projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: pack ioend structure better
[deliverable/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 (card@masi.ibp.fr)
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 (davem@caip.rutgers.edu), 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;
f7ad6d2e 945 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 946 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 947 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
b30ab0e0 948#ifdef CONFIG_EXT4_FS_ENCRYPTION
b7236e21 949 ei->i_crypt_info = NULL;
b30ab0e0 950#endif
ac27a0ec
DK		 951 return &ei->vfs_inode;
952}
953
7ff9c073
TT		 954static int ext4_drop_inode(struct inode *inode)
955{
956 int drop = generic_drop_inode(inode);
957
958 trace_ext4_drop_inode(inode, drop);
959 return drop;
960}
961
fa0d7e3d
NP		 962static void ext4_i_callback(struct rcu_head *head)
963{
964 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 965 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
966}
967
617ba13b 968static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 969{
9f7dd93d 970 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 971 ext4_msg(inode->i_sb, KERN_ERR,
972 "Inode %lu (%p): orphan list check failed!",
973 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 974 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
975 EXT4_I(inode), sizeof(struct ext4_inode_info),
976 true);
977 dump_stack();
978 }
fa0d7e3d 979 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 980}
981
51cc5068 982static void init_once(void *foo)
ac27a0ec 983{
617ba13b 984 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 985
a35afb83 986 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 987 init_rwsem(&ei->xattr_sem);
0e855ac8 988 init_rwsem(&ei->i_data_sem);
ea3d7209 989 init_rwsem(&ei->i_mmap_sem);
a35afb83 990 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 991}
992
e67bc2b3 993static int __init init_inodecache(void)
ac27a0ec 994{
617ba13b
MC		 995 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
996 sizeof(struct ext4_inode_info),
ac27a0ec 997 0, (SLAB_RECLAIM_ACCOUNT|
5d097056 998 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
20c2df83 999 init_once);
617ba13b 1000 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1001 return -ENOMEM;
1002 return 0;
1003}
1004
1005static void destroy_inodecache(void)
1006{
8c0a8537
KS		 1007 /*
1008 * Make sure all delayed rcu free inodes are flushed before we
1009 * destroy cache.
1010 */
1011 rcu_barrier();
617ba13b 1012 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1013}
1014
0930fcc1 1015void ext4_clear_inode(struct inode *inode)
ac27a0ec 1016{
0930fcc1 1017 invalidate_inode_buffers(inode);
dbd5768f 1018 clear_inode(inode);
9f754758 1019 dquot_drop(inode);
c2ea3fde 1020 ext4_discard_preallocations(inode);
51865fda 1021 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1022 if (EXT4_I(inode)->jinode) {
1023 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1024 EXT4_I(inode)->jinode);
1025 jbd2_free_inode(EXT4_I(inode)->jinode);
1026 EXT4_I(inode)->jinode = NULL;
1027 }
b7236e21
TT		 1028#ifdef CONFIG_EXT4_FS_ENCRYPTION
1029 if (EXT4_I(inode)->i_crypt_info)
c936e1ec 1030 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
b7236e21 1031#endif
ac27a0ec
DK		 1032}
1033
1b961ac0 1034static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1035 u64 ino, u32 generation)
ac27a0ec 1036{
ac27a0ec 1037 struct inode *inode;
ac27a0ec 1038
617ba13b 1039 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1040 return ERR_PTR(-ESTALE);
617ba13b 1041 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1042 return ERR_PTR(-ESTALE);
1043
1044 /* iget isn't really right if the inode is currently unallocated!!
1045 *
617ba13b 1046 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1047 * deleted, so we should be safe.
1048 *
1049 * Currently we don't know the generation for parent directory, so
1050 * a generation of 0 means "accept any"
1051 */
f4bb2981 1052 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1053 if (IS_ERR(inode))
1054 return ERR_CAST(inode);
1055 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1056 iput(inode);
1057 return ERR_PTR(-ESTALE);
1058 }
1b961ac0
CH		 1059
1060 return inode;
1061}
1062
1063static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1064 int fh_len, int fh_type)
1b961ac0
CH		 1065{
1066 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1067 ext4_nfs_get_inode);
1068}
1069
1070static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1071 int fh_len, int fh_type)
1b961ac0
CH		 1072{
1073 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1074 ext4_nfs_get_inode);
ac27a0ec
DK		 1075}
1076
c39a7f84
TO		 1077/*
1078 * Try to release metadata pages (indirect blocks, directories) which are
1079 * mapped via the block device. Since these pages could have journal heads
1080 * which would prevent try_to_free_buffers() from freeing them, we must use
1081 * jbd2 layer's try_to_free_buffers() function to release them.
1082 */
0b8e58a1
AD		 1083static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1084 gfp_t wait)
c39a7f84
TO		 1085{
1086 journal_t *journal = EXT4_SB(sb)->s_journal;
1087
1088 WARN_ON(PageChecked(page));
1089 if (!page_has_buffers(page))
1090 return 0;
1091 if (journal)
1092 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1093 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1094 return try_to_free_buffers(page);
1095}
1096
ac27a0ec 1097#ifdef CONFIG_QUOTA
689c958c
LX		 1098static char *quotatypes[] = INITQFNAMES;
1099#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1100
617ba13b
MC		 1101static int ext4_write_dquot(struct dquot *dquot);
1102static int ext4_acquire_dquot(struct dquot *dquot);
1103static int ext4_release_dquot(struct dquot *dquot);
1104static int ext4_mark_dquot_dirty(struct dquot *dquot);
1105static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1106static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1107 struct path *path);
ca0e05e4 1108static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1109static int ext4_quota_on_mount(struct super_block *sb, int type);
1110static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1111 size_t len, loff_t off);
617ba13b 1112static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1113 const char *data, size_t len, loff_t off);
7c319d32
AK		 1114static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1115 unsigned int flags);
1116static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1117
96c7e0d9
JK		 1118static struct dquot **ext4_get_dquots(struct inode *inode)
1119{
1120 return EXT4_I(inode)->i_dquot;
1121}
1122
61e225dc 1123static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1124 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1125 .write_dquot = ext4_write_dquot,
1126 .acquire_dquot = ext4_acquire_dquot,
1127 .release_dquot = ext4_release_dquot,
1128 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1129 .write_info = ext4_write_info,
1130 .alloc_dquot = dquot_alloc,
1131 .destroy_dquot = dquot_destroy,
040cb378 1132 .get_projid = ext4_get_projid,
ac27a0ec
DK		 1133};
1134
0d54b217 1135static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1136 .quota_on = ext4_quota_on,
ca0e05e4 1137 .quota_off = ext4_quota_off,
287a8095 1138 .quota_sync = dquot_quota_sync,
0a240339 1139 .get_state = dquot_get_state,
287a8095
CH		 1140 .set_info = dquot_set_dqinfo,
1141 .get_dqblk = dquot_get_dqblk,
1142 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1143};
1144#endif
1145
ee9b6d61 1146static const struct super_operations ext4_sops = {
617ba13b
MC		 1147 .alloc_inode = ext4_alloc_inode,
1148 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1149 .write_inode = ext4_write_inode,
1150 .dirty_inode = ext4_dirty_inode,
7ff9c073 1151 .drop_inode = ext4_drop_inode,
0930fcc1 1152 .evict_inode = ext4_evict_inode,
617ba13b 1153 .put_super = ext4_put_super,
617ba13b 1154 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1155 .freeze_fs = ext4_freeze,
1156 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1157 .statfs = ext4_statfs,
1158 .remount_fs = ext4_remount,
617ba13b 1159 .show_options = ext4_show_options,
ac27a0ec 1160#ifdef CONFIG_QUOTA
617ba13b
MC		 1161 .quota_read = ext4_quota_read,
1162 .quota_write = ext4_quota_write,
96c7e0d9 1163 .get_dquots = ext4_get_dquots,
ac27a0ec 1164#endif
c39a7f84 1165 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1166};
1167
39655164 1168static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1169 .fh_to_dentry = ext4_fh_to_dentry,
1170 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1171 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1172};
1173
1174enum {
1175 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1176 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1177 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1178 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1179 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1180 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1181 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1182 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1183 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1184 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1185 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1186 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
923ae0ff 1187 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1449032b 1188 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1189 Opt_lazytime, Opt_nolazytime,
1449032b 1190 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1191 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1192 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1193 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1194 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1195};
1196
a447c093 1197static const match_table_t tokens = {
ac27a0ec
DK		 1198 {Opt_bsd_df, "bsddf"},
1199 {Opt_minix_df, "minixdf"},
1200 {Opt_grpid, "grpid"},
1201 {Opt_grpid, "bsdgroups"},
1202 {Opt_nogrpid, "nogrpid"},
1203 {Opt_nogrpid, "sysvgroups"},
1204 {Opt_resgid, "resgid=%u"},
1205 {Opt_resuid, "resuid=%u"},
1206 {Opt_sb, "sb=%u"},
1207 {Opt_err_cont, "errors=continue"},
1208 {Opt_err_panic, "errors=panic"},
1209 {Opt_err_ro, "errors=remount-ro"},
1210 {Opt_nouid32, "nouid32"},
ac27a0ec 1211 {Opt_debug, "debug"},
72578c33
TT		 1212 {Opt_removed, "oldalloc"},
1213 {Opt_removed, "orlov"},
ac27a0ec
DK		 1214 {Opt_user_xattr, "user_xattr"},
1215 {Opt_nouser_xattr, "nouser_xattr"},
1216 {Opt_acl, "acl"},
1217 {Opt_noacl, "noacl"},
e3bb52ae 1218 {Opt_noload, "norecovery"},
5a916be1 1219 {Opt_noload, "noload"},
72578c33
TT		 1220 {Opt_removed, "nobh"},
1221 {Opt_removed, "bh"},
ac27a0ec 1222 {Opt_commit, "commit=%u"},
30773840
TT		 1223 {Opt_min_batch_time, "min_batch_time=%u"},
1224 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1225 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1226 {Opt_journal_path, "journal_path=%s"},
818d276c 1227 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1228 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1229 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1230 {Opt_abort, "abort"},
1231 {Opt_data_journal, "data=journal"},
1232 {Opt_data_ordered, "data=ordered"},
1233 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1234 {Opt_data_err_abort, "data_err=abort"},
1235 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1236 {Opt_offusrjquota, "usrjquota="},
1237 {Opt_usrjquota, "usrjquota=%s"},
1238 {Opt_offgrpjquota, "grpjquota="},
1239 {Opt_grpjquota, "grpjquota=%s"},
1240 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1241 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1242 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1243 {Opt_grpquota, "grpquota"},
1244 {Opt_noquota, "noquota"},
1245 {Opt_quota, "quota"},
1246 {Opt_usrquota, "usrquota"},
1247 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1248 {Opt_barrier, "barrier"},
1249 {Opt_nobarrier, "nobarrier"},
25ec56b5 1250 {Opt_i_version, "i_version"},
923ae0ff 1251 {Opt_dax, "dax"},
c9de560d 1252 {Opt_stripe, "stripe=%u"},
64769240 1253 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1254 {Opt_lazytime, "lazytime"},
1255 {Opt_nolazytime, "nolazytime"},
dd919b98 1256 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1257 {Opt_removed, "mblk_io_submit"},
1258 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1259 {Opt_block_validity, "block_validity"},
1260 {Opt_noblock_validity, "noblock_validity"},
240799cd 1261 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1262 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1263 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1264 {Opt_auto_da_alloc, "auto_da_alloc"},
1265 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1266 {Opt_dioread_nolock, "dioread_nolock"},
1267 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1268 {Opt_discard, "discard"},
1269 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1270 {Opt_init_itable, "init_itable=%u"},
1271 {Opt_init_itable, "init_itable"},
1272 {Opt_noinit_itable, "noinit_itable"},
df981d03 1273 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1274 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1275 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1276 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1277 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1278 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1279 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1280 {Opt_err, NULL},
ac27a0ec
DK		 1281};
1282
617ba13b 1283static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1284{
617ba13b 1285 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1286 char *options = (char *) *data;
1287
1288 if (!options || strncmp(options, "sb=", 3) != 0)
1289 return 1; /* Default location */
0b8e58a1 1290
ac27a0ec 1291 options += 3;
0b8e58a1 1292 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1293 sb_block = simple_strtoul(options, &options, 0);
1294 if (*options && *options != ',') {
4776004f 1295 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1296 (char *) *data);
1297 return 1;
1298 }
1299 if (*options == ',')
1300 options++;
1301 *data = (void *) options;
0b8e58a1 1302
ac27a0ec
DK		 1303 return sb_block;
1304}
1305
b3881f74 1306#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1307static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1308 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1309
56c50f11
DM		 1310#ifdef CONFIG_QUOTA
1311static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1312{
1313 struct ext4_sb_info *sbi = EXT4_SB(sb);
1314 char *qname;
03dafb5f 1315 int ret = -1;
56c50f11
DM		 1316
1317 if (sb_any_quota_loaded(sb) &&
1318 !sbi->s_qf_names[qtype]) {
1319 ext4_msg(sb, KERN_ERR,
1320 "Cannot change journaled "
1321 "quota options when quota turned on");
57f73c2c 1322 return -1;
56c50f11 1323 }
e2b911c5 1324 if (ext4_has_feature_quota(sb)) {
262b4662
JK		 1325 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1326 "when QUOTA feature is enabled");
1327 return -1;
1328 }
56c50f11
DM		 1329 qname = match_strdup(args);
1330 if (!qname) {
1331 ext4_msg(sb, KERN_ERR,
1332 "Not enough memory for storing quotafile name");
57f73c2c 1333 return -1;
56c50f11 1334 }
03dafb5f
CG		 1335 if (sbi->s_qf_names[qtype]) {
1336 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1337 ret = 1;
1338 else
1339 ext4_msg(sb, KERN_ERR,
1340 "%s quota file already specified",
1341 QTYPE2NAME(qtype));
1342 goto errout;
56c50f11 1343 }
03dafb5f 1344 if (strchr(qname, '/')) {
56c50f11
DM		 1345 ext4_msg(sb, KERN_ERR,
1346 "quotafile must be on filesystem root");
03dafb5f 1347 goto errout;
56c50f11 1348 }
03dafb5f 1349 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1350 set_opt(sb, QUOTA);
56c50f11 1351 return 1;
03dafb5f
CG		 1352errout:
1353 kfree(qname);
1354 return ret;
56c50f11
DM		 1355}
1356
1357static int clear_qf_name(struct super_block *sb, int qtype)
1358{
1359
1360 struct ext4_sb_info *sbi = EXT4_SB(sb);
1361
1362 if (sb_any_quota_loaded(sb) &&
1363 sbi->s_qf_names[qtype]) {
1364 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1365 " when quota turned on");
57f73c2c 1366 return -1;
56c50f11 1367 }
03dafb5f 1368 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1369 sbi->s_qf_names[qtype] = NULL;
1370 return 1;
1371}
1372#endif
1373
26092bf5
TT		 1374#define MOPT_SET 0x0001
1375#define MOPT_CLEAR 0x0002
1376#define MOPT_NOSUPPORT 0x0004
1377#define MOPT_EXPLICIT 0x0008
1378#define MOPT_CLEAR_ERR 0x0010
1379#define MOPT_GTE0 0x0020
ac27a0ec 1380#ifdef CONFIG_QUOTA
26092bf5
TT		 1381#define MOPT_Q 0
1382#define MOPT_QFMT 0x0040
1383#else
1384#define MOPT_Q MOPT_NOSUPPORT
1385#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1386#endif
26092bf5 1387#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1388#define MOPT_NO_EXT2 0x0100
1389#define MOPT_NO_EXT3 0x0200
1390#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1391#define MOPT_STRING 0x0400
26092bf5
TT		 1392
1393static const struct mount_opts {
1394 int token;
1395 int mount_opt;
1396 int flags;
1397} ext4_mount_opts[] = {
1398 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1399 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1400 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1401 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1402 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1403 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1404 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1405 MOPT_EXT4_ONLY | MOPT_SET},
1406 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1407 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1408 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1409 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1410 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1411 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1412 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1413 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1414 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1415 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1416 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1417 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1418 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1419 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1420 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1421 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1422 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1423 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1424 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1425 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1426 MOPT_NO_EXT2 | MOPT_SET},
1427 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1428 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1429 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1430 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1431 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1432 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1433 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1434 {Opt_commit, 0, MOPT_GTE0},
1435 {Opt_max_batch_time, 0, MOPT_GTE0},
1436 {Opt_min_batch_time, 0, MOPT_GTE0},
1437 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1438 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1439 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1440 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1441 {Opt_resuid, 0, MOPT_GTE0},
1442 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1443 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1444 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1445 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1446 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1447 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1448 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1449 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1450 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1451 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1452#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1453 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1454 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1455#else
26092bf5
TT		 1456 {Opt_acl, 0, MOPT_NOSUPPORT},
1457 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1458#endif
26092bf5
TT		 1459 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1460 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1461 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1462 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1463 MOPT_SET | MOPT_Q},
1464 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1465 MOPT_SET | MOPT_Q},
1466 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1467 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1468 {Opt_usrjquota, 0, MOPT_Q},
1469 {Opt_grpjquota, 0, MOPT_Q},
1470 {Opt_offusrjquota, 0, MOPT_Q},
1471 {Opt_offgrpjquota, 0, MOPT_Q},
1472 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1473 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1474 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1475 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1476 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1477 {Opt_err, 0, 0}
1478};
1479
1480static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1481 substring_t *args, unsigned long *journal_devnum,
1482 unsigned int *journal_ioprio, int is_remount)
1483{
1484 struct ext4_sb_info *sbi = EXT4_SB(sb);
1485 const struct mount_opts *m;
08cefc7a
EB		 1486 kuid_t uid;
1487 kgid_t gid;
26092bf5
TT		 1488 int arg = 0;
1489
57f73c2c
TT		 1490#ifdef CONFIG_QUOTA
1491 if (token == Opt_usrjquota)
1492 return set_qf_name(sb, USRQUOTA, &args[0]);
1493 else if (token == Opt_grpjquota)
1494 return set_qf_name(sb, GRPQUOTA, &args[0]);
1495 else if (token == Opt_offusrjquota)
1496 return clear_qf_name(sb, USRQUOTA);
1497 else if (token == Opt_offgrpjquota)
1498 return clear_qf_name(sb, GRPQUOTA);
1499#endif
26092bf5 1500 switch (token) {
f7048605
TT		 1501 case Opt_noacl:
1502 case Opt_nouser_xattr:
1503 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1504 break;
26092bf5
TT		 1505 case Opt_sb:
1506 return 1; /* handled by get_sb_block() */
1507 case Opt_removed:
5f3633e3 1508 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1509 return 1;
26092bf5
TT		 1510 case Opt_abort:
1511 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1512 return 1;
1513 case Opt_i_version:
1514 sb->s_flags |= MS_I_VERSION;
1515 return 1;
a26f4992
TT		 1516 case Opt_lazytime:
1517 sb->s_flags |= MS_LAZYTIME;
1518 return 1;
1519 case Opt_nolazytime:
1520 sb->s_flags &= ~MS_LAZYTIME;
1521 return 1;
26092bf5
TT		 1522 }
1523
5f3633e3
JK		 1524 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1525 if (token == m->token)
1526 break;
1527
1528 if (m->token == Opt_err) {
1529 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1530 "or missing value", opt);
1531 return -1;
1532 }
1533
8dc0aa8c
TT		 1534 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1535 ext4_msg(sb, KERN_ERR,
1536 "Mount option \"%s\" incompatible with ext2", opt);
1537 return -1;
1538 }
1539 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1540 ext4_msg(sb, KERN_ERR,
1541 "Mount option \"%s\" incompatible with ext3", opt);
1542 return -1;
1543 }
1544
ad4eec61 1545 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1546 return -1;
1547 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1548 return -1;
c93cf2d7
DM		 1549 if (m->flags & MOPT_EXPLICIT) {
1550 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1551 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1552 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1553 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1554 } else
1555 return -1;
1556 }
5f3633e3
JK		 1557 if (m->flags & MOPT_CLEAR_ERR)
1558 clear_opt(sb, ERRORS_MASK);
1559 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1560 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1561 "options when quota turned on");
1562 return -1;
1563 }
1564
1565 if (m->flags & MOPT_NOSUPPORT) {
1566 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1567 } else if (token == Opt_commit) {
1568 if (arg == 0)
1569 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1570 sbi->s_commit_interval = HZ * arg;
1571 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1572 sbi->s_max_batch_time = arg;
1573 } else if (token == Opt_min_batch_time) {
1574 sbi->s_min_batch_time = arg;
1575 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1576 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1577 ext4_msg(sb, KERN_ERR,
1578 "EXT4-fs: inode_readahead_blks must be "
1579 "0 or a power of 2 smaller than 2^31");
26092bf5 1580 return -1;
5f3633e3
JK		 1581 }
1582 sbi->s_inode_readahead_blks = arg;
1583 } else if (token == Opt_init_itable) {
1584 set_opt(sb, INIT_INODE_TABLE);
1585 if (!args->from)
1586 arg = EXT4_DEF_LI_WAIT_MULT;
1587 sbi->s_li_wait_mult = arg;
1588 } else if (token == Opt_max_dir_size_kb) {
1589 sbi->s_max_dir_size_kb = arg;
1590 } else if (token == Opt_stripe) {
1591 sbi->s_stripe = arg;
1592 } else if (token == Opt_resuid) {
1593 uid = make_kuid(current_user_ns(), arg);
1594 if (!uid_valid(uid)) {
1595 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1596 return -1;
1597 }
5f3633e3
JK		 1598 sbi->s_resuid = uid;
1599 } else if (token == Opt_resgid) {
1600 gid = make_kgid(current_user_ns(), arg);
1601 if (!gid_valid(gid)) {
1602 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1603 return -1;
1604 }
1605 sbi->s_resgid = gid;
1606 } else if (token == Opt_journal_dev) {
1607 if (is_remount) {
1608 ext4_msg(sb, KERN_ERR,
1609 "Cannot specify journal on remount");
1610 return -1;
1611 }
1612 *journal_devnum = arg;
ad4eec61
ES		 1613 } else if (token == Opt_journal_path) {
1614 char *journal_path;
1615 struct inode *journal_inode;
1616 struct path path;
1617 int error;
1618
1619 if (is_remount) {
1620 ext4_msg(sb, KERN_ERR,
1621 "Cannot specify journal on remount");
1622 return -1;
1623 }
1624 journal_path = match_strdup(&args[0]);
1625 if (!journal_path) {
1626 ext4_msg(sb, KERN_ERR, "error: could not dup "
1627 "journal device string");
1628 return -1;
1629 }
1630
1631 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1632 if (error) {
1633 ext4_msg(sb, KERN_ERR, "error: could not find "
1634 "journal device path: error %d", error);
1635 kfree(journal_path);
1636 return -1;
1637 }
1638
2b0143b5 1639 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1640 if (!S_ISBLK(journal_inode->i_mode)) {
1641 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1642 "is not a block device", journal_path);
1643 path_put(&path);
1644 kfree(journal_path);
1645 return -1;
1646 }
1647
1648 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1649 path_put(&path);
1650 kfree(journal_path);
5f3633e3
JK		 1651 } else if (token == Opt_journal_ioprio) {
1652 if (arg > 7) {
1653 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1654 " (must be 0-7)");
1655 return -1;
1656 }
1657 *journal_ioprio =
1658 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1659 } else if (token == Opt_test_dummy_encryption) {
1660#ifdef CONFIG_EXT4_FS_ENCRYPTION
1661 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1662 ext4_msg(sb, KERN_WARNING,
1663 "Test dummy encryption mode enabled");
1664#else
1665 ext4_msg(sb, KERN_WARNING,
1666 "Test dummy encryption mount option ignored");
1667#endif
5f3633e3
JK		 1668 } else if (m->flags & MOPT_DATAJ) {
1669 if (is_remount) {
1670 if (!sbi->s_journal)
1671 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1672 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1673 ext4_msg(sb, KERN_ERR,
26092bf5 1674 "Cannot change data mode on remount");
26092bf5 1675 return -1;
ac27a0ec 1676 }
26092bf5 1677 } else {
5f3633e3
JK		 1678 clear_opt(sb, DATA_FLAGS);
1679 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1680 }
5f3633e3
JK		 1681#ifdef CONFIG_QUOTA
1682 } else if (m->flags & MOPT_QFMT) {
1683 if (sb_any_quota_loaded(sb) &&
1684 sbi->s_jquota_fmt != m->mount_opt) {
1685 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1686 "quota options when quota turned on");
1687 return -1;
1688 }
e2b911c5 1689 if (ext4_has_feature_quota(sb)) {
262b4662
JK		 1690 ext4_msg(sb, KERN_ERR,
1691 "Cannot set journaled quota options "
1692 "when QUOTA feature is enabled");
1693 return -1;
1694 }
5f3633e3 1695 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1696#endif
923ae0ff 1697 } else if (token == Opt_dax) {
ef83b6e8
DW		 1698#ifdef CONFIG_FS_DAX
1699 ext4_msg(sb, KERN_WARNING,
1700 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1701 sbi->s_mount_opt |= m->mount_opt;
1702#else
923ae0ff
RZ		 1703 ext4_msg(sb, KERN_INFO, "dax option not supported");
1704 return -1;
5f3633e3
JK		 1705#endif
1706 } else {
1707 if (!args->from)
1708 arg = 1;
1709 if (m->flags & MOPT_CLEAR)
1710 arg = !arg;
1711 else if (unlikely(!(m->flags & MOPT_SET))) {
1712 ext4_msg(sb, KERN_WARNING,
1713 "buggy handling of option %s", opt);
1714 WARN_ON(1);
1715 return -1;
1716 }
1717 if (arg != 0)
1718 sbi->s_mount_opt |= m->mount_opt;
1719 else
1720 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1721 }
5f3633e3 1722 return 1;
26092bf5
TT		 1723}
1724
1725static int parse_options(char *options, struct super_block *sb,
1726 unsigned long *journal_devnum,
1727 unsigned int *journal_ioprio,
1728 int is_remount)
1729{
1730 struct ext4_sb_info *sbi = EXT4_SB(sb);
1731 char *p;
1732 substring_t args[MAX_OPT_ARGS];
1733 int token;
1734
1735 if (!options)
1736 return 1;
1737
1738 while ((p = strsep(&options, ",")) != NULL) {
1739 if (!*p)
1740 continue;
1741 /*
1742 * Initialize args struct so we know whether arg was
1743 * found; some options take optional arguments.
1744 */
caecd0af 1745 args[0].to = args[0].from = NULL;
26092bf5
TT		 1746 token = match_token(p, tokens, args);
1747 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1748 journal_ioprio, is_remount) < 0)
1749 return 0;
ac27a0ec
DK		 1750 }
1751#ifdef CONFIG_QUOTA
e2b911c5 1752 if (ext4_has_feature_quota(sb) &&
262b4662
JK		 1753 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1754 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1755 "feature is enabled");
1756 return 0;
1757 }
ac27a0ec 1758 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1759 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1760 clear_opt(sb, USRQUOTA);
ac27a0ec 1761
482a7425 1762 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1763 clear_opt(sb, GRPQUOTA);
ac27a0ec 1764
56c50f11 1765 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1766 ext4_msg(sb, KERN_ERR, "old and new quota "
1767 "format mixing");
ac27a0ec
DK		 1768 return 0;
1769 }
1770
1771 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1772 ext4_msg(sb, KERN_ERR, "journaled quota format "
1773 "not specified");
ac27a0ec
DK		 1774 return 0;
1775 }
ac27a0ec
DK		 1776 }
1777#endif
261cb20c
JK		 1778 if (test_opt(sb, DIOREAD_NOLOCK)) {
1779 int blocksize =
1780 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1781
1782 if (blocksize < PAGE_CACHE_SIZE) {
1783 ext4_msg(sb, KERN_ERR, "can't mount with "
1784 "dioread_nolock if block size != PAGE_SIZE");
1785 return 0;
1786 }
1787 }
d4f76107
JK		 1788 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1789 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1790 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1791 "in data=ordered mode");
1792 return 0;
1793 }
ac27a0ec
DK		 1794 return 1;
1795}
1796
2adf6da8
TT		 1797static inline void ext4_show_quota_options(struct seq_file *seq,
1798 struct super_block *sb)
1799{
1800#if defined(CONFIG_QUOTA)
1801 struct ext4_sb_info *sbi = EXT4_SB(sb);
1802
1803 if (sbi->s_jquota_fmt) {
1804 char *fmtname = "";
1805
1806 switch (sbi->s_jquota_fmt) {
1807 case QFMT_VFS_OLD:
1808 fmtname = "vfsold";
1809 break;
1810 case QFMT_VFS_V0:
1811 fmtname = "vfsv0";
1812 break;
1813 case QFMT_VFS_V1:
1814 fmtname = "vfsv1";
1815 break;
1816 }
1817 seq_printf(seq, ",jqfmt=%s", fmtname);
1818 }
1819
1820 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1821 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1822
1823 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1824 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1825#endif
1826}
1827
5a916be1
TT		 1828static const char *token2str(int token)
1829{
50df9fd5 1830 const struct match_token *t;
5a916be1
TT		 1831
1832 for (t = tokens; t->token != Opt_err; t++)
1833 if (t->token == token && !strchr(t->pattern, '='))
1834 break;
1835 return t->pattern;
1836}
1837
2adf6da8
TT		 1838/*
1839 * Show an option if
1840 * - it's set to a non-default value OR
1841 * - if the per-sb default is different from the global default
1842 */
66acdcf4
TT		 1843static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1844 int nodefs)
2adf6da8 1845{
2adf6da8
TT		 1846 struct ext4_sb_info *sbi = EXT4_SB(sb);
1847 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1848 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1849 const struct mount_opts *m;
66acdcf4 1850 char sep = nodefs ? '\n' : ',';
2adf6da8 1851
66acdcf4
TT		 1852#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1853#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1854
1855 if (sbi->s_sb_block != 1)
5a916be1
TT		 1856 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1857
1858 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1859 int want_set = m->flags & MOPT_SET;
1860 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1861 (m->flags & MOPT_CLEAR_ERR))
1862 continue;
66acdcf4 1863 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1864 continue; /* skip if same as the default */
1865 if ((want_set &&
1866 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1867 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1868 continue; /* select Opt_noFoo vs Opt_Foo */
1869 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1870 }
5a916be1 1871
08cefc7a 1872 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1873 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1874 SEQ_OPTS_PRINT("resuid=%u",
1875 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1876 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1877 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1878 SEQ_OPTS_PRINT("resgid=%u",
1879 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1880 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1881 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1882 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1883 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1884 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1885 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1886 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1887 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1888 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1889 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1890 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1891 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1892 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1893 if (sb->s_flags & MS_I_VERSION)
5a916be1 1894 SEQ_OPTS_PUTS("i_version");
66acdcf4 1895 if (nodefs || sbi->s_stripe)
5a916be1 1896 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1897 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1898 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1899 SEQ_OPTS_PUTS("data=journal");
1900 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1901 SEQ_OPTS_PUTS("data=ordered");
1902 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1903 SEQ_OPTS_PUTS("data=writeback");
1904 }
66acdcf4
TT		 1905 if (nodefs ||
1906 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1907 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1908 sbi->s_inode_readahead_blks);
2adf6da8 1909
66acdcf4
TT		 1910 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1911 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1912 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1913 if (nodefs || sbi->s_max_dir_size_kb)
1914 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1915
1916 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1917 return 0;
1918}
1919
66acdcf4
TT		 1920static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1921{
1922 return _ext4_show_options(seq, root->d_sb, 0);
1923}
1924
ebd173be 1925int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 1926{
1927 struct super_block *sb = seq->private;
1928 int rc;
1929
1930 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1931 rc = _ext4_show_options(seq, sb, 1);
1932 seq_puts(seq, "\n");
1933 return rc;
1934}
1935
617ba13b 1936static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1937 int read_only)
1938{
617ba13b 1939 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1940 int res = 0;
1941
617ba13b 1942 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1943 ext4_msg(sb, KERN_ERR, "revision level too high, "
1944 "forcing read-only mode");
ac27a0ec
DK		 1945 res = MS_RDONLY;
1946 }
1947 if (read_only)
281b5995 1948 goto done;
617ba13b 1949 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1950 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1951 "running e2fsck is recommended");
c8b459f4 1952 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1953 ext4_msg(sb, KERN_WARNING,
1954 "warning: mounting fs with errors, "
1955 "running e2fsck is recommended");
ed3ce80a 1956 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1957 le16_to_cpu(es->s_mnt_count) >=
1958 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1959 ext4_msg(sb, KERN_WARNING,
1960 "warning: maximal mount count reached, "
1961 "running e2fsck is recommended");
ac27a0ec
DK		 1962 else if (le32_to_cpu(es->s_checkinterval) &&
1963 (le32_to_cpu(es->s_lastcheck) +
1964 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1965 ext4_msg(sb, KERN_WARNING,
1966 "warning: checktime reached, "
1967 "running e2fsck is recommended");
0b8e58a1 1968 if (!sbi->s_journal)
0390131b 1969 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1970 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1971 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1972 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1973 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1974 ext4_update_dynamic_rev(sb);
0390131b 1975 if (sbi->s_journal)
e2b911c5 1976 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 1977
e2d67052 1978 ext4_commit_super(sb, 1);
281b5995 1979done:
ac27a0ec 1980 if (test_opt(sb, DEBUG))
a9df9a49 1981 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1982 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1983 sb->s_blocksize,
1984 sbi->s_groups_count,
617ba13b
MC		 1985 EXT4_BLOCKS_PER_GROUP(sb),
1986 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1987 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1988
7abc52c2 1989 cleancache_init_fs(sb);
ac27a0ec
DK		 1990 return res;
1991}
1992
117fff10
TT		 1993int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1994{
1995 struct ext4_sb_info *sbi = EXT4_SB(sb);
1996 struct flex_groups *new_groups;
1997 int size;
1998
1999 if (!sbi->s_log_groups_per_flex)
2000 return 0;
2001
2002 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2003 if (size <= sbi->s_flex_groups_allocated)
2004 return 0;
2005
2006 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2007 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2008 if (!new_groups) {
2009 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2010 size / (int) sizeof(struct flex_groups));
2011 return -ENOMEM;
2012 }
2013
2014 if (sbi->s_flex_groups) {
2015 memcpy(new_groups, sbi->s_flex_groups,
2016 (sbi->s_flex_groups_allocated *
2017 sizeof(struct flex_groups)));
b93b41d4 2018 kvfree(sbi->s_flex_groups);
117fff10
TT		 2019 }
2020 sbi->s_flex_groups = new_groups;
2021 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2022 return 0;
2023}
2024
772cb7c8
JS		 2025static int ext4_fill_flex_info(struct super_block *sb)
2026{
2027 struct ext4_sb_info *sbi = EXT4_SB(sb);
2028 struct ext4_group_desc *gdp = NULL;
772cb7c8 2029 ext4_group_t flex_group;
117fff10 2030 int i, err;
772cb7c8 2031
503358ae 2032 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2033 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2034 sbi->s_log_groups_per_flex = 0;
2035 return 1;
2036 }
2037
117fff10
TT		 2038 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2039 if (err)
9933fc0a 2040 goto failed;
772cb7c8 2041
772cb7c8 2042 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2043 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2044
2045 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2046 atomic_add(ext4_free_inodes_count(sb, gdp),
2047 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2048 atomic64_add(ext4_free_group_clusters(sb, gdp),
2049 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2050 atomic_add(ext4_used_dirs_count(sb, gdp),
2051 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2052 }
2053
2054 return 1;
2055failed:
2056 return 0;
2057}
2058
e2b911c5 2059static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2060 struct ext4_group_desc *gdp)
717d50e4 2061{
feb0ab32 2062 int offset;
717d50e4 2063 __u16 crc = 0;
feb0ab32 2064 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2065 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2066
9aa5d32b 2067 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2068 /* Use new metadata_csum algorithm */
d6a77105 2069 __le16 save_csum;
feb0ab32
DW		 2070 __u32 csum32;
2071
d6a77105 2072 save_csum = gdp->bg_checksum;
feb0ab32
DW		 2073 gdp->bg_checksum = 0;
2074 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2075 sizeof(le_group));
2076 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2077 sbi->s_desc_size);
d6a77105 2078 gdp->bg_checksum = save_csum;
feb0ab32
DW		 2079
2080 crc = csum32 & 0xFFFF;
2081 goto out;
717d50e4
AD		 2082 }
2083
feb0ab32 2084 /* old crc16 code */
e2b911c5 2085 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2086 return 0;
2087
feb0ab32
DW		 2088 offset = offsetof(struct ext4_group_desc, bg_checksum);
2089
2090 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2091 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2092 crc = crc16(crc, (__u8 *)gdp, offset);
2093 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2094 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2095 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2096 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2097 crc = crc16(crc, (__u8 *)gdp + offset,
2098 le16_to_cpu(sbi->s_es->s_desc_size) -
2099 offset);
2100
2101out:
717d50e4
AD		 2102 return cpu_to_le16(crc);
2103}
2104
feb0ab32 2105int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2106 struct ext4_group_desc *gdp)
2107{
feb0ab32 2108 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2109 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2110 return 0;
2111
2112 return 1;
2113}
2114
feb0ab32
DW		 2115void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2116 struct ext4_group_desc *gdp)
2117{
2118 if (!ext4_has_group_desc_csum(sb))
2119 return;
e2b911c5 2120 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2121}
2122
ac27a0ec 2123/* Called at mount-time, super-block is locked */
bfff6873
LC		 2124static int ext4_check_descriptors(struct super_block *sb,
2125 ext4_group_t *first_not_zeroed)
ac27a0ec 2126{
617ba13b
MC		 2127 struct ext4_sb_info *sbi = EXT4_SB(sb);
2128 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2129 ext4_fsblk_t last_block;
bd81d8ee
LV		 2130 ext4_fsblk_t block_bitmap;
2131 ext4_fsblk_t inode_bitmap;
2132 ext4_fsblk_t inode_table;
ce421581 2133 int flexbg_flag = 0;
bfff6873 2134 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2135
e2b911c5 2136 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2137 flexbg_flag = 1;
2138
af5bc92d 2139 ext4_debug("Checking group descriptors");
ac27a0ec 2140
197cd65a
AM		 2141 for (i = 0; i < sbi->s_groups_count; i++) {
2142 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2143
ce421581 2144 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2145 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2146 else
2147 last_block = first_block +
617ba13b 2148 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2149
bfff6873
LC		 2150 if ((grp == sbi->s_groups_count) &&
2151 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2152 grp = i;
2153
8fadc143 2154 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2155 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2156 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2157 "Block bitmap for group %u not in group "
b31e1552 2158 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2159 return 0;
2160 }
8fadc143 2161 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2162 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2163 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2164 "Inode bitmap for group %u not in group "
b31e1552 2165 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2166 return 0;
2167 }
8fadc143 2168 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2169 if (inode_table < first_block ||
2b2d6d01 2170 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2171 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2172 "Inode table for group %u not in group "
b31e1552 2173 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2174 return 0;
2175 }
955ce5f5 2176 ext4_lock_group(sb, i);
feb0ab32 2177 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2178 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2179 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2180 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2181 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2182 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2183 ext4_unlock_group(sb, i);
8a266467 2184 return 0;
7ee1ec4c 2185 }
717d50e4 2186 }
955ce5f5 2187 ext4_unlock_group(sb, i);
ce421581
JS		 2188 if (!flexbg_flag)
2189 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2190 }
bfff6873
LC		 2191 if (NULL != first_not_zeroed)
2192 *first_not_zeroed = grp;
ac27a0ec
DK		 2193 return 1;
2194}
2195
617ba13b 2196/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2197 * the superblock) which were deleted from all directories, but held open by
2198 * a process at the time of a crash. We walk the list and try to delete these
2199 * inodes at recovery time (only with a read-write filesystem).
2200 *
2201 * In order to keep the orphan inode chain consistent during traversal (in
2202 * case of crash during recovery), we link each inode into the superblock
2203 * orphan list_head and handle it the same way as an inode deletion during
2204 * normal operation (which journals the operations for us).
2205 *
2206 * We only do an iget() and an iput() on each inode, which is very safe if we
2207 * accidentally point at an in-use or already deleted inode. The worst that
2208 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2209 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2210 * e2fsck was run on this filesystem, and it must have already done the orphan
2211 * inode cleanup for us, so we can safely abort without any further action.
2212 */
2b2d6d01
TT		 2213static void ext4_orphan_cleanup(struct super_block *sb,
2214 struct ext4_super_block *es)
ac27a0ec
DK		 2215{
2216 unsigned int s_flags = sb->s_flags;
2217 int nr_orphans = 0, nr_truncates = 0;
2218#ifdef CONFIG_QUOTA
2219 int i;
2220#endif
2221 if (!es->s_last_orphan) {
2222 jbd_debug(4, "no orphan inodes to clean up\n");
2223 return;
2224 }
2225
a8f48a95 2226 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2227 ext4_msg(sb, KERN_ERR, "write access "
2228 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2229 return;
2230 }
2231
d39195c3
AG		 2232 /* Check if feature set would not allow a r/w mount */
2233 if (!ext4_feature_set_ok(sb, 0)) {
2234 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2235 "unknown ROCOMPAT features");
2236 return;
2237 }
2238
617ba13b 2239 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2240 /* don't clear list on RO mount w/ errors */
2241 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2242 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2243 "clearing orphan list.\n");
c25f9bc6
ES		 2244 es->s_last_orphan = 0;
2245 }
ac27a0ec
DK		 2246 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2247 return;
2248 }
2249
2250 if (s_flags & MS_RDONLY) {
b31e1552 2251 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2252 sb->s_flags &= ~MS_RDONLY;
2253 }
2254#ifdef CONFIG_QUOTA
2255 /* Needed for iput() to work correctly and not trash data */
2256 sb->s_flags |= MS_ACTIVE;
2257 /* Turn on quotas so that they are updated correctly */
a2d4a646 2258 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2259 if (EXT4_SB(sb)->s_qf_names[i]) {
2260 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2261 if (ret < 0)
b31e1552
ES		 2262 ext4_msg(sb, KERN_ERR,
2263 "Cannot turn on journaled "
2264 "quota: error %d", ret);
ac27a0ec
DK		 2265 }
2266 }
2267#endif
2268
2269 while (es->s_last_orphan) {
2270 struct inode *inode;
2271
97bd42b9
JB		 2272 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2273 if (IS_ERR(inode)) {
ac27a0ec
DK		 2274 es->s_last_orphan = 0;
2275 break;
2276 }
2277
617ba13b 2278 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2279 dquot_initialize(inode);
ac27a0ec 2280 if (inode->i_nlink) {
566370a2
PT		 2281 if (test_opt(sb, DEBUG))
2282 ext4_msg(sb, KERN_DEBUG,
2283 "%s: truncating inode %lu to %lld bytes",
2284 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2285 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2286 inode->i_ino, inode->i_size);
5955102c 2287 inode_lock(inode);
55f252c9 2288 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2289 ext4_truncate(inode);
5955102c 2290 inode_unlock(inode);
ac27a0ec
DK		 2291 nr_truncates++;
2292 } else {
566370a2
PT		 2293 if (test_opt(sb, DEBUG))
2294 ext4_msg(sb, KERN_DEBUG,
2295 "%s: deleting unreferenced inode %lu",
2296 __func__, inode->i_ino);
ac27a0ec
DK		 2297 jbd_debug(2, "deleting unreferenced inode %lu\n",
2298 inode->i_ino);
2299 nr_orphans++;
2300 }
2301 iput(inode); /* The delete magic happens here! */
2302 }
2303
2b2d6d01 2304#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2305
2306 if (nr_orphans)
b31e1552
ES		 2307 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2308 PLURAL(nr_orphans));
ac27a0ec 2309 if (nr_truncates)
b31e1552
ES		 2310 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2311 PLURAL(nr_truncates));
ac27a0ec
DK		 2312#ifdef CONFIG_QUOTA
2313 /* Turn quotas off */
a2d4a646 2314 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2315 if (sb_dqopt(sb)->files[i])
287a8095 2316 dquot_quota_off(sb, i);
ac27a0ec
DK		 2317 }
2318#endif
2319 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2320}
0b8e58a1 2321
cd2291a4
ES		 2322/*
2323 * Maximal extent format file size.
2324 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2325 * extent format containers, within a sector_t, and within i_blocks
2326 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2327 * so that won't be a limiting factor.
2328 *
f17722f9
LC		 2329 * However there is other limiting factor. We do store extents in the form
2330 * of starting block and length, hence the resulting length of the extent
2331 * covering maximum file size must fit into on-disk format containers as
2332 * well. Given that length is always by 1 unit bigger than max unit (because
2333 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2334 *
cd2291a4
ES		 2335 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2336 */
f287a1a5 2337static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2338{
2339 loff_t res;
2340 loff_t upper_limit = MAX_LFS_FILESIZE;
2341
2342 /* small i_blocks in vfs inode? */
f287a1a5 2343 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2344 /*
90c699a9 2345 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2346 * i_block represent total blocks in 512 bytes
2347 * 32 == size of vfs inode i_blocks * 8
2348 */
2349 upper_limit = (1LL << 32) - 1;
2350
2351 /* total blocks in file system block size */
2352 upper_limit >>= (blkbits - 9);
2353 upper_limit <<= blkbits;
2354 }
2355
f17722f9
LC		 2356 /*
2357 * 32-bit extent-start container, ee_block. We lower the maxbytes
2358 * by one fs block, so ee_len can cover the extent of maximum file
2359 * size
2360 */
2361 res = (1LL << 32) - 1;
cd2291a4 2362 res <<= blkbits;
cd2291a4
ES		 2363
2364 /* Sanity check against vm- & vfs- imposed limits */
2365 if (res > upper_limit)
2366 res = upper_limit;
2367
2368 return res;
2369}
ac27a0ec 2370
ac27a0ec 2371/*
cd2291a4 2372 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2373 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2374 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2375 */
f287a1a5 2376static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2377{
617ba13b 2378 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2379 int meta_blocks;
2380 loff_t upper_limit;
0b8e58a1
AD		 2381 /* This is calculated to be the largest file size for a dense, block
2382 * mapped file such that the file's total number of 512-byte sectors,
2383 * including data and all indirect blocks, does not exceed (2^48 - 1).
2384 *
2385 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2386 * number of 512-byte sectors of the file.
0fc1b451
AK		 2387 */
2388
f287a1a5 2389 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2390 /*
90c699a9 2391 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2392 * the inode i_block field represents total file blocks in
2393 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2394 */
2395 upper_limit = (1LL << 32) - 1;
2396
2397 /* total blocks in file system block size */
2398 upper_limit >>= (bits - 9);
2399
2400 } else {
8180a562
AK		 2401 /*
2402 * We use 48 bit ext4_inode i_blocks
2403 * With EXT4_HUGE_FILE_FL set the i_blocks
2404 * represent total number of blocks in
2405 * file system block size
2406 */
0fc1b451
AK		 2407 upper_limit = (1LL << 48) - 1;
2408
0fc1b451
AK		 2409 }
2410
2411 /* indirect blocks */
2412 meta_blocks = 1;
2413 /* double indirect blocks */
2414 meta_blocks += 1 + (1LL << (bits-2));
2415 /* tripple indirect blocks */
2416 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2417
2418 upper_limit -= meta_blocks;
2419 upper_limit <<= bits;
ac27a0ec
DK		 2420
2421 res += 1LL << (bits-2);
2422 res += 1LL << (2*(bits-2));
2423 res += 1LL << (3*(bits-2));
2424 res <<= bits;
2425 if (res > upper_limit)
2426 res = upper_limit;
0fc1b451
AK		 2427
2428 if (res > MAX_LFS_FILESIZE)
2429 res = MAX_LFS_FILESIZE;
2430
ac27a0ec
DK		 2431 return res;
2432}
2433
617ba13b 2434static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2435 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2436{
617ba13b 2437 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2438 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2439 int has_super = 0;
2440
2441 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2442
e2b911c5 2443 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2444 return logical_sb_block + nr + 1;
ac27a0ec 2445 bg = sbi->s_desc_per_block * nr;
617ba13b 2446 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2447 has_super = 1;
0b8e58a1 2448
bd63f6b0
DW		 2449 /*
2450 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2451 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2452 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2453 * compensate.
2454 */
2455 if (sb->s_blocksize == 1024 && nr == 0 &&
2456 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2457 has_super++;
2458
617ba13b 2459 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2460}
2461
c9de560d
AT		 2462/**
2463 * ext4_get_stripe_size: Get the stripe size.
2464 * @sbi: In memory super block info
2465 *
2466 * If we have specified it via mount option, then
2467 * use the mount option value. If the value specified at mount time is
2468 * greater than the blocks per group use the super block value.
2469 * If the super block value is greater than blocks per group return 0.
2470 * Allocator needs it be less than blocks per group.
2471 *
2472 */
2473static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2474{
2475 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2476 unsigned long stripe_width =
2477 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2478 int ret;
c9de560d
AT		 2479
2480 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2481 ret = sbi->s_stripe;
2482 else if (stripe_width <= sbi->s_blocks_per_group)
2483 ret = stripe_width;
2484 else if (stride <= sbi->s_blocks_per_group)
2485 ret = stride;
2486 else
2487 ret = 0;
c9de560d 2488
3eb08658
DE		 2489 /*
2490 * If the stripe width is 1, this makes no sense and
2491 * we set it to 0 to turn off stripe handling code.
2492 */
2493 if (ret <= 1)
2494 ret = 0;
c9de560d 2495
3eb08658 2496 return ret;
c9de560d 2497}
ac27a0ec 2498
a13fb1a4
ES		 2499/*
2500 * Check whether this filesystem can be mounted based on
2501 * the features present and the RDONLY/RDWR mount requested.
2502 * Returns 1 if this filesystem can be mounted as requested,
2503 * 0 if it cannot be.
2504 */
2505static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2506{
e2b911c5 2507 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2508 ext4_msg(sb, KERN_ERR,
2509 "Couldn't mount because of "
2510 "unsupported optional features (%x)",
2511 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2512 ~EXT4_FEATURE_INCOMPAT_SUPP));
2513 return 0;
2514 }
2515
2516 if (readonly)
2517 return 1;
2518
e2b911c5 2519 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b
DW		 2520 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2521 sb->s_flags |= MS_RDONLY;
2522 return 1;
2523 }
2524
a13fb1a4 2525 /* Check that feature set is OK for a read-write mount */
e2b911c5 2526 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2527 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2528 "unsupported optional features (%x)",
2529 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2530 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2531 return 0;
2532 }
2533 /*
2534 * Large file size enabled file system can only be mounted
2535 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2536 */
e2b911c5 2537 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2538 if (sizeof(blkcnt_t) < sizeof(u64)) {
2539 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2540 "cannot be mounted RDWR without "
2541 "CONFIG_LBDAF");
2542 return 0;
2543 }
2544 }
e2b911c5 2545 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2546 ext4_msg(sb, KERN_ERR,
2547 "Can't support bigalloc feature without "
2548 "extents feature\n");
2549 return 0;
2550 }
7c319d32
AK		 2551
2552#ifndef CONFIG_QUOTA
e2b911c5 2553 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2554 ext4_msg(sb, KERN_ERR,
2555 "Filesystem with quota feature cannot be mounted RDWR "
2556 "without CONFIG_QUOTA");
2557 return 0;
2558 }
689c958c
LX		 2559 if (ext4_has_feature_project(sb) && !readonly) {
2560 ext4_msg(sb, KERN_ERR,
2561 "Filesystem with project quota feature cannot be mounted RDWR "
2562 "without CONFIG_QUOTA");
2563 return 0;
2564 }
7c319d32 2565#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2566 return 1;
2567}
2568
66e61a9e
TT		 2569/*
2570 * This function is called once a day if we have errors logged
2571 * on the file system
2572 */
2573static void print_daily_error_info(unsigned long arg)
2574{
2575 struct super_block *sb = (struct super_block *) arg;
2576 struct ext4_sb_info *sbi;
2577 struct ext4_super_block *es;
2578
2579 sbi = EXT4_SB(sb);
2580 es = sbi->s_es;
2581
2582 if (es->s_error_count)
ae0f78de
TT		 2583 /* fsck newer than v1.41.13 is needed to clean this condition. */
2584 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2585 le32_to_cpu(es->s_error_count));
2586 if (es->s_first_error_time) {
ae0f78de 2587 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2588 sb->s_id, le32_to_cpu(es->s_first_error_time),
2589 (int) sizeof(es->s_first_error_func),
2590 es->s_first_error_func,
2591 le32_to_cpu(es->s_first_error_line));
2592 if (es->s_first_error_ino)
2593 printk(": inode %u",
2594 le32_to_cpu(es->s_first_error_ino));
2595 if (es->s_first_error_block)
2596 printk(": block %llu", (unsigned long long)
2597 le64_to_cpu(es->s_first_error_block));
2598 printk("\n");
2599 }
2600 if (es->s_last_error_time) {
ae0f78de 2601 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2602 sb->s_id, le32_to_cpu(es->s_last_error_time),
2603 (int) sizeof(es->s_last_error_func),
2604 es->s_last_error_func,
2605 le32_to_cpu(es->s_last_error_line));
2606 if (es->s_last_error_ino)
2607 printk(": inode %u",
2608 le32_to_cpu(es->s_last_error_ino));
2609 if (es->s_last_error_block)
2610 printk(": block %llu", (unsigned long long)
2611 le64_to_cpu(es->s_last_error_block));
2612 printk("\n");
2613 }
2614 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2615}
2616
bfff6873
LC		 2617/* Find next suitable group and run ext4_init_inode_table */
2618static int ext4_run_li_request(struct ext4_li_request *elr)
2619{
2620 struct ext4_group_desc *gdp = NULL;
2621 ext4_group_t group, ngroups;
2622 struct super_block *sb;
2623 unsigned long timeout = 0;
2624 int ret = 0;
2625
2626 sb = elr->lr_super;
2627 ngroups = EXT4_SB(sb)->s_groups_count;
2628
8e8ad8a5 2629 sb_start_write(sb);
bfff6873
LC		 2630 for (group = elr->lr_next_group; group < ngroups; group++) {
2631 gdp = ext4_get_group_desc(sb, group, NULL);
2632 if (!gdp) {
2633 ret = 1;
2634 break;
2635 }
2636
2637 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2638 break;
2639 }
2640
7f511862 2641 if (group >= ngroups)
bfff6873
LC		 2642 ret = 1;
2643
2644 if (!ret) {
2645 timeout = jiffies;
2646 ret = ext4_init_inode_table(sb, group,
2647 elr->lr_timeout ? 0 : 1);
2648 if (elr->lr_timeout == 0) {
51ce6511
LC		 2649 timeout = (jiffies - timeout) *
2650 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2651 elr->lr_timeout = timeout;
2652 }
2653 elr->lr_next_sched = jiffies + elr->lr_timeout;
2654 elr->lr_next_group = group + 1;
2655 }
8e8ad8a5 2656 sb_end_write(sb);
bfff6873
LC		 2657
2658 return ret;
2659}
2660
2661/*
2662 * Remove lr_request from the list_request and free the
4ed5c033 2663 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2664 */
2665static void ext4_remove_li_request(struct ext4_li_request *elr)
2666{
2667 struct ext4_sb_info *sbi;
2668
2669 if (!elr)
2670 return;
2671
2672 sbi = elr->lr_sbi;
2673
2674 list_del(&elr->lr_request);
2675 sbi->s_li_request = NULL;
2676 kfree(elr);
2677}
2678
2679static void ext4_unregister_li_request(struct super_block *sb)
2680{
1bb933fb
LC		 2681 mutex_lock(&ext4_li_mtx);
2682 if (!ext4_li_info) {
2683 mutex_unlock(&ext4_li_mtx);
bfff6873 2684 return;
1bb933fb 2685 }
bfff6873
LC		 2686
2687 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2688 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2689 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2690 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2691}
2692
8f1f7453
ES		 2693static struct task_struct *ext4_lazyinit_task;
2694
bfff6873
LC		 2695/*
2696 * This is the function where ext4lazyinit thread lives. It walks
2697 * through the request list searching for next scheduled filesystem.
2698 * When such a fs is found, run the lazy initialization request
2699 * (ext4_rn_li_request) and keep track of the time spend in this
2700 * function. Based on that time we compute next schedule time of
2701 * the request. When walking through the list is complete, compute
2702 * next waking time and put itself into sleep.
2703 */
2704static int ext4_lazyinit_thread(void *arg)
2705{
2706 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2707 struct list_head *pos, *n;
2708 struct ext4_li_request *elr;
4ed5c033 2709 unsigned long next_wakeup, cur;
bfff6873
LC		 2710
2711 BUG_ON(NULL == eli);
2712
bfff6873
LC		 2713cont_thread:
2714 while (true) {
2715 next_wakeup = MAX_JIFFY_OFFSET;
2716
2717 mutex_lock(&eli->li_list_mtx);
2718 if (list_empty(&eli->li_request_list)) {
2719 mutex_unlock(&eli->li_list_mtx);
2720 goto exit_thread;
2721 }
2722
2723 list_for_each_safe(pos, n, &eli->li_request_list) {
2724 elr = list_entry(pos, struct ext4_li_request,
2725 lr_request);
2726
b2c78cd0
TT		 2727 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2728 if (ext4_run_li_request(elr) != 0) {
2729 /* error, remove the lazy_init job */
2730 ext4_remove_li_request(elr);
2731 continue;
2732 }
bfff6873
LC		 2733 }
2734
2735 if (time_before(elr->lr_next_sched, next_wakeup))
2736 next_wakeup = elr->lr_next_sched;
2737 }
2738 mutex_unlock(&eli->li_list_mtx);
2739
a0acae0e 2740 try_to_freeze();
bfff6873 2741
4ed5c033
LC		 2742 cur = jiffies;
2743 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2744 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2745 cond_resched();
2746 continue;
2747 }
2748
4ed5c033
LC		 2749 schedule_timeout_interruptible(next_wakeup - cur);
2750
8f1f7453
ES		 2751 if (kthread_should_stop()) {
2752 ext4_clear_request_list();
2753 goto exit_thread;
2754 }
bfff6873
LC		 2755 }
2756
2757exit_thread:
2758 /*
2759 * It looks like the request list is empty, but we need
2760 * to check it under the li_list_mtx lock, to prevent any
2761 * additions into it, and of course we should lock ext4_li_mtx
2762 * to atomically free the list and ext4_li_info, because at
2763 * this point another ext4 filesystem could be registering
2764 * new one.
2765 */
2766 mutex_lock(&ext4_li_mtx);
2767 mutex_lock(&eli->li_list_mtx);
2768 if (!list_empty(&eli->li_request_list)) {
2769 mutex_unlock(&eli->li_list_mtx);
2770 mutex_unlock(&ext4_li_mtx);
2771 goto cont_thread;
2772 }
2773 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2774 kfree(ext4_li_info);
2775 ext4_li_info = NULL;
2776 mutex_unlock(&ext4_li_mtx);
2777
2778 return 0;
2779}
2780
2781static void ext4_clear_request_list(void)
2782{
2783 struct list_head *pos, *n;
2784 struct ext4_li_request *elr;
2785
2786 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2787 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2788 elr = list_entry(pos, struct ext4_li_request,
2789 lr_request);
2790 ext4_remove_li_request(elr);
2791 }
2792 mutex_unlock(&ext4_li_info->li_list_mtx);
2793}
2794
2795static int ext4_run_lazyinit_thread(void)
2796{
8f1f7453
ES		 2797 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2798 ext4_li_info, "ext4lazyinit");
2799 if (IS_ERR(ext4_lazyinit_task)) {
2800 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2801 ext4_clear_request_list();
bfff6873
LC		 2802 kfree(ext4_li_info);
2803 ext4_li_info = NULL;
92b97816 2804 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2805 "initialization thread\n",
2806 err);
2807 return err;
2808 }
2809 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2810 return 0;
2811}
2812
2813/*
2814 * Check whether it make sense to run itable init. thread or not.
2815 * If there is at least one uninitialized inode table, return
2816 * corresponding group number, else the loop goes through all
2817 * groups and return total number of groups.
2818 */
2819static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2820{
2821 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2822 struct ext4_group_desc *gdp = NULL;
2823
2824 for (group = 0; group < ngroups; group++) {
2825 gdp = ext4_get_group_desc(sb, group, NULL);
2826 if (!gdp)
2827 continue;
2828
2829 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2830 break;
2831 }
2832
2833 return group;
2834}
2835
2836static int ext4_li_info_new(void)
2837{
2838 struct ext4_lazy_init *eli = NULL;
2839
2840 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2841 if (!eli)
2842 return -ENOMEM;
2843
bfff6873
LC		 2844 INIT_LIST_HEAD(&eli->li_request_list);
2845 mutex_init(&eli->li_list_mtx);
2846
bfff6873
LC		 2847 eli->li_state |= EXT4_LAZYINIT_QUIT;
2848
2849 ext4_li_info = eli;
2850
2851 return 0;
2852}
2853
2854static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2855 ext4_group_t start)
2856{
2857 struct ext4_sb_info *sbi = EXT4_SB(sb);
2858 struct ext4_li_request *elr;
bfff6873
LC		 2859
2860 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2861 if (!elr)
2862 return NULL;
2863
2864 elr->lr_super = sb;
2865 elr->lr_sbi = sbi;
2866 elr->lr_next_group = start;
2867
2868 /*
2869 * Randomize first schedule time of the request to
2870 * spread the inode table initialization requests
2871 * better.
2872 */
dd1f723b
TT		 2873 elr->lr_next_sched = jiffies + (prandom_u32() %
2874 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 2875 return elr;
2876}
2877
7f511862
TT		 2878int ext4_register_li_request(struct super_block *sb,
2879 ext4_group_t first_not_zeroed)
bfff6873
LC		 2880{
2881 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 2882 struct ext4_li_request *elr = NULL;
bfff6873 2883 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2884 int ret = 0;
bfff6873 2885
7f511862 2886 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 2887 if (sbi->s_li_request != NULL) {
2888 /*
2889 * Reset timeout so it can be computed again, because
2890 * s_li_wait_mult might have changed.
2891 */
2892 sbi->s_li_request->lr_timeout = 0;
7f511862 2893 goto out;
51ce6511 2894 }
bfff6873
LC		 2895
2896 if (first_not_zeroed == ngroups ||
2897 (sb->s_flags & MS_RDONLY) ||
55ff3840 2898 !test_opt(sb, INIT_INODE_TABLE))
7f511862 2899 goto out;
bfff6873
LC		 2900
2901 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 2902 if (!elr) {
2903 ret = -ENOMEM;
2904 goto out;
2905 }
bfff6873
LC		 2906
2907 if (NULL == ext4_li_info) {
2908 ret = ext4_li_info_new();
2909 if (ret)
2910 goto out;
2911 }
2912
2913 mutex_lock(&ext4_li_info->li_list_mtx);
2914 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2915 mutex_unlock(&ext4_li_info->li_list_mtx);
2916
2917 sbi->s_li_request = elr;
46e4690b
TM		 2918 /*
2919 * set elr to NULL here since it has been inserted to
2920 * the request_list and the removal and free of it is
2921 * handled by ext4_clear_request_list from now on.
2922 */
2923 elr = NULL;
bfff6873
LC		 2924
2925 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2926 ret = ext4_run_lazyinit_thread();
2927 if (ret)
2928 goto out;
2929 }
bfff6873 2930out:
beed5ecb
NK		 2931 mutex_unlock(&ext4_li_mtx);
2932 if (ret)
bfff6873 2933 kfree(elr);
bfff6873
LC		 2934 return ret;
2935}
2936
2937/*
2938 * We do not need to lock anything since this is called on
2939 * module unload.
2940 */
2941static void ext4_destroy_lazyinit_thread(void)
2942{
2943 /*
2944 * If thread exited earlier
2945 * there's nothing to be done.
2946 */
8f1f7453 2947 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2948 return;
2949
8f1f7453 2950 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2951}
2952
25ed6e8a
DW		 2953static int set_journal_csum_feature_set(struct super_block *sb)
2954{
2955 int ret = 1;
2956 int compat, incompat;
2957 struct ext4_sb_info *sbi = EXT4_SB(sb);
2958
9aa5d32b 2959 if (ext4_has_metadata_csum(sb)) {
db9ee220 2960 /* journal checksum v3 */
25ed6e8a 2961 compat = 0;
db9ee220 2962 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 2963 } else {
2964 /* journal checksum v1 */
2965 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2966 incompat = 0;
2967 }
2968
feb8c6d3
DW		 2969 jbd2_journal_clear_features(sbi->s_journal,
2970 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2971 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2972 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 2973 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2974 ret = jbd2_journal_set_features(sbi->s_journal,
2975 compat, 0,
2976 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2977 incompat);
2978 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2979 ret = jbd2_journal_set_features(sbi->s_journal,
2980 compat, 0,
2981 incompat);
2982 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2983 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2984 } else {
feb8c6d3
DW		 2985 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2986 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 2987 }
2988
2989 return ret;
2990}
2991
952fc18e
TT		 2992/*
2993 * Note: calculating the overhead so we can be compatible with
2994 * historical BSD practice is quite difficult in the face of
2995 * clusters/bigalloc. This is because multiple metadata blocks from
2996 * different block group can end up in the same allocation cluster.
2997 * Calculating the exact overhead in the face of clustered allocation
2998 * requires either O(all block bitmaps) in memory or O(number of block
2999 * groups**2) in time. We will still calculate the superblock for
3000 * older file systems --- and if we come across with a bigalloc file
3001 * system with zero in s_overhead_clusters the estimate will be close to
3002 * correct especially for very large cluster sizes --- but for newer
3003 * file systems, it's better to calculate this figure once at mkfs
3004 * time, and store it in the superblock. If the superblock value is
3005 * present (even for non-bigalloc file systems), we will use it.
3006 */
3007static int count_overhead(struct super_block *sb, ext4_group_t grp,
3008 char *buf)
3009{
3010 struct ext4_sb_info *sbi = EXT4_SB(sb);
3011 struct ext4_group_desc *gdp;
3012 ext4_fsblk_t first_block, last_block, b;
3013 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3014 int s, j, count = 0;
3015
e2b911c5 3016 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3017 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3018 sbi->s_itb_per_group + 2);
3019
952fc18e
TT		 3020 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3021 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3022 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3023 for (i = 0; i < ngroups; i++) {
3024 gdp = ext4_get_group_desc(sb, i, NULL);
3025 b = ext4_block_bitmap(sb, gdp);
3026 if (b >= first_block && b <= last_block) {
3027 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3028 count++;
3029 }
3030 b = ext4_inode_bitmap(sb, gdp);
3031 if (b >= first_block && b <= last_block) {
3032 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3033 count++;
3034 }
3035 b = ext4_inode_table(sb, gdp);
3036 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3037 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3038 int c = EXT4_B2C(sbi, b - first_block);
3039 ext4_set_bit(c, buf);
3040 count++;
3041 }
3042 if (i != grp)
3043 continue;
3044 s = 0;
3045 if (ext4_bg_has_super(sb, grp)) {
3046 ext4_set_bit(s++, buf);
3047 count++;
3048 }
3049 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3050 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3051 count++;
3052 }
3053 }
3054 if (!count)
3055 return 0;
3056 return EXT4_CLUSTERS_PER_GROUP(sb) -
3057 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3058}
3059
3060/*
3061 * Compute the overhead and stash it in sbi->s_overhead
3062 */
3063int ext4_calculate_overhead(struct super_block *sb)
3064{
3065 struct ext4_sb_info *sbi = EXT4_SB(sb);
3066 struct ext4_super_block *es = sbi->s_es;
3067 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3068 ext4_fsblk_t overhead = 0;
4fdb5543 3069 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3070
952fc18e
TT		 3071 if (!buf)
3072 return -ENOMEM;
3073
3074 /*
3075 * Compute the overhead (FS structures). This is constant
3076 * for a given filesystem unless the number of block groups
3077 * changes so we cache the previous value until it does.
3078 */
3079
3080 /*
3081 * All of the blocks before first_data_block are overhead
3082 */
3083 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3084
3085 /*
3086 * Add the overhead found in each block group
3087 */
3088 for (i = 0; i < ngroups; i++) {
3089 int blks;
3090
3091 blks = count_overhead(sb, i, buf);
3092 overhead += blks;
3093 if (blks)
3094 memset(buf, 0, PAGE_SIZE);
3095 cond_resched();
3096 }
b003b524
ES		 3097 /* Add the internal journal blocks as well */
3098 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3099 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3100
952fc18e
TT		 3101 sbi->s_overhead = overhead;
3102 smp_wmb();
3103 free_page((unsigned long) buf);
3104 return 0;
3105}
3106
b5799018 3107static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3108{
3109 ext4_fsblk_t resv_clusters;
b5799018 3110 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3111
30fac0f7
JK		 3112 /*
3113 * There's no need to reserve anything when we aren't using extents.
3114 * The space estimates are exact, there are no unwritten extents,
3115 * hole punching doesn't need new metadata... This is needed especially
3116 * to keep ext2/3 backward compatibility.
3117 */
e2b911c5 3118 if (!ext4_has_feature_extents(sb))
b5799018 3119 return;
27dd4385
LC		 3120 /*
3121 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3122 * This should cover the situations where we can not afford to run
3123 * out of space like for example punch hole, or converting
556615dc 3124 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3125 * allocation would require 1, or 2 blocks, higher numbers are
3126 * very rare.
3127 */
b5799018
TT		 3128 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3129 sbi->s_cluster_bits);
27dd4385
LC		 3130
3131 do_div(resv_clusters, 50);
3132 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3133
b5799018 3134 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3135}
3136
2b2d6d01 3137static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3138{
d4c402d9 3139 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3140 struct buffer_head *bh;
617ba13b
MC		 3141 struct ext4_super_block *es = NULL;
3142 struct ext4_sb_info *sbi;
3143 ext4_fsblk_t block;
3144 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3145 ext4_fsblk_t logical_sb_block;
ac27a0ec 3146 unsigned long offset = 0;
ac27a0ec
DK		 3147 unsigned long journal_devnum = 0;
3148 unsigned long def_mount_opts;
3149 struct inode *root;
0390131b 3150 const char *descr;
dcc7dae3 3151 int ret = -ENOMEM;
281b5995 3152 int blocksize, clustersize;
4ec11028
TT		 3153 unsigned int db_count;
3154 unsigned int i;
281b5995 3155 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3156 __u64 blocks_count;
07aa2ea1 3157 int err = 0;
b3881f74 3158 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3159 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3160
3161 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3162 if (!sbi)
dcc7dae3 3163 goto out_free_orig;
705895b6
PE		 3164
3165 sbi->s_blockgroup_lock =
3166 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3167 if (!sbi->s_blockgroup_lock) {
3168 kfree(sbi);
dcc7dae3 3169 goto out_free_orig;
705895b6 3170 }
ac27a0ec 3171 sb->s_fs_info = sbi;
2c0544b2 3172 sbi->s_sb = sb;
240799cd 3173 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3174 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3175 if (sb->s_bdev->bd_part)
3176 sbi->s_sectors_written_start =
3177 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3178
9f6200bb 3179 /* Cleanup superblock name */
ec3904dc 3180 strreplace(sb->s_id, '/', '!');
9f6200bb 3181
07aa2ea1 3182 /* -EINVAL is default */
dcc7dae3 3183 ret = -EINVAL;
617ba13b 3184 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3185 if (!blocksize) {
b31e1552 3186 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3187 goto out_fail;
3188 }
3189
3190 /*
617ba13b 3191 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3192 * block sizes. We need to calculate the offset from buffer start.
3193 */
617ba13b 3194 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3195 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3196 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3197 } else {
70bbb3e0 3198 logical_sb_block = sb_block;
ac27a0ec
DK		 3199 }
3200
a8ac900b 3201 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3202 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3203 goto out_fail;
3204 }
3205 /*
3206 * Note: s_es must be initialized as soon as possible because
617ba13b 3207 * some ext4 macro-instructions depend on its value
ac27a0ec 3208 */
2716b802 3209 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3210 sbi->s_es = es;
3211 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3212 if (sb->s_magic != EXT4_SUPER_MAGIC)
3213 goto cantfind_ext4;
afc32f7e 3214 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3215
feb0ab32 3216 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3217 if (ext4_has_feature_metadata_csum(sb) &&
3218 ext4_has_feature_gdt_csum(sb))
363307e6 3219 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3220 "redundant flags; please run fsck.");
3221
d25425f8
DW		 3222 /* Check for a known checksum algorithm */
3223 if (!ext4_verify_csum_type(sb, es)) {
3224 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3225 "unknown checksum algorithm.");
3226 silent = 1;
3227 goto cantfind_ext4;
3228 }
3229
0441984a 3230 /* Load the checksum driver */
e2b911c5 3231 if (ext4_has_feature_metadata_csum(sb)) {
0441984a
DW		 3232 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3233 if (IS_ERR(sbi->s_chksum_driver)) {
3234 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3235 ret = PTR_ERR(sbi->s_chksum_driver);
3236 sbi->s_chksum_driver = NULL;
3237 goto failed_mount;
3238 }
3239 }
3240
a9c47317
DW		 3241 /* Check superblock checksum */
3242 if (!ext4_superblock_csum_verify(sb, es)) {
3243 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3244 "invalid superblock checksum. Run e2fsck?");
3245 silent = 1;
6a797d27 3246 ret = -EFSBADCRC;
a9c47317
DW		 3247 goto cantfind_ext4;
3248 }
3249
3250 /* Precompute checksum seed for all metadata */
e2b911c5 3251 if (ext4_has_feature_csum_seed(sb))
8c81bd8f
DW		 3252 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3253 else if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3254 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3255 sizeof(es->s_uuid));
3256
ac27a0ec
DK		 3257 /* Set defaults before we parse the mount options */
3258 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3259 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3260 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3261 set_opt(sb, DEBUG);
87f26807 3262 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3263 set_opt(sb, GRPID);
617ba13b 3264 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3265 set_opt(sb, NO_UID32);
ea663336 3266 /* xattr user namespace & acls are now defaulted on */
ea663336 3267 set_opt(sb, XATTR_USER);
03010a33 3268#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3269 set_opt(sb, POSIX_ACL);
2e7842b8 3270#endif
98c1a759
DW		 3271 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3272 if (ext4_has_metadata_csum(sb))
3273 set_opt(sb, JOURNAL_CHECKSUM);
3274
617ba13b 3275 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3276 set_opt(sb, JOURNAL_DATA);
617ba13b 3277 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3278 set_opt(sb, ORDERED_DATA);
617ba13b 3279 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3280 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3281
3282 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3283 set_opt(sb, ERRORS_PANIC);
bb4f397a 3284 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3285 set_opt(sb, ERRORS_CONT);
bb4f397a 3286 else
fd8c37ec 3287 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3288 /* block_validity enabled by default; disable with noblock_validity */
3289 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3290 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3291 set_opt(sb, DISCARD);
ac27a0ec 3292
08cefc7a
EB		 3293 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3294 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3295 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3296 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3297 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3298
8b67f04a 3299 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3300 set_opt(sb, BARRIER);
ac27a0ec 3301
dd919b98
AK		 3302 /*
3303 * enable delayed allocation by default
3304 * Use -o nodelalloc to turn it off
3305 */
bc0b75f7 3306 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3307 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3308 set_opt(sb, DELALLOC);
dd919b98 3309
51ce6511
LC		 3310 /*
3311 * set default s_li_wait_mult for lazyinit, for the case there is
3312 * no mount option specified.
3313 */
3314 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3315
8b67f04a 3316 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3317 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3318 ext4_msg(sb, KERN_WARNING,
3319 "failed to parse options in superblock: %s",
3320 sbi->s_es->s_mount_opts);
3321 }
5a916be1 3322 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3323 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3324 &journal_ioprio, 0))
ac27a0ec
DK		 3325 goto failed_mount;
3326
56889787
TT		 3327 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3328 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3329 "with data=journal disables delayed "
3330 "allocation and O_DIRECT support!\n");
3331 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3332 ext4_msg(sb, KERN_ERR, "can't mount with "
3333 "both data=journal and delalloc");
3334 goto failed_mount;
3335 }
3336 if (test_opt(sb, DIOREAD_NOLOCK)) {
3337 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3338 "both data=journal and dioread_nolock");
56889787
TT		 3339 goto failed_mount;
3340 }
923ae0ff
RZ		 3341 if (test_opt(sb, DAX)) {
3342 ext4_msg(sb, KERN_ERR, "can't mount with "
3343 "both data=journal and dax");
3344 goto failed_mount;
3345 }
56889787
TT		 3346 if (test_opt(sb, DELALLOC))
3347 clear_opt(sb, DELALLOC);
001e4a87
TH		 3348 } else {
3349 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3350 }
3351
ac27a0ec 3352 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3353 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3354
617ba13b 3355 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3356 (ext4_has_compat_features(sb) ||
3357 ext4_has_ro_compat_features(sb) ||
3358 ext4_has_incompat_features(sb)))
b31e1552
ES		 3359 ext4_msg(sb, KERN_WARNING,
3360 "feature flags set on rev 0 fs, "
3361 "running e2fsck is recommended");
469108ff 3362
ed3654eb
TT		 3363 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3364 set_opt2(sb, HURD_COMPAT);
e2b911c5 3365 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3366 ext4_msg(sb, KERN_ERR,
3367 "The Hurd can't support 64-bit file systems");
3368 goto failed_mount;
3369 }
3370 }
3371
2035e776
TT		 3372 if (IS_EXT2_SB(sb)) {
3373 if (ext2_feature_set_ok(sb))
3374 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3375 "using the ext4 subsystem");
3376 else {
3377 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3378 "to feature incompatibilities");
3379 goto failed_mount;
3380 }
3381 }
3382
3383 if (IS_EXT3_SB(sb)) {
3384 if (ext3_feature_set_ok(sb))
3385 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3386 "using the ext4 subsystem");
3387 else {
3388 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3389 "to feature incompatibilities");
3390 goto failed_mount;
3391 }
3392 }
3393
ac27a0ec
DK		 3394 /*
3395 * Check feature flags regardless of the revision level, since we
3396 * previously didn't change the revision level when setting the flags,
3397 * so there is a chance incompat flags are set on a rev 0 filesystem.
3398 */
a13fb1a4 3399 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3400 goto failed_mount;
a13fb1a4 3401
261cb20c 3402 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3403 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3404 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3405 ext4_msg(sb, KERN_ERR,
3406 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3407 goto failed_mount;
3408 }
3409
923ae0ff
RZ		 3410 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3411 if (blocksize != PAGE_SIZE) {
3412 ext4_msg(sb, KERN_ERR,
3413 "error: unsupported blocksize for dax");
3414 goto failed_mount;
3415 }
3416 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3417 ext4_msg(sb, KERN_ERR,
3418 "error: device does not support dax");
3419 goto failed_mount;
3420 }
3421 }
3422
e2b911c5 3423 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3424 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3425 es->s_encryption_level);
3426 goto failed_mount;
3427 }
3428
ac27a0ec 3429 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3430 /* Validate the filesystem blocksize */
3431 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3432 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3433 blocksize);
ac27a0ec
DK		 3434 goto failed_mount;
3435 }
3436
2b2d6d01 3437 brelse(bh);
70bbb3e0
AM		 3438 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3439 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3440 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3441 if (!bh) {
b31e1552
ES		 3442 ext4_msg(sb, KERN_ERR,
3443 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3444 goto failed_mount;
3445 }
2716b802 3446 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3447 sbi->s_es = es;
617ba13b 3448 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3449 ext4_msg(sb, KERN_ERR,
3450 "Magic mismatch, very weird!");
ac27a0ec
DK		 3451 goto failed_mount;
3452 }
3453 }
3454
e2b911c5 3455 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3456 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3457 has_huge_files);
3458 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3459
617ba13b
MC		 3460 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3461 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3462 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3463 } else {
3464 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3465 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3466 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3467 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3468 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3469 ext4_msg(sb, KERN_ERR,
3470 "unsupported inode size: %d",
2b2d6d01 3471 sbi->s_inode_size);
ac27a0ec
DK		 3472 goto failed_mount;
3473 }
ef7f3835
KS		 3474 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3475 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3476 }
0b8e58a1 3477
0d1ee42f 3478 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3479 if (ext4_has_feature_64bit(sb)) {
8fadc143 3480 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3481 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3482 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3483 ext4_msg(sb, KERN_ERR,
3484 "unsupported descriptor size %lu",
0d1ee42f
AR		 3485 sbi->s_desc_size);
3486 goto failed_mount;
3487 }
3488 } else
3489 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3490
ac27a0ec 3491 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3492 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3493 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3494 goto cantfind_ext4;
0b8e58a1 3495
617ba13b 3496 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3497 if (sbi->s_inodes_per_block == 0)
617ba13b 3498 goto cantfind_ext4;
ac27a0ec
DK		 3499 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3500 sbi->s_inodes_per_block;
0d1ee42f 3501 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3502 sbi->s_sbh = bh;
3503 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3504 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3505 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3506
2b2d6d01 3507 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3508 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3509 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3510 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3511 i = le32_to_cpu(es->s_flags);
3512 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3513 sbi->s_hash_unsigned = 3;
3514 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3515#ifdef __CHAR_UNSIGNED__
23301410
TT		 3516 if (!(sb->s_flags & MS_RDONLY))
3517 es->s_flags |=
3518 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3519 sbi->s_hash_unsigned = 3;
f99b2589 3520#else
23301410
TT		 3521 if (!(sb->s_flags & MS_RDONLY))
3522 es->s_flags |=
3523 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3524#endif
23301410 3525 }
f99b2589 3526 }
ac27a0ec 3527
281b5995
TT		 3528 /* Handle clustersize */
3529 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3530 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3531 if (has_bigalloc) {
3532 if (clustersize < blocksize) {
3533 ext4_msg(sb, KERN_ERR,
3534 "cluster size (%d) smaller than "
3535 "block size (%d)", clustersize, blocksize);
3536 goto failed_mount;
3537 }
3538 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3539 le32_to_cpu(es->s_log_block_size);
3540 sbi->s_clusters_per_group =
3541 le32_to_cpu(es->s_clusters_per_group);
3542 if (sbi->s_clusters_per_group > blocksize * 8) {
3543 ext4_msg(sb, KERN_ERR,
3544 "#clusters per group too big: %lu",
3545 sbi->s_clusters_per_group);
3546 goto failed_mount;
3547 }
3548 if (sbi->s_blocks_per_group !=
3549 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3550 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3551 "clusters per group (%lu) inconsistent",
3552 sbi->s_blocks_per_group,
3553 sbi->s_clusters_per_group);
3554 goto failed_mount;
3555 }
3556 } else {
3557 if (clustersize != blocksize) {
3558 ext4_warning(sb, "fragment/cluster size (%d) != "
3559 "block size (%d)", clustersize,
3560 blocksize);
3561 clustersize = blocksize;
3562 }
3563 if (sbi->s_blocks_per_group > blocksize * 8) {
3564 ext4_msg(sb, KERN_ERR,
3565 "#blocks per group too big: %lu",
3566 sbi->s_blocks_per_group);
3567 goto failed_mount;
3568 }
3569 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3570 sbi->s_cluster_bits = 0;
ac27a0ec 3571 }
281b5995
TT		 3572 sbi->s_cluster_ratio = clustersize / blocksize;
3573
ac27a0ec 3574 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3575 ext4_msg(sb, KERN_ERR,
3576 "#inodes per group too big: %lu",
2b2d6d01 3577 sbi->s_inodes_per_group);
ac27a0ec
DK		 3578 goto failed_mount;
3579 }
3580
960fd856
TT		 3581 /* Do we have standard group size of clustersize * 8 blocks ? */
3582 if (sbi->s_blocks_per_group == clustersize << 3)
3583 set_opt2(sb, STD_GROUP_SIZE);
3584
bf43d84b
ES		 3585 /*
3586 * Test whether we have more sectors than will fit in sector_t,
3587 * and whether the max offset is addressable by the page cache.
3588 */
5a9ae68a 3589 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3590 ext4_blocks_count(es));
5a9ae68a 3591 if (err) {
b31e1552 3592 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3593 " too large to mount safely on this system");
ac27a0ec 3594 if (sizeof(sector_t) < 8)
90c699a9 3595 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3596 goto failed_mount;
3597 }
3598
617ba13b
MC		 3599 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3600 goto cantfind_ext4;
e7c95593 3601
0f2ddca6
FTN		 3602 /* check blocks count against device size */
3603 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3604 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3605 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3606 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3607 ext4_blocks_count(es), blocks_count);
3608 goto failed_mount;
3609 }
3610
0b8e58a1
AD		 3611 /*
3612 * It makes no sense for the first data block to be beyond the end
3613 * of the filesystem.
3614 */
3615 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3616 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3617 "block %u is beyond end of filesystem (%llu)",
3618 le32_to_cpu(es->s_first_data_block),
3619 ext4_blocks_count(es));
e7c95593
ES		 3620 goto failed_mount;
3621 }
bd81d8ee
LV		 3622 blocks_count = (ext4_blocks_count(es) -
3623 le32_to_cpu(es->s_first_data_block) +
3624 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3625 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3626 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3627 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3628 "(block count %llu, first data block %u, "
b31e1552 3629 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3630 ext4_blocks_count(es),
3631 le32_to_cpu(es->s_first_data_block),
3632 EXT4_BLOCKS_PER_GROUP(sb));
3633 goto failed_mount;
3634 }
bd81d8ee 3635 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3636 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3637 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3638 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3639 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3640 sbi->s_group_desc = ext4_kvmalloc(db_count *
3641 sizeof(struct buffer_head *),
3642 GFP_KERNEL);
ac27a0ec 3643 if (sbi->s_group_desc == NULL) {
b31e1552 3644 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3645 ret = -ENOMEM;
ac27a0ec
DK		 3646 goto failed_mount;
3647 }
3648
705895b6 3649 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3650
3651 for (i = 0; i < db_count; i++) {
70bbb3e0 3652 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3653 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3654 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3655 ext4_msg(sb, KERN_ERR,
3656 "can't read group descriptor %d", i);
ac27a0ec
DK		 3657 db_count = i;
3658 goto failed_mount2;
3659 }
3660 }
bfff6873 3661 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3662 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 3663 ret = -EFSCORRUPTED;
f9ae9cf5 3664 goto failed_mount2;
ac27a0ec 3665 }
772cb7c8 3666
f9ae9cf5 3667 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3668 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3669 spin_lock_init(&sbi->s_next_gen_lock);
3670
04ecddb7
JM		 3671 setup_timer(&sbi->s_err_report, print_daily_error_info,
3672 (unsigned long) sb);
04496411 3673
a75ae78f 3674 /* Register extent status tree shrinker */
eb68d0e2 3675 if (ext4_es_register_shrinker(sbi))
ce7e010a 3676 goto failed_mount3;
ce7e010a 3677
c9de560d 3678 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3679 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3680
f9ae9cf5
TT		 3681 /*
3682 * set up enough so that it can read an inode
3683 */
f6e63f90 3684 sb->s_op = &ext4_sops;
617ba13b
MC		 3685 sb->s_export_op = &ext4_export_ops;
3686 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3687#ifdef CONFIG_QUOTA
617ba13b 3688 sb->dq_op = &ext4_quota_operations;
e2b911c5 3689 if (ext4_has_feature_quota(sb))
1fa5efe3 3690 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3691 else
3692 sb->s_qcop = &ext4_qctl_operations;
689c958c 3693 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 3694#endif
f2fa2ffc
AK		 3695 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3696
ac27a0ec 3697 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3698 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3699
3700 sb->s_root = NULL;
3701
3702 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 3703 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 3704
e2b911c5 3705 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
c5e06d10 3706 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3707 goto failed_mount3a;
c5e06d10 3708
ac27a0ec
DK		 3709 /*
3710 * The first inode we look at is the journal inode. Don't try
3711 * root first: it may be modified in the journal!
3712 */
e2b911c5 3713 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
617ba13b 3714 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3715 goto failed_mount3a;
0390131b 3716 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
e2b911c5 3717 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 3718 ext4_msg(sb, KERN_ERR, "required journal recovery "
3719 "suppressed and not mounted read-only");
744692dc 3720 goto failed_mount_wq;
ac27a0ec 3721 } else {
1e381f60
DM		 3722 /* Nojournal mode, all journal mount options are illegal */
3723 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3724 ext4_msg(sb, KERN_ERR, "can't mount with "
3725 "journal_checksum, fs mounted w/o journal");
3726 goto failed_mount_wq;
3727 }
3728 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3729 ext4_msg(sb, KERN_ERR, "can't mount with "
3730 "journal_async_commit, fs mounted w/o journal");
3731 goto failed_mount_wq;
3732 }
3733 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3734 ext4_msg(sb, KERN_ERR, "can't mount with "
3735 "commit=%lu, fs mounted w/o journal",
3736 sbi->s_commit_interval / HZ);
3737 goto failed_mount_wq;
3738 }
3739 if (EXT4_MOUNT_DATA_FLAGS &
3740 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3741 ext4_msg(sb, KERN_ERR, "can't mount with "
3742 "data=, fs mounted w/o journal");
3743 goto failed_mount_wq;
3744 }
3745 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3746 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 3747 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3748 sbi->s_journal = NULL;
3749 needs_recovery = 0;
3750 goto no_journal;
ac27a0ec
DK		 3751 }
3752
e2b911c5 3753 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 3754 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3755 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3756 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3757 goto failed_mount_wq;
eb40a09c
JS		 3758 }
3759
25ed6e8a
DW		 3760 if (!set_journal_csum_feature_set(sb)) {
3761 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3762 "feature set");
3763 goto failed_mount_wq;
d4da6c9c 3764 }
818d276c 3765
ac27a0ec
DK		 3766 /* We have now updated the journal if required, so we can
3767 * validate the data journaling mode. */
3768 switch (test_opt(sb, DATA_FLAGS)) {
3769 case 0:
3770 /* No mode set, assume a default based on the journal
63f57933
AM		 3771 * capabilities: ORDERED_DATA if the journal can
3772 * cope, else JOURNAL_DATA
3773 */
dab291af
MC		 3774 if (jbd2_journal_check_available_features
3775 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3776 set_opt(sb, ORDERED_DATA);
ac27a0ec 3777 else
fd8c37ec 3778 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3779 break;
3780
617ba13b
MC		 3781 case EXT4_MOUNT_ORDERED_DATA:
3782 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3783 if (!jbd2_journal_check_available_features
3784 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3785 ext4_msg(sb, KERN_ERR, "Journal does not support "
3786 "requested data journaling mode");
744692dc 3787 goto failed_mount_wq;
ac27a0ec
DK		 3788 }
3789 default:
3790 break;
3791 }
b3881f74 3792 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3793
18aadd47
BJ		 3794 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3795
ce7e010a 3796no_journal:
2335d05f
AG		 3797 sbi->s_mb_cache = ext4_xattr_create_cache();
3798 if (!sbi->s_mb_cache) {
3799 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3800 goto failed_mount_wq;
9c191f70
M		 3801 }
3802
e2b911c5 3803 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
1cb767cd
TT		 3804 (blocksize != PAGE_CACHE_SIZE)) {
3805 ext4_msg(sb, KERN_ERR,
3806 "Unsupported blocksize for fs encryption");
3807 goto failed_mount_wq;
3808 }
3809
e2b911c5
DW		 3810 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3811 !ext4_has_feature_encrypt(sb)) {
3812 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 3813 ext4_commit_super(sb, 1);
3814 }
3815
952fc18e
TT		 3816 /*
3817 * Get the # of file system overhead blocks from the
3818 * superblock if present.
3819 */
3820 if (es->s_overhead_clusters)
3821 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3822 else {
07aa2ea1
LC		 3823 err = ext4_calculate_overhead(sb);
3824 if (err)
952fc18e
TT		 3825 goto failed_mount_wq;
3826 }
3827
fd89d5f2
TH		 3828 /*
3829 * The maximum number of concurrent works can be high and
3830 * concurrency isn't really necessary. Limit it to 1.
3831 */
2e8fa54e
JK		 3832 EXT4_SB(sb)->rsv_conversion_wq =
3833 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3834 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3835 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 3836 ret = -ENOMEM;
2e8fa54e
JK		 3837 goto failed_mount4;
3838 }
3839
ac27a0ec 3840 /*
dab291af 3841 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3842 * so we can safely mount the rest of the filesystem now.
3843 */
3844
1d1fe1ee
DH		 3845 root = ext4_iget(sb, EXT4_ROOT_INO);
3846 if (IS_ERR(root)) {
b31e1552 3847 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3848 ret = PTR_ERR(root);
32a9bb57 3849 root = NULL;
ac27a0ec
DK		 3850 goto failed_mount4;
3851 }
3852 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3853 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3854 iput(root);
ac27a0ec
DK		 3855 goto failed_mount4;
3856 }
48fde701 3857 sb->s_root = d_make_root(root);
1d1fe1ee 3858 if (!sb->s_root) {
b31e1552 3859 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3860 ret = -ENOMEM;
3861 goto failed_mount4;
3862 }
ac27a0ec 3863
7e84b621
ES		 3864 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3865 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3866
3867 /* determine the minimum size of new large inodes, if present */
3868 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3869 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3870 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 3871 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 3872 if (sbi->s_want_extra_isize <
3873 le16_to_cpu(es->s_want_extra_isize))
3874 sbi->s_want_extra_isize =
3875 le16_to_cpu(es->s_want_extra_isize);
3876 if (sbi->s_want_extra_isize <
3877 le16_to_cpu(es->s_min_extra_isize))
3878 sbi->s_want_extra_isize =
3879 le16_to_cpu(es->s_min_extra_isize);
3880 }
3881 }
3882 /* Check if enough inode space is available */
3883 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3884 sbi->s_inode_size) {
3885 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3886 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3887 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3888 "available");
ef7f3835
KS		 3889 }
3890
b5799018 3891 ext4_set_resv_clusters(sb);
27dd4385 3892
6fd058f7
TT		 3893 err = ext4_setup_system_zone(sb);
3894 if (err) {
b31e1552 3895 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3896 "zone (%d)", err);
f9ae9cf5
TT		 3897 goto failed_mount4a;
3898 }
3899
3900 ext4_ext_init(sb);
3901 err = ext4_mb_init(sb);
3902 if (err) {
3903 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3904 err);
dcf2d804 3905 goto failed_mount5;
c2774d84
AK		 3906 }
3907
d5e03cbb
TT		 3908 block = ext4_count_free_clusters(sb);
3909 ext4_free_blocks_count_set(sbi->s_es,
3910 EXT4_C2B(sbi, block));
908c7f19
TH		 3911 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3912 GFP_KERNEL);
d5e03cbb
TT		 3913 if (!err) {
3914 unsigned long freei = ext4_count_free_inodes(sb);
3915 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 3916 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3917 GFP_KERNEL);
d5e03cbb
TT		 3918 }
3919 if (!err)
3920 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 3921 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 3922 if (!err)
908c7f19
TH		 3923 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3924 GFP_KERNEL);
d5e03cbb
TT		 3925 if (err) {
3926 ext4_msg(sb, KERN_ERR, "insufficient memory");
3927 goto failed_mount6;
3928 }
3929
e2b911c5 3930 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 3931 if (!ext4_fill_flex_info(sb)) {
3932 ext4_msg(sb, KERN_ERR,
3933 "unable to initialize "
3934 "flex_bg meta info!");
3935 goto failed_mount6;
3936 }
3937
bfff6873
LC		 3938 err = ext4_register_li_request(sb, first_not_zeroed);
3939 if (err)
dcf2d804 3940 goto failed_mount6;
bfff6873 3941
b5799018 3942 err = ext4_register_sysfs(sb);
dcf2d804
TM		 3943 if (err)
3944 goto failed_mount7;
3197ebdb 3945
9b2ff357
JK		 3946#ifdef CONFIG_QUOTA
3947 /* Enable quota usage during mount. */
e2b911c5 3948 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
9b2ff357
JK		 3949 err = ext4_enable_quotas(sb);
3950 if (err)
3951 goto failed_mount8;
3952 }
3953#endif /* CONFIG_QUOTA */
3954
617ba13b
MC		 3955 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3956 ext4_orphan_cleanup(sb, es);
3957 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3958 if (needs_recovery) {
b31e1552 3959 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3960 ext4_mark_recovery_complete(sb, es);
3961 }
3962 if (EXT4_SB(sb)->s_journal) {
3963 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3964 descr = " journalled data mode";
3965 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3966 descr = " ordered data mode";
3967 else
3968 descr = " writeback data mode";
3969 } else
3970 descr = "out journal";
3971
79add3a3
LC		 3972 if (test_opt(sb, DISCARD)) {
3973 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3974 if (!blk_queue_discard(q))
3975 ext4_msg(sb, KERN_WARNING,
3976 "mounting with \"discard\" option, but "
3977 "the device does not support discard");
3978 }
3979
e294a537
TT		 3980 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3981 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3982 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3983 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3984
66e61a9e
TT		 3985 if (es->s_error_count)
3986 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3987
efbed4dc
TT		 3988 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3989 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
3990 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
3991 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
3992
d4c402d9 3993 kfree(orig_data);
ac27a0ec
DK		 3994 return 0;
3995
617ba13b 3996cantfind_ext4:
ac27a0ec 3997 if (!silent)
b31e1552 3998 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3999 goto failed_mount;
4000
72ba7450
TT		 4001#ifdef CONFIG_QUOTA
4002failed_mount8:
ebd173be 4003 ext4_unregister_sysfs(sb);
72ba7450 4004#endif
dcf2d804
TM		 4005failed_mount7:
4006 ext4_unregister_li_request(sb);
4007failed_mount6:
f9ae9cf5 4008 ext4_mb_release(sb);
d5e03cbb 4009 if (sbi->s_flex_groups)
b93b41d4 4010 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4011 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4012 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4013 percpu_counter_destroy(&sbi->s_dirs_counter);
4014 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4015failed_mount5:
f9ae9cf5
TT		 4016 ext4_ext_release(sb);
4017 ext4_release_system_zone(sb);
4018failed_mount4a:
94bf608a 4019 dput(sb->s_root);
32a9bb57 4020 sb->s_root = NULL;
94bf608a 4021failed_mount4:
b31e1552 4022 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4023 if (EXT4_SB(sb)->rsv_conversion_wq)
4024 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4025failed_mount_wq:
82939d79
JK		 4026 if (sbi->s_mb_cache) {
4027 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4028 sbi->s_mb_cache = NULL;
4029 }
0390131b
FM		 4030 if (sbi->s_journal) {
4031 jbd2_journal_destroy(sbi->s_journal);
4032 sbi->s_journal = NULL;
4033 }
50460fe8 4034failed_mount3a:
d3922a77 4035 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4036failed_mount3:
9105bb14 4037 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4038 if (sbi->s_mmp_tsk)
4039 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4040failed_mount2:
4041 for (i = 0; i < db_count; i++)
4042 brelse(sbi->s_group_desc[i]);
b93b41d4 4043 kvfree(sbi->s_group_desc);
ac27a0ec 4044failed_mount:
0441984a
DW		 4045 if (sbi->s_chksum_driver)
4046 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4047#ifdef CONFIG_QUOTA
a2d4a646 4048 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4049 kfree(sbi->s_qf_names[i]);
4050#endif
617ba13b 4051 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4052 brelse(bh);
4053out_fail:
4054 sb->s_fs_info = NULL;
f6830165 4055 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4056 kfree(sbi);
dcc7dae3 4057out_free_orig:
d4c402d9 4058 kfree(orig_data);
07aa2ea1 4059 return err ? err : ret;
ac27a0ec
DK		 4060}
4061
4062/*
4063 * Setup any per-fs journal parameters now. We'll do this both on
4064 * initial mount, once the journal has been initialised but before we've
4065 * done any recovery; and again on any subsequent remount.
4066 */
617ba13b 4067static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4068{
617ba13b 4069 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4070
30773840
TT		 4071 journal->j_commit_interval = sbi->s_commit_interval;
4072 journal->j_min_batch_time = sbi->s_min_batch_time;
4073 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4074
a931da6a 4075 write_lock(&journal->j_state_lock);
ac27a0ec 4076 if (test_opt(sb, BARRIER))
dab291af 4077 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4078 else
dab291af 4079 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4080 if (test_opt(sb, DATA_ERR_ABORT))
4081 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4082 else
4083 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4084 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4085}
4086
617ba13b 4087static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4088 unsigned int journal_inum)
4089{
4090 struct inode *journal_inode;
4091 journal_t *journal;
4092
e2b911c5 4093 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4094
ac27a0ec
DK		 4095 /* First, test for the existence of a valid inode on disk. Bad
4096 * things happen if we iget() an unused inode, as the subsequent
4097 * iput() will try to delete it. */
4098
1d1fe1ee
DH		 4099 journal_inode = ext4_iget(sb, journal_inum);
4100 if (IS_ERR(journal_inode)) {
b31e1552 4101 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4102 return NULL;
4103 }
4104 if (!journal_inode->i_nlink) {
4105 make_bad_inode(journal_inode);
4106 iput(journal_inode);
b31e1552 4107 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4108 return NULL;
4109 }
4110
e5f8eab8 4111 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4112 journal_inode, journal_inode->i_size);
1d1fe1ee 4113 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4114 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4115 iput(journal_inode);
4116 return NULL;
4117 }
4118
dab291af 4119 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4120 if (!journal) {
b31e1552 4121 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4122 iput(journal_inode);
4123 return NULL;
4124 }
4125 journal->j_private = sb;
617ba13b 4126 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4127 return journal;
4128}
4129
617ba13b 4130static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4131 dev_t j_dev)
4132{
2b2d6d01 4133 struct buffer_head *bh;
ac27a0ec 4134 journal_t *journal;
617ba13b
MC		 4135 ext4_fsblk_t start;
4136 ext4_fsblk_t len;
ac27a0ec 4137 int hblock, blocksize;
617ba13b 4138 ext4_fsblk_t sb_block;
ac27a0ec 4139 unsigned long offset;
2b2d6d01 4140 struct ext4_super_block *es;
ac27a0ec
DK		 4141 struct block_device *bdev;
4142
e2b911c5 4143 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4144
b31e1552 4145 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4146 if (bdev == NULL)
4147 return NULL;
4148
ac27a0ec 4149 blocksize = sb->s_blocksize;
e1defc4f 4150 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4151 if (blocksize < hblock) {
b31e1552
ES		 4152 ext4_msg(sb, KERN_ERR,
4153 "blocksize too small for journal device");
ac27a0ec
DK		 4154 goto out_bdev;
4155 }
4156
617ba13b
MC		 4157 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4158 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4159 set_blocksize(bdev, blocksize);
4160 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4161 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4162 "external journal");
ac27a0ec
DK		 4163 goto out_bdev;
4164 }
4165
2716b802 4166 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4167 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4168 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4169 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4170 ext4_msg(sb, KERN_ERR, "external journal has "
4171 "bad superblock");
ac27a0ec
DK		 4172 brelse(bh);
4173 goto out_bdev;
4174 }
4175
df4763be
DW		 4176 if ((le32_to_cpu(es->s_feature_ro_compat) &
4177 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4178 es->s_checksum != ext4_superblock_csum(sb, es)) {
4179 ext4_msg(sb, KERN_ERR, "external journal has "
4180 "corrupt superblock");
4181 brelse(bh);
4182 goto out_bdev;
4183 }
4184
617ba13b 4185 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4186 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4187 brelse(bh);
4188 goto out_bdev;
4189 }
4190
bd81d8ee 4191 len = ext4_blocks_count(es);
ac27a0ec
DK		 4192 start = sb_block + 1;
4193 brelse(bh); /* we're done with the superblock */
4194
dab291af 4195 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4196 start, len, blocksize);
4197 if (!journal) {
b31e1552 4198 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4199 goto out_bdev;
4200 }
4201 journal->j_private = sb;
9f203507 4202 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4203 wait_on_buffer(journal->j_sb_buffer);
4204 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4205 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4206 goto out_journal;
4207 }
4208 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4209 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4210 "user (unsupported) - %d",
ac27a0ec
DK		 4211 be32_to_cpu(journal->j_superblock->s_nr_users));
4212 goto out_journal;
4213 }
617ba13b
MC		 4214 EXT4_SB(sb)->journal_bdev = bdev;
4215 ext4_init_journal_params(sb, journal);
ac27a0ec 4216 return journal;
0b8e58a1 4217
ac27a0ec 4218out_journal:
dab291af 4219 jbd2_journal_destroy(journal);
ac27a0ec 4220out_bdev:
617ba13b 4221 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4222 return NULL;
4223}
4224
617ba13b
MC		 4225static int ext4_load_journal(struct super_block *sb,
4226 struct ext4_super_block *es,
ac27a0ec
DK		 4227 unsigned long journal_devnum)
4228{
4229 journal_t *journal;
4230 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4231 dev_t journal_dev;
4232 int err = 0;
4233 int really_read_only;
4234
e2b911c5 4235 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4236
ac27a0ec
DK		 4237 if (journal_devnum &&
4238 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4239 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4240 "numbers have changed");
ac27a0ec
DK		 4241 journal_dev = new_decode_dev(journal_devnum);
4242 } else
4243 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4244
4245 really_read_only = bdev_read_only(sb->s_bdev);
4246
4247 /*
4248 * Are we loading a blank journal or performing recovery after a
4249 * crash? For recovery, we need to check in advance whether we
4250 * can get read-write access to the device.
4251 */
e2b911c5 4252 if (ext4_has_feature_journal_needs_recovery(sb)) {
ac27a0ec 4253 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4254 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4255 "required on readonly filesystem");
ac27a0ec 4256 if (really_read_only) {
b31e1552
ES		 4257 ext4_msg(sb, KERN_ERR, "write access "
4258 "unavailable, cannot proceed");
ac27a0ec
DK		 4259 return -EROFS;
4260 }
b31e1552
ES		 4261 ext4_msg(sb, KERN_INFO, "write access will "
4262 "be enabled during recovery");
ac27a0ec
DK		 4263 }
4264 }
4265
4266 if (journal_inum && journal_dev) {
b31e1552
ES		 4267 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4268 "and inode journals!");
ac27a0ec
DK		 4269 return -EINVAL;
4270 }
4271
4272 if (journal_inum) {
617ba13b 4273 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4274 return -EINVAL;
4275 } else {
617ba13b 4276 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4277 return -EINVAL;
4278 }
4279
90576c0b 4280 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4281 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4282
e2b911c5 4283 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4284 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4285 if (!err) {
4286 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4287 if (save)
4288 memcpy(save, ((char *) es) +
4289 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4290 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4291 if (save)
4292 memcpy(((char *) es) + EXT4_S_ERR_START,
4293 save, EXT4_S_ERR_LEN);
4294 kfree(save);
4295 }
ac27a0ec
DK		 4296
4297 if (err) {
b31e1552 4298 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4299 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4300 return err;
4301 }
4302
617ba13b
MC		 4303 EXT4_SB(sb)->s_journal = journal;
4304 ext4_clear_journal_err(sb, es);
ac27a0ec 4305
c41303ce 4306 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4307 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4308 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4309
4310 /* Make sure we flush the recovery flag to disk. */
e2d67052 4311 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4312 }
4313
4314 return 0;
4315}
4316
e2d67052 4317static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4318{
e2d67052 4319 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4320 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4321 int error = 0;
ac27a0ec 4322
bdfe0cbd 4323 if (!sbh || block_device_ejected(sb))
c4be0c1d 4324 return error;
914258bf
TT		 4325 if (buffer_write_io_error(sbh)) {
4326 /*
4327 * Oh, dear. A previous attempt to write the
4328 * superblock failed. This could happen because the
4329 * USB device was yanked out. Or it could happen to
4330 * be a transient write error and maybe the block will
4331 * be remapped. Nothing we can do but to retry the
4332 * write and hope for the best.
4333 */
b31e1552
ES		 4334 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4335 "superblock detected");
914258bf
TT		 4336 clear_buffer_write_io_error(sbh);
4337 set_buffer_uptodate(sbh);
4338 }
71290b36
TT		 4339 /*
4340 * If the file system is mounted read-only, don't update the
4341 * superblock write time. This avoids updating the superblock
4342 * write time when we are mounting the root file system
4343 * read/only but we need to replay the journal; at that point,
4344 * for people who are east of GMT and who make their clock
4345 * tick in localtime for Windows bug-for-bug compatibility,
4346 * the clock is set in the future, and this will cause e2fsck
4347 * to complain and force a full file system check.
4348 */
4349 if (!(sb->s_flags & MS_RDONLY))
4350 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4351 if (sb->s_bdev->bd_part)
4352 es->s_kbytes_written =
4353 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4354 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4355 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4356 else
4357 es->s_kbytes_written =
4358 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4359 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4360 ext4_free_blocks_count_set(es,
57042651
TT		 4361 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4362 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4363 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4364 es->s_free_inodes_count =
4365 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4366 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4367 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4368 ext4_superblock_csum_set(sb);
ac27a0ec 4369 mark_buffer_dirty(sbh);
914258bf 4370 if (sync) {
564bc402
DJ		 4371 error = __sync_dirty_buffer(sbh,
4372 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
c4be0c1d
TS		 4373 if (error)
4374 return error;
4375
4376 error = buffer_write_io_error(sbh);
4377 if (error) {
b31e1552
ES		 4378 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4379 "superblock");
914258bf
TT		 4380 clear_buffer_write_io_error(sbh);
4381 set_buffer_uptodate(sbh);
4382 }
4383 }
c4be0c1d 4384 return error;
ac27a0ec
DK		 4385}
4386
ac27a0ec
DK		 4387/*
4388 * Have we just finished recovery? If so, and if we are mounting (or
4389 * remounting) the filesystem readonly, then we will end up with a
4390 * consistent fs on disk. Record that fact.
4391 */
2b2d6d01
TT		 4392static void ext4_mark_recovery_complete(struct super_block *sb,
4393 struct ext4_super_block *es)
ac27a0ec 4394{
617ba13b 4395 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4396
e2b911c5 4397 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4398 BUG_ON(journal != NULL);
4399 return;
4400 }
dab291af 4401 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4402 if (jbd2_journal_flush(journal) < 0)
4403 goto out;
4404
e2b911c5 4405 if (ext4_has_feature_journal_needs_recovery(sb) &&
ac27a0ec 4406 sb->s_flags & MS_RDONLY) {
e2b911c5 4407 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4408 ext4_commit_super(sb, 1);
ac27a0ec 4409 }
7ffe1ea8
HK		 4410
4411out:
dab291af 4412 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4413}
4414
4415/*
4416 * If we are mounting (or read-write remounting) a filesystem whose journal
4417 * has recorded an error from a previous lifetime, move that error to the
4418 * main filesystem now.
4419 */
2b2d6d01
TT		 4420static void ext4_clear_journal_err(struct super_block *sb,
4421 struct ext4_super_block *es)
ac27a0ec
DK		 4422{
4423 journal_t *journal;
4424 int j_errno;
4425 const char *errstr;
4426
e2b911c5 4427 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4428
617ba13b 4429 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4430
4431 /*
4432 * Now check for any error status which may have been recorded in the
617ba13b 4433 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4434 */
4435
dab291af 4436 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4437 if (j_errno) {
4438 char nbuf[16];
4439
617ba13b 4440 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4441 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4442 "from previous mount: %s", errstr);
12062ddd 4443 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4444
617ba13b
MC		 4445 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4446 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4447 ext4_commit_super(sb, 1);
ac27a0ec 4448
dab291af 4449 jbd2_journal_clear_err(journal);
d796c52e 4450 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4451 }
4452}
4453
4454/*
4455 * Force the running and committing transactions to commit,
4456 * and wait on the commit.
4457 */
617ba13b 4458int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4459{
4460 journal_t *journal;
ac27a0ec
DK		 4461
4462 if (sb->s_flags & MS_RDONLY)
4463 return 0;
4464
617ba13b 4465 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4466 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4467}
4468
617ba13b 4469static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4470{
14ce0cb4 4471 int ret = 0;
9eddacf9 4472 tid_t target;
06a407f1 4473 bool needs_barrier = false;
8d5d02e6 4474 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4475
9bffad1e 4476 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4477 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4478 /*
4479 * Writeback quota in non-journalled quota case - journalled quota has
4480 * no dirty dquots
4481 */
4482 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4483 /*
4484 * Data writeback is possible w/o journal transaction, so barrier must
4485 * being sent at the end of the function. But we can skip it if
4486 * transaction_commit will do it for us.
4487 */
bda32530
TT		 4488 if (sbi->s_journal) {
4489 target = jbd2_get_latest_transaction(sbi->s_journal);
4490 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4491 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4492 needs_barrier = true;
4493
4494 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4495 if (wait)
4496 ret = jbd2_log_wait_commit(sbi->s_journal,
4497 target);
4498 }
4499 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4500 needs_barrier = true;
06a407f1
DM		 4501 if (needs_barrier) {
4502 int err;
4503 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4504 if (!ret)
4505 ret = err;
0390131b 4506 }
06a407f1
DM		 4507
4508 return ret;
4509}
4510
ac27a0ec
DK		 4511/*
4512 * LVM calls this function before a (read-only) snapshot is created. This
4513 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4514 *
4515 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4516 * state independently. It relies on upper layer to stop all data & metadata
4517 * modifications.
ac27a0ec 4518 */
c4be0c1d 4519static int ext4_freeze(struct super_block *sb)
ac27a0ec 4520{
c4be0c1d
TS		 4521 int error = 0;
4522 journal_t *journal;
ac27a0ec 4523
9ca92389
TT		 4524 if (sb->s_flags & MS_RDONLY)
4525 return 0;
ac27a0ec 4526
9ca92389 4527 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4528
bb044576
TT		 4529 if (journal) {
4530 /* Now we set up the journal barrier. */
4531 jbd2_journal_lock_updates(journal);
ac27a0ec 4532
bb044576
TT		 4533 /*
4534 * Don't clear the needs_recovery flag if we failed to
4535 * flush the journal.
4536 */
4537 error = jbd2_journal_flush(journal);
4538 if (error < 0)
4539 goto out;
c642dc9e
ES		 4540
4541 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4542 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4543 }
9ca92389 4544
9ca92389 4545 error = ext4_commit_super(sb, 1);
6b0310fb 4546out:
bb044576
TT		 4547 if (journal)
4548 /* we rely on upper layer to stop further updates */
4549 jbd2_journal_unlock_updates(journal);
6b0310fb 4550 return error;
ac27a0ec
DK		 4551}
4552
4553/*
4554 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4555 * flag here, even though the filesystem is not technically dirty yet.
4556 */
c4be0c1d 4557static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4558{
9ca92389
TT		 4559 if (sb->s_flags & MS_RDONLY)
4560 return 0;
4561
c642dc9e
ES		 4562 if (EXT4_SB(sb)->s_journal) {
4563 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4564 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4565 }
4566
9ca92389 4567 ext4_commit_super(sb, 1);
c4be0c1d 4568 return 0;
ac27a0ec
DK		 4569}
4570
673c6100
TT		 4571/*
4572 * Structure to save mount options for ext4_remount's benefit
4573 */
4574struct ext4_mount_options {
4575 unsigned long s_mount_opt;
a2595b8a 4576 unsigned long s_mount_opt2;
08cefc7a
EB		 4577 kuid_t s_resuid;
4578 kgid_t s_resgid;
673c6100
TT		 4579 unsigned long s_commit_interval;
4580 u32 s_min_batch_time, s_max_batch_time;
4581#ifdef CONFIG_QUOTA
4582 int s_jquota_fmt;
a2d4a646 4583 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4584#endif
4585};
4586
2b2d6d01 4587static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4588{
2b2d6d01 4589 struct ext4_super_block *es;
617ba13b 4590 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4591 unsigned long old_sb_flags;
617ba13b 4592 struct ext4_mount_options old_opts;
c79d967d 4593 int enable_quota = 0;
8a266467 4594 ext4_group_t g;
b3881f74 4595 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4596 int err = 0;
ac27a0ec 4597#ifdef CONFIG_QUOTA
03dafb5f 4598 int i, j;
ac27a0ec 4599#endif
d4c402d9 4600 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4601
4602 /* Store the original options */
4603 old_sb_flags = sb->s_flags;
4604 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4605 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4606 old_opts.s_resuid = sbi->s_resuid;
4607 old_opts.s_resgid = sbi->s_resgid;
4608 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4609 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4610 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4611#ifdef CONFIG_QUOTA
4612 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4613 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4614 if (sbi->s_qf_names[i]) {
4615 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4616 GFP_KERNEL);
4617 if (!old_opts.s_qf_names[i]) {
4618 for (j = 0; j < i; j++)
4619 kfree(old_opts.s_qf_names[j]);
3e36a163 4620 kfree(orig_data);
03dafb5f
CG		 4621 return -ENOMEM;
4622 }
4623 } else
4624 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4625#endif
b3881f74
TT		 4626 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4627 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4628
661aa520 4629 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4630 err = -EINVAL;
4631 goto restore_opts;
4632 }
4633
6b992ff2 4634 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4635 test_opt(sb, JOURNAL_CHECKSUM)) {
4636 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4637 "during remount not supported; ignoring");
4638 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4639 }
4640
6ae6514b
PS		 4641 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4642 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4643 ext4_msg(sb, KERN_ERR, "can't mount with "
4644 "both data=journal and delalloc");
4645 err = -EINVAL;
4646 goto restore_opts;
4647 }
4648 if (test_opt(sb, DIOREAD_NOLOCK)) {
4649 ext4_msg(sb, KERN_ERR, "can't mount with "
4650 "both data=journal and dioread_nolock");
4651 err = -EINVAL;
4652 goto restore_opts;
4653 }
923ae0ff
RZ		 4654 if (test_opt(sb, DAX)) {
4655 ext4_msg(sb, KERN_ERR, "can't mount with "
4656 "both data=journal and dax");
4657 err = -EINVAL;
4658 goto restore_opts;
4659 }
4660 }
4661
4662 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4663 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4664 "dax flag with busy inodes while remounting");
4665 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4666 }
4667
4ab2f15b 4668 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4669 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4670
4671 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4672 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4673
4674 es = sbi->s_es;
4675
b3881f74 4676 if (sbi->s_journal) {
0390131b 4677 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4678 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4679 }
ac27a0ec 4680
a2fd66d0
TT		 4681 if (*flags & MS_LAZYTIME)
4682 sb->s_flags |= MS_LAZYTIME;
4683
661aa520 4684 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4685 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4686 err = -EROFS;
4687 goto restore_opts;
4688 }
4689
4690 if (*flags & MS_RDONLY) {
38c03b34
TT		 4691 err = sync_filesystem(sb);
4692 if (err < 0)
4693 goto restore_opts;
0f0dd62f
CH		 4694 err = dquot_suspend(sb, -1);
4695 if (err < 0)
c79d967d 4696 goto restore_opts;
c79d967d 4697
ac27a0ec
DK		 4698 /*
4699 * First of all, the unconditional stuff we have to do
4700 * to disable replay of the journal when we next remount
4701 */
4702 sb->s_flags |= MS_RDONLY;
4703
4704 /*
4705 * OK, test if we are remounting a valid rw partition
4706 * readonly, and if so set the rdonly flag and then
4707 * mark the partition as valid again.
4708 */
617ba13b
MC		 4709 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4710 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4711 es->s_state = cpu_to_le16(sbi->s_mount_state);
4712
a63c9eb2 4713 if (sbi->s_journal)
0390131b 4714 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4715 } else {
a13fb1a4 4716 /* Make sure we can mount this feature set readwrite */
e2b911c5 4717 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 4718 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4719 err = -EROFS;
4720 goto restore_opts;
4721 }
8a266467
TT		 4722 /*
4723 * Make sure the group descriptor checksums
0b8e58a1 4724 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4725 */
4726 for (g = 0; g < sbi->s_groups_count; g++) {
4727 struct ext4_group_desc *gdp =
4728 ext4_get_group_desc(sb, g, NULL);
4729
feb0ab32 4730 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4731 ext4_msg(sb, KERN_ERR,
4732 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 4733 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 4734 le16_to_cpu(gdp->bg_checksum));
6a797d27 4735 err = -EFSBADCRC;
8a266467
TT		 4736 goto restore_opts;
4737 }
4738 }
4739
ead6596b
ES		 4740 /*
4741 * If we have an unprocessed orphan list hanging
4742 * around from a previously readonly bdev mount,
4743 * require a full umount/remount for now.
4744 */
4745 if (es->s_last_orphan) {
b31e1552 4746 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4747 "remount RDWR because of unprocessed "
4748 "orphan inode list. Please "
b31e1552 4749 "umount/remount instead");
ead6596b
ES		 4750 err = -EINVAL;
4751 goto restore_opts;
4752 }
4753
ac27a0ec
DK		 4754 /*
4755 * Mounting a RDONLY partition read-write, so reread
4756 * and store the current valid flag. (It may have
4757 * been changed by e2fsck since we originally mounted
4758 * the partition.)
4759 */
0390131b
FM		 4760 if (sbi->s_journal)
4761 ext4_clear_journal_err(sb, es);
ac27a0ec 4762 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4763 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4764 sb->s_flags &= ~MS_RDONLY;
e2b911c5 4765 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 4766 if (ext4_multi_mount_protect(sb,
4767 le64_to_cpu(es->s_mmp_block))) {
4768 err = -EROFS;
4769 goto restore_opts;
4770 }
c79d967d 4771 enable_quota = 1;
ac27a0ec
DK		 4772 }
4773 }
bfff6873
LC		 4774
4775 /*
4776 * Reinitialize lazy itable initialization thread based on
4777 * current settings
4778 */
4779 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4780 ext4_unregister_li_request(sb);
4781 else {
4782 ext4_group_t first_not_zeroed;
4783 first_not_zeroed = ext4_has_uninit_itable(sb);
4784 ext4_register_li_request(sb, first_not_zeroed);
4785 }
4786
6fd058f7 4787 ext4_setup_system_zone(sb);
d096ad0f 4788 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4789 ext4_commit_super(sb, 1);
0390131b 4790
ac27a0ec
DK		 4791#ifdef CONFIG_QUOTA
4792 /* Release old quota file names */
a2d4a646 4793 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 4794 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4795 if (enable_quota) {
4796 if (sb_any_quota_suspended(sb))
4797 dquot_resume(sb, -1);
e2b911c5 4798 else if (ext4_has_feature_quota(sb)) {
7c319d32 4799 err = ext4_enable_quotas(sb);
07724f98 4800 if (err)
7c319d32 4801 goto restore_opts;
7c319d32
AK		 4802 }
4803 }
ac27a0ec 4804#endif
d4c402d9 4805
a26f4992 4806 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 4807 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4808 kfree(orig_data);
ac27a0ec 4809 return 0;
0b8e58a1 4810
ac27a0ec
DK		 4811restore_opts:
4812 sb->s_flags = old_sb_flags;
4813 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4814 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4815 sbi->s_resuid = old_opts.s_resuid;
4816 sbi->s_resgid = old_opts.s_resgid;
4817 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4818 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4819 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4820#ifdef CONFIG_QUOTA
4821 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 4822 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 4823 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4824 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4825 }
4826#endif
d4c402d9 4827 kfree(orig_data);
ac27a0ec
DK		 4828 return err;
4829}
4830
689c958c
LX		 4831#ifdef CONFIG_QUOTA
4832static int ext4_statfs_project(struct super_block *sb,
4833 kprojid_t projid, struct kstatfs *buf)
4834{
4835 struct kqid qid;
4836 struct dquot *dquot;
4837 u64 limit;
4838 u64 curblock;
4839
4840 qid = make_kqid_projid(projid);
4841 dquot = dqget(sb, qid);
4842 if (IS_ERR(dquot))
4843 return PTR_ERR(dquot);
4844 spin_lock(&dq_data_lock);
4845
4846 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
4847 dquot->dq_dqb.dqb_bsoftlimit :
4848 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
4849 if (limit && buf->f_blocks > limit) {
4850 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
4851 buf->f_blocks = limit;
4852 buf->f_bfree = buf->f_bavail =
4853 (buf->f_blocks > curblock) ?
4854 (buf->f_blocks - curblock) : 0;
4855 }
4856
4857 limit = dquot->dq_dqb.dqb_isoftlimit ?
4858 dquot->dq_dqb.dqb_isoftlimit :
4859 dquot->dq_dqb.dqb_ihardlimit;
4860 if (limit && buf->f_files > limit) {
4861 buf->f_files = limit;
4862 buf->f_ffree =
4863 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
4864 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
4865 }
4866
4867 spin_unlock(&dq_data_lock);
4868 dqput(dquot);
4869 return 0;
4870}
4871#endif
4872
2b2d6d01 4873static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4874{
4875 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4876 struct ext4_sb_info *sbi = EXT4_SB(sb);
4877 struct ext4_super_block *es = sbi->s_es;
27dd4385 4878 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 4879 u64 fsid;
d02a9391 4880 s64 bfree;
27dd4385 4881 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 4882
952fc18e
TT		 4883 if (!test_opt(sb, MINIX_DF))
4884 overhead = sbi->s_overhead;
ac27a0ec 4885
617ba13b 4886 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4887 buf->f_bsize = sb->s_blocksize;
b72f78cb 4888 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4889 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4890 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4891 /* prevent underflow in case that few free space is available */
57042651 4892 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 4893 buf->f_bavail = buf->f_bfree -
4894 (ext4_r_blocks_count(es) + resv_blocks);
4895 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 4896 buf->f_bavail = 0;
4897 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4898 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4899 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4900 fsid = le64_to_cpup((void *)es->s_uuid) ^
4901 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4902 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4903 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4904
689c958c
LX		 4905#ifdef CONFIG_QUOTA
4906 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
4907 sb_has_quota_limits_enabled(sb, PRJQUOTA))
4908 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
4909#endif
ac27a0ec
DK		 4910 return 0;
4911}
4912
0b8e58a1
AD		 4913/* Helper function for writing quotas on sync - we need to start transaction
4914 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4915 * Process 1 Process 2
617ba13b 4916 * ext4_create() quota_sync()
a269eb18 4917 * jbd2_journal_start() write_dquot()
871a2931 4918 * dquot_initialize() down(dqio_mutex)
dab291af 4919 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4920 *
4921 */
4922
4923#ifdef CONFIG_QUOTA
4924
4925static inline struct inode *dquot_to_inode(struct dquot *dquot)
4926{
4c376dca 4927 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4928}
4929
617ba13b 4930static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4931{
4932 int ret, err;
4933 handle_t *handle;
4934 struct inode *inode;
4935
4936 inode = dquot_to_inode(dquot);
9924a92a 4937 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 4938 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4939 if (IS_ERR(handle))
4940 return PTR_ERR(handle);
4941 ret = dquot_commit(dquot);
617ba13b 4942 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4943 if (!ret)
4944 ret = err;
4945 return ret;
4946}
4947
617ba13b 4948static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4949{
4950 int ret, err;
4951 handle_t *handle;
4952
9924a92a 4953 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4954 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4955 if (IS_ERR(handle))
4956 return PTR_ERR(handle);
4957 ret = dquot_acquire(dquot);
617ba13b 4958 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4959 if (!ret)
4960 ret = err;
4961 return ret;
4962}
4963
617ba13b 4964static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4965{
4966 int ret, err;
4967 handle_t *handle;
4968
9924a92a 4969 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4970 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4971 if (IS_ERR(handle)) {
4972 /* Release dquot anyway to avoid endless cycle in dqput() */
4973 dquot_release(dquot);
ac27a0ec 4974 return PTR_ERR(handle);
9c3013e9 4975 }
ac27a0ec 4976 ret = dquot_release(dquot);
617ba13b 4977 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4978 if (!ret)
4979 ret = err;
4980 return ret;
4981}
4982
617ba13b 4983static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4984{
262b4662
JK		 4985 struct super_block *sb = dquot->dq_sb;
4986 struct ext4_sb_info *sbi = EXT4_SB(sb);
4987
2c8be6b2 4988 /* Are we journaling quotas? */
e2b911c5 4989 if (ext4_has_feature_quota(sb) ||
262b4662 4990 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 4991 dquot_mark_dquot_dirty(dquot);
617ba13b 4992 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4993 } else {
4994 return dquot_mark_dquot_dirty(dquot);
4995 }
4996}
4997
617ba13b 4998static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4999{
5000 int ret, err;
5001 handle_t *handle;
5002
5003 /* Data block + inode block */
2b0143b5 5004 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5005 if (IS_ERR(handle))
5006 return PTR_ERR(handle);
5007 ret = dquot_commit_info(sb, type);
617ba13b 5008 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5009 if (!ret)
5010 ret = err;
5011 return ret;
5012}
5013
5014/*
5015 * Turn on quotas during mount time - we need to find
5016 * the quota file and such...
5017 */
617ba13b 5018static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5019{
287a8095
CH		 5020 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5021 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5022}
5023
5024/*
5025 * Standard function to be called on quota_on
5026 */
617ba13b 5027static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5028 struct path *path)
ac27a0ec
DK		 5029{
5030 int err;
ac27a0ec
DK		 5031
5032 if (!test_opt(sb, QUOTA))
5033 return -EINVAL;
0623543b 5034
ac27a0ec 5035 /* Quotafile not on the same filesystem? */
d8c9584e 5036 if (path->dentry->d_sb != sb)
ac27a0ec 5037 return -EXDEV;
0623543b
JK		 5038 /* Journaling quota? */
5039 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5040 /* Quotafile not in fs root? */
f00c9e44 5041 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5042 ext4_msg(sb, KERN_WARNING,
5043 "Quota file not on filesystem root. "
5044 "Journaled quota will not work");
2b2d6d01 5045 }
0623543b
JK		 5046
5047 /*
5048 * When we journal data on quota file, we have to flush journal to see
5049 * all updates to the file when we bypass pagecache...
5050 */
0390131b 5051 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5052 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5053 /*
5054 * We don't need to lock updates but journal_flush() could
5055 * otherwise be livelocked...
5056 */
5057 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5058 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5059 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5060 if (err)
7ffe1ea8 5061 return err;
0623543b
JK		 5062 }
5063
f00c9e44 5064 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5065}
5066
7c319d32
AK		 5067static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5068 unsigned int flags)
5069{
5070 int err;
5071 struct inode *qf_inode;
a2d4a646 5072 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5073 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5074 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5075 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5076 };
5077
e2b911c5 5078 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5079
5080 if (!qf_inums[type])
5081 return -EPERM;
5082
5083 qf_inode = ext4_iget(sb, qf_inums[type]);
5084 if (IS_ERR(qf_inode)) {
5085 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5086 return PTR_ERR(qf_inode);
5087 }
5088
bcb13850
JK		 5089 /* Don't account quota for quota files to avoid recursion */
5090 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5091 err = dquot_enable(qf_inode, type, format_id, flags);
5092 iput(qf_inode);
5093
5094 return err;
5095}
5096
5097/* Enable usage tracking for all quota types. */
5098static int ext4_enable_quotas(struct super_block *sb)
5099{
5100 int type, err = 0;
a2d4a646 5101 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5102 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5103 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5104 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5105 };
5106
5107 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5108 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5109 if (qf_inums[type]) {
5110 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5111 DQUOT_USAGE_ENABLED);
5112 if (err) {
5113 ext4_warning(sb,
72ba7450
TT		 5114 "Failed to enable quota tracking "
5115 "(type=%d, err=%d). Please run "
5116 "e2fsck to fix.", type, err);
7c319d32
AK		 5117 return err;
5118 }
5119 }
5120 }
5121 return 0;
5122}
5123
ca0e05e4
DM		 5124static int ext4_quota_off(struct super_block *sb, int type)
5125{
21f97697
JK		 5126 struct inode *inode = sb_dqopt(sb)->files[type];
5127 handle_t *handle;
5128
87009d86
DM		 5129 /* Force all delayed allocation blocks to be allocated.
5130 * Caller already holds s_umount sem */
5131 if (test_opt(sb, DELALLOC))
ca0e05e4 5132 sync_filesystem(sb);
ca0e05e4 5133
0b268590
AG		 5134 if (!inode)
5135 goto out;
5136
21f97697
JK		 5137 /* Update modification times of quota files when userspace can
5138 * start looking at them */
9924a92a 5139 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5140 if (IS_ERR(handle))
5141 goto out;
5142 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5143 ext4_mark_inode_dirty(handle, inode);
5144 ext4_journal_stop(handle);
5145
5146out:
ca0e05e4
DM		 5147 return dquot_quota_off(sb, type);
5148}
5149
ac27a0ec
DK		 5150/* Read data from quotafile - avoid pagecache and such because we cannot afford
5151 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5152 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5153 * we don't have to be afraid of races */
617ba13b 5154static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5155 size_t len, loff_t off)
5156{
5157 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5158 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5159 int offset = off & (sb->s_blocksize - 1);
5160 int tocopy;
5161 size_t toread;
5162 struct buffer_head *bh;
5163 loff_t i_size = i_size_read(inode);
5164
5165 if (off > i_size)
5166 return 0;
5167 if (off+len > i_size)
5168 len = i_size-off;
5169 toread = len;
5170 while (toread > 0) {
5171 tocopy = sb->s_blocksize - offset < toread ?
5172 sb->s_blocksize - offset : toread;
1c215028
TT		 5173 bh = ext4_bread(NULL, inode, blk, 0);
5174 if (IS_ERR(bh))
5175 return PTR_ERR(bh);
ac27a0ec
DK		 5176 if (!bh) /* A hole? */
5177 memset(data, 0, tocopy);
5178 else
5179 memcpy(data, bh->b_data+offset, tocopy);
5180 brelse(bh);
5181 offset = 0;
5182 toread -= tocopy;
5183 data += tocopy;
5184 blk++;
5185 }
5186 return len;
5187}
5188
5189/* Write to quotafile (we know the transaction is already started and has
5190 * enough credits) */
617ba13b 5191static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5192 const char *data, size_t len, loff_t off)
5193{
5194 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5195 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5196 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5197 int retries = 0;
ac27a0ec
DK		 5198 struct buffer_head *bh;
5199 handle_t *handle = journal_current_handle();
5200
0390131b 5201 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5202 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5203 " cancelled because transaction is not started",
9c3013e9
JK		 5204 (unsigned long long)off, (unsigned long long)len);
5205 return -EIO;
5206 }
67eeb568
DM		 5207 /*
5208 * Since we account only one data block in transaction credits,
5209 * then it is impossible to cross a block boundary.
5210 */
5211 if (sb->s_blocksize - offset < len) {
5212 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5213 " cancelled because not block aligned",
5214 (unsigned long long)off, (unsigned long long)len);
5215 return -EIO;
5216 }
5217
c5e298ae
TT		 5218 do {
5219 bh = ext4_bread(handle, inode, blk,
5220 EXT4_GET_BLOCKS_CREATE |
5221 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5222 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5223 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5224 if (IS_ERR(bh))
5225 return PTR_ERR(bh);
67eeb568
DM		 5226 if (!bh)
5227 goto out;
5d601255 5228 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5229 err = ext4_journal_get_write_access(handle, bh);
5230 if (err) {
5231 brelse(bh);
1c215028 5232 return err;
ac27a0ec 5233 }
67eeb568
DM		 5234 lock_buffer(bh);
5235 memcpy(bh->b_data+offset, data, len);
5236 flush_dcache_page(bh->b_page);
5237 unlock_buffer(bh);
62d2b5f2 5238 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5239 brelse(bh);
ac27a0ec 5240out:
67eeb568
DM		 5241 if (inode->i_size < off + len) {
5242 i_size_write(inode, off + len);
617ba13b 5243 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5244 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5245 }
67eeb568 5246 return len;
ac27a0ec
DK		 5247}
5248
5249#endif
5250
152a0836
AV		 5251static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5252 const char *dev_name, void *data)
ac27a0ec 5253{
152a0836 5254 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5255}
5256
c290ea01 5257#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5258static inline void register_as_ext2(void)
5259{
5260 int err = register_filesystem(&ext2_fs_type);
5261 if (err)
5262 printk(KERN_WARNING
5263 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5264}
5265
5266static inline void unregister_as_ext2(void)
5267{
5268 unregister_filesystem(&ext2_fs_type);
5269}
2035e776
TT		 5270
5271static inline int ext2_feature_set_ok(struct super_block *sb)
5272{
e2b911c5 5273 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776
TT		 5274 return 0;
5275 if (sb->s_flags & MS_RDONLY)
5276 return 1;
e2b911c5 5277 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5278 return 0;
5279 return 1;
5280}
24b58424
TT		 5281#else
5282static inline void register_as_ext2(void) { }
5283static inline void unregister_as_ext2(void) { }
2035e776 5284static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5285#endif
5286
24b58424
TT		 5287static inline void register_as_ext3(void)
5288{
5289 int err = register_filesystem(&ext3_fs_type);
5290 if (err)
5291 printk(KERN_WARNING
5292 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5293}
5294
5295static inline void unregister_as_ext3(void)
5296{
5297 unregister_filesystem(&ext3_fs_type);
5298}
2035e776
TT		 5299
5300static inline int ext3_feature_set_ok(struct super_block *sb)
5301{
e2b911c5 5302 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5303 return 0;
e2b911c5 5304 if (!ext4_has_feature_journal(sb))
2035e776
TT		 5305 return 0;
5306 if (sb->s_flags & MS_RDONLY)
5307 return 1;
e2b911c5 5308 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5309 return 0;
5310 return 1;
5311}
24b58424 5312
03010a33
TT		 5313static struct file_system_type ext4_fs_type = {
5314 .owner = THIS_MODULE,
5315 .name = "ext4",
152a0836 5316 .mount = ext4_mount,
03010a33
TT		 5317 .kill_sb = kill_block_super,
5318 .fs_flags = FS_REQUIRES_DEV,
5319};
7f78e035 5320MODULE_ALIAS_FS("ext4");
03010a33 5321
e9e3bcec
ES		 5322/* Shared across all ext4 file systems */
5323wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5324struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5325
5dabfc78 5326static int __init ext4_init_fs(void)
ac27a0ec 5327{
e9e3bcec 5328 int i, err;
c9de560d 5329
e294a537 5330 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5331 ext4_li_info = NULL;
5332 mutex_init(&ext4_li_mtx);
5333
9a4c8019 5334 /* Build-time check for flags consistency */
12e9b892 5335 ext4_check_flag_values();
e9e3bcec
ES		 5336
5337 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5338 mutex_init(&ext4__aio_mutex[i]);
5339 init_waitqueue_head(&ext4__ioend_wq[i]);
5340 }
5341
51865fda 5342 err = ext4_init_es();
6fd058f7
TT		 5343 if (err)
5344 return err;
51865fda
ZL		 5345
5346 err = ext4_init_pageio();
5347 if (err)
b5799018 5348 goto out5;
51865fda 5349
5dabfc78 5350 err = ext4_init_system_zone();
bd2d0210 5351 if (err)
b5799018 5352 goto out4;
857ac889 5353
b5799018 5354 err = ext4_init_sysfs();
dd68314c 5355 if (err)
b5799018 5356 goto out3;
857ac889 5357
5dabfc78 5358 err = ext4_init_mballoc();
c9de560d
AT		 5359 if (err)
5360 goto out2;
ac27a0ec
DK		 5361 err = init_inodecache();
5362 if (err)
5363 goto out1;
24b58424 5364 register_as_ext3();
2035e776 5365 register_as_ext2();
03010a33 5366 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5367 if (err)
5368 goto out;
bfff6873 5369
ac27a0ec
DK		 5370 return 0;
5371out:
24b58424
TT		 5372 unregister_as_ext2();
5373 unregister_as_ext3();
ac27a0ec
DK		 5374 destroy_inodecache();
5375out1:
5dabfc78 5376 ext4_exit_mballoc();
9c191f70 5377out2:
b5799018
TT		 5378 ext4_exit_sysfs();
5379out3:
5dabfc78 5380 ext4_exit_system_zone();
b5799018 5381out4:
5dabfc78 5382 ext4_exit_pageio();
b5799018 5383out5:
51865fda
ZL		 5384 ext4_exit_es();
5385
ac27a0ec
DK		 5386 return err;
5387}
5388
5dabfc78 5389static void __exit ext4_exit_fs(void)
ac27a0ec 5390{
e298e73b 5391 ext4_exit_crypto();
bfff6873 5392 ext4_destroy_lazyinit_thread();
24b58424
TT		 5393 unregister_as_ext2();
5394 unregister_as_ext3();
03010a33 5395 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5396 destroy_inodecache();
5dabfc78 5397 ext4_exit_mballoc();
b5799018 5398 ext4_exit_sysfs();
5dabfc78
TT		 5399 ext4_exit_system_zone();
5400 ext4_exit_pageio();
dd12ed14 5401 ext4_exit_es();
ac27a0ec
DK		 5402}
5403
5404MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5405MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5406MODULE_LICENSE("GPL");
5dabfc78
TT		 5407module_init(ext4_init_fs)
5408module_exit(ext4_exit_fs)
Linux kernel - Deliverables
This page took 1.356931 seconds and 5 git commands to generate.