3765 "running e2fsck is recommended");
469108ff 3766
ed3654eb
TT		 3767 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3768 set_opt2(sb, HURD_COMPAT);
e2b911c5 3769 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3770 ext4_msg(sb, KERN_ERR,
3771 "The Hurd can't support 64-bit file systems");
3772 goto failed_mount;
3773 }
dec214d0
TE		 3774
3775 /*
3776 * ea_inode feature uses l_i_version field which is not
3777 * available in HURD_COMPAT mode.
3778 */
3779 if (ext4_has_feature_ea_inode(sb)) {
3780 ext4_msg(sb, KERN_ERR,
3781 "ea_inode feature is not supported for Hurd");
3782 goto failed_mount;
3783 }
ed3654eb
TT		 3784 }
3785
2035e776
TT		 3786 if (IS_EXT2_SB(sb)) {
3787 if (ext2_feature_set_ok(sb))
3788 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3789 "using the ext4 subsystem");
3790 else {
0d9366d6
ES		 3791 /*
3792 * If we're probing be silent, if this looks like
3793 * it's actually an ext[34] filesystem.
3794 */
3795 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3796 goto failed_mount;
2035e776
TT		 3797 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3798 "to feature incompatibilities");
3799 goto failed_mount;
3800 }
3801 }
3802
3803 if (IS_EXT3_SB(sb)) {
3804 if (ext3_feature_set_ok(sb))
3805 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3806 "using the ext4 subsystem");
3807 else {
0d9366d6
ES		 3808 /*
3809 * If we're probing be silent, if this looks like
3810 * it's actually an ext4 filesystem.
3811 */
3812 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3813 goto failed_mount;
2035e776
TT		 3814 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3815 "to feature incompatibilities");
3816 goto failed_mount;
3817 }
3818 }
3819
ac27a0ec
DK		 3820 /*
3821 * Check feature flags regardless of the revision level, since we
3822 * previously didn't change the revision level when setting the flags,
3823 * so there is a chance incompat flags are set on a rev 0 filesystem.
3824 */
bc98a42c 3825 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
ac27a0ec 3826 goto failed_mount;
a13fb1a4 3827
261cb20c 3828 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3829 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3830 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3831 ext4_msg(sb, KERN_ERR,
8cdf3372
TT		 3832 "Unsupported filesystem blocksize %d (%d log_block_size)",
3833 blocksize, le32_to_cpu(es->s_log_block_size));
3834 goto failed_mount;
3835 }
3836 if (le32_to_cpu(es->s_log_block_size) >
3837 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3838 ext4_msg(sb, KERN_ERR,
3839 "Invalid log block size: %u",
3840 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3841 goto failed_mount;
3842 }
bfe0a5f4
TT		 3843 if (le32_to_cpu(es->s_log_cluster_size) >
3844 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3845 ext4_msg(sb, KERN_ERR,
3846 "Invalid log cluster size: %u",
3847 le32_to_cpu(es->s_log_cluster_size));
3848 goto failed_mount;
3849 }
ac27a0ec 3850
5b9554dc
TT		 3851 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3852 ext4_msg(sb, KERN_ERR,
3853 "Number of reserved GDT blocks insanely large: %d",
3854 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3855 goto failed_mount;
3856 }
3857
923ae0ff 3858 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
559db4c6
RZ		 3859 if (ext4_has_feature_inline_data(sb)) {
3860 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3861 " that may contain inline data");
361d24d4 3862 goto failed_mount;
559db4c6 3863 }
80660f20 3864 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
24f3478d 3865 ext4_msg(sb, KERN_ERR,
361d24d4
ES		 3866 "DAX unsupported by block device.");
3867 goto failed_mount;
24f3478d 3868 }
923ae0ff
RZ		 3869 }
3870
e2b911c5 3871 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3872 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3873 es->s_encryption_level);
3874 goto failed_mount;
3875 }
3876
ac27a0ec 3877 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3878 /* Validate the filesystem blocksize */
3879 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3880 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3881 blocksize);
ac27a0ec
DK		 3882 goto failed_mount;
3883 }
3884
2b2d6d01 3885 brelse(bh);
70bbb3e0
AM		 3886 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3887 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3888 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3889 if (!bh) {
b31e1552
ES		 3890 ext4_msg(sb, KERN_ERR,
3891 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3892 goto failed_mount;
3893 }
2716b802 3894 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3895 sbi->s_es = es;
617ba13b 3896 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3897 ext4_msg(sb, KERN_ERR,
3898 "Magic mismatch, very weird!");
ac27a0ec
DK		 3899 goto failed_mount;
3900 }
3901 }
3902
e2b911c5 3903 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3904 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3905 has_huge_files);
3906 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3907
617ba13b
MC		 3908 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3909 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3910 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3911 } else {
3912 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3913 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
c37e9e01
TT		 3914 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3915 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3916 sbi->s_first_ino);
3917 goto failed_mount;
3918 }
617ba13b 3919 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3920 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3921 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3922 ext4_msg(sb, KERN_ERR,
3923 "unsupported inode size: %d",
2b2d6d01 3924 sbi->s_inode_size);
ac27a0ec
DK		 3925 goto failed_mount;
3926 }
ef7f3835
KS		 3927 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3928 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3929 }
0b8e58a1 3930
0d1ee42f 3931 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3932 if (ext4_has_feature_64bit(sb)) {
8fadc143 3933 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3934 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3935 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3936 ext4_msg(sb, KERN_ERR,
3937 "unsupported descriptor size %lu",
0d1ee42f
AR		 3938 sbi->s_desc_size);
3939 goto failed_mount;
3940 }
3941 } else
3942 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3943
ac27a0ec 3944 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3945 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 3946
617ba13b 3947 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3948 if (sbi->s_inodes_per_block == 0)
617ba13b 3949 goto cantfind_ext4;
cd6bb35b
TT		 3950 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3951 sbi->s_inodes_per_group > blocksize * 8) {
3952 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3953 sbi->s_blocks_per_group);
3954 goto failed_mount;
3955 }
ac27a0ec
DK		 3956 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3957 sbi->s_inodes_per_block;
0d1ee42f 3958 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3959 sbi->s_sbh = bh;
3960 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3961 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3962 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3963
2b2d6d01 3964 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3965 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3966 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3967 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3968 i = le32_to_cpu(es->s_flags);
3969 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3970 sbi->s_hash_unsigned = 3;
3971 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3972#ifdef __CHAR_UNSIGNED__
bc98a42c 3973 if (!sb_rdonly(sb))
23301410
TT		 3974 es->s_flags |=
3975 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3976 sbi->s_hash_unsigned = 3;
f99b2589 3977#else
bc98a42c 3978 if (!sb_rdonly(sb))
23301410
TT		 3979 es->s_flags |=
3980 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3981#endif
23301410 3982 }
f99b2589 3983 }
ac27a0ec 3984
281b5995
TT		 3985 /* Handle clustersize */
3986 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3987 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3988 if (has_bigalloc) {
3989 if (clustersize < blocksize) {
3990 ext4_msg(sb, KERN_ERR,
3991 "cluster size (%d) smaller than "
3992 "block size (%d)", clustersize, blocksize);
3993 goto failed_mount;
3994 }
3995 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3996 le32_to_cpu(es->s_log_block_size);
3997 sbi->s_clusters_per_group =
3998 le32_to_cpu(es->s_clusters_per_group);
3999 if (sbi->s_clusters_per_group > blocksize * 8) {
4000 ext4_msg(sb, KERN_ERR,
4001 "#clusters per group too big: %lu",
4002 sbi->s_clusters_per_group);
4003 goto failed_mount;
4004 }
4005 if (sbi->s_blocks_per_group !=
4006 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4007 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4008 "clusters per group (%lu) inconsistent",
4009 sbi->s_blocks_per_group,
4010 sbi->s_clusters_per_group);
4011 goto failed_mount;
4012 }
4013 } else {
4014 if (clustersize != blocksize) {
bfe0a5f4
TT		 4015 ext4_msg(sb, KERN_ERR,
4016 "fragment/cluster size (%d) != "
4017 "block size (%d)", clustersize, blocksize);
4018 goto failed_mount;
281b5995
TT		 4019 }
4020 if (sbi->s_blocks_per_group > blocksize * 8) {
4021 ext4_msg(sb, KERN_ERR,
4022 "#blocks per group too big: %lu",
4023 sbi->s_blocks_per_group);
4024 goto failed_mount;
4025 }
4026 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4027 sbi->s_cluster_bits = 0;
ac27a0ec 4028 }
281b5995
TT		 4029 sbi->s_cluster_ratio = clustersize / blocksize;
4030
960fd856
TT		 4031 /* Do we have standard group size of clustersize * 8 blocks ? */
4032 if (sbi->s_blocks_per_group == clustersize << 3)
4033 set_opt2(sb, STD_GROUP_SIZE);
4034
bf43d84b
ES		 4035 /*
4036 * Test whether we have more sectors than will fit in sector_t,
4037 * and whether the max offset is addressable by the page cache.
4038 */
5a9ae68a 4039 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 4040 ext4_blocks_count(es));
5a9ae68a 4041 if (err) {
b31e1552 4042 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 4043 " too large to mount safely on this system");
ac27a0ec
DK		 4044 goto failed_mount;
4045 }
4046
617ba13b
MC		 4047 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4048 goto cantfind_ext4;
e7c95593 4049
0f2ddca6
FTN		 4050 /* check blocks count against device size */
4051 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4052 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 4053 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4054 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 4055 ext4_blocks_count(es), blocks_count);
4056 goto failed_mount;
4057 }
4058
0b8e58a1
AD		 4059 /*
4060 * It makes no sense for the first data block to be beyond the end
4061 * of the filesystem.
4062 */
4063 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 4064 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 4065 "block %u is beyond end of filesystem (%llu)",
4066 le32_to_cpu(es->s_first_data_block),
4067 ext4_blocks_count(es));
e7c95593
ES		 4068 goto failed_mount;
4069 }
bfe0a5f4
TT		 4070 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4071 (sbi->s_cluster_ratio == 1)) {
4072 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4073 "block is 0 with a 1k block and cluster size");
4074 goto failed_mount;
4075 }
4076
bd81d8ee
LV		 4077 blocks_count = (ext4_blocks_count(es) -
4078 le32_to_cpu(es->s_first_data_block) +
4079 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4080 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 4081 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 4082 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 4083 "(block count %llu, first data block %u, "
b31e1552 4084 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 4085 ext4_blocks_count(es),
4086 le32_to_cpu(es->s_first_data_block),
4087 EXT4_BLOCKS_PER_GROUP(sb));
4088 goto failed_mount;
4089 }
bd81d8ee 4090 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 4091 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4092 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
9e463084
TT		 4093 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4094 le32_to_cpu(es->s_inodes_count)) {
4095 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4096 le32_to_cpu(es->s_inodes_count),
4097 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4098 ret = -EINVAL;
4099 goto failed_mount;
4100 }
617ba13b
MC		 4101 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4102 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 4103 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 4104 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 4105 ext4_msg(sb, KERN_WARNING,
4106 "first meta block group too large: %u "
4107 "(group descriptor block count %u)",
4108 le32_to_cpu(es->s_first_meta_bg), db_count);
4109 goto failed_mount;
4110 }
4111 }
344476e1
KC		 4112 sbi->s_group_desc = kvmalloc_array(db_count,
4113 sizeof(struct buffer_head *),
4114 GFP_KERNEL);
ac27a0ec 4115 if (sbi->s_group_desc == NULL) {
b31e1552 4116 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 4117 ret = -ENOMEM;
ac27a0ec
DK		 4118 goto failed_mount;
4119 }
4120
705895b6 4121 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 4122
85c8f176
AP		 4123 /* Pre-read the descriptors into the buffer cache */
4124 for (i = 0; i < db_count; i++) {
4125 block = descriptor_loc(sb, logical_sb_block, i);
4126 sb_breadahead(sb, block);
4127 }
4128
ac27a0ec 4129 for (i = 0; i < db_count; i++) {
70bbb3e0 4130 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 4131 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 4132 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 4133 ext4_msg(sb, KERN_ERR,
4134 "can't read group descriptor %d", i);
ac27a0ec
DK		 4135 db_count = i;
4136 goto failed_mount2;
4137 }
4138 }
44de022c 4139 sbi->s_gdb_count = db_count;
829fa70d 4140 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 4141 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 4142 ret = -EFSCORRUPTED;
f9ae9cf5 4143 goto failed_mount2;
ac27a0ec 4144 }
772cb7c8 4145
235699a8 4146 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
04496411 4147
a75ae78f 4148 /* Register extent status tree shrinker */
eb68d0e2 4149 if (ext4_es_register_shrinker(sbi))
ce7e010a 4150 goto failed_mount3;
ce7e010a 4151
c9de560d 4152 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 4153 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 4154
f9ae9cf5
TT		 4155 /*
4156 * set up enough so that it can read an inode
4157 */
f6e63f90 4158 sb->s_op = &ext4_sops;
617ba13b
MC		 4159 sb->s_export_op = &ext4_export_ops;
4160 sb->s_xattr = ext4_xattr_handlers;
643fa961 4161#ifdef CONFIG_FS_ENCRYPTION
a7550b30 4162 sb->s_cop = &ext4_cryptops;
ffcc4182 4163#endif
ac27a0ec 4164#ifdef CONFIG_QUOTA
617ba13b 4165 sb->dq_op = &ext4_quota_operations;
e2b911c5 4166 if (ext4_has_feature_quota(sb))
1fa5efe3 4167 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 4168 else
4169 sb->s_qcop = &ext4_qctl_operations;
689c958c 4170 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 4171#endif
85787090 4172 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 4173
ac27a0ec 4174 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4175 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 4176
4177 sb->s_root = NULL;
4178
4179 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 4180 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 4181
bc98a42c 4182 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 4183 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4184 goto failed_mount3a;
c5e06d10 4185
ac27a0ec
DK		 4186 /*
4187 * The first inode we look at is the journal inode. Don't try
4188 * root first: it may be modified in the journal!
4189 */
e2b911c5 4190 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 4191 err = ext4_load_journal(sb, es, journal_devnum);
4192 if (err)
50460fe8 4193 goto failed_mount3a;
bc98a42c 4194 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 4195 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4196 ext4_msg(sb, KERN_ERR, "required journal recovery "
4197 "suppressed and not mounted read-only");
744692dc 4198 goto failed_mount_wq;
ac27a0ec 4199 } else {
1e381f60
DM		 4200 /* Nojournal mode, all journal mount options are illegal */
4201 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4202 ext4_msg(sb, KERN_ERR, "can't mount with "
4203 "journal_checksum, fs mounted w/o journal");
4204 goto failed_mount_wq;
4205 }
4206 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4207 ext4_msg(sb, KERN_ERR, "can't mount with "
4208 "journal_async_commit, fs mounted w/o journal");
4209 goto failed_mount_wq;
4210 }
4211 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4212 ext4_msg(sb, KERN_ERR, "can't mount with "
4213 "commit=%lu, fs mounted w/o journal",
4214 sbi->s_commit_interval / HZ);
4215 goto failed_mount_wq;
4216 }
4217 if (EXT4_MOUNT_DATA_FLAGS &
4218 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4219 ext4_msg(sb, KERN_ERR, "can't mount with "
4220 "data=, fs mounted w/o journal");
4221 goto failed_mount_wq;
4222 }
4223 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4224 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4225 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4226 sbi->s_journal = NULL;
4227 needs_recovery = 0;
4228 goto no_journal;
ac27a0ec
DK		 4229 }
4230
e2b911c5 4231 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4232 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4233 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4234 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4235 goto failed_mount_wq;
eb40a09c
JS		 4236 }
4237
25ed6e8a
DW		 4238 if (!set_journal_csum_feature_set(sb)) {
4239 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4240 "feature set");
4241 goto failed_mount_wq;
d4da6c9c 4242 }
818d276c 4243
ac27a0ec
DK		 4244 /* We have now updated the journal if required, so we can
4245 * validate the data journaling mode. */
4246 switch (test_opt(sb, DATA_FLAGS)) {
4247 case 0:
4248 /* No mode set, assume a default based on the journal
63f57933
AM		 4249 * capabilities: ORDERED_DATA if the journal can
4250 * cope, else JOURNAL_DATA
4251 */
dab291af 4252 if (jbd2_journal_check_available_features
27f394a7 4253 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
fd8c37ec 4254 set_opt(sb, ORDERED_DATA);
27f394a7
TN		 4255 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4256 } else {
fd8c37ec 4257 set_opt(sb, JOURNAL_DATA);
27f394a7
TN		 4258 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4259 }
ac27a0ec
DK		 4260 break;
4261
617ba13b
MC		 4262 case EXT4_MOUNT_ORDERED_DATA:
4263 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4264 if (!jbd2_journal_check_available_features
4265 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4266 ext4_msg(sb, KERN_ERR, "Journal does not support "
4267 "requested data journaling mode");
744692dc 4268 goto failed_mount_wq;
ac27a0ec
DK		 4269 }
4270 default:
4271 break;
4272 }
ab04df78
JK		 4273
4274 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4275 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4276 ext4_msg(sb, KERN_ERR, "can't mount with "
4277 "journal_async_commit in data=ordered mode");
4278 goto failed_mount_wq;
4279 }
4280
b3881f74 4281 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4282
18aadd47
BJ		 4283 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4284
ce7e010a 4285no_journal:
cdb7ee4c
TE		 4286 if (!test_opt(sb, NO_MBCACHE)) {
4287 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4288 if (!sbi->s_ea_block_cache) {
dec214d0 4289 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4290 "Failed to create ea_block_cache");
dec214d0
TE		 4291 goto failed_mount_wq;
4292 }
cdb7ee4c
TE		 4293
4294 if (ext4_has_feature_ea_inode(sb)) {
4295 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4296 if (!sbi->s_ea_inode_cache) {
4297 ext4_msg(sb, KERN_ERR,
4298 "Failed to create ea_inode_cache");
4299 goto failed_mount_wq;
4300 }
4301 }
9c191f70
M		 4302 }
4303
e2b911c5 4304 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 4305 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 4306 ext4_msg(sb, KERN_ERR,
4307 "Unsupported blocksize for fs encryption");
4308 goto failed_mount_wq;
4309 }
4310
bc98a42c 4311 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
e2b911c5
DW		 4312 !ext4_has_feature_encrypt(sb)) {
4313 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4314 ext4_commit_super(sb, 1);
4315 }
4316
952fc18e
TT		 4317 /*
4318 * Get the # of file system overhead blocks from the
4319 * superblock if present.
4320 */
4321 if (es->s_overhead_clusters)
4322 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4323 else {
07aa2ea1
LC		 4324 err = ext4_calculate_overhead(sb);
4325 if (err)
952fc18e
TT		 4326 goto failed_mount_wq;
4327 }
4328
fd89d5f2
TH		 4329 /*
4330 * The maximum number of concurrent works can be high and
4331 * concurrency isn't really necessary. Limit it to 1.
4332 */
2e8fa54e
JK		 4333 EXT4_SB(sb)->rsv_conversion_wq =
4334 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4335 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4336 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4337 ret = -ENOMEM;
2e8fa54e
JK		 4338 goto failed_mount4;
4339 }
4340
ac27a0ec 4341 /*
dab291af 4342 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4343 * so we can safely mount the rest of the filesystem now.
4344 */
4345
8a363970 4346 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
1d1fe1ee 4347 if (IS_ERR(root)) {
b31e1552 4348 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4349 ret = PTR_ERR(root);
32a9bb57 4350 root = NULL;
ac27a0ec
DK		 4351 goto failed_mount4;
4352 }
4353 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4354 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4355 iput(root);
ac27a0ec
DK		 4356 goto failed_mount4;
4357 }
48fde701 4358 sb->s_root = d_make_root(root);
1d1fe1ee 4359 if (!sb->s_root) {
b31e1552 4360 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4361 ret = -ENOMEM;
4362 goto failed_mount4;
4363 }
ac27a0ec 4364
c89128a0
JK		 4365 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4366 if (ret == -EROFS) {
1751e8a6 4367 sb->s_flags |= SB_RDONLY;
c89128a0
JK		 4368 ret = 0;
4369 } else if (ret)
4370 goto failed_mount4a;
ef7f3835
KS		 4371
4372 /* determine the minimum size of new large inodes, if present */
670e9875
TT		 4373 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4374 sbi->s_want_extra_isize == 0) {
ef7f3835
KS		 4375 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4376 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4377 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4378 if (sbi->s_want_extra_isize <
4379 le16_to_cpu(es->s_want_extra_isize))
4380 sbi->s_want_extra_isize =
4381 le16_to_cpu(es->s_want_extra_isize);
4382 if (sbi->s_want_extra_isize <
4383 le16_to_cpu(es->s_min_extra_isize))
4384 sbi->s_want_extra_isize =
4385 le16_to_cpu(es->s_min_extra_isize);
4386 }
4387 }
4388 /* Check if enough inode space is available */
4389 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4390 sbi->s_inode_size) {
4391 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4392 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4393 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4394 "available");
ef7f3835
KS		 4395 }
4396
b5799018 4397 ext4_set_resv_clusters(sb);
27dd4385 4398
6fd058f7
TT		 4399 err = ext4_setup_system_zone(sb);
4400 if (err) {
b31e1552 4401 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4402 "zone (%d)", err);
f9ae9cf5
TT		 4403 goto failed_mount4a;
4404 }
4405
4406 ext4_ext_init(sb);
4407 err = ext4_mb_init(sb);
4408 if (err) {
4409 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4410 err);
dcf2d804 4411 goto failed_mount5;
c2774d84
AK		 4412 }
4413
d5e03cbb
TT		 4414 block = ext4_count_free_clusters(sb);
4415 ext4_free_blocks_count_set(sbi->s_es,
4416 EXT4_C2B(sbi, block));
4274f516 4417 ext4_superblock_csum_set(sb);
908c7f19
TH		 4418 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4419 GFP_KERNEL);
d5e03cbb
TT		 4420 if (!err) {
4421 unsigned long freei = ext4_count_free_inodes(sb);
4422 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4274f516 4423 ext4_superblock_csum_set(sb);
908c7f19
TH		 4424 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4425 GFP_KERNEL);
d5e03cbb
TT		 4426 }
4427 if (!err)
4428 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4429 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4430 if (!err)
908c7f19
TH		 4431 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4432 GFP_KERNEL);
c8585c6f
DJ		 4433 if (!err)
4434 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4435
d5e03cbb
TT		 4436 if (err) {
4437 ext4_msg(sb, KERN_ERR, "insufficient memory");
4438 goto failed_mount6;
4439 }
4440
e2b911c5 4441 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4442 if (!ext4_fill_flex_info(sb)) {
4443 ext4_msg(sb, KERN_ERR,
4444 "unable to initialize "
4445 "flex_bg meta info!");
4446 goto failed_mount6;
4447 }
4448
bfff6873
LC		 4449 err = ext4_register_li_request(sb, first_not_zeroed);
4450 if (err)
dcf2d804 4451 goto failed_mount6;
bfff6873 4452
b5799018 4453 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4454 if (err)
4455 goto failed_mount7;
3197ebdb 4456
9b2ff357
JK		 4457#ifdef CONFIG_QUOTA
4458 /* Enable quota usage during mount. */
bc98a42c 4459 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 4460 err = ext4_enable_quotas(sb);
4461 if (err)
4462 goto failed_mount8;
4463 }
4464#endif /* CONFIG_QUOTA */
4465
617ba13b
MC		 4466 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4467 ext4_orphan_cleanup(sb, es);
4468 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4469 if (needs_recovery) {
b31e1552 4470 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4471 ext4_mark_recovery_complete(sb, es);
4472 }
4473 if (EXT4_SB(sb)->s_journal) {
4474 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4475 descr = " journalled data mode";
4476 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4477 descr = " ordered data mode";
4478 else
4479 descr = " writeback data mode";
4480 } else
4481 descr = "out journal";
4482
79add3a3
LC		 4483 if (test_opt(sb, DISCARD)) {
4484 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4485 if (!blk_queue_discard(q))
4486 ext4_msg(sb, KERN_WARNING,
4487 "mounting with \"discard\" option, but "
4488 "the device does not support discard");
4489 }
4490
e294a537
TT		 4491 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4492 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4493 "Opts: %.*s%s%s", descr,
4494 (int) sizeof(sbi->s_es->s_mount_opts),
4495 sbi->s_es->s_mount_opts,
e294a537 4496 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4497
66e61a9e
TT		 4498 if (es->s_error_count)
4499 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4500
efbed4dc
TT		 4501 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4502 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4503 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4504 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4505
d4c402d9 4506 kfree(orig_data);
ac27a0ec
DK		 4507 return 0;
4508
617ba13b 4509cantfind_ext4:
ac27a0ec 4510 if (!silent)
b31e1552 4511 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4512 goto failed_mount;
4513
72ba7450
TT		 4514#ifdef CONFIG_QUOTA
4515failed_mount8:
ebd173be 4516 ext4_unregister_sysfs(sb);
72ba7450 4517#endif
dcf2d804
TM		 4518failed_mount7:
4519 ext4_unregister_li_request(sb);
4520failed_mount6:
f9ae9cf5 4521 ext4_mb_release(sb);
d5e03cbb 4522 if (sbi->s_flex_groups)
b93b41d4 4523 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4524 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4525 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4526 percpu_counter_destroy(&sbi->s_dirs_counter);
4527 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
af18e35b 4528 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
00764937 4529failed_mount5:
f9ae9cf5
TT		 4530 ext4_ext_release(sb);
4531 ext4_release_system_zone(sb);
4532failed_mount4a:
94bf608a 4533 dput(sb->s_root);
32a9bb57 4534 sb->s_root = NULL;
94bf608a 4535failed_mount4:
b31e1552 4536 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4537 if (EXT4_SB(sb)->rsv_conversion_wq)
4538 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4539failed_mount_wq:
50c15df6
CX		 4540 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4541 sbi->s_ea_inode_cache = NULL;
4542
4543 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4544 sbi->s_ea_block_cache = NULL;
4545
0390131b
FM		 4546 if (sbi->s_journal) {
4547 jbd2_journal_destroy(sbi->s_journal);
4548 sbi->s_journal = NULL;
4549 }
50460fe8 4550failed_mount3a:
d3922a77 4551 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4552failed_mount3:
9105bb14 4553 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4554 if (sbi->s_mmp_tsk)
4555 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4556failed_mount2:
4557 for (i = 0; i < db_count; i++)
4558 brelse(sbi->s_group_desc[i]);
b93b41d4 4559 kvfree(sbi->s_group_desc);
ac27a0ec 4560failed_mount:
0441984a
DW		 4561 if (sbi->s_chksum_driver)
4562 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4563#ifdef CONFIG_QUOTA
a2d4a646 4564 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4565 kfree(sbi->s_qf_names[i]);
4566#endif
617ba13b 4567 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4568 brelse(bh);
4569out_fail:
4570 sb->s_fs_info = NULL;
f6830165 4571 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4572out_free_base:
ac27a0ec 4573 kfree(sbi);
d4c402d9 4574 kfree(orig_data);
5e405595 4575 fs_put_dax(dax_dev);
07aa2ea1 4576 return err ? err : ret;
ac27a0ec
DK		 4577}
4578
4579/*
4580 * Setup any per-fs journal parameters now. We'll do this both on
4581 * initial mount, once the journal has been initialised but before we've
4582 * done any recovery; and again on any subsequent remount.
4583 */
617ba13b 4584static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4585{
617ba13b 4586 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4587
30773840
TT		 4588 journal->j_commit_interval = sbi->s_commit_interval;
4589 journal->j_min_batch_time = sbi->s_min_batch_time;
4590 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4591
a931da6a 4592 write_lock(&journal->j_state_lock);
ac27a0ec 4593 if (test_opt(sb, BARRIER))
dab291af 4594 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4595 else
dab291af 4596 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4597 if (test_opt(sb, DATA_ERR_ABORT))
4598 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4599 else
4600 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4601 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4602}
4603
c6cb7e77
EW		 4604static struct inode *ext4_get_journal_inode(struct super_block *sb,
4605 unsigned int journal_inum)
ac27a0ec
DK		 4606{
4607 struct inode *journal_inode;
ac27a0ec 4608
c6cb7e77
EW		 4609 /*
4610 * Test for the existence of a valid inode on disk. Bad things
4611 * happen if we iget() an unused inode, as the subsequent iput()
4612 * will try to delete it.
4613 */
8a363970 4614 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
1d1fe1ee 4615 if (IS_ERR(journal_inode)) {
b31e1552 4616 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4617 return NULL;
4618 }
4619 if (!journal_inode->i_nlink) {
4620 make_bad_inode(journal_inode);
4621 iput(journal_inode);
b31e1552 4622 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4623 return NULL;
4624 }
4625
e5f8eab8 4626 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4627 journal_inode, journal_inode->i_size);
1d1fe1ee 4628 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4629 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4630 iput(journal_inode);
4631 return NULL;
4632 }
c6cb7e77
EW		 4633 return journal_inode;
4634}
4635
4636static journal_t *ext4_get_journal(struct super_block *sb,
4637 unsigned int journal_inum)
4638{
4639 struct inode *journal_inode;
4640 journal_t *journal;
4641
4642 BUG_ON(!ext4_has_feature_journal(sb));
4643
4644 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4645 if (!journal_inode)
4646 return NULL;
ac27a0ec 4647
dab291af 4648 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4649 if (!journal) {
b31e1552 4650 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4651 iput(journal_inode);
4652 return NULL;
4653 }
4654 journal->j_private = sb;
617ba13b 4655 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4656 return journal;
4657}
4658
617ba13b 4659static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4660 dev_t j_dev)
4661{
2b2d6d01 4662 struct buffer_head *bh;
ac27a0ec 4663 journal_t *journal;
617ba13b
MC		 4664 ext4_fsblk_t start;
4665 ext4_fsblk_t len;
ac27a0ec 4666 int hblock, blocksize;
617ba13b 4667 ext4_fsblk_t sb_block;
ac27a0ec 4668 unsigned long offset;
2b2d6d01 4669 struct ext4_super_block *es;
ac27a0ec
DK		 4670 struct block_device *bdev;
4671
e2b911c5 4672 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4673
b31e1552 4674 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4675 if (bdev == NULL)
4676 return NULL;
4677
ac27a0ec 4678 blocksize = sb->s_blocksize;
e1defc4f 4679 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4680 if (blocksize < hblock) {
b31e1552
ES		 4681 ext4_msg(sb, KERN_ERR,
4682 "blocksize too small for journal device");
ac27a0ec
DK		 4683 goto out_bdev;
4684 }
4685
617ba13b
MC		 4686 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4687 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4688 set_blocksize(bdev, blocksize);
4689 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4690 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4691 "external journal");
ac27a0ec
DK		 4692 goto out_bdev;
4693 }
4694
2716b802 4695 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4696 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4697 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4698 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4699 ext4_msg(sb, KERN_ERR, "external journal has "
4700 "bad superblock");
ac27a0ec
DK		 4701 brelse(bh);
4702 goto out_bdev;
4703 }
4704
df4763be
DW		 4705 if ((le32_to_cpu(es->s_feature_ro_compat) &
4706 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4707 es->s_checksum != ext4_superblock_csum(sb, es)) {
4708 ext4_msg(sb, KERN_ERR, "external journal has "
4709 "corrupt superblock");
4710 brelse(bh);
4711 goto out_bdev;
4712 }
4713
617ba13b 4714 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4715 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4716 brelse(bh);
4717 goto out_bdev;
4718 }
4719
bd81d8ee 4720 len = ext4_blocks_count(es);
ac27a0ec
DK		 4721 start = sb_block + 1;
4722 brelse(bh); /* we're done with the superblock */
4723
dab291af 4724 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4725 start, len, blocksize);
4726 if (!journal) {
b31e1552 4727 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4728 goto out_bdev;
4729 }
4730 journal->j_private = sb;
dfec8a14 4731 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4732 wait_on_buffer(journal->j_sb_buffer);
4733 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4734 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4735 goto out_journal;
4736 }
4737 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4738 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4739 "user (unsupported) - %d",
ac27a0ec
DK		 4740 be32_to_cpu(journal->j_superblock->s_nr_users));
4741 goto out_journal;
4742 }
617ba13b
MC		 4743 EXT4_SB(sb)->journal_bdev = bdev;
4744 ext4_init_journal_params(sb, journal);
ac27a0ec 4745 return journal;
0b8e58a1 4746
ac27a0ec 4747out_journal:
dab291af 4748 jbd2_journal_destroy(journal);
ac27a0ec 4749out_bdev:
617ba13b 4750 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4751 return NULL;
4752}
4753
617ba13b
MC		 4754static int ext4_load_journal(struct super_block *sb,
4755 struct ext4_super_block *es,
ac27a0ec
DK		 4756 unsigned long journal_devnum)
4757{
4758 journal_t *journal;
4759 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4760 dev_t journal_dev;
4761 int err = 0;
4762 int really_read_only;
4763
e2b911c5 4764 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4765
ac27a0ec
DK		 4766 if (journal_devnum &&
4767 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4768 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4769 "numbers have changed");
ac27a0ec
DK		 4770 journal_dev = new_decode_dev(journal_devnum);
4771 } else
4772 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4773
4774 really_read_only = bdev_read_only(sb->s_bdev);
4775
4776 /*
4777 * Are we loading a blank journal or performing recovery after a
4778 * crash? For recovery, we need to check in advance whether we
4779 * can get read-write access to the device.
4780 */
e2b911c5 4781 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 4782 if (sb_rdonly(sb)) {
b31e1552
ES		 4783 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4784 "required on readonly filesystem");
ac27a0ec 4785 if (really_read_only) {
b31e1552 4786 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 4787 "unavailable, cannot proceed "
4788 "(try mounting with noload)");
ac27a0ec
DK		 4789 return -EROFS;
4790 }
b31e1552
ES		 4791 ext4_msg(sb, KERN_INFO, "write access will "
4792 "be enabled during recovery");
ac27a0ec
DK		 4793 }
4794 }
4795
4796 if (journal_inum && journal_dev) {
b31e1552
ES		 4797 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4798 "and inode journals!");
ac27a0ec
DK		 4799 return -EINVAL;
4800 }
4801
4802 if (journal_inum) {
617ba13b 4803 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4804 return -EINVAL;
4805 } else {
617ba13b 4806 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4807 return -EINVAL;
4808 }
4809
90576c0b 4810 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4811 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4812
e2b911c5 4813 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4814 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4815 if (!err) {
4816 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4817 if (save)
4818 memcpy(save, ((char *) es) +
4819 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4820 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4821 if (save)
4822 memcpy(((char *) es) + EXT4_S_ERR_START,
4823 save, EXT4_S_ERR_LEN);
4824 kfree(save);
4825 }
ac27a0ec
DK		 4826
4827 if (err) {
b31e1552 4828 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4829 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4830 return err;
4831 }
4832
617ba13b
MC		 4833 EXT4_SB(sb)->s_journal = journal;
4834 ext4_clear_journal_err(sb, es);
ac27a0ec 4835
c41303ce 4836 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4837 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4838 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4839
4840 /* Make sure we flush the recovery flag to disk. */
e2d67052 4841 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4842 }
4843
4844 return 0;
4845}
4846
e2d67052 4847static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4848{
e2d67052 4849 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4850 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4851 int error = 0;
ac27a0ec 4852
bdfe0cbd 4853 if (!sbh || block_device_ejected(sb))
c4be0c1d 4854 return error;
a17712c8
JD		 4855
4856 /*
4857 * The superblock bh should be mapped, but it might not be if the
4858 * device was hot-removed. Not much we can do but fail the I/O.
4859 */
4860 if (!buffer_mapped(sbh))
4861 return error;
4862
71290b36
TT		 4863 /*
4864 * If the file system is mounted read-only, don't update the
4865 * superblock write time. This avoids updating the superblock
4866 * write time when we are mounting the root file system
4867 * read/only but we need to replay the journal; at that point,
4868 * for people who are east of GMT and who make their clock
4869 * tick in localtime for Windows bug-for-bug compatibility,
4870 * the clock is set in the future, and this will cause e2fsck
4871 * to complain and force a full file system check.
4872 */
1751e8a6 4873 if (!(sb->s_flags & SB_RDONLY))
6a0678a7 4874 ext4_update_tstamp(es, s_wtime);
f613dfcb
TT		 4875 if (sb->s_bdev->bd_part)
4876 es->s_kbytes_written =
4877 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
dbae2c55
MC		 4878 ((part_stat_read(sb->s_bdev->bd_part,
4879 sectors[STAT_WRITE]) -
afc32f7e 4880 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4881 else
4882 es->s_kbytes_written =
4883 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4884 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4885 ext4_free_blocks_count_set(es,
57042651
TT		 4886 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4887 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4888 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4889 es->s_free_inodes_count =
4890 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4891 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4892 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4893 ext4_superblock_csum_set(sb);
1566a48a
TT		 4894 if (sync)
4895 lock_buffer(sbh);
e8680786 4896 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4743f839
PKS		 4897 /*
4898 * Oh, dear. A previous attempt to write the
4899 * superblock failed. This could happen because the
4900 * USB device was yanked out. Or it could happen to
4901 * be a transient write error and maybe the block will
4902 * be remapped. Nothing we can do but to retry the
4903 * write and hope for the best.
4904 */
4905 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4906 "superblock detected");
4907 clear_buffer_write_io_error(sbh);
4908 set_buffer_uptodate(sbh);
4909 }
ac27a0ec 4910 mark_buffer_dirty(sbh);
914258bf 4911 if (sync) {
1566a48a 4912 unlock_buffer(sbh);
564bc402 4913 error = __sync_dirty_buffer(sbh,
00473374 4914 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
c89128a0 4915 if (buffer_write_io_error(sbh)) {
b31e1552
ES		 4916 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4917 "superblock");
914258bf
TT		 4918 clear_buffer_write_io_error(sbh);
4919 set_buffer_uptodate(sbh);
4920 }
4921 }
c4be0c1d 4922 return error;
ac27a0ec
DK		 4923}
4924
ac27a0ec
DK		 4925/*
4926 * Have we just finished recovery? If so, and if we are mounting (or
4927 * remounting) the filesystem readonly, then we will end up with a
4928 * consistent fs on disk. Record that fact.
4929 */
2b2d6d01
TT		 4930static void ext4_mark_recovery_complete(struct super_block *sb,
4931 struct ext4_super_block *es)
ac27a0ec 4932{
617ba13b 4933 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4934
e2b911c5 4935 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4936 BUG_ON(journal != NULL);
4937 return;
4938 }
dab291af 4939 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4940 if (jbd2_journal_flush(journal) < 0)
4941 goto out;
4942
bc98a42c 4943 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
e2b911c5 4944 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4945 ext4_commit_super(sb, 1);
ac27a0ec 4946 }
7ffe1ea8
HK		 4947
4948out:
dab291af 4949 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4950}
4951
4952/*
4953 * If we are mounting (or read-write remounting) a filesystem whose journal
4954 * has recorded an error from a previous lifetime, move that error to the
4955 * main filesystem now.
4956 */
2b2d6d01
TT		 4957static void ext4_clear_journal_err(struct super_block *sb,
4958 struct ext4_super_block *es)
ac27a0ec
DK		 4959{
4960 journal_t *journal;
4961 int j_errno;
4962 const char *errstr;
4963
e2b911c5 4964 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4965
617ba13b 4966 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4967
4968 /*
4969 * Now check for any error status which may have been recorded in the
617ba13b 4970 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4971 */
4972
dab291af 4973 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4974 if (j_errno) {
4975 char nbuf[16];
4976
617ba13b 4977 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4978 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4979 "from previous mount: %s", errstr);
12062ddd 4980 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4981
617ba13b
MC		 4982 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4983 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4984 ext4_commit_super(sb, 1);
ac27a0ec 4985
dab291af 4986 jbd2_journal_clear_err(journal);
d796c52e 4987 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4988 }
4989}
4990
4991/*
4992 * Force the running and committing transactions to commit,
4993 * and wait on the commit.
4994 */
617ba13b 4995int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4996{
4997 journal_t *journal;
ac27a0ec 4998
bc98a42c 4999 if (sb_rdonly(sb))
ac27a0ec
DK		 5000 return 0;
5001
617ba13b 5002 journal = EXT4_SB(sb)->s_journal;
b1deefc9 5003 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 5004}
5005
617ba13b 5006static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 5007{
14ce0cb4 5008 int ret = 0;
9eddacf9 5009 tid_t target;
06a407f1 5010 bool needs_barrier = false;
8d5d02e6 5011 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5012
49598e04 5013 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 5014 return 0;
5015
9bffad1e 5016 trace_ext4_sync_fs(sb, wait);
2e8fa54e 5017 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 5018 /*
5019 * Writeback quota in non-journalled quota case - journalled quota has
5020 * no dirty dquots
5021 */
5022 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 5023 /*
5024 * Data writeback is possible w/o journal transaction, so barrier must
5025 * being sent at the end of the function. But we can skip it if
5026 * transaction_commit will do it for us.
5027 */
bda32530
TT		 5028 if (sbi->s_journal) {
5029 target = jbd2_get_latest_transaction(sbi->s_journal);
5030 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5031 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5032 needs_barrier = true;
5033
5034 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5035 if (wait)
5036 ret = jbd2_log_wait_commit(sbi->s_journal,
5037 target);
5038 }
5039 } else if (wait && test_opt(sb, BARRIER))
06a407f1 5040 needs_barrier = true;
06a407f1
DM		 5041 if (needs_barrier) {
5042 int err;
5043 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5044 if (!ret)
5045 ret = err;
0390131b 5046 }
06a407f1
DM		 5047
5048 return ret;
5049}
5050
ac27a0ec
DK		 5051/*
5052 * LVM calls this function before a (read-only) snapshot is created. This
5053 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 5054 *
5055 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 5056 * state independently. It relies on upper layer to stop all data & metadata
5057 * modifications.
ac27a0ec 5058 */
c4be0c1d 5059static int ext4_freeze(struct super_block *sb)
ac27a0ec 5060{
c4be0c1d
TS		 5061 int error = 0;
5062 journal_t *journal;
ac27a0ec 5063
bc98a42c 5064 if (sb_rdonly(sb))
9ca92389 5065 return 0;
ac27a0ec 5066
9ca92389 5067 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 5068
bb044576
TT		 5069 if (journal) {
5070 /* Now we set up the journal barrier. */
5071 jbd2_journal_lock_updates(journal);
ac27a0ec 5072
bb044576
TT		 5073 /*
5074 * Don't clear the needs_recovery flag if we failed to
5075 * flush the journal.
5076 */
5077 error = jbd2_journal_flush(journal);
5078 if (error < 0)
5079 goto out;
c642dc9e
ES		 5080
5081 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 5082 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 5083 }
9ca92389 5084
9ca92389 5085 error = ext4_commit_super(sb, 1);
6b0310fb 5086out:
bb044576
TT		 5087 if (journal)
5088 /* we rely on upper layer to stop further updates */
5089 jbd2_journal_unlock_updates(journal);
6b0310fb 5090 return error;
ac27a0ec
DK		 5091}
5092
5093/*
5094 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5095 * flag here, even though the filesystem is not technically dirty yet.
5096 */
c4be0c1d 5097static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 5098{
bc98a42c 5099 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 5100 return 0;
5101
c642dc9e
ES		 5102 if (EXT4_SB(sb)->s_journal) {
5103 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 5104 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 5105 }
5106
9ca92389 5107 ext4_commit_super(sb, 1);
c4be0c1d 5108 return 0;
ac27a0ec
DK		 5109}
5110
673c6100
TT		 5111/*
5112 * Structure to save mount options for ext4_remount's benefit
5113 */
5114struct ext4_mount_options {
5115 unsigned long s_mount_opt;
a2595b8a 5116 unsigned long s_mount_opt2;
08cefc7a
EB		 5117 kuid_t s_resuid;
5118 kgid_t s_resgid;
673c6100
TT		 5119 unsigned long s_commit_interval;
5120 u32 s_min_batch_time, s_max_batch_time;
5121#ifdef CONFIG_QUOTA
5122 int s_jquota_fmt;
a2d4a646 5123 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 5124#endif
5125};
5126
2b2d6d01 5127static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 5128{
2b2d6d01 5129 struct ext4_super_block *es;
617ba13b 5130 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5131 unsigned long old_sb_flags;
617ba13b 5132 struct ext4_mount_options old_opts;
c79d967d 5133 int enable_quota = 0;
8a266467 5134 ext4_group_t g;
b3881f74 5135 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 5136 int err = 0;
ac27a0ec 5137#ifdef CONFIG_QUOTA
03dafb5f 5138 int i, j;
33458eab 5139 char *to_free[EXT4_MAXQUOTAS];
ac27a0ec 5140#endif
d4c402d9 5141 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec 5142
21ac738e
CX		 5143 if (data && !orig_data)
5144 return -ENOMEM;
5145
ac27a0ec
DK		 5146 /* Store the original options */
5147 old_sb_flags = sb->s_flags;
5148 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 5149 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 5150 old_opts.s_resuid = sbi->s_resuid;
5151 old_opts.s_resgid = sbi->s_resgid;
5152 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 5153 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5154 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 5155#ifdef CONFIG_QUOTA
5156 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 5157 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5158 if (sbi->s_qf_names[i]) {
33458eab
TT		 5159 char *qf_name = get_qf_name(sb, sbi, i);
5160
5161 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
03dafb5f
CG		 5162 if (!old_opts.s_qf_names[i]) {
5163 for (j = 0; j < i; j++)
5164 kfree(old_opts.s_qf_names[j]);
3e36a163 5165 kfree(orig_data);
03dafb5f
CG		 5166 return -ENOMEM;
5167 }
5168 } else
5169 old_opts.s_qf_names[i] = NULL;
ac27a0ec 5170#endif
b3881f74
TT		 5171 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5172 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 5173
661aa520 5174 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 5175 err = -EINVAL;
5176 goto restore_opts;
5177 }
5178
6b992ff2 5179 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 5180 test_opt(sb, JOURNAL_CHECKSUM)) {
5181 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 5182 "during remount not supported; ignoring");
5183 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 5184 }
5185
6ae6514b
PS		 5186 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5187 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5188 ext4_msg(sb, KERN_ERR, "can't mount with "
5189 "both data=journal and delalloc");
5190 err = -EINVAL;
5191 goto restore_opts;
5192 }
5193 if (test_opt(sb, DIOREAD_NOLOCK)) {
5194 ext4_msg(sb, KERN_ERR, "can't mount with "
5195 "both data=journal and dioread_nolock");
5196 err = -EINVAL;
5197 goto restore_opts;
5198 }
923ae0ff
RZ		 5199 if (test_opt(sb, DAX)) {
5200 ext4_msg(sb, KERN_ERR, "can't mount with "
5201 "both data=journal and dax");
5202 err = -EINVAL;
5203 goto restore_opts;
5204 }
ab04df78
JK		 5205 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5206 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5207 ext4_msg(sb, KERN_ERR, "can't mount with "
5208 "journal_async_commit in data=ordered mode");
5209 err = -EINVAL;
5210 goto restore_opts;
5211 }
923ae0ff
RZ		 5212 }
5213
cdb7ee4c
TE		 5214 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5215 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5216 err = -EINVAL;
5217 goto restore_opts;
5218 }
5219
923ae0ff
RZ		 5220 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5221 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5222 "dax flag with busy inodes while remounting");
5223 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 5224 }
5225
4ab2f15b 5226 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 5227 ext4_abort(sb, "Abort forced by user");
ac27a0ec 5228
1751e8a6
LT		 5229 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5230 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 5231
5232 es = sbi->s_es;
5233
b3881f74 5234 if (sbi->s_journal) {
0390131b 5235 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5236 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5237 }
ac27a0ec 5238
1751e8a6
LT		 5239 if (*flags & SB_LAZYTIME)
5240 sb->s_flags |= SB_LAZYTIME;
a2fd66d0 5241
1751e8a6 5242 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
4ab2f15b 5243 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 5244 err = -EROFS;
5245 goto restore_opts;
5246 }
5247
1751e8a6 5248 if (*flags & SB_RDONLY) {
38c03b34
TT		 5249 err = sync_filesystem(sb);
5250 if (err < 0)
5251 goto restore_opts;
0f0dd62f
CH		 5252 err = dquot_suspend(sb, -1);
5253 if (err < 0)
c79d967d 5254 goto restore_opts;
c79d967d 5255
ac27a0ec
DK		 5256 /*
5257 * First of all, the unconditional stuff we have to do
5258 * to disable replay of the journal when we next remount
5259 */
1751e8a6 5260 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 5261
5262 /*
5263 * OK, test if we are remounting a valid rw partition
5264 * readonly, and if so set the rdonly flag and then
5265 * mark the partition as valid again.
5266 */
617ba13b
MC		 5267 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5268 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5269 es->s_state = cpu_to_le16(sbi->s_mount_state);
5270
a63c9eb2 5271 if (sbi->s_journal)
0390131b 5272 ext4_mark_recovery_complete(sb, es);
2dca60d9
TT		 5273 if (sbi->s_mmp_tsk)
5274 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 5275 } else {
a13fb1a4 5276 /* Make sure we can mount this feature set readwrite */
e2b911c5 5277 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5278 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5279 err = -EROFS;
5280 goto restore_opts;
5281 }
8a266467
TT		 5282 /*
5283 * Make sure the group descriptor checksums
0b8e58a1 5284 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5285 */
5286 for (g = 0; g < sbi->s_groups_count; g++) {
5287 struct ext4_group_desc *gdp =
5288 ext4_get_group_desc(sb, g, NULL);
5289
feb0ab32 5290 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5291 ext4_msg(sb, KERN_ERR,
5292 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5293 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5294 le16_to_cpu(gdp->bg_checksum));
6a797d27 5295 err = -EFSBADCRC;
8a266467
TT		 5296 goto restore_opts;
5297 }
5298 }
5299
ead6596b
ES		 5300 /*
5301 * If we have an unprocessed orphan list hanging
5302 * around from a previously readonly bdev mount,
5303 * require a full umount/remount for now.
5304 */
5305 if (es->s_last_orphan) {
b31e1552 5306 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5307 "remount RDWR because of unprocessed "
5308 "orphan inode list. Please "
b31e1552 5309 "umount/remount instead");
ead6596b
ES		 5310 err = -EINVAL;
5311 goto restore_opts;
5312 }
5313
ac27a0ec
DK		 5314 /*
5315 * Mounting a RDONLY partition read-write, so reread
5316 * and store the current valid flag. (It may have
5317 * been changed by e2fsck since we originally mounted
5318 * the partition.)
5319 */
0390131b
FM		 5320 if (sbi->s_journal)
5321 ext4_clear_journal_err(sb, es);
ac27a0ec 5322 sbi->s_mount_state = le16_to_cpu(es->s_state);
c89128a0
JK		 5323
5324 err = ext4_setup_super(sb, es, 0);
5325 if (err)
5326 goto restore_opts;
5327
5328 sb->s_flags &= ~SB_RDONLY;
e2b911c5 5329 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5330 if (ext4_multi_mount_protect(sb,
5331 le64_to_cpu(es->s_mmp_block))) {
5332 err = -EROFS;
5333 goto restore_opts;
5334 }
c79d967d 5335 enable_quota = 1;
ac27a0ec
DK		 5336 }
5337 }
bfff6873
LC		 5338
5339 /*
5340 * Reinitialize lazy itable initialization thread based on
5341 * current settings
5342 */
bc98a42c 5343 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 5344 ext4_unregister_li_request(sb);
5345 else {
5346 ext4_group_t first_not_zeroed;
5347 first_not_zeroed = ext4_has_uninit_itable(sb);
5348 ext4_register_li_request(sb, first_not_zeroed);
5349 }
5350
6fd058f7 5351 ext4_setup_system_zone(sb);
c89128a0
JK		 5352 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5353 err = ext4_commit_super(sb, 1);
5354 if (err)
5355 goto restore_opts;
5356 }
0390131b 5357
ac27a0ec
DK		 5358#ifdef CONFIG_QUOTA
5359 /* Release old quota file names */
a2d4a646 5360 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5361 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5362 if (enable_quota) {
5363 if (sb_any_quota_suspended(sb))
5364 dquot_resume(sb, -1);
e2b911c5 5365 else if (ext4_has_feature_quota(sb)) {
7c319d32 5366 err = ext4_enable_quotas(sb);
07724f98 5367 if (err)
7c319d32 5368 goto restore_opts;
7c319d32
AK		 5369 }
5370 }
ac27a0ec 5371#endif
d4c402d9 5372
1751e8a6 5373 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
d4c402d9
CW		 5374 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5375 kfree(orig_data);
ac27a0ec 5376 return 0;
0b8e58a1 5377
ac27a0ec
DK		 5378restore_opts:
5379 sb->s_flags = old_sb_flags;
5380 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5381 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5382 sbi->s_resuid = old_opts.s_resuid;
5383 sbi->s_resgid = old_opts.s_resgid;
5384 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5385 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5386 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5387#ifdef CONFIG_QUOTA
5388 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5389 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
33458eab
TT		 5390 to_free[i] = get_qf_name(sb, sbi, i);
5391 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
ac27a0ec 5392 }
33458eab
TT		 5393 synchronize_rcu();
5394 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5395 kfree(to_free[i]);
ac27a0ec 5396#endif
d4c402d9 5397 kfree(orig_data);
ac27a0ec
DK		 5398 return err;
5399}
5400
689c958c
LX		 5401#ifdef CONFIG_QUOTA
5402static int ext4_statfs_project(struct super_block *sb,
5403 kprojid_t projid, struct kstatfs *buf)
5404{
5405 struct kqid qid;
5406 struct dquot *dquot;
5407 u64 limit;
5408 u64 curblock;
5409
5410 qid = make_kqid_projid(projid);
5411 dquot = dqget(sb, qid);
5412 if (IS_ERR(dquot))
5413 return PTR_ERR(dquot);
7b9ca4c6 5414 spin_lock(&dquot->dq_dqb_lock);
689c958c
LX		 5415
5416 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5417 dquot->dq_dqb.dqb_bsoftlimit :
5418 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5419 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 5420 curblock = (dquot->dq_dqb.dqb_curspace +
5421 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 5422 buf->f_blocks = limit;
5423 buf->f_bfree = buf->f_bavail =
5424 (buf->f_blocks > curblock) ?
5425 (buf->f_blocks - curblock) : 0;
5426 }
5427
5428 limit = dquot->dq_dqb.dqb_isoftlimit ?
5429 dquot->dq_dqb.dqb_isoftlimit :
5430 dquot->dq_dqb.dqb_ihardlimit;
5431 if (limit && buf->f_files > limit) {
5432 buf->f_files = limit;
5433 buf->f_ffree =
5434 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5435 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5436 }
5437
7b9ca4c6 5438 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 5439 dqput(dquot);
5440 return 0;
5441}
5442#endif
5443
2b2d6d01 5444static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5445{
5446 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5447 struct ext4_sb_info *sbi = EXT4_SB(sb);
5448 struct ext4_super_block *es = sbi->s_es;
27dd4385 5449 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5450 u64 fsid;
d02a9391 5451 s64 bfree;
27dd4385 5452 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5453
952fc18e
TT		 5454 if (!test_opt(sb, MINIX_DF))
5455 overhead = sbi->s_overhead;
ac27a0ec 5456
617ba13b 5457 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5458 buf->f_bsize = sb->s_blocksize;
b72f78cb 5459 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5460 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5461 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5462 /* prevent underflow in case that few free space is available */
57042651 5463 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5464 buf->f_bavail = buf->f_bfree -
5465 (ext4_r_blocks_count(es) + resv_blocks);
5466 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5467 buf->f_bavail = 0;
5468 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5469 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5470 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5471 fsid = le64_to_cpup((void *)es->s_uuid) ^
5472 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5473 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5474 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5475
689c958c
LX		 5476#ifdef CONFIG_QUOTA
5477 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5478 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5479 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5480#endif
ac27a0ec
DK		 5481 return 0;
5482}
5483
ac27a0ec
DK		 5484
5485#ifdef CONFIG_QUOTA
5486
bc8230ee
JK		 5487/*
5488 * Helper functions so that transaction is started before we acquire dqio_sem
5489 * to keep correct lock ordering of transaction > dqio_sem
5490 */
ac27a0ec
DK		 5491static inline struct inode *dquot_to_inode(struct dquot *dquot)
5492{
4c376dca 5493 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5494}
5495
617ba13b 5496static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5497{
5498 int ret, err;
5499 handle_t *handle;
5500 struct inode *inode;
5501
5502 inode = dquot_to_inode(dquot);
9924a92a 5503 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5504 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5505 if (IS_ERR(handle))
5506 return PTR_ERR(handle);
5507 ret = dquot_commit(dquot);
617ba13b 5508 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5509 if (!ret)
5510 ret = err;
5511 return ret;
5512}
5513
617ba13b 5514static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5515{
5516 int ret, err;
5517 handle_t *handle;
5518
9924a92a 5519 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5520 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5521 if (IS_ERR(handle))
5522 return PTR_ERR(handle);
5523 ret = dquot_acquire(dquot);
617ba13b 5524 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5525 if (!ret)
5526 ret = err;
5527 return ret;
5528}
5529
617ba13b 5530static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5531{
5532 int ret, err;
5533 handle_t *handle;
5534
9924a92a 5535 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5536 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5537 if (IS_ERR(handle)) {
5538 /* Release dquot anyway to avoid endless cycle in dqput() */
5539 dquot_release(dquot);
ac27a0ec 5540 return PTR_ERR(handle);
9c3013e9 5541 }
ac27a0ec 5542 ret = dquot_release(dquot);
617ba13b 5543 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5544 if (!ret)
5545 ret = err;
5546 return ret;
5547}
5548
617ba13b 5549static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5550{
262b4662
JK		 5551 struct super_block *sb = dquot->dq_sb;
5552 struct ext4_sb_info *sbi = EXT4_SB(sb);
5553
2c8be6b2 5554 /* Are we journaling quotas? */
e2b911c5 5555 if (ext4_has_feature_quota(sb) ||
262b4662 5556 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5557 dquot_mark_dquot_dirty(dquot);
617ba13b 5558 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5559 } else {
5560 return dquot_mark_dquot_dirty(dquot);
5561 }
5562}
5563
617ba13b 5564static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5565{
5566 int ret, err;
5567 handle_t *handle;
5568
5569 /* Data block + inode block */
2b0143b5 5570 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5571 if (IS_ERR(handle))
5572 return PTR_ERR(handle);
5573 ret = dquot_commit_info(sb, type);
617ba13b 5574 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5575 if (!ret)
5576 ret = err;
5577 return ret;
5578}
5579
5580/*
5581 * Turn on quotas during mount time - we need to find
5582 * the quota file and such...
5583 */
617ba13b 5584static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5585{
33458eab 5586 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
287a8095 5587 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5588}
5589
daf647d2
TT		 5590static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5591{
5592 struct ext4_inode_info *ei = EXT4_I(inode);
5593
5594 /* The first argument of lockdep_set_subclass has to be
5595 * *exactly* the same as the argument to init_rwsem() --- in
5596 * this case, in init_once() --- or lockdep gets unhappy
5597 * because the name of the lock is set using the
5598 * stringification of the argument to init_rwsem().
5599 */
5600 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5601 lockdep_set_subclass(&ei->i_data_sem, subclass);
5602}
5603
ac27a0ec
DK		 5604/*
5605 * Standard function to be called on quota_on
5606 */
617ba13b 5607static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5608 const struct path *path)
ac27a0ec
DK		 5609{
5610 int err;
ac27a0ec
DK		 5611
5612 if (!test_opt(sb, QUOTA))
5613 return -EINVAL;
0623543b 5614
ac27a0ec 5615 /* Quotafile not on the same filesystem? */
d8c9584e 5616 if (path->dentry->d_sb != sb)
ac27a0ec 5617 return -EXDEV;
0623543b
JK		 5618 /* Journaling quota? */
5619 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5620 /* Quotafile not in fs root? */
f00c9e44 5621 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5622 ext4_msg(sb, KERN_WARNING,
5623 "Quota file not on filesystem root. "
5624 "Journaled quota will not work");
91389240
JK		 5625 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5626 } else {
5627 /*
5628 * Clear the flag just in case mount options changed since
5629 * last time.
5630 */
5631 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 5632 }
0623543b
JK		 5633
5634 /*
5635 * When we journal data on quota file, we have to flush journal to see
5636 * all updates to the file when we bypass pagecache...
5637 */
0390131b 5638 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5639 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5640 /*
5641 * We don't need to lock updates but journal_flush() could
5642 * otherwise be livelocked...
5643 */
5644 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5645 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5646 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5647 if (err)
7ffe1ea8 5648 return err;
0623543b 5649 }
957153fc 5650
daf647d2
TT		 5651 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5652 err = dquot_quota_on(sb, type, format_id, path);
957153fc 5653 if (err) {
daf647d2
TT		 5654 lockdep_set_quota_inode(path->dentry->d_inode,
5655 I_DATA_SEM_NORMAL);
957153fc
JK		 5656 } else {
5657 struct inode *inode = d_inode(path->dentry);
5658 handle_t *handle;
5659
61a92987
JK		 5660 /*
5661 * Set inode flags to prevent userspace from messing with quota
5662 * files. If this fails, we return success anyway since quotas
5663 * are already enabled and this is not a hard failure.
5664 */
957153fc
JK		 5665 inode_lock(inode);
5666 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5667 if (IS_ERR(handle))
5668 goto unlock_inode;
5669 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5670 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5671 S_NOATIME | S_IMMUTABLE);
5672 ext4_mark_inode_dirty(handle, inode);
5673 ext4_journal_stop(handle);
5674 unlock_inode:
5675 inode_unlock(inode);
5676 }
daf647d2 5677 return err;
ac27a0ec
DK		 5678}
5679
7c319d32
AK		 5680static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5681 unsigned int flags)
5682{
5683 int err;
5684 struct inode *qf_inode;
a2d4a646 5685 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5686 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5687 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5688 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5689 };
5690
e2b911c5 5691 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5692
5693 if (!qf_inums[type])
5694 return -EPERM;
5695
8a363970 5696 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7c319d32
AK		 5697 if (IS_ERR(qf_inode)) {
5698 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5699 return PTR_ERR(qf_inode);
5700 }
5701
bcb13850
JK		 5702 /* Don't account quota for quota files to avoid recursion */
5703 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5704 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32 5705 err = dquot_enable(qf_inode, type, format_id, flags);
daf647d2
TT		 5706 if (err)
5707 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
61157b24 5708 iput(qf_inode);
7c319d32
AK		 5709
5710 return err;
5711}
5712
5713/* Enable usage tracking for all quota types. */
5714static int ext4_enable_quotas(struct super_block *sb)
5715{
5716 int type, err = 0;
a2d4a646 5717 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5718 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5719 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5720 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5721 };
49da9392
JK		 5722 bool quota_mopt[EXT4_MAXQUOTAS] = {
5723 test_opt(sb, USRQUOTA),
5724 test_opt(sb, GRPQUOTA),
5725 test_opt(sb, PRJQUOTA),
5726 };
7c319d32 5727
91389240 5728 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 5729 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5730 if (qf_inums[type]) {
5731 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5732 DQUOT_USAGE_ENABLED |
5733 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 5734 if (err) {
5735 ext4_warning(sb,
72ba7450
TT		 5736 "Failed to enable quota tracking "
5737 "(type=%d, err=%d). Please run "
5738 "e2fsck to fix.", type, err);
7f144fd0
JU		 5739 for (type--; type >= 0; type--)
5740 dquot_quota_off(sb, type);
5741
7c319d32
AK		 5742 return err;
5743 }
5744 }
5745 }
5746 return 0;
5747}
5748
ca0e05e4
DM		 5749static int ext4_quota_off(struct super_block *sb, int type)
5750{
21f97697
JK		 5751 struct inode *inode = sb_dqopt(sb)->files[type];
5752 handle_t *handle;
957153fc 5753 int err;
21f97697 5754
87009d86
DM		 5755 /* Force all delayed allocation blocks to be allocated.
5756 * Caller already holds s_umount sem */
5757 if (test_opt(sb, DELALLOC))
ca0e05e4 5758 sync_filesystem(sb);
ca0e05e4 5759
957153fc 5760 if (!inode || !igrab(inode))
0b268590
AG		 5761 goto out;
5762
957153fc 5763 err = dquot_quota_off(sb, type);
964edf66 5764 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 5765 goto out_put;
5766
5767 inode_lock(inode);
61a92987
JK		 5768 /*
5769 * Update modification times of quota files when userspace can
5770 * start looking at them. If we fail, we return success anyway since
5771 * this is not a hard failure and quotas are already disabled.
5772 */
9924a92a 5773 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697 5774 if (IS_ERR(handle))
957153fc
JK		 5775 goto out_unlock;
5776 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5777 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 5778 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5779 ext4_mark_inode_dirty(handle, inode);
5780 ext4_journal_stop(handle);
957153fc
JK		 5781out_unlock:
5782 inode_unlock(inode);
5783out_put:
964edf66 5784 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 5785 iput(inode);
5786 return err;
21f97697 5787out:
ca0e05e4
DM		 5788 return dquot_quota_off(sb, type);
5789}
5790
ac27a0ec
DK		 5791/* Read data from quotafile - avoid pagecache and such because we cannot afford
5792 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5793 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5794 * we don't have to be afraid of races */
617ba13b 5795static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5796 size_t len, loff_t off)
5797{
5798 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5799 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5800 int offset = off & (sb->s_blocksize - 1);
5801 int tocopy;
5802 size_t toread;
5803 struct buffer_head *bh;
5804 loff_t i_size = i_size_read(inode);
5805
5806 if (off > i_size)
5807 return 0;
5808 if (off+len > i_size)
5809 len = i_size-off;
5810 toread = len;
5811 while (toread > 0) {
5812 tocopy = sb->s_blocksize - offset < toread ?
5813 sb->s_blocksize - offset : toread;
1c215028
TT		 5814 bh = ext4_bread(NULL, inode, blk, 0);
5815 if (IS_ERR(bh))
5816 return PTR_ERR(bh);
ac27a0ec
DK		 5817 if (!bh) /* A hole? */
5818 memset(data, 0, tocopy);
5819 else
5820 memcpy(data, bh->b_data+offset, tocopy);
5821 brelse(bh);
5822 offset = 0;
5823 toread -= tocopy;
5824 data += tocopy;
5825 blk++;
5826 }
5827 return len;
5828}
5829
5830/* Write to quotafile (we know the transaction is already started and has
5831 * enough credits) */
617ba13b 5832static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5833 const char *data, size_t len, loff_t off)
5834{
5835 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5836 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5837 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5838 int retries = 0;
ac27a0ec
DK		 5839 struct buffer_head *bh;
5840 handle_t *handle = journal_current_handle();
5841
0390131b 5842 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5843 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5844 " cancelled because transaction is not started",
9c3013e9
JK		 5845 (unsigned long long)off, (unsigned long long)len);
5846 return -EIO;
5847 }
67eeb568
DM		 5848 /*
5849 * Since we account only one data block in transaction credits,
5850 * then it is impossible to cross a block boundary.
5851 */
5852 if (sb->s_blocksize - offset < len) {
5853 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5854 " cancelled because not block aligned",
5855 (unsigned long long)off, (unsigned long long)len);
5856 return -EIO;
5857 }
5858
c5e298ae
TT		 5859 do {
5860 bh = ext4_bread(handle, inode, blk,
5861 EXT4_GET_BLOCKS_CREATE |
5862 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5863 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5864 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5865 if (IS_ERR(bh))
5866 return PTR_ERR(bh);
67eeb568
DM		 5867 if (!bh)
5868 goto out;
5d601255 5869 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5870 err = ext4_journal_get_write_access(handle, bh);
5871 if (err) {
5872 brelse(bh);
1c215028 5873 return err;
ac27a0ec 5874 }
67eeb568
DM		 5875 lock_buffer(bh);
5876 memcpy(bh->b_data+offset, data, len);
5877 flush_dcache_page(bh->b_page);
5878 unlock_buffer(bh);
62d2b5f2 5879 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5880 brelse(bh);
ac27a0ec 5881out:
67eeb568
DM		 5882 if (inode->i_size < off + len) {
5883 i_size_write(inode, off + len);
617ba13b 5884 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5885 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5886 }
67eeb568 5887 return len;
ac27a0ec
DK		 5888}
5889
8f0e8746
TT		 5890static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5891{
5892 const struct quota_format_ops *ops;
5893
5894 if (!sb_has_quota_loaded(sb, qid->type))
5895 return -ESRCH;
5896 ops = sb_dqopt(sb)->ops[qid->type];
5897 if (!ops || !ops->get_next_id)
5898 return -ENOSYS;
5899 return dquot_get_next_id(sb, qid);
5900}
ac27a0ec
DK		 5901#endif
5902
152a0836
AV		 5903static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5904 const char *dev_name, void *data)
ac27a0ec 5905{
152a0836 5906 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5907}
5908
c290ea01 5909#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5910static inline void register_as_ext2(void)
5911{
5912 int err = register_filesystem(&ext2_fs_type);
5913 if (err)
5914 printk(KERN_WARNING
5915 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5916}
5917
5918static inline void unregister_as_ext2(void)
5919{
5920 unregister_filesystem(&ext2_fs_type);
5921}
2035e776
TT		 5922
5923static inline int ext2_feature_set_ok(struct super_block *sb)
5924{
e2b911c5 5925 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 5926 return 0;
bc98a42c 5927 if (sb_rdonly(sb))
2035e776 5928 return 1;
e2b911c5 5929 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5930 return 0;
5931 return 1;
5932}
24b58424
TT		 5933#else
5934static inline void register_as_ext2(void) { }
5935static inline void unregister_as_ext2(void) { }
2035e776 5936static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5937#endif
5938
24b58424
TT		 5939static inline void register_as_ext3(void)
5940{
5941 int err = register_filesystem(&ext3_fs_type);
5942 if (err)
5943 printk(KERN_WARNING
5944 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5945}
5946
5947static inline void unregister_as_ext3(void)
5948{
5949 unregister_filesystem(&ext3_fs_type);
5950}
2035e776
TT		 5951
5952static inline int ext3_feature_set_ok(struct super_block *sb)
5953{
e2b911c5 5954 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5955 return 0;
e2b911c5 5956 if (!ext4_has_feature_journal(sb))
2035e776 5957 return 0;
bc98a42c 5958 if (sb_rdonly(sb))
2035e776 5959 return 1;
e2b911c5 5960 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5961 return 0;
5962 return 1;
5963}
24b58424 5964
03010a33
TT		 5965static struct file_system_type ext4_fs_type = {
5966 .owner = THIS_MODULE,
5967 .name = "ext4",
152a0836 5968 .mount = ext4_mount,
03010a33
TT		 5969 .kill_sb = kill_block_super,
5970 .fs_flags = FS_REQUIRES_DEV,
5971};
7f78e035 5972MODULE_ALIAS_FS("ext4");
03010a33 5973
e9e3bcec
ES		 5974/* Shared across all ext4 file systems */
5975wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5976
5dabfc78 5977static int __init ext4_init_fs(void)
ac27a0ec 5978{
e9e3bcec 5979 int i, err;
c9de560d 5980
e294a537 5981 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5982 ext4_li_info = NULL;
5983 mutex_init(&ext4_li_mtx);
5984
9a4c8019 5985 /* Build-time check for flags consistency */
12e9b892 5986 ext4_check_flag_values();
e9e3bcec 5987
e142d052 5988 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5989 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5990
51865fda 5991 err = ext4_init_es();
6fd058f7
TT		 5992 if (err)
5993 return err;
51865fda 5994
1dc0aa46
EW		 5995 err = ext4_init_pending();
5996 if (err)
5997 goto out6;
5998
51865fda
ZL		 5999 err = ext4_init_pageio();
6000 if (err)
b5799018 6001 goto out5;
51865fda 6002
5dabfc78 6003 err = ext4_init_system_zone();
bd2d0210 6004 if (err)
b5799018 6005 goto out4;
857ac889 6006
b5799018 6007 err = ext4_init_sysfs();
dd68314c 6008 if (err)
b5799018 6009 goto out3;
857ac889 6010
5dabfc78 6011 err = ext4_init_mballoc();
c9de560d
AT		 6012 if (err)
6013 goto out2;
ac27a0ec
DK		 6014 err = init_inodecache();
6015 if (err)
6016 goto out1;
24b58424 6017 register_as_ext3();
2035e776 6018 register_as_ext2();
03010a33 6019 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 6020 if (err)
6021 goto out;
bfff6873 6022
ac27a0ec
DK		 6023 return 0;
6024out:
24b58424
TT		 6025 unregister_as_ext2();
6026 unregister_as_ext3();
ac27a0ec
DK		 6027 destroy_inodecache();
6028out1:
5dabfc78 6029 ext4_exit_mballoc();
9c191f70 6030out2:
b5799018
TT		 6031 ext4_exit_sysfs();
6032out3:
5dabfc78 6033 ext4_exit_system_zone();
b5799018 6034out4:
5dabfc78 6035 ext4_exit_pageio();
b5799018 6036out5:
1dc0aa46
EW		 6037 ext4_exit_pending();
6038out6:
51865fda
ZL		 6039 ext4_exit_es();
6040
ac27a0ec
DK		 6041 return err;
6042}
6043
5dabfc78 6044static void __exit ext4_exit_fs(void)
ac27a0ec 6045{
bfff6873 6046 ext4_destroy_lazyinit_thread();
24b58424
TT		 6047 unregister_as_ext2();
6048 unregister_as_ext3();
03010a33 6049 unregister_filesystem(&ext4_fs_type);
ac27a0ec 6050 destroy_inodecache();
5dabfc78 6051 ext4_exit_mballoc();
b5799018 6052 ext4_exit_sysfs();
5dabfc78
TT		 6053 ext4_exit_system_zone();
6054 ext4_exit_pageio();
dd12ed14 6055 ext4_exit_es();
1dc0aa46 6056 ext4_exit_pending();
ac27a0ec
DK		 6057}
6058
6059MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 6060MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 6061MODULE_LICENSE("GPL");
7ef79ad5 6062MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 6063module_init(ext4_init_fs)
6064module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository