]> Git Repo - linux.git/blame - fs/ext4/super.c
projects / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: fix NULL pointer dereference in ext4_mark_inode_dirty()
[linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card ([email protected])
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller ([email protected]), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
66114cad 27#include <linux/backing-dev.h>
ac27a0ec 28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
3197ebdb 37#include <linux/ctype.h>
1330593e 38#include <linux/log2.h>
717d50e4 39#include <linux/crc16.h>
7abc52c2 40#include <linux/cleancache.h>
ac27a0ec
DK		 41#include <asm/uaccess.h>
42
bfff6873
LC		 43#include <linux/kthread.h>
44#include <linux/freezer.h>
45
3dcf5451 46#include "ext4.h"
4a092d73 47#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 48#include "ext4_jbd2.h"
ac27a0ec
DK		 49#include "xattr.h"
50#include "acl.h"
3661d286 51#include "mballoc.h"
ac27a0ec 52
9bffad1e
TT		 53#define CREATE_TRACE_POINTS
54#include <trace/events/ext4.h>
55
0b75a840
LC		 56static struct ext4_lazy_init *ext4_li_info;
57static struct mutex ext4_li_mtx;
e294a537 58static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 59
617ba13b 60static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 61 unsigned long journal_devnum);
2adf6da8 62static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 63static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 64static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
617ba13b 68static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 69static int ext4_remount(struct super_block *sb, int *flags, char *data);
70static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 71static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 72static int ext4_freeze(struct super_block *sb);
152a0836
AV		 73static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74 const char *dev_name, void *data);
2035e776
TT		 75static inline int ext2_feature_set_ok(struct super_block *sb);
76static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 77static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 78static void ext4_destroy_lazyinit_thread(void);
79static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 80static void ext4_clear_request_list(void);
ac27a0ec 81
e74031fd
JK		 82/*
83 * Lock ordering
84 *
85 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
86 * i_mmap_rwsem (inode->i_mmap_rwsem)!
87 *
88 * page fault path:
89 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
90 * page lock -> i_data_sem (rw)
91 *
92 * buffered write path:
93 * sb_start_write -> i_mutex -> mmap_sem
94 * sb_start_write -> i_mutex -> transaction start -> page lock ->
95 * i_data_sem (rw)
96 *
97 * truncate:
98 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
99 * i_mmap_rwsem (w) -> page lock
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * transaction start -> i_data_sem (rw)
102 *
103 * direct IO:
104 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
105 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
106 * transaction start -> i_data_sem (rw)
107 *
108 * writepages:
109 * transaction start -> page lock(s) -> i_data_sem (rw)
110 */
111
c290ea01 112#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 113static struct file_system_type ext2_fs_type = {
114 .owner = THIS_MODULE,
115 .name = "ext2",
116 .mount = ext4_mount,
117 .kill_sb = kill_block_super,
118 .fs_flags = FS_REQUIRES_DEV,
119};
7f78e035 120MODULE_ALIAS_FS("ext2");
fa7614dd 121MODULE_ALIAS("ext2");
2035e776
TT		 122#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
123#else
124#define IS_EXT2_SB(sb) (0)
125#endif
126
127
ba69f9ab
JK		 128static struct file_system_type ext3_fs_type = {
129 .owner = THIS_MODULE,
130 .name = "ext3",
152a0836 131 .mount = ext4_mount,
ba69f9ab
JK		 132 .kill_sb = kill_block_super,
133 .fs_flags = FS_REQUIRES_DEV,
134};
7f78e035 135MODULE_ALIAS_FS("ext3");
fa7614dd 136MODULE_ALIAS("ext3");
ba69f9ab 137#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 138
d25425f8
DW		 139static int ext4_verify_csum_type(struct super_block *sb,
140 struct ext4_super_block *es)
141{
e2b911c5 142 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 143 return 1;
144
145 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
146}
147
a9c47317
DW		 148static __le32 ext4_superblock_csum(struct super_block *sb,
149 struct ext4_super_block *es)
150{
151 struct ext4_sb_info *sbi = EXT4_SB(sb);
152 int offset = offsetof(struct ext4_super_block, s_checksum);
153 __u32 csum;
154
155 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
156
157 return cpu_to_le32(csum);
158}
159
c197855e
SH		 160static int ext4_superblock_csum_verify(struct super_block *sb,
161 struct ext4_super_block *es)
a9c47317 162{
9aa5d32b 163 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 164 return 1;
165
166 return es->s_checksum == ext4_superblock_csum(sb, es);
167}
168
06db49e6 169void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 170{
06db49e6
TT		 171 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
172
9aa5d32b 173 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 174 return;
175
176 es->s_checksum = ext4_superblock_csum(sb, es);
177}
178
9933fc0a
TT		 179void *ext4_kvmalloc(size_t size, gfp_t flags)
180{
181 void *ret;
182
8be04b93 183 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 184 if (!ret)
185 ret = __vmalloc(size, flags, PAGE_KERNEL);
186 return ret;
187}
188
189void *ext4_kvzalloc(size_t size, gfp_t flags)
190{
191 void *ret;
192
8be04b93 193 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 194 if (!ret)
195 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
196 return ret;
197}
198
8fadc143
AR		 199ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
200 struct ext4_group_desc *bg)
bd81d8ee 201{
3a14589c 202 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 203 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 204 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 205}
206
8fadc143
AR		 207ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
208 struct ext4_group_desc *bg)
bd81d8ee 209{
5272f837 210 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 212 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 213}
214
8fadc143
AR		 215ext4_fsblk_t ext4_inode_table(struct super_block *sb,
216 struct ext4_group_desc *bg)
bd81d8ee 217{
5272f837 218 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 220 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 221}
222
021b65bb
TT		 223__u32 ext4_free_group_clusters(struct super_block *sb,
224 struct ext4_group_desc *bg)
560671a0
AK		 225{
226 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 228 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 229}
230
231__u32 ext4_free_inodes_count(struct super_block *sb,
232 struct ext4_group_desc *bg)
233{
234 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 236 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 237}
238
239__u32 ext4_used_dirs_count(struct super_block *sb,
240 struct ext4_group_desc *bg)
241{
242 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 244 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 245}
246
247__u32 ext4_itable_unused_count(struct super_block *sb,
248 struct ext4_group_desc *bg)
249{
250 return le16_to_cpu(bg->bg_itable_unused_lo) |
251 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 252 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 253}
254
8fadc143
AR		 255void ext4_block_bitmap_set(struct super_block *sb,
256 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 257{
3a14589c 258 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 259 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
260 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 261}
262
8fadc143
AR		 263void ext4_inode_bitmap_set(struct super_block *sb,
264 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 265{
5272f837 266 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 269}
270
8fadc143
AR		 271void ext4_inode_table_set(struct super_block *sb,
272 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 273{
5272f837 274 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 277}
278
021b65bb
TT		 279void ext4_free_group_clusters_set(struct super_block *sb,
280 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 281{
282 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
285}
286
287void ext4_free_inodes_set(struct super_block *sb,
288 struct ext4_group_desc *bg, __u32 count)
289{
290 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
293}
294
295void ext4_used_dirs_set(struct super_block *sb,
296 struct ext4_group_desc *bg, __u32 count)
297{
298 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
301}
302
303void ext4_itable_unused_set(struct super_block *sb,
304 struct ext4_group_desc *bg, __u32 count)
305{
306 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
307 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
308 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
309}
310
d3d1faf6 311
1c13d5c0
TT		 312static void __save_error_info(struct super_block *sb, const char *func,
313 unsigned int line)
314{
315 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
316
317 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 318 if (bdev_read_only(sb->s_bdev))
319 return;
1c13d5c0
TT		 320 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
321 es->s_last_error_time = cpu_to_le32(get_seconds());
322 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
323 es->s_last_error_line = cpu_to_le32(line);
324 if (!es->s_first_error_time) {
325 es->s_first_error_time = es->s_last_error_time;
326 strncpy(es->s_first_error_func, func,
327 sizeof(es->s_first_error_func));
328 es->s_first_error_line = cpu_to_le32(line);
329 es->s_first_error_ino = es->s_last_error_ino;
330 es->s_first_error_block = es->s_last_error_block;
331 }
66e61a9e
TT		 332 /*
333 * Start the daily error reporting function if it hasn't been
334 * started already
335 */
336 if (!es->s_error_count)
337 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 338 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 339}
340
341static void save_error_info(struct super_block *sb, const char *func,
342 unsigned int line)
343{
344 __save_error_info(sb, func, line);
345 ext4_commit_super(sb, 1);
346}
347
bdfe0cbd
TT		 348/*
349 * The del_gendisk() function uninitializes the disk-specific data
350 * structures, including the bdi structure, without telling anyone
351 * else. Once this happens, any attempt to call mark_buffer_dirty()
352 * (for example, by ext4_commit_super), will cause a kernel OOPS.
353 * This is a kludge to prevent these oops until we can put in a proper
354 * hook in del_gendisk() to inform the VFS and file system layers.
355 */
356static int block_device_ejected(struct super_block *sb)
357{
358 struct inode *bd_inode = sb->s_bdev->bd_inode;
359 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
360
361 return bdi->dev == NULL;
362}
363
18aadd47
BJ		 364static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
365{
366 struct super_block *sb = journal->j_private;
367 struct ext4_sb_info *sbi = EXT4_SB(sb);
368 int error = is_journal_aborted(journal);
5d3ee208 369 struct ext4_journal_cb_entry *jce;
18aadd47 370
5d3ee208 371 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 372 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 373 while (!list_empty(&txn->t_private_list)) {
374 jce = list_entry(txn->t_private_list.next,
375 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 376 list_del_init(&jce->jce_list);
377 spin_unlock(&sbi->s_md_lock);
378 jce->jce_func(sb, jce, error);
379 spin_lock(&sbi->s_md_lock);
380 }
381 spin_unlock(&sbi->s_md_lock);
382}
1c13d5c0 383
ac27a0ec
DK		 384/* Deal with the reporting of failure conditions on a filesystem such as
385 * inconsistencies detected or read IO failures.
386 *
387 * On ext2, we can store the error state of the filesystem in the
617ba13b 388 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 389 * write ordering constraints on the superblock which prevent us from
390 * writing it out straight away; and given that the journal is about to
391 * be aborted, we can't rely on the current, or future, transactions to
392 * write out the superblock safely.
393 *
dab291af 394 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 395 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 396 * that error until we've noted it down and cleared it.
397 */
398
617ba13b 399static void ext4_handle_error(struct super_block *sb)
ac27a0ec 400{
ac27a0ec
DK		 401 if (sb->s_flags & MS_RDONLY)
402 return;
403
2b2d6d01 404 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 405 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 406
4ab2f15b 407 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 408 if (journal)
dab291af 409 jbd2_journal_abort(journal, -EIO);
ac27a0ec 410 }
2b2d6d01 411 if (test_opt(sb, ERRORS_RO)) {
b31e1552 412 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 413 /*
414 * Make sure updated value of ->s_mount_flags will be visible
415 * before ->s_flags update
416 */
417 smp_wmb();
ac27a0ec
DK		 418 sb->s_flags |= MS_RDONLY;
419 }
4327ba52
DJ		 420 if (test_opt(sb, ERRORS_PANIC)) {
421 if (EXT4_SB(sb)->s_journal &&
422 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
423 return;
617ba13b 424 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 425 sb->s_id);
4327ba52 426 }
ac27a0ec
DK		 427}
428
efbed4dc
TT		 429#define ext4_error_ratelimit(sb) \
430 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
431 "EXT4-fs error")
432
12062ddd 433void __ext4_error(struct super_block *sb, const char *function,
c398eda0 434 unsigned int line, const char *fmt, ...)
ac27a0ec 435{
0ff2ea7d 436 struct va_format vaf;
ac27a0ec
DK		 437 va_list args;
438
efbed4dc
TT		 439 if (ext4_error_ratelimit(sb)) {
440 va_start(args, fmt);
441 vaf.fmt = fmt;
442 vaf.va = &args;
443 printk(KERN_CRIT
444 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
445 sb->s_id, function, line, current->comm, &vaf);
446 va_end(args);
447 }
f3fc0210 448 save_error_info(sb, function, line);
617ba13b 449 ext4_handle_error(sb);
ac27a0ec
DK		 450}
451
e7c96e8e
JP		 452void __ext4_error_inode(struct inode *inode, const char *function,
453 unsigned int line, ext4_fsblk_t block,
454 const char *fmt, ...)
273df556
FM		 455{
456 va_list args;
f7c21177 457 struct va_format vaf;
1c13d5c0 458 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 459
1c13d5c0
TT		 460 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
461 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 462 if (ext4_error_ratelimit(inode->i_sb)) {
463 va_start(args, fmt);
464 vaf.fmt = fmt;
465 vaf.va = &args;
466 if (block)
467 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
468 "inode #%lu: block %llu: comm %s: %pV\n",
469 inode->i_sb->s_id, function, line, inode->i_ino,
470 block, current->comm, &vaf);
471 else
472 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
473 "inode #%lu: comm %s: %pV\n",
474 inode->i_sb->s_id, function, line, inode->i_ino,
475 current->comm, &vaf);
476 va_end(args);
477 }
1c13d5c0 478 save_error_info(inode->i_sb, function, line);
273df556
FM		 479 ext4_handle_error(inode->i_sb);
480}
481
e7c96e8e
JP		 482void __ext4_error_file(struct file *file, const char *function,
483 unsigned int line, ext4_fsblk_t block,
484 const char *fmt, ...)
273df556
FM		 485{
486 va_list args;
f7c21177 487 struct va_format vaf;
1c13d5c0 488 struct ext4_super_block *es;
496ad9aa 489 struct inode *inode = file_inode(file);
273df556
FM		 490 char pathname[80], *path;
491
1c13d5c0
TT		 492 es = EXT4_SB(inode->i_sb)->s_es;
493 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 494 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 495 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 496 if (IS_ERR(path))
497 path = "(unknown)";
498 va_start(args, fmt);
499 vaf.fmt = fmt;
500 vaf.va = &args;
501 if (block)
502 printk(KERN_CRIT
503 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
504 "block %llu: comm %s: path %s: %pV\n",
505 inode->i_sb->s_id, function, line, inode->i_ino,
506 block, current->comm, path, &vaf);
507 else
508 printk(KERN_CRIT
509 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
510 "comm %s: path %s: %pV\n",
511 inode->i_sb->s_id, function, line, inode->i_ino,
512 current->comm, path, &vaf);
513 va_end(args);
514 }
1c13d5c0 515 save_error_info(inode->i_sb, function, line);
273df556
FM		 516 ext4_handle_error(inode->i_sb);
517}
518
722887dd
TT		 519const char *ext4_decode_error(struct super_block *sb, int errno,
520 char nbuf[16])
ac27a0ec
DK		 521{
522 char *errstr = NULL;
523
524 switch (errno) {
6a797d27
DW		 525 case -EFSCORRUPTED:
526 errstr = "Corrupt filesystem";
527 break;
528 case -EFSBADCRC:
529 errstr = "Filesystem failed CRC";
530 break;
ac27a0ec
DK		 531 case -EIO:
532 errstr = "IO failure";
533 break;
534 case -ENOMEM:
535 errstr = "Out of memory";
536 break;
537 case -EROFS:
78f1ddbb
TT		 538 if (!sb || (EXT4_SB(sb)->s_journal &&
539 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 540 errstr = "Journal has aborted";
541 else
542 errstr = "Readonly filesystem";
543 break;
544 default:
545 /* If the caller passed in an extra buffer for unknown
546 * errors, textualise them now. Else we just return
547 * NULL. */
548 if (nbuf) {
549 /* Check for truncated error codes... */
550 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
551 errstr = nbuf;
552 }
553 break;
554 }
555
556 return errstr;
557}
558
617ba13b 559/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 560 * automatically and invokes the appropriate error response. */
561
c398eda0
TT		 562void __ext4_std_error(struct super_block *sb, const char *function,
563 unsigned int line, int errno)
ac27a0ec
DK		 564{
565 char nbuf[16];
566 const char *errstr;
567
568 /* Special case: if the error is EROFS, and we're not already
569 * inside a transaction, then there's really no point in logging
570 * an error. */
571 if (errno == -EROFS && journal_current_handle() == NULL &&
572 (sb->s_flags & MS_RDONLY))
573 return;
574
efbed4dc
TT		 575 if (ext4_error_ratelimit(sb)) {
576 errstr = ext4_decode_error(sb, errno, nbuf);
577 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
578 sb->s_id, function, line, errstr);
579 }
ac27a0ec 580
efbed4dc 581 save_error_info(sb, function, line);
617ba13b 582 ext4_handle_error(sb);
ac27a0ec
DK		 583}
584
585/*
617ba13b 586 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 587 * abort function may be used to deal with unrecoverable failures such
588 * as journal IO errors or ENOMEM at a critical moment in log management.
589 *
590 * We unconditionally force the filesystem into an ABORT|READONLY state,
591 * unless the error response on the fs has been set to panic in which
592 * case we take the easy way out and panic immediately.
593 */
594
c67d859e 595void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 596 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 597{
598 va_list args;
599
1c13d5c0 600 save_error_info(sb, function, line);
ac27a0ec 601 va_start(args, fmt);
c398eda0
TT		 602 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
603 function, line);
ac27a0ec
DK		 604 vprintk(fmt, args);
605 printk("\n");
606 va_end(args);
607
1c13d5c0
TT		 608 if ((sb->s_flags & MS_RDONLY) == 0) {
609 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 610 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 611 /*
612 * Make sure updated value of ->s_mount_flags will be visible
613 * before ->s_flags update
614 */
615 smp_wmb();
616 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 617 if (EXT4_SB(sb)->s_journal)
618 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
619 save_error_info(sb, function, line);
620 }
4327ba52
DJ		 621 if (test_opt(sb, ERRORS_PANIC)) {
622 if (EXT4_SB(sb)->s_journal &&
623 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
624 return;
617ba13b 625 panic("EXT4-fs panic from previous error\n");
4327ba52 626 }
ac27a0ec
DK		 627}
628
e7c96e8e
JP		 629void __ext4_msg(struct super_block *sb,
630 const char *prefix, const char *fmt, ...)
b31e1552 631{
0ff2ea7d 632 struct va_format vaf;
b31e1552
ES		 633 va_list args;
634
efbed4dc
TT		 635 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
636 return;
637
b31e1552 638 va_start(args, fmt);
0ff2ea7d
JP		 639 vaf.fmt = fmt;
640 vaf.va = &args;
641 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 642 va_end(args);
643}
644
b03a2f7e
AD		 645#define ext4_warning_ratelimit(sb) \
646 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
647 "EXT4-fs warning")
648
12062ddd 649void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 650 unsigned int line, const char *fmt, ...)
ac27a0ec 651{
0ff2ea7d 652 struct va_format vaf;
ac27a0ec
DK		 653 va_list args;
654
b03a2f7e 655 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 656 return;
657
ac27a0ec 658 va_start(args, fmt);
0ff2ea7d
JP		 659 vaf.fmt = fmt;
660 vaf.va = &args;
661 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
662 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 663 va_end(args);
664}
665
b03a2f7e
AD		 666void __ext4_warning_inode(const struct inode *inode, const char *function,
667 unsigned int line, const char *fmt, ...)
668{
669 struct va_format vaf;
670 va_list args;
671
672 if (!ext4_warning_ratelimit(inode->i_sb))
673 return;
674
675 va_start(args, fmt);
676 vaf.fmt = fmt;
677 vaf.va = &args;
678 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
679 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
680 function, line, inode->i_ino, current->comm, &vaf);
681 va_end(args);
682}
683
e29136f8
TT		 684void __ext4_grp_locked_error(const char *function, unsigned int line,
685 struct super_block *sb, ext4_group_t grp,
686 unsigned long ino, ext4_fsblk_t block,
687 const char *fmt, ...)
5d1b1b3f
AK		 688__releases(bitlock)
689__acquires(bitlock)
690{
0ff2ea7d 691 struct va_format vaf;
5d1b1b3f
AK		 692 va_list args;
693 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
694
1c13d5c0
TT		 695 es->s_last_error_ino = cpu_to_le32(ino);
696 es->s_last_error_block = cpu_to_le64(block);
697 __save_error_info(sb, function, line);
0ff2ea7d 698
efbed4dc
TT		 699 if (ext4_error_ratelimit(sb)) {
700 va_start(args, fmt);
701 vaf.fmt = fmt;
702 vaf.va = &args;
703 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
704 sb->s_id, function, line, grp);
705 if (ino)
706 printk(KERN_CONT "inode %lu: ", ino);
707 if (block)
708 printk(KERN_CONT "block %llu:",
709 (unsigned long long) block);
710 printk(KERN_CONT "%pV\n", &vaf);
711 va_end(args);
712 }
5d1b1b3f
AK		 713
714 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 715 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 716 return;
717 }
1c13d5c0 718
5d1b1b3f
AK		 719 ext4_unlock_group(sb, grp);
720 ext4_handle_error(sb);
721 /*
722 * We only get here in the ERRORS_RO case; relocking the group
723 * may be dangerous, but nothing bad will happen since the
724 * filesystem will have already been marked read/only and the
725 * journal has been aborted. We return 1 as a hint to callers
726 * who might what to use the return value from
25985edc 727 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 728 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
729 * aggressively from the ext4 function in question, with a
730 * more appropriate error code.
731 */
732 ext4_lock_group(sb, grp);
733 return;
734}
735
617ba13b 736void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 737{
617ba13b 738 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 739
617ba13b 740 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 741 return;
742
12062ddd 743 ext4_warning(sb,
ac27a0ec
DK		 744 "updating to rev %d because of new feature flag, "
745 "running e2fsck is recommended",
617ba13b 746 EXT4_DYNAMIC_REV);
ac27a0ec 747
617ba13b
MC		 748 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
749 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
750 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 751 /* leave es->s_feature_*compat flags alone */
752 /* es->s_uuid will be set by e2fsck if empty */
753
754 /*
755 * The rest of the superblock fields should be zero, and if not it
756 * means they are likely already in use, so leave them alone. We
757 * can leave it up to e2fsck to clean up any inconsistencies there.
758 */
759}
760
761/*
762 * Open the external journal device
763 */
b31e1552 764static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 765{
766 struct block_device *bdev;
767 char b[BDEVNAME_SIZE];
768
d4d77629 769 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 770 if (IS_ERR(bdev))
771 goto fail;
772 return bdev;
773
774fail:
b31e1552 775 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 776 __bdevname(dev, b), PTR_ERR(bdev));
777 return NULL;
778}
779
780/*
781 * Release the journal device
782 */
4385bab1 783static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 784{
4385bab1 785 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 786}
787
4385bab1 788static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 789{
790 struct block_device *bdev;
ac27a0ec
DK		 791 bdev = sbi->journal_bdev;
792 if (bdev) {
4385bab1 793 ext4_blkdev_put(bdev);
ac27a0ec
DK		 794 sbi->journal_bdev = NULL;
795 }
ac27a0ec
DK		 796}
797
798static inline struct inode *orphan_list_entry(struct list_head *l)
799{
617ba13b 800 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 801}
802
617ba13b 803static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 804{
805 struct list_head *l;
806
b31e1552
ES		 807 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
808 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 809
810 printk(KERN_ERR "sb_info orphan list:\n");
811 list_for_each(l, &sbi->s_orphan) {
812 struct inode *inode = orphan_list_entry(l);
813 printk(KERN_ERR " "
814 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
815 inode->i_sb->s_id, inode->i_ino, inode,
816 inode->i_mode, inode->i_nlink,
817 NEXT_ORPHAN(inode));
818 }
819}
820
2b2d6d01 821static void ext4_put_super(struct super_block *sb)
ac27a0ec 822{
617ba13b
MC		 823 struct ext4_sb_info *sbi = EXT4_SB(sb);
824 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 825 int i, err;
ac27a0ec 826
857ac889 827 ext4_unregister_li_request(sb);
e0ccfd95
CH		 828 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
829
2e8fa54e 830 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 831 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 832
0390131b
FM		 833 if (sbi->s_journal) {
834 err = jbd2_journal_destroy(sbi->s_journal);
835 sbi->s_journal = NULL;
836 if (err < 0)
c67d859e 837 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 838 }
d4edac31 839
ebd173be 840 ext4_unregister_sysfs(sb);
d3922a77 841 ext4_es_unregister_shrinker(sbi);
9105bb14 842 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 843 ext4_release_system_zone(sb);
844 ext4_mb_release(sb);
845 ext4_ext_release(sb);
d4edac31 846
ac27a0ec 847 if (!(sb->s_flags & MS_RDONLY)) {
e2b911c5 848 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 849 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 850 }
58c5873a 851 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 852 ext4_commit_super(sb, 1);
853
ac27a0ec
DK		 854 for (i = 0; i < sbi->s_gdb_count; i++)
855 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 856 kvfree(sbi->s_group_desc);
857 kvfree(sbi->s_flex_groups);
57042651 858 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 859 percpu_counter_destroy(&sbi->s_freeinodes_counter);
860 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 861 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 862 brelse(sbi->s_sbh);
863#ifdef CONFIG_QUOTA
a2d4a646 864 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 865 kfree(sbi->s_qf_names[i]);
866#endif
867
868 /* Debugging code just in case the in-memory inode orphan list
869 * isn't empty. The on-disk one can be non-empty if we've
870 * detected an error and taken the fs readonly, but the
871 * in-memory list had better be clean by this point. */
872 if (!list_empty(&sbi->s_orphan))
873 dump_orphan_list(sb, sbi);
874 J_ASSERT(list_empty(&sbi->s_orphan));
875
89d96a6f 876 sync_blockdev(sb->s_bdev);
f98393a6 877 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 878 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
879 /*
880 * Invalidate the journal device's buffers. We don't want them
881 * floating about in memory - the physical journal device may
882 * hotswapped, and it breaks the `ro-after' testing code.
883 */
884 sync_blockdev(sbi->journal_bdev);
f98393a6 885 invalidate_bdev(sbi->journal_bdev);
617ba13b 886 ext4_blkdev_remove(sbi);
ac27a0ec 887 }
9c191f70
M		 888 if (sbi->s_mb_cache) {
889 ext4_xattr_destroy_cache(sbi->s_mb_cache);
890 sbi->s_mb_cache = NULL;
891 }
c5e06d10
JL		 892 if (sbi->s_mmp_tsk)
893 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 894 sb->s_fs_info = NULL;
3197ebdb
TT		 895 /*
896 * Now that we are completely done shutting down the
897 * superblock, we need to actually destroy the kobject.
898 */
3197ebdb
TT		 899 kobject_put(&sbi->s_kobj);
900 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 901 if (sbi->s_chksum_driver)
902 crypto_free_shash(sbi->s_chksum_driver);
705895b6 903 kfree(sbi->s_blockgroup_lock);
ac27a0ec 904 kfree(sbi);
ac27a0ec
DK		 905}
906
e18b890b 907static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 908
909/*
910 * Called inside transaction, so use GFP_NOFS
911 */
617ba13b 912static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 913{
617ba13b 914 struct ext4_inode_info *ei;
ac27a0ec 915
e6b4f8da 916 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 917 if (!ei)
918 return NULL;
0b8e58a1 919
ac27a0ec 920 ei->vfs_inode.i_version = 1;
202ee5df 921 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 922 INIT_LIST_HEAD(&ei->i_prealloc_list);
923 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 924 ext4_es_init_tree(&ei->i_es_tree);
925 rwlock_init(&ei->i_es_lock);
edaa53ca 926 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 927 ei->i_es_all_nr = 0;
edaa53ca 928 ei->i_es_shk_nr = 0;
dd475925 929 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 930 ei->i_reserved_data_blocks = 0;
931 ei->i_reserved_meta_blocks = 0;
932 ei->i_allocated_meta_blocks = 0;
9d0be502 933 ei->i_da_metadata_calc_len = 0;
7e731bc9 934 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 935 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 936#ifdef CONFIG_QUOTA
937 ei->i_reserved_quota = 0;
96c7e0d9 938 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 939#endif
8aefcd55 940 ei->jinode = NULL;
2e8fa54e 941 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 942 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 943 ei->i_sync_tid = 0;
944 ei->i_datasync_tid = 0;
e27f41e1 945 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 946 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
b30ab0e0 947#ifdef CONFIG_EXT4_FS_ENCRYPTION
b7236e21 948 ei->i_crypt_info = NULL;
b30ab0e0 949#endif
ac27a0ec
DK		 950 return &ei->vfs_inode;
951}
952
7ff9c073
TT		 953static int ext4_drop_inode(struct inode *inode)
954{
955 int drop = generic_drop_inode(inode);
956
957 trace_ext4_drop_inode(inode, drop);
958 return drop;
959}
960
fa0d7e3d
NP		 961static void ext4_i_callback(struct rcu_head *head)
962{
963 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 964 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
965}
966
617ba13b 967static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 968{
9f7dd93d 969 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 970 ext4_msg(inode->i_sb, KERN_ERR,
971 "Inode %lu (%p): orphan list check failed!",
972 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 973 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
974 EXT4_I(inode), sizeof(struct ext4_inode_info),
975 true);
976 dump_stack();
977 }
fa0d7e3d 978 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 979}
980
51cc5068 981static void init_once(void *foo)
ac27a0ec 982{
617ba13b 983 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 984
a35afb83 985 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 986 init_rwsem(&ei->xattr_sem);
0e855ac8 987 init_rwsem(&ei->i_data_sem);
ea3d7209 988 init_rwsem(&ei->i_mmap_sem);
a35afb83 989 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 990}
991
e67bc2b3 992static int __init init_inodecache(void)
ac27a0ec 993{
617ba13b
MC		 994 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
995 sizeof(struct ext4_inode_info),
ac27a0ec 996 0, (SLAB_RECLAIM_ACCOUNT|
5d097056 997 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
20c2df83 998 init_once);
617ba13b 999 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1000 return -ENOMEM;
1001 return 0;
1002}
1003
1004static void destroy_inodecache(void)
1005{
8c0a8537
KS		 1006 /*
1007 * Make sure all delayed rcu free inodes are flushed before we
1008 * destroy cache.
1009 */
1010 rcu_barrier();
617ba13b 1011 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1012}
1013
0930fcc1 1014void ext4_clear_inode(struct inode *inode)
ac27a0ec 1015{
0930fcc1 1016 invalidate_inode_buffers(inode);
dbd5768f 1017 clear_inode(inode);
9f754758 1018 dquot_drop(inode);
c2ea3fde 1019 ext4_discard_preallocations(inode);
51865fda 1020 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1021 if (EXT4_I(inode)->jinode) {
1022 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1023 EXT4_I(inode)->jinode);
1024 jbd2_free_inode(EXT4_I(inode)->jinode);
1025 EXT4_I(inode)->jinode = NULL;
1026 }
b7236e21
TT		 1027#ifdef CONFIG_EXT4_FS_ENCRYPTION
1028 if (EXT4_I(inode)->i_crypt_info)
c936e1ec 1029 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
b7236e21 1030#endif
ac27a0ec
DK		 1031}
1032
1b961ac0 1033static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1034 u64 ino, u32 generation)
ac27a0ec 1035{
ac27a0ec 1036 struct inode *inode;
ac27a0ec 1037
617ba13b 1038 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1039 return ERR_PTR(-ESTALE);
617ba13b 1040 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1041 return ERR_PTR(-ESTALE);
1042
1043 /* iget isn't really right if the inode is currently unallocated!!
1044 *
617ba13b 1045 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1046 * deleted, so we should be safe.
1047 *
1048 * Currently we don't know the generation for parent directory, so
1049 * a generation of 0 means "accept any"
1050 */
f4bb2981 1051 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1052 if (IS_ERR(inode))
1053 return ERR_CAST(inode);
1054 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1055 iput(inode);
1056 return ERR_PTR(-ESTALE);
1057 }
1b961ac0
CH		 1058
1059 return inode;
1060}
1061
1062static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1063 int fh_len, int fh_type)
1b961ac0
CH		 1064{
1065 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1066 ext4_nfs_get_inode);
1067}
1068
1069static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1070 int fh_len, int fh_type)
1b961ac0
CH		 1071{
1072 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1073 ext4_nfs_get_inode);
ac27a0ec
DK		 1074}
1075
c39a7f84
TO		 1076/*
1077 * Try to release metadata pages (indirect blocks, directories) which are
1078 * mapped via the block device. Since these pages could have journal heads
1079 * which would prevent try_to_free_buffers() from freeing them, we must use
1080 * jbd2 layer's try_to_free_buffers() function to release them.
1081 */
0b8e58a1
AD		 1082static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1083 gfp_t wait)
c39a7f84
TO		 1084{
1085 journal_t *journal = EXT4_SB(sb)->s_journal;
1086
1087 WARN_ON(PageChecked(page));
1088 if (!page_has_buffers(page))
1089 return 0;
1090 if (journal)
1091 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1092 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1093 return try_to_free_buffers(page);
1094}
1095
ac27a0ec 1096#ifdef CONFIG_QUOTA
689c958c
LX		 1097static char *quotatypes[] = INITQFNAMES;
1098#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1099
617ba13b
MC		 1100static int ext4_write_dquot(struct dquot *dquot);
1101static int ext4_acquire_dquot(struct dquot *dquot);
1102static int ext4_release_dquot(struct dquot *dquot);
1103static int ext4_mark_dquot_dirty(struct dquot *dquot);
1104static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1105static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1106 struct path *path);
ca0e05e4 1107static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1108static int ext4_quota_on_mount(struct super_block *sb, int type);
1109static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1110 size_t len, loff_t off);
617ba13b 1111static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1112 const char *data, size_t len, loff_t off);
7c319d32
AK		 1113static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1114 unsigned int flags);
1115static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1116
96c7e0d9
JK		 1117static struct dquot **ext4_get_dquots(struct inode *inode)
1118{
1119 return EXT4_I(inode)->i_dquot;
1120}
1121
61e225dc 1122static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1123 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1124 .write_dquot = ext4_write_dquot,
1125 .acquire_dquot = ext4_acquire_dquot,
1126 .release_dquot = ext4_release_dquot,
1127 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1128 .write_info = ext4_write_info,
1129 .alloc_dquot = dquot_alloc,
1130 .destroy_dquot = dquot_destroy,
040cb378 1131 .get_projid = ext4_get_projid,
ac27a0ec
DK		 1132};
1133
0d54b217 1134static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1135 .quota_on = ext4_quota_on,
ca0e05e4 1136 .quota_off = ext4_quota_off,
287a8095 1137 .quota_sync = dquot_quota_sync,
0a240339 1138 .get_state = dquot_get_state,
287a8095
CH		 1139 .set_info = dquot_set_dqinfo,
1140 .get_dqblk = dquot_get_dqblk,
1141 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1142};
1143#endif
1144
ee9b6d61 1145static const struct super_operations ext4_sops = {
617ba13b
MC		 1146 .alloc_inode = ext4_alloc_inode,
1147 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1148 .write_inode = ext4_write_inode,
1149 .dirty_inode = ext4_dirty_inode,
7ff9c073 1150 .drop_inode = ext4_drop_inode,
0930fcc1 1151 .evict_inode = ext4_evict_inode,
617ba13b 1152 .put_super = ext4_put_super,
617ba13b 1153 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1154 .freeze_fs = ext4_freeze,
1155 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1156 .statfs = ext4_statfs,
1157 .remount_fs = ext4_remount,
617ba13b 1158 .show_options = ext4_show_options,
ac27a0ec 1159#ifdef CONFIG_QUOTA
617ba13b
MC		 1160 .quota_read = ext4_quota_read,
1161 .quota_write = ext4_quota_write,
96c7e0d9 1162 .get_dquots = ext4_get_dquots,
ac27a0ec 1163#endif
c39a7f84 1164 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1165};
1166
39655164 1167static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1168 .fh_to_dentry = ext4_fh_to_dentry,
1169 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1170 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1171};
1172
1173enum {
1174 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1175 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1176 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1177 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1178 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1179 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1180 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1181 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1182 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1183 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1184 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1185 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
923ae0ff 1186 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1449032b 1187 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1188 Opt_lazytime, Opt_nolazytime,
1449032b 1189 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1190 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1191 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1192 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1193 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1194};
1195
a447c093 1196static const match_table_t tokens = {
ac27a0ec
DK		 1197 {Opt_bsd_df, "bsddf"},
1198 {Opt_minix_df, "minixdf"},
1199 {Opt_grpid, "grpid"},
1200 {Opt_grpid, "bsdgroups"},
1201 {Opt_nogrpid, "nogrpid"},
1202 {Opt_nogrpid, "sysvgroups"},
1203 {Opt_resgid, "resgid=%u"},
1204 {Opt_resuid, "resuid=%u"},
1205 {Opt_sb, "sb=%u"},
1206 {Opt_err_cont, "errors=continue"},
1207 {Opt_err_panic, "errors=panic"},
1208 {Opt_err_ro, "errors=remount-ro"},
1209 {Opt_nouid32, "nouid32"},
ac27a0ec 1210 {Opt_debug, "debug"},
72578c33
TT		 1211 {Opt_removed, "oldalloc"},
1212 {Opt_removed, "orlov"},
ac27a0ec
DK		 1213 {Opt_user_xattr, "user_xattr"},
1214 {Opt_nouser_xattr, "nouser_xattr"},
1215 {Opt_acl, "acl"},
1216 {Opt_noacl, "noacl"},
e3bb52ae 1217 {Opt_noload, "norecovery"},
5a916be1 1218 {Opt_noload, "noload"},
72578c33
TT		 1219 {Opt_removed, "nobh"},
1220 {Opt_removed, "bh"},
ac27a0ec 1221 {Opt_commit, "commit=%u"},
30773840
TT		 1222 {Opt_min_batch_time, "min_batch_time=%u"},
1223 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1224 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1225 {Opt_journal_path, "journal_path=%s"},
818d276c 1226 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1227 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1228 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1229 {Opt_abort, "abort"},
1230 {Opt_data_journal, "data=journal"},
1231 {Opt_data_ordered, "data=ordered"},
1232 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1233 {Opt_data_err_abort, "data_err=abort"},
1234 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1235 {Opt_offusrjquota, "usrjquota="},
1236 {Opt_usrjquota, "usrjquota=%s"},
1237 {Opt_offgrpjquota, "grpjquota="},
1238 {Opt_grpjquota, "grpjquota=%s"},
1239 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1240 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1241 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1242 {Opt_grpquota, "grpquota"},
1243 {Opt_noquota, "noquota"},
1244 {Opt_quota, "quota"},
1245 {Opt_usrquota, "usrquota"},
1246 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1247 {Opt_barrier, "barrier"},
1248 {Opt_nobarrier, "nobarrier"},
25ec56b5 1249 {Opt_i_version, "i_version"},
923ae0ff 1250 {Opt_dax, "dax"},
c9de560d 1251 {Opt_stripe, "stripe=%u"},
64769240 1252 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1253 {Opt_lazytime, "lazytime"},
1254 {Opt_nolazytime, "nolazytime"},
dd919b98 1255 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1256 {Opt_removed, "mblk_io_submit"},
1257 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1258 {Opt_block_validity, "block_validity"},
1259 {Opt_noblock_validity, "noblock_validity"},
240799cd 1260 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1261 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1262 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1263 {Opt_auto_da_alloc, "auto_da_alloc"},
1264 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1265 {Opt_dioread_nolock, "dioread_nolock"},
1266 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1267 {Opt_discard, "discard"},
1268 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1269 {Opt_init_itable, "init_itable=%u"},
1270 {Opt_init_itable, "init_itable"},
1271 {Opt_noinit_itable, "noinit_itable"},
df981d03 1272 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1273 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1274 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1275 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1276 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1277 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1278 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1279 {Opt_err, NULL},
ac27a0ec
DK		 1280};
1281
617ba13b 1282static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1283{
617ba13b 1284 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1285 char *options = (char *) *data;
1286
1287 if (!options || strncmp(options, "sb=", 3) != 0)
1288 return 1; /* Default location */
0b8e58a1 1289
ac27a0ec 1290 options += 3;
0b8e58a1 1291 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1292 sb_block = simple_strtoul(options, &options, 0);
1293 if (*options && *options != ',') {
4776004f 1294 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1295 (char *) *data);
1296 return 1;
1297 }
1298 if (*options == ',')
1299 options++;
1300 *data = (void *) options;
0b8e58a1 1301
ac27a0ec
DK		 1302 return sb_block;
1303}
1304
b3881f74 1305#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1306static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1307 "Contact [email protected] if you think we should keep it.\n";
b3881f74 1308
56c50f11
DM		 1309#ifdef CONFIG_QUOTA
1310static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1311{
1312 struct ext4_sb_info *sbi = EXT4_SB(sb);
1313 char *qname;
03dafb5f 1314 int ret = -1;
56c50f11
DM		 1315
1316 if (sb_any_quota_loaded(sb) &&
1317 !sbi->s_qf_names[qtype]) {
1318 ext4_msg(sb, KERN_ERR,
1319 "Cannot change journaled "
1320 "quota options when quota turned on");
57f73c2c 1321 return -1;
56c50f11 1322 }
e2b911c5 1323 if (ext4_has_feature_quota(sb)) {
262b4662
JK		 1324 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1325 "when QUOTA feature is enabled");
1326 return -1;
1327 }
56c50f11
DM		 1328 qname = match_strdup(args);
1329 if (!qname) {
1330 ext4_msg(sb, KERN_ERR,
1331 "Not enough memory for storing quotafile name");
57f73c2c 1332 return -1;
56c50f11 1333 }
03dafb5f
CG		 1334 if (sbi->s_qf_names[qtype]) {
1335 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1336 ret = 1;
1337 else
1338 ext4_msg(sb, KERN_ERR,
1339 "%s quota file already specified",
1340 QTYPE2NAME(qtype));
1341 goto errout;
56c50f11 1342 }
03dafb5f 1343 if (strchr(qname, '/')) {
56c50f11
DM		 1344 ext4_msg(sb, KERN_ERR,
1345 "quotafile must be on filesystem root");
03dafb5f 1346 goto errout;
56c50f11 1347 }
03dafb5f 1348 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1349 set_opt(sb, QUOTA);
56c50f11 1350 return 1;
03dafb5f
CG		 1351errout:
1352 kfree(qname);
1353 return ret;
56c50f11
DM		 1354}
1355
1356static int clear_qf_name(struct super_block *sb, int qtype)
1357{
1358
1359 struct ext4_sb_info *sbi = EXT4_SB(sb);
1360
1361 if (sb_any_quota_loaded(sb) &&
1362 sbi->s_qf_names[qtype]) {
1363 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1364 " when quota turned on");
57f73c2c 1365 return -1;
56c50f11 1366 }
03dafb5f 1367 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1368 sbi->s_qf_names[qtype] = NULL;
1369 return 1;
1370}
1371#endif
1372
26092bf5
TT		 1373#define MOPT_SET 0x0001
1374#define MOPT_CLEAR 0x0002
1375#define MOPT_NOSUPPORT 0x0004
1376#define MOPT_EXPLICIT 0x0008
1377#define MOPT_CLEAR_ERR 0x0010
1378#define MOPT_GTE0 0x0020
ac27a0ec 1379#ifdef CONFIG_QUOTA
26092bf5
TT		 1380#define MOPT_Q 0
1381#define MOPT_QFMT 0x0040
1382#else
1383#define MOPT_Q MOPT_NOSUPPORT
1384#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1385#endif
26092bf5 1386#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1387#define MOPT_NO_EXT2 0x0100
1388#define MOPT_NO_EXT3 0x0200
1389#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1390#define MOPT_STRING 0x0400
26092bf5
TT		 1391
1392static const struct mount_opts {
1393 int token;
1394 int mount_opt;
1395 int flags;
1396} ext4_mount_opts[] = {
1397 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1398 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1399 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1400 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1401 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1402 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1403 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1404 MOPT_EXT4_ONLY | MOPT_SET},
1405 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1406 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1407 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1408 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1409 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1410 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1411 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1412 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1413 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1414 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1415 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1416 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1417 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1418 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1419 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1420 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1421 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1422 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1423 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1424 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1425 MOPT_NO_EXT2 | MOPT_SET},
1426 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1427 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1428 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1429 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1430 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1431 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1432 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1433 {Opt_commit, 0, MOPT_GTE0},
1434 {Opt_max_batch_time, 0, MOPT_GTE0},
1435 {Opt_min_batch_time, 0, MOPT_GTE0},
1436 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1437 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1438 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1439 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1440 {Opt_resuid, 0, MOPT_GTE0},
1441 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1442 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1443 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1444 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1445 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1446 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1447 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1448 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1449 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1450 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1451#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1452 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1453 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1454#else
26092bf5
TT		 1455 {Opt_acl, 0, MOPT_NOSUPPORT},
1456 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1457#endif
26092bf5
TT		 1458 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1459 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1460 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1461 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1462 MOPT_SET | MOPT_Q},
1463 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1464 MOPT_SET | MOPT_Q},
1465 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1466 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1467 {Opt_usrjquota, 0, MOPT_Q},
1468 {Opt_grpjquota, 0, MOPT_Q},
1469 {Opt_offusrjquota, 0, MOPT_Q},
1470 {Opt_offgrpjquota, 0, MOPT_Q},
1471 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1472 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1473 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1474 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1475 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1476 {Opt_err, 0, 0}
1477};
1478
1479static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1480 substring_t *args, unsigned long *journal_devnum,
1481 unsigned int *journal_ioprio, int is_remount)
1482{
1483 struct ext4_sb_info *sbi = EXT4_SB(sb);
1484 const struct mount_opts *m;
08cefc7a
EB		 1485 kuid_t uid;
1486 kgid_t gid;
26092bf5
TT		 1487 int arg = 0;
1488
57f73c2c
TT		 1489#ifdef CONFIG_QUOTA
1490 if (token == Opt_usrjquota)
1491 return set_qf_name(sb, USRQUOTA, &args[0]);
1492 else if (token == Opt_grpjquota)
1493 return set_qf_name(sb, GRPQUOTA, &args[0]);
1494 else if (token == Opt_offusrjquota)
1495 return clear_qf_name(sb, USRQUOTA);
1496 else if (token == Opt_offgrpjquota)
1497 return clear_qf_name(sb, GRPQUOTA);
1498#endif
26092bf5 1499 switch (token) {
f7048605
TT		 1500 case Opt_noacl:
1501 case Opt_nouser_xattr:
1502 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1503 break;
26092bf5
TT		 1504 case Opt_sb:
1505 return 1; /* handled by get_sb_block() */
1506 case Opt_removed:
5f3633e3 1507 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1508 return 1;
26092bf5
TT		 1509 case Opt_abort:
1510 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1511 return 1;
1512 case Opt_i_version:
1513 sb->s_flags |= MS_I_VERSION;
1514 return 1;
a26f4992
TT		 1515 case Opt_lazytime:
1516 sb->s_flags |= MS_LAZYTIME;
1517 return 1;
1518 case Opt_nolazytime:
1519 sb->s_flags &= ~MS_LAZYTIME;
1520 return 1;
26092bf5
TT		 1521 }
1522
5f3633e3
JK		 1523 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1524 if (token == m->token)
1525 break;
1526
1527 if (m->token == Opt_err) {
1528 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1529 "or missing value", opt);
1530 return -1;
1531 }
1532
8dc0aa8c
TT		 1533 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1534 ext4_msg(sb, KERN_ERR,
1535 "Mount option \"%s\" incompatible with ext2", opt);
1536 return -1;
1537 }
1538 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1539 ext4_msg(sb, KERN_ERR,
1540 "Mount option \"%s\" incompatible with ext3", opt);
1541 return -1;
1542 }
1543
ad4eec61 1544 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1545 return -1;
1546 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1547 return -1;
c93cf2d7
DM		 1548 if (m->flags & MOPT_EXPLICIT) {
1549 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1550 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1551 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1552 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1553 } else
1554 return -1;
1555 }
5f3633e3
JK		 1556 if (m->flags & MOPT_CLEAR_ERR)
1557 clear_opt(sb, ERRORS_MASK);
1558 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1559 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1560 "options when quota turned on");
1561 return -1;
1562 }
1563
1564 if (m->flags & MOPT_NOSUPPORT) {
1565 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1566 } else if (token == Opt_commit) {
1567 if (arg == 0)
1568 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1569 sbi->s_commit_interval = HZ * arg;
1570 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1571 sbi->s_max_batch_time = arg;
1572 } else if (token == Opt_min_batch_time) {
1573 sbi->s_min_batch_time = arg;
1574 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1575 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1576 ext4_msg(sb, KERN_ERR,
1577 "EXT4-fs: inode_readahead_blks must be "
1578 "0 or a power of 2 smaller than 2^31");
26092bf5 1579 return -1;
5f3633e3
JK		 1580 }
1581 sbi->s_inode_readahead_blks = arg;
1582 } else if (token == Opt_init_itable) {
1583 set_opt(sb, INIT_INODE_TABLE);
1584 if (!args->from)
1585 arg = EXT4_DEF_LI_WAIT_MULT;
1586 sbi->s_li_wait_mult = arg;
1587 } else if (token == Opt_max_dir_size_kb) {
1588 sbi->s_max_dir_size_kb = arg;
1589 } else if (token == Opt_stripe) {
1590 sbi->s_stripe = arg;
1591 } else if (token == Opt_resuid) {
1592 uid = make_kuid(current_user_ns(), arg);
1593 if (!uid_valid(uid)) {
1594 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1595 return -1;
1596 }
5f3633e3
JK		 1597 sbi->s_resuid = uid;
1598 } else if (token == Opt_resgid) {
1599 gid = make_kgid(current_user_ns(), arg);
1600 if (!gid_valid(gid)) {
1601 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1602 return -1;
1603 }
1604 sbi->s_resgid = gid;
1605 } else if (token == Opt_journal_dev) {
1606 if (is_remount) {
1607 ext4_msg(sb, KERN_ERR,
1608 "Cannot specify journal on remount");
1609 return -1;
1610 }
1611 *journal_devnum = arg;
ad4eec61
ES		 1612 } else if (token == Opt_journal_path) {
1613 char *journal_path;
1614 struct inode *journal_inode;
1615 struct path path;
1616 int error;
1617
1618 if (is_remount) {
1619 ext4_msg(sb, KERN_ERR,
1620 "Cannot specify journal on remount");
1621 return -1;
1622 }
1623 journal_path = match_strdup(&args[0]);
1624 if (!journal_path) {
1625 ext4_msg(sb, KERN_ERR, "error: could not dup "
1626 "journal device string");
1627 return -1;
1628 }
1629
1630 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1631 if (error) {
1632 ext4_msg(sb, KERN_ERR, "error: could not find "
1633 "journal device path: error %d", error);
1634 kfree(journal_path);
1635 return -1;
1636 }
1637
2b0143b5 1638 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1639 if (!S_ISBLK(journal_inode->i_mode)) {
1640 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1641 "is not a block device", journal_path);
1642 path_put(&path);
1643 kfree(journal_path);
1644 return -1;
1645 }
1646
1647 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1648 path_put(&path);
1649 kfree(journal_path);
5f3633e3
JK		 1650 } else if (token == Opt_journal_ioprio) {
1651 if (arg > 7) {
1652 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1653 " (must be 0-7)");
1654 return -1;
1655 }
1656 *journal_ioprio =
1657 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1658 } else if (token == Opt_test_dummy_encryption) {
1659#ifdef CONFIG_EXT4_FS_ENCRYPTION
1660 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1661 ext4_msg(sb, KERN_WARNING,
1662 "Test dummy encryption mode enabled");
1663#else
1664 ext4_msg(sb, KERN_WARNING,
1665 "Test dummy encryption mount option ignored");
1666#endif
5f3633e3
JK		 1667 } else if (m->flags & MOPT_DATAJ) {
1668 if (is_remount) {
1669 if (!sbi->s_journal)
1670 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1671 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1672 ext4_msg(sb, KERN_ERR,
26092bf5 1673 "Cannot change data mode on remount");
26092bf5 1674 return -1;
ac27a0ec 1675 }
26092bf5 1676 } else {
5f3633e3
JK		 1677 clear_opt(sb, DATA_FLAGS);
1678 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1679 }
5f3633e3
JK		 1680#ifdef CONFIG_QUOTA
1681 } else if (m->flags & MOPT_QFMT) {
1682 if (sb_any_quota_loaded(sb) &&
1683 sbi->s_jquota_fmt != m->mount_opt) {
1684 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1685 "quota options when quota turned on");
1686 return -1;
1687 }
e2b911c5 1688 if (ext4_has_feature_quota(sb)) {
262b4662
JK		 1689 ext4_msg(sb, KERN_ERR,
1690 "Cannot set journaled quota options "
1691 "when QUOTA feature is enabled");
1692 return -1;
1693 }
5f3633e3 1694 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1695#endif
923ae0ff 1696 } else if (token == Opt_dax) {
ef83b6e8
DW		 1697#ifdef CONFIG_FS_DAX
1698 ext4_msg(sb, KERN_WARNING,
1699 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1700 sbi->s_mount_opt |= m->mount_opt;
1701#else
923ae0ff
RZ		 1702 ext4_msg(sb, KERN_INFO, "dax option not supported");
1703 return -1;
5f3633e3
JK		 1704#endif
1705 } else {
1706 if (!args->from)
1707 arg = 1;
1708 if (m->flags & MOPT_CLEAR)
1709 arg = !arg;
1710 else if (unlikely(!(m->flags & MOPT_SET))) {
1711 ext4_msg(sb, KERN_WARNING,
1712 "buggy handling of option %s", opt);
1713 WARN_ON(1);
1714 return -1;
1715 }
1716 if (arg != 0)
1717 sbi->s_mount_opt |= m->mount_opt;
1718 else
1719 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1720 }
5f3633e3 1721 return 1;
26092bf5
TT		 1722}
1723
1724static int parse_options(char *options, struct super_block *sb,
1725 unsigned long *journal_devnum,
1726 unsigned int *journal_ioprio,
1727 int is_remount)
1728{
1729 struct ext4_sb_info *sbi = EXT4_SB(sb);
1730 char *p;
1731 substring_t args[MAX_OPT_ARGS];
1732 int token;
1733
1734 if (!options)
1735 return 1;
1736
1737 while ((p = strsep(&options, ",")) != NULL) {
1738 if (!*p)
1739 continue;
1740 /*
1741 * Initialize args struct so we know whether arg was
1742 * found; some options take optional arguments.
1743 */
caecd0af 1744 args[0].to = args[0].from = NULL;
26092bf5
TT		 1745 token = match_token(p, tokens, args);
1746 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1747 journal_ioprio, is_remount) < 0)
1748 return 0;
ac27a0ec
DK		 1749 }
1750#ifdef CONFIG_QUOTA
e2b911c5 1751 if (ext4_has_feature_quota(sb) &&
262b4662
JK		 1752 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1753 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1754 "feature is enabled");
1755 return 0;
1756 }
ac27a0ec 1757 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1758 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1759 clear_opt(sb, USRQUOTA);
ac27a0ec 1760
482a7425 1761 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1762 clear_opt(sb, GRPQUOTA);
ac27a0ec 1763
56c50f11 1764 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1765 ext4_msg(sb, KERN_ERR, "old and new quota "
1766 "format mixing");
ac27a0ec
DK		 1767 return 0;
1768 }
1769
1770 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1771 ext4_msg(sb, KERN_ERR, "journaled quota format "
1772 "not specified");
ac27a0ec
DK		 1773 return 0;
1774 }
ac27a0ec
DK		 1775 }
1776#endif
261cb20c
JK		 1777 if (test_opt(sb, DIOREAD_NOLOCK)) {
1778 int blocksize =
1779 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1780
1781 if (blocksize < PAGE_CACHE_SIZE) {
1782 ext4_msg(sb, KERN_ERR, "can't mount with "
1783 "dioread_nolock if block size != PAGE_SIZE");
1784 return 0;
1785 }
1786 }
d4f76107
JK		 1787 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1788 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1789 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1790 "in data=ordered mode");
1791 return 0;
1792 }
ac27a0ec
DK		 1793 return 1;
1794}
1795
2adf6da8
TT		 1796static inline void ext4_show_quota_options(struct seq_file *seq,
1797 struct super_block *sb)
1798{
1799#if defined(CONFIG_QUOTA)
1800 struct ext4_sb_info *sbi = EXT4_SB(sb);
1801
1802 if (sbi->s_jquota_fmt) {
1803 char *fmtname = "";
1804
1805 switch (sbi->s_jquota_fmt) {
1806 case QFMT_VFS_OLD:
1807 fmtname = "vfsold";
1808 break;
1809 case QFMT_VFS_V0:
1810 fmtname = "vfsv0";
1811 break;
1812 case QFMT_VFS_V1:
1813 fmtname = "vfsv1";
1814 break;
1815 }
1816 seq_printf(seq, ",jqfmt=%s", fmtname);
1817 }
1818
1819 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1820 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1821
1822 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1823 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1824#endif
1825}
1826
5a916be1
TT		 1827static const char *token2str(int token)
1828{
50df9fd5 1829 const struct match_token *t;
5a916be1
TT		 1830
1831 for (t = tokens; t->token != Opt_err; t++)
1832 if (t->token == token && !strchr(t->pattern, '='))
1833 break;
1834 return t->pattern;
1835}
1836
2adf6da8
TT		 1837/*
1838 * Show an option if
1839 * - it's set to a non-default value OR
1840 * - if the per-sb default is different from the global default
1841 */
66acdcf4
TT		 1842static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1843 int nodefs)
2adf6da8 1844{
2adf6da8
TT		 1845 struct ext4_sb_info *sbi = EXT4_SB(sb);
1846 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1847 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1848 const struct mount_opts *m;
66acdcf4 1849 char sep = nodefs ? '\n' : ',';
2adf6da8 1850
66acdcf4
TT		 1851#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1852#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1853
1854 if (sbi->s_sb_block != 1)
5a916be1
TT		 1855 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1856
1857 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1858 int want_set = m->flags & MOPT_SET;
1859 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1860 (m->flags & MOPT_CLEAR_ERR))
1861 continue;
66acdcf4 1862 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1863 continue; /* skip if same as the default */
1864 if ((want_set &&
1865 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1866 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1867 continue; /* select Opt_noFoo vs Opt_Foo */
1868 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1869 }
5a916be1 1870
08cefc7a 1871 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1872 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1873 SEQ_OPTS_PRINT("resuid=%u",
1874 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1875 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1876 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1877 SEQ_OPTS_PRINT("resgid=%u",
1878 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1879 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1880 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1881 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1882 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1883 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1884 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1885 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1886 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1887 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1888 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1889 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1890 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1891 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1892 if (sb->s_flags & MS_I_VERSION)
5a916be1 1893 SEQ_OPTS_PUTS("i_version");
66acdcf4 1894 if (nodefs || sbi->s_stripe)
5a916be1 1895 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1896 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1897 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1898 SEQ_OPTS_PUTS("data=journal");
1899 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1900 SEQ_OPTS_PUTS("data=ordered");
1901 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1902 SEQ_OPTS_PUTS("data=writeback");
1903 }
66acdcf4
TT		 1904 if (nodefs ||
1905 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1906 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1907 sbi->s_inode_readahead_blks);
2adf6da8 1908
66acdcf4
TT		 1909 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1910 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1911 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1912 if (nodefs || sbi->s_max_dir_size_kb)
1913 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1914
1915 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1916 return 0;
1917}
1918
66acdcf4
TT		 1919static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1920{
1921 return _ext4_show_options(seq, root->d_sb, 0);
1922}
1923
ebd173be 1924int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 1925{
1926 struct super_block *sb = seq->private;
1927 int rc;
1928
1929 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1930 rc = _ext4_show_options(seq, sb, 1);
1931 seq_puts(seq, "\n");
1932 return rc;
1933}
1934
617ba13b 1935static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1936 int read_only)
1937{
617ba13b 1938 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1939 int res = 0;
1940
617ba13b 1941 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1942 ext4_msg(sb, KERN_ERR, "revision level too high, "
1943 "forcing read-only mode");
ac27a0ec
DK		 1944 res = MS_RDONLY;
1945 }
1946 if (read_only)
281b5995 1947 goto done;
617ba13b 1948 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1949 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1950 "running e2fsck is recommended");
c8b459f4 1951 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1952 ext4_msg(sb, KERN_WARNING,
1953 "warning: mounting fs with errors, "
1954 "running e2fsck is recommended");
ed3ce80a 1955 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1956 le16_to_cpu(es->s_mnt_count) >=
1957 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1958 ext4_msg(sb, KERN_WARNING,
1959 "warning: maximal mount count reached, "
1960 "running e2fsck is recommended");
ac27a0ec
DK		 1961 else if (le32_to_cpu(es->s_checkinterval) &&
1962 (le32_to_cpu(es->s_lastcheck) +
1963 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1964 ext4_msg(sb, KERN_WARNING,
1965 "warning: checktime reached, "
1966 "running e2fsck is recommended");
0b8e58a1 1967 if (!sbi->s_journal)
0390131b 1968 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1969 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1970 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1971 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1972 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1973 ext4_update_dynamic_rev(sb);
0390131b 1974 if (sbi->s_journal)
e2b911c5 1975 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 1976
e2d67052 1977 ext4_commit_super(sb, 1);
281b5995 1978done:
ac27a0ec 1979 if (test_opt(sb, DEBUG))
a9df9a49 1980 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1981 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1982 sb->s_blocksize,
1983 sbi->s_groups_count,
617ba13b
MC		 1984 EXT4_BLOCKS_PER_GROUP(sb),
1985 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1986 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1987
7abc52c2 1988 cleancache_init_fs(sb);
ac27a0ec
DK		 1989 return res;
1990}
1991
117fff10
TT		 1992int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1993{
1994 struct ext4_sb_info *sbi = EXT4_SB(sb);
1995 struct flex_groups *new_groups;
1996 int size;
1997
1998 if (!sbi->s_log_groups_per_flex)
1999 return 0;
2000
2001 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2002 if (size <= sbi->s_flex_groups_allocated)
2003 return 0;
2004
2005 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2006 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2007 if (!new_groups) {
2008 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2009 size / (int) sizeof(struct flex_groups));
2010 return -ENOMEM;
2011 }
2012
2013 if (sbi->s_flex_groups) {
2014 memcpy(new_groups, sbi->s_flex_groups,
2015 (sbi->s_flex_groups_allocated *
2016 sizeof(struct flex_groups)));
b93b41d4 2017 kvfree(sbi->s_flex_groups);
117fff10
TT		 2018 }
2019 sbi->s_flex_groups = new_groups;
2020 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2021 return 0;
2022}
2023
772cb7c8
JS		 2024static int ext4_fill_flex_info(struct super_block *sb)
2025{
2026 struct ext4_sb_info *sbi = EXT4_SB(sb);
2027 struct ext4_group_desc *gdp = NULL;
772cb7c8 2028 ext4_group_t flex_group;
117fff10 2029 int i, err;
772cb7c8 2030
503358ae 2031 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2032 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2033 sbi->s_log_groups_per_flex = 0;
2034 return 1;
2035 }
2036
117fff10
TT		 2037 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2038 if (err)
9933fc0a 2039 goto failed;
772cb7c8 2040
772cb7c8 2041 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2042 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2043
2044 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2045 atomic_add(ext4_free_inodes_count(sb, gdp),
2046 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2047 atomic64_add(ext4_free_group_clusters(sb, gdp),
2048 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2049 atomic_add(ext4_used_dirs_count(sb, gdp),
2050 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2051 }
2052
2053 return 1;
2054failed:
2055 return 0;
2056}
2057
e2b911c5 2058static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2059 struct ext4_group_desc *gdp)
717d50e4 2060{
feb0ab32 2061 int offset;
717d50e4 2062 __u16 crc = 0;
feb0ab32 2063 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2064 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2065
9aa5d32b 2066 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2067 /* Use new metadata_csum algorithm */
d6a77105 2068 __le16 save_csum;
feb0ab32
DW		 2069 __u32 csum32;
2070
d6a77105 2071 save_csum = gdp->bg_checksum;
feb0ab32
DW		 2072 gdp->bg_checksum = 0;
2073 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2074 sizeof(le_group));
2075 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2076 sbi->s_desc_size);
d6a77105 2077 gdp->bg_checksum = save_csum;
feb0ab32
DW		 2078
2079 crc = csum32 & 0xFFFF;
2080 goto out;
717d50e4
AD		 2081 }
2082
feb0ab32 2083 /* old crc16 code */
e2b911c5 2084 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2085 return 0;
2086
feb0ab32
DW		 2087 offset = offsetof(struct ext4_group_desc, bg_checksum);
2088
2089 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2090 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2091 crc = crc16(crc, (__u8 *)gdp, offset);
2092 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2093 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2094 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2095 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2096 crc = crc16(crc, (__u8 *)gdp + offset,
2097 le16_to_cpu(sbi->s_es->s_desc_size) -
2098 offset);
2099
2100out:
717d50e4
AD		 2101 return cpu_to_le16(crc);
2102}
2103
feb0ab32 2104int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2105 struct ext4_group_desc *gdp)
2106{
feb0ab32 2107 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2108 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2109 return 0;
2110
2111 return 1;
2112}
2113
feb0ab32
DW		 2114void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2115 struct ext4_group_desc *gdp)
2116{
2117 if (!ext4_has_group_desc_csum(sb))
2118 return;
e2b911c5 2119 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2120}
2121
ac27a0ec 2122/* Called at mount-time, super-block is locked */
bfff6873
LC		 2123static int ext4_check_descriptors(struct super_block *sb,
2124 ext4_group_t *first_not_zeroed)
ac27a0ec 2125{
617ba13b
MC		 2126 struct ext4_sb_info *sbi = EXT4_SB(sb);
2127 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2128 ext4_fsblk_t last_block;
bd81d8ee
LV		 2129 ext4_fsblk_t block_bitmap;
2130 ext4_fsblk_t inode_bitmap;
2131 ext4_fsblk_t inode_table;
ce421581 2132 int flexbg_flag = 0;
bfff6873 2133 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2134
e2b911c5 2135 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2136 flexbg_flag = 1;
2137
af5bc92d 2138 ext4_debug("Checking group descriptors");
ac27a0ec 2139
197cd65a
AM		 2140 for (i = 0; i < sbi->s_groups_count; i++) {
2141 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2142
ce421581 2143 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2144 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2145 else
2146 last_block = first_block +
617ba13b 2147 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2148
bfff6873
LC		 2149 if ((grp == sbi->s_groups_count) &&
2150 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2151 grp = i;
2152
8fadc143 2153 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2154 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2155 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2156 "Block bitmap for group %u not in group "
b31e1552 2157 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2158 return 0;
2159 }
8fadc143 2160 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2161 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2162 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2163 "Inode bitmap for group %u not in group "
b31e1552 2164 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2165 return 0;
2166 }
8fadc143 2167 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2168 if (inode_table < first_block ||
2b2d6d01 2169 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2170 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2171 "Inode table for group %u not in group "
b31e1552 2172 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2173 return 0;
2174 }
955ce5f5 2175 ext4_lock_group(sb, i);
feb0ab32 2176 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2177 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2178 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2179 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2180 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2181 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2182 ext4_unlock_group(sb, i);
8a266467 2183 return 0;
7ee1ec4c 2184 }
717d50e4 2185 }
955ce5f5 2186 ext4_unlock_group(sb, i);
ce421581
JS		 2187 if (!flexbg_flag)
2188 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2189 }
bfff6873
LC		 2190 if (NULL != first_not_zeroed)
2191 *first_not_zeroed = grp;
ac27a0ec
DK		 2192 return 1;
2193}
2194
617ba13b 2195/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2196 * the superblock) which were deleted from all directories, but held open by
2197 * a process at the time of a crash. We walk the list and try to delete these
2198 * inodes at recovery time (only with a read-write filesystem).
2199 *
2200 * In order to keep the orphan inode chain consistent during traversal (in
2201 * case of crash during recovery), we link each inode into the superblock
2202 * orphan list_head and handle it the same way as an inode deletion during
2203 * normal operation (which journals the operations for us).
2204 *
2205 * We only do an iget() and an iput() on each inode, which is very safe if we
2206 * accidentally point at an in-use or already deleted inode. The worst that
2207 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2208 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2209 * e2fsck was run on this filesystem, and it must have already done the orphan
2210 * inode cleanup for us, so we can safely abort without any further action.
2211 */
2b2d6d01
TT		 2212static void ext4_orphan_cleanup(struct super_block *sb,
2213 struct ext4_super_block *es)
ac27a0ec
DK		 2214{
2215 unsigned int s_flags = sb->s_flags;
2216 int nr_orphans = 0, nr_truncates = 0;
2217#ifdef CONFIG_QUOTA
2218 int i;
2219#endif
2220 if (!es->s_last_orphan) {
2221 jbd_debug(4, "no orphan inodes to clean up\n");
2222 return;
2223 }
2224
a8f48a95 2225 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2226 ext4_msg(sb, KERN_ERR, "write access "
2227 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2228 return;
2229 }
2230
d39195c3
AG		 2231 /* Check if feature set would not allow a r/w mount */
2232 if (!ext4_feature_set_ok(sb, 0)) {
2233 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2234 "unknown ROCOMPAT features");
2235 return;
2236 }
2237
617ba13b 2238 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2239 /* don't clear list on RO mount w/ errors */
2240 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2241 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2242 "clearing orphan list.\n");
c25f9bc6
ES		 2243 es->s_last_orphan = 0;
2244 }
ac27a0ec
DK		 2245 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2246 return;
2247 }
2248
2249 if (s_flags & MS_RDONLY) {
b31e1552 2250 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2251 sb->s_flags &= ~MS_RDONLY;
2252 }
2253#ifdef CONFIG_QUOTA
2254 /* Needed for iput() to work correctly and not trash data */
2255 sb->s_flags |= MS_ACTIVE;
2256 /* Turn on quotas so that they are updated correctly */
a2d4a646 2257 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2258 if (EXT4_SB(sb)->s_qf_names[i]) {
2259 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2260 if (ret < 0)
b31e1552
ES		 2261 ext4_msg(sb, KERN_ERR,
2262 "Cannot turn on journaled "
2263 "quota: error %d", ret);
ac27a0ec
DK		 2264 }
2265 }
2266#endif
2267
2268 while (es->s_last_orphan) {
2269 struct inode *inode;
2270
97bd42b9
JB		 2271 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2272 if (IS_ERR(inode)) {
ac27a0ec
DK		 2273 es->s_last_orphan = 0;
2274 break;
2275 }
2276
617ba13b 2277 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2278 dquot_initialize(inode);
ac27a0ec 2279 if (inode->i_nlink) {
566370a2
PT		 2280 if (test_opt(sb, DEBUG))
2281 ext4_msg(sb, KERN_DEBUG,
2282 "%s: truncating inode %lu to %lld bytes",
2283 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2284 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2285 inode->i_ino, inode->i_size);
5955102c 2286 inode_lock(inode);
55f252c9 2287 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2288 ext4_truncate(inode);
5955102c 2289 inode_unlock(inode);
ac27a0ec
DK		 2290 nr_truncates++;
2291 } else {
566370a2
PT		 2292 if (test_opt(sb, DEBUG))
2293 ext4_msg(sb, KERN_DEBUG,
2294 "%s: deleting unreferenced inode %lu",
2295 __func__, inode->i_ino);
ac27a0ec
DK		 2296 jbd_debug(2, "deleting unreferenced inode %lu\n",
2297 inode->i_ino);
2298 nr_orphans++;
2299 }
2300 iput(inode); /* The delete magic happens here! */
2301 }
2302
2b2d6d01 2303#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2304
2305 if (nr_orphans)
b31e1552
ES		 2306 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2307 PLURAL(nr_orphans));
ac27a0ec 2308 if (nr_truncates)
b31e1552
ES		 2309 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2310 PLURAL(nr_truncates));
ac27a0ec
DK		 2311#ifdef CONFIG_QUOTA
2312 /* Turn quotas off */
a2d4a646 2313 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2314 if (sb_dqopt(sb)->files[i])
287a8095 2315 dquot_quota_off(sb, i);
ac27a0ec
DK		 2316 }
2317#endif
2318 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2319}
0b8e58a1 2320
cd2291a4
ES		 2321/*
2322 * Maximal extent format file size.
2323 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2324 * extent format containers, within a sector_t, and within i_blocks
2325 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2326 * so that won't be a limiting factor.
2327 *
f17722f9
LC		 2328 * However there is other limiting factor. We do store extents in the form
2329 * of starting block and length, hence the resulting length of the extent
2330 * covering maximum file size must fit into on-disk format containers as
2331 * well. Given that length is always by 1 unit bigger than max unit (because
2332 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2333 *
cd2291a4
ES		 2334 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2335 */
f287a1a5 2336static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2337{
2338 loff_t res;
2339 loff_t upper_limit = MAX_LFS_FILESIZE;
2340
2341 /* small i_blocks in vfs inode? */
f287a1a5 2342 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2343 /*
90c699a9 2344 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2345 * i_block represent total blocks in 512 bytes
2346 * 32 == size of vfs inode i_blocks * 8
2347 */
2348 upper_limit = (1LL << 32) - 1;
2349
2350 /* total blocks in file system block size */
2351 upper_limit >>= (blkbits - 9);
2352 upper_limit <<= blkbits;
2353 }
2354
f17722f9
LC		 2355 /*
2356 * 32-bit extent-start container, ee_block. We lower the maxbytes
2357 * by one fs block, so ee_len can cover the extent of maximum file
2358 * size
2359 */
2360 res = (1LL << 32) - 1;
cd2291a4 2361 res <<= blkbits;
cd2291a4
ES		 2362
2363 /* Sanity check against vm- & vfs- imposed limits */
2364 if (res > upper_limit)
2365 res = upper_limit;
2366
2367 return res;
2368}
ac27a0ec 2369
ac27a0ec 2370/*
cd2291a4 2371 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2372 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2373 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2374 */
f287a1a5 2375static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2376{
617ba13b 2377 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2378 int meta_blocks;
2379 loff_t upper_limit;
0b8e58a1
AD		 2380 /* This is calculated to be the largest file size for a dense, block
2381 * mapped file such that the file's total number of 512-byte sectors,
2382 * including data and all indirect blocks, does not exceed (2^48 - 1).
2383 *
2384 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2385 * number of 512-byte sectors of the file.
0fc1b451
AK		 2386 */
2387
f287a1a5 2388 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2389 /*
90c699a9 2390 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2391 * the inode i_block field represents total file blocks in
2392 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2393 */
2394 upper_limit = (1LL << 32) - 1;
2395
2396 /* total blocks in file system block size */
2397 upper_limit >>= (bits - 9);
2398
2399 } else {
8180a562
AK		 2400 /*
2401 * We use 48 bit ext4_inode i_blocks
2402 * With EXT4_HUGE_FILE_FL set the i_blocks
2403 * represent total number of blocks in
2404 * file system block size
2405 */
0fc1b451
AK		 2406 upper_limit = (1LL << 48) - 1;
2407
0fc1b451
AK		 2408 }
2409
2410 /* indirect blocks */
2411 meta_blocks = 1;
2412 /* double indirect blocks */
2413 meta_blocks += 1 + (1LL << (bits-2));
2414 /* tripple indirect blocks */
2415 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2416
2417 upper_limit -= meta_blocks;
2418 upper_limit <<= bits;
ac27a0ec
DK		 2419
2420 res += 1LL << (bits-2);
2421 res += 1LL << (2*(bits-2));
2422 res += 1LL << (3*(bits-2));
2423 res <<= bits;
2424 if (res > upper_limit)
2425 res = upper_limit;
0fc1b451
AK		 2426
2427 if (res > MAX_LFS_FILESIZE)
2428 res = MAX_LFS_FILESIZE;
2429
ac27a0ec
DK		 2430 return res;
2431}
2432
617ba13b 2433static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2434 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2435{
617ba13b 2436 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2437 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2438 int has_super = 0;
2439
2440 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2441
e2b911c5 2442 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2443 return logical_sb_block + nr + 1;
ac27a0ec 2444 bg = sbi->s_desc_per_block * nr;
617ba13b 2445 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2446 has_super = 1;
0b8e58a1 2447
bd63f6b0
DW		 2448 /*
2449 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2450 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2451 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2452 * compensate.
2453 */
2454 if (sb->s_blocksize == 1024 && nr == 0 &&
2455 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2456 has_super++;
2457
617ba13b 2458 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2459}
2460
c9de560d
AT		 2461/**
2462 * ext4_get_stripe_size: Get the stripe size.
2463 * @sbi: In memory super block info
2464 *
2465 * If we have specified it via mount option, then
2466 * use the mount option value. If the value specified at mount time is
2467 * greater than the blocks per group use the super block value.
2468 * If the super block value is greater than blocks per group return 0.
2469 * Allocator needs it be less than blocks per group.
2470 *
2471 */
2472static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2473{
2474 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2475 unsigned long stripe_width =
2476 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2477 int ret;
c9de560d
AT		 2478
2479 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2480 ret = sbi->s_stripe;
2481 else if (stripe_width <= sbi->s_blocks_per_group)
2482 ret = stripe_width;
2483 else if (stride <= sbi->s_blocks_per_group)
2484 ret = stride;
2485 else
2486 ret = 0;
c9de560d 2487
3eb08658
DE		 2488 /*
2489 * If the stripe width is 1, this makes no sense and
2490 * we set it to 0 to turn off stripe handling code.
2491 */
2492 if (ret <= 1)
2493 ret = 0;
c9de560d 2494
3eb08658 2495 return ret;
c9de560d 2496}
ac27a0ec 2497
a13fb1a4
ES		 2498/*
2499 * Check whether this filesystem can be mounted based on
2500 * the features present and the RDONLY/RDWR mount requested.
2501 * Returns 1 if this filesystem can be mounted as requested,
2502 * 0 if it cannot be.
2503 */
2504static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2505{
e2b911c5 2506 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2507 ext4_msg(sb, KERN_ERR,
2508 "Couldn't mount because of "
2509 "unsupported optional features (%x)",
2510 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2511 ~EXT4_FEATURE_INCOMPAT_SUPP));
2512 return 0;
2513 }
2514
2515 if (readonly)
2516 return 1;
2517
e2b911c5 2518 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b
DW		 2519 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2520 sb->s_flags |= MS_RDONLY;
2521 return 1;
2522 }
2523
a13fb1a4 2524 /* Check that feature set is OK for a read-write mount */
e2b911c5 2525 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2526 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2527 "unsupported optional features (%x)",
2528 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2529 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2530 return 0;
2531 }
2532 /*
2533 * Large file size enabled file system can only be mounted
2534 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2535 */
e2b911c5 2536 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2537 if (sizeof(blkcnt_t) < sizeof(u64)) {
2538 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2539 "cannot be mounted RDWR without "
2540 "CONFIG_LBDAF");
2541 return 0;
2542 }
2543 }
e2b911c5 2544 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2545 ext4_msg(sb, KERN_ERR,
2546 "Can't support bigalloc feature without "
2547 "extents feature\n");
2548 return 0;
2549 }
7c319d32
AK		 2550
2551#ifndef CONFIG_QUOTA
e2b911c5 2552 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2553 ext4_msg(sb, KERN_ERR,
2554 "Filesystem with quota feature cannot be mounted RDWR "
2555 "without CONFIG_QUOTA");
2556 return 0;
2557 }
689c958c
LX		 2558 if (ext4_has_feature_project(sb) && !readonly) {
2559 ext4_msg(sb, KERN_ERR,
2560 "Filesystem with project quota feature cannot be mounted RDWR "
2561 "without CONFIG_QUOTA");
2562 return 0;
2563 }
7c319d32 2564#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2565 return 1;
2566}
2567
66e61a9e
TT		 2568/*
2569 * This function is called once a day if we have errors logged
2570 * on the file system
2571 */
2572static void print_daily_error_info(unsigned long arg)
2573{
2574 struct super_block *sb = (struct super_block *) arg;
2575 struct ext4_sb_info *sbi;
2576 struct ext4_super_block *es;
2577
2578 sbi = EXT4_SB(sb);
2579 es = sbi->s_es;
2580
2581 if (es->s_error_count)
ae0f78de
TT		 2582 /* fsck newer than v1.41.13 is needed to clean this condition. */
2583 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2584 le32_to_cpu(es->s_error_count));
2585 if (es->s_first_error_time) {
ae0f78de 2586 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2587 sb->s_id, le32_to_cpu(es->s_first_error_time),
2588 (int) sizeof(es->s_first_error_func),
2589 es->s_first_error_func,
2590 le32_to_cpu(es->s_first_error_line));
2591 if (es->s_first_error_ino)
2592 printk(": inode %u",
2593 le32_to_cpu(es->s_first_error_ino));
2594 if (es->s_first_error_block)
2595 printk(": block %llu", (unsigned long long)
2596 le64_to_cpu(es->s_first_error_block));
2597 printk("\n");
2598 }
2599 if (es->s_last_error_time) {
ae0f78de 2600 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2601 sb->s_id, le32_to_cpu(es->s_last_error_time),
2602 (int) sizeof(es->s_last_error_func),
2603 es->s_last_error_func,
2604 le32_to_cpu(es->s_last_error_line));
2605 if (es->s_last_error_ino)
2606 printk(": inode %u",
2607 le32_to_cpu(es->s_last_error_ino));
2608 if (es->s_last_error_block)
2609 printk(": block %llu", (unsigned long long)
2610 le64_to_cpu(es->s_last_error_block));
2611 printk("\n");
2612 }
2613 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2614}
2615
bfff6873
LC		 2616/* Find next suitable group and run ext4_init_inode_table */
2617static int ext4_run_li_request(struct ext4_li_request *elr)
2618{
2619 struct ext4_group_desc *gdp = NULL;
2620 ext4_group_t group, ngroups;
2621 struct super_block *sb;
2622 unsigned long timeout = 0;
2623 int ret = 0;
2624
2625 sb = elr->lr_super;
2626 ngroups = EXT4_SB(sb)->s_groups_count;
2627
8e8ad8a5 2628 sb_start_write(sb);
bfff6873
LC		 2629 for (group = elr->lr_next_group; group < ngroups; group++) {
2630 gdp = ext4_get_group_desc(sb, group, NULL);
2631 if (!gdp) {
2632 ret = 1;
2633 break;
2634 }
2635
2636 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2637 break;
2638 }
2639
7f511862 2640 if (group >= ngroups)
bfff6873
LC		 2641 ret = 1;
2642
2643 if (!ret) {
2644 timeout = jiffies;
2645 ret = ext4_init_inode_table(sb, group,
2646 elr->lr_timeout ? 0 : 1);
2647 if (elr->lr_timeout == 0) {
51ce6511
LC		 2648 timeout = (jiffies - timeout) *
2649 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2650 elr->lr_timeout = timeout;
2651 }
2652 elr->lr_next_sched = jiffies + elr->lr_timeout;
2653 elr->lr_next_group = group + 1;
2654 }
8e8ad8a5 2655 sb_end_write(sb);
bfff6873
LC		 2656
2657 return ret;
2658}
2659
2660/*
2661 * Remove lr_request from the list_request and free the
4ed5c033 2662 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2663 */
2664static void ext4_remove_li_request(struct ext4_li_request *elr)
2665{
2666 struct ext4_sb_info *sbi;
2667
2668 if (!elr)
2669 return;
2670
2671 sbi = elr->lr_sbi;
2672
2673 list_del(&elr->lr_request);
2674 sbi->s_li_request = NULL;
2675 kfree(elr);
2676}
2677
2678static void ext4_unregister_li_request(struct super_block *sb)
2679{
1bb933fb
LC		 2680 mutex_lock(&ext4_li_mtx);
2681 if (!ext4_li_info) {
2682 mutex_unlock(&ext4_li_mtx);
bfff6873 2683 return;
1bb933fb 2684 }
bfff6873
LC		 2685
2686 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2687 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2688 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2689 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2690}
2691
8f1f7453
ES		 2692static struct task_struct *ext4_lazyinit_task;
2693
bfff6873
LC		 2694/*
2695 * This is the function where ext4lazyinit thread lives. It walks
2696 * through the request list searching for next scheduled filesystem.
2697 * When such a fs is found, run the lazy initialization request
2698 * (ext4_rn_li_request) and keep track of the time spend in this
2699 * function. Based on that time we compute next schedule time of
2700 * the request. When walking through the list is complete, compute
2701 * next waking time and put itself into sleep.
2702 */
2703static int ext4_lazyinit_thread(void *arg)
2704{
2705 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2706 struct list_head *pos, *n;
2707 struct ext4_li_request *elr;
4ed5c033 2708 unsigned long next_wakeup, cur;
bfff6873
LC		 2709
2710 BUG_ON(NULL == eli);
2711
bfff6873
LC		 2712cont_thread:
2713 while (true) {
2714 next_wakeup = MAX_JIFFY_OFFSET;
2715
2716 mutex_lock(&eli->li_list_mtx);
2717 if (list_empty(&eli->li_request_list)) {
2718 mutex_unlock(&eli->li_list_mtx);
2719 goto exit_thread;
2720 }
2721
2722 list_for_each_safe(pos, n, &eli->li_request_list) {
2723 elr = list_entry(pos, struct ext4_li_request,
2724 lr_request);
2725
b2c78cd0
TT		 2726 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2727 if (ext4_run_li_request(elr) != 0) {
2728 /* error, remove the lazy_init job */
2729 ext4_remove_li_request(elr);
2730 continue;
2731 }
bfff6873
LC		 2732 }
2733
2734 if (time_before(elr->lr_next_sched, next_wakeup))
2735 next_wakeup = elr->lr_next_sched;
2736 }
2737 mutex_unlock(&eli->li_list_mtx);
2738
a0acae0e 2739 try_to_freeze();
bfff6873 2740
4ed5c033
LC		 2741 cur = jiffies;
2742 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2743 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2744 cond_resched();
2745 continue;
2746 }
2747
4ed5c033
LC		 2748 schedule_timeout_interruptible(next_wakeup - cur);
2749
8f1f7453
ES		 2750 if (kthread_should_stop()) {
2751 ext4_clear_request_list();
2752 goto exit_thread;
2753 }
bfff6873
LC		 2754 }
2755
2756exit_thread:
2757 /*
2758 * It looks like the request list is empty, but we need
2759 * to check it under the li_list_mtx lock, to prevent any
2760 * additions into it, and of course we should lock ext4_li_mtx
2761 * to atomically free the list and ext4_li_info, because at
2762 * this point another ext4 filesystem could be registering
2763 * new one.
2764 */
2765 mutex_lock(&ext4_li_mtx);
2766 mutex_lock(&eli->li_list_mtx);
2767 if (!list_empty(&eli->li_request_list)) {
2768 mutex_unlock(&eli->li_list_mtx);
2769 mutex_unlock(&ext4_li_mtx);
2770 goto cont_thread;
2771 }
2772 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2773 kfree(ext4_li_info);
2774 ext4_li_info = NULL;
2775 mutex_unlock(&ext4_li_mtx);
2776
2777 return 0;
2778}
2779
2780static void ext4_clear_request_list(void)
2781{
2782 struct list_head *pos, *n;
2783 struct ext4_li_request *elr;
2784
2785 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2786 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2787 elr = list_entry(pos, struct ext4_li_request,
2788 lr_request);
2789 ext4_remove_li_request(elr);
2790 }
2791 mutex_unlock(&ext4_li_info->li_list_mtx);
2792}
2793
2794static int ext4_run_lazyinit_thread(void)
2795{
8f1f7453
ES		 2796 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2797 ext4_li_info, "ext4lazyinit");
2798 if (IS_ERR(ext4_lazyinit_task)) {
2799 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2800 ext4_clear_request_list();
bfff6873
LC		 2801 kfree(ext4_li_info);
2802 ext4_li_info = NULL;
92b97816 2803 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2804 "initialization thread\n",
2805 err);
2806 return err;
2807 }
2808 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2809 return 0;
2810}
2811
2812/*
2813 * Check whether it make sense to run itable init. thread or not.
2814 * If there is at least one uninitialized inode table, return
2815 * corresponding group number, else the loop goes through all
2816 * groups and return total number of groups.
2817 */
2818static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2819{
2820 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2821 struct ext4_group_desc *gdp = NULL;
2822
2823 for (group = 0; group < ngroups; group++) {
2824 gdp = ext4_get_group_desc(sb, group, NULL);
2825 if (!gdp)
2826 continue;
2827
2828 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2829 break;
2830 }
2831
2832 return group;
2833}
2834
2835static int ext4_li_info_new(void)
2836{
2837 struct ext4_lazy_init *eli = NULL;
2838
2839 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2840 if (!eli)
2841 return -ENOMEM;
2842
bfff6873
LC		 2843 INIT_LIST_HEAD(&eli->li_request_list);
2844 mutex_init(&eli->li_list_mtx);
2845
bfff6873
LC		 2846 eli->li_state |= EXT4_LAZYINIT_QUIT;
2847
2848 ext4_li_info = eli;
2849
2850 return 0;
2851}
2852
2853static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2854 ext4_group_t start)
2855{
2856 struct ext4_sb_info *sbi = EXT4_SB(sb);
2857 struct ext4_li_request *elr;
bfff6873
LC		 2858
2859 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2860 if (!elr)
2861 return NULL;
2862
2863 elr->lr_super = sb;
2864 elr->lr_sbi = sbi;
2865 elr->lr_next_group = start;
2866
2867 /*
2868 * Randomize first schedule time of the request to
2869 * spread the inode table initialization requests
2870 * better.
2871 */
dd1f723b
TT		 2872 elr->lr_next_sched = jiffies + (prandom_u32() %
2873 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 2874 return elr;
2875}
2876
7f511862
TT		 2877int ext4_register_li_request(struct super_block *sb,
2878 ext4_group_t first_not_zeroed)
bfff6873
LC		 2879{
2880 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 2881 struct ext4_li_request *elr = NULL;
bfff6873 2882 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2883 int ret = 0;
bfff6873 2884
7f511862 2885 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 2886 if (sbi->s_li_request != NULL) {
2887 /*
2888 * Reset timeout so it can be computed again, because
2889 * s_li_wait_mult might have changed.
2890 */
2891 sbi->s_li_request->lr_timeout = 0;
7f511862 2892 goto out;
51ce6511 2893 }
bfff6873
LC		 2894
2895 if (first_not_zeroed == ngroups ||
2896 (sb->s_flags & MS_RDONLY) ||
55ff3840 2897 !test_opt(sb, INIT_INODE_TABLE))
7f511862 2898 goto out;
bfff6873
LC		 2899
2900 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 2901 if (!elr) {
2902 ret = -ENOMEM;
2903 goto out;
2904 }
bfff6873
LC		 2905
2906 if (NULL == ext4_li_info) {
2907 ret = ext4_li_info_new();
2908 if (ret)
2909 goto out;
2910 }
2911
2912 mutex_lock(&ext4_li_info->li_list_mtx);
2913 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2914 mutex_unlock(&ext4_li_info->li_list_mtx);
2915
2916 sbi->s_li_request = elr;
46e4690b
TM		 2917 /*
2918 * set elr to NULL here since it has been inserted to
2919 * the request_list and the removal and free of it is
2920 * handled by ext4_clear_request_list from now on.
2921 */
2922 elr = NULL;
bfff6873
LC		 2923
2924 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2925 ret = ext4_run_lazyinit_thread();
2926 if (ret)
2927 goto out;
2928 }
bfff6873 2929out:
beed5ecb
NK		 2930 mutex_unlock(&ext4_li_mtx);
2931 if (ret)
bfff6873 2932 kfree(elr);
bfff6873
LC		 2933 return ret;
2934}
2935
2936/*
2937 * We do not need to lock anything since this is called on
2938 * module unload.
2939 */
2940static void ext4_destroy_lazyinit_thread(void)
2941{
2942 /*
2943 * If thread exited earlier
2944 * there's nothing to be done.
2945 */
8f1f7453 2946 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2947 return;
2948
8f1f7453 2949 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2950}
2951
25ed6e8a
DW		 2952static int set_journal_csum_feature_set(struct super_block *sb)
2953{
2954 int ret = 1;
2955 int compat, incompat;
2956 struct ext4_sb_info *sbi = EXT4_SB(sb);
2957
9aa5d32b 2958 if (ext4_has_metadata_csum(sb)) {
db9ee220 2959 /* journal checksum v3 */
25ed6e8a 2960 compat = 0;
db9ee220 2961 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 2962 } else {
2963 /* journal checksum v1 */
2964 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2965 incompat = 0;
2966 }
2967
feb8c6d3
DW		 2968 jbd2_journal_clear_features(sbi->s_journal,
2969 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2970 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2971 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 2972 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2973 ret = jbd2_journal_set_features(sbi->s_journal,
2974 compat, 0,
2975 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2976 incompat);
2977 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2978 ret = jbd2_journal_set_features(sbi->s_journal,
2979 compat, 0,
2980 incompat);
2981 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2982 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2983 } else {
feb8c6d3
DW		 2984 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2985 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 2986 }
2987
2988 return ret;
2989}
2990
952fc18e
TT		 2991/*
2992 * Note: calculating the overhead so we can be compatible with
2993 * historical BSD practice is quite difficult in the face of
2994 * clusters/bigalloc. This is because multiple metadata blocks from
2995 * different block group can end up in the same allocation cluster.
2996 * Calculating the exact overhead in the face of clustered allocation
2997 * requires either O(all block bitmaps) in memory or O(number of block
2998 * groups**2) in time. We will still calculate the superblock for
2999 * older file systems --- and if we come across with a bigalloc file
3000 * system with zero in s_overhead_clusters the estimate will be close to
3001 * correct especially for very large cluster sizes --- but for newer
3002 * file systems, it's better to calculate this figure once at mkfs
3003 * time, and store it in the superblock. If the superblock value is
3004 * present (even for non-bigalloc file systems), we will use it.
3005 */
3006static int count_overhead(struct super_block *sb, ext4_group_t grp,
3007 char *buf)
3008{
3009 struct ext4_sb_info *sbi = EXT4_SB(sb);
3010 struct ext4_group_desc *gdp;
3011 ext4_fsblk_t first_block, last_block, b;
3012 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3013 int s, j, count = 0;
3014
e2b911c5 3015 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3016 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3017 sbi->s_itb_per_group + 2);
3018
952fc18e
TT		 3019 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3020 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3021 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3022 for (i = 0; i < ngroups; i++) {
3023 gdp = ext4_get_group_desc(sb, i, NULL);
3024 b = ext4_block_bitmap(sb, gdp);
3025 if (b >= first_block && b <= last_block) {
3026 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3027 count++;
3028 }
3029 b = ext4_inode_bitmap(sb, gdp);
3030 if (b >= first_block && b <= last_block) {
3031 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3032 count++;
3033 }
3034 b = ext4_inode_table(sb, gdp);
3035 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3036 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3037 int c = EXT4_B2C(sbi, b - first_block);
3038 ext4_set_bit(c, buf);
3039 count++;
3040 }
3041 if (i != grp)
3042 continue;
3043 s = 0;
3044 if (ext4_bg_has_super(sb, grp)) {
3045 ext4_set_bit(s++, buf);
3046 count++;
3047 }
3048 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3049 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3050 count++;
3051 }
3052 }
3053 if (!count)
3054 return 0;
3055 return EXT4_CLUSTERS_PER_GROUP(sb) -
3056 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3057}
3058
3059/*
3060 * Compute the overhead and stash it in sbi->s_overhead
3061 */
3062int ext4_calculate_overhead(struct super_block *sb)
3063{
3064 struct ext4_sb_info *sbi = EXT4_SB(sb);
3065 struct ext4_super_block *es = sbi->s_es;
3066 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3067 ext4_fsblk_t overhead = 0;
4fdb5543 3068 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3069
952fc18e
TT		 3070 if (!buf)
3071 return -ENOMEM;
3072
3073 /*
3074 * Compute the overhead (FS structures). This is constant
3075 * for a given filesystem unless the number of block groups
3076 * changes so we cache the previous value until it does.
3077 */
3078
3079 /*
3080 * All of the blocks before first_data_block are overhead
3081 */
3082 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3083
3084 /*
3085 * Add the overhead found in each block group
3086 */
3087 for (i = 0; i < ngroups; i++) {
3088 int blks;
3089
3090 blks = count_overhead(sb, i, buf);
3091 overhead += blks;
3092 if (blks)
3093 memset(buf, 0, PAGE_SIZE);
3094 cond_resched();
3095 }
b003b524
ES		 3096 /* Add the internal journal blocks as well */
3097 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3098 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3099
952fc18e
TT		 3100 sbi->s_overhead = overhead;
3101 smp_wmb();
3102 free_page((unsigned long) buf);
3103 return 0;
3104}
3105
b5799018 3106static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3107{
3108 ext4_fsblk_t resv_clusters;
b5799018 3109 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3110
30fac0f7
JK		 3111 /*
3112 * There's no need to reserve anything when we aren't using extents.
3113 * The space estimates are exact, there are no unwritten extents,
3114 * hole punching doesn't need new metadata... This is needed especially
3115 * to keep ext2/3 backward compatibility.
3116 */
e2b911c5 3117 if (!ext4_has_feature_extents(sb))
b5799018 3118 return;
27dd4385
LC		 3119 /*
3120 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3121 * This should cover the situations where we can not afford to run
3122 * out of space like for example punch hole, or converting
556615dc 3123 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3124 * allocation would require 1, or 2 blocks, higher numbers are
3125 * very rare.
3126 */
b5799018
TT		 3127 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3128 sbi->s_cluster_bits);
27dd4385
LC		 3129
3130 do_div(resv_clusters, 50);
3131 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3132
b5799018 3133 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3134}
3135
2b2d6d01 3136static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3137{
d4c402d9 3138 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3139 struct buffer_head *bh;
617ba13b
MC		 3140 struct ext4_super_block *es = NULL;
3141 struct ext4_sb_info *sbi;
3142 ext4_fsblk_t block;
3143 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3144 ext4_fsblk_t logical_sb_block;
ac27a0ec 3145 unsigned long offset = 0;
ac27a0ec
DK		 3146 unsigned long journal_devnum = 0;
3147 unsigned long def_mount_opts;
3148 struct inode *root;
0390131b 3149 const char *descr;
dcc7dae3 3150 int ret = -ENOMEM;
281b5995 3151 int blocksize, clustersize;
4ec11028
TT		 3152 unsigned int db_count;
3153 unsigned int i;
281b5995 3154 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3155 __u64 blocks_count;
07aa2ea1 3156 int err = 0;
b3881f74 3157 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3158 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3159
3160 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3161 if (!sbi)
dcc7dae3 3162 goto out_free_orig;
705895b6
PE		 3163
3164 sbi->s_blockgroup_lock =
3165 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3166 if (!sbi->s_blockgroup_lock) {
3167 kfree(sbi);
dcc7dae3 3168 goto out_free_orig;
705895b6 3169 }
ac27a0ec 3170 sb->s_fs_info = sbi;
2c0544b2 3171 sbi->s_sb = sb;
240799cd 3172 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3173 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3174 if (sb->s_bdev->bd_part)
3175 sbi->s_sectors_written_start =
3176 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3177
9f6200bb 3178 /* Cleanup superblock name */
ec3904dc 3179 strreplace(sb->s_id, '/', '!');
9f6200bb 3180
07aa2ea1 3181 /* -EINVAL is default */
dcc7dae3 3182 ret = -EINVAL;
617ba13b 3183 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3184 if (!blocksize) {
b31e1552 3185 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3186 goto out_fail;
3187 }
3188
3189 /*
617ba13b 3190 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3191 * block sizes. We need to calculate the offset from buffer start.
3192 */
617ba13b 3193 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3194 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3195 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3196 } else {
70bbb3e0 3197 logical_sb_block = sb_block;
ac27a0ec
DK		 3198 }
3199
a8ac900b 3200 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3201 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3202 goto out_fail;
3203 }
3204 /*
3205 * Note: s_es must be initialized as soon as possible because
617ba13b 3206 * some ext4 macro-instructions depend on its value
ac27a0ec 3207 */
2716b802 3208 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3209 sbi->s_es = es;
3210 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3211 if (sb->s_magic != EXT4_SUPER_MAGIC)
3212 goto cantfind_ext4;
afc32f7e 3213 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3214
feb0ab32 3215 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3216 if (ext4_has_feature_metadata_csum(sb) &&
3217 ext4_has_feature_gdt_csum(sb))
363307e6 3218 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3219 "redundant flags; please run fsck.");
3220
d25425f8
DW		 3221 /* Check for a known checksum algorithm */
3222 if (!ext4_verify_csum_type(sb, es)) {
3223 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3224 "unknown checksum algorithm.");
3225 silent = 1;
3226 goto cantfind_ext4;
3227 }
3228
0441984a 3229 /* Load the checksum driver */
e2b911c5 3230 if (ext4_has_feature_metadata_csum(sb)) {
0441984a
DW		 3231 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3232 if (IS_ERR(sbi->s_chksum_driver)) {
3233 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3234 ret = PTR_ERR(sbi->s_chksum_driver);
3235 sbi->s_chksum_driver = NULL;
3236 goto failed_mount;
3237 }
3238 }
3239
a9c47317
DW		 3240 /* Check superblock checksum */
3241 if (!ext4_superblock_csum_verify(sb, es)) {
3242 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3243 "invalid superblock checksum. Run e2fsck?");
3244 silent = 1;
6a797d27 3245 ret = -EFSBADCRC;
a9c47317
DW		 3246 goto cantfind_ext4;
3247 }
3248
3249 /* Precompute checksum seed for all metadata */
e2b911c5 3250 if (ext4_has_feature_csum_seed(sb))
8c81bd8f
DW		 3251 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3252 else if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3253 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3254 sizeof(es->s_uuid));
3255
ac27a0ec
DK		 3256 /* Set defaults before we parse the mount options */
3257 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3258 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3259 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3260 set_opt(sb, DEBUG);
87f26807 3261 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3262 set_opt(sb, GRPID);
617ba13b 3263 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3264 set_opt(sb, NO_UID32);
ea663336 3265 /* xattr user namespace & acls are now defaulted on */
ea663336 3266 set_opt(sb, XATTR_USER);
03010a33 3267#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3268 set_opt(sb, POSIX_ACL);
2e7842b8 3269#endif
98c1a759
DW		 3270 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3271 if (ext4_has_metadata_csum(sb))
3272 set_opt(sb, JOURNAL_CHECKSUM);
3273
617ba13b 3274 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3275 set_opt(sb, JOURNAL_DATA);
617ba13b 3276 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3277 set_opt(sb, ORDERED_DATA);
617ba13b 3278 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3279 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3280
3281 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3282 set_opt(sb, ERRORS_PANIC);
bb4f397a 3283 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3284 set_opt(sb, ERRORS_CONT);
bb4f397a 3285 else
fd8c37ec 3286 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3287 /* block_validity enabled by default; disable with noblock_validity */
3288 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3289 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3290 set_opt(sb, DISCARD);
ac27a0ec 3291
08cefc7a
EB		 3292 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3293 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3294 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3295 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3296 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3297
8b67f04a 3298 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3299 set_opt(sb, BARRIER);
ac27a0ec 3300
dd919b98
AK		 3301 /*
3302 * enable delayed allocation by default
3303 * Use -o nodelalloc to turn it off
3304 */
bc0b75f7 3305 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3306 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3307 set_opt(sb, DELALLOC);
dd919b98 3308
51ce6511
LC		 3309 /*
3310 * set default s_li_wait_mult for lazyinit, for the case there is
3311 * no mount option specified.
3312 */
3313 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3314
8b67f04a 3315 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3316 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3317 ext4_msg(sb, KERN_WARNING,
3318 "failed to parse options in superblock: %s",
3319 sbi->s_es->s_mount_opts);
3320 }
5a916be1 3321 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3322 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3323 &journal_ioprio, 0))
ac27a0ec
DK		 3324 goto failed_mount;
3325
56889787
TT		 3326 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3327 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3328 "with data=journal disables delayed "
3329 "allocation and O_DIRECT support!\n");
3330 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3331 ext4_msg(sb, KERN_ERR, "can't mount with "
3332 "both data=journal and delalloc");
3333 goto failed_mount;
3334 }
3335 if (test_opt(sb, DIOREAD_NOLOCK)) {
3336 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3337 "both data=journal and dioread_nolock");
56889787
TT		 3338 goto failed_mount;
3339 }
923ae0ff
RZ		 3340 if (test_opt(sb, DAX)) {
3341 ext4_msg(sb, KERN_ERR, "can't mount with "
3342 "both data=journal and dax");
3343 goto failed_mount;
3344 }
56889787
TT		 3345 if (test_opt(sb, DELALLOC))
3346 clear_opt(sb, DELALLOC);
001e4a87
TH		 3347 } else {
3348 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3349 }
3350
ac27a0ec 3351 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3352 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3353
617ba13b 3354 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3355 (ext4_has_compat_features(sb) ||
3356 ext4_has_ro_compat_features(sb) ||
3357 ext4_has_incompat_features(sb)))
b31e1552
ES		 3358 ext4_msg(sb, KERN_WARNING,
3359 "feature flags set on rev 0 fs, "
3360 "running e2fsck is recommended");
469108ff 3361
ed3654eb
TT		 3362 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3363 set_opt2(sb, HURD_COMPAT);
e2b911c5 3364 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3365 ext4_msg(sb, KERN_ERR,
3366 "The Hurd can't support 64-bit file systems");
3367 goto failed_mount;
3368 }
3369 }
3370
2035e776
TT		 3371 if (IS_EXT2_SB(sb)) {
3372 if (ext2_feature_set_ok(sb))
3373 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3374 "using the ext4 subsystem");
3375 else {
3376 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3377 "to feature incompatibilities");
3378 goto failed_mount;
3379 }
3380 }
3381
3382 if (IS_EXT3_SB(sb)) {
3383 if (ext3_feature_set_ok(sb))
3384 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3385 "using the ext4 subsystem");
3386 else {
3387 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3388 "to feature incompatibilities");
3389 goto failed_mount;
3390 }
3391 }
3392
ac27a0ec
DK		 3393 /*
3394 * Check feature flags regardless of the revision level, since we
3395 * previously didn't change the revision level when setting the flags,
3396 * so there is a chance incompat flags are set on a rev 0 filesystem.
3397 */
a13fb1a4 3398 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3399 goto failed_mount;
a13fb1a4 3400
261cb20c 3401 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3402 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3403 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3404 ext4_msg(sb, KERN_ERR,
3405 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3406 goto failed_mount;
3407 }
3408
923ae0ff
RZ		 3409 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3410 if (blocksize != PAGE_SIZE) {
3411 ext4_msg(sb, KERN_ERR,
3412 "error: unsupported blocksize for dax");
3413 goto failed_mount;
3414 }
3415 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3416 ext4_msg(sb, KERN_ERR,
3417 "error: device does not support dax");
3418 goto failed_mount;
3419 }
3420 }
3421
e2b911c5 3422 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3423 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3424 es->s_encryption_level);
3425 goto failed_mount;
3426 }
3427
ac27a0ec 3428 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3429 /* Validate the filesystem blocksize */
3430 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3431 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3432 blocksize);
ac27a0ec
DK		 3433 goto failed_mount;
3434 }
3435
2b2d6d01 3436 brelse(bh);
70bbb3e0
AM		 3437 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3438 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3439 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3440 if (!bh) {
b31e1552
ES		 3441 ext4_msg(sb, KERN_ERR,
3442 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3443 goto failed_mount;
3444 }
2716b802 3445 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3446 sbi->s_es = es;
617ba13b 3447 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3448 ext4_msg(sb, KERN_ERR,
3449 "Magic mismatch, very weird!");
ac27a0ec
DK		 3450 goto failed_mount;
3451 }
3452 }
3453
e2b911c5 3454 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3455 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3456 has_huge_files);
3457 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3458
617ba13b
MC		 3459 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3460 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3461 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3462 } else {
3463 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3464 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3465 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3466 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3467 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3468 ext4_msg(sb, KERN_ERR,
3469 "unsupported inode size: %d",
2b2d6d01 3470 sbi->s_inode_size);
ac27a0ec
DK		 3471 goto failed_mount;
3472 }
ef7f3835
KS		 3473 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3474 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3475 }
0b8e58a1 3476
0d1ee42f 3477 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3478 if (ext4_has_feature_64bit(sb)) {
8fadc143 3479 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3480 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3481 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3482 ext4_msg(sb, KERN_ERR,
3483 "unsupported descriptor size %lu",
0d1ee42f
AR		 3484 sbi->s_desc_size);
3485 goto failed_mount;
3486 }
3487 } else
3488 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3489
ac27a0ec 3490 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3491 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3492 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3493 goto cantfind_ext4;
0b8e58a1 3494
617ba13b 3495 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3496 if (sbi->s_inodes_per_block == 0)
617ba13b 3497 goto cantfind_ext4;
ac27a0ec
DK		 3498 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3499 sbi->s_inodes_per_block;
0d1ee42f 3500 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3501 sbi->s_sbh = bh;
3502 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3503 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3504 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3505
2b2d6d01 3506 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3507 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3508 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3509 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3510 i = le32_to_cpu(es->s_flags);
3511 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3512 sbi->s_hash_unsigned = 3;
3513 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3514#ifdef __CHAR_UNSIGNED__
23301410
TT		 3515 if (!(sb->s_flags & MS_RDONLY))
3516 es->s_flags |=
3517 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3518 sbi->s_hash_unsigned = 3;
f99b2589 3519#else
23301410
TT		 3520 if (!(sb->s_flags & MS_RDONLY))
3521 es->s_flags |=
3522 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3523#endif
23301410 3524 }
f99b2589 3525 }
ac27a0ec 3526
281b5995
TT		 3527 /* Handle clustersize */
3528 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3529 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3530 if (has_bigalloc) {
3531 if (clustersize < blocksize) {
3532 ext4_msg(sb, KERN_ERR,
3533 "cluster size (%d) smaller than "
3534 "block size (%d)", clustersize, blocksize);
3535 goto failed_mount;
3536 }
3537 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3538 le32_to_cpu(es->s_log_block_size);
3539 sbi->s_clusters_per_group =
3540 le32_to_cpu(es->s_clusters_per_group);
3541 if (sbi->s_clusters_per_group > blocksize * 8) {
3542 ext4_msg(sb, KERN_ERR,
3543 "#clusters per group too big: %lu",
3544 sbi->s_clusters_per_group);
3545 goto failed_mount;
3546 }
3547 if (sbi->s_blocks_per_group !=
3548 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3549 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3550 "clusters per group (%lu) inconsistent",
3551 sbi->s_blocks_per_group,
3552 sbi->s_clusters_per_group);
3553 goto failed_mount;
3554 }
3555 } else {
3556 if (clustersize != blocksize) {
3557 ext4_warning(sb, "fragment/cluster size (%d) != "
3558 "block size (%d)", clustersize,
3559 blocksize);
3560 clustersize = blocksize;
3561 }
3562 if (sbi->s_blocks_per_group > blocksize * 8) {
3563 ext4_msg(sb, KERN_ERR,
3564 "#blocks per group too big: %lu",
3565 sbi->s_blocks_per_group);
3566 goto failed_mount;
3567 }
3568 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3569 sbi->s_cluster_bits = 0;
ac27a0ec 3570 }
281b5995
TT		 3571 sbi->s_cluster_ratio = clustersize / blocksize;
3572
ac27a0ec 3573 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3574 ext4_msg(sb, KERN_ERR,
3575 "#inodes per group too big: %lu",
2b2d6d01 3576 sbi->s_inodes_per_group);
ac27a0ec
DK		 3577 goto failed_mount;
3578 }
3579
960fd856
TT		 3580 /* Do we have standard group size of clustersize * 8 blocks ? */
3581 if (sbi->s_blocks_per_group == clustersize << 3)
3582 set_opt2(sb, STD_GROUP_SIZE);
3583
bf43d84b
ES		 3584 /*
3585 * Test whether we have more sectors than will fit in sector_t,
3586 * and whether the max offset is addressable by the page cache.
3587 */
5a9ae68a 3588 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3589 ext4_blocks_count(es));
5a9ae68a 3590 if (err) {
b31e1552 3591 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3592 " too large to mount safely on this system");
ac27a0ec 3593 if (sizeof(sector_t) < 8)
90c699a9 3594 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3595 goto failed_mount;
3596 }
3597
617ba13b
MC		 3598 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3599 goto cantfind_ext4;
e7c95593 3600
0f2ddca6
FTN		 3601 /* check blocks count against device size */
3602 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3603 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3604 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3605 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3606 ext4_blocks_count(es), blocks_count);
3607 goto failed_mount;
3608 }
3609
0b8e58a1
AD		 3610 /*
3611 * It makes no sense for the first data block to be beyond the end
3612 * of the filesystem.
3613 */
3614 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3615 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3616 "block %u is beyond end of filesystem (%llu)",
3617 le32_to_cpu(es->s_first_data_block),
3618 ext4_blocks_count(es));
e7c95593
ES		 3619 goto failed_mount;
3620 }
bd81d8ee
LV		 3621 blocks_count = (ext4_blocks_count(es) -
3622 le32_to_cpu(es->s_first_data_block) +
3623 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3624 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3625 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3626 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3627 "(block count %llu, first data block %u, "
b31e1552 3628 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3629 ext4_blocks_count(es),
3630 le32_to_cpu(es->s_first_data_block),
3631 EXT4_BLOCKS_PER_GROUP(sb));
3632 goto failed_mount;
3633 }
bd81d8ee 3634 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3635 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3636 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3637 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3638 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3639 sbi->s_group_desc = ext4_kvmalloc(db_count *
3640 sizeof(struct buffer_head *),
3641 GFP_KERNEL);
ac27a0ec 3642 if (sbi->s_group_desc == NULL) {
b31e1552 3643 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3644 ret = -ENOMEM;
ac27a0ec
DK		 3645 goto failed_mount;
3646 }
3647
705895b6 3648 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3649
3650 for (i = 0; i < db_count; i++) {
70bbb3e0 3651 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3652 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3653 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3654 ext4_msg(sb, KERN_ERR,
3655 "can't read group descriptor %d", i);
ac27a0ec
DK		 3656 db_count = i;
3657 goto failed_mount2;
3658 }
3659 }
bfff6873 3660 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3661 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 3662 ret = -EFSCORRUPTED;
f9ae9cf5 3663 goto failed_mount2;
ac27a0ec 3664 }
772cb7c8 3665
f9ae9cf5 3666 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3667 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3668 spin_lock_init(&sbi->s_next_gen_lock);
3669
04ecddb7
JM		 3670 setup_timer(&sbi->s_err_report, print_daily_error_info,
3671 (unsigned long) sb);
04496411 3672
a75ae78f 3673 /* Register extent status tree shrinker */
eb68d0e2 3674 if (ext4_es_register_shrinker(sbi))
ce7e010a 3675 goto failed_mount3;
ce7e010a 3676
c9de560d 3677 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3678 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3679
f9ae9cf5
TT		 3680 /*
3681 * set up enough so that it can read an inode
3682 */
f6e63f90 3683 sb->s_op = &ext4_sops;
617ba13b
MC		 3684 sb->s_export_op = &ext4_export_ops;
3685 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3686#ifdef CONFIG_QUOTA
617ba13b 3687 sb->dq_op = &ext4_quota_operations;
e2b911c5 3688 if (ext4_has_feature_quota(sb))
1fa5efe3 3689 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3690 else
3691 sb->s_qcop = &ext4_qctl_operations;
689c958c 3692 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 3693#endif
f2fa2ffc
AK		 3694 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3695
ac27a0ec 3696 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3697 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3698
3699 sb->s_root = NULL;
3700
3701 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 3702 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 3703
e2b911c5 3704 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
c5e06d10 3705 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3706 goto failed_mount3a;
c5e06d10 3707
ac27a0ec
DK		 3708 /*
3709 * The first inode we look at is the journal inode. Don't try
3710 * root first: it may be modified in the journal!
3711 */
e2b911c5 3712 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
617ba13b 3713 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3714 goto failed_mount3a;
0390131b 3715 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
e2b911c5 3716 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 3717 ext4_msg(sb, KERN_ERR, "required journal recovery "
3718 "suppressed and not mounted read-only");
744692dc 3719 goto failed_mount_wq;
ac27a0ec 3720 } else {
1e381f60
DM		 3721 /* Nojournal mode, all journal mount options are illegal */
3722 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3723 ext4_msg(sb, KERN_ERR, "can't mount with "
3724 "journal_checksum, fs mounted w/o journal");
3725 goto failed_mount_wq;
3726 }
3727 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3728 ext4_msg(sb, KERN_ERR, "can't mount with "
3729 "journal_async_commit, fs mounted w/o journal");
3730 goto failed_mount_wq;
3731 }
3732 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3733 ext4_msg(sb, KERN_ERR, "can't mount with "
3734 "commit=%lu, fs mounted w/o journal",
3735 sbi->s_commit_interval / HZ);
3736 goto failed_mount_wq;
3737 }
3738 if (EXT4_MOUNT_DATA_FLAGS &
3739 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3740 ext4_msg(sb, KERN_ERR, "can't mount with "
3741 "data=, fs mounted w/o journal");
3742 goto failed_mount_wq;
3743 }
3744 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3745 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 3746 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3747 sbi->s_journal = NULL;
3748 needs_recovery = 0;
3749 goto no_journal;
ac27a0ec
DK		 3750 }
3751
e2b911c5 3752 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 3753 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3754 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3755 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3756 goto failed_mount_wq;
eb40a09c
JS		 3757 }
3758
25ed6e8a
DW		 3759 if (!set_journal_csum_feature_set(sb)) {
3760 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3761 "feature set");
3762 goto failed_mount_wq;
d4da6c9c 3763 }
818d276c 3764
ac27a0ec
DK		 3765 /* We have now updated the journal if required, so we can
3766 * validate the data journaling mode. */
3767 switch (test_opt(sb, DATA_FLAGS)) {
3768 case 0:
3769 /* No mode set, assume a default based on the journal
63f57933
AM		 3770 * capabilities: ORDERED_DATA if the journal can
3771 * cope, else JOURNAL_DATA
3772 */
dab291af
MC		 3773 if (jbd2_journal_check_available_features
3774 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3775 set_opt(sb, ORDERED_DATA);
ac27a0ec 3776 else
fd8c37ec 3777 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3778 break;
3779
617ba13b
MC		 3780 case EXT4_MOUNT_ORDERED_DATA:
3781 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3782 if (!jbd2_journal_check_available_features
3783 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3784 ext4_msg(sb, KERN_ERR, "Journal does not support "
3785 "requested data journaling mode");
744692dc 3786 goto failed_mount_wq;
ac27a0ec
DK		 3787 }
3788 default:
3789 break;
3790 }
b3881f74 3791 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3792
18aadd47
BJ		 3793 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3794
ce7e010a 3795no_journal:
2335d05f
AG		 3796 sbi->s_mb_cache = ext4_xattr_create_cache();
3797 if (!sbi->s_mb_cache) {
3798 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3799 goto failed_mount_wq;
9c191f70
M		 3800 }
3801
e2b911c5 3802 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
1cb767cd
TT		 3803 (blocksize != PAGE_CACHE_SIZE)) {
3804 ext4_msg(sb, KERN_ERR,
3805 "Unsupported blocksize for fs encryption");
3806 goto failed_mount_wq;
3807 }
3808
e2b911c5
DW		 3809 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3810 !ext4_has_feature_encrypt(sb)) {
3811 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 3812 ext4_commit_super(sb, 1);
3813 }
3814
952fc18e
TT		 3815 /*
3816 * Get the # of file system overhead blocks from the
3817 * superblock if present.
3818 */
3819 if (es->s_overhead_clusters)
3820 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3821 else {
07aa2ea1
LC		 3822 err = ext4_calculate_overhead(sb);
3823 if (err)
952fc18e
TT		 3824 goto failed_mount_wq;
3825 }
3826
fd89d5f2
TH		 3827 /*
3828 * The maximum number of concurrent works can be high and
3829 * concurrency isn't really necessary. Limit it to 1.
3830 */
2e8fa54e
JK		 3831 EXT4_SB(sb)->rsv_conversion_wq =
3832 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3833 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3834 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 3835 ret = -ENOMEM;
2e8fa54e
JK		 3836 goto failed_mount4;
3837 }
3838
ac27a0ec 3839 /*
dab291af 3840 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3841 * so we can safely mount the rest of the filesystem now.
3842 */
3843
1d1fe1ee
DH		 3844 root = ext4_iget(sb, EXT4_ROOT_INO);
3845 if (IS_ERR(root)) {
b31e1552 3846 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3847 ret = PTR_ERR(root);
32a9bb57 3848 root = NULL;
ac27a0ec
DK		 3849 goto failed_mount4;
3850 }
3851 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3852 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3853 iput(root);
ac27a0ec
DK		 3854 goto failed_mount4;
3855 }
48fde701 3856 sb->s_root = d_make_root(root);
1d1fe1ee 3857 if (!sb->s_root) {
b31e1552 3858 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3859 ret = -ENOMEM;
3860 goto failed_mount4;
3861 }
ac27a0ec 3862
7e84b621
ES		 3863 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3864 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3865
3866 /* determine the minimum size of new large inodes, if present */
3867 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3868 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3869 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 3870 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 3871 if (sbi->s_want_extra_isize <
3872 le16_to_cpu(es->s_want_extra_isize))
3873 sbi->s_want_extra_isize =
3874 le16_to_cpu(es->s_want_extra_isize);
3875 if (sbi->s_want_extra_isize <
3876 le16_to_cpu(es->s_min_extra_isize))
3877 sbi->s_want_extra_isize =
3878 le16_to_cpu(es->s_min_extra_isize);
3879 }
3880 }
3881 /* Check if enough inode space is available */
3882 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3883 sbi->s_inode_size) {
3884 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3885 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3886 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3887 "available");
ef7f3835
KS		 3888 }
3889
b5799018 3890 ext4_set_resv_clusters(sb);
27dd4385 3891
6fd058f7
TT		 3892 err = ext4_setup_system_zone(sb);
3893 if (err) {
b31e1552 3894 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3895 "zone (%d)", err);
f9ae9cf5
TT		 3896 goto failed_mount4a;
3897 }
3898
3899 ext4_ext_init(sb);
3900 err = ext4_mb_init(sb);
3901 if (err) {
3902 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3903 err);
dcf2d804 3904 goto failed_mount5;
c2774d84
AK		 3905 }
3906
d5e03cbb
TT		 3907 block = ext4_count_free_clusters(sb);
3908 ext4_free_blocks_count_set(sbi->s_es,
3909 EXT4_C2B(sbi, block));
908c7f19
TH		 3910 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3911 GFP_KERNEL);
d5e03cbb
TT		 3912 if (!err) {
3913 unsigned long freei = ext4_count_free_inodes(sb);
3914 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 3915 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3916 GFP_KERNEL);
d5e03cbb
TT		 3917 }
3918 if (!err)
3919 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 3920 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 3921 if (!err)
908c7f19
TH		 3922 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3923 GFP_KERNEL);
d5e03cbb
TT		 3924 if (err) {
3925 ext4_msg(sb, KERN_ERR, "insufficient memory");
3926 goto failed_mount6;
3927 }
3928
e2b911c5 3929 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 3930 if (!ext4_fill_flex_info(sb)) {
3931 ext4_msg(sb, KERN_ERR,
3932 "unable to initialize "
3933 "flex_bg meta info!");
3934 goto failed_mount6;
3935 }
3936
bfff6873
LC		 3937 err = ext4_register_li_request(sb, first_not_zeroed);
3938 if (err)
dcf2d804 3939 goto failed_mount6;
bfff6873 3940
b5799018 3941 err = ext4_register_sysfs(sb);
dcf2d804
TM		 3942 if (err)
3943 goto failed_mount7;
3197ebdb 3944
9b2ff357
JK		 3945#ifdef CONFIG_QUOTA
3946 /* Enable quota usage during mount. */
e2b911c5 3947 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
9b2ff357
JK		 3948 err = ext4_enable_quotas(sb);
3949 if (err)
3950 goto failed_mount8;
3951 }
3952#endif /* CONFIG_QUOTA */
3953
617ba13b
MC		 3954 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3955 ext4_orphan_cleanup(sb, es);
3956 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3957 if (needs_recovery) {
b31e1552 3958 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3959 ext4_mark_recovery_complete(sb, es);
3960 }
3961 if (EXT4_SB(sb)->s_journal) {
3962 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3963 descr = " journalled data mode";
3964 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3965 descr = " ordered data mode";
3966 else
3967 descr = " writeback data mode";
3968 } else
3969 descr = "out journal";
3970
79add3a3
LC		 3971 if (test_opt(sb, DISCARD)) {
3972 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3973 if (!blk_queue_discard(q))
3974 ext4_msg(sb, KERN_WARNING,
3975 "mounting with \"discard\" option, but "
3976 "the device does not support discard");
3977 }
3978
e294a537
TT		 3979 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3980 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3981 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3982 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3983
66e61a9e
TT		 3984 if (es->s_error_count)
3985 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3986
efbed4dc
TT		 3987 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3988 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
3989 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
3990 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
3991
d4c402d9 3992 kfree(orig_data);
ac27a0ec
DK		 3993 return 0;
3994
617ba13b 3995cantfind_ext4:
ac27a0ec 3996 if (!silent)
b31e1552 3997 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3998 goto failed_mount;
3999
72ba7450
TT		 4000#ifdef CONFIG_QUOTA
4001failed_mount8:
ebd173be 4002 ext4_unregister_sysfs(sb);
72ba7450 4003#endif
dcf2d804
TM		 4004failed_mount7:
4005 ext4_unregister_li_request(sb);
4006failed_mount6:
f9ae9cf5 4007 ext4_mb_release(sb);
d5e03cbb 4008 if (sbi->s_flex_groups)
b93b41d4 4009 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4010 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4011 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4012 percpu_counter_destroy(&sbi->s_dirs_counter);
4013 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4014failed_mount5:
f9ae9cf5
TT		 4015 ext4_ext_release(sb);
4016 ext4_release_system_zone(sb);
4017failed_mount4a:
94bf608a 4018 dput(sb->s_root);
32a9bb57 4019 sb->s_root = NULL;
94bf608a 4020failed_mount4:
b31e1552 4021 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4022 if (EXT4_SB(sb)->rsv_conversion_wq)
4023 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4024failed_mount_wq:
82939d79
JK		 4025 if (sbi->s_mb_cache) {
4026 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4027 sbi->s_mb_cache = NULL;
4028 }
0390131b
FM		 4029 if (sbi->s_journal) {
4030 jbd2_journal_destroy(sbi->s_journal);
4031 sbi->s_journal = NULL;
4032 }
50460fe8 4033failed_mount3a:
d3922a77 4034 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4035failed_mount3:
9105bb14 4036 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4037 if (sbi->s_mmp_tsk)
4038 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4039failed_mount2:
4040 for (i = 0; i < db_count; i++)
4041 brelse(sbi->s_group_desc[i]);
b93b41d4 4042 kvfree(sbi->s_group_desc);
ac27a0ec 4043failed_mount:
0441984a
DW		 4044 if (sbi->s_chksum_driver)
4045 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4046#ifdef CONFIG_QUOTA
a2d4a646 4047 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4048 kfree(sbi->s_qf_names[i]);
4049#endif
617ba13b 4050 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4051 brelse(bh);
4052out_fail:
4053 sb->s_fs_info = NULL;
f6830165 4054 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4055 kfree(sbi);
dcc7dae3 4056out_free_orig:
d4c402d9 4057 kfree(orig_data);
07aa2ea1 4058 return err ? err : ret;
ac27a0ec
DK		 4059}
4060
4061/*
4062 * Setup any per-fs journal parameters now. We'll do this both on
4063 * initial mount, once the journal has been initialised but before we've
4064 * done any recovery; and again on any subsequent remount.
4065 */
617ba13b 4066static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4067{
617ba13b 4068 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4069
30773840
TT		 4070 journal->j_commit_interval = sbi->s_commit_interval;
4071 journal->j_min_batch_time = sbi->s_min_batch_time;
4072 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4073
a931da6a 4074 write_lock(&journal->j_state_lock);
ac27a0ec 4075 if (test_opt(sb, BARRIER))
dab291af 4076 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4077 else
dab291af 4078 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4079 if (test_opt(sb, DATA_ERR_ABORT))
4080 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4081 else
4082 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4083 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4084}
4085
617ba13b 4086static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4087 unsigned int journal_inum)
4088{
4089 struct inode *journal_inode;
4090 journal_t *journal;
4091
e2b911c5 4092 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4093
ac27a0ec
DK		 4094 /* First, test for the existence of a valid inode on disk. Bad
4095 * things happen if we iget() an unused inode, as the subsequent
4096 * iput() will try to delete it. */
4097
1d1fe1ee
DH		 4098 journal_inode = ext4_iget(sb, journal_inum);
4099 if (IS_ERR(journal_inode)) {
b31e1552 4100 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4101 return NULL;
4102 }
4103 if (!journal_inode->i_nlink) {
4104 make_bad_inode(journal_inode);
4105 iput(journal_inode);
b31e1552 4106 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4107 return NULL;
4108 }
4109
e5f8eab8 4110 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4111 journal_inode, journal_inode->i_size);
1d1fe1ee 4112 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4113 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4114 iput(journal_inode);
4115 return NULL;
4116 }
4117
dab291af 4118 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4119 if (!journal) {
b31e1552 4120 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4121 iput(journal_inode);
4122 return NULL;
4123 }
4124 journal->j_private = sb;
617ba13b 4125 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4126 return journal;
4127}
4128
617ba13b 4129static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4130 dev_t j_dev)
4131{
2b2d6d01 4132 struct buffer_head *bh;
ac27a0ec 4133 journal_t *journal;
617ba13b
MC		 4134 ext4_fsblk_t start;
4135 ext4_fsblk_t len;
ac27a0ec 4136 int hblock, blocksize;
617ba13b 4137 ext4_fsblk_t sb_block;
ac27a0ec 4138 unsigned long offset;
2b2d6d01 4139 struct ext4_super_block *es;
ac27a0ec
DK		 4140 struct block_device *bdev;
4141
e2b911c5 4142 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4143
b31e1552 4144 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4145 if (bdev == NULL)
4146 return NULL;
4147
ac27a0ec 4148 blocksize = sb->s_blocksize;
e1defc4f 4149 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4150 if (blocksize < hblock) {
b31e1552
ES		 4151 ext4_msg(sb, KERN_ERR,
4152 "blocksize too small for journal device");
ac27a0ec
DK		 4153 goto out_bdev;
4154 }
4155
617ba13b
MC		 4156 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4157 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4158 set_blocksize(bdev, blocksize);
4159 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4160 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4161 "external journal");
ac27a0ec
DK		 4162 goto out_bdev;
4163 }
4164
2716b802 4165 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4166 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4167 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4168 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4169 ext4_msg(sb, KERN_ERR, "external journal has "
4170 "bad superblock");
ac27a0ec
DK		 4171 brelse(bh);
4172 goto out_bdev;
4173 }
4174
df4763be
DW		 4175 if ((le32_to_cpu(es->s_feature_ro_compat) &
4176 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4177 es->s_checksum != ext4_superblock_csum(sb, es)) {
4178 ext4_msg(sb, KERN_ERR, "external journal has "
4179 "corrupt superblock");
4180 brelse(bh);
4181 goto out_bdev;
4182 }
4183
617ba13b 4184 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4185 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4186 brelse(bh);
4187 goto out_bdev;
4188 }
4189
bd81d8ee 4190 len = ext4_blocks_count(es);
ac27a0ec
DK		 4191 start = sb_block + 1;
4192 brelse(bh); /* we're done with the superblock */
4193
dab291af 4194 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4195 start, len, blocksize);
4196 if (!journal) {
b31e1552 4197 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4198 goto out_bdev;
4199 }
4200 journal->j_private = sb;
9f203507 4201 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4202 wait_on_buffer(journal->j_sb_buffer);
4203 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4204 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4205 goto out_journal;
4206 }
4207 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4208 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4209 "user (unsupported) - %d",
ac27a0ec
DK		 4210 be32_to_cpu(journal->j_superblock->s_nr_users));
4211 goto out_journal;
4212 }
617ba13b
MC		 4213 EXT4_SB(sb)->journal_bdev = bdev;
4214 ext4_init_journal_params(sb, journal);
ac27a0ec 4215 return journal;
0b8e58a1 4216
ac27a0ec 4217out_journal:
dab291af 4218 jbd2_journal_destroy(journal);
ac27a0ec 4219out_bdev:
617ba13b 4220 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4221 return NULL;
4222}
4223
617ba13b
MC		 4224static int ext4_load_journal(struct super_block *sb,
4225 struct ext4_super_block *es,
ac27a0ec
DK		 4226 unsigned long journal_devnum)
4227{
4228 journal_t *journal;
4229 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4230 dev_t journal_dev;
4231 int err = 0;
4232 int really_read_only;
4233
e2b911c5 4234 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4235
ac27a0ec
DK		 4236 if (journal_devnum &&
4237 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4238 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4239 "numbers have changed");
ac27a0ec
DK		 4240 journal_dev = new_decode_dev(journal_devnum);
4241 } else
4242 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4243
4244 really_read_only = bdev_read_only(sb->s_bdev);
4245
4246 /*
4247 * Are we loading a blank journal or performing recovery after a
4248 * crash? For recovery, we need to check in advance whether we
4249 * can get read-write access to the device.
4250 */
e2b911c5 4251 if (ext4_has_feature_journal_needs_recovery(sb)) {
ac27a0ec 4252 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4253 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4254 "required on readonly filesystem");
ac27a0ec 4255 if (really_read_only) {
b31e1552
ES		 4256 ext4_msg(sb, KERN_ERR, "write access "
4257 "unavailable, cannot proceed");
ac27a0ec
DK		 4258 return -EROFS;
4259 }
b31e1552
ES		 4260 ext4_msg(sb, KERN_INFO, "write access will "
4261 "be enabled during recovery");
ac27a0ec
DK		 4262 }
4263 }
4264
4265 if (journal_inum && journal_dev) {
b31e1552
ES		 4266 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4267 "and inode journals!");
ac27a0ec
DK		 4268 return -EINVAL;
4269 }
4270
4271 if (journal_inum) {
617ba13b 4272 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4273 return -EINVAL;
4274 } else {
617ba13b 4275 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4276 return -EINVAL;
4277 }
4278
90576c0b 4279 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4280 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4281
e2b911c5 4282 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4283 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4284 if (!err) {
4285 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4286 if (save)
4287 memcpy(save, ((char *) es) +
4288 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4289 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4290 if (save)
4291 memcpy(((char *) es) + EXT4_S_ERR_START,
4292 save, EXT4_S_ERR_LEN);
4293 kfree(save);
4294 }
ac27a0ec
DK		 4295
4296 if (err) {
b31e1552 4297 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4298 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4299 return err;
4300 }
4301
617ba13b
MC		 4302 EXT4_SB(sb)->s_journal = journal;
4303 ext4_clear_journal_err(sb, es);
ac27a0ec 4304
c41303ce 4305 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4306 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4307 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4308
4309 /* Make sure we flush the recovery flag to disk. */
e2d67052 4310 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4311 }
4312
4313 return 0;
4314}
4315
e2d67052 4316static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4317{
e2d67052 4318 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4319 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4320 int error = 0;
ac27a0ec 4321
bdfe0cbd 4322 if (!sbh || block_device_ejected(sb))
c4be0c1d 4323 return error;
914258bf
TT		 4324 if (buffer_write_io_error(sbh)) {
4325 /*
4326 * Oh, dear. A previous attempt to write the
4327 * superblock failed. This could happen because the
4328 * USB device was yanked out. Or it could happen to
4329 * be a transient write error and maybe the block will
4330 * be remapped. Nothing we can do but to retry the
4331 * write and hope for the best.
4332 */
b31e1552
ES		 4333 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4334 "superblock detected");
914258bf
TT		 4335 clear_buffer_write_io_error(sbh);
4336 set_buffer_uptodate(sbh);
4337 }
71290b36
TT		 4338 /*
4339 * If the file system is mounted read-only, don't update the
4340 * superblock write time. This avoids updating the superblock
4341 * write time when we are mounting the root file system
4342 * read/only but we need to replay the journal; at that point,
4343 * for people who are east of GMT and who make their clock
4344 * tick in localtime for Windows bug-for-bug compatibility,
4345 * the clock is set in the future, and this will cause e2fsck
4346 * to complain and force a full file system check.
4347 */
4348 if (!(sb->s_flags & MS_RDONLY))
4349 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4350 if (sb->s_bdev->bd_part)
4351 es->s_kbytes_written =
4352 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4353 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4354 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4355 else
4356 es->s_kbytes_written =
4357 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4358 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4359 ext4_free_blocks_count_set(es,
57042651
TT		 4360 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4361 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4362 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4363 es->s_free_inodes_count =
4364 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4365 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4366 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4367 ext4_superblock_csum_set(sb);
ac27a0ec 4368 mark_buffer_dirty(sbh);
914258bf 4369 if (sync) {
564bc402
DJ		 4370 error = __sync_dirty_buffer(sbh,
4371 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
c4be0c1d
TS		 4372 if (error)
4373 return error;
4374
4375 error = buffer_write_io_error(sbh);
4376 if (error) {
b31e1552
ES		 4377 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4378 "superblock");
914258bf
TT		 4379 clear_buffer_write_io_error(sbh);
4380 set_buffer_uptodate(sbh);
4381 }
4382 }
c4be0c1d 4383 return error;
ac27a0ec
DK		 4384}
4385
ac27a0ec
DK		 4386/*
4387 * Have we just finished recovery? If so, and if we are mounting (or
4388 * remounting) the filesystem readonly, then we will end up with a
4389 * consistent fs on disk. Record that fact.
4390 */
2b2d6d01
TT		 4391static void ext4_mark_recovery_complete(struct super_block *sb,
4392 struct ext4_super_block *es)
ac27a0ec 4393{
617ba13b 4394 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4395
e2b911c5 4396 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4397 BUG_ON(journal != NULL);
4398 return;
4399 }
dab291af 4400 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4401 if (jbd2_journal_flush(journal) < 0)
4402 goto out;
4403
e2b911c5 4404 if (ext4_has_feature_journal_needs_recovery(sb) &&
ac27a0ec 4405 sb->s_flags & MS_RDONLY) {
e2b911c5 4406 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4407 ext4_commit_super(sb, 1);
ac27a0ec 4408 }
7ffe1ea8
HK		 4409
4410out:
dab291af 4411 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4412}
4413
4414/*
4415 * If we are mounting (or read-write remounting) a filesystem whose journal
4416 * has recorded an error from a previous lifetime, move that error to the
4417 * main filesystem now.
4418 */
2b2d6d01
TT		 4419static void ext4_clear_journal_err(struct super_block *sb,
4420 struct ext4_super_block *es)
ac27a0ec
DK		 4421{
4422 journal_t *journal;
4423 int j_errno;
4424 const char *errstr;
4425
e2b911c5 4426 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4427
617ba13b 4428 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4429
4430 /*
4431 * Now check for any error status which may have been recorded in the
617ba13b 4432 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4433 */
4434
dab291af 4435 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4436 if (j_errno) {
4437 char nbuf[16];
4438
617ba13b 4439 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4440 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4441 "from previous mount: %s", errstr);
12062ddd 4442 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4443
617ba13b
MC		 4444 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4445 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4446 ext4_commit_super(sb, 1);
ac27a0ec 4447
dab291af 4448 jbd2_journal_clear_err(journal);
d796c52e 4449 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4450 }
4451}
4452
4453/*
4454 * Force the running and committing transactions to commit,
4455 * and wait on the commit.
4456 */
617ba13b 4457int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4458{
4459 journal_t *journal;
ac27a0ec
DK		 4460
4461 if (sb->s_flags & MS_RDONLY)
4462 return 0;
4463
617ba13b 4464 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4465 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4466}
4467
617ba13b 4468static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4469{
14ce0cb4 4470 int ret = 0;
9eddacf9 4471 tid_t target;
06a407f1 4472 bool needs_barrier = false;
8d5d02e6 4473 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4474
9bffad1e 4475 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4476 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4477 /*
4478 * Writeback quota in non-journalled quota case - journalled quota has
4479 * no dirty dquots
4480 */
4481 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4482 /*
4483 * Data writeback is possible w/o journal transaction, so barrier must
4484 * being sent at the end of the function. But we can skip it if
4485 * transaction_commit will do it for us.
4486 */
bda32530
TT		 4487 if (sbi->s_journal) {
4488 target = jbd2_get_latest_transaction(sbi->s_journal);
4489 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4490 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4491 needs_barrier = true;
4492
4493 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4494 if (wait)
4495 ret = jbd2_log_wait_commit(sbi->s_journal,
4496 target);
4497 }
4498 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4499 needs_barrier = true;
06a407f1
DM		 4500 if (needs_barrier) {
4501 int err;
4502 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4503 if (!ret)
4504 ret = err;
0390131b 4505 }
06a407f1
DM		 4506
4507 return ret;
4508}
4509
ac27a0ec
DK		 4510/*
4511 * LVM calls this function before a (read-only) snapshot is created. This
4512 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4513 *
4514 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4515 * state independently. It relies on upper layer to stop all data & metadata
4516 * modifications.
ac27a0ec 4517 */
c4be0c1d 4518static int ext4_freeze(struct super_block *sb)
ac27a0ec 4519{
c4be0c1d
TS		 4520 int error = 0;
4521 journal_t *journal;
ac27a0ec 4522
9ca92389
TT		 4523 if (sb->s_flags & MS_RDONLY)
4524 return 0;
ac27a0ec 4525
9ca92389 4526 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4527
bb044576
TT		 4528 if (journal) {
4529 /* Now we set up the journal barrier. */
4530 jbd2_journal_lock_updates(journal);
ac27a0ec 4531
bb044576
TT		 4532 /*
4533 * Don't clear the needs_recovery flag if we failed to
4534 * flush the journal.
4535 */
4536 error = jbd2_journal_flush(journal);
4537 if (error < 0)
4538 goto out;
c642dc9e
ES		 4539
4540 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4541 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4542 }
9ca92389 4543
9ca92389 4544 error = ext4_commit_super(sb, 1);
6b0310fb 4545out:
bb044576
TT		 4546 if (journal)
4547 /* we rely on upper layer to stop further updates */
4548 jbd2_journal_unlock_updates(journal);
6b0310fb 4549 return error;
ac27a0ec
DK		 4550}
4551
4552/*
4553 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4554 * flag here, even though the filesystem is not technically dirty yet.
4555 */
c4be0c1d 4556static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4557{
9ca92389
TT		 4558 if (sb->s_flags & MS_RDONLY)
4559 return 0;
4560
c642dc9e
ES		 4561 if (EXT4_SB(sb)->s_journal) {
4562 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4563 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4564 }
4565
9ca92389 4566 ext4_commit_super(sb, 1);
c4be0c1d 4567 return 0;
ac27a0ec
DK		 4568}
4569
673c6100
TT		 4570/*
4571 * Structure to save mount options for ext4_remount's benefit
4572 */
4573struct ext4_mount_options {
4574 unsigned long s_mount_opt;
a2595b8a 4575 unsigned long s_mount_opt2;
08cefc7a
EB		 4576 kuid_t s_resuid;
4577 kgid_t s_resgid;
673c6100
TT		 4578 unsigned long s_commit_interval;
4579 u32 s_min_batch_time, s_max_batch_time;
4580#ifdef CONFIG_QUOTA
4581 int s_jquota_fmt;
a2d4a646 4582 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4583#endif
4584};
4585
2b2d6d01 4586static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4587{
2b2d6d01 4588 struct ext4_super_block *es;
617ba13b 4589 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4590 unsigned long old_sb_flags;
617ba13b 4591 struct ext4_mount_options old_opts;
c79d967d 4592 int enable_quota = 0;
8a266467 4593 ext4_group_t g;
b3881f74 4594 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4595 int err = 0;
ac27a0ec 4596#ifdef CONFIG_QUOTA
03dafb5f 4597 int i, j;
ac27a0ec 4598#endif
d4c402d9 4599 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4600
4601 /* Store the original options */
4602 old_sb_flags = sb->s_flags;
4603 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4604 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4605 old_opts.s_resuid = sbi->s_resuid;
4606 old_opts.s_resgid = sbi->s_resgid;
4607 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4608 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4609 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4610#ifdef CONFIG_QUOTA
4611 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4612 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4613 if (sbi->s_qf_names[i]) {
4614 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4615 GFP_KERNEL);
4616 if (!old_opts.s_qf_names[i]) {
4617 for (j = 0; j < i; j++)
4618 kfree(old_opts.s_qf_names[j]);
3e36a163 4619 kfree(orig_data);
03dafb5f
CG		 4620 return -ENOMEM;
4621 }
4622 } else
4623 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4624#endif
b3881f74
TT		 4625 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4626 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4627
661aa520 4628 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4629 err = -EINVAL;
4630 goto restore_opts;
4631 }
4632
6b992ff2 4633 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4634 test_opt(sb, JOURNAL_CHECKSUM)) {
4635 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4636 "during remount not supported; ignoring");
4637 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4638 }
4639
6ae6514b
PS		 4640 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4641 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4642 ext4_msg(sb, KERN_ERR, "can't mount with "
4643 "both data=journal and delalloc");
4644 err = -EINVAL;
4645 goto restore_opts;
4646 }
4647 if (test_opt(sb, DIOREAD_NOLOCK)) {
4648 ext4_msg(sb, KERN_ERR, "can't mount with "
4649 "both data=journal and dioread_nolock");
4650 err = -EINVAL;
4651 goto restore_opts;
4652 }
923ae0ff
RZ		 4653 if (test_opt(sb, DAX)) {
4654 ext4_msg(sb, KERN_ERR, "can't mount with "
4655 "both data=journal and dax");
4656 err = -EINVAL;
4657 goto restore_opts;
4658 }
4659 }
4660
4661 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4662 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4663 "dax flag with busy inodes while remounting");
4664 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4665 }
4666
4ab2f15b 4667 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4668 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4669
4670 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4671 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4672
4673 es = sbi->s_es;
4674
b3881f74 4675 if (sbi->s_journal) {
0390131b 4676 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4677 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4678 }
ac27a0ec 4679
a2fd66d0
TT		 4680 if (*flags & MS_LAZYTIME)
4681 sb->s_flags |= MS_LAZYTIME;
4682
661aa520 4683 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4684 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4685 err = -EROFS;
4686 goto restore_opts;
4687 }
4688
4689 if (*flags & MS_RDONLY) {
38c03b34
TT		 4690 err = sync_filesystem(sb);
4691 if (err < 0)
4692 goto restore_opts;
0f0dd62f
CH		 4693 err = dquot_suspend(sb, -1);
4694 if (err < 0)
c79d967d 4695 goto restore_opts;
c79d967d 4696
ac27a0ec
DK		 4697 /*
4698 * First of all, the unconditional stuff we have to do
4699 * to disable replay of the journal when we next remount
4700 */
4701 sb->s_flags |= MS_RDONLY;
4702
4703 /*
4704 * OK, test if we are remounting a valid rw partition
4705 * readonly, and if so set the rdonly flag and then
4706 * mark the partition as valid again.
4707 */
617ba13b
MC		 4708 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4709 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4710 es->s_state = cpu_to_le16(sbi->s_mount_state);
4711
a63c9eb2 4712 if (sbi->s_journal)
0390131b 4713 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4714 } else {
a13fb1a4 4715 /* Make sure we can mount this feature set readwrite */
e2b911c5 4716 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 4717 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4718 err = -EROFS;
4719 goto restore_opts;
4720 }
8a266467
TT		 4721 /*
4722 * Make sure the group descriptor checksums
0b8e58a1 4723 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4724 */
4725 for (g = 0; g < sbi->s_groups_count; g++) {
4726 struct ext4_group_desc *gdp =
4727 ext4_get_group_desc(sb, g, NULL);
4728
feb0ab32 4729 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4730 ext4_msg(sb, KERN_ERR,
4731 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 4732 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 4733 le16_to_cpu(gdp->bg_checksum));
6a797d27 4734 err = -EFSBADCRC;
8a266467
TT		 4735 goto restore_opts;
4736 }
4737 }
4738
ead6596b
ES		 4739 /*
4740 * If we have an unprocessed orphan list hanging
4741 * around from a previously readonly bdev mount,
4742 * require a full umount/remount for now.
4743 */
4744 if (es->s_last_orphan) {
b31e1552 4745 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4746 "remount RDWR because of unprocessed "
4747 "orphan inode list. Please "
b31e1552 4748 "umount/remount instead");
ead6596b
ES		 4749 err = -EINVAL;
4750 goto restore_opts;
4751 }
4752
ac27a0ec
DK		 4753 /*
4754 * Mounting a RDONLY partition read-write, so reread
4755 * and store the current valid flag. (It may have
4756 * been changed by e2fsck since we originally mounted
4757 * the partition.)
4758 */
0390131b
FM		 4759 if (sbi->s_journal)
4760 ext4_clear_journal_err(sb, es);
ac27a0ec 4761 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4762 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4763 sb->s_flags &= ~MS_RDONLY;
e2b911c5 4764 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 4765 if (ext4_multi_mount_protect(sb,
4766 le64_to_cpu(es->s_mmp_block))) {
4767 err = -EROFS;
4768 goto restore_opts;
4769 }
c79d967d 4770 enable_quota = 1;
ac27a0ec
DK		 4771 }
4772 }
bfff6873
LC		 4773
4774 /*
4775 * Reinitialize lazy itable initialization thread based on
4776 * current settings
4777 */
4778 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4779 ext4_unregister_li_request(sb);
4780 else {
4781 ext4_group_t first_not_zeroed;
4782 first_not_zeroed = ext4_has_uninit_itable(sb);
4783 ext4_register_li_request(sb, first_not_zeroed);
4784 }
4785
6fd058f7 4786 ext4_setup_system_zone(sb);
d096ad0f 4787 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4788 ext4_commit_super(sb, 1);
0390131b 4789
ac27a0ec
DK		 4790#ifdef CONFIG_QUOTA
4791 /* Release old quota file names */
a2d4a646 4792 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 4793 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4794 if (enable_quota) {
4795 if (sb_any_quota_suspended(sb))
4796 dquot_resume(sb, -1);
e2b911c5 4797 else if (ext4_has_feature_quota(sb)) {
7c319d32 4798 err = ext4_enable_quotas(sb);
07724f98 4799 if (err)
7c319d32 4800 goto restore_opts;
7c319d32
AK		 4801 }
4802 }
ac27a0ec 4803#endif
d4c402d9 4804
a26f4992 4805 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 4806 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4807 kfree(orig_data);
ac27a0ec 4808 return 0;
0b8e58a1 4809
ac27a0ec
DK		 4810restore_opts:
4811 sb->s_flags = old_sb_flags;
4812 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4813 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4814 sbi->s_resuid = old_opts.s_resuid;
4815 sbi->s_resgid = old_opts.s_resgid;
4816 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4817 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4818 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4819#ifdef CONFIG_QUOTA
4820 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 4821 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 4822 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4823 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4824 }
4825#endif
d4c402d9 4826 kfree(orig_data);
ac27a0ec
DK		 4827 return err;
4828}
4829
689c958c
LX		 4830#ifdef CONFIG_QUOTA
4831static int ext4_statfs_project(struct super_block *sb,
4832 kprojid_t projid, struct kstatfs *buf)
4833{
4834 struct kqid qid;
4835 struct dquot *dquot;
4836 u64 limit;
4837 u64 curblock;
4838
4839 qid = make_kqid_projid(projid);
4840 dquot = dqget(sb, qid);
4841 if (IS_ERR(dquot))
4842 return PTR_ERR(dquot);
4843 spin_lock(&dq_data_lock);
4844
4845 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
4846 dquot->dq_dqb.dqb_bsoftlimit :
4847 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
4848 if (limit && buf->f_blocks > limit) {
4849 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
4850 buf->f_blocks = limit;
4851 buf->f_bfree = buf->f_bavail =
4852 (buf->f_blocks > curblock) ?
4853 (buf->f_blocks - curblock) : 0;
4854 }
4855
4856 limit = dquot->dq_dqb.dqb_isoftlimit ?
4857 dquot->dq_dqb.dqb_isoftlimit :
4858 dquot->dq_dqb.dqb_ihardlimit;
4859 if (limit && buf->f_files > limit) {
4860 buf->f_files = limit;
4861 buf->f_ffree =
4862 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
4863 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
4864 }
4865
4866 spin_unlock(&dq_data_lock);
4867 dqput(dquot);
4868 return 0;
4869}
4870#endif
4871
2b2d6d01 4872static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4873{
4874 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4875 struct ext4_sb_info *sbi = EXT4_SB(sb);
4876 struct ext4_super_block *es = sbi->s_es;
27dd4385 4877 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 4878 u64 fsid;
d02a9391 4879 s64 bfree;
27dd4385 4880 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 4881
952fc18e
TT		 4882 if (!test_opt(sb, MINIX_DF))
4883 overhead = sbi->s_overhead;
ac27a0ec 4884
617ba13b 4885 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4886 buf->f_bsize = sb->s_blocksize;
b72f78cb 4887 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4888 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4889 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4890 /* prevent underflow in case that few free space is available */
57042651 4891 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 4892 buf->f_bavail = buf->f_bfree -
4893 (ext4_r_blocks_count(es) + resv_blocks);
4894 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 4895 buf->f_bavail = 0;
4896 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4897 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4898 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4899 fsid = le64_to_cpup((void *)es->s_uuid) ^
4900 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4901 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4902 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4903
689c958c
LX		 4904#ifdef CONFIG_QUOTA
4905 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
4906 sb_has_quota_limits_enabled(sb, PRJQUOTA))
4907 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
4908#endif
ac27a0ec
DK		 4909 return 0;
4910}
4911
0b8e58a1
AD		 4912/* Helper function for writing quotas on sync - we need to start transaction
4913 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4914 * Process 1 Process 2
617ba13b 4915 * ext4_create() quota_sync()
a269eb18 4916 * jbd2_journal_start() write_dquot()
871a2931 4917 * dquot_initialize() down(dqio_mutex)
dab291af 4918 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4919 *
4920 */
4921
4922#ifdef CONFIG_QUOTA
4923
4924static inline struct inode *dquot_to_inode(struct dquot *dquot)
4925{
4c376dca 4926 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4927}
4928
617ba13b 4929static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4930{
4931 int ret, err;
4932 handle_t *handle;
4933 struct inode *inode;
4934
4935 inode = dquot_to_inode(dquot);
9924a92a 4936 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 4937 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4938 if (IS_ERR(handle))
4939 return PTR_ERR(handle);
4940 ret = dquot_commit(dquot);
617ba13b 4941 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4942 if (!ret)
4943 ret = err;
4944 return ret;
4945}
4946
617ba13b 4947static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4948{
4949 int ret, err;
4950 handle_t *handle;
4951
9924a92a 4952 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4953 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4954 if (IS_ERR(handle))
4955 return PTR_ERR(handle);
4956 ret = dquot_acquire(dquot);
617ba13b 4957 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4958 if (!ret)
4959 ret = err;
4960 return ret;
4961}
4962
617ba13b 4963static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4964{
4965 int ret, err;
4966 handle_t *handle;
4967
9924a92a 4968 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4969 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4970 if (IS_ERR(handle)) {
4971 /* Release dquot anyway to avoid endless cycle in dqput() */
4972 dquot_release(dquot);
ac27a0ec 4973 return PTR_ERR(handle);
9c3013e9 4974 }
ac27a0ec 4975 ret = dquot_release(dquot);
617ba13b 4976 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4977 if (!ret)
4978 ret = err;
4979 return ret;
4980}
4981
617ba13b 4982static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4983{
262b4662
JK		 4984 struct super_block *sb = dquot->dq_sb;
4985 struct ext4_sb_info *sbi = EXT4_SB(sb);
4986
2c8be6b2 4987 /* Are we journaling quotas? */
e2b911c5 4988 if (ext4_has_feature_quota(sb) ||
262b4662 4989 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 4990 dquot_mark_dquot_dirty(dquot);
617ba13b 4991 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4992 } else {
4993 return dquot_mark_dquot_dirty(dquot);
4994 }
4995}
4996
617ba13b 4997static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4998{
4999 int ret, err;
5000 handle_t *handle;
5001
5002 /* Data block + inode block */
2b0143b5 5003 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5004 if (IS_ERR(handle))
5005 return PTR_ERR(handle);
5006 ret = dquot_commit_info(sb, type);
617ba13b 5007 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5008 if (!ret)
5009 ret = err;
5010 return ret;
5011}
5012
5013/*
5014 * Turn on quotas during mount time - we need to find
5015 * the quota file and such...
5016 */
617ba13b 5017static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5018{
287a8095
CH		 5019 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5020 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5021}
5022
5023/*
5024 * Standard function to be called on quota_on
5025 */
617ba13b 5026static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5027 struct path *path)
ac27a0ec
DK		 5028{
5029 int err;
ac27a0ec
DK		 5030
5031 if (!test_opt(sb, QUOTA))
5032 return -EINVAL;
0623543b 5033
ac27a0ec 5034 /* Quotafile not on the same filesystem? */
d8c9584e 5035 if (path->dentry->d_sb != sb)
ac27a0ec 5036 return -EXDEV;
0623543b
JK		 5037 /* Journaling quota? */
5038 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5039 /* Quotafile not in fs root? */
f00c9e44 5040 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5041 ext4_msg(sb, KERN_WARNING,
5042 "Quota file not on filesystem root. "
5043 "Journaled quota will not work");
2b2d6d01 5044 }
0623543b
JK		 5045
5046 /*
5047 * When we journal data on quota file, we have to flush journal to see
5048 * all updates to the file when we bypass pagecache...
5049 */
0390131b 5050 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5051 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5052 /*
5053 * We don't need to lock updates but journal_flush() could
5054 * otherwise be livelocked...
5055 */
5056 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5057 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5058 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5059 if (err)
7ffe1ea8 5060 return err;
0623543b
JK		 5061 }
5062
f00c9e44 5063 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5064}
5065
7c319d32
AK		 5066static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5067 unsigned int flags)
5068{
5069 int err;
5070 struct inode *qf_inode;
a2d4a646 5071 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5072 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5073 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5074 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5075 };
5076
e2b911c5 5077 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5078
5079 if (!qf_inums[type])
5080 return -EPERM;
5081
5082 qf_inode = ext4_iget(sb, qf_inums[type]);
5083 if (IS_ERR(qf_inode)) {
5084 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5085 return PTR_ERR(qf_inode);
5086 }
5087
bcb13850
JK		 5088 /* Don't account quota for quota files to avoid recursion */
5089 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5090 err = dquot_enable(qf_inode, type, format_id, flags);
5091 iput(qf_inode);
5092
5093 return err;
5094}
5095
5096/* Enable usage tracking for all quota types. */
5097static int ext4_enable_quotas(struct super_block *sb)
5098{
5099 int type, err = 0;
a2d4a646 5100 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5101 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5102 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5103 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5104 };
5105
5106 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5107 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5108 if (qf_inums[type]) {
5109 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5110 DQUOT_USAGE_ENABLED);
5111 if (err) {
5112 ext4_warning(sb,
72ba7450
TT		 5113 "Failed to enable quota tracking "
5114 "(type=%d, err=%d). Please run "
5115 "e2fsck to fix.", type, err);
7c319d32
AK		 5116 return err;
5117 }
5118 }
5119 }
5120 return 0;
5121}
5122
ca0e05e4
DM		 5123static int ext4_quota_off(struct super_block *sb, int type)
5124{
21f97697
JK		 5125 struct inode *inode = sb_dqopt(sb)->files[type];
5126 handle_t *handle;
5127
87009d86
DM		 5128 /* Force all delayed allocation blocks to be allocated.
5129 * Caller already holds s_umount sem */
5130 if (test_opt(sb, DELALLOC))
ca0e05e4 5131 sync_filesystem(sb);
ca0e05e4 5132
0b268590
AG		 5133 if (!inode)
5134 goto out;
5135
21f97697
JK		 5136 /* Update modification times of quota files when userspace can
5137 * start looking at them */
9924a92a 5138 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5139 if (IS_ERR(handle))
5140 goto out;
5141 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5142 ext4_mark_inode_dirty(handle, inode);
5143 ext4_journal_stop(handle);
5144
5145out:
ca0e05e4
DM		 5146 return dquot_quota_off(sb, type);
5147}
5148
ac27a0ec
DK		 5149/* Read data from quotafile - avoid pagecache and such because we cannot afford
5150 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5151 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5152 * we don't have to be afraid of races */
617ba13b 5153static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5154 size_t len, loff_t off)
5155{
5156 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5157 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5158 int offset = off & (sb->s_blocksize - 1);
5159 int tocopy;
5160 size_t toread;
5161 struct buffer_head *bh;
5162 loff_t i_size = i_size_read(inode);
5163
5164 if (off > i_size)
5165 return 0;
5166 if (off+len > i_size)
5167 len = i_size-off;
5168 toread = len;
5169 while (toread > 0) {
5170 tocopy = sb->s_blocksize - offset < toread ?
5171 sb->s_blocksize - offset : toread;
1c215028
TT		 5172 bh = ext4_bread(NULL, inode, blk, 0);
5173 if (IS_ERR(bh))
5174 return PTR_ERR(bh);
ac27a0ec
DK		 5175 if (!bh) /* A hole? */
5176 memset(data, 0, tocopy);
5177 else
5178 memcpy(data, bh->b_data+offset, tocopy);
5179 brelse(bh);
5180 offset = 0;
5181 toread -= tocopy;
5182 data += tocopy;
5183 blk++;
5184 }
5185 return len;
5186}
5187
5188/* Write to quotafile (we know the transaction is already started and has
5189 * enough credits) */
617ba13b 5190static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5191 const char *data, size_t len, loff_t off)
5192{
5193 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5194 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5195 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5196 int retries = 0;
ac27a0ec
DK		 5197 struct buffer_head *bh;
5198 handle_t *handle = journal_current_handle();
5199
0390131b 5200 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5201 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5202 " cancelled because transaction is not started",
9c3013e9
JK		 5203 (unsigned long long)off, (unsigned long long)len);
5204 return -EIO;
5205 }
67eeb568
DM		 5206 /*
5207 * Since we account only one data block in transaction credits,
5208 * then it is impossible to cross a block boundary.
5209 */
5210 if (sb->s_blocksize - offset < len) {
5211 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5212 " cancelled because not block aligned",
5213 (unsigned long long)off, (unsigned long long)len);
5214 return -EIO;
5215 }
5216
c5e298ae
TT		 5217 do {
5218 bh = ext4_bread(handle, inode, blk,
5219 EXT4_GET_BLOCKS_CREATE |
5220 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5221 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5222 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5223 if (IS_ERR(bh))
5224 return PTR_ERR(bh);
67eeb568
DM		 5225 if (!bh)
5226 goto out;
5d601255 5227 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5228 err = ext4_journal_get_write_access(handle, bh);
5229 if (err) {
5230 brelse(bh);
1c215028 5231 return err;
ac27a0ec 5232 }
67eeb568
DM		 5233 lock_buffer(bh);
5234 memcpy(bh->b_data+offset, data, len);
5235 flush_dcache_page(bh->b_page);
5236 unlock_buffer(bh);
62d2b5f2 5237 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5238 brelse(bh);
ac27a0ec 5239out:
67eeb568
DM		 5240 if (inode->i_size < off + len) {
5241 i_size_write(inode, off + len);
617ba13b 5242 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5243 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5244 }
67eeb568 5245 return len;
ac27a0ec
DK		 5246}
5247
5248#endif
5249
152a0836
AV		 5250static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5251 const char *dev_name, void *data)
ac27a0ec 5252{
152a0836 5253 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5254}
5255
c290ea01 5256#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5257static inline void register_as_ext2(void)
5258{
5259 int err = register_filesystem(&ext2_fs_type);
5260 if (err)
5261 printk(KERN_WARNING
5262 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5263}
5264
5265static inline void unregister_as_ext2(void)
5266{
5267 unregister_filesystem(&ext2_fs_type);
5268}
2035e776
TT		 5269
5270static inline int ext2_feature_set_ok(struct super_block *sb)
5271{
e2b911c5 5272 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776
TT		 5273 return 0;
5274 if (sb->s_flags & MS_RDONLY)
5275 return 1;
e2b911c5 5276 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5277 return 0;
5278 return 1;
5279}
24b58424
TT		 5280#else
5281static inline void register_as_ext2(void) { }
5282static inline void unregister_as_ext2(void) { }
2035e776 5283static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5284#endif
5285
24b58424
TT		 5286static inline void register_as_ext3(void)
5287{
5288 int err = register_filesystem(&ext3_fs_type);
5289 if (err)
5290 printk(KERN_WARNING
5291 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5292}
5293
5294static inline void unregister_as_ext3(void)
5295{
5296 unregister_filesystem(&ext3_fs_type);
5297}
2035e776
TT		 5298
5299static inline int ext3_feature_set_ok(struct super_block *sb)
5300{
e2b911c5 5301 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5302 return 0;
e2b911c5 5303 if (!ext4_has_feature_journal(sb))
2035e776
TT		 5304 return 0;
5305 if (sb->s_flags & MS_RDONLY)
5306 return 1;
e2b911c5 5307 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5308 return 0;
5309 return 1;
5310}
24b58424 5311
03010a33
TT		 5312static struct file_system_type ext4_fs_type = {
5313 .owner = THIS_MODULE,
5314 .name = "ext4",
152a0836 5315 .mount = ext4_mount,
03010a33
TT		 5316 .kill_sb = kill_block_super,
5317 .fs_flags = FS_REQUIRES_DEV,
5318};
7f78e035 5319MODULE_ALIAS_FS("ext4");
03010a33 5320
e9e3bcec
ES		 5321/* Shared across all ext4 file systems */
5322wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5323
5dabfc78 5324static int __init ext4_init_fs(void)
ac27a0ec 5325{
e9e3bcec 5326 int i, err;
c9de560d 5327
e294a537 5328 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5329 ext4_li_info = NULL;
5330 mutex_init(&ext4_li_mtx);
5331
9a4c8019 5332 /* Build-time check for flags consistency */
12e9b892 5333 ext4_check_flag_values();
e9e3bcec 5334
e142d052 5335 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5336 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5337
51865fda 5338 err = ext4_init_es();
6fd058f7
TT		 5339 if (err)
5340 return err;
51865fda
ZL		 5341
5342 err = ext4_init_pageio();
5343 if (err)
b5799018 5344 goto out5;
51865fda 5345
5dabfc78 5346 err = ext4_init_system_zone();
bd2d0210 5347 if (err)
b5799018 5348 goto out4;
857ac889 5349
b5799018 5350 err = ext4_init_sysfs();
dd68314c 5351 if (err)
b5799018 5352 goto out3;
857ac889 5353
5dabfc78 5354 err = ext4_init_mballoc();
c9de560d
AT		 5355 if (err)
5356 goto out2;
ac27a0ec
DK		 5357 err = init_inodecache();
5358 if (err)
5359 goto out1;
24b58424 5360 register_as_ext3();
2035e776 5361 register_as_ext2();
03010a33 5362 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5363 if (err)
5364 goto out;
bfff6873 5365
ac27a0ec
DK		 5366 return 0;
5367out:
24b58424
TT		 5368 unregister_as_ext2();
5369 unregister_as_ext3();
ac27a0ec
DK		 5370 destroy_inodecache();
5371out1:
5dabfc78 5372 ext4_exit_mballoc();
9c191f70 5373out2:
b5799018
TT		 5374 ext4_exit_sysfs();
5375out3:
5dabfc78 5376 ext4_exit_system_zone();
b5799018 5377out4:
5dabfc78 5378 ext4_exit_pageio();
b5799018 5379out5:
51865fda
ZL		 5380 ext4_exit_es();
5381
ac27a0ec
DK		 5382 return err;
5383}
5384
5dabfc78 5385static void __exit ext4_exit_fs(void)
ac27a0ec 5386{
e298e73b 5387 ext4_exit_crypto();
bfff6873 5388 ext4_destroy_lazyinit_thread();
24b58424
TT		 5389 unregister_as_ext2();
5390 unregister_as_ext3();
03010a33 5391 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5392 destroy_inodecache();
5dabfc78 5393 ext4_exit_mballoc();
b5799018 5394 ext4_exit_sysfs();
5dabfc78
TT		 5395 ext4_exit_system_zone();
5396 ext4_exit_pageio();
dd12ed14 5397 ext4_exit_es();
ac27a0ec
DK		 5398}
5399
5400MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5401MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5402MODULE_LICENSE("GPL");
5dabfc78
TT		 5403module_init(ext4_init_fs)
5404module_exit(ext4_exit_fs)
Empty description
This page took 1.938985 seconds and 4 git commands to generate.