]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-hirsute-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'asoc-v5.7' of https://git.kernel.org/pub/scm/linux/kernel/git/broonie...
[mirror_ubuntu-hirsute-kernel.git] / fs / ext4 / super.c
CommitLineData
f5166768 1// SPDX-License-Identifier: GPL-2.0
ac27a0ec 2/*
617ba13b 3 * linux/fs/ext4/super.c
ac27a0ec
DK		 4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/fs/minix/inode.c
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 *
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
18 */
19
20#include <linux/module.h>
21#include <linux/string.h>
22#include <linux/fs.h>
23#include <linux/time.h>
c5ca7c76 24#include <linux/vmalloc.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
66114cad 28#include <linux/backing-dev.h>
ac27a0ec 29#include <linux/parser.h>
ac27a0ec 30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ef510424 41#include <linux/dax.h>
7abc52c2 42#include <linux/cleancache.h>
7c0f6ba6 43#include <linux/uaccess.h>
ee73f9a5 44#include <linux/iversion.h>
c83ad55e 45#include <linux/unicode.h>
ac27a0ec 46
bfff6873
LC		 47#include <linux/kthread.h>
48#include <linux/freezer.h>
49
3dcf5451 50#include "ext4.h"
4a092d73 51#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 52#include "ext4_jbd2.h"
ac27a0ec
DK		 53#include "xattr.h"
54#include "acl.h"
3661d286 55#include "mballoc.h"
0c9ec4be 56#include "fsmap.h"
ac27a0ec 57
9bffad1e
TT		 58#define CREATE_TRACE_POINTS
59#include <trace/events/ext4.h>
60
0b75a840
LC		 61static struct ext4_lazy_init *ext4_li_info;
62static struct mutex ext4_li_mtx;
e294a537 63static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 64
617ba13b 65static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 66 unsigned long journal_devnum);
2adf6da8 67static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 68static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 69static void ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71static void ext4_clear_journal_err(struct super_block *sb,
72 struct ext4_super_block *es);
617ba13b 73static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 74static int ext4_remount(struct super_block *sb, int *flags, char *data);
75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 76static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 77static int ext4_freeze(struct super_block *sb);
152a0836
AV		 78static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
2035e776
TT		 80static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 82static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 83static void ext4_destroy_lazyinit_thread(void);
84static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 85static void ext4_clear_request_list(void);
c6cb7e77
EW		 86static struct inode *ext4_get_journal_inode(struct super_block *sb,
87 unsigned int journal_inum);
ac27a0ec 88
e74031fd
JK		 89/*
90 * Lock ordering
91 *
92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
93 * i_mmap_rwsem (inode->i_mmap_rwsem)!
94 *
95 * page fault path:
96 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
97 * page lock -> i_data_sem (rw)
98 *
99 * buffered write path:
100 * sb_start_write -> i_mutex -> mmap_sem
101 * sb_start_write -> i_mutex -> transaction start -> page lock ->
102 * i_data_sem (rw)
103 *
104 * truncate:
1d39834f
NB		 105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
107 * i_data_sem (rw)
e74031fd
JK		 108 *
109 * direct IO:
1d39834f
NB		 110 * sb_start_write -> i_mutex -> mmap_sem
111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
e74031fd
JK		 112 *
113 * writepages:
114 * transaction start -> page lock(s) -> i_data_sem (rw)
115 */
116
c290ea01 117#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 118static struct file_system_type ext2_fs_type = {
119 .owner = THIS_MODULE,
120 .name = "ext2",
121 .mount = ext4_mount,
122 .kill_sb = kill_block_super,
123 .fs_flags = FS_REQUIRES_DEV,
124};
7f78e035 125MODULE_ALIAS_FS("ext2");
fa7614dd 126MODULE_ALIAS("ext2");
2035e776
TT		 127#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
128#else
129#define IS_EXT2_SB(sb) (0)
130#endif
131
132
ba69f9ab
JK		 133static struct file_system_type ext3_fs_type = {
134 .owner = THIS_MODULE,
135 .name = "ext3",
152a0836 136 .mount = ext4_mount,
ba69f9ab
JK		 137 .kill_sb = kill_block_super,
138 .fs_flags = FS_REQUIRES_DEV,
139};
7f78e035 140MODULE_ALIAS_FS("ext3");
fa7614dd 141MODULE_ALIAS("ext3");
ba69f9ab 142#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 143
fb265c9c
TT		 144/*
145 * This works like sb_bread() except it uses ERR_PTR for error
146 * returns. Currently with sb_bread it's impossible to distinguish
147 * between ENOMEM and EIO situations (since both result in a NULL
148 * return.
149 */
150struct buffer_head *
151ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags)
152{
153 struct buffer_head *bh = sb_getblk(sb, block);
154
155 if (bh == NULL)
156 return ERR_PTR(-ENOMEM);
cf2834a5 157 if (ext4_buffer_uptodate(bh))
fb265c9c
TT		 158 return bh;
159 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh);
160 wait_on_buffer(bh);
161 if (buffer_uptodate(bh))
162 return bh;
163 put_bh(bh);
164 return ERR_PTR(-EIO);
165}
166
d25425f8
DW		 167static int ext4_verify_csum_type(struct super_block *sb,
168 struct ext4_super_block *es)
169{
e2b911c5 170 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 171 return 1;
172
173 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
174}
175
a9c47317
DW		 176static __le32 ext4_superblock_csum(struct super_block *sb,
177 struct ext4_super_block *es)
178{
179 struct ext4_sb_info *sbi = EXT4_SB(sb);
180 int offset = offsetof(struct ext4_super_block, s_checksum);
181 __u32 csum;
182
183 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
184
185 return cpu_to_le32(csum);
186}
187
c197855e
SH		 188static int ext4_superblock_csum_verify(struct super_block *sb,
189 struct ext4_super_block *es)
a9c47317 190{
9aa5d32b 191 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 192 return 1;
193
194 return es->s_checksum == ext4_superblock_csum(sb, es);
195}
196
06db49e6 197void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 198{
06db49e6
TT		 199 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
200
9aa5d32b 201 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 202 return;
203
204 es->s_checksum = ext4_superblock_csum(sb, es);
205}
206
8fadc143
AR		 207ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
208 struct ext4_group_desc *bg)
bd81d8ee 209{
3a14589c 210 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 212 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 213}
214
8fadc143
AR		 215ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
216 struct ext4_group_desc *bg)
bd81d8ee 217{
5272f837 218 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 220 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 221}
222
8fadc143
AR		 223ext4_fsblk_t ext4_inode_table(struct super_block *sb,
224 struct ext4_group_desc *bg)
bd81d8ee 225{
5272f837 226 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 228 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 229}
230
021b65bb
TT		 231__u32 ext4_free_group_clusters(struct super_block *sb,
232 struct ext4_group_desc *bg)
560671a0
AK		 233{
234 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 236 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 237}
238
239__u32 ext4_free_inodes_count(struct super_block *sb,
240 struct ext4_group_desc *bg)
241{
242 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 244 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 245}
246
247__u32 ext4_used_dirs_count(struct super_block *sb,
248 struct ext4_group_desc *bg)
249{
250 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
251 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 252 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 253}
254
255__u32 ext4_itable_unused_count(struct super_block *sb,
256 struct ext4_group_desc *bg)
257{
258 return le16_to_cpu(bg->bg_itable_unused_lo) |
259 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 260 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 261}
262
8fadc143
AR		 263void ext4_block_bitmap_set(struct super_block *sb,
264 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 265{
3a14589c 266 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 269}
270
8fadc143
AR		 271void ext4_inode_bitmap_set(struct super_block *sb,
272 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 273{
5272f837 274 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 277}
278
8fadc143
AR		 279void ext4_inode_table_set(struct super_block *sb,
280 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 281{
5272f837 282 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 285}
286
021b65bb
TT		 287void ext4_free_group_clusters_set(struct super_block *sb,
288 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 289{
290 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
293}
294
295void ext4_free_inodes_set(struct super_block *sb,
296 struct ext4_group_desc *bg, __u32 count)
297{
298 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
301}
302
303void ext4_used_dirs_set(struct super_block *sb,
304 struct ext4_group_desc *bg, __u32 count)
305{
306 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
307 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
308 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
309}
310
311void ext4_itable_unused_set(struct super_block *sb,
312 struct ext4_group_desc *bg, __u32 count)
313{
314 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
315 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
316 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
317}
318
6a0678a7
AB		 319static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
320{
321 time64_t now = ktime_get_real_seconds();
322
323 now = clamp_val(now, 0, (1ull << 40) - 1);
324
325 *lo = cpu_to_le32(lower_32_bits(now));
326 *hi = upper_32_bits(now);
327}
328
329static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
330{
331 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
332}
333#define ext4_update_tstamp(es, tstamp) \
334 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
335#define ext4_get_tstamp(es, tstamp) \
336 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
d3d1faf6 337
1c13d5c0
TT		 338static void __save_error_info(struct super_block *sb, const char *func,
339 unsigned int line)
340{
341 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
342
343 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 344 if (bdev_read_only(sb->s_bdev))
345 return;
1c13d5c0 346 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6a0678a7 347 ext4_update_tstamp(es, s_last_error_time);
1c13d5c0
TT		 348 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
349 es->s_last_error_line = cpu_to_le32(line);
878520ac
TT		 350 if (es->s_last_error_errcode == 0)
351 es->s_last_error_errcode = EXT4_ERR_EFSCORRUPTED;
1c13d5c0
TT		 352 if (!es->s_first_error_time) {
353 es->s_first_error_time = es->s_last_error_time;
6a0678a7 354 es->s_first_error_time_hi = es->s_last_error_time_hi;
1c13d5c0
TT		 355 strncpy(es->s_first_error_func, func,
356 sizeof(es->s_first_error_func));
357 es->s_first_error_line = cpu_to_le32(line);
358 es->s_first_error_ino = es->s_last_error_ino;
359 es->s_first_error_block = es->s_last_error_block;
878520ac 360 es->s_first_error_errcode = es->s_last_error_errcode;
1c13d5c0 361 }
66e61a9e
TT		 362 /*
363 * Start the daily error reporting function if it hasn't been
364 * started already
365 */
366 if (!es->s_error_count)
367 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 368 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 369}
370
371static void save_error_info(struct super_block *sb, const char *func,
372 unsigned int line)
373{
374 __save_error_info(sb, func, line);
375 ext4_commit_super(sb, 1);
376}
377
bdfe0cbd
TT		 378/*
379 * The del_gendisk() function uninitializes the disk-specific data
380 * structures, including the bdi structure, without telling anyone
381 * else. Once this happens, any attempt to call mark_buffer_dirty()
382 * (for example, by ext4_commit_super), will cause a kernel OOPS.
383 * This is a kludge to prevent these oops until we can put in a proper
384 * hook in del_gendisk() to inform the VFS and file system layers.
385 */
386static int block_device_ejected(struct super_block *sb)
387{
388 struct inode *bd_inode = sb->s_bdev->bd_inode;
389 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
390
391 return bdi->dev == NULL;
392}
393
18aadd47
BJ		 394static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
395{
396 struct super_block *sb = journal->j_private;
397 struct ext4_sb_info *sbi = EXT4_SB(sb);
398 int error = is_journal_aborted(journal);
5d3ee208 399 struct ext4_journal_cb_entry *jce;
18aadd47 400
5d3ee208 401 BUG_ON(txn->t_state == T_FINISHED);
a0154344
DJ		 402
403 ext4_process_freed_data(sb, txn->t_tid);
404
18aadd47 405 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 406 while (!list_empty(&txn->t_private_list)) {
407 jce = list_entry(txn->t_private_list.next,
408 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 409 list_del_init(&jce->jce_list);
410 spin_unlock(&sbi->s_md_lock);
411 jce->jce_func(sb, jce, error);
412 spin_lock(&sbi->s_md_lock);
413 }
414 spin_unlock(&sbi->s_md_lock);
415}
1c13d5c0 416
1dc1097f
JK		 417static bool system_going_down(void)
418{
419 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
420 || system_state == SYSTEM_RESTART;
421}
422
ac27a0ec
DK		 423/* Deal with the reporting of failure conditions on a filesystem such as
424 * inconsistencies detected or read IO failures.
425 *
426 * On ext2, we can store the error state of the filesystem in the
617ba13b 427 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 428 * write ordering constraints on the superblock which prevent us from
429 * writing it out straight away; and given that the journal is about to
430 * be aborted, we can't rely on the current, or future, transactions to
431 * write out the superblock safely.
432 *
dab291af 433 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 434 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 435 * that error until we've noted it down and cleared it.
436 */
437
617ba13b 438static void ext4_handle_error(struct super_block *sb)
ac27a0ec 439{
327eaf73
TT		 440 if (test_opt(sb, WARN_ON_ERROR))
441 WARN_ON_ONCE(1);
442
bc98a42c 443 if (sb_rdonly(sb))
ac27a0ec
DK		 444 return;
445
2b2d6d01 446 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 447 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 448
4ab2f15b 449 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 450 if (journal)
dab291af 451 jbd2_journal_abort(journal, -EIO);
ac27a0ec 452 }
1dc1097f
JK		 453 /*
454 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
455 * could panic during 'reboot -f' as the underlying device got already
456 * disabled.
457 */
458 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
b31e1552 459 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 460 /*
461 * Make sure updated value of ->s_mount_flags will be visible
462 * before ->s_flags update
463 */
464 smp_wmb();
1751e8a6 465 sb->s_flags |= SB_RDONLY;
1dc1097f 466 } else if (test_opt(sb, ERRORS_PANIC)) {
4327ba52
DJ		 467 if (EXT4_SB(sb)->s_journal &&
468 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
469 return;
617ba13b 470 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 471 sb->s_id);
4327ba52 472 }
ac27a0ec
DK		 473}
474
efbed4dc
TT		 475#define ext4_error_ratelimit(sb) \
476 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
477 "EXT4-fs error")
478
12062ddd 479void __ext4_error(struct super_block *sb, const char *function,
c398eda0 480 unsigned int line, const char *fmt, ...)
ac27a0ec 481{
0ff2ea7d 482 struct va_format vaf;
ac27a0ec
DK		 483 va_list args;
484
0db1ff22
TT		 485 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
486 return;
487
ccf0f32a 488 trace_ext4_error(sb, function, line);
efbed4dc
TT		 489 if (ext4_error_ratelimit(sb)) {
490 va_start(args, fmt);
491 vaf.fmt = fmt;
492 vaf.va = &args;
493 printk(KERN_CRIT
494 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
495 sb->s_id, function, line, current->comm, &vaf);
496 va_end(args);
497 }
f3fc0210 498 save_error_info(sb, function, line);
617ba13b 499 ext4_handle_error(sb);
ac27a0ec
DK		 500}
501
e7c96e8e
JP		 502void __ext4_error_inode(struct inode *inode, const char *function,
503 unsigned int line, ext4_fsblk_t block,
504 const char *fmt, ...)
273df556
FM		 505{
506 va_list args;
f7c21177 507 struct va_format vaf;
1c13d5c0 508 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 509
0db1ff22
TT		 510 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
511 return;
512
ccf0f32a 513 trace_ext4_error(inode->i_sb, function, line);
1c13d5c0
TT		 514 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
515 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 516 if (ext4_error_ratelimit(inode->i_sb)) {
517 va_start(args, fmt);
518 vaf.fmt = fmt;
519 vaf.va = &args;
520 if (block)
521 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
522 "inode #%lu: block %llu: comm %s: %pV\n",
523 inode->i_sb->s_id, function, line, inode->i_ino,
524 block, current->comm, &vaf);
525 else
526 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
527 "inode #%lu: comm %s: %pV\n",
528 inode->i_sb->s_id, function, line, inode->i_ino,
529 current->comm, &vaf);
530 va_end(args);
531 }
1c13d5c0 532 save_error_info(inode->i_sb, function, line);
273df556
FM		 533 ext4_handle_error(inode->i_sb);
534}
535
e7c96e8e
JP		 536void __ext4_error_file(struct file *file, const char *function,
537 unsigned int line, ext4_fsblk_t block,
538 const char *fmt, ...)
273df556
FM		 539{
540 va_list args;
f7c21177 541 struct va_format vaf;
1c13d5c0 542 struct ext4_super_block *es;
496ad9aa 543 struct inode *inode = file_inode(file);
273df556
FM		 544 char pathname[80], *path;
545
0db1ff22
TT		 546 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
547 return;
548
ccf0f32a 549 trace_ext4_error(inode->i_sb, function, line);
1c13d5c0
TT		 550 es = EXT4_SB(inode->i_sb)->s_es;
551 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 552 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 553 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 554 if (IS_ERR(path))
555 path = "(unknown)";
556 va_start(args, fmt);
557 vaf.fmt = fmt;
558 vaf.va = &args;
559 if (block)
560 printk(KERN_CRIT
561 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
562 "block %llu: comm %s: path %s: %pV\n",
563 inode->i_sb->s_id, function, line, inode->i_ino,
564 block, current->comm, path, &vaf);
565 else
566 printk(KERN_CRIT
567 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
568 "comm %s: path %s: %pV\n",
569 inode->i_sb->s_id, function, line, inode->i_ino,
570 current->comm, path, &vaf);
571 va_end(args);
572 }
1c13d5c0 573 save_error_info(inode->i_sb, function, line);
273df556
FM		 574 ext4_handle_error(inode->i_sb);
575}
576
722887dd
TT		 577const char *ext4_decode_error(struct super_block *sb, int errno,
578 char nbuf[16])
ac27a0ec
DK		 579{
580 char *errstr = NULL;
581
582 switch (errno) {
6a797d27
DW		 583 case -EFSCORRUPTED:
584 errstr = "Corrupt filesystem";
585 break;
586 case -EFSBADCRC:
587 errstr = "Filesystem failed CRC";
588 break;
ac27a0ec
DK		 589 case -EIO:
590 errstr = "IO failure";
591 break;
592 case -ENOMEM:
593 errstr = "Out of memory";
594 break;
595 case -EROFS:
78f1ddbb
TT		 596 if (!sb || (EXT4_SB(sb)->s_journal &&
597 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 598 errstr = "Journal has aborted";
599 else
600 errstr = "Readonly filesystem";
601 break;
602 default:
603 /* If the caller passed in an extra buffer for unknown
604 * errors, textualise them now. Else we just return
605 * NULL. */
606 if (nbuf) {
607 /* Check for truncated error codes... */
608 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
609 errstr = nbuf;
610 }
611 break;
612 }
613
614 return errstr;
615}
616
878520ac
TT		 617void ext4_set_errno(struct super_block *sb, int err)
618{
619 if (err < 0)
620 err = -err;
621
622 switch (err) {
623 case EIO:
624 err = EXT4_ERR_EIO;
625 break;
626 case ENOMEM:
627 err = EXT4_ERR_ENOMEM;
628 break;
629 case EFSBADCRC:
630 err = EXT4_ERR_EFSBADCRC;
631 break;
632 case EFSCORRUPTED:
633 err = EXT4_ERR_EFSCORRUPTED;
634 break;
635 case ENOSPC:
636 err = EXT4_ERR_ENOSPC;
637 break;
638 case ENOKEY:
639 err = EXT4_ERR_ENOKEY;
640 break;
641 case EROFS:
642 err = EXT4_ERR_EROFS;
643 break;
644 case EFBIG:
645 err = EXT4_ERR_EFBIG;
646 break;
647 case EEXIST:
648 err = EXT4_ERR_EEXIST;
649 break;
650 case ERANGE:
651 err = EXT4_ERR_ERANGE;
652 break;
653 case EOVERFLOW:
654 err = EXT4_ERR_EOVERFLOW;
655 break;
656 case EBUSY:
657 err = EXT4_ERR_EBUSY;
658 break;
659 case ENOTDIR:
660 err = EXT4_ERR_ENOTDIR;
661 break;
662 case ENOTEMPTY:
663 err = EXT4_ERR_ENOTEMPTY;
664 break;
665 case ESHUTDOWN:
666 err = EXT4_ERR_ESHUTDOWN;
667 break;
668 case EFAULT:
669 err = EXT4_ERR_EFAULT;
670 break;
671 default:
672 err = EXT4_ERR_UNKNOWN;
673 }
674 EXT4_SB(sb)->s_es->s_last_error_errcode = err;
675}
676
617ba13b 677/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 678 * automatically and invokes the appropriate error response. */
679
c398eda0
TT		 680void __ext4_std_error(struct super_block *sb, const char *function,
681 unsigned int line, int errno)
ac27a0ec
DK		 682{
683 char nbuf[16];
684 const char *errstr;
685
0db1ff22
TT		 686 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
687 return;
688
ac27a0ec
DK		 689 /* Special case: if the error is EROFS, and we're not already
690 * inside a transaction, then there's really no point in logging
691 * an error. */
bc98a42c 692 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
ac27a0ec
DK		 693 return;
694
efbed4dc
TT		 695 if (ext4_error_ratelimit(sb)) {
696 errstr = ext4_decode_error(sb, errno, nbuf);
697 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
698 sb->s_id, function, line, errstr);
699 }
ac27a0ec 700
878520ac 701 ext4_set_errno(sb, -errno);
efbed4dc 702 save_error_info(sb, function, line);
617ba13b 703 ext4_handle_error(sb);
ac27a0ec
DK		 704}
705
706/*
617ba13b 707 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 708 * abort function may be used to deal with unrecoverable failures such
709 * as journal IO errors or ENOMEM at a critical moment in log management.
710 *
711 * We unconditionally force the filesystem into an ABORT|READONLY state,
712 * unless the error response on the fs has been set to panic in which
713 * case we take the easy way out and panic immediately.
714 */
715
c67d859e 716void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 717 unsigned int line, const char *fmt, ...)
ac27a0ec 718{
651e1c3b 719 struct va_format vaf;
ac27a0ec
DK		 720 va_list args;
721
0db1ff22
TT		 722 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
723 return;
724
1c13d5c0 725 save_error_info(sb, function, line);
ac27a0ec 726 va_start(args, fmt);
651e1c3b
JP		 727 vaf.fmt = fmt;
728 vaf.va = &args;
729 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
730 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 731 va_end(args);
732
bc98a42c 733 if (sb_rdonly(sb) == 0) {
1c13d5c0 734 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 735 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 736 /*
737 * Make sure updated value of ->s_mount_flags will be visible
738 * before ->s_flags update
739 */
740 smp_wmb();
1751e8a6 741 sb->s_flags |= SB_RDONLY;
1c13d5c0
TT		 742 if (EXT4_SB(sb)->s_journal)
743 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
744 save_error_info(sb, function, line);
745 }
2c1d0e36 746 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
4327ba52
DJ		 747 if (EXT4_SB(sb)->s_journal &&
748 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
749 return;
617ba13b 750 panic("EXT4-fs panic from previous error\n");
4327ba52 751 }
ac27a0ec
DK		 752}
753
e7c96e8e
JP		 754void __ext4_msg(struct super_block *sb,
755 const char *prefix, const char *fmt, ...)
b31e1552 756{
0ff2ea7d 757 struct va_format vaf;
b31e1552
ES		 758 va_list args;
759
efbed4dc
TT		 760 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
761 return;
762
b31e1552 763 va_start(args, fmt);
0ff2ea7d
JP		 764 vaf.fmt = fmt;
765 vaf.va = &args;
766 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 767 va_end(args);
768}
769
b03a2f7e
AD		 770#define ext4_warning_ratelimit(sb) \
771 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
772 "EXT4-fs warning")
773
12062ddd 774void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 775 unsigned int line, const char *fmt, ...)
ac27a0ec 776{
0ff2ea7d 777 struct va_format vaf;
ac27a0ec
DK		 778 va_list args;
779
b03a2f7e 780 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 781 return;
782
ac27a0ec 783 va_start(args, fmt);
0ff2ea7d
JP		 784 vaf.fmt = fmt;
785 vaf.va = &args;
786 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
787 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 788 va_end(args);
789}
790
b03a2f7e
AD		 791void __ext4_warning_inode(const struct inode *inode, const char *function,
792 unsigned int line, const char *fmt, ...)
793{
794 struct va_format vaf;
795 va_list args;
796
797 if (!ext4_warning_ratelimit(inode->i_sb))
798 return;
799
800 va_start(args, fmt);
801 vaf.fmt = fmt;
802 vaf.va = &args;
803 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
804 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
805 function, line, inode->i_ino, current->comm, &vaf);
806 va_end(args);
807}
808
e29136f8
TT		 809void __ext4_grp_locked_error(const char *function, unsigned int line,
810 struct super_block *sb, ext4_group_t grp,
811 unsigned long ino, ext4_fsblk_t block,
812 const char *fmt, ...)
5d1b1b3f
AK		 813__releases(bitlock)
814__acquires(bitlock)
815{
0ff2ea7d 816 struct va_format vaf;
5d1b1b3f
AK		 817 va_list args;
818 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
819
0db1ff22
TT		 820 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
821 return;
822
ccf0f32a 823 trace_ext4_error(sb, function, line);
1c13d5c0
TT		 824 es->s_last_error_ino = cpu_to_le32(ino);
825 es->s_last_error_block = cpu_to_le64(block);
826 __save_error_info(sb, function, line);
0ff2ea7d 827
efbed4dc
TT		 828 if (ext4_error_ratelimit(sb)) {
829 va_start(args, fmt);
830 vaf.fmt = fmt;
831 vaf.va = &args;
832 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
833 sb->s_id, function, line, grp);
834 if (ino)
835 printk(KERN_CONT "inode %lu: ", ino);
836 if (block)
837 printk(KERN_CONT "block %llu:",
838 (unsigned long long) block);
839 printk(KERN_CONT "%pV\n", &vaf);
840 va_end(args);
841 }
5d1b1b3f 842
327eaf73
TT		 843 if (test_opt(sb, WARN_ON_ERROR))
844 WARN_ON_ONCE(1);
845
5d1b1b3f 846 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 847 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 848 return;
849 }
1c13d5c0 850
5d1b1b3f 851 ext4_unlock_group(sb, grp);
06f29cc8 852 ext4_commit_super(sb, 1);
5d1b1b3f
AK		 853 ext4_handle_error(sb);
854 /*
855 * We only get here in the ERRORS_RO case; relocking the group
856 * may be dangerous, but nothing bad will happen since the
857 * filesystem will have already been marked read/only and the
858 * journal has been aborted. We return 1 as a hint to callers
859 * who might what to use the return value from
25985edc 860 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 861 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
862 * aggressively from the ext4 function in question, with a
863 * more appropriate error code.
864 */
865 ext4_lock_group(sb, grp);
866 return;
867}
868
db79e6d1
WS		 869void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
870 ext4_group_t group,
871 unsigned int flags)
872{
873 struct ext4_sb_info *sbi = EXT4_SB(sb);
874 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
875 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
9af0b3d1
WS		 876 int ret;
877
878 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
879 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
880 &grp->bb_state);
881 if (!ret)
882 percpu_counter_sub(&sbi->s_freeclusters_counter,
883 grp->bb_free);
db79e6d1
WS		 884 }
885
9af0b3d1
WS		 886 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
887 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
888 &grp->bb_state);
889 if (!ret && gdp) {
db79e6d1
WS		 890 int count;
891
892 count = ext4_free_inodes_count(sb, gdp);
893 percpu_counter_sub(&sbi->s_freeinodes_counter,
894 count);
895 }
db79e6d1
WS		 896 }
897}
898
617ba13b 899void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 900{
617ba13b 901 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 902
617ba13b 903 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 904 return;
905
12062ddd 906 ext4_warning(sb,
ac27a0ec
DK		 907 "updating to rev %d because of new feature flag, "
908 "running e2fsck is recommended",
617ba13b 909 EXT4_DYNAMIC_REV);
ac27a0ec 910
617ba13b
MC		 911 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
912 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
913 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 914 /* leave es->s_feature_*compat flags alone */
915 /* es->s_uuid will be set by e2fsck if empty */
916
917 /*
918 * The rest of the superblock fields should be zero, and if not it
919 * means they are likely already in use, so leave them alone. We
920 * can leave it up to e2fsck to clean up any inconsistencies there.
921 */
922}
923
924/*
925 * Open the external journal device
926 */
b31e1552 927static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 928{
929 struct block_device *bdev;
930 char b[BDEVNAME_SIZE];
931
d4d77629 932 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 933 if (IS_ERR(bdev))
934 goto fail;
935 return bdev;
936
937fail:
b31e1552 938 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 939 __bdevname(dev, b), PTR_ERR(bdev));
940 return NULL;
941}
942
943/*
944 * Release the journal device
945 */
4385bab1 946static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 947{
4385bab1 948 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 949}
950
4385bab1 951static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 952{
953 struct block_device *bdev;
ac27a0ec
DK		 954 bdev = sbi->journal_bdev;
955 if (bdev) {
4385bab1 956 ext4_blkdev_put(bdev);
ac27a0ec
DK		 957 sbi->journal_bdev = NULL;
958 }
ac27a0ec
DK		 959}
960
961static inline struct inode *orphan_list_entry(struct list_head *l)
962{
617ba13b 963 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 964}
965
617ba13b 966static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 967{
968 struct list_head *l;
969
b31e1552
ES		 970 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
971 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 972
973 printk(KERN_ERR "sb_info orphan list:\n");
974 list_for_each(l, &sbi->s_orphan) {
975 struct inode *inode = orphan_list_entry(l);
976 printk(KERN_ERR " "
977 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
978 inode->i_sb->s_id, inode->i_ino, inode,
979 inode->i_mode, inode->i_nlink,
980 NEXT_ORPHAN(inode));
981 }
982}
983
957153fc
JK		 984#ifdef CONFIG_QUOTA
985static int ext4_quota_off(struct super_block *sb, int type);
986
987static inline void ext4_quota_off_umount(struct super_block *sb)
988{
989 int type;
990
964edf66
JK		 991 /* Use our quota_off function to clear inode flags etc. */
992 for (type = 0; type < EXT4_MAXQUOTAS; type++)
993 ext4_quota_off(sb, type);
957153fc 994}
33458eab
TT		 995
996/*
997 * This is a helper function which is used in the mount/remount
998 * codepaths (which holds s_umount) to fetch the quota file name.
999 */
1000static inline char *get_qf_name(struct super_block *sb,
1001 struct ext4_sb_info *sbi,
1002 int type)
1003{
1004 return rcu_dereference_protected(sbi->s_qf_names[type],
1005 lockdep_is_held(&sb->s_umount));
1006}
957153fc
JK		 1007#else
1008static inline void ext4_quota_off_umount(struct super_block *sb)
1009{
1010}
1011#endif
1012
2b2d6d01 1013static void ext4_put_super(struct super_block *sb)
ac27a0ec 1014{
617ba13b
MC		 1015 struct ext4_sb_info *sbi = EXT4_SB(sb);
1016 struct ext4_super_block *es = sbi->s_es;
1d0c3924 1017 struct buffer_head **group_desc;
7c990728 1018 struct flex_groups **flex_groups;
97abd7d4 1019 int aborted = 0;
ef2cabf7 1020 int i, err;
ac27a0ec 1021
857ac889 1022 ext4_unregister_li_request(sb);
957153fc 1023 ext4_quota_off_umount(sb);
e0ccfd95 1024
2e8fa54e 1025 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 1026
0390131b 1027 if (sbi->s_journal) {
97abd7d4 1028 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 1029 err = jbd2_journal_destroy(sbi->s_journal);
1030 sbi->s_journal = NULL;
878520ac
TT		 1031 if ((err < 0) && !aborted) {
1032 ext4_set_errno(sb, -err);
c67d859e 1033 ext4_abort(sb, "Couldn't clean up the journal");
878520ac 1034 }
0390131b 1035 }
d4edac31 1036
ebd173be 1037 ext4_unregister_sysfs(sb);
d3922a77 1038 ext4_es_unregister_shrinker(sbi);
9105bb14 1039 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 1040 ext4_release_system_zone(sb);
1041 ext4_mb_release(sb);
1042 ext4_ext_release(sb);
d4edac31 1043
bc98a42c 1044 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 1045 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 1046 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 1047 }
bc98a42c 1048 if (!sb_rdonly(sb))
a8e25a83
AB		 1049 ext4_commit_super(sb, 1);
1050
1d0c3924
TT		 1051 rcu_read_lock();
1052 group_desc = rcu_dereference(sbi->s_group_desc);
ac27a0ec 1053 for (i = 0; i < sbi->s_gdb_count; i++)
1d0c3924
TT		 1054 brelse(group_desc[i]);
1055 kvfree(group_desc);
7c990728
SJS		 1056 flex_groups = rcu_dereference(sbi->s_flex_groups);
1057 if (flex_groups) {
1058 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1059 kvfree(flex_groups[i]);
1060 kvfree(flex_groups);
1061 }
1d0c3924 1062 rcu_read_unlock();
57042651 1063 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 1064 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1065 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 1066 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
bbd55937 1067 percpu_free_rwsem(&sbi->s_writepages_rwsem);
ac27a0ec 1068#ifdef CONFIG_QUOTA
a2d4a646 1069 for (i = 0; i < EXT4_MAXQUOTAS; i++)
33458eab 1070 kfree(get_qf_name(sb, sbi, i));
ac27a0ec
DK		 1071#endif
1072
1073 /* Debugging code just in case the in-memory inode orphan list
1074 * isn't empty. The on-disk one can be non-empty if we've
1075 * detected an error and taken the fs readonly, but the
1076 * in-memory list had better be clean by this point. */
1077 if (!list_empty(&sbi->s_orphan))
1078 dump_orphan_list(sb, sbi);
1079 J_ASSERT(list_empty(&sbi->s_orphan));
1080
89d96a6f 1081 sync_blockdev(sb->s_bdev);
f98393a6 1082 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 1083 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
1084 /*
1085 * Invalidate the journal device's buffers. We don't want them
1086 * floating about in memory - the physical journal device may
1087 * hotswapped, and it breaks the `ro-after' testing code.
1088 */
1089 sync_blockdev(sbi->journal_bdev);
f98393a6 1090 invalidate_bdev(sbi->journal_bdev);
617ba13b 1091 ext4_blkdev_remove(sbi);
ac27a0ec 1092 }
50c15df6
CX		 1093
1094 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1095 sbi->s_ea_inode_cache = NULL;
1096
1097 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1098 sbi->s_ea_block_cache = NULL;
1099
c5e06d10
JL		 1100 if (sbi->s_mmp_tsk)
1101 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 1102 brelse(sbi->s_sbh);
ac27a0ec 1103 sb->s_fs_info = NULL;
3197ebdb
TT		 1104 /*
1105 * Now that we are completely done shutting down the
1106 * superblock, we need to actually destroy the kobject.
1107 */
3197ebdb
TT		 1108 kobject_put(&sbi->s_kobj);
1109 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 1110 if (sbi->s_chksum_driver)
1111 crypto_free_shash(sbi->s_chksum_driver);
705895b6 1112 kfree(sbi->s_blockgroup_lock);
5e405595 1113 fs_put_dax(sbi->s_daxdev);
c83ad55e
GKB		 1114#ifdef CONFIG_UNICODE
1115 utf8_unload(sbi->s_encoding);
1116#endif
ac27a0ec 1117 kfree(sbi);
ac27a0ec
DK		 1118}
1119
e18b890b 1120static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 1121
1122/*
1123 * Called inside transaction, so use GFP_NOFS
1124 */
617ba13b 1125static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 1126{
617ba13b 1127 struct ext4_inode_info *ei;
ac27a0ec 1128
e6b4f8da 1129 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 1130 if (!ei)
1131 return NULL;
0b8e58a1 1132
ee73f9a5 1133 inode_set_iversion(&ei->vfs_inode, 1);
202ee5df 1134 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 1135 INIT_LIST_HEAD(&ei->i_prealloc_list);
1136 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 1137 ext4_es_init_tree(&ei->i_es_tree);
1138 rwlock_init(&ei->i_es_lock);
edaa53ca 1139 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 1140 ei->i_es_all_nr = 0;
edaa53ca 1141 ei->i_es_shk_nr = 0;
dd475925 1142 ei->i_es_shrink_lblk = 0;
d2a17637 1143 ei->i_reserved_data_blocks = 0;
d2a17637 1144 spin_lock_init(&(ei->i_block_reservation_lock));
1dc0aa46 1145 ext4_init_pending_tree(&ei->i_pending_tree);
a9e7f447
DM		 1146#ifdef CONFIG_QUOTA
1147 ei->i_reserved_quota = 0;
96c7e0d9 1148 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 1149#endif
8aefcd55 1150 ei->jinode = NULL;
2e8fa54e 1151 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 1152 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 1153 ei->i_sync_tid = 0;
1154 ei->i_datasync_tid = 0;
e27f41e1 1155 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 1156 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 1157 return &ei->vfs_inode;
1158}
1159
7ff9c073
TT		 1160static int ext4_drop_inode(struct inode *inode)
1161{
1162 int drop = generic_drop_inode(inode);
1163
29b3692e
EB		 1164 if (!drop)
1165 drop = fscrypt_drop_inode(inode);
1166
7ff9c073
TT		 1167 trace_ext4_drop_inode(inode, drop);
1168 return drop;
1169}
1170
94053139 1171static void ext4_free_in_core_inode(struct inode *inode)
fa0d7e3d 1172{
2c58d548 1173 fscrypt_free_inode(inode);
fa0d7e3d
NP		 1174 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1175}
1176
617ba13b 1177static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1178{
9f7dd93d 1179 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1180 ext4_msg(inode->i_sb, KERN_ERR,
1181 "Inode %lu (%p): orphan list check failed!",
1182 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1183 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1184 EXT4_I(inode), sizeof(struct ext4_inode_info),
1185 true);
1186 dump_stack();
1187 }
ac27a0ec
DK		 1188}
1189
51cc5068 1190static void init_once(void *foo)
ac27a0ec 1191{
617ba13b 1192 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1193
a35afb83 1194 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1195 init_rwsem(&ei->xattr_sem);
0e855ac8 1196 init_rwsem(&ei->i_data_sem);
ea3d7209 1197 init_rwsem(&ei->i_mmap_sem);
a35afb83 1198 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1199}
1200
e67bc2b3 1201static int __init init_inodecache(void)
ac27a0ec 1202{
f8dd7c70
DW		 1203 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1204 sizeof(struct ext4_inode_info), 0,
1205 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1206 SLAB_ACCOUNT),
1207 offsetof(struct ext4_inode_info, i_data),
1208 sizeof_field(struct ext4_inode_info, i_data),
1209 init_once);
617ba13b 1210 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1211 return -ENOMEM;
1212 return 0;
1213}
1214
1215static void destroy_inodecache(void)
1216{
8c0a8537
KS		 1217 /*
1218 * Make sure all delayed rcu free inodes are flushed before we
1219 * destroy cache.
1220 */
1221 rcu_barrier();
617ba13b 1222 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1223}
1224
0930fcc1 1225void ext4_clear_inode(struct inode *inode)
ac27a0ec 1226{
0930fcc1 1227 invalidate_inode_buffers(inode);
dbd5768f 1228 clear_inode(inode);
c2ea3fde 1229 ext4_discard_preallocations(inode);
51865fda 1230 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
f4c2d372 1231 dquot_drop(inode);
8aefcd55
TT		 1232 if (EXT4_I(inode)->jinode) {
1233 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1234 EXT4_I(inode)->jinode);
1235 jbd2_free_inode(EXT4_I(inode)->jinode);
1236 EXT4_I(inode)->jinode = NULL;
1237 }
3d204e24 1238 fscrypt_put_encryption_info(inode);
c93d8f88 1239 fsverity_cleanup_inode(inode);
ac27a0ec
DK		 1240}
1241
1b961ac0 1242static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1243 u64 ino, u32 generation)
ac27a0ec 1244{
ac27a0ec 1245 struct inode *inode;
ac27a0ec 1246
8a363970 1247 /*
ac27a0ec
DK		 1248 * Currently we don't know the generation for parent directory, so
1249 * a generation of 0 means "accept any"
1250 */
8a363970 1251 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1d1fe1ee
DH		 1252 if (IS_ERR(inode))
1253 return ERR_CAST(inode);
1254 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1255 iput(inode);
1256 return ERR_PTR(-ESTALE);
1257 }
1b961ac0
CH		 1258
1259 return inode;
1260}
1261
1262static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1263 int fh_len, int fh_type)
1b961ac0
CH		 1264{
1265 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1266 ext4_nfs_get_inode);
1267}
1268
1269static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1270 int fh_len, int fh_type)
1b961ac0
CH		 1271{
1272 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1273 ext4_nfs_get_inode);
ac27a0ec
DK		 1274}
1275
fde87268
TT		 1276static int ext4_nfs_commit_metadata(struct inode *inode)
1277{
1278 struct writeback_control wbc = {
1279 .sync_mode = WB_SYNC_ALL
1280 };
1281
1282 trace_ext4_nfs_commit_metadata(inode);
1283 return ext4_write_inode(inode, &wbc);
1284}
1285
c39a7f84
TO		 1286/*
1287 * Try to release metadata pages (indirect blocks, directories) which are
1288 * mapped via the block device. Since these pages could have journal heads
1289 * which would prevent try_to_free_buffers() from freeing them, we must use
1290 * jbd2 layer's try_to_free_buffers() function to release them.
1291 */
0b8e58a1
AD		 1292static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1293 gfp_t wait)
c39a7f84
TO		 1294{
1295 journal_t *journal = EXT4_SB(sb)->s_journal;
1296
1297 WARN_ON(PageChecked(page));
1298 if (!page_has_buffers(page))
1299 return 0;
1300 if (journal)
1301 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1302 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1303 return try_to_free_buffers(page);
1304}
1305
643fa961 1306#ifdef CONFIG_FS_ENCRYPTION
a7550b30
JK		 1307static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1308{
1309 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1310 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1311}
1312
a7550b30
JK		 1313static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1314 void *fs_data)
1315{
2f8f5e76 1316 handle_t *handle = fs_data;
c1a5d5f6 1317 int res, res2, credits, retries = 0;
2f8f5e76 1318
9ce0151a
EB		 1319 /*
1320 * Encrypting the root directory is not allowed because e2fsck expects
1321 * lost+found to exist and be unencrypted, and encrypting the root
1322 * directory would imply encrypting the lost+found directory as well as
1323 * the filename "lost+found" itself.
1324 */
1325 if (inode->i_ino == EXT4_ROOT_INO)
1326 return -EPERM;
2f8f5e76 1327
7d3e06a8
RZ		 1328 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1329 return -EINVAL;
1330
94840e3c
EB		 1331 res = ext4_convert_inline_data(inode);
1332 if (res)
1333 return res;
1334
2f8f5e76
EB		 1335 /*
1336 * If a journal handle was specified, then the encryption context is
1337 * being set on a new inode via inheritance and is part of a larger
1338 * transaction to create the inode. Otherwise the encryption context is
1339 * being set on an existing inode in its own transaction. Only in the
1340 * latter case should the "retry on ENOSPC" logic be used.
1341 */
a7550b30 1342
2f8f5e76
EB		 1343 if (handle) {
1344 res = ext4_xattr_set_handle(handle, inode,
1345 EXT4_XATTR_INDEX_ENCRYPTION,
1346 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1347 ctx, len, 0);
a7550b30
JK		 1348 if (!res) {
1349 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1350 ext4_clear_inode_state(inode,
1351 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b 1352 /*
2ee6a576
EB		 1353 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1354 * S_DAX may be disabled
a3caa24b
JK		 1355 */
1356 ext4_set_inode_flags(inode);
a7550b30
JK		 1357 }
1358 return res;
1359 }
1360
b8cb5a54
TE		 1361 res = dquot_initialize(inode);
1362 if (res)
1363 return res;
2f8f5e76 1364retry:
af65207c
TE		 1365 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1366 &credits);
dec214d0
TE		 1367 if (res)
1368 return res;
1369
c1a5d5f6 1370 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
a7550b30
JK		 1371 if (IS_ERR(handle))
1372 return PTR_ERR(handle);
1373
2f8f5e76
EB		 1374 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1375 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1376 ctx, len, 0);
a7550b30
JK		 1377 if (!res) {
1378 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
2ee6a576
EB		 1379 /*
1380 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1381 * S_DAX may be disabled
1382 */
a3caa24b 1383 ext4_set_inode_flags(inode);
a7550b30
JK		 1384 res = ext4_mark_inode_dirty(handle, inode);
1385 if (res)
1386 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1387 }
1388 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1389
1390 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1391 goto retry;
a7550b30
JK		 1392 if (!res)
1393 res = res2;
1394 return res;
1395}
1396
c250b7dd 1397static bool ext4_dummy_context(struct inode *inode)
a7550b30
JK		 1398{
1399 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1400}
1401
b925acb8
EB		 1402static bool ext4_has_stable_inodes(struct super_block *sb)
1403{
1404 return ext4_has_feature_stable_inodes(sb);
1405}
1406
1407static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1408 int *ino_bits_ret, int *lblk_bits_ret)
1409{
1410 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1411 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1412}
1413
6f69f0ed 1414static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1415 .key_prefix = "ext4:",
a7550b30 1416 .get_context = ext4_get_context,
a7550b30
JK		 1417 .set_context = ext4_set_context,
1418 .dummy_context = ext4_dummy_context,
a7550b30 1419 .empty_dir = ext4_empty_dir,
e12ee683 1420 .max_namelen = EXT4_NAME_LEN,
b925acb8
EB		 1421 .has_stable_inodes = ext4_has_stable_inodes,
1422 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
a7550b30 1423};
a7550b30
JK		 1424#endif
1425
ac27a0ec 1426#ifdef CONFIG_QUOTA
d6006186 1427static const char * const quotatypes[] = INITQFNAMES;
689c958c 1428#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1429
617ba13b
MC		 1430static int ext4_write_dquot(struct dquot *dquot);
1431static int ext4_acquire_dquot(struct dquot *dquot);
1432static int ext4_release_dquot(struct dquot *dquot);
1433static int ext4_mark_dquot_dirty(struct dquot *dquot);
1434static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1435static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1436 const struct path *path);
617ba13b
MC		 1437static int ext4_quota_on_mount(struct super_block *sb, int type);
1438static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1439 size_t len, loff_t off);
617ba13b 1440static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1441 const char *data, size_t len, loff_t off);
7c319d32
AK		 1442static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1443 unsigned int flags);
1444static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1445
96c7e0d9
JK		 1446static struct dquot **ext4_get_dquots(struct inode *inode)
1447{
1448 return EXT4_I(inode)->i_dquot;
1449}
1450
61e225dc 1451static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1452 .get_reserved_space = ext4_get_reserved_space,
1453 .write_dquot = ext4_write_dquot,
1454 .acquire_dquot = ext4_acquire_dquot,
1455 .release_dquot = ext4_release_dquot,
1456 .mark_dirty = ext4_mark_dquot_dirty,
1457 .write_info = ext4_write_info,
1458 .alloc_dquot = dquot_alloc,
1459 .destroy_dquot = dquot_destroy,
1460 .get_projid = ext4_get_projid,
1461 .get_inode_usage = ext4_get_inode_usage,
ebc11f7b 1462 .get_next_id = dquot_get_next_id,
ac27a0ec
DK		 1463};
1464
0d54b217 1465static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1466 .quota_on = ext4_quota_on,
ca0e05e4 1467 .quota_off = ext4_quota_off,
287a8095 1468 .quota_sync = dquot_quota_sync,
0a240339 1469 .get_state = dquot_get_state,
287a8095
CH		 1470 .set_info = dquot_set_dqinfo,
1471 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1472 .set_dqblk = dquot_set_dqblk,
1473 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1474};
1475#endif
1476
ee9b6d61 1477static const struct super_operations ext4_sops = {
617ba13b 1478 .alloc_inode = ext4_alloc_inode,
94053139 1479 .free_inode = ext4_free_in_core_inode,
617ba13b 1480 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1481 .write_inode = ext4_write_inode,
1482 .dirty_inode = ext4_dirty_inode,
7ff9c073 1483 .drop_inode = ext4_drop_inode,
0930fcc1 1484 .evict_inode = ext4_evict_inode,
617ba13b 1485 .put_super = ext4_put_super,
617ba13b 1486 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1487 .freeze_fs = ext4_freeze,
1488 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1489 .statfs = ext4_statfs,
1490 .remount_fs = ext4_remount,
617ba13b 1491 .show_options = ext4_show_options,
ac27a0ec 1492#ifdef CONFIG_QUOTA
617ba13b
MC		 1493 .quota_read = ext4_quota_read,
1494 .quota_write = ext4_quota_write,
96c7e0d9 1495 .get_dquots = ext4_get_dquots,
ac27a0ec 1496#endif
c39a7f84 1497 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1498};
1499
39655164 1500static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1501 .fh_to_dentry = ext4_fh_to_dentry,
1502 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1503 .get_parent = ext4_get_parent,
fde87268 1504 .commit_metadata = ext4_nfs_commit_metadata,
ac27a0ec
DK		 1505};
1506
1507enum {
1508 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1509 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1510 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1511 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1512 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1513 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1514 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1515 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1516 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1517 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1518 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1519 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1520 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
327eaf73
TT		 1521 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1522 Opt_nowarn_on_error, Opt_mblk_io_submit,
670e9875 1523 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1524 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1525 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1526 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1527 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1528 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
ac27a0ec
DK		 1529};
1530
a447c093 1531static const match_table_t tokens = {
ac27a0ec
DK		 1532 {Opt_bsd_df, "bsddf"},
1533 {Opt_minix_df, "minixdf"},
1534 {Opt_grpid, "grpid"},
1535 {Opt_grpid, "bsdgroups"},
1536 {Opt_nogrpid, "nogrpid"},
1537 {Opt_nogrpid, "sysvgroups"},
1538 {Opt_resgid, "resgid=%u"},
1539 {Opt_resuid, "resuid=%u"},
1540 {Opt_sb, "sb=%u"},
1541 {Opt_err_cont, "errors=continue"},
1542 {Opt_err_panic, "errors=panic"},
1543 {Opt_err_ro, "errors=remount-ro"},
1544 {Opt_nouid32, "nouid32"},
ac27a0ec 1545 {Opt_debug, "debug"},
72578c33
TT		 1546 {Opt_removed, "oldalloc"},
1547 {Opt_removed, "orlov"},
ac27a0ec
DK		 1548 {Opt_user_xattr, "user_xattr"},
1549 {Opt_nouser_xattr, "nouser_xattr"},
1550 {Opt_acl, "acl"},
1551 {Opt_noacl, "noacl"},
e3bb52ae 1552 {Opt_noload, "norecovery"},
5a916be1 1553 {Opt_noload, "noload"},
72578c33
TT		 1554 {Opt_removed, "nobh"},
1555 {Opt_removed, "bh"},
ac27a0ec 1556 {Opt_commit, "commit=%u"},
30773840
TT		 1557 {Opt_min_batch_time, "min_batch_time=%u"},
1558 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1559 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1560 {Opt_journal_path, "journal_path=%s"},
818d276c 1561 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1562 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1563 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1564 {Opt_abort, "abort"},
1565 {Opt_data_journal, "data=journal"},
1566 {Opt_data_ordered, "data=ordered"},
1567 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1568 {Opt_data_err_abort, "data_err=abort"},
1569 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1570 {Opt_offusrjquota, "usrjquota="},
1571 {Opt_usrjquota, "usrjquota=%s"},
1572 {Opt_offgrpjquota, "grpjquota="},
1573 {Opt_grpjquota, "grpjquota=%s"},
1574 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1575 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1576 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1577 {Opt_grpquota, "grpquota"},
1578 {Opt_noquota, "noquota"},
1579 {Opt_quota, "quota"},
1580 {Opt_usrquota, "usrquota"},
49da9392 1581 {Opt_prjquota, "prjquota"},
ac27a0ec 1582 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1583 {Opt_barrier, "barrier"},
1584 {Opt_nobarrier, "nobarrier"},
25ec56b5 1585 {Opt_i_version, "i_version"},
923ae0ff 1586 {Opt_dax, "dax"},
c9de560d 1587 {Opt_stripe, "stripe=%u"},
64769240 1588 {Opt_delalloc, "delalloc"},
327eaf73
TT		 1589 {Opt_warn_on_error, "warn_on_error"},
1590 {Opt_nowarn_on_error, "nowarn_on_error"},
a26f4992
TT		 1591 {Opt_lazytime, "lazytime"},
1592 {Opt_nolazytime, "nolazytime"},
670e9875 1593 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1594 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1595 {Opt_removed, "mblk_io_submit"},
1596 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1597 {Opt_block_validity, "block_validity"},
1598 {Opt_noblock_validity, "noblock_validity"},
240799cd 1599 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1600 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1601 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1602 {Opt_auto_da_alloc, "auto_da_alloc"},
1603 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc 1604 {Opt_dioread_nolock, "dioread_nolock"},
244adf64 1605 {Opt_dioread_lock, "nodioread_nolock"},
744692dc 1606 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1607 {Opt_discard, "discard"},
1608 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1609 {Opt_init_itable, "init_itable=%u"},
1610 {Opt_init_itable, "init_itable"},
1611 {Opt_noinit_itable, "noinit_itable"},
df981d03 1612 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1613 {Opt_test_dummy_encryption, "test_dummy_encryption"},
cdb7ee4c
TE		 1614 {Opt_nombcache, "nombcache"},
1615 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
c7198b9c
TT		 1616 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1617 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1618 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1619 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1620 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1621 {Opt_err, NULL},
ac27a0ec
DK		 1622};
1623
617ba13b 1624static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1625{
617ba13b 1626 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1627 char *options = (char *) *data;
1628
1629 if (!options || strncmp(options, "sb=", 3) != 0)
1630 return 1; /* Default location */
0b8e58a1 1631
ac27a0ec 1632 options += 3;
0b8e58a1 1633 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1634 sb_block = simple_strtoul(options, &options, 0);
1635 if (*options && *options != ',') {
4776004f 1636 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1637 (char *) *data);
1638 return 1;
1639 }
1640 if (*options == ',')
1641 options++;
1642 *data = (void *) options;
0b8e58a1 1643
ac27a0ec
DK		 1644 return sb_block;
1645}
1646
b3881f74 1647#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1648static const char deprecated_msg[] =
1649 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1650 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1651
56c50f11
DM		 1652#ifdef CONFIG_QUOTA
1653static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1654{
1655 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 1656 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
03dafb5f 1657 int ret = -1;
56c50f11 1658
33458eab 1659 if (sb_any_quota_loaded(sb) && !old_qname) {
56c50f11
DM		 1660 ext4_msg(sb, KERN_ERR,
1661 "Cannot change journaled "
1662 "quota options when quota turned on");
57f73c2c 1663 return -1;
56c50f11 1664 }
e2b911c5 1665 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1666 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1667 "ignored when QUOTA feature is enabled");
1668 return 1;
262b4662 1669 }
56c50f11
DM		 1670 qname = match_strdup(args);
1671 if (!qname) {
1672 ext4_msg(sb, KERN_ERR,
1673 "Not enough memory for storing quotafile name");
57f73c2c 1674 return -1;
56c50f11 1675 }
33458eab
TT		 1676 if (old_qname) {
1677 if (strcmp(old_qname, qname) == 0)
03dafb5f
CG		 1678 ret = 1;
1679 else
1680 ext4_msg(sb, KERN_ERR,
1681 "%s quota file already specified",
1682 QTYPE2NAME(qtype));
1683 goto errout;
56c50f11 1684 }
03dafb5f 1685 if (strchr(qname, '/')) {
56c50f11
DM		 1686 ext4_msg(sb, KERN_ERR,
1687 "quotafile must be on filesystem root");
03dafb5f 1688 goto errout;
56c50f11 1689 }
33458eab 1690 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
fd8c37ec 1691 set_opt(sb, QUOTA);
56c50f11 1692 return 1;
03dafb5f
CG		 1693errout:
1694 kfree(qname);
1695 return ret;
56c50f11
DM		 1696}
1697
1698static int clear_qf_name(struct super_block *sb, int qtype)
1699{
1700
1701 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 1702 char *old_qname = get_qf_name(sb, sbi, qtype);
56c50f11 1703
33458eab 1704 if (sb_any_quota_loaded(sb) && old_qname) {
56c50f11
DM		 1705 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1706 " when quota turned on");
57f73c2c 1707 return -1;
56c50f11 1708 }
33458eab
TT		 1709 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1710 synchronize_rcu();
1711 kfree(old_qname);
56c50f11
DM		 1712 return 1;
1713}
1714#endif
1715
26092bf5
TT		 1716#define MOPT_SET 0x0001
1717#define MOPT_CLEAR 0x0002
1718#define MOPT_NOSUPPORT 0x0004
1719#define MOPT_EXPLICIT 0x0008
1720#define MOPT_CLEAR_ERR 0x0010
1721#define MOPT_GTE0 0x0020
ac27a0ec 1722#ifdef CONFIG_QUOTA
26092bf5
TT		 1723#define MOPT_Q 0
1724#define MOPT_QFMT 0x0040
1725#else
1726#define MOPT_Q MOPT_NOSUPPORT
1727#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1728#endif
26092bf5 1729#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1730#define MOPT_NO_EXT2 0x0100
1731#define MOPT_NO_EXT3 0x0200
1732#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1733#define MOPT_STRING 0x0400
26092bf5
TT		 1734
1735static const struct mount_opts {
1736 int token;
1737 int mount_opt;
1738 int flags;
1739} ext4_mount_opts[] = {
1740 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1741 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1742 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1743 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1744 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1745 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1746 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1747 MOPT_EXT4_ONLY | MOPT_SET},
1748 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1749 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1750 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1751 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1752 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1753 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1754 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1755 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1756 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1757 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
c6d3d56d
DW		 1758 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1759 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1760 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1761 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1762 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1763 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1764 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1765 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1766 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1767 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1768 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1769 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1770 MOPT_NO_EXT2},
8dc0aa8c 1771 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1772 MOPT_NO_EXT2},
26092bf5
TT		 1773 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1774 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1775 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1776 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1777 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1778 {Opt_commit, 0, MOPT_GTE0},
1779 {Opt_max_batch_time, 0, MOPT_GTE0},
1780 {Opt_min_batch_time, 0, MOPT_GTE0},
1781 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1782 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1783 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1784 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1785 {Opt_resuid, 0, MOPT_GTE0},
1786 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1787 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1788 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1789 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1790 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1791 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1792 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1793 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1794 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1795 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1796#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1797 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1798 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1799#else
26092bf5
TT		 1800 {Opt_acl, 0, MOPT_NOSUPPORT},
1801 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1802#endif
26092bf5
TT		 1803 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1804 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1805 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1806 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1807 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1808 MOPT_SET | MOPT_Q},
1809 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1810 MOPT_SET | MOPT_Q},
49da9392
JK		 1811 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1812 MOPT_SET | MOPT_Q},
26092bf5 1813 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1814 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1815 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1816 {Opt_usrjquota, 0, MOPT_Q},
1817 {Opt_grpjquota, 0, MOPT_Q},
1818 {Opt_offusrjquota, 0, MOPT_Q},
1819 {Opt_offgrpjquota, 0, MOPT_Q},
1820 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1821 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1822 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1823 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1824 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
cdb7ee4c 1825 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
26092bf5
TT		 1826 {Opt_err, 0, 0}
1827};
1828
c83ad55e
GKB		 1829#ifdef CONFIG_UNICODE
1830static const struct ext4_sb_encodings {
1831 __u16 magic;
1832 char *name;
1833 char *version;
1834} ext4_sb_encoding_map[] = {
1835 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1836};
1837
1838static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1839 const struct ext4_sb_encodings **encoding,
1840 __u16 *flags)
1841{
1842 __u16 magic = le16_to_cpu(es->s_encoding);
1843 int i;
1844
1845 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1846 if (magic == ext4_sb_encoding_map[i].magic)
1847 break;
1848
1849 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1850 return -EINVAL;
1851
1852 *encoding = &ext4_sb_encoding_map[i];
1853 *flags = le16_to_cpu(es->s_encoding_flags);
1854
1855 return 0;
1856}
1857#endif
1858
26092bf5
TT		 1859static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1860 substring_t *args, unsigned long *journal_devnum,
1861 unsigned int *journal_ioprio, int is_remount)
1862{
1863 struct ext4_sb_info *sbi = EXT4_SB(sb);
1864 const struct mount_opts *m;
08cefc7a
EB		 1865 kuid_t uid;
1866 kgid_t gid;
26092bf5
TT		 1867 int arg = 0;
1868
57f73c2c
TT		 1869#ifdef CONFIG_QUOTA
1870 if (token == Opt_usrjquota)
1871 return set_qf_name(sb, USRQUOTA, &args[0]);
1872 else if (token == Opt_grpjquota)
1873 return set_qf_name(sb, GRPQUOTA, &args[0]);
1874 else if (token == Opt_offusrjquota)
1875 return clear_qf_name(sb, USRQUOTA);
1876 else if (token == Opt_offgrpjquota)
1877 return clear_qf_name(sb, GRPQUOTA);
1878#endif
26092bf5 1879 switch (token) {
f7048605
TT		 1880 case Opt_noacl:
1881 case Opt_nouser_xattr:
1882 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1883 break;
26092bf5
TT		 1884 case Opt_sb:
1885 return 1; /* handled by get_sb_block() */
1886 case Opt_removed:
5f3633e3 1887 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1888 return 1;
26092bf5
TT		 1889 case Opt_abort:
1890 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1891 return 1;
1892 case Opt_i_version:
357fdad0 1893 sb->s_flags |= SB_I_VERSION;
26092bf5 1894 return 1;
a26f4992 1895 case Opt_lazytime:
1751e8a6 1896 sb->s_flags |= SB_LAZYTIME;
a26f4992
TT		 1897 return 1;
1898 case Opt_nolazytime:
1751e8a6 1899 sb->s_flags &= ~SB_LAZYTIME;
a26f4992 1900 return 1;
26092bf5
TT		 1901 }
1902
5f3633e3
JK		 1903 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1904 if (token == m->token)
1905 break;
1906
1907 if (m->token == Opt_err) {
1908 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1909 "or missing value", opt);
1910 return -1;
1911 }
1912
8dc0aa8c
TT		 1913 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1914 ext4_msg(sb, KERN_ERR,
1915 "Mount option \"%s\" incompatible with ext2", opt);
1916 return -1;
1917 }
1918 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1919 ext4_msg(sb, KERN_ERR,
1920 "Mount option \"%s\" incompatible with ext3", opt);
1921 return -1;
1922 }
1923
ad4eec61 1924 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1925 return -1;
1926 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1927 return -1;
c93cf2d7
DM		 1928 if (m->flags & MOPT_EXPLICIT) {
1929 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1930 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1931 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1932 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1933 } else
1934 return -1;
1935 }
5f3633e3
JK		 1936 if (m->flags & MOPT_CLEAR_ERR)
1937 clear_opt(sb, ERRORS_MASK);
1938 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1939 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1940 "options when quota turned on");
1941 return -1;
1942 }
1943
1944 if (m->flags & MOPT_NOSUPPORT) {
1945 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1946 } else if (token == Opt_commit) {
1947 if (arg == 0)
1948 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
9ba55543 1949 else if (arg > INT_MAX / HZ) {
1950 ext4_msg(sb, KERN_ERR,
1951 "Invalid commit interval %d, "
1952 "must be smaller than %d",
1953 arg, INT_MAX / HZ);
1954 return -1;
1955 }
5f3633e3 1956 sbi->s_commit_interval = HZ * arg;
670e9875 1957 } else if (token == Opt_debug_want_extra_isize) {
9803387c
TT		 1958 if ((arg & 1) ||
1959 (arg < 4) ||
1960 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
1961 ext4_msg(sb, KERN_ERR,
1962 "Invalid want_extra_isize %d", arg);
1963 return -1;
1964 }
670e9875 1965 sbi->s_want_extra_isize = arg;
5f3633e3 1966 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1967 sbi->s_max_batch_time = arg;
1968 } else if (token == Opt_min_batch_time) {
1969 sbi->s_min_batch_time = arg;
1970 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1971 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1972 ext4_msg(sb, KERN_ERR,
1973 "EXT4-fs: inode_readahead_blks must be "
1974 "0 or a power of 2 smaller than 2^31");
26092bf5 1975 return -1;
5f3633e3
JK		 1976 }
1977 sbi->s_inode_readahead_blks = arg;
1978 } else if (token == Opt_init_itable) {
1979 set_opt(sb, INIT_INODE_TABLE);
1980 if (!args->from)
1981 arg = EXT4_DEF_LI_WAIT_MULT;
1982 sbi->s_li_wait_mult = arg;
1983 } else if (token == Opt_max_dir_size_kb) {
1984 sbi->s_max_dir_size_kb = arg;
1985 } else if (token == Opt_stripe) {
1986 sbi->s_stripe = arg;
1987 } else if (token == Opt_resuid) {
1988 uid = make_kuid(current_user_ns(), arg);
1989 if (!uid_valid(uid)) {
1990 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1991 return -1;
1992 }
5f3633e3
JK		 1993 sbi->s_resuid = uid;
1994 } else if (token == Opt_resgid) {
1995 gid = make_kgid(current_user_ns(), arg);
1996 if (!gid_valid(gid)) {
1997 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1998 return -1;
1999 }
2000 sbi->s_resgid = gid;
2001 } else if (token == Opt_journal_dev) {
2002 if (is_remount) {
2003 ext4_msg(sb, KERN_ERR,
2004 "Cannot specify journal on remount");
2005 return -1;
2006 }
2007 *journal_devnum = arg;
ad4eec61
ES		 2008 } else if (token == Opt_journal_path) {
2009 char *journal_path;
2010 struct inode *journal_inode;
2011 struct path path;
2012 int error;
2013
2014 if (is_remount) {
2015 ext4_msg(sb, KERN_ERR,
2016 "Cannot specify journal on remount");
2017 return -1;
2018 }
2019 journal_path = match_strdup(&args[0]);
2020 if (!journal_path) {
2021 ext4_msg(sb, KERN_ERR, "error: could not dup "
2022 "journal device string");
2023 return -1;
2024 }
2025
2026 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2027 if (error) {
2028 ext4_msg(sb, KERN_ERR, "error: could not find "
2029 "journal device path: error %d", error);
2030 kfree(journal_path);
2031 return -1;
2032 }
2033
2b0143b5 2034 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 2035 if (!S_ISBLK(journal_inode->i_mode)) {
2036 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2037 "is not a block device", journal_path);
2038 path_put(&path);
2039 kfree(journal_path);
2040 return -1;
2041 }
2042
2043 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2044 path_put(&path);
2045 kfree(journal_path);
5f3633e3
JK		 2046 } else if (token == Opt_journal_ioprio) {
2047 if (arg > 7) {
2048 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2049 " (must be 0-7)");
2050 return -1;
2051 }
2052 *journal_ioprio =
2053 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447 2054 } else if (token == Opt_test_dummy_encryption) {
643fa961 2055#ifdef CONFIG_FS_ENCRYPTION
6ddb2447
TT		 2056 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
2057 ext4_msg(sb, KERN_WARNING,
2058 "Test dummy encryption mode enabled");
2059#else
2060 ext4_msg(sb, KERN_WARNING,
2061 "Test dummy encryption mount option ignored");
2062#endif
5f3633e3
JK		 2063 } else if (m->flags & MOPT_DATAJ) {
2064 if (is_remount) {
2065 if (!sbi->s_journal)
2066 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2067 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 2068 ext4_msg(sb, KERN_ERR,
26092bf5 2069 "Cannot change data mode on remount");
26092bf5 2070 return -1;
ac27a0ec 2071 }
26092bf5 2072 } else {
5f3633e3
JK		 2073 clear_opt(sb, DATA_FLAGS);
2074 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 2075 }
5f3633e3
JK		 2076#ifdef CONFIG_QUOTA
2077 } else if (m->flags & MOPT_QFMT) {
2078 if (sb_any_quota_loaded(sb) &&
2079 sbi->s_jquota_fmt != m->mount_opt) {
2080 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2081 "quota options when quota turned on");
2082 return -1;
2083 }
e2b911c5 2084 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 2085 ext4_msg(sb, KERN_INFO,
2086 "Quota format mount options ignored "
262b4662 2087 "when QUOTA feature is enabled");
c325a67c 2088 return 1;
262b4662 2089 }
5f3633e3 2090 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 2091#endif
923ae0ff 2092 } else if (token == Opt_dax) {
ef83b6e8
DW		 2093#ifdef CONFIG_FS_DAX
2094 ext4_msg(sb, KERN_WARNING,
2095 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
a92abd73 2096 sbi->s_mount_opt |= m->mount_opt;
ef83b6e8 2097#else
923ae0ff
RZ		 2098 ext4_msg(sb, KERN_INFO, "dax option not supported");
2099 return -1;
5f3633e3 2100#endif
7915a861
AN		 2101 } else if (token == Opt_data_err_abort) {
2102 sbi->s_mount_opt |= m->mount_opt;
2103 } else if (token == Opt_data_err_ignore) {
2104 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 2105 } else {
2106 if (!args->from)
2107 arg = 1;
2108 if (m->flags & MOPT_CLEAR)
2109 arg = !arg;
2110 else if (unlikely(!(m->flags & MOPT_SET))) {
2111 ext4_msg(sb, KERN_WARNING,
2112 "buggy handling of option %s", opt);
2113 WARN_ON(1);
2114 return -1;
2115 }
2116 if (arg != 0)
2117 sbi->s_mount_opt |= m->mount_opt;
2118 else
2119 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 2120 }
5f3633e3 2121 return 1;
26092bf5
TT		 2122}
2123
2124static int parse_options(char *options, struct super_block *sb,
2125 unsigned long *journal_devnum,
2126 unsigned int *journal_ioprio,
2127 int is_remount)
2128{
1e1a76ed 2129 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
33458eab 2130 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
26092bf5
TT		 2131 substring_t args[MAX_OPT_ARGS];
2132 int token;
2133
2134 if (!options)
2135 return 1;
2136
2137 while ((p = strsep(&options, ",")) != NULL) {
2138 if (!*p)
2139 continue;
2140 /*
2141 * Initialize args struct so we know whether arg was
2142 * found; some options take optional arguments.
2143 */
caecd0af 2144 args[0].to = args[0].from = NULL;
26092bf5
TT		 2145 token = match_token(p, tokens, args);
2146 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2147 journal_ioprio, is_remount) < 0)
2148 return 0;
ac27a0ec
DK		 2149 }
2150#ifdef CONFIG_QUOTA
49da9392
JK		 2151 /*
2152 * We do the test below only for project quotas. 'usrquota' and
2153 * 'grpquota' mount options are allowed even without quota feature
2154 * to support legacy quotas in quota files.
2155 */
2156 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2157 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2158 "Cannot enable project quota enforcement.");
2159 return 0;
2160 }
33458eab
TT		 2161 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2162 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2163 if (usr_qf_name || grp_qf_name) {
2164 if (test_opt(sb, USRQUOTA) && usr_qf_name)
fd8c37ec 2165 clear_opt(sb, USRQUOTA);
ac27a0ec 2166
33458eab 2167 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
fd8c37ec 2168 clear_opt(sb, GRPQUOTA);
ac27a0ec 2169
56c50f11 2170 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 2171 ext4_msg(sb, KERN_ERR, "old and new quota "
2172 "format mixing");
ac27a0ec
DK		 2173 return 0;
2174 }
2175
2176 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 2177 ext4_msg(sb, KERN_ERR, "journaled quota format "
2178 "not specified");
ac27a0ec
DK		 2179 return 0;
2180 }
ac27a0ec
DK		 2181 }
2182#endif
2183 return 1;
2184}
2185
2adf6da8
TT		 2186static inline void ext4_show_quota_options(struct seq_file *seq,
2187 struct super_block *sb)
2188{
2189#if defined(CONFIG_QUOTA)
2190 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 2191 char *usr_qf_name, *grp_qf_name;
2adf6da8
TT		 2192
2193 if (sbi->s_jquota_fmt) {
2194 char *fmtname = "";
2195
2196 switch (sbi->s_jquota_fmt) {
2197 case QFMT_VFS_OLD:
2198 fmtname = "vfsold";
2199 break;
2200 case QFMT_VFS_V0:
2201 fmtname = "vfsv0";
2202 break;
2203 case QFMT_VFS_V1:
2204 fmtname = "vfsv1";
2205 break;
2206 }
2207 seq_printf(seq, ",jqfmt=%s", fmtname);
2208 }
2209
33458eab
TT		 2210 rcu_read_lock();
2211 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2212 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2213 if (usr_qf_name)
2214 seq_show_option(seq, "usrjquota", usr_qf_name);
2215 if (grp_qf_name)
2216 seq_show_option(seq, "grpjquota", grp_qf_name);
2217 rcu_read_unlock();
2adf6da8
TT		 2218#endif
2219}
2220
5a916be1
TT		 2221static const char *token2str(int token)
2222{
50df9fd5 2223 const struct match_token *t;
5a916be1
TT		 2224
2225 for (t = tokens; t->token != Opt_err; t++)
2226 if (t->token == token && !strchr(t->pattern, '='))
2227 break;
2228 return t->pattern;
2229}
2230
2adf6da8
TT		 2231/*
2232 * Show an option if
2233 * - it's set to a non-default value OR
2234 * - if the per-sb default is different from the global default
2235 */
66acdcf4
TT		 2236static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2237 int nodefs)
2adf6da8 2238{
2adf6da8
TT		 2239 struct ext4_sb_info *sbi = EXT4_SB(sb);
2240 struct ext4_super_block *es = sbi->s_es;
68afa7e0 2241 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
5a916be1 2242 const struct mount_opts *m;
66acdcf4 2243 char sep = nodefs ? '\n' : ',';
2adf6da8 2244
66acdcf4
TT		 2245#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2246#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2247
2248 if (sbi->s_sb_block != 1)
5a916be1
TT		 2249 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2250
2251 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2252 int want_set = m->flags & MOPT_SET;
2253 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2254 (m->flags & MOPT_CLEAR_ERR))
2255 continue;
68afa7e0 2256 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2257 continue; /* skip if same as the default */
2258 if ((want_set &&
2259 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2260 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2261 continue; /* select Opt_noFoo vs Opt_Foo */
2262 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2263 }
5a916be1 2264
08cefc7a 2265 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2266 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2267 SEQ_OPTS_PRINT("resuid=%u",
2268 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2269 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2270 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2271 SEQ_OPTS_PRINT("resgid=%u",
2272 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2273 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2274 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2275 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2276 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2277 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2278 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2279 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2280 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2281 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2282 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2283 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2284 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2285 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
357fdad0 2286 if (sb->s_flags & SB_I_VERSION)
5a916be1 2287 SEQ_OPTS_PUTS("i_version");
66acdcf4 2288 if (nodefs || sbi->s_stripe)
5a916be1 2289 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0
TN		 2290 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2291 (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2292 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2293 SEQ_OPTS_PUTS("data=journal");
2294 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2295 SEQ_OPTS_PUTS("data=ordered");
2296 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2297 SEQ_OPTS_PUTS("data=writeback");
2298 }
66acdcf4
TT		 2299 if (nodefs ||
2300 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2301 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2302 sbi->s_inode_readahead_blks);
2adf6da8 2303
ceec0376 2304 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 2305 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2306 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2307 if (nodefs || sbi->s_max_dir_size_kb)
2308 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2309 if (test_opt(sb, DATA_ERR_ABORT))
2310 SEQ_OPTS_PUTS("data_err=abort");
338affb5
EB		 2311 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2312 SEQ_OPTS_PUTS("test_dummy_encryption");
2adf6da8
TT		 2313
2314 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2315 return 0;
2316}
2317
66acdcf4
TT		 2318static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2319{
2320 return _ext4_show_options(seq, root->d_sb, 0);
2321}
2322
ebd173be 2323int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2324{
2325 struct super_block *sb = seq->private;
2326 int rc;
2327
bc98a42c 2328 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 2329 rc = _ext4_show_options(seq, sb, 1);
2330 seq_puts(seq, "\n");
2331 return rc;
2332}
2333
617ba13b 2334static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2335 int read_only)
2336{
617ba13b 2337 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 2338 int err = 0;
ac27a0ec 2339
617ba13b 2340 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2341 ext4_msg(sb, KERN_ERR, "revision level too high, "
2342 "forcing read-only mode");
c89128a0 2343 err = -EROFS;
ac27a0ec
DK		 2344 }
2345 if (read_only)
281b5995 2346 goto done;
617ba13b 2347 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2348 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2349 "running e2fsck is recommended");
c8b459f4 2350 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2351 ext4_msg(sb, KERN_WARNING,
2352 "warning: mounting fs with errors, "
2353 "running e2fsck is recommended");
ed3ce80a 2354 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2355 le16_to_cpu(es->s_mnt_count) >=
2356 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2357 ext4_msg(sb, KERN_WARNING,
2358 "warning: maximal mount count reached, "
2359 "running e2fsck is recommended");
ac27a0ec 2360 else if (le32_to_cpu(es->s_checkinterval) &&
6a0678a7
AB		 2361 (ext4_get_tstamp(es, s_lastcheck) +
2362 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
b31e1552
ES		 2363 ext4_msg(sb, KERN_WARNING,
2364 "warning: checktime reached, "
2365 "running e2fsck is recommended");
0b8e58a1 2366 if (!sbi->s_journal)
0390131b 2367 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2368 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2369 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2370 le16_add_cpu(&es->s_mnt_count, 1);
6a0678a7 2371 ext4_update_tstamp(es, s_mtime);
0390131b 2372 if (sbi->s_journal)
e2b911c5 2373 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2374
c89128a0 2375 err = ext4_commit_super(sb, 1);
281b5995 2376done:
ac27a0ec 2377 if (test_opt(sb, DEBUG))
a9df9a49 2378 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2379 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2380 sb->s_blocksize,
2381 sbi->s_groups_count,
617ba13b
MC		 2382 EXT4_BLOCKS_PER_GROUP(sb),
2383 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2384 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2385
7abc52c2 2386 cleancache_init_fs(sb);
c89128a0 2387 return err;
ac27a0ec
DK		 2388}
2389
117fff10
TT		 2390int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2391{
2392 struct ext4_sb_info *sbi = EXT4_SB(sb);
7c990728 2393 struct flex_groups **old_groups, **new_groups;
37b0b6b8 2394 int size, i, j;
117fff10
TT		 2395
2396 if (!sbi->s_log_groups_per_flex)
2397 return 0;
2398
2399 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2400 if (size <= sbi->s_flex_groups_allocated)
2401 return 0;
2402
7c990728
SJS		 2403 new_groups = kvzalloc(roundup_pow_of_two(size *
2404 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
117fff10 2405 if (!new_groups) {
7c990728
SJS		 2406 ext4_msg(sb, KERN_ERR,
2407 "not enough memory for %d flex group pointers", size);
117fff10
TT		 2408 return -ENOMEM;
2409 }
7c990728
SJS		 2410 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2411 new_groups[i] = kvzalloc(roundup_pow_of_two(
2412 sizeof(struct flex_groups)),
2413 GFP_KERNEL);
2414 if (!new_groups[i]) {
37b0b6b8
DC		 2415 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2416 kvfree(new_groups[j]);
7c990728
SJS		 2417 kvfree(new_groups);
2418 ext4_msg(sb, KERN_ERR,
2419 "not enough memory for %d flex groups", size);
2420 return -ENOMEM;
2421 }
117fff10 2422 }
7c990728
SJS		 2423 rcu_read_lock();
2424 old_groups = rcu_dereference(sbi->s_flex_groups);
2425 if (old_groups)
2426 memcpy(new_groups, old_groups,
2427 (sbi->s_flex_groups_allocated *
2428 sizeof(struct flex_groups *)));
2429 rcu_read_unlock();
2430 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2431 sbi->s_flex_groups_allocated = size;
2432 if (old_groups)
2433 ext4_kvfree_array_rcu(old_groups);
117fff10
TT		 2434 return 0;
2435}
2436
772cb7c8
JS		 2437static int ext4_fill_flex_info(struct super_block *sb)
2438{
2439 struct ext4_sb_info *sbi = EXT4_SB(sb);
2440 struct ext4_group_desc *gdp = NULL;
7c990728 2441 struct flex_groups *fg;
772cb7c8 2442 ext4_group_t flex_group;
117fff10 2443 int i, err;
772cb7c8 2444
503358ae 2445 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2446 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2447 sbi->s_log_groups_per_flex = 0;
2448 return 1;
2449 }
2450
117fff10
TT		 2451 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2452 if (err)
9933fc0a 2453 goto failed;
772cb7c8 2454
772cb7c8 2455 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2456 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2457
2458 flex_group = ext4_flex_group(sbi, i);
7c990728
SJS		 2459 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2460 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
90ba983f 2461 atomic64_add(ext4_free_group_clusters(sb, gdp),
7c990728
SJS		 2462 &fg->free_clusters);
2463 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
772cb7c8
JS		 2464 }
2465
2466 return 1;
2467failed:
2468 return 0;
2469}
2470
e2b911c5 2471static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2472 struct ext4_group_desc *gdp)
717d50e4 2473{
b47820ed 2474 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2475 __u16 crc = 0;
feb0ab32 2476 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2477 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2478
9aa5d32b 2479 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2480 /* Use new metadata_csum algorithm */
feb0ab32 2481 __u32 csum32;
b47820ed 2482 __u16 dummy_csum = 0;
feb0ab32 2483
feb0ab32
DW		 2484 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2485 sizeof(le_group));
b47820ed
DJ		 2486 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2487 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2488 sizeof(dummy_csum));
2489 offset += sizeof(dummy_csum);
2490 if (offset < sbi->s_desc_size)
2491 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2492 sbi->s_desc_size - offset);
feb0ab32
DW		 2493
2494 crc = csum32 & 0xFFFF;
2495 goto out;
717d50e4
AD		 2496 }
2497
feb0ab32 2498 /* old crc16 code */
e2b911c5 2499 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2500 return 0;
2501
feb0ab32
DW		 2502 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2503 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2504 crc = crc16(crc, (__u8 *)gdp, offset);
2505 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2506 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2507 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2508 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2509 crc = crc16(crc, (__u8 *)gdp + offset,
2510 le16_to_cpu(sbi->s_es->s_desc_size) -
2511 offset);
2512
2513out:
717d50e4
AD		 2514 return cpu_to_le16(crc);
2515}
2516
feb0ab32 2517int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2518 struct ext4_group_desc *gdp)
2519{
feb0ab32 2520 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2521 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2522 return 0;
2523
2524 return 1;
2525}
2526
feb0ab32
DW		 2527void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2528 struct ext4_group_desc *gdp)
2529{
2530 if (!ext4_has_group_desc_csum(sb))
2531 return;
e2b911c5 2532 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2533}
2534
ac27a0ec 2535/* Called at mount-time, super-block is locked */
bfff6873 2536static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2537 ext4_fsblk_t sb_block,
bfff6873 2538 ext4_group_t *first_not_zeroed)
ac27a0ec 2539{
617ba13b
MC		 2540 struct ext4_sb_info *sbi = EXT4_SB(sb);
2541 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2542 ext4_fsblk_t last_block;
44de022c 2543 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
bd81d8ee
LV		 2544 ext4_fsblk_t block_bitmap;
2545 ext4_fsblk_t inode_bitmap;
2546 ext4_fsblk_t inode_table;
ce421581 2547 int flexbg_flag = 0;
bfff6873 2548 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2549
e2b911c5 2550 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2551 flexbg_flag = 1;
2552
af5bc92d 2553 ext4_debug("Checking group descriptors");
ac27a0ec 2554
197cd65a
AM		 2555 for (i = 0; i < sbi->s_groups_count; i++) {
2556 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2557
ce421581 2558 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2559 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2560 else
2561 last_block = first_block +
617ba13b 2562 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2563
bfff6873
LC		 2564 if ((grp == sbi->s_groups_count) &&
2565 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2566 grp = i;
2567
8fadc143 2568 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2569 if (block_bitmap == sb_block) {
2570 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2571 "Block bitmap for group %u overlaps "
2572 "superblock", i);
18db4b4e
TT		 2573 if (!sb_rdonly(sb))
2574 return 0;
829fa70d 2575 }
77260807
TT		 2576 if (block_bitmap >= sb_block + 1 &&
2577 block_bitmap <= last_bg_block) {
2578 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2579 "Block bitmap for group %u overlaps "
2580 "block group descriptors", i);
2581 if (!sb_rdonly(sb))
2582 return 0;
2583 }
2b2d6d01 2584 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2585 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2586 "Block bitmap for group %u not in group "
b31e1552 2587 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2588 return 0;
2589 }
8fadc143 2590 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2591 if (inode_bitmap == sb_block) {
2592 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2593 "Inode bitmap for group %u overlaps "
2594 "superblock", i);
18db4b4e
TT		 2595 if (!sb_rdonly(sb))
2596 return 0;
829fa70d 2597 }
77260807
TT		 2598 if (inode_bitmap >= sb_block + 1 &&
2599 inode_bitmap <= last_bg_block) {
2600 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2601 "Inode bitmap for group %u overlaps "
2602 "block group descriptors", i);
2603 if (!sb_rdonly(sb))
2604 return 0;
2605 }
2b2d6d01 2606 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2607 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2608 "Inode bitmap for group %u not in group "
b31e1552 2609 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2610 return 0;
2611 }
8fadc143 2612 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2613 if (inode_table == sb_block) {
2614 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2615 "Inode table for group %u overlaps "
2616 "superblock", i);
18db4b4e
TT		 2617 if (!sb_rdonly(sb))
2618 return 0;
829fa70d 2619 }
77260807
TT		 2620 if (inode_table >= sb_block + 1 &&
2621 inode_table <= last_bg_block) {
2622 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2623 "Inode table for group %u overlaps "
2624 "block group descriptors", i);
2625 if (!sb_rdonly(sb))
2626 return 0;
2627 }
bd81d8ee 2628 if (inode_table < first_block ||
2b2d6d01 2629 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2630 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2631 "Inode table for group %u not in group "
b31e1552 2632 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2633 return 0;
2634 }
955ce5f5 2635 ext4_lock_group(sb, i);
feb0ab32 2636 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2637 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2638 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2639 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2640 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 2641 if (!sb_rdonly(sb)) {
955ce5f5 2642 ext4_unlock_group(sb, i);
8a266467 2643 return 0;
7ee1ec4c 2644 }
717d50e4 2645 }
955ce5f5 2646 ext4_unlock_group(sb, i);
ce421581
JS		 2647 if (!flexbg_flag)
2648 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2649 }
bfff6873
LC		 2650 if (NULL != first_not_zeroed)
2651 *first_not_zeroed = grp;
ac27a0ec
DK		 2652 return 1;
2653}
2654
617ba13b 2655/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2656 * the superblock) which were deleted from all directories, but held open by
2657 * a process at the time of a crash. We walk the list and try to delete these
2658 * inodes at recovery time (only with a read-write filesystem).
2659 *
2660 * In order to keep the orphan inode chain consistent during traversal (in
2661 * case of crash during recovery), we link each inode into the superblock
2662 * orphan list_head and handle it the same way as an inode deletion during
2663 * normal operation (which journals the operations for us).
2664 *
2665 * We only do an iget() and an iput() on each inode, which is very safe if we
2666 * accidentally point at an in-use or already deleted inode. The worst that
2667 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2668 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2669 * e2fsck was run on this filesystem, and it must have already done the orphan
2670 * inode cleanup for us, so we can safely abort without any further action.
2671 */
2b2d6d01
TT		 2672static void ext4_orphan_cleanup(struct super_block *sb,
2673 struct ext4_super_block *es)
ac27a0ec
DK		 2674{
2675 unsigned int s_flags = sb->s_flags;
2c98eb5e 2676 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec 2677#ifdef CONFIG_QUOTA
95f1fda4 2678 int quota_update = 0;
ac27a0ec
DK		 2679 int i;
2680#endif
2681 if (!es->s_last_orphan) {
2682 jbd_debug(4, "no orphan inodes to clean up\n");
2683 return;
2684 }
2685
a8f48a95 2686 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2687 ext4_msg(sb, KERN_ERR, "write access "
2688 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2689 return;
2690 }
2691
d39195c3
AG		 2692 /* Check if feature set would not allow a r/w mount */
2693 if (!ext4_feature_set_ok(sb, 0)) {
2694 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2695 "unknown ROCOMPAT features");
2696 return;
2697 }
2698
617ba13b 2699 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6 2700 /* don't clear list on RO mount w/ errors */
1751e8a6 2701 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
84474976 2702 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2703 "clearing orphan list.\n");
c25f9bc6
ES		 2704 es->s_last_orphan = 0;
2705 }
ac27a0ec
DK		 2706 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2707 return;
2708 }
2709
1751e8a6 2710 if (s_flags & SB_RDONLY) {
b31e1552 2711 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1751e8a6 2712 sb->s_flags &= ~SB_RDONLY;
ac27a0ec
DK		 2713 }
2714#ifdef CONFIG_QUOTA
2715 /* Needed for iput() to work correctly and not trash data */
1751e8a6 2716 sb->s_flags |= SB_ACTIVE;
95f1fda4 2717
2718 /*
2719 * Turn on quotas which were not enabled for read-only mounts if
2720 * filesystem has quota feature, so that they are updated correctly.
2721 */
1751e8a6 2722 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
95f1fda4 2723 int ret = ext4_enable_quotas(sb);
2724
2725 if (!ret)
2726 quota_update = 1;
2727 else
2728 ext4_msg(sb, KERN_ERR,
2729 "Cannot turn on quotas: error %d", ret);
2730 }
2731
2732 /* Turn on journaled quotas used for old sytle */
a2d4a646 2733 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2734 if (EXT4_SB(sb)->s_qf_names[i]) {
2735 int ret = ext4_quota_on_mount(sb, i);
95f1fda4 2736
2737 if (!ret)
2738 quota_update = 1;
2739 else
b31e1552
ES		 2740 ext4_msg(sb, KERN_ERR,
2741 "Cannot turn on journaled "
95f1fda4 2742 "quota: type %d: error %d", i, ret);
ac27a0ec
DK		 2743 }
2744 }
2745#endif
2746
2747 while (es->s_last_orphan) {
2748 struct inode *inode;
2749
c65d5c6c
VN		 2750 /*
2751 * We may have encountered an error during cleanup; if
2752 * so, skip the rest.
2753 */
2754 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2755 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2756 es->s_last_orphan = 0;
2757 break;
2758 }
2759
97bd42b9
JB		 2760 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2761 if (IS_ERR(inode)) {
ac27a0ec
DK		 2762 es->s_last_orphan = 0;
2763 break;
2764 }
2765
617ba13b 2766 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2767 dquot_initialize(inode);
ac27a0ec 2768 if (inode->i_nlink) {
566370a2
PT		 2769 if (test_opt(sb, DEBUG))
2770 ext4_msg(sb, KERN_DEBUG,
2771 "%s: truncating inode %lu to %lld bytes",
2772 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2773 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2774 inode->i_ino, inode->i_size);
5955102c 2775 inode_lock(inode);
55f252c9 2776 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2777 ret = ext4_truncate(inode);
2778 if (ret)
2779 ext4_std_error(inode->i_sb, ret);
5955102c 2780 inode_unlock(inode);
ac27a0ec
DK		 2781 nr_truncates++;
2782 } else {
566370a2
PT		 2783 if (test_opt(sb, DEBUG))
2784 ext4_msg(sb, KERN_DEBUG,
2785 "%s: deleting unreferenced inode %lu",
2786 __func__, inode->i_ino);
ac27a0ec
DK		 2787 jbd_debug(2, "deleting unreferenced inode %lu\n",
2788 inode->i_ino);
2789 nr_orphans++;
2790 }
2791 iput(inode); /* The delete magic happens here! */
2792 }
2793
2b2d6d01 2794#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2795
2796 if (nr_orphans)
b31e1552
ES		 2797 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2798 PLURAL(nr_orphans));
ac27a0ec 2799 if (nr_truncates)
b31e1552
ES		 2800 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2801 PLURAL(nr_truncates));
ac27a0ec 2802#ifdef CONFIG_QUOTA
95f1fda4 2803 /* Turn off quotas if they were enabled for orphan cleanup */
2804 if (quota_update) {
2805 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2806 if (sb_dqopt(sb)->files[i])
2807 dquot_quota_off(sb, i);
2808 }
ac27a0ec
DK		 2809 }
2810#endif
1751e8a6 2811 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
ac27a0ec 2812}
0b8e58a1 2813
cd2291a4
ES		 2814/*
2815 * Maximal extent format file size.
2816 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2817 * extent format containers, within a sector_t, and within i_blocks
2818 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2819 * so that won't be a limiting factor.
2820 *
f17722f9
LC		 2821 * However there is other limiting factor. We do store extents in the form
2822 * of starting block and length, hence the resulting length of the extent
2823 * covering maximum file size must fit into on-disk format containers as
2824 * well. Given that length is always by 1 unit bigger than max unit (because
2825 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2826 *
cd2291a4
ES		 2827 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2828 */
f287a1a5 2829static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2830{
2831 loff_t res;
2832 loff_t upper_limit = MAX_LFS_FILESIZE;
2833
72deb455
CH		 2834 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2835
2836 if (!has_huge_files) {
cd2291a4
ES		 2837 upper_limit = (1LL << 32) - 1;
2838
2839 /* total blocks in file system block size */
2840 upper_limit >>= (blkbits - 9);
2841 upper_limit <<= blkbits;
2842 }
2843
f17722f9
LC		 2844 /*
2845 * 32-bit extent-start container, ee_block. We lower the maxbytes
2846 * by one fs block, so ee_len can cover the extent of maximum file
2847 * size
2848 */
2849 res = (1LL << 32) - 1;
cd2291a4 2850 res <<= blkbits;
cd2291a4
ES		 2851
2852 /* Sanity check against vm- & vfs- imposed limits */
2853 if (res > upper_limit)
2854 res = upper_limit;
2855
2856 return res;
2857}
ac27a0ec 2858
ac27a0ec 2859/*
cd2291a4 2860 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2861 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2862 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2863 */
f287a1a5 2864static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2865{
617ba13b 2866 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2867 int meta_blocks;
2868 loff_t upper_limit;
0b8e58a1
AD		 2869 /* This is calculated to be the largest file size for a dense, block
2870 * mapped file such that the file's total number of 512-byte sectors,
2871 * including data and all indirect blocks, does not exceed (2^48 - 1).
2872 *
2873 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2874 * number of 512-byte sectors of the file.
0fc1b451
AK		 2875 */
2876
72deb455 2877 if (!has_huge_files) {
0fc1b451 2878 /*
72deb455
CH		 2879 * !has_huge_files or implies that the inode i_block field
2880 * represents total file blocks in 2^32 512-byte sectors ==
2881 * size of vfs inode i_blocks * 8
0fc1b451
AK		 2882 */
2883 upper_limit = (1LL << 32) - 1;
2884
2885 /* total blocks in file system block size */
2886 upper_limit >>= (bits - 9);
2887
2888 } else {
8180a562
AK		 2889 /*
2890 * We use 48 bit ext4_inode i_blocks
2891 * With EXT4_HUGE_FILE_FL set the i_blocks
2892 * represent total number of blocks in
2893 * file system block size
2894 */
0fc1b451
AK		 2895 upper_limit = (1LL << 48) - 1;
2896
0fc1b451
AK		 2897 }
2898
2899 /* indirect blocks */
2900 meta_blocks = 1;
2901 /* double indirect blocks */
2902 meta_blocks += 1 + (1LL << (bits-2));
2903 /* tripple indirect blocks */
2904 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2905
2906 upper_limit -= meta_blocks;
2907 upper_limit <<= bits;
ac27a0ec
DK		 2908
2909 res += 1LL << (bits-2);
2910 res += 1LL << (2*(bits-2));
2911 res += 1LL << (3*(bits-2));
2912 res <<= bits;
2913 if (res > upper_limit)
2914 res = upper_limit;
0fc1b451
AK		 2915
2916 if (res > MAX_LFS_FILESIZE)
2917 res = MAX_LFS_FILESIZE;
2918
ac27a0ec
DK		 2919 return res;
2920}
2921
617ba13b 2922static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2923 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2924{
617ba13b 2925 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2926 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2927 int has_super = 0;
2928
2929 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2930
e2b911c5 2931 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2932 return logical_sb_block + nr + 1;
ac27a0ec 2933 bg = sbi->s_desc_per_block * nr;
617ba13b 2934 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2935 has_super = 1;
0b8e58a1 2936
bd63f6b0
DW		 2937 /*
2938 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2939 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2940 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2941 * compensate.
2942 */
2943 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 2944 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 2945 has_super++;
2946
617ba13b 2947 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2948}
2949
c9de560d
AT		 2950/**
2951 * ext4_get_stripe_size: Get the stripe size.
2952 * @sbi: In memory super block info
2953 *
2954 * If we have specified it via mount option, then
2955 * use the mount option value. If the value specified at mount time is
2956 * greater than the blocks per group use the super block value.
2957 * If the super block value is greater than blocks per group return 0.
2958 * Allocator needs it be less than blocks per group.
2959 *
2960 */
2961static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2962{
2963 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2964 unsigned long stripe_width =
2965 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2966 int ret;
c9de560d
AT		 2967
2968 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 2969 ret = sbi->s_stripe;
5469d7c3 2970 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 2971 ret = stripe_width;
5469d7c3 2972 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 2973 ret = stride;
2974 else
2975 ret = 0;
c9de560d 2976
3eb08658
DE		 2977 /*
2978 * If the stripe width is 1, this makes no sense and
2979 * we set it to 0 to turn off stripe handling code.
2980 */
2981 if (ret <= 1)
2982 ret = 0;
c9de560d 2983
3eb08658 2984 return ret;
c9de560d 2985}
ac27a0ec 2986
a13fb1a4
ES		 2987/*
2988 * Check whether this filesystem can be mounted based on
2989 * the features present and the RDONLY/RDWR mount requested.
2990 * Returns 1 if this filesystem can be mounted as requested,
2991 * 0 if it cannot be.
2992 */
2993static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2994{
e2b911c5 2995 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2996 ext4_msg(sb, KERN_ERR,
2997 "Couldn't mount because of "
2998 "unsupported optional features (%x)",
2999 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3000 ~EXT4_FEATURE_INCOMPAT_SUPP));
3001 return 0;
3002 }
3003
c83ad55e
GKB		 3004#ifndef CONFIG_UNICODE
3005 if (ext4_has_feature_casefold(sb)) {
3006 ext4_msg(sb, KERN_ERR,
3007 "Filesystem with casefold feature cannot be "
3008 "mounted without CONFIG_UNICODE");
3009 return 0;
3010 }
3011#endif
3012
a13fb1a4
ES		 3013 if (readonly)
3014 return 1;
3015
e2b911c5 3016 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 3017 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 3018 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 3019 return 1;
3020 }
3021
a13fb1a4 3022 /* Check that feature set is OK for a read-write mount */
e2b911c5 3023 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 3024 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3025 "unsupported optional features (%x)",
3026 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3027 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3028 return 0;
3029 }
e2b911c5 3030 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 3031 ext4_msg(sb, KERN_ERR,
3032 "Can't support bigalloc feature without "
3033 "extents feature\n");
3034 return 0;
3035 }
7c319d32 3036
9db176bc 3037#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
d65d87a0
TT		 3038 if (!readonly && (ext4_has_feature_quota(sb) ||
3039 ext4_has_feature_project(sb))) {
7c319d32 3040 ext4_msg(sb, KERN_ERR,
d65d87a0 3041 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
689c958c
LX		 3042 return 0;
3043 }
7c319d32 3044#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 3045 return 1;
3046}
3047
66e61a9e
TT		 3048/*
3049 * This function is called once a day if we have errors logged
3050 * on the file system
3051 */
235699a8 3052static void print_daily_error_info(struct timer_list *t)
66e61a9e 3053{
235699a8
KC		 3054 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3055 struct super_block *sb = sbi->s_sb;
3056 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 3057
3058 if (es->s_error_count)
ae0f78de
TT		 3059 /* fsck newer than v1.41.13 is needed to clean this condition. */
3060 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 3061 le32_to_cpu(es->s_error_count));
3062 if (es->s_first_error_time) {
6a0678a7
AB		 3063 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3064 sb->s_id,
3065 ext4_get_tstamp(es, s_first_error_time),
66e61a9e
TT		 3066 (int) sizeof(es->s_first_error_func),
3067 es->s_first_error_func,
3068 le32_to_cpu(es->s_first_error_line));
3069 if (es->s_first_error_ino)
651e1c3b 3070 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3071 le32_to_cpu(es->s_first_error_ino));
3072 if (es->s_first_error_block)
651e1c3b 3073 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3074 le64_to_cpu(es->s_first_error_block));
651e1c3b 3075 printk(KERN_CONT "\n");
66e61a9e
TT		 3076 }
3077 if (es->s_last_error_time) {
6a0678a7
AB		 3078 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3079 sb->s_id,
3080 ext4_get_tstamp(es, s_last_error_time),
66e61a9e
TT		 3081 (int) sizeof(es->s_last_error_func),
3082 es->s_last_error_func,
3083 le32_to_cpu(es->s_last_error_line));
3084 if (es->s_last_error_ino)
651e1c3b 3085 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3086 le32_to_cpu(es->s_last_error_ino));
3087 if (es->s_last_error_block)
651e1c3b 3088 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3089 le64_to_cpu(es->s_last_error_block));
651e1c3b 3090 printk(KERN_CONT "\n");
66e61a9e
TT		 3091 }
3092 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3093}
3094
bfff6873
LC		 3095/* Find next suitable group and run ext4_init_inode_table */
3096static int ext4_run_li_request(struct ext4_li_request *elr)
3097{
3098 struct ext4_group_desc *gdp = NULL;
3099 ext4_group_t group, ngroups;
3100 struct super_block *sb;
3101 unsigned long timeout = 0;
3102 int ret = 0;
3103
3104 sb = elr->lr_super;
3105 ngroups = EXT4_SB(sb)->s_groups_count;
3106
3107 for (group = elr->lr_next_group; group < ngroups; group++) {
3108 gdp = ext4_get_group_desc(sb, group, NULL);
3109 if (!gdp) {
3110 ret = 1;
3111 break;
3112 }
3113
3114 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3115 break;
3116 }
3117
7f511862 3118 if (group >= ngroups)
bfff6873
LC		 3119 ret = 1;
3120
3121 if (!ret) {
3122 timeout = jiffies;
3123 ret = ext4_init_inode_table(sb, group,
3124 elr->lr_timeout ? 0 : 1);
3125 if (elr->lr_timeout == 0) {
51ce6511
LC		 3126 timeout = (jiffies - timeout) *
3127 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 3128 elr->lr_timeout = timeout;
3129 }
3130 elr->lr_next_sched = jiffies + elr->lr_timeout;
3131 elr->lr_next_group = group + 1;
3132 }
bfff6873
LC		 3133 return ret;
3134}
3135
3136/*
3137 * Remove lr_request from the list_request and free the
4ed5c033 3138 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 3139 */
3140static void ext4_remove_li_request(struct ext4_li_request *elr)
3141{
3142 struct ext4_sb_info *sbi;
3143
3144 if (!elr)
3145 return;
3146
3147 sbi = elr->lr_sbi;
3148
3149 list_del(&elr->lr_request);
3150 sbi->s_li_request = NULL;
3151 kfree(elr);
3152}
3153
3154static void ext4_unregister_li_request(struct super_block *sb)
3155{
1bb933fb
LC		 3156 mutex_lock(&ext4_li_mtx);
3157 if (!ext4_li_info) {
3158 mutex_unlock(&ext4_li_mtx);
bfff6873 3159 return;
1bb933fb 3160 }
bfff6873
LC		 3161
3162 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 3163 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 3164 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 3165 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 3166}
3167
8f1f7453
ES		 3168static struct task_struct *ext4_lazyinit_task;
3169
bfff6873
LC		 3170/*
3171 * This is the function where ext4lazyinit thread lives. It walks
3172 * through the request list searching for next scheduled filesystem.
3173 * When such a fs is found, run the lazy initialization request
3174 * (ext4_rn_li_request) and keep track of the time spend in this
3175 * function. Based on that time we compute next schedule time of
3176 * the request. When walking through the list is complete, compute
3177 * next waking time and put itself into sleep.
3178 */
3179static int ext4_lazyinit_thread(void *arg)
3180{
3181 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3182 struct list_head *pos, *n;
3183 struct ext4_li_request *elr;
4ed5c033 3184 unsigned long next_wakeup, cur;
bfff6873
LC		 3185
3186 BUG_ON(NULL == eli);
3187
bfff6873
LC		 3188cont_thread:
3189 while (true) {
3190 next_wakeup = MAX_JIFFY_OFFSET;
3191
3192 mutex_lock(&eli->li_list_mtx);
3193 if (list_empty(&eli->li_request_list)) {
3194 mutex_unlock(&eli->li_list_mtx);
3195 goto exit_thread;
3196 }
bfff6873 3197 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3198 int err = 0;
3199 int progress = 0;
bfff6873
LC		 3200 elr = list_entry(pos, struct ext4_li_request,
3201 lr_request);
3202
e22834f0
DM		 3203 if (time_before(jiffies, elr->lr_next_sched)) {
3204 if (time_before(elr->lr_next_sched, next_wakeup))
3205 next_wakeup = elr->lr_next_sched;
3206 continue;
3207 }
3208 if (down_read_trylock(&elr->lr_super->s_umount)) {
3209 if (sb_start_write_trylock(elr->lr_super)) {
3210 progress = 1;
3211 /*
3212 * We hold sb->s_umount, sb can not
3213 * be removed from the list, it is
3214 * now safe to drop li_list_mtx
3215 */
3216 mutex_unlock(&eli->li_list_mtx);
3217 err = ext4_run_li_request(elr);
3218 sb_end_write(elr->lr_super);
3219 mutex_lock(&eli->li_list_mtx);
3220 n = pos->next;
b2c78cd0 3221 }
e22834f0
DM		 3222 up_read((&elr->lr_super->s_umount));
3223 }
3224 /* error, remove the lazy_init job */
3225 if (err) {
3226 ext4_remove_li_request(elr);
3227 continue;
3228 }
3229 if (!progress) {
3230 elr->lr_next_sched = jiffies +
3231 (prandom_u32()
3232 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 3233 }
bfff6873
LC		 3234 if (time_before(elr->lr_next_sched, next_wakeup))
3235 next_wakeup = elr->lr_next_sched;
3236 }
3237 mutex_unlock(&eli->li_list_mtx);
3238
a0acae0e 3239 try_to_freeze();
bfff6873 3240
4ed5c033
LC		 3241 cur = jiffies;
3242 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3243 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3244 cond_resched();
3245 continue;
3246 }
3247
4ed5c033
LC		 3248 schedule_timeout_interruptible(next_wakeup - cur);
3249
8f1f7453
ES		 3250 if (kthread_should_stop()) {
3251 ext4_clear_request_list();
3252 goto exit_thread;
3253 }
bfff6873
LC		 3254 }
3255
3256exit_thread:
3257 /*
3258 * It looks like the request list is empty, but we need
3259 * to check it under the li_list_mtx lock, to prevent any
3260 * additions into it, and of course we should lock ext4_li_mtx
3261 * to atomically free the list and ext4_li_info, because at
3262 * this point another ext4 filesystem could be registering
3263 * new one.
3264 */
3265 mutex_lock(&ext4_li_mtx);
3266 mutex_lock(&eli->li_list_mtx);
3267 if (!list_empty(&eli->li_request_list)) {
3268 mutex_unlock(&eli->li_list_mtx);
3269 mutex_unlock(&ext4_li_mtx);
3270 goto cont_thread;
3271 }
3272 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3273 kfree(ext4_li_info);
3274 ext4_li_info = NULL;
3275 mutex_unlock(&ext4_li_mtx);
3276
3277 return 0;
3278}
3279
3280static void ext4_clear_request_list(void)
3281{
3282 struct list_head *pos, *n;
3283 struct ext4_li_request *elr;
3284
3285 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3286 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3287 elr = list_entry(pos, struct ext4_li_request,
3288 lr_request);
3289 ext4_remove_li_request(elr);
3290 }
3291 mutex_unlock(&ext4_li_info->li_list_mtx);
3292}
3293
3294static int ext4_run_lazyinit_thread(void)
3295{
8f1f7453
ES		 3296 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3297 ext4_li_info, "ext4lazyinit");
3298 if (IS_ERR(ext4_lazyinit_task)) {
3299 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3300 ext4_clear_request_list();
bfff6873
LC		 3301 kfree(ext4_li_info);
3302 ext4_li_info = NULL;
92b97816 3303 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3304 "initialization thread\n",
3305 err);
3306 return err;
3307 }
3308 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3309 return 0;
3310}
3311
3312/*
3313 * Check whether it make sense to run itable init. thread or not.
3314 * If there is at least one uninitialized inode table, return
3315 * corresponding group number, else the loop goes through all
3316 * groups and return total number of groups.
3317 */
3318static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3319{
3320 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3321 struct ext4_group_desc *gdp = NULL;
3322
8844618d
TT		 3323 if (!ext4_has_group_desc_csum(sb))
3324 return ngroups;
3325
bfff6873
LC		 3326 for (group = 0; group < ngroups; group++) {
3327 gdp = ext4_get_group_desc(sb, group, NULL);
3328 if (!gdp)
3329 continue;
3330
50122847 3331 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
bfff6873
LC		 3332 break;
3333 }
3334
3335 return group;
3336}
3337
3338static int ext4_li_info_new(void)
3339{
3340 struct ext4_lazy_init *eli = NULL;
3341
3342 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3343 if (!eli)
3344 return -ENOMEM;
3345
bfff6873
LC		 3346 INIT_LIST_HEAD(&eli->li_request_list);
3347 mutex_init(&eli->li_list_mtx);
3348
bfff6873
LC		 3349 eli->li_state |= EXT4_LAZYINIT_QUIT;
3350
3351 ext4_li_info = eli;
3352
3353 return 0;
3354}
3355
3356static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3357 ext4_group_t start)
3358{
3359 struct ext4_sb_info *sbi = EXT4_SB(sb);
3360 struct ext4_li_request *elr;
bfff6873
LC		 3361
3362 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3363 if (!elr)
3364 return NULL;
3365
3366 elr->lr_super = sb;
3367 elr->lr_sbi = sbi;
3368 elr->lr_next_group = start;
3369
3370 /*
3371 * Randomize first schedule time of the request to
3372 * spread the inode table initialization requests
3373 * better.
3374 */
dd1f723b
TT		 3375 elr->lr_next_sched = jiffies + (prandom_u32() %
3376 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3377 return elr;
3378}
3379
7f511862
TT		 3380int ext4_register_li_request(struct super_block *sb,
3381 ext4_group_t first_not_zeroed)
bfff6873
LC		 3382{
3383 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3384 struct ext4_li_request *elr = NULL;
49598e04 3385 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 3386 int ret = 0;
bfff6873 3387
7f511862 3388 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3389 if (sbi->s_li_request != NULL) {
3390 /*
3391 * Reset timeout so it can be computed again, because
3392 * s_li_wait_mult might have changed.
3393 */
3394 sbi->s_li_request->lr_timeout = 0;
7f511862 3395 goto out;
51ce6511 3396 }
bfff6873 3397
bc98a42c 3398 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
55ff3840 3399 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3400 goto out;
bfff6873
LC		 3401
3402 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3403 if (!elr) {
3404 ret = -ENOMEM;
3405 goto out;
3406 }
bfff6873
LC		 3407
3408 if (NULL == ext4_li_info) {
3409 ret = ext4_li_info_new();
3410 if (ret)
3411 goto out;
3412 }
3413
3414 mutex_lock(&ext4_li_info->li_list_mtx);
3415 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3416 mutex_unlock(&ext4_li_info->li_list_mtx);
3417
3418 sbi->s_li_request = elr;
46e4690b
TM		 3419 /*
3420 * set elr to NULL here since it has been inserted to
3421 * the request_list and the removal and free of it is
3422 * handled by ext4_clear_request_list from now on.
3423 */
3424 elr = NULL;
bfff6873
LC		 3425
3426 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3427 ret = ext4_run_lazyinit_thread();
3428 if (ret)
3429 goto out;
3430 }
bfff6873 3431out:
beed5ecb
NK		 3432 mutex_unlock(&ext4_li_mtx);
3433 if (ret)
bfff6873 3434 kfree(elr);
bfff6873
LC		 3435 return ret;
3436}
3437
3438/*
3439 * We do not need to lock anything since this is called on
3440 * module unload.
3441 */
3442static void ext4_destroy_lazyinit_thread(void)
3443{
3444 /*
3445 * If thread exited earlier
3446 * there's nothing to be done.
3447 */
8f1f7453 3448 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3449 return;
3450
8f1f7453 3451 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3452}
3453
25ed6e8a
DW		 3454static int set_journal_csum_feature_set(struct super_block *sb)
3455{
3456 int ret = 1;
3457 int compat, incompat;
3458 struct ext4_sb_info *sbi = EXT4_SB(sb);
3459
9aa5d32b 3460 if (ext4_has_metadata_csum(sb)) {
db9ee220 3461 /* journal checksum v3 */
25ed6e8a 3462 compat = 0;
db9ee220 3463 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3464 } else {
3465 /* journal checksum v1 */
3466 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3467 incompat = 0;
3468 }
3469
feb8c6d3
DW		 3470 jbd2_journal_clear_features(sbi->s_journal,
3471 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3472 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3473 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3474 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3475 ret = jbd2_journal_set_features(sbi->s_journal,
3476 compat, 0,
3477 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3478 incompat);
3479 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3480 ret = jbd2_journal_set_features(sbi->s_journal,
3481 compat, 0,
3482 incompat);
3483 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3484 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3485 } else {
feb8c6d3
DW		 3486 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3487 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3488 }
3489
3490 return ret;
3491}
3492
952fc18e
TT		 3493/*
3494 * Note: calculating the overhead so we can be compatible with
3495 * historical BSD practice is quite difficult in the face of
3496 * clusters/bigalloc. This is because multiple metadata blocks from
3497 * different block group can end up in the same allocation cluster.
3498 * Calculating the exact overhead in the face of clustered allocation
3499 * requires either O(all block bitmaps) in memory or O(number of block
3500 * groups**2) in time. We will still calculate the superblock for
3501 * older file systems --- and if we come across with a bigalloc file
3502 * system with zero in s_overhead_clusters the estimate will be close to
3503 * correct especially for very large cluster sizes --- but for newer
3504 * file systems, it's better to calculate this figure once at mkfs
3505 * time, and store it in the superblock. If the superblock value is
3506 * present (even for non-bigalloc file systems), we will use it.
3507 */
3508static int count_overhead(struct super_block *sb, ext4_group_t grp,
3509 char *buf)
3510{
3511 struct ext4_sb_info *sbi = EXT4_SB(sb);
3512 struct ext4_group_desc *gdp;
3513 ext4_fsblk_t first_block, last_block, b;
3514 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3515 int s, j, count = 0;
3516
e2b911c5 3517 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3518 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3519 sbi->s_itb_per_group + 2);
3520
952fc18e
TT		 3521 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3522 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3523 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3524 for (i = 0; i < ngroups; i++) {
3525 gdp = ext4_get_group_desc(sb, i, NULL);
3526 b = ext4_block_bitmap(sb, gdp);
3527 if (b >= first_block && b <= last_block) {
3528 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3529 count++;
3530 }
3531 b = ext4_inode_bitmap(sb, gdp);
3532 if (b >= first_block && b <= last_block) {
3533 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3534 count++;
3535 }
3536 b = ext4_inode_table(sb, gdp);
3537 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3538 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3539 int c = EXT4_B2C(sbi, b - first_block);
3540 ext4_set_bit(c, buf);
3541 count++;
3542 }
3543 if (i != grp)
3544 continue;
3545 s = 0;
3546 if (ext4_bg_has_super(sb, grp)) {
3547 ext4_set_bit(s++, buf);
3548 count++;
3549 }
c48ae41b
TT		 3550 j = ext4_bg_num_gdb(sb, grp);
3551 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3552 ext4_error(sb, "Invalid number of block group "
3553 "descriptor blocks: %d", j);
3554 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3555 }
c48ae41b
TT		 3556 count += j;
3557 for (; j > 0; j--)
3558 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3559 }
3560 if (!count)
3561 return 0;
3562 return EXT4_CLUSTERS_PER_GROUP(sb) -
3563 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3564}
3565
3566/*
3567 * Compute the overhead and stash it in sbi->s_overhead
3568 */
3569int ext4_calculate_overhead(struct super_block *sb)
3570{
3571 struct ext4_sb_info *sbi = EXT4_SB(sb);
3572 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3573 struct inode *j_inode;
3574 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3575 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3576 ext4_fsblk_t overhead = 0;
4fdb5543 3577 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3578
952fc18e
TT		 3579 if (!buf)
3580 return -ENOMEM;
3581
3582 /*
3583 * Compute the overhead (FS structures). This is constant
3584 * for a given filesystem unless the number of block groups
3585 * changes so we cache the previous value until it does.
3586 */
3587
3588 /*
3589 * All of the blocks before first_data_block are overhead
3590 */
3591 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3592
3593 /*
3594 * Add the overhead found in each block group
3595 */
3596 for (i = 0; i < ngroups; i++) {
3597 int blks;
3598
3599 blks = count_overhead(sb, i, buf);
3600 overhead += blks;
3601 if (blks)
3602 memset(buf, 0, PAGE_SIZE);
3603 cond_resched();
3604 }
3c816ded
EW		 3605
3606 /*
3607 * Add the internal journal blocks whether the journal has been
3608 * loaded or not
3609 */
b003b524 3610 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3611 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3612 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3613 j_inode = ext4_get_journal_inode(sb, j_inum);
3614 if (j_inode) {
3615 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3616 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3617 iput(j_inode);
3618 } else {
3619 ext4_msg(sb, KERN_ERR, "can't get journal size");
3620 }
3621 }
952fc18e
TT		 3622 sbi->s_overhead = overhead;
3623 smp_wmb();
3624 free_page((unsigned long) buf);
3625 return 0;
3626}
3627
b5799018 3628static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3629{
3630 ext4_fsblk_t resv_clusters;
b5799018 3631 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3632
30fac0f7
JK		 3633 /*
3634 * There's no need to reserve anything when we aren't using extents.
3635 * The space estimates are exact, there are no unwritten extents,
3636 * hole punching doesn't need new metadata... This is needed especially
3637 * to keep ext2/3 backward compatibility.
3638 */
e2b911c5 3639 if (!ext4_has_feature_extents(sb))
b5799018 3640 return;
27dd4385
LC		 3641 /*
3642 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3643 * This should cover the situations where we can not afford to run
3644 * out of space like for example punch hole, or converting
556615dc 3645 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3646 * allocation would require 1, or 2 blocks, higher numbers are
3647 * very rare.
3648 */
b5799018
TT		 3649 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3650 sbi->s_cluster_bits);
27dd4385
LC		 3651
3652 do_div(resv_clusters, 50);
3653 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3654
b5799018 3655 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3656}
3657
2b2d6d01 3658static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3659{
5e405595 3660 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
d4c402d9 3661 char *orig_data = kstrdup(data, GFP_KERNEL);
1d0c3924 3662 struct buffer_head *bh, **group_desc;
617ba13b 3663 struct ext4_super_block *es = NULL;
5aee0f8a 3664 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
7c990728 3665 struct flex_groups **flex_groups;
617ba13b
MC		 3666 ext4_fsblk_t block;
3667 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3668 ext4_fsblk_t logical_sb_block;
ac27a0ec 3669 unsigned long offset = 0;
ac27a0ec
DK		 3670 unsigned long journal_devnum = 0;
3671 unsigned long def_mount_opts;
3672 struct inode *root;
0390131b 3673 const char *descr;
dcc7dae3 3674 int ret = -ENOMEM;
281b5995 3675 int blocksize, clustersize;
4ec11028
TT		 3676 unsigned int db_count;
3677 unsigned int i;
281b5995 3678 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3679 __u64 blocks_count;
07aa2ea1 3680 int err = 0;
b3881f74 3681 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3682 ext4_group_t first_not_zeroed;
ac27a0ec 3683
5aee0f8a
TT		 3684 if ((data && !orig_data) || !sbi)
3685 goto out_free_base;
705895b6 3686
aed9eb1b 3687 sbi->s_daxdev = dax_dev;
705895b6
PE		 3688 sbi->s_blockgroup_lock =
3689 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3690 if (!sbi->s_blockgroup_lock)
3691 goto out_free_base;
3692
ac27a0ec 3693 sb->s_fs_info = sbi;
2c0544b2 3694 sbi->s_sb = sb;
240799cd 3695 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3696 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3697 if (sb->s_bdev->bd_part)
3698 sbi->s_sectors_written_start =
dbae2c55 3699 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
ac27a0ec 3700
9f6200bb 3701 /* Cleanup superblock name */
ec3904dc 3702 strreplace(sb->s_id, '/', '!');
9f6200bb 3703
07aa2ea1 3704 /* -EINVAL is default */
dcc7dae3 3705 ret = -EINVAL;
617ba13b 3706 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3707 if (!blocksize) {
b31e1552 3708 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3709 goto out_fail;
3710 }
3711
3712 /*
617ba13b 3713 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3714 * block sizes. We need to calculate the offset from buffer start.
3715 */
617ba13b 3716 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3717 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3718 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3719 } else {
70bbb3e0 3720 logical_sb_block = sb_block;
ac27a0ec
DK		 3721 }
3722
a8ac900b 3723 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3724 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3725 goto out_fail;
3726 }
3727 /*
3728 * Note: s_es must be initialized as soon as possible because
617ba13b 3729 * some ext4 macro-instructions depend on its value
ac27a0ec 3730 */
2716b802 3731 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3732 sbi->s_es = es;
3733 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3734 if (sb->s_magic != EXT4_SUPER_MAGIC)
3735 goto cantfind_ext4;
afc32f7e 3736 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3737
feb0ab32 3738 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3739 if (ext4_has_feature_metadata_csum(sb) &&
3740 ext4_has_feature_gdt_csum(sb))
363307e6 3741 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3742 "redundant flags; please run fsck.");
3743
d25425f8
DW		 3744 /* Check for a known checksum algorithm */
3745 if (!ext4_verify_csum_type(sb, es)) {
3746 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3747 "unknown checksum algorithm.");
3748 silent = 1;
3749 goto cantfind_ext4;
3750 }
3751
0441984a 3752 /* Load the checksum driver */
a45403b5
TT		 3753 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3754 if (IS_ERR(sbi->s_chksum_driver)) {
3755 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3756 ret = PTR_ERR(sbi->s_chksum_driver);
3757 sbi->s_chksum_driver = NULL;
3758 goto failed_mount;
0441984a
DW		 3759 }
3760
a9c47317
DW		 3761 /* Check superblock checksum */
3762 if (!ext4_superblock_csum_verify(sb, es)) {
3763 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3764 "invalid superblock checksum. Run e2fsck?");
3765 silent = 1;
6a797d27 3766 ret = -EFSBADCRC;
a9c47317
DW		 3767 goto cantfind_ext4;
3768 }
3769
3770 /* Precompute checksum seed for all metadata */
e2b911c5 3771 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 3772 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 3773 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 3774 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3775 sizeof(es->s_uuid));
3776
ac27a0ec
DK		 3777 /* Set defaults before we parse the mount options */
3778 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3779 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3780 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3781 set_opt(sb, DEBUG);
87f26807 3782 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3783 set_opt(sb, GRPID);
617ba13b 3784 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3785 set_opt(sb, NO_UID32);
ea663336 3786 /* xattr user namespace & acls are now defaulted on */
ea663336 3787 set_opt(sb, XATTR_USER);
244adf64 3788 set_opt(sb, DIOREAD_NOLOCK);
03010a33 3789#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3790 set_opt(sb, POSIX_ACL);
2e7842b8 3791#endif
98c1a759
DW		 3792 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3793 if (ext4_has_metadata_csum(sb))
3794 set_opt(sb, JOURNAL_CHECKSUM);
3795
617ba13b 3796 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3797 set_opt(sb, JOURNAL_DATA);
617ba13b 3798 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3799 set_opt(sb, ORDERED_DATA);
617ba13b 3800 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3801 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3802
3803 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3804 set_opt(sb, ERRORS_PANIC);
bb4f397a 3805 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3806 set_opt(sb, ERRORS_CONT);
bb4f397a 3807 else
fd8c37ec 3808 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3809 /* block_validity enabled by default; disable with noblock_validity */
3810 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3811 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3812 set_opt(sb, DISCARD);
ac27a0ec 3813
08cefc7a
EB		 3814 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3815 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3816 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3817 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3818 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3819
8b67f04a 3820 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3821 set_opt(sb, BARRIER);
ac27a0ec 3822
dd919b98
AK		 3823 /*
3824 * enable delayed allocation by default
3825 * Use -o nodelalloc to turn it off
3826 */
bc0b75f7 3827 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3828 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3829 set_opt(sb, DELALLOC);
dd919b98 3830
51ce6511
LC		 3831 /*
3832 * set default s_li_wait_mult for lazyinit, for the case there is
3833 * no mount option specified.
3834 */
3835 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3836
4f97a681
TT		 3837 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3838 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3839 blocksize > EXT4_MAX_BLOCK_SIZE) {
3840 ext4_msg(sb, KERN_ERR,
3841 "Unsupported filesystem blocksize %d (%d log_block_size)",
3842 blocksize, le32_to_cpu(es->s_log_block_size));
3843 goto failed_mount;
3844 }
3845
9803387c
TT		 3846 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3847 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3848 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3849 } else {
3850 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3851 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3852 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3853 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3854 sbi->s_first_ino);
3855 goto failed_mount;
3856 }
3857 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3858 (!is_power_of_2(sbi->s_inode_size)) ||
3859 (sbi->s_inode_size > blocksize)) {
3860 ext4_msg(sb, KERN_ERR,
3861 "unsupported inode size: %d",
3862 sbi->s_inode_size);
4f97a681 3863 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
9803387c
TT		 3864 goto failed_mount;
3865 }
3866 /*
3867 * i_atime_extra is the last extra field available for
3868 * [acm]times in struct ext4_inode. Checking for that
3869 * field should suffice to ensure we have extra space
3870 * for all three.
3871 */
3872 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
3873 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
3874 sb->s_time_gran = 1;
3875 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
3876 } else {
3877 sb->s_time_gran = NSEC_PER_SEC;
3878 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
3879 }
3880 sb->s_time_min = EXT4_TIMESTAMP_MIN;
3881 }
3882 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3883 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3884 EXT4_GOOD_OLD_INODE_SIZE;
3885 if (ext4_has_feature_extra_isize(sb)) {
3886 unsigned v, max = (sbi->s_inode_size -
3887 EXT4_GOOD_OLD_INODE_SIZE);
3888
3889 v = le16_to_cpu(es->s_want_extra_isize);
3890 if (v > max) {
3891 ext4_msg(sb, KERN_ERR,
3892 "bad s_want_extra_isize: %d", v);
3893 goto failed_mount;
3894 }
3895 if (sbi->s_want_extra_isize < v)
3896 sbi->s_want_extra_isize = v;
3897
3898 v = le16_to_cpu(es->s_min_extra_isize);
3899 if (v > max) {
3900 ext4_msg(sb, KERN_ERR,
3901 "bad s_min_extra_isize: %d", v);
3902 goto failed_mount;
3903 }
3904 if (sbi->s_want_extra_isize < v)
3905 sbi->s_want_extra_isize = v;
3906 }
3907 }
3908
5aee0f8a
TT		 3909 if (sbi->s_es->s_mount_opts[0]) {
3910 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3911 sizeof(sbi->s_es->s_mount_opts),
3912 GFP_KERNEL);
3913 if (!s_mount_opts)
3914 goto failed_mount;
3915 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3916 &journal_ioprio, 0)) {
3917 ext4_msg(sb, KERN_WARNING,
3918 "failed to parse options in superblock: %s",
3919 s_mount_opts);
3920 }
3921 kfree(s_mount_opts);
8b67f04a 3922 }
5a916be1 3923 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3924 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3925 &journal_ioprio, 0))
ac27a0ec
DK		 3926 goto failed_mount;
3927
c83ad55e
GKB		 3928#ifdef CONFIG_UNICODE
3929 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
3930 const struct ext4_sb_encodings *encoding_info;
3931 struct unicode_map *encoding;
3932 __u16 encoding_flags;
3933
3934 if (ext4_has_feature_encrypt(sb)) {
3935 ext4_msg(sb, KERN_ERR,
3936 "Can't mount with encoding and encryption");
3937 goto failed_mount;
3938 }
3939
3940 if (ext4_sb_read_encoding(es, &encoding_info,
3941 &encoding_flags)) {
3942 ext4_msg(sb, KERN_ERR,
3943 "Encoding requested by superblock is unknown");
3944 goto failed_mount;
3945 }
3946
3947 encoding = utf8_load(encoding_info->version);
3948 if (IS_ERR(encoding)) {
3949 ext4_msg(sb, KERN_ERR,
3950 "can't mount with superblock charset: %s-%s "
3951 "not supported by the kernel. flags: 0x%x.",
3952 encoding_info->name, encoding_info->version,
3953 encoding_flags);
3954 goto failed_mount;
3955 }
3956 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
3957 "%s-%s with flags 0x%hx", encoding_info->name,
3958 encoding_info->version?:"\b", encoding_flags);
3959
3960 sbi->s_encoding = encoding;
3961 sbi->s_encoding_flags = encoding_flags;
3962 }
3963#endif
3964
56889787 3965 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
244adf64
TT		 3966 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n");
3967 clear_opt(sb, DIOREAD_NOLOCK);
56889787
TT		 3968 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3969 ext4_msg(sb, KERN_ERR, "can't mount with "
3970 "both data=journal and delalloc");
3971 goto failed_mount;
3972 }
3973 if (test_opt(sb, DIOREAD_NOLOCK)) {
3974 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3975 "both data=journal and dioread_nolock");
56889787
TT		 3976 goto failed_mount;
3977 }
923ae0ff
RZ		 3978 if (test_opt(sb, DAX)) {
3979 ext4_msg(sb, KERN_ERR, "can't mount with "
3980 "both data=journal and dax");
3981 goto failed_mount;
3982 }
73b92a2a
SK		 3983 if (ext4_has_feature_encrypt(sb)) {
3984 ext4_msg(sb, KERN_WARNING,
3985 "encrypted files will use data=ordered "
3986 "instead of data journaling mode");
3987 }
56889787
TT		 3988 if (test_opt(sb, DELALLOC))
3989 clear_opt(sb, DELALLOC);
001e4a87
TH		 3990 } else {
3991 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3992 }
3993
1751e8a6
LT		 3994 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3995 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 3996
617ba13b 3997 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3998 (ext4_has_compat_features(sb) ||
3999 ext4_has_ro_compat_features(sb) ||
4000 ext4_has_incompat_features(sb)))
b31e1552
ES		 4001 ext4_msg(sb, KERN_WARNING,
4002 "feature flags set on rev 0 fs, "
4003 "running e2fsck is recommended");
469108ff 4004
ed3654eb
TT		 4005 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4006 set_opt2(sb, HURD_COMPAT);
e2b911c5 4007 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 4008 ext4_msg(sb, KERN_ERR,
4009 "The Hurd can't support 64-bit file systems");
4010 goto failed_mount;
4011 }
dec214d0
TE		 4012
4013 /*
4014 * ea_inode feature uses l_i_version field which is not
4015 * available in HURD_COMPAT mode.
4016 */
4017 if (ext4_has_feature_ea_inode(sb)) {
4018 ext4_msg(sb, KERN_ERR,
4019 "ea_inode feature is not supported for Hurd");
4020 goto failed_mount;
4021 }
ed3654eb
TT		 4022 }
4023
2035e776
TT		 4024 if (IS_EXT2_SB(sb)) {
4025 if (ext2_feature_set_ok(sb))
4026 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4027 "using the ext4 subsystem");
4028 else {
0d9366d6
ES		 4029 /*
4030 * If we're probing be silent, if this looks like
4031 * it's actually an ext[34] filesystem.
4032 */
4033 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4034 goto failed_mount;
2035e776
TT		 4035 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4036 "to feature incompatibilities");
4037 goto failed_mount;
4038 }
4039 }
4040
4041 if (IS_EXT3_SB(sb)) {
4042 if (ext3_feature_set_ok(sb))
4043 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4044 "using the ext4 subsystem");
4045 else {
0d9366d6
ES		 4046 /*
4047 * If we're probing be silent, if this looks like
4048 * it's actually an ext4 filesystem.
4049 */
4050 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4051 goto failed_mount;
2035e776
TT		 4052 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4053 "to feature incompatibilities");
4054 goto failed_mount;
4055 }
4056 }
4057
ac27a0ec
DK		 4058 /*
4059 * Check feature flags regardless of the revision level, since we
4060 * previously didn't change the revision level when setting the flags,
4061 * so there is a chance incompat flags are set on a rev 0 filesystem.
4062 */
bc98a42c 4063 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
ac27a0ec 4064 goto failed_mount;
a13fb1a4 4065
8cdf3372
TT		 4066 if (le32_to_cpu(es->s_log_block_size) >
4067 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4068 ext4_msg(sb, KERN_ERR,
4069 "Invalid log block size: %u",
4070 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 4071 goto failed_mount;
4072 }
bfe0a5f4
TT		 4073 if (le32_to_cpu(es->s_log_cluster_size) >
4074 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4075 ext4_msg(sb, KERN_ERR,
4076 "Invalid log cluster size: %u",
4077 le32_to_cpu(es->s_log_cluster_size));
4078 goto failed_mount;
4079 }
ac27a0ec 4080
5b9554dc
TT		 4081 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4082 ext4_msg(sb, KERN_ERR,
4083 "Number of reserved GDT blocks insanely large: %d",
4084 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4085 goto failed_mount;
4086 }
4087
923ae0ff 4088 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
559db4c6
RZ		 4089 if (ext4_has_feature_inline_data(sb)) {
4090 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4091 " that may contain inline data");
361d24d4 4092 goto failed_mount;
559db4c6 4093 }
80660f20 4094 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
24f3478d 4095 ext4_msg(sb, KERN_ERR,
361d24d4
ES		 4096 "DAX unsupported by block device.");
4097 goto failed_mount;
24f3478d 4098 }
923ae0ff
RZ		 4099 }
4100
e2b911c5 4101 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 4102 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4103 es->s_encryption_level);
4104 goto failed_mount;
4105 }
4106
ac27a0ec 4107 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 4108 /* Validate the filesystem blocksize */
4109 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 4110 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 4111 blocksize);
ac27a0ec
DK		 4112 goto failed_mount;
4113 }
4114
2b2d6d01 4115 brelse(bh);
70bbb3e0
AM		 4116 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4117 offset = do_div(logical_sb_block, blocksize);
a8ac900b 4118 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 4119 if (!bh) {
b31e1552
ES		 4120 ext4_msg(sb, KERN_ERR,
4121 "Can't read superblock on 2nd try");
ac27a0ec
DK		 4122 goto failed_mount;
4123 }
2716b802 4124 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 4125 sbi->s_es = es;
617ba13b 4126 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 4127 ext4_msg(sb, KERN_ERR,
4128 "Magic mismatch, very weird!");
ac27a0ec
DK		 4129 goto failed_mount;
4130 }
4131 }
4132
e2b911c5 4133 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 4134 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4135 has_huge_files);
4136 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 4137
0d1ee42f 4138 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 4139 if (ext4_has_feature_64bit(sb)) {
8fadc143 4140 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 4141 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 4142 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 4143 ext4_msg(sb, KERN_ERR,
4144 "unsupported descriptor size %lu",
0d1ee42f
AR		 4145 sbi->s_desc_size);
4146 goto failed_mount;
4147 }
4148 } else
4149 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 4150
ac27a0ec 4151 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 4152 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 4153
617ba13b 4154 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 4155 if (sbi->s_inodes_per_block == 0)
617ba13b 4156 goto cantfind_ext4;
cd6bb35b
TT		 4157 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4158 sbi->s_inodes_per_group > blocksize * 8) {
4159 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4160 sbi->s_blocks_per_group);
4161 goto failed_mount;
4162 }
ac27a0ec
DK		 4163 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4164 sbi->s_inodes_per_block;
0d1ee42f 4165 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 4166 sbi->s_sbh = bh;
4167 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 4168 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4169 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 4170
2b2d6d01 4171 for (i = 0; i < 4; i++)
ac27a0ec
DK		 4172 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4173 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 4174 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 4175 i = le32_to_cpu(es->s_flags);
4176 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4177 sbi->s_hash_unsigned = 3;
4178 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 4179#ifdef __CHAR_UNSIGNED__
bc98a42c 4180 if (!sb_rdonly(sb))
23301410
TT		 4181 es->s_flags |=
4182 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4183 sbi->s_hash_unsigned = 3;
f99b2589 4184#else
bc98a42c 4185 if (!sb_rdonly(sb))
23301410
TT		 4186 es->s_flags |=
4187 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 4188#endif
23301410 4189 }
f99b2589 4190 }
ac27a0ec 4191
281b5995
TT		 4192 /* Handle clustersize */
4193 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 4194 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 4195 if (has_bigalloc) {
4196 if (clustersize < blocksize) {
4197 ext4_msg(sb, KERN_ERR,
4198 "cluster size (%d) smaller than "
4199 "block size (%d)", clustersize, blocksize);
4200 goto failed_mount;
4201 }
4202 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4203 le32_to_cpu(es->s_log_block_size);
4204 sbi->s_clusters_per_group =
4205 le32_to_cpu(es->s_clusters_per_group);
4206 if (sbi->s_clusters_per_group > blocksize * 8) {
4207 ext4_msg(sb, KERN_ERR,
4208 "#clusters per group too big: %lu",
4209 sbi->s_clusters_per_group);
4210 goto failed_mount;
4211 }
4212 if (sbi->s_blocks_per_group !=
4213 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4214 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4215 "clusters per group (%lu) inconsistent",
4216 sbi->s_blocks_per_group,
4217 sbi->s_clusters_per_group);
4218 goto failed_mount;
4219 }
4220 } else {
4221 if (clustersize != blocksize) {
bfe0a5f4
TT		 4222 ext4_msg(sb, KERN_ERR,
4223 "fragment/cluster size (%d) != "
4224 "block size (%d)", clustersize, blocksize);
4225 goto failed_mount;
281b5995
TT		 4226 }
4227 if (sbi->s_blocks_per_group > blocksize * 8) {
4228 ext4_msg(sb, KERN_ERR,
4229 "#blocks per group too big: %lu",
4230 sbi->s_blocks_per_group);
4231 goto failed_mount;
4232 }
4233 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4234 sbi->s_cluster_bits = 0;
ac27a0ec 4235 }
281b5995
TT		 4236 sbi->s_cluster_ratio = clustersize / blocksize;
4237
960fd856
TT		 4238 /* Do we have standard group size of clustersize * 8 blocks ? */
4239 if (sbi->s_blocks_per_group == clustersize << 3)
4240 set_opt2(sb, STD_GROUP_SIZE);
4241
bf43d84b
ES		 4242 /*
4243 * Test whether we have more sectors than will fit in sector_t,
4244 * and whether the max offset is addressable by the page cache.
4245 */
5a9ae68a 4246 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 4247 ext4_blocks_count(es));
5a9ae68a 4248 if (err) {
b31e1552 4249 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 4250 " too large to mount safely on this system");
ac27a0ec
DK		 4251 goto failed_mount;
4252 }
4253
617ba13b
MC		 4254 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4255 goto cantfind_ext4;
e7c95593 4256
0f2ddca6
FTN		 4257 /* check blocks count against device size */
4258 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4259 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 4260 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4261 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 4262 ext4_blocks_count(es), blocks_count);
4263 goto failed_mount;
4264 }
4265
0b8e58a1
AD		 4266 /*
4267 * It makes no sense for the first data block to be beyond the end
4268 * of the filesystem.
4269 */
4270 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 4271 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 4272 "block %u is beyond end of filesystem (%llu)",
4273 le32_to_cpu(es->s_first_data_block),
4274 ext4_blocks_count(es));
e7c95593
ES		 4275 goto failed_mount;
4276 }
bfe0a5f4
TT		 4277 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4278 (sbi->s_cluster_ratio == 1)) {
4279 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4280 "block is 0 with a 1k block and cluster size");
4281 goto failed_mount;
4282 }
4283
bd81d8ee
LV		 4284 blocks_count = (ext4_blocks_count(es) -
4285 le32_to_cpu(es->s_first_data_block) +
4286 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4287 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 4288 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 4289 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 4290 "(block count %llu, first data block %u, "
b31e1552 4291 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 4292 ext4_blocks_count(es),
4293 le32_to_cpu(es->s_first_data_block),
4294 EXT4_BLOCKS_PER_GROUP(sb));
4295 goto failed_mount;
4296 }
bd81d8ee 4297 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 4298 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4299 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
9e463084
TT		 4300 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4301 le32_to_cpu(es->s_inodes_count)) {
4302 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4303 le32_to_cpu(es->s_inodes_count),
4304 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4305 ret = -EINVAL;
4306 goto failed_mount;
4307 }
617ba13b
MC		 4308 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4309 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 4310 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 4311 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 4312 ext4_msg(sb, KERN_WARNING,
4313 "first meta block group too large: %u "
4314 "(group descriptor block count %u)",
4315 le32_to_cpu(es->s_first_meta_bg), db_count);
4316 goto failed_mount;
4317 }
4318 }
1d0c3924
TT		 4319 rcu_assign_pointer(sbi->s_group_desc,
4320 kvmalloc_array(db_count,
4321 sizeof(struct buffer_head *),
4322 GFP_KERNEL));
ac27a0ec 4323 if (sbi->s_group_desc == NULL) {
b31e1552 4324 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 4325 ret = -ENOMEM;
ac27a0ec
DK		 4326 goto failed_mount;
4327 }
4328
705895b6 4329 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 4330
85c8f176
AP		 4331 /* Pre-read the descriptors into the buffer cache */
4332 for (i = 0; i < db_count; i++) {
4333 block = descriptor_loc(sb, logical_sb_block, i);
4334 sb_breadahead(sb, block);
4335 }
4336
ac27a0ec 4337 for (i = 0; i < db_count; i++) {
1d0c3924
TT		 4338 struct buffer_head *bh;
4339
70bbb3e0 4340 block = descriptor_loc(sb, logical_sb_block, i);
1d0c3924
TT		 4341 bh = sb_bread_unmovable(sb, block);
4342 if (!bh) {
b31e1552
ES		 4343 ext4_msg(sb, KERN_ERR,
4344 "can't read group descriptor %d", i);
ac27a0ec
DK		 4345 db_count = i;
4346 goto failed_mount2;
4347 }
1d0c3924
TT		 4348 rcu_read_lock();
4349 rcu_dereference(sbi->s_group_desc)[i] = bh;
4350 rcu_read_unlock();
ac27a0ec 4351 }
44de022c 4352 sbi->s_gdb_count = db_count;
829fa70d 4353 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 4354 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 4355 ret = -EFSCORRUPTED;
f9ae9cf5 4356 goto failed_mount2;
ac27a0ec 4357 }
772cb7c8 4358
235699a8 4359 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
04496411 4360
a75ae78f 4361 /* Register extent status tree shrinker */
eb68d0e2 4362 if (ext4_es_register_shrinker(sbi))
ce7e010a 4363 goto failed_mount3;
ce7e010a 4364
c9de560d 4365 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 4366 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 4367
f9ae9cf5
TT		 4368 /*
4369 * set up enough so that it can read an inode
4370 */
f6e63f90 4371 sb->s_op = &ext4_sops;
617ba13b
MC		 4372 sb->s_export_op = &ext4_export_ops;
4373 sb->s_xattr = ext4_xattr_handlers;
643fa961 4374#ifdef CONFIG_FS_ENCRYPTION
a7550b30 4375 sb->s_cop = &ext4_cryptops;
ffcc4182 4376#endif
c93d8f88
EB		 4377#ifdef CONFIG_FS_VERITY
4378 sb->s_vop = &ext4_verityops;
4379#endif
ac27a0ec 4380#ifdef CONFIG_QUOTA
617ba13b 4381 sb->dq_op = &ext4_quota_operations;
e2b911c5 4382 if (ext4_has_feature_quota(sb))
1fa5efe3 4383 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 4384 else
4385 sb->s_qcop = &ext4_qctl_operations;
689c958c 4386 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 4387#endif
85787090 4388 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 4389
ac27a0ec 4390 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4391 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 4392
4393 sb->s_root = NULL;
4394
4395 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 4396 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 4397
bc98a42c 4398 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 4399 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4400 goto failed_mount3a;
c5e06d10 4401
ac27a0ec
DK		 4402 /*
4403 * The first inode we look at is the journal inode. Don't try
4404 * root first: it may be modified in the journal!
4405 */
e2b911c5 4406 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 4407 err = ext4_load_journal(sb, es, journal_devnum);
4408 if (err)
50460fe8 4409 goto failed_mount3a;
bc98a42c 4410 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 4411 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4412 ext4_msg(sb, KERN_ERR, "required journal recovery "
4413 "suppressed and not mounted read-only");
744692dc 4414 goto failed_mount_wq;
ac27a0ec 4415 } else {
1e381f60
DM		 4416 /* Nojournal mode, all journal mount options are illegal */
4417 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4418 ext4_msg(sb, KERN_ERR, "can't mount with "
4419 "journal_checksum, fs mounted w/o journal");
4420 goto failed_mount_wq;
4421 }
4422 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4423 ext4_msg(sb, KERN_ERR, "can't mount with "
4424 "journal_async_commit, fs mounted w/o journal");
4425 goto failed_mount_wq;
4426 }
4427 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4428 ext4_msg(sb, KERN_ERR, "can't mount with "
4429 "commit=%lu, fs mounted w/o journal",
4430 sbi->s_commit_interval / HZ);
4431 goto failed_mount_wq;
4432 }
4433 if (EXT4_MOUNT_DATA_FLAGS &
4434 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4435 ext4_msg(sb, KERN_ERR, "can't mount with "
4436 "data=, fs mounted w/o journal");
4437 goto failed_mount_wq;
4438 }
50b29d8f 4439 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
1e381f60 4440 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4441 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4442 sbi->s_journal = NULL;
4443 needs_recovery = 0;
4444 goto no_journal;
ac27a0ec
DK		 4445 }
4446
e2b911c5 4447 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4448 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4449 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4450 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4451 goto failed_mount_wq;
eb40a09c
JS		 4452 }
4453
25ed6e8a
DW		 4454 if (!set_journal_csum_feature_set(sb)) {
4455 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4456 "feature set");
4457 goto failed_mount_wq;
d4da6c9c 4458 }
818d276c 4459
ac27a0ec
DK		 4460 /* We have now updated the journal if required, so we can
4461 * validate the data journaling mode. */
4462 switch (test_opt(sb, DATA_FLAGS)) {
4463 case 0:
4464 /* No mode set, assume a default based on the journal
63f57933
AM		 4465 * capabilities: ORDERED_DATA if the journal can
4466 * cope, else JOURNAL_DATA
4467 */
dab291af 4468 if (jbd2_journal_check_available_features
27f394a7 4469 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
fd8c37ec 4470 set_opt(sb, ORDERED_DATA);
27f394a7
TN		 4471 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4472 } else {
fd8c37ec 4473 set_opt(sb, JOURNAL_DATA);
27f394a7
TN		 4474 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4475 }
ac27a0ec
DK		 4476 break;
4477
617ba13b
MC		 4478 case EXT4_MOUNT_ORDERED_DATA:
4479 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4480 if (!jbd2_journal_check_available_features
4481 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4482 ext4_msg(sb, KERN_ERR, "Journal does not support "
4483 "requested data journaling mode");
744692dc 4484 goto failed_mount_wq;
ac27a0ec
DK		 4485 }
4486 default:
4487 break;
4488 }
ab04df78
JK		 4489
4490 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4491 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4492 ext4_msg(sb, KERN_ERR, "can't mount with "
4493 "journal_async_commit in data=ordered mode");
4494 goto failed_mount_wq;
4495 }
4496
b3881f74 4497 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4498
18aadd47
BJ		 4499 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4500
ce7e010a 4501no_journal:
cdb7ee4c
TE		 4502 if (!test_opt(sb, NO_MBCACHE)) {
4503 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4504 if (!sbi->s_ea_block_cache) {
dec214d0 4505 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4506 "Failed to create ea_block_cache");
dec214d0
TE		 4507 goto failed_mount_wq;
4508 }
cdb7ee4c
TE		 4509
4510 if (ext4_has_feature_ea_inode(sb)) {
4511 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4512 if (!sbi->s_ea_inode_cache) {
4513 ext4_msg(sb, KERN_ERR,
4514 "Failed to create ea_inode_cache");
4515 goto failed_mount_wq;
4516 }
4517 }
9c191f70
M		 4518 }
4519
c93d8f88
EB		 4520 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4521 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4522 goto failed_mount_wq;
4523 }
4524
bc98a42c 4525 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
e2b911c5
DW		 4526 !ext4_has_feature_encrypt(sb)) {
4527 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4528 ext4_commit_super(sb, 1);
4529 }
4530
952fc18e
TT		 4531 /*
4532 * Get the # of file system overhead blocks from the
4533 * superblock if present.
4534 */
4535 if (es->s_overhead_clusters)
4536 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4537 else {
07aa2ea1
LC		 4538 err = ext4_calculate_overhead(sb);
4539 if (err)
952fc18e
TT		 4540 goto failed_mount_wq;
4541 }
4542
fd89d5f2
TH		 4543 /*
4544 * The maximum number of concurrent works can be high and
4545 * concurrency isn't really necessary. Limit it to 1.
4546 */
2e8fa54e
JK		 4547 EXT4_SB(sb)->rsv_conversion_wq =
4548 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4549 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4550 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4551 ret = -ENOMEM;
2e8fa54e
JK		 4552 goto failed_mount4;
4553 }
4554
ac27a0ec 4555 /*
dab291af 4556 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4557 * so we can safely mount the rest of the filesystem now.
4558 */
4559
8a363970 4560 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
1d1fe1ee 4561 if (IS_ERR(root)) {
b31e1552 4562 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4563 ret = PTR_ERR(root);
32a9bb57 4564 root = NULL;
ac27a0ec
DK		 4565 goto failed_mount4;
4566 }
4567 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4568 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4569 iput(root);
ac27a0ec
DK		 4570 goto failed_mount4;
4571 }
b886ee3e
GKB		 4572
4573#ifdef CONFIG_UNICODE
4574 if (sbi->s_encoding)
4575 sb->s_d_op = &ext4_dentry_ops;
4576#endif
4577
48fde701 4578 sb->s_root = d_make_root(root);
1d1fe1ee 4579 if (!sb->s_root) {
b31e1552 4580 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4581 ret = -ENOMEM;
4582 goto failed_mount4;
4583 }
ac27a0ec 4584
c89128a0
JK		 4585 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4586 if (ret == -EROFS) {
1751e8a6 4587 sb->s_flags |= SB_RDONLY;
c89128a0
JK		 4588 ret = 0;
4589 } else if (ret)
4590 goto failed_mount4a;
ef7f3835 4591
b5799018 4592 ext4_set_resv_clusters(sb);
27dd4385 4593
6fd058f7
TT		 4594 err = ext4_setup_system_zone(sb);
4595 if (err) {
b31e1552 4596 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4597 "zone (%d)", err);
f9ae9cf5
TT		 4598 goto failed_mount4a;
4599 }
4600
4601 ext4_ext_init(sb);
4602 err = ext4_mb_init(sb);
4603 if (err) {
4604 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4605 err);
dcf2d804 4606 goto failed_mount5;
c2774d84
AK		 4607 }
4608
d5e03cbb
TT		 4609 block = ext4_count_free_clusters(sb);
4610 ext4_free_blocks_count_set(sbi->s_es,
4611 EXT4_C2B(sbi, block));
4274f516 4612 ext4_superblock_csum_set(sb);
908c7f19
TH		 4613 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4614 GFP_KERNEL);
d5e03cbb
TT		 4615 if (!err) {
4616 unsigned long freei = ext4_count_free_inodes(sb);
4617 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4274f516 4618 ext4_superblock_csum_set(sb);
908c7f19
TH		 4619 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4620 GFP_KERNEL);
d5e03cbb
TT		 4621 }
4622 if (!err)
4623 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4624 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4625 if (!err)
908c7f19
TH		 4626 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4627 GFP_KERNEL);
c8585c6f 4628 if (!err)
bbd55937 4629 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
c8585c6f 4630
d5e03cbb
TT		 4631 if (err) {
4632 ext4_msg(sb, KERN_ERR, "insufficient memory");
4633 goto failed_mount6;
4634 }
4635
e2b911c5 4636 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4637 if (!ext4_fill_flex_info(sb)) {
4638 ext4_msg(sb, KERN_ERR,
4639 "unable to initialize "
4640 "flex_bg meta info!");
4641 goto failed_mount6;
4642 }
4643
bfff6873
LC		 4644 err = ext4_register_li_request(sb, first_not_zeroed);
4645 if (err)
dcf2d804 4646 goto failed_mount6;
bfff6873 4647
b5799018 4648 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4649 if (err)
4650 goto failed_mount7;
3197ebdb 4651
9b2ff357
JK		 4652#ifdef CONFIG_QUOTA
4653 /* Enable quota usage during mount. */
bc98a42c 4654 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 4655 err = ext4_enable_quotas(sb);
4656 if (err)
4657 goto failed_mount8;
4658 }
4659#endif /* CONFIG_QUOTA */
4660
617ba13b
MC		 4661 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4662 ext4_orphan_cleanup(sb, es);
4663 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4664 if (needs_recovery) {
b31e1552 4665 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4666 ext4_mark_recovery_complete(sb, es);
4667 }
4668 if (EXT4_SB(sb)->s_journal) {
4669 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4670 descr = " journalled data mode";
4671 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4672 descr = " ordered data mode";
4673 else
4674 descr = " writeback data mode";
4675 } else
4676 descr = "out journal";
4677
79add3a3
LC		 4678 if (test_opt(sb, DISCARD)) {
4679 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4680 if (!blk_queue_discard(q))
4681 ext4_msg(sb, KERN_WARNING,
4682 "mounting with \"discard\" option, but "
4683 "the device does not support discard");
4684 }
4685
e294a537
TT		 4686 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4687 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4688 "Opts: %.*s%s%s", descr,
4689 (int) sizeof(sbi->s_es->s_mount_opts),
4690 sbi->s_es->s_mount_opts,
e294a537 4691 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4692
66e61a9e
TT		 4693 if (es->s_error_count)
4694 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4695
efbed4dc
TT		 4696 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4697 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4698 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4699 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4700
d4c402d9 4701 kfree(orig_data);
ac27a0ec
DK		 4702 return 0;
4703
617ba13b 4704cantfind_ext4:
ac27a0ec 4705 if (!silent)
b31e1552 4706 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4707 goto failed_mount;
4708
72ba7450
TT		 4709#ifdef CONFIG_QUOTA
4710failed_mount8:
ebd173be 4711 ext4_unregister_sysfs(sb);
72ba7450 4712#endif
dcf2d804
TM		 4713failed_mount7:
4714 ext4_unregister_li_request(sb);
4715failed_mount6:
f9ae9cf5 4716 ext4_mb_release(sb);
7c990728
SJS		 4717 rcu_read_lock();
4718 flex_groups = rcu_dereference(sbi->s_flex_groups);
4719 if (flex_groups) {
4720 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4721 kvfree(flex_groups[i]);
4722 kvfree(flex_groups);
4723 }
4724 rcu_read_unlock();
d5e03cbb
TT		 4725 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4726 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4727 percpu_counter_destroy(&sbi->s_dirs_counter);
4728 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
bbd55937 4729 percpu_free_rwsem(&sbi->s_writepages_rwsem);
00764937 4730failed_mount5:
f9ae9cf5
TT		 4731 ext4_ext_release(sb);
4732 ext4_release_system_zone(sb);
4733failed_mount4a:
94bf608a 4734 dput(sb->s_root);
32a9bb57 4735 sb->s_root = NULL;
94bf608a 4736failed_mount4:
b31e1552 4737 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4738 if (EXT4_SB(sb)->rsv_conversion_wq)
4739 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4740failed_mount_wq:
50c15df6
CX		 4741 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4742 sbi->s_ea_inode_cache = NULL;
4743
4744 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4745 sbi->s_ea_block_cache = NULL;
4746
0390131b
FM		 4747 if (sbi->s_journal) {
4748 jbd2_journal_destroy(sbi->s_journal);
4749 sbi->s_journal = NULL;
4750 }
50460fe8 4751failed_mount3a:
d3922a77 4752 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4753failed_mount3:
9105bb14 4754 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4755 if (sbi->s_mmp_tsk)
4756 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 4757failed_mount2:
1d0c3924
TT		 4758 rcu_read_lock();
4759 group_desc = rcu_dereference(sbi->s_group_desc);
ac27a0ec 4760 for (i = 0; i < db_count; i++)
1d0c3924
TT		 4761 brelse(group_desc[i]);
4762 kvfree(group_desc);
4763 rcu_read_unlock();
ac27a0ec 4764failed_mount:
0441984a
DW		 4765 if (sbi->s_chksum_driver)
4766 crypto_free_shash(sbi->s_chksum_driver);
c83ad55e
GKB		 4767
4768#ifdef CONFIG_UNICODE
4769 utf8_unload(sbi->s_encoding);
4770#endif
4771
ac27a0ec 4772#ifdef CONFIG_QUOTA
a2d4a646 4773 for (i = 0; i < EXT4_MAXQUOTAS; i++)
0ba33fac 4774 kfree(get_qf_name(sb, sbi, i));
ac27a0ec 4775#endif
617ba13b 4776 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4777 brelse(bh);
4778out_fail:
4779 sb->s_fs_info = NULL;
f6830165 4780 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4781out_free_base:
ac27a0ec 4782 kfree(sbi);
d4c402d9 4783 kfree(orig_data);
5e405595 4784 fs_put_dax(dax_dev);
07aa2ea1 4785 return err ? err : ret;
ac27a0ec
DK		 4786}
4787
4788/*
4789 * Setup any per-fs journal parameters now. We'll do this both on
4790 * initial mount, once the journal has been initialised but before we've
4791 * done any recovery; and again on any subsequent remount.
4792 */
617ba13b 4793static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4794{
617ba13b 4795 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4796
30773840
TT		 4797 journal->j_commit_interval = sbi->s_commit_interval;
4798 journal->j_min_batch_time = sbi->s_min_batch_time;
4799 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4800
a931da6a 4801 write_lock(&journal->j_state_lock);
ac27a0ec 4802 if (test_opt(sb, BARRIER))
dab291af 4803 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4804 else
dab291af 4805 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4806 if (test_opt(sb, DATA_ERR_ABORT))
4807 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4808 else
4809 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4810 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4811}
4812
c6cb7e77
EW		 4813static struct inode *ext4_get_journal_inode(struct super_block *sb,
4814 unsigned int journal_inum)
ac27a0ec
DK		 4815{
4816 struct inode *journal_inode;
ac27a0ec 4817
c6cb7e77
EW		 4818 /*
4819 * Test for the existence of a valid inode on disk. Bad things
4820 * happen if we iget() an unused inode, as the subsequent iput()
4821 * will try to delete it.
4822 */
8a363970 4823 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
1d1fe1ee 4824 if (IS_ERR(journal_inode)) {
b31e1552 4825 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4826 return NULL;
4827 }
4828 if (!journal_inode->i_nlink) {
4829 make_bad_inode(journal_inode);
4830 iput(journal_inode);
b31e1552 4831 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4832 return NULL;
4833 }
4834
e5f8eab8 4835 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4836 journal_inode, journal_inode->i_size);
1d1fe1ee 4837 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4838 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4839 iput(journal_inode);
4840 return NULL;
4841 }
c6cb7e77
EW		 4842 return journal_inode;
4843}
4844
4845static journal_t *ext4_get_journal(struct super_block *sb,
4846 unsigned int journal_inum)
4847{
4848 struct inode *journal_inode;
4849 journal_t *journal;
4850
4851 BUG_ON(!ext4_has_feature_journal(sb));
4852
4853 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4854 if (!journal_inode)
4855 return NULL;
ac27a0ec 4856
dab291af 4857 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4858 if (!journal) {
b31e1552 4859 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4860 iput(journal_inode);
4861 return NULL;
4862 }
4863 journal->j_private = sb;
617ba13b 4864 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4865 return journal;
4866}
4867
617ba13b 4868static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4869 dev_t j_dev)
4870{
2b2d6d01 4871 struct buffer_head *bh;
ac27a0ec 4872 journal_t *journal;
617ba13b
MC		 4873 ext4_fsblk_t start;
4874 ext4_fsblk_t len;
ac27a0ec 4875 int hblock, blocksize;
617ba13b 4876 ext4_fsblk_t sb_block;
ac27a0ec 4877 unsigned long offset;
2b2d6d01 4878 struct ext4_super_block *es;
ac27a0ec
DK		 4879 struct block_device *bdev;
4880
e2b911c5 4881 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4882
b31e1552 4883 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4884 if (bdev == NULL)
4885 return NULL;
4886
ac27a0ec 4887 blocksize = sb->s_blocksize;
e1defc4f 4888 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4889 if (blocksize < hblock) {
b31e1552
ES		 4890 ext4_msg(sb, KERN_ERR,
4891 "blocksize too small for journal device");
ac27a0ec
DK		 4892 goto out_bdev;
4893 }
4894
617ba13b
MC		 4895 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4896 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4897 set_blocksize(bdev, blocksize);
4898 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4899 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4900 "external journal");
ac27a0ec
DK		 4901 goto out_bdev;
4902 }
4903
2716b802 4904 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4905 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4906 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4907 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4908 ext4_msg(sb, KERN_ERR, "external journal has "
4909 "bad superblock");
ac27a0ec
DK		 4910 brelse(bh);
4911 goto out_bdev;
4912 }
4913
df4763be
DW		 4914 if ((le32_to_cpu(es->s_feature_ro_compat) &
4915 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4916 es->s_checksum != ext4_superblock_csum(sb, es)) {
4917 ext4_msg(sb, KERN_ERR, "external journal has "
4918 "corrupt superblock");
4919 brelse(bh);
4920 goto out_bdev;
4921 }
4922
617ba13b 4923 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4924 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4925 brelse(bh);
4926 goto out_bdev;
4927 }
4928
bd81d8ee 4929 len = ext4_blocks_count(es);
ac27a0ec
DK		 4930 start = sb_block + 1;
4931 brelse(bh); /* we're done with the superblock */
4932
dab291af 4933 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4934 start, len, blocksize);
4935 if (!journal) {
b31e1552 4936 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4937 goto out_bdev;
4938 }
4939 journal->j_private = sb;
dfec8a14 4940 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4941 wait_on_buffer(journal->j_sb_buffer);
4942 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4943 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4944 goto out_journal;
4945 }
4946 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4947 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4948 "user (unsupported) - %d",
ac27a0ec
DK		 4949 be32_to_cpu(journal->j_superblock->s_nr_users));
4950 goto out_journal;
4951 }
617ba13b
MC		 4952 EXT4_SB(sb)->journal_bdev = bdev;
4953 ext4_init_journal_params(sb, journal);
ac27a0ec 4954 return journal;
0b8e58a1 4955
ac27a0ec 4956out_journal:
dab291af 4957 jbd2_journal_destroy(journal);
ac27a0ec 4958out_bdev:
617ba13b 4959 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4960 return NULL;
4961}
4962
617ba13b
MC		 4963static int ext4_load_journal(struct super_block *sb,
4964 struct ext4_super_block *es,
ac27a0ec
DK		 4965 unsigned long journal_devnum)
4966{
4967 journal_t *journal;
4968 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4969 dev_t journal_dev;
4970 int err = 0;
4971 int really_read_only;
4972
e2b911c5 4973 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4974
ac27a0ec
DK		 4975 if (journal_devnum &&
4976 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4977 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4978 "numbers have changed");
ac27a0ec
DK		 4979 journal_dev = new_decode_dev(journal_devnum);
4980 } else
4981 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4982
4983 really_read_only = bdev_read_only(sb->s_bdev);
4984
4985 /*
4986 * Are we loading a blank journal or performing recovery after a
4987 * crash? For recovery, we need to check in advance whether we
4988 * can get read-write access to the device.
4989 */
e2b911c5 4990 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 4991 if (sb_rdonly(sb)) {
b31e1552
ES		 4992 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4993 "required on readonly filesystem");
ac27a0ec 4994 if (really_read_only) {
b31e1552 4995 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 4996 "unavailable, cannot proceed "
4997 "(try mounting with noload)");
ac27a0ec
DK		 4998 return -EROFS;
4999 }
b31e1552
ES		 5000 ext4_msg(sb, KERN_INFO, "write access will "
5001 "be enabled during recovery");
ac27a0ec
DK		 5002 }
5003 }
5004
5005 if (journal_inum && journal_dev) {
b31e1552
ES		 5006 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
5007 "and inode journals!");
ac27a0ec
DK		 5008 return -EINVAL;
5009 }
5010
5011 if (journal_inum) {
617ba13b 5012 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 5013 return -EINVAL;
5014 } else {
617ba13b 5015 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 5016 return -EINVAL;
5017 }
5018
90576c0b 5019 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 5020 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 5021
e2b911c5 5022 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 5023 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 5024 if (!err) {
5025 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5026 if (save)
5027 memcpy(save, ((char *) es) +
5028 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 5029 err = jbd2_journal_load(journal);
1c13d5c0
TT		 5030 if (save)
5031 memcpy(((char *) es) + EXT4_S_ERR_START,
5032 save, EXT4_S_ERR_LEN);
5033 kfree(save);
5034 }
ac27a0ec
DK		 5035
5036 if (err) {
b31e1552 5037 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 5038 jbd2_journal_destroy(journal);
ac27a0ec
DK		 5039 return err;
5040 }
5041
617ba13b
MC		 5042 EXT4_SB(sb)->s_journal = journal;
5043 ext4_clear_journal_err(sb, es);
ac27a0ec 5044
c41303ce 5045 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 5046 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5047 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 5048
5049 /* Make sure we flush the recovery flag to disk. */
e2d67052 5050 ext4_commit_super(sb, 1);
ac27a0ec
DK		 5051 }
5052
5053 return 0;
5054}
5055
e2d67052 5056static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 5057{
e2d67052 5058 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 5059 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 5060 int error = 0;
ac27a0ec 5061
bdfe0cbd 5062 if (!sbh || block_device_ejected(sb))
c4be0c1d 5063 return error;
a17712c8
JD		 5064
5065 /*
5066 * The superblock bh should be mapped, but it might not be if the
5067 * device was hot-removed. Not much we can do but fail the I/O.
5068 */
5069 if (!buffer_mapped(sbh))
5070 return error;
5071
71290b36
TT		 5072 /*
5073 * If the file system is mounted read-only, don't update the
5074 * superblock write time. This avoids updating the superblock
5075 * write time when we are mounting the root file system
5076 * read/only but we need to replay the journal; at that point,
5077 * for people who are east of GMT and who make their clock
5078 * tick in localtime for Windows bug-for-bug compatibility,
5079 * the clock is set in the future, and this will cause e2fsck
5080 * to complain and force a full file system check.
5081 */
1751e8a6 5082 if (!(sb->s_flags & SB_RDONLY))
6a0678a7 5083 ext4_update_tstamp(es, s_wtime);
f613dfcb
TT		 5084 if (sb->s_bdev->bd_part)
5085 es->s_kbytes_written =
5086 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
dbae2c55
MC		 5087 ((part_stat_read(sb->s_bdev->bd_part,
5088 sectors[STAT_WRITE]) -
afc32f7e 5089 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 5090 else
5091 es->s_kbytes_written =
5092 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 5093 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5094 ext4_free_blocks_count_set(es,
57042651
TT		 5095 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5096 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 5097 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5098 es->s_free_inodes_count =
5099 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 5100 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 5101 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 5102 ext4_superblock_csum_set(sb);
1566a48a
TT		 5103 if (sync)
5104 lock_buffer(sbh);
e8680786 5105 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4743f839
PKS		 5106 /*
5107 * Oh, dear. A previous attempt to write the
5108 * superblock failed. This could happen because the
5109 * USB device was yanked out. Or it could happen to
5110 * be a transient write error and maybe the block will
5111 * be remapped. Nothing we can do but to retry the
5112 * write and hope for the best.
5113 */
5114 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5115 "superblock detected");
5116 clear_buffer_write_io_error(sbh);
5117 set_buffer_uptodate(sbh);
5118 }
ac27a0ec 5119 mark_buffer_dirty(sbh);
914258bf 5120 if (sync) {
1566a48a 5121 unlock_buffer(sbh);
564bc402 5122 error = __sync_dirty_buffer(sbh,
00473374 5123 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
c89128a0 5124 if (buffer_write_io_error(sbh)) {
b31e1552
ES		 5125 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5126 "superblock");
914258bf
TT		 5127 clear_buffer_write_io_error(sbh);
5128 set_buffer_uptodate(sbh);
5129 }
5130 }
c4be0c1d 5131 return error;
ac27a0ec
DK		 5132}
5133
ac27a0ec
DK		 5134/*
5135 * Have we just finished recovery? If so, and if we are mounting (or
5136 * remounting) the filesystem readonly, then we will end up with a
5137 * consistent fs on disk. Record that fact.
5138 */
2b2d6d01
TT		 5139static void ext4_mark_recovery_complete(struct super_block *sb,
5140 struct ext4_super_block *es)
ac27a0ec 5141{
617ba13b 5142 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 5143
e2b911c5 5144 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 5145 BUG_ON(journal != NULL);
5146 return;
5147 }
dab291af 5148 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 5149 if (jbd2_journal_flush(journal) < 0)
5150 goto out;
5151
bc98a42c 5152 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
e2b911c5 5153 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 5154 ext4_commit_super(sb, 1);
ac27a0ec 5155 }
7ffe1ea8
HK		 5156
5157out:
dab291af 5158 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 5159}
5160
5161/*
5162 * If we are mounting (or read-write remounting) a filesystem whose journal
5163 * has recorded an error from a previous lifetime, move that error to the
5164 * main filesystem now.
5165 */
2b2d6d01
TT		 5166static void ext4_clear_journal_err(struct super_block *sb,
5167 struct ext4_super_block *es)
ac27a0ec
DK		 5168{
5169 journal_t *journal;
5170 int j_errno;
5171 const char *errstr;
5172
e2b911c5 5173 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 5174
617ba13b 5175 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 5176
5177 /*
5178 * Now check for any error status which may have been recorded in the
617ba13b 5179 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 5180 */
5181
dab291af 5182 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 5183 if (j_errno) {
5184 char nbuf[16];
5185
617ba13b 5186 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 5187 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 5188 "from previous mount: %s", errstr);
12062ddd 5189 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 5190
617ba13b
MC		 5191 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5192 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 5193 ext4_commit_super(sb, 1);
ac27a0ec 5194
dab291af 5195 jbd2_journal_clear_err(journal);
d796c52e 5196 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 5197 }
5198}
5199
5200/*
5201 * Force the running and committing transactions to commit,
5202 * and wait on the commit.
5203 */
617ba13b 5204int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 5205{
5206 journal_t *journal;
ac27a0ec 5207
bc98a42c 5208 if (sb_rdonly(sb))
ac27a0ec
DK		 5209 return 0;
5210
617ba13b 5211 journal = EXT4_SB(sb)->s_journal;
b1deefc9 5212 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 5213}
5214
617ba13b 5215static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 5216{
14ce0cb4 5217 int ret = 0;
9eddacf9 5218 tid_t target;
06a407f1 5219 bool needs_barrier = false;
8d5d02e6 5220 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5221
49598e04 5222 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 5223 return 0;
5224
9bffad1e 5225 trace_ext4_sync_fs(sb, wait);
2e8fa54e 5226 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 5227 /*
5228 * Writeback quota in non-journalled quota case - journalled quota has
5229 * no dirty dquots
5230 */
5231 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 5232 /*
5233 * Data writeback is possible w/o journal transaction, so barrier must
5234 * being sent at the end of the function. But we can skip it if
5235 * transaction_commit will do it for us.
5236 */
bda32530
TT		 5237 if (sbi->s_journal) {
5238 target = jbd2_get_latest_transaction(sbi->s_journal);
5239 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5240 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5241 needs_barrier = true;
5242
5243 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5244 if (wait)
5245 ret = jbd2_log_wait_commit(sbi->s_journal,
5246 target);
5247 }
5248 } else if (wait && test_opt(sb, BARRIER))
06a407f1 5249 needs_barrier = true;
06a407f1
DM		 5250 if (needs_barrier) {
5251 int err;
5252 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5253 if (!ret)
5254 ret = err;
0390131b 5255 }
06a407f1
DM		 5256
5257 return ret;
5258}
5259
ac27a0ec
DK		 5260/*
5261 * LVM calls this function before a (read-only) snapshot is created. This
5262 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 5263 *
5264 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 5265 * state independently. It relies on upper layer to stop all data & metadata
5266 * modifications.
ac27a0ec 5267 */
c4be0c1d 5268static int ext4_freeze(struct super_block *sb)
ac27a0ec 5269{
c4be0c1d
TS		 5270 int error = 0;
5271 journal_t *journal;
ac27a0ec 5272
bc98a42c 5273 if (sb_rdonly(sb))
9ca92389 5274 return 0;
ac27a0ec 5275
9ca92389 5276 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 5277
bb044576
TT		 5278 if (journal) {
5279 /* Now we set up the journal barrier. */
5280 jbd2_journal_lock_updates(journal);
ac27a0ec 5281
bb044576
TT		 5282 /*
5283 * Don't clear the needs_recovery flag if we failed to
5284 * flush the journal.
5285 */
5286 error = jbd2_journal_flush(journal);
5287 if (error < 0)
5288 goto out;
c642dc9e
ES		 5289
5290 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 5291 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 5292 }
9ca92389 5293
9ca92389 5294 error = ext4_commit_super(sb, 1);
6b0310fb 5295out:
bb044576
TT		 5296 if (journal)
5297 /* we rely on upper layer to stop further updates */
5298 jbd2_journal_unlock_updates(journal);
6b0310fb 5299 return error;
ac27a0ec
DK		 5300}
5301
5302/*
5303 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5304 * flag here, even though the filesystem is not technically dirty yet.
5305 */
c4be0c1d 5306static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 5307{
bc98a42c 5308 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 5309 return 0;
5310
c642dc9e
ES		 5311 if (EXT4_SB(sb)->s_journal) {
5312 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 5313 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 5314 }
5315
9ca92389 5316 ext4_commit_super(sb, 1);
c4be0c1d 5317 return 0;
ac27a0ec
DK		 5318}
5319
673c6100
TT		 5320/*
5321 * Structure to save mount options for ext4_remount's benefit
5322 */
5323struct ext4_mount_options {
5324 unsigned long s_mount_opt;
a2595b8a 5325 unsigned long s_mount_opt2;
08cefc7a
EB		 5326 kuid_t s_resuid;
5327 kgid_t s_resgid;
673c6100
TT		 5328 unsigned long s_commit_interval;
5329 u32 s_min_batch_time, s_max_batch_time;
5330#ifdef CONFIG_QUOTA
5331 int s_jquota_fmt;
a2d4a646 5332 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 5333#endif
5334};
5335
2b2d6d01 5336static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 5337{
2b2d6d01 5338 struct ext4_super_block *es;
617ba13b 5339 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5340 unsigned long old_sb_flags;
617ba13b 5341 struct ext4_mount_options old_opts;
c79d967d 5342 int enable_quota = 0;
8a266467 5343 ext4_group_t g;
b3881f74 5344 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 5345 int err = 0;
ac27a0ec 5346#ifdef CONFIG_QUOTA
03dafb5f 5347 int i, j;
33458eab 5348 char *to_free[EXT4_MAXQUOTAS];
ac27a0ec 5349#endif
d4c402d9 5350 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec 5351
21ac738e
CX		 5352 if (data && !orig_data)
5353 return -ENOMEM;
5354
ac27a0ec
DK		 5355 /* Store the original options */
5356 old_sb_flags = sb->s_flags;
5357 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 5358 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 5359 old_opts.s_resuid = sbi->s_resuid;
5360 old_opts.s_resgid = sbi->s_resgid;
5361 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 5362 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5363 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 5364#ifdef CONFIG_QUOTA
5365 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 5366 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5367 if (sbi->s_qf_names[i]) {
33458eab
TT		 5368 char *qf_name = get_qf_name(sb, sbi, i);
5369
5370 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
03dafb5f
CG		 5371 if (!old_opts.s_qf_names[i]) {
5372 for (j = 0; j < i; j++)
5373 kfree(old_opts.s_qf_names[j]);
3e36a163 5374 kfree(orig_data);
03dafb5f
CG		 5375 return -ENOMEM;
5376 }
5377 } else
5378 old_opts.s_qf_names[i] = NULL;
ac27a0ec 5379#endif
b3881f74
TT		 5380 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5381 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 5382
661aa520 5383 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 5384 err = -EINVAL;
5385 goto restore_opts;
5386 }
5387
6b992ff2 5388 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 5389 test_opt(sb, JOURNAL_CHECKSUM)) {
5390 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 5391 "during remount not supported; ignoring");
5392 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 5393 }
5394
6ae6514b
PS		 5395 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5396 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5397 ext4_msg(sb, KERN_ERR, "can't mount with "
5398 "both data=journal and delalloc");
5399 err = -EINVAL;
5400 goto restore_opts;
5401 }
5402 if (test_opt(sb, DIOREAD_NOLOCK)) {
5403 ext4_msg(sb, KERN_ERR, "can't mount with "
5404 "both data=journal and dioread_nolock");
5405 err = -EINVAL;
5406 goto restore_opts;
5407 }
923ae0ff
RZ		 5408 if (test_opt(sb, DAX)) {
5409 ext4_msg(sb, KERN_ERR, "can't mount with "
5410 "both data=journal and dax");
5411 err = -EINVAL;
5412 goto restore_opts;
5413 }
ab04df78
JK		 5414 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5415 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5416 ext4_msg(sb, KERN_ERR, "can't mount with "
5417 "journal_async_commit in data=ordered mode");
5418 err = -EINVAL;
5419 goto restore_opts;
5420 }
923ae0ff
RZ		 5421 }
5422
cdb7ee4c
TE		 5423 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5424 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5425 err = -EINVAL;
5426 goto restore_opts;
5427 }
5428
923ae0ff
RZ		 5429 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5430 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5431 "dax flag with busy inodes while remounting");
5432 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 5433 }
5434
4ab2f15b 5435 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 5436 ext4_abort(sb, "Abort forced by user");
ac27a0ec 5437
1751e8a6
LT		 5438 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5439 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 5440
5441 es = sbi->s_es;
5442
b3881f74 5443 if (sbi->s_journal) {
0390131b 5444 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5445 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5446 }
ac27a0ec 5447
1751e8a6
LT		 5448 if (*flags & SB_LAZYTIME)
5449 sb->s_flags |= SB_LAZYTIME;
a2fd66d0 5450
1751e8a6 5451 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
4ab2f15b 5452 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 5453 err = -EROFS;
5454 goto restore_opts;
5455 }
5456
1751e8a6 5457 if (*flags & SB_RDONLY) {
38c03b34
TT		 5458 err = sync_filesystem(sb);
5459 if (err < 0)
5460 goto restore_opts;
0f0dd62f
CH		 5461 err = dquot_suspend(sb, -1);
5462 if (err < 0)
c79d967d 5463 goto restore_opts;
c79d967d 5464
ac27a0ec
DK		 5465 /*
5466 * First of all, the unconditional stuff we have to do
5467 * to disable replay of the journal when we next remount
5468 */
1751e8a6 5469 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 5470
5471 /*
5472 * OK, test if we are remounting a valid rw partition
5473 * readonly, and if so set the rdonly flag and then
5474 * mark the partition as valid again.
5475 */
617ba13b
MC		 5476 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5477 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5478 es->s_state = cpu_to_le16(sbi->s_mount_state);
5479
a63c9eb2 5480 if (sbi->s_journal)
0390131b 5481 ext4_mark_recovery_complete(sb, es);
2dca60d9
TT		 5482 if (sbi->s_mmp_tsk)
5483 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 5484 } else {
a13fb1a4 5485 /* Make sure we can mount this feature set readwrite */
e2b911c5 5486 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5487 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5488 err = -EROFS;
5489 goto restore_opts;
5490 }
8a266467
TT		 5491 /*
5492 * Make sure the group descriptor checksums
0b8e58a1 5493 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5494 */
5495 for (g = 0; g < sbi->s_groups_count; g++) {
5496 struct ext4_group_desc *gdp =
5497 ext4_get_group_desc(sb, g, NULL);
5498
feb0ab32 5499 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5500 ext4_msg(sb, KERN_ERR,
5501 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5502 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5503 le16_to_cpu(gdp->bg_checksum));
6a797d27 5504 err = -EFSBADCRC;
8a266467
TT		 5505 goto restore_opts;
5506 }
5507 }
5508
ead6596b
ES		 5509 /*
5510 * If we have an unprocessed orphan list hanging
5511 * around from a previously readonly bdev mount,
5512 * require a full umount/remount for now.
5513 */
5514 if (es->s_last_orphan) {
b31e1552 5515 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5516 "remount RDWR because of unprocessed "
5517 "orphan inode list. Please "
b31e1552 5518 "umount/remount instead");
ead6596b
ES		 5519 err = -EINVAL;
5520 goto restore_opts;
5521 }
5522
ac27a0ec
DK		 5523 /*
5524 * Mounting a RDONLY partition read-write, so reread
5525 * and store the current valid flag. (It may have
5526 * been changed by e2fsck since we originally mounted
5527 * the partition.)
5528 */
0390131b
FM		 5529 if (sbi->s_journal)
5530 ext4_clear_journal_err(sb, es);
ac27a0ec 5531 sbi->s_mount_state = le16_to_cpu(es->s_state);
c89128a0
JK		 5532
5533 err = ext4_setup_super(sb, es, 0);
5534 if (err)
5535 goto restore_opts;
5536
5537 sb->s_flags &= ~SB_RDONLY;
e2b911c5 5538 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5539 if (ext4_multi_mount_protect(sb,
5540 le64_to_cpu(es->s_mmp_block))) {
5541 err = -EROFS;
5542 goto restore_opts;
5543 }
c79d967d 5544 enable_quota = 1;
ac27a0ec
DK		 5545 }
5546 }
bfff6873
LC		 5547
5548 /*
5549 * Reinitialize lazy itable initialization thread based on
5550 * current settings
5551 */
bc98a42c 5552 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 5553 ext4_unregister_li_request(sb);
5554 else {
5555 ext4_group_t first_not_zeroed;
5556 first_not_zeroed = ext4_has_uninit_itable(sb);
5557 ext4_register_li_request(sb, first_not_zeroed);
5558 }
5559
6fd058f7 5560 ext4_setup_system_zone(sb);
c89128a0
JK		 5561 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5562 err = ext4_commit_super(sb, 1);
5563 if (err)
5564 goto restore_opts;
5565 }
0390131b 5566
ac27a0ec
DK		 5567#ifdef CONFIG_QUOTA
5568 /* Release old quota file names */
a2d4a646 5569 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5570 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5571 if (enable_quota) {
5572 if (sb_any_quota_suspended(sb))
5573 dquot_resume(sb, -1);
e2b911c5 5574 else if (ext4_has_feature_quota(sb)) {
7c319d32 5575 err = ext4_enable_quotas(sb);
07724f98 5576 if (err)
7c319d32 5577 goto restore_opts;
7c319d32
AK		 5578 }
5579 }
ac27a0ec 5580#endif
d4c402d9 5581
1751e8a6 5582 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
d4c402d9
CW		 5583 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5584 kfree(orig_data);
ac27a0ec 5585 return 0;
0b8e58a1 5586
ac27a0ec
DK		 5587restore_opts:
5588 sb->s_flags = old_sb_flags;
5589 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5590 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5591 sbi->s_resuid = old_opts.s_resuid;
5592 sbi->s_resgid = old_opts.s_resgid;
5593 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5594 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5595 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5596#ifdef CONFIG_QUOTA
5597 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5598 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
33458eab
TT		 5599 to_free[i] = get_qf_name(sb, sbi, i);
5600 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
ac27a0ec 5601 }
33458eab
TT		 5602 synchronize_rcu();
5603 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5604 kfree(to_free[i]);
ac27a0ec 5605#endif
d4c402d9 5606 kfree(orig_data);
ac27a0ec
DK		 5607 return err;
5608}
5609
689c958c
LX		 5610#ifdef CONFIG_QUOTA
5611static int ext4_statfs_project(struct super_block *sb,
5612 kprojid_t projid, struct kstatfs *buf)
5613{
5614 struct kqid qid;
5615 struct dquot *dquot;
5616 u64 limit;
5617 u64 curblock;
5618
5619 qid = make_kqid_projid(projid);
5620 dquot = dqget(sb, qid);
5621 if (IS_ERR(dquot))
5622 return PTR_ERR(dquot);
7b9ca4c6 5623 spin_lock(&dquot->dq_dqb_lock);
689c958c 5624
46d36880 5625 limit = dquot->dq_dqb.dqb_bsoftlimit;
57c32ea4
CX		 5626 if (dquot->dq_dqb.dqb_bhardlimit &&
5627 (!limit || dquot->dq_dqb.dqb_bhardlimit < limit))
5628 limit = dquot->dq_dqb.dqb_bhardlimit;
5629 limit >>= sb->s_blocksize_bits;
5630
689c958c 5631 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 5632 curblock = (dquot->dq_dqb.dqb_curspace +
5633 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 5634 buf->f_blocks = limit;
5635 buf->f_bfree = buf->f_bavail =
5636 (buf->f_blocks > curblock) ?
5637 (buf->f_blocks - curblock) : 0;
5638 }
5639
46d36880 5640 limit = dquot->dq_dqb.dqb_isoftlimit;
57c32ea4
CX		 5641 if (dquot->dq_dqb.dqb_ihardlimit &&
5642 (!limit || dquot->dq_dqb.dqb_ihardlimit < limit))
5643 limit = dquot->dq_dqb.dqb_ihardlimit;
5644
689c958c
LX		 5645 if (limit && buf->f_files > limit) {
5646 buf->f_files = limit;
5647 buf->f_ffree =
5648 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5649 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5650 }
5651
7b9ca4c6 5652 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 5653 dqput(dquot);
5654 return 0;
5655}
5656#endif
5657
2b2d6d01 5658static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5659{
5660 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5661 struct ext4_sb_info *sbi = EXT4_SB(sb);
5662 struct ext4_super_block *es = sbi->s_es;
27dd4385 5663 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5664 u64 fsid;
d02a9391 5665 s64 bfree;
27dd4385 5666 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5667
952fc18e
TT		 5668 if (!test_opt(sb, MINIX_DF))
5669 overhead = sbi->s_overhead;
ac27a0ec 5670
617ba13b 5671 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5672 buf->f_bsize = sb->s_blocksize;
b72f78cb 5673 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5674 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5675 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5676 /* prevent underflow in case that few free space is available */
57042651 5677 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5678 buf->f_bavail = buf->f_bfree -
5679 (ext4_r_blocks_count(es) + resv_blocks);
5680 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5681 buf->f_bavail = 0;
5682 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5683 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5684 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5685 fsid = le64_to_cpup((void *)es->s_uuid) ^
5686 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5687 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5688 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5689
689c958c
LX		 5690#ifdef CONFIG_QUOTA
5691 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5692 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5693 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5694#endif
ac27a0ec
DK		 5695 return 0;
5696}
5697
ac27a0ec
DK		 5698
5699#ifdef CONFIG_QUOTA
5700
bc8230ee
JK		 5701/*
5702 * Helper functions so that transaction is started before we acquire dqio_sem
5703 * to keep correct lock ordering of transaction > dqio_sem
5704 */
ac27a0ec
DK		 5705static inline struct inode *dquot_to_inode(struct dquot *dquot)
5706{
4c376dca 5707 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5708}
5709
617ba13b 5710static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5711{
5712 int ret, err;
5713 handle_t *handle;
5714 struct inode *inode;
5715
5716 inode = dquot_to_inode(dquot);
9924a92a 5717 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5718 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5719 if (IS_ERR(handle))
5720 return PTR_ERR(handle);
5721 ret = dquot_commit(dquot);
617ba13b 5722 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5723 if (!ret)
5724 ret = err;
5725 return ret;
5726}
5727
617ba13b 5728static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5729{
5730 int ret, err;
5731 handle_t *handle;
5732
9924a92a 5733 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5734 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5735 if (IS_ERR(handle))
5736 return PTR_ERR(handle);
5737 ret = dquot_acquire(dquot);
617ba13b 5738 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5739 if (!ret)
5740 ret = err;
5741 return ret;
5742}
5743
617ba13b 5744static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5745{
5746 int ret, err;
5747 handle_t *handle;
5748
9924a92a 5749 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5750 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5751 if (IS_ERR(handle)) {
5752 /* Release dquot anyway to avoid endless cycle in dqput() */
5753 dquot_release(dquot);
ac27a0ec 5754 return PTR_ERR(handle);
9c3013e9 5755 }
ac27a0ec 5756 ret = dquot_release(dquot);
617ba13b 5757 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5758 if (!ret)
5759 ret = err;
5760 return ret;
5761}
5762
617ba13b 5763static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5764{
262b4662
JK		 5765 struct super_block *sb = dquot->dq_sb;
5766 struct ext4_sb_info *sbi = EXT4_SB(sb);
5767
2c8be6b2 5768 /* Are we journaling quotas? */
e2b911c5 5769 if (ext4_has_feature_quota(sb) ||
262b4662 5770 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5771 dquot_mark_dquot_dirty(dquot);
617ba13b 5772 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5773 } else {
5774 return dquot_mark_dquot_dirty(dquot);
5775 }
5776}
5777
617ba13b 5778static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5779{
5780 int ret, err;
5781 handle_t *handle;
5782
5783 /* Data block + inode block */
2b0143b5 5784 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5785 if (IS_ERR(handle))
5786 return PTR_ERR(handle);
5787 ret = dquot_commit_info(sb, type);
617ba13b 5788 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5789 if (!ret)
5790 ret = err;
5791 return ret;
5792}
5793
5794/*
5795 * Turn on quotas during mount time - we need to find
5796 * the quota file and such...
5797 */
617ba13b 5798static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5799{
33458eab 5800 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
287a8095 5801 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5802}
5803
daf647d2
TT		 5804static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5805{
5806 struct ext4_inode_info *ei = EXT4_I(inode);
5807
5808 /* The first argument of lockdep_set_subclass has to be
5809 * *exactly* the same as the argument to init_rwsem() --- in
5810 * this case, in init_once() --- or lockdep gets unhappy
5811 * because the name of the lock is set using the
5812 * stringification of the argument to init_rwsem().
5813 */
5814 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5815 lockdep_set_subclass(&ei->i_data_sem, subclass);
5816}
5817
ac27a0ec
DK		 5818/*
5819 * Standard function to be called on quota_on
5820 */
617ba13b 5821static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5822 const struct path *path)
ac27a0ec
DK		 5823{
5824 int err;
ac27a0ec
DK		 5825
5826 if (!test_opt(sb, QUOTA))
5827 return -EINVAL;
0623543b 5828
ac27a0ec 5829 /* Quotafile not on the same filesystem? */
d8c9584e 5830 if (path->dentry->d_sb != sb)
ac27a0ec 5831 return -EXDEV;
0623543b
JK		 5832 /* Journaling quota? */
5833 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5834 /* Quotafile not in fs root? */
f00c9e44 5835 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5836 ext4_msg(sb, KERN_WARNING,
5837 "Quota file not on filesystem root. "
5838 "Journaled quota will not work");
91389240
JK		 5839 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5840 } else {
5841 /*
5842 * Clear the flag just in case mount options changed since
5843 * last time.
5844 */
5845 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 5846 }
0623543b
JK		 5847
5848 /*
5849 * When we journal data on quota file, we have to flush journal to see
5850 * all updates to the file when we bypass pagecache...
5851 */
0390131b 5852 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5853 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5854 /*
5855 * We don't need to lock updates but journal_flush() could
5856 * otherwise be livelocked...
5857 */
5858 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5859 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5860 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5861 if (err)
7ffe1ea8 5862 return err;
0623543b 5863 }
957153fc 5864
daf647d2
TT		 5865 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5866 err = dquot_quota_on(sb, type, format_id, path);
957153fc 5867 if (err) {
daf647d2
TT		 5868 lockdep_set_quota_inode(path->dentry->d_inode,
5869 I_DATA_SEM_NORMAL);
957153fc
JK		 5870 } else {
5871 struct inode *inode = d_inode(path->dentry);
5872 handle_t *handle;
5873
61a92987
JK		 5874 /*
5875 * Set inode flags to prevent userspace from messing with quota
5876 * files. If this fails, we return success anyway since quotas
5877 * are already enabled and this is not a hard failure.
5878 */
957153fc
JK		 5879 inode_lock(inode);
5880 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5881 if (IS_ERR(handle))
5882 goto unlock_inode;
5883 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5884 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5885 S_NOATIME | S_IMMUTABLE);
5886 ext4_mark_inode_dirty(handle, inode);
5887 ext4_journal_stop(handle);
5888 unlock_inode:
5889 inode_unlock(inode);
5890 }
daf647d2 5891 return err;
ac27a0ec
DK		 5892}
5893
7c319d32
AK		 5894static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5895 unsigned int flags)
5896{
5897 int err;
5898 struct inode *qf_inode;
a2d4a646 5899 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5900 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5901 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5902 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5903 };
5904
e2b911c5 5905 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5906
5907 if (!qf_inums[type])
5908 return -EPERM;
5909
8a363970 5910 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7c319d32
AK		 5911 if (IS_ERR(qf_inode)) {
5912 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5913 return PTR_ERR(qf_inode);
5914 }
5915
bcb13850
JK		 5916 /* Don't account quota for quota files to avoid recursion */
5917 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5918 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7212b95e 5919 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
daf647d2
TT		 5920 if (err)
5921 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
61157b24 5922 iput(qf_inode);
7c319d32
AK		 5923
5924 return err;
5925}
5926
5927/* Enable usage tracking for all quota types. */
5928static int ext4_enable_quotas(struct super_block *sb)
5929{
5930 int type, err = 0;
a2d4a646 5931 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5932 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5933 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5934 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5935 };
49da9392
JK		 5936 bool quota_mopt[EXT4_MAXQUOTAS] = {
5937 test_opt(sb, USRQUOTA),
5938 test_opt(sb, GRPQUOTA),
5939 test_opt(sb, PRJQUOTA),
5940 };
7c319d32 5941
91389240 5942 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 5943 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5944 if (qf_inums[type]) {
5945 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5946 DQUOT_USAGE_ENABLED |
5947 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 5948 if (err) {
5949 ext4_warning(sb,
72ba7450
TT		 5950 "Failed to enable quota tracking "
5951 "(type=%d, err=%d). Please run "
5952 "e2fsck to fix.", type, err);
7f144fd0
JU		 5953 for (type--; type >= 0; type--)
5954 dquot_quota_off(sb, type);
5955
7c319d32
AK		 5956 return err;
5957 }
5958 }
5959 }
5960 return 0;
5961}
5962
ca0e05e4
DM		 5963static int ext4_quota_off(struct super_block *sb, int type)
5964{
21f97697
JK		 5965 struct inode *inode = sb_dqopt(sb)->files[type];
5966 handle_t *handle;
957153fc 5967 int err;
21f97697 5968
87009d86
DM		 5969 /* Force all delayed allocation blocks to be allocated.
5970 * Caller already holds s_umount sem */
5971 if (test_opt(sb, DELALLOC))
ca0e05e4 5972 sync_filesystem(sb);
ca0e05e4 5973
957153fc 5974 if (!inode || !igrab(inode))
0b268590
AG		 5975 goto out;
5976
957153fc 5977 err = dquot_quota_off(sb, type);
964edf66 5978 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 5979 goto out_put;
5980
5981 inode_lock(inode);
61a92987
JK		 5982 /*
5983 * Update modification times of quota files when userspace can
5984 * start looking at them. If we fail, we return success anyway since
5985 * this is not a hard failure and quotas are already disabled.
5986 */
9924a92a 5987 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697 5988 if (IS_ERR(handle))
957153fc
JK		 5989 goto out_unlock;
5990 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5991 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 5992 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5993 ext4_mark_inode_dirty(handle, inode);
5994 ext4_journal_stop(handle);
957153fc
JK		 5995out_unlock:
5996 inode_unlock(inode);
5997out_put:
964edf66 5998 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 5999 iput(inode);
6000 return err;
21f97697 6001out:
ca0e05e4
DM		 6002 return dquot_quota_off(sb, type);
6003}
6004
ac27a0ec
DK		 6005/* Read data from quotafile - avoid pagecache and such because we cannot afford
6006 * acquiring the locks... As quota files are never truncated and quota code
25985edc 6007 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 6008 * we don't have to be afraid of races */
617ba13b 6009static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 6010 size_t len, loff_t off)
6011{
6012 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 6013 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 6014 int offset = off & (sb->s_blocksize - 1);
6015 int tocopy;
6016 size_t toread;
6017 struct buffer_head *bh;
6018 loff_t i_size = i_size_read(inode);
6019
6020 if (off > i_size)
6021 return 0;
6022 if (off+len > i_size)
6023 len = i_size-off;
6024 toread = len;
6025 while (toread > 0) {
6026 tocopy = sb->s_blocksize - offset < toread ?
6027 sb->s_blocksize - offset : toread;
1c215028
TT		 6028 bh = ext4_bread(NULL, inode, blk, 0);
6029 if (IS_ERR(bh))
6030 return PTR_ERR(bh);
ac27a0ec
DK		 6031 if (!bh) /* A hole? */
6032 memset(data, 0, tocopy);
6033 else
6034 memcpy(data, bh->b_data+offset, tocopy);
6035 brelse(bh);
6036 offset = 0;
6037 toread -= tocopy;
6038 data += tocopy;
6039 blk++;
6040 }
6041 return len;
6042}
6043
6044/* Write to quotafile (we know the transaction is already started and has
6045 * enough credits) */
617ba13b 6046static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 6047 const char *data, size_t len, loff_t off)
6048{
6049 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 6050 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 6051 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 6052 int retries = 0;
ac27a0ec
DK		 6053 struct buffer_head *bh;
6054 handle_t *handle = journal_current_handle();
6055
0390131b 6056 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 6057 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6058 " cancelled because transaction is not started",
9c3013e9
JK		 6059 (unsigned long long)off, (unsigned long long)len);
6060 return -EIO;
6061 }
67eeb568
DM		 6062 /*
6063 * Since we account only one data block in transaction credits,
6064 * then it is impossible to cross a block boundary.
6065 */
6066 if (sb->s_blocksize - offset < len) {
6067 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6068 " cancelled because not block aligned",
6069 (unsigned long long)off, (unsigned long long)len);
6070 return -EIO;
6071 }
6072
c5e298ae
TT		 6073 do {
6074 bh = ext4_bread(handle, inode, blk,
6075 EXT4_GET_BLOCKS_CREATE |
6076 EXT4_GET_BLOCKS_METADATA_NOFAIL);
45586c70 6077 } while (PTR_ERR(bh) == -ENOSPC &&
c5e298ae 6078 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 6079 if (IS_ERR(bh))
6080 return PTR_ERR(bh);
67eeb568
DM		 6081 if (!bh)
6082 goto out;
5d601255 6083 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 6084 err = ext4_journal_get_write_access(handle, bh);
6085 if (err) {
6086 brelse(bh);
1c215028 6087 return err;
ac27a0ec 6088 }
67eeb568
DM		 6089 lock_buffer(bh);
6090 memcpy(bh->b_data+offset, data, len);
6091 flush_dcache_page(bh->b_page);
6092 unlock_buffer(bh);
62d2b5f2 6093 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 6094 brelse(bh);
ac27a0ec 6095out:
67eeb568
DM		 6096 if (inode->i_size < off + len) {
6097 i_size_write(inode, off + len);
617ba13b 6098 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 6099 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 6100 }
67eeb568 6101 return len;
ac27a0ec 6102}
ac27a0ec
DK		 6103#endif
6104
152a0836
AV		 6105static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6106 const char *dev_name, void *data)
ac27a0ec 6107{
152a0836 6108 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 6109}
6110
c290ea01 6111#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 6112static inline void register_as_ext2(void)
6113{
6114 int err = register_filesystem(&ext2_fs_type);
6115 if (err)
6116 printk(KERN_WARNING
6117 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6118}
6119
6120static inline void unregister_as_ext2(void)
6121{
6122 unregister_filesystem(&ext2_fs_type);
6123}
2035e776
TT		 6124
6125static inline int ext2_feature_set_ok(struct super_block *sb)
6126{
e2b911c5 6127 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 6128 return 0;
bc98a42c 6129 if (sb_rdonly(sb))
2035e776 6130 return 1;
e2b911c5 6131 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 6132 return 0;
6133 return 1;
6134}
24b58424
TT		 6135#else
6136static inline void register_as_ext2(void) { }
6137static inline void unregister_as_ext2(void) { }
2035e776 6138static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 6139#endif
6140
24b58424
TT		 6141static inline void register_as_ext3(void)
6142{
6143 int err = register_filesystem(&ext3_fs_type);
6144 if (err)
6145 printk(KERN_WARNING
6146 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6147}
6148
6149static inline void unregister_as_ext3(void)
6150{
6151 unregister_filesystem(&ext3_fs_type);
6152}
2035e776
TT		 6153
6154static inline int ext3_feature_set_ok(struct super_block *sb)
6155{
e2b911c5 6156 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 6157 return 0;
e2b911c5 6158 if (!ext4_has_feature_journal(sb))
2035e776 6159 return 0;
bc98a42c 6160 if (sb_rdonly(sb))
2035e776 6161 return 1;
e2b911c5 6162 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 6163 return 0;
6164 return 1;
6165}
24b58424 6166
03010a33
TT		 6167static struct file_system_type ext4_fs_type = {
6168 .owner = THIS_MODULE,
6169 .name = "ext4",
152a0836 6170 .mount = ext4_mount,
03010a33
TT		 6171 .kill_sb = kill_block_super,
6172 .fs_flags = FS_REQUIRES_DEV,
6173};
7f78e035 6174MODULE_ALIAS_FS("ext4");
03010a33 6175
e9e3bcec
ES		 6176/* Shared across all ext4 file systems */
6177wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 6178
5dabfc78 6179static int __init ext4_init_fs(void)
ac27a0ec 6180{
e9e3bcec 6181 int i, err;
c9de560d 6182
e294a537 6183 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 6184 ext4_li_info = NULL;
6185 mutex_init(&ext4_li_mtx);
6186
9a4c8019 6187 /* Build-time check for flags consistency */
12e9b892 6188 ext4_check_flag_values();
e9e3bcec 6189
e142d052 6190 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 6191 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 6192
51865fda 6193 err = ext4_init_es();
6fd058f7
TT		 6194 if (err)
6195 return err;
51865fda 6196
1dc0aa46 6197 err = ext4_init_pending();
22cfe4b4
EB		 6198 if (err)
6199 goto out7;
6200
6201 err = ext4_init_post_read_processing();
1dc0aa46
EW		 6202 if (err)
6203 goto out6;
6204
51865fda
ZL		 6205 err = ext4_init_pageio();
6206 if (err)
b5799018 6207 goto out5;
51865fda 6208
5dabfc78 6209 err = ext4_init_system_zone();
bd2d0210 6210 if (err)
b5799018 6211 goto out4;
857ac889 6212
b5799018 6213 err = ext4_init_sysfs();
dd68314c 6214 if (err)
b5799018 6215 goto out3;
857ac889 6216
5dabfc78 6217 err = ext4_init_mballoc();
c9de560d
AT		 6218 if (err)
6219 goto out2;
ac27a0ec
DK		 6220 err = init_inodecache();
6221 if (err)
6222 goto out1;
24b58424 6223 register_as_ext3();
2035e776 6224 register_as_ext2();
03010a33 6225 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 6226 if (err)
6227 goto out;
bfff6873 6228
ac27a0ec
DK		 6229 return 0;
6230out:
24b58424
TT		 6231 unregister_as_ext2();
6232 unregister_as_ext3();
ac27a0ec
DK		 6233 destroy_inodecache();
6234out1:
5dabfc78 6235 ext4_exit_mballoc();
9c191f70 6236out2:
b5799018
TT		 6237 ext4_exit_sysfs();
6238out3:
5dabfc78 6239 ext4_exit_system_zone();
b5799018 6240out4:
5dabfc78 6241 ext4_exit_pageio();
b5799018 6242out5:
22cfe4b4 6243 ext4_exit_post_read_processing();
1dc0aa46 6244out6:
22cfe4b4
EB		 6245 ext4_exit_pending();
6246out7:
51865fda
ZL		 6247 ext4_exit_es();
6248
ac27a0ec
DK		 6249 return err;
6250}
6251
5dabfc78 6252static void __exit ext4_exit_fs(void)
ac27a0ec 6253{
bfff6873 6254 ext4_destroy_lazyinit_thread();
24b58424
TT		 6255 unregister_as_ext2();
6256 unregister_as_ext3();
03010a33 6257 unregister_filesystem(&ext4_fs_type);
ac27a0ec 6258 destroy_inodecache();
5dabfc78 6259 ext4_exit_mballoc();
b5799018 6260 ext4_exit_sysfs();
5dabfc78
TT		 6261 ext4_exit_system_zone();
6262 ext4_exit_pageio();
22cfe4b4 6263 ext4_exit_post_read_processing();
dd12ed14 6264 ext4_exit_es();
1dc0aa46 6265 ext4_exit_pending();
ac27a0ec
DK		 6266}
6267
6268MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 6269MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 6270MODULE_LICENSE("GPL");
7ef79ad5 6271MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 6272module_init(ext4_init_fs)
6273module_exit(ext4_exit_fs)
Ubuntu Hirsute Linux kernel mirror