]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: fix deadlock during page writeback
[thirdparty/kernel/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);
c8585c6f 862 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec
DK		 863 brelse(sbi->s_sbh);
864#ifdef CONFIG_QUOTA
a2d4a646 865 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 866 kfree(sbi->s_qf_names[i]);
867#endif
868
869 /* Debugging code just in case the in-memory inode orphan list
870 * isn't empty. The on-disk one can be non-empty if we've
871 * detected an error and taken the fs readonly, but the
872 * in-memory list had better be clean by this point. */
873 if (!list_empty(&sbi->s_orphan))
874 dump_orphan_list(sb, sbi);
875 J_ASSERT(list_empty(&sbi->s_orphan));
876
89d96a6f 877 sync_blockdev(sb->s_bdev);
f98393a6 878 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 879 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
880 /*
881 * Invalidate the journal device's buffers. We don't want them
882 * floating about in memory - the physical journal device may
883 * hotswapped, and it breaks the `ro-after' testing code.
884 */
885 sync_blockdev(sbi->journal_bdev);
f98393a6 886 invalidate_bdev(sbi->journal_bdev);
617ba13b 887 ext4_blkdev_remove(sbi);
ac27a0ec 888 }
9c191f70
M		 889 if (sbi->s_mb_cache) {
890 ext4_xattr_destroy_cache(sbi->s_mb_cache);
891 sbi->s_mb_cache = NULL;
892 }
c5e06d10
JL		 893 if (sbi->s_mmp_tsk)
894 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 895 sb->s_fs_info = NULL;
3197ebdb
TT		 896 /*
897 * Now that we are completely done shutting down the
898 * superblock, we need to actually destroy the kobject.
899 */
3197ebdb
TT		 900 kobject_put(&sbi->s_kobj);
901 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 902 if (sbi->s_chksum_driver)
903 crypto_free_shash(sbi->s_chksum_driver);
705895b6 904 kfree(sbi->s_blockgroup_lock);
ac27a0ec 905 kfree(sbi);
ac27a0ec
DK		 906}
907
e18b890b 908static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 909
910/*
911 * Called inside transaction, so use GFP_NOFS
912 */
617ba13b 913static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 914{
617ba13b 915 struct ext4_inode_info *ei;
ac27a0ec 916
e6b4f8da 917 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 918 if (!ei)
919 return NULL;
0b8e58a1 920
ac27a0ec 921 ei->vfs_inode.i_version = 1;
202ee5df 922 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 923 INIT_LIST_HEAD(&ei->i_prealloc_list);
924 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 925 ext4_es_init_tree(&ei->i_es_tree);
926 rwlock_init(&ei->i_es_lock);
edaa53ca 927 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 928 ei->i_es_all_nr = 0;
edaa53ca 929 ei->i_es_shk_nr = 0;
dd475925 930 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 931 ei->i_reserved_data_blocks = 0;
932 ei->i_reserved_meta_blocks = 0;
933 ei->i_allocated_meta_blocks = 0;
9d0be502 934 ei->i_da_metadata_calc_len = 0;
7e731bc9 935 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 936 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 937#ifdef CONFIG_QUOTA
938 ei->i_reserved_quota = 0;
96c7e0d9 939 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 940#endif
8aefcd55 941 ei->jinode = NULL;
2e8fa54e 942 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 943 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 944 ei->i_sync_tid = 0;
945 ei->i_datasync_tid = 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);
8f0e8746 1117static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1118
96c7e0d9
JK		 1119static struct dquot **ext4_get_dquots(struct inode *inode)
1120{
1121 return EXT4_I(inode)->i_dquot;
1122}
1123
61e225dc 1124static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1125 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1126 .write_dquot = ext4_write_dquot,
1127 .acquire_dquot = ext4_acquire_dquot,
1128 .release_dquot = ext4_release_dquot,
1129 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1130 .write_info = ext4_write_info,
1131 .alloc_dquot = dquot_alloc,
1132 .destroy_dquot = dquot_destroy,
040cb378 1133 .get_projid = ext4_get_projid,
8f0e8746 1134 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1135};
1136
0d54b217 1137static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1138 .quota_on = ext4_quota_on,
ca0e05e4 1139 .quota_off = ext4_quota_off,
287a8095 1140 .quota_sync = dquot_quota_sync,
0a240339 1141 .get_state = dquot_get_state,
287a8095
CH		 1142 .set_info = dquot_set_dqinfo,
1143 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1144 .set_dqblk = dquot_set_dqblk,
1145 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1146};
1147#endif
1148
ee9b6d61 1149static const struct super_operations ext4_sops = {
617ba13b
MC		 1150 .alloc_inode = ext4_alloc_inode,
1151 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1152 .write_inode = ext4_write_inode,
1153 .dirty_inode = ext4_dirty_inode,
7ff9c073 1154 .drop_inode = ext4_drop_inode,
0930fcc1 1155 .evict_inode = ext4_evict_inode,
617ba13b 1156 .put_super = ext4_put_super,
617ba13b 1157 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1158 .freeze_fs = ext4_freeze,
1159 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1160 .statfs = ext4_statfs,
1161 .remount_fs = ext4_remount,
617ba13b 1162 .show_options = ext4_show_options,
ac27a0ec 1163#ifdef CONFIG_QUOTA
617ba13b
MC		 1164 .quota_read = ext4_quota_read,
1165 .quota_write = ext4_quota_write,
96c7e0d9 1166 .get_dquots = ext4_get_dquots,
ac27a0ec 1167#endif
c39a7f84 1168 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1169};
1170
39655164 1171static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1172 .fh_to_dentry = ext4_fh_to_dentry,
1173 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1174 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1175};
1176
1177enum {
1178 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1179 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1180 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1181 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1182 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1183 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1184 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1185 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1186 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1187 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1188 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1189 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
923ae0ff 1190 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1449032b 1191 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1192 Opt_lazytime, Opt_nolazytime,
1449032b 1193 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1194 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1195 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1196 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1197 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1198};
1199
a447c093 1200static const match_table_t tokens = {
ac27a0ec
DK		 1201 {Opt_bsd_df, "bsddf"},
1202 {Opt_minix_df, "minixdf"},
1203 {Opt_grpid, "grpid"},
1204 {Opt_grpid, "bsdgroups"},
1205 {Opt_nogrpid, "nogrpid"},
1206 {Opt_nogrpid, "sysvgroups"},
1207 {Opt_resgid, "resgid=%u"},
1208 {Opt_resuid, "resuid=%u"},
1209 {Opt_sb, "sb=%u"},
1210 {Opt_err_cont, "errors=continue"},
1211 {Opt_err_panic, "errors=panic"},
1212 {Opt_err_ro, "errors=remount-ro"},
1213 {Opt_nouid32, "nouid32"},
ac27a0ec 1214 {Opt_debug, "debug"},
72578c33
TT		 1215 {Opt_removed, "oldalloc"},
1216 {Opt_removed, "orlov"},
ac27a0ec
DK		 1217 {Opt_user_xattr, "user_xattr"},
1218 {Opt_nouser_xattr, "nouser_xattr"},
1219 {Opt_acl, "acl"},
1220 {Opt_noacl, "noacl"},
e3bb52ae 1221 {Opt_noload, "norecovery"},
5a916be1 1222 {Opt_noload, "noload"},
72578c33
TT		 1223 {Opt_removed, "nobh"},
1224 {Opt_removed, "bh"},
ac27a0ec 1225 {Opt_commit, "commit=%u"},
30773840
TT		 1226 {Opt_min_batch_time, "min_batch_time=%u"},
1227 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1228 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1229 {Opt_journal_path, "journal_path=%s"},
818d276c 1230 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1231 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1232 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1233 {Opt_abort, "abort"},
1234 {Opt_data_journal, "data=journal"},
1235 {Opt_data_ordered, "data=ordered"},
1236 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1237 {Opt_data_err_abort, "data_err=abort"},
1238 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1239 {Opt_offusrjquota, "usrjquota="},
1240 {Opt_usrjquota, "usrjquota=%s"},
1241 {Opt_offgrpjquota, "grpjquota="},
1242 {Opt_grpjquota, "grpjquota=%s"},
1243 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1244 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1245 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1246 {Opt_grpquota, "grpquota"},
1247 {Opt_noquota, "noquota"},
1248 {Opt_quota, "quota"},
1249 {Opt_usrquota, "usrquota"},
1250 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1251 {Opt_barrier, "barrier"},
1252 {Opt_nobarrier, "nobarrier"},
25ec56b5 1253 {Opt_i_version, "i_version"},
923ae0ff 1254 {Opt_dax, "dax"},
c9de560d 1255 {Opt_stripe, "stripe=%u"},
64769240 1256 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1257 {Opt_lazytime, "lazytime"},
1258 {Opt_nolazytime, "nolazytime"},
dd919b98 1259 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1260 {Opt_removed, "mblk_io_submit"},
1261 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1262 {Opt_block_validity, "block_validity"},
1263 {Opt_noblock_validity, "noblock_validity"},
240799cd 1264 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1265 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1266 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1267 {Opt_auto_da_alloc, "auto_da_alloc"},
1268 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1269 {Opt_dioread_nolock, "dioread_nolock"},
1270 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1271 {Opt_discard, "discard"},
1272 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1273 {Opt_init_itable, "init_itable=%u"},
1274 {Opt_init_itable, "init_itable"},
1275 {Opt_noinit_itable, "noinit_itable"},
df981d03 1276 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1277 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1278 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1279 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1280 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1281 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1282 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1283 {Opt_err, NULL},
ac27a0ec
DK		 1284};
1285
617ba13b 1286static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1287{
617ba13b 1288 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1289 char *options = (char *) *data;
1290
1291 if (!options || strncmp(options, "sb=", 3) != 0)
1292 return 1; /* Default location */
0b8e58a1 1293
ac27a0ec 1294 options += 3;
0b8e58a1 1295 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1296 sb_block = simple_strtoul(options, &options, 0);
1297 if (*options && *options != ',') {
4776004f 1298 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1299 (char *) *data);
1300 return 1;
1301 }
1302 if (*options == ',')
1303 options++;
1304 *data = (void *) options;
0b8e58a1 1305
ac27a0ec
DK		 1306 return sb_block;
1307}
1308
b3881f74 1309#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1310static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1311 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1312
56c50f11
DM		 1313#ifdef CONFIG_QUOTA
1314static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1315{
1316 struct ext4_sb_info *sbi = EXT4_SB(sb);
1317 char *qname;
03dafb5f 1318 int ret = -1;
56c50f11
DM		 1319
1320 if (sb_any_quota_loaded(sb) &&
1321 !sbi->s_qf_names[qtype]) {
1322 ext4_msg(sb, KERN_ERR,
1323 "Cannot change journaled "
1324 "quota options when quota turned on");
57f73c2c 1325 return -1;
56c50f11 1326 }
e2b911c5 1327 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1328 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1329 "ignored when QUOTA feature is enabled");
1330 return 1;
262b4662 1331 }
56c50f11
DM		 1332 qname = match_strdup(args);
1333 if (!qname) {
1334 ext4_msg(sb, KERN_ERR,
1335 "Not enough memory for storing quotafile name");
57f73c2c 1336 return -1;
56c50f11 1337 }
03dafb5f
CG		 1338 if (sbi->s_qf_names[qtype]) {
1339 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1340 ret = 1;
1341 else
1342 ext4_msg(sb, KERN_ERR,
1343 "%s quota file already specified",
1344 QTYPE2NAME(qtype));
1345 goto errout;
56c50f11 1346 }
03dafb5f 1347 if (strchr(qname, '/')) {
56c50f11
DM		 1348 ext4_msg(sb, KERN_ERR,
1349 "quotafile must be on filesystem root");
03dafb5f 1350 goto errout;
56c50f11 1351 }
03dafb5f 1352 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1353 set_opt(sb, QUOTA);
56c50f11 1354 return 1;
03dafb5f
CG		 1355errout:
1356 kfree(qname);
1357 return ret;
56c50f11
DM		 1358}
1359
1360static int clear_qf_name(struct super_block *sb, int qtype)
1361{
1362
1363 struct ext4_sb_info *sbi = EXT4_SB(sb);
1364
1365 if (sb_any_quota_loaded(sb) &&
1366 sbi->s_qf_names[qtype]) {
1367 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1368 " when quota turned on");
57f73c2c 1369 return -1;
56c50f11 1370 }
03dafb5f 1371 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1372 sbi->s_qf_names[qtype] = NULL;
1373 return 1;
1374}
1375#endif
1376
26092bf5
TT		 1377#define MOPT_SET 0x0001
1378#define MOPT_CLEAR 0x0002
1379#define MOPT_NOSUPPORT 0x0004
1380#define MOPT_EXPLICIT 0x0008
1381#define MOPT_CLEAR_ERR 0x0010
1382#define MOPT_GTE0 0x0020
ac27a0ec 1383#ifdef CONFIG_QUOTA
26092bf5
TT		 1384#define MOPT_Q 0
1385#define MOPT_QFMT 0x0040
1386#else
1387#define MOPT_Q MOPT_NOSUPPORT
1388#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1389#endif
26092bf5 1390#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1391#define MOPT_NO_EXT2 0x0100
1392#define MOPT_NO_EXT3 0x0200
1393#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1394#define MOPT_STRING 0x0400
26092bf5
TT		 1395
1396static const struct mount_opts {
1397 int token;
1398 int mount_opt;
1399 int flags;
1400} ext4_mount_opts[] = {
1401 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1402 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1403 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1404 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1405 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1406 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1407 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1408 MOPT_EXT4_ONLY | MOPT_SET},
1409 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1410 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1411 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1412 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1413 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1414 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1415 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1416 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1417 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1418 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1419 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1420 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1421 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1422 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1423 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1424 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1425 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1426 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1427 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1428 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1429 MOPT_NO_EXT2},
8dc0aa8c 1430 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1431 MOPT_NO_EXT2},
26092bf5
TT		 1432 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1433 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1434 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1435 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1436 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1437 {Opt_commit, 0, MOPT_GTE0},
1438 {Opt_max_batch_time, 0, MOPT_GTE0},
1439 {Opt_min_batch_time, 0, MOPT_GTE0},
1440 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1441 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1442 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1443 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1444 {Opt_resuid, 0, MOPT_GTE0},
1445 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1446 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1447 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1448 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1449 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1450 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1451 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1452 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1453 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1454 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1455#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1456 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1457 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1458#else
26092bf5
TT		 1459 {Opt_acl, 0, MOPT_NOSUPPORT},
1460 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1461#endif
26092bf5
TT		 1462 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1463 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1464 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1465 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1466 MOPT_SET | MOPT_Q},
1467 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1468 MOPT_SET | MOPT_Q},
1469 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1470 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1471 {Opt_usrjquota, 0, MOPT_Q},
1472 {Opt_grpjquota, 0, MOPT_Q},
1473 {Opt_offusrjquota, 0, MOPT_Q},
1474 {Opt_offgrpjquota, 0, MOPT_Q},
1475 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1476 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1477 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1478 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1479 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1480 {Opt_err, 0, 0}
1481};
1482
1483static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1484 substring_t *args, unsigned long *journal_devnum,
1485 unsigned int *journal_ioprio, int is_remount)
1486{
1487 struct ext4_sb_info *sbi = EXT4_SB(sb);
1488 const struct mount_opts *m;
08cefc7a
EB		 1489 kuid_t uid;
1490 kgid_t gid;
26092bf5
TT		 1491 int arg = 0;
1492
57f73c2c
TT		 1493#ifdef CONFIG_QUOTA
1494 if (token == Opt_usrjquota)
1495 return set_qf_name(sb, USRQUOTA, &args[0]);
1496 else if (token == Opt_grpjquota)
1497 return set_qf_name(sb, GRPQUOTA, &args[0]);
1498 else if (token == Opt_offusrjquota)
1499 return clear_qf_name(sb, USRQUOTA);
1500 else if (token == Opt_offgrpjquota)
1501 return clear_qf_name(sb, GRPQUOTA);
1502#endif
26092bf5 1503 switch (token) {
f7048605
TT		 1504 case Opt_noacl:
1505 case Opt_nouser_xattr:
1506 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1507 break;
26092bf5
TT		 1508 case Opt_sb:
1509 return 1; /* handled by get_sb_block() */
1510 case Opt_removed:
5f3633e3 1511 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1512 return 1;
26092bf5
TT		 1513 case Opt_abort:
1514 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1515 return 1;
1516 case Opt_i_version:
1517 sb->s_flags |= MS_I_VERSION;
1518 return 1;
a26f4992
TT		 1519 case Opt_lazytime:
1520 sb->s_flags |= MS_LAZYTIME;
1521 return 1;
1522 case Opt_nolazytime:
1523 sb->s_flags &= ~MS_LAZYTIME;
1524 return 1;
26092bf5
TT		 1525 }
1526
5f3633e3
JK		 1527 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1528 if (token == m->token)
1529 break;
1530
1531 if (m->token == Opt_err) {
1532 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1533 "or missing value", opt);
1534 return -1;
1535 }
1536
8dc0aa8c
TT		 1537 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1538 ext4_msg(sb, KERN_ERR,
1539 "Mount option \"%s\" incompatible with ext2", opt);
1540 return -1;
1541 }
1542 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1543 ext4_msg(sb, KERN_ERR,
1544 "Mount option \"%s\" incompatible with ext3", opt);
1545 return -1;
1546 }
1547
ad4eec61 1548 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1549 return -1;
1550 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1551 return -1;
c93cf2d7
DM		 1552 if (m->flags & MOPT_EXPLICIT) {
1553 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1554 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1555 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1556 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1557 } else
1558 return -1;
1559 }
5f3633e3
JK		 1560 if (m->flags & MOPT_CLEAR_ERR)
1561 clear_opt(sb, ERRORS_MASK);
1562 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1563 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1564 "options when quota turned on");
1565 return -1;
1566 }
1567
1568 if (m->flags & MOPT_NOSUPPORT) {
1569 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1570 } else if (token == Opt_commit) {
1571 if (arg == 0)
1572 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1573 sbi->s_commit_interval = HZ * arg;
1574 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1575 sbi->s_max_batch_time = arg;
1576 } else if (token == Opt_min_batch_time) {
1577 sbi->s_min_batch_time = arg;
1578 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1579 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1580 ext4_msg(sb, KERN_ERR,
1581 "EXT4-fs: inode_readahead_blks must be "
1582 "0 or a power of 2 smaller than 2^31");
26092bf5 1583 return -1;
5f3633e3
JK		 1584 }
1585 sbi->s_inode_readahead_blks = arg;
1586 } else if (token == Opt_init_itable) {
1587 set_opt(sb, INIT_INODE_TABLE);
1588 if (!args->from)
1589 arg = EXT4_DEF_LI_WAIT_MULT;
1590 sbi->s_li_wait_mult = arg;
1591 } else if (token == Opt_max_dir_size_kb) {
1592 sbi->s_max_dir_size_kb = arg;
1593 } else if (token == Opt_stripe) {
1594 sbi->s_stripe = arg;
1595 } else if (token == Opt_resuid) {
1596 uid = make_kuid(current_user_ns(), arg);
1597 if (!uid_valid(uid)) {
1598 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1599 return -1;
1600 }
5f3633e3
JK		 1601 sbi->s_resuid = uid;
1602 } else if (token == Opt_resgid) {
1603 gid = make_kgid(current_user_ns(), arg);
1604 if (!gid_valid(gid)) {
1605 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1606 return -1;
1607 }
1608 sbi->s_resgid = gid;
1609 } else if (token == Opt_journal_dev) {
1610 if (is_remount) {
1611 ext4_msg(sb, KERN_ERR,
1612 "Cannot specify journal on remount");
1613 return -1;
1614 }
1615 *journal_devnum = arg;
ad4eec61
ES		 1616 } else if (token == Opt_journal_path) {
1617 char *journal_path;
1618 struct inode *journal_inode;
1619 struct path path;
1620 int error;
1621
1622 if (is_remount) {
1623 ext4_msg(sb, KERN_ERR,
1624 "Cannot specify journal on remount");
1625 return -1;
1626 }
1627 journal_path = match_strdup(&args[0]);
1628 if (!journal_path) {
1629 ext4_msg(sb, KERN_ERR, "error: could not dup "
1630 "journal device string");
1631 return -1;
1632 }
1633
1634 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1635 if (error) {
1636 ext4_msg(sb, KERN_ERR, "error: could not find "
1637 "journal device path: error %d", error);
1638 kfree(journal_path);
1639 return -1;
1640 }
1641
2b0143b5 1642 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1643 if (!S_ISBLK(journal_inode->i_mode)) {
1644 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1645 "is not a block device", journal_path);
1646 path_put(&path);
1647 kfree(journal_path);
1648 return -1;
1649 }
1650
1651 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1652 path_put(&path);
1653 kfree(journal_path);
5f3633e3
JK		 1654 } else if (token == Opt_journal_ioprio) {
1655 if (arg > 7) {
1656 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1657 " (must be 0-7)");
1658 return -1;
1659 }
1660 *journal_ioprio =
1661 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1662 } else if (token == Opt_test_dummy_encryption) {
1663#ifdef CONFIG_EXT4_FS_ENCRYPTION
1664 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1665 ext4_msg(sb, KERN_WARNING,
1666 "Test dummy encryption mode enabled");
1667#else
1668 ext4_msg(sb, KERN_WARNING,
1669 "Test dummy encryption mount option ignored");
1670#endif
5f3633e3
JK		 1671 } else if (m->flags & MOPT_DATAJ) {
1672 if (is_remount) {
1673 if (!sbi->s_journal)
1674 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1675 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1676 ext4_msg(sb, KERN_ERR,
26092bf5 1677 "Cannot change data mode on remount");
26092bf5 1678 return -1;
ac27a0ec 1679 }
26092bf5 1680 } else {
5f3633e3
JK		 1681 clear_opt(sb, DATA_FLAGS);
1682 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1683 }
5f3633e3
JK		 1684#ifdef CONFIG_QUOTA
1685 } else if (m->flags & MOPT_QFMT) {
1686 if (sb_any_quota_loaded(sb) &&
1687 sbi->s_jquota_fmt != m->mount_opt) {
1688 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1689 "quota options when quota turned on");
1690 return -1;
1691 }
e2b911c5 1692 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1693 ext4_msg(sb, KERN_INFO,
1694 "Quota format mount options ignored "
262b4662 1695 "when QUOTA feature is enabled");
c325a67c 1696 return 1;
262b4662 1697 }
5f3633e3 1698 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1699#endif
923ae0ff 1700 } else if (token == Opt_dax) {
ef83b6e8
DW		 1701#ifdef CONFIG_FS_DAX
1702 ext4_msg(sb, KERN_WARNING,
1703 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1704 sbi->s_mount_opt |= m->mount_opt;
1705#else
923ae0ff
RZ		 1706 ext4_msg(sb, KERN_INFO, "dax option not supported");
1707 return -1;
5f3633e3 1708#endif
7915a861
AN		 1709 } else if (token == Opt_data_err_abort) {
1710 sbi->s_mount_opt |= m->mount_opt;
1711 } else if (token == Opt_data_err_ignore) {
1712 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1713 } else {
1714 if (!args->from)
1715 arg = 1;
1716 if (m->flags & MOPT_CLEAR)
1717 arg = !arg;
1718 else if (unlikely(!(m->flags & MOPT_SET))) {
1719 ext4_msg(sb, KERN_WARNING,
1720 "buggy handling of option %s", opt);
1721 WARN_ON(1);
1722 return -1;
1723 }
1724 if (arg != 0)
1725 sbi->s_mount_opt |= m->mount_opt;
1726 else
1727 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1728 }
5f3633e3 1729 return 1;
26092bf5
TT		 1730}
1731
1732static int parse_options(char *options, struct super_block *sb,
1733 unsigned long *journal_devnum,
1734 unsigned int *journal_ioprio,
1735 int is_remount)
1736{
1737 struct ext4_sb_info *sbi = EXT4_SB(sb);
1738 char *p;
1739 substring_t args[MAX_OPT_ARGS];
1740 int token;
1741
1742 if (!options)
1743 return 1;
1744
1745 while ((p = strsep(&options, ",")) != NULL) {
1746 if (!*p)
1747 continue;
1748 /*
1749 * Initialize args struct so we know whether arg was
1750 * found; some options take optional arguments.
1751 */
caecd0af 1752 args[0].to = args[0].from = NULL;
26092bf5
TT		 1753 token = match_token(p, tokens, args);
1754 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1755 journal_ioprio, is_remount) < 0)
1756 return 0;
ac27a0ec
DK		 1757 }
1758#ifdef CONFIG_QUOTA
e2b911c5 1759 if (ext4_has_feature_quota(sb) &&
262b4662 1760 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
c325a67c
TT		 1761 ext4_msg(sb, KERN_INFO, "Quota feature enabled, usrquota and grpquota "
1762 "mount options ignored.");
1763 clear_opt(sb, USRQUOTA);
1764 clear_opt(sb, GRPQUOTA);
1765 } else if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1766 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1767 clear_opt(sb, USRQUOTA);
ac27a0ec 1768
482a7425 1769 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1770 clear_opt(sb, GRPQUOTA);
ac27a0ec 1771
56c50f11 1772 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1773 ext4_msg(sb, KERN_ERR, "old and new quota "
1774 "format mixing");
ac27a0ec
DK		 1775 return 0;
1776 }
1777
1778 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1779 ext4_msg(sb, KERN_ERR, "journaled quota format "
1780 "not specified");
ac27a0ec
DK		 1781 return 0;
1782 }
ac27a0ec
DK		 1783 }
1784#endif
261cb20c
JK		 1785 if (test_opt(sb, DIOREAD_NOLOCK)) {
1786 int blocksize =
1787 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1788
09cbfeaf 1789 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 1790 ext4_msg(sb, KERN_ERR, "can't mount with "
1791 "dioread_nolock if block size != PAGE_SIZE");
1792 return 0;
1793 }
1794 }
d4f76107
JK		 1795 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1796 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1797 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1798 "in data=ordered mode");
1799 return 0;
1800 }
ac27a0ec
DK		 1801 return 1;
1802}
1803
2adf6da8
TT		 1804static inline void ext4_show_quota_options(struct seq_file *seq,
1805 struct super_block *sb)
1806{
1807#if defined(CONFIG_QUOTA)
1808 struct ext4_sb_info *sbi = EXT4_SB(sb);
1809
1810 if (sbi->s_jquota_fmt) {
1811 char *fmtname = "";
1812
1813 switch (sbi->s_jquota_fmt) {
1814 case QFMT_VFS_OLD:
1815 fmtname = "vfsold";
1816 break;
1817 case QFMT_VFS_V0:
1818 fmtname = "vfsv0";
1819 break;
1820 case QFMT_VFS_V1:
1821 fmtname = "vfsv1";
1822 break;
1823 }
1824 seq_printf(seq, ",jqfmt=%s", fmtname);
1825 }
1826
1827 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1828 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1829
1830 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1831 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1832#endif
1833}
1834
5a916be1
TT		 1835static const char *token2str(int token)
1836{
50df9fd5 1837 const struct match_token *t;
5a916be1
TT		 1838
1839 for (t = tokens; t->token != Opt_err; t++)
1840 if (t->token == token && !strchr(t->pattern, '='))
1841 break;
1842 return t->pattern;
1843}
1844
2adf6da8
TT		 1845/*
1846 * Show an option if
1847 * - it's set to a non-default value OR
1848 * - if the per-sb default is different from the global default
1849 */
66acdcf4
TT		 1850static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1851 int nodefs)
2adf6da8 1852{
2adf6da8
TT		 1853 struct ext4_sb_info *sbi = EXT4_SB(sb);
1854 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1855 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1856 const struct mount_opts *m;
66acdcf4 1857 char sep = nodefs ? '\n' : ',';
2adf6da8 1858
66acdcf4
TT		 1859#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1860#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1861
1862 if (sbi->s_sb_block != 1)
5a916be1
TT		 1863 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1864
1865 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1866 int want_set = m->flags & MOPT_SET;
1867 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1868 (m->flags & MOPT_CLEAR_ERR))
1869 continue;
66acdcf4 1870 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1871 continue; /* skip if same as the default */
1872 if ((want_set &&
1873 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1874 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1875 continue; /* select Opt_noFoo vs Opt_Foo */
1876 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1877 }
5a916be1 1878
08cefc7a 1879 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1880 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1881 SEQ_OPTS_PRINT("resuid=%u",
1882 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1883 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1884 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1885 SEQ_OPTS_PRINT("resgid=%u",
1886 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1887 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1888 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1889 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1890 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1891 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1892 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1893 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1894 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1895 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1896 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1897 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1898 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1899 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1900 if (sb->s_flags & MS_I_VERSION)
5a916be1 1901 SEQ_OPTS_PUTS("i_version");
66acdcf4 1902 if (nodefs || sbi->s_stripe)
5a916be1 1903 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1904 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1905 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1906 SEQ_OPTS_PUTS("data=journal");
1907 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1908 SEQ_OPTS_PUTS("data=ordered");
1909 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1910 SEQ_OPTS_PUTS("data=writeback");
1911 }
66acdcf4
TT		 1912 if (nodefs ||
1913 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1914 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1915 sbi->s_inode_readahead_blks);
2adf6da8 1916
66acdcf4
TT		 1917 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1918 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1919 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1920 if (nodefs || sbi->s_max_dir_size_kb)
1921 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 1922 if (test_opt(sb, DATA_ERR_ABORT))
1923 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 1924
1925 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1926 return 0;
1927}
1928
66acdcf4
TT		 1929static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1930{
1931 return _ext4_show_options(seq, root->d_sb, 0);
1932}
1933
ebd173be 1934int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 1935{
1936 struct super_block *sb = seq->private;
1937 int rc;
1938
1939 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1940 rc = _ext4_show_options(seq, sb, 1);
1941 seq_puts(seq, "\n");
1942 return rc;
1943}
1944
617ba13b 1945static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1946 int read_only)
1947{
617ba13b 1948 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1949 int res = 0;
1950
617ba13b 1951 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1952 ext4_msg(sb, KERN_ERR, "revision level too high, "
1953 "forcing read-only mode");
ac27a0ec
DK		 1954 res = MS_RDONLY;
1955 }
1956 if (read_only)
281b5995 1957 goto done;
617ba13b 1958 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1959 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1960 "running e2fsck is recommended");
c8b459f4 1961 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1962 ext4_msg(sb, KERN_WARNING,
1963 "warning: mounting fs with errors, "
1964 "running e2fsck is recommended");
ed3ce80a 1965 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1966 le16_to_cpu(es->s_mnt_count) >=
1967 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1968 ext4_msg(sb, KERN_WARNING,
1969 "warning: maximal mount count reached, "
1970 "running e2fsck is recommended");
ac27a0ec
DK		 1971 else if (le32_to_cpu(es->s_checkinterval) &&
1972 (le32_to_cpu(es->s_lastcheck) +
1973 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1974 ext4_msg(sb, KERN_WARNING,
1975 "warning: checktime reached, "
1976 "running e2fsck is recommended");
0b8e58a1 1977 if (!sbi->s_journal)
0390131b 1978 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1979 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1980 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1981 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1982 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1983 ext4_update_dynamic_rev(sb);
0390131b 1984 if (sbi->s_journal)
e2b911c5 1985 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 1986
e2d67052 1987 ext4_commit_super(sb, 1);
281b5995 1988done:
ac27a0ec 1989 if (test_opt(sb, DEBUG))
a9df9a49 1990 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1991 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1992 sb->s_blocksize,
1993 sbi->s_groups_count,
617ba13b
MC		 1994 EXT4_BLOCKS_PER_GROUP(sb),
1995 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1996 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1997
7abc52c2 1998 cleancache_init_fs(sb);
ac27a0ec
DK		 1999 return res;
2000}
2001
117fff10
TT		 2002int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2003{
2004 struct ext4_sb_info *sbi = EXT4_SB(sb);
2005 struct flex_groups *new_groups;
2006 int size;
2007
2008 if (!sbi->s_log_groups_per_flex)
2009 return 0;
2010
2011 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2012 if (size <= sbi->s_flex_groups_allocated)
2013 return 0;
2014
2015 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2016 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2017 if (!new_groups) {
2018 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2019 size / (int) sizeof(struct flex_groups));
2020 return -ENOMEM;
2021 }
2022
2023 if (sbi->s_flex_groups) {
2024 memcpy(new_groups, sbi->s_flex_groups,
2025 (sbi->s_flex_groups_allocated *
2026 sizeof(struct flex_groups)));
b93b41d4 2027 kvfree(sbi->s_flex_groups);
117fff10
TT		 2028 }
2029 sbi->s_flex_groups = new_groups;
2030 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2031 return 0;
2032}
2033
772cb7c8
JS		 2034static int ext4_fill_flex_info(struct super_block *sb)
2035{
2036 struct ext4_sb_info *sbi = EXT4_SB(sb);
2037 struct ext4_group_desc *gdp = NULL;
772cb7c8 2038 ext4_group_t flex_group;
117fff10 2039 int i, err;
772cb7c8 2040
503358ae 2041 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2042 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2043 sbi->s_log_groups_per_flex = 0;
2044 return 1;
2045 }
2046
117fff10
TT		 2047 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2048 if (err)
9933fc0a 2049 goto failed;
772cb7c8 2050
772cb7c8 2051 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2052 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2053
2054 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2055 atomic_add(ext4_free_inodes_count(sb, gdp),
2056 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2057 atomic64_add(ext4_free_group_clusters(sb, gdp),
2058 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2059 atomic_add(ext4_used_dirs_count(sb, gdp),
2060 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2061 }
2062
2063 return 1;
2064failed:
2065 return 0;
2066}
2067
e2b911c5 2068static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2069 struct ext4_group_desc *gdp)
717d50e4 2070{
b47820ed 2071 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2072 __u16 crc = 0;
feb0ab32 2073 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2074 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2075
9aa5d32b 2076 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2077 /* Use new metadata_csum algorithm */
feb0ab32 2078 __u32 csum32;
b47820ed 2079 __u16 dummy_csum = 0;
feb0ab32 2080
feb0ab32
DW		 2081 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2082 sizeof(le_group));
b47820ed
DJ		 2083 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2084 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2085 sizeof(dummy_csum));
2086 offset += sizeof(dummy_csum);
2087 if (offset < sbi->s_desc_size)
2088 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2089 sbi->s_desc_size - offset);
feb0ab32
DW		 2090
2091 crc = csum32 & 0xFFFF;
2092 goto out;
717d50e4
AD		 2093 }
2094
feb0ab32 2095 /* old crc16 code */
e2b911c5 2096 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2097 return 0;
2098
feb0ab32
DW		 2099 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2100 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2101 crc = crc16(crc, (__u8 *)gdp, offset);
2102 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2103 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2104 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2105 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2106 crc = crc16(crc, (__u8 *)gdp + offset,
2107 le16_to_cpu(sbi->s_es->s_desc_size) -
2108 offset);
2109
2110out:
717d50e4
AD		 2111 return cpu_to_le16(crc);
2112}
2113
feb0ab32 2114int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2115 struct ext4_group_desc *gdp)
2116{
feb0ab32 2117 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2118 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2119 return 0;
2120
2121 return 1;
2122}
2123
feb0ab32
DW		 2124void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2125 struct ext4_group_desc *gdp)
2126{
2127 if (!ext4_has_group_desc_csum(sb))
2128 return;
e2b911c5 2129 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2130}
2131
ac27a0ec 2132/* Called at mount-time, super-block is locked */
bfff6873
LC		 2133static int ext4_check_descriptors(struct super_block *sb,
2134 ext4_group_t *first_not_zeroed)
ac27a0ec 2135{
617ba13b
MC		 2136 struct ext4_sb_info *sbi = EXT4_SB(sb);
2137 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2138 ext4_fsblk_t last_block;
bd81d8ee
LV		 2139 ext4_fsblk_t block_bitmap;
2140 ext4_fsblk_t inode_bitmap;
2141 ext4_fsblk_t inode_table;
ce421581 2142 int flexbg_flag = 0;
bfff6873 2143 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2144
e2b911c5 2145 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2146 flexbg_flag = 1;
2147
af5bc92d 2148 ext4_debug("Checking group descriptors");
ac27a0ec 2149
197cd65a
AM		 2150 for (i = 0; i < sbi->s_groups_count; i++) {
2151 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2152
ce421581 2153 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2154 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2155 else
2156 last_block = first_block +
617ba13b 2157 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2158
bfff6873
LC		 2159 if ((grp == sbi->s_groups_count) &&
2160 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2161 grp = i;
2162
8fadc143 2163 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2164 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2165 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2166 "Block bitmap for group %u not in group "
b31e1552 2167 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2168 return 0;
2169 }
8fadc143 2170 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2171 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2172 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2173 "Inode bitmap for group %u not in group "
b31e1552 2174 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2175 return 0;
2176 }
8fadc143 2177 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2178 if (inode_table < first_block ||
2b2d6d01 2179 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2180 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2181 "Inode table for group %u not in group "
b31e1552 2182 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2183 return 0;
2184 }
955ce5f5 2185 ext4_lock_group(sb, i);
feb0ab32 2186 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2187 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2188 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2189 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2190 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2191 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2192 ext4_unlock_group(sb, i);
8a266467 2193 return 0;
7ee1ec4c 2194 }
717d50e4 2195 }
955ce5f5 2196 ext4_unlock_group(sb, i);
ce421581
JS		 2197 if (!flexbg_flag)
2198 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2199 }
bfff6873
LC		 2200 if (NULL != first_not_zeroed)
2201 *first_not_zeroed = grp;
ac27a0ec
DK		 2202 return 1;
2203}
2204
617ba13b 2205/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2206 * the superblock) which were deleted from all directories, but held open by
2207 * a process at the time of a crash. We walk the list and try to delete these
2208 * inodes at recovery time (only with a read-write filesystem).
2209 *
2210 * In order to keep the orphan inode chain consistent during traversal (in
2211 * case of crash during recovery), we link each inode into the superblock
2212 * orphan list_head and handle it the same way as an inode deletion during
2213 * normal operation (which journals the operations for us).
2214 *
2215 * We only do an iget() and an iput() on each inode, which is very safe if we
2216 * accidentally point at an in-use or already deleted inode. The worst that
2217 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2218 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2219 * e2fsck was run on this filesystem, and it must have already done the orphan
2220 * inode cleanup for us, so we can safely abort without any further action.
2221 */
2b2d6d01
TT		 2222static void ext4_orphan_cleanup(struct super_block *sb,
2223 struct ext4_super_block *es)
ac27a0ec
DK		 2224{
2225 unsigned int s_flags = sb->s_flags;
2226 int nr_orphans = 0, nr_truncates = 0;
2227#ifdef CONFIG_QUOTA
2228 int i;
2229#endif
2230 if (!es->s_last_orphan) {
2231 jbd_debug(4, "no orphan inodes to clean up\n");
2232 return;
2233 }
2234
a8f48a95 2235 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2236 ext4_msg(sb, KERN_ERR, "write access "
2237 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2238 return;
2239 }
2240
d39195c3
AG		 2241 /* Check if feature set would not allow a r/w mount */
2242 if (!ext4_feature_set_ok(sb, 0)) {
2243 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2244 "unknown ROCOMPAT features");
2245 return;
2246 }
2247
617ba13b 2248 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2249 /* don't clear list on RO mount w/ errors */
2250 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2251 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2252 "clearing orphan list.\n");
c25f9bc6
ES		 2253 es->s_last_orphan = 0;
2254 }
ac27a0ec
DK		 2255 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2256 return;
2257 }
2258
2259 if (s_flags & MS_RDONLY) {
b31e1552 2260 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2261 sb->s_flags &= ~MS_RDONLY;
2262 }
2263#ifdef CONFIG_QUOTA
2264 /* Needed for iput() to work correctly and not trash data */
2265 sb->s_flags |= MS_ACTIVE;
2266 /* Turn on quotas so that they are updated correctly */
a2d4a646 2267 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2268 if (EXT4_SB(sb)->s_qf_names[i]) {
2269 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2270 if (ret < 0)
b31e1552
ES		 2271 ext4_msg(sb, KERN_ERR,
2272 "Cannot turn on journaled "
2273 "quota: error %d", ret);
ac27a0ec
DK		 2274 }
2275 }
2276#endif
2277
2278 while (es->s_last_orphan) {
2279 struct inode *inode;
2280
97bd42b9
JB		 2281 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2282 if (IS_ERR(inode)) {
ac27a0ec
DK		 2283 es->s_last_orphan = 0;
2284 break;
2285 }
2286
617ba13b 2287 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2288 dquot_initialize(inode);
ac27a0ec 2289 if (inode->i_nlink) {
566370a2
PT		 2290 if (test_opt(sb, DEBUG))
2291 ext4_msg(sb, KERN_DEBUG,
2292 "%s: truncating inode %lu to %lld bytes",
2293 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2294 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2295 inode->i_ino, inode->i_size);
5955102c 2296 inode_lock(inode);
55f252c9 2297 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2298 ext4_truncate(inode);
5955102c 2299 inode_unlock(inode);
ac27a0ec
DK		 2300 nr_truncates++;
2301 } else {
566370a2
PT		 2302 if (test_opt(sb, DEBUG))
2303 ext4_msg(sb, KERN_DEBUG,
2304 "%s: deleting unreferenced inode %lu",
2305 __func__, inode->i_ino);
ac27a0ec
DK		 2306 jbd_debug(2, "deleting unreferenced inode %lu\n",
2307 inode->i_ino);
2308 nr_orphans++;
2309 }
2310 iput(inode); /* The delete magic happens here! */
2311 }
2312
2b2d6d01 2313#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2314
2315 if (nr_orphans)
b31e1552
ES		 2316 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2317 PLURAL(nr_orphans));
ac27a0ec 2318 if (nr_truncates)
b31e1552
ES		 2319 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2320 PLURAL(nr_truncates));
ac27a0ec
DK		 2321#ifdef CONFIG_QUOTA
2322 /* Turn quotas off */
a2d4a646 2323 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2324 if (sb_dqopt(sb)->files[i])
287a8095 2325 dquot_quota_off(sb, i);
ac27a0ec
DK		 2326 }
2327#endif
2328 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2329}
0b8e58a1 2330
cd2291a4
ES		 2331/*
2332 * Maximal extent format file size.
2333 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2334 * extent format containers, within a sector_t, and within i_blocks
2335 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2336 * so that won't be a limiting factor.
2337 *
f17722f9
LC		 2338 * However there is other limiting factor. We do store extents in the form
2339 * of starting block and length, hence the resulting length of the extent
2340 * covering maximum file size must fit into on-disk format containers as
2341 * well. Given that length is always by 1 unit bigger than max unit (because
2342 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2343 *
cd2291a4
ES		 2344 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2345 */
f287a1a5 2346static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2347{
2348 loff_t res;
2349 loff_t upper_limit = MAX_LFS_FILESIZE;
2350
2351 /* small i_blocks in vfs inode? */
f287a1a5 2352 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2353 /*
90c699a9 2354 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2355 * i_block represent total blocks in 512 bytes
2356 * 32 == size of vfs inode i_blocks * 8
2357 */
2358 upper_limit = (1LL << 32) - 1;
2359
2360 /* total blocks in file system block size */
2361 upper_limit >>= (blkbits - 9);
2362 upper_limit <<= blkbits;
2363 }
2364
f17722f9
LC		 2365 /*
2366 * 32-bit extent-start container, ee_block. We lower the maxbytes
2367 * by one fs block, so ee_len can cover the extent of maximum file
2368 * size
2369 */
2370 res = (1LL << 32) - 1;
cd2291a4 2371 res <<= blkbits;
cd2291a4
ES		 2372
2373 /* Sanity check against vm- & vfs- imposed limits */
2374 if (res > upper_limit)
2375 res = upper_limit;
2376
2377 return res;
2378}
ac27a0ec 2379
ac27a0ec 2380/*
cd2291a4 2381 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2382 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2383 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2384 */
f287a1a5 2385static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2386{
617ba13b 2387 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2388 int meta_blocks;
2389 loff_t upper_limit;
0b8e58a1
AD		 2390 /* This is calculated to be the largest file size for a dense, block
2391 * mapped file such that the file's total number of 512-byte sectors,
2392 * including data and all indirect blocks, does not exceed (2^48 - 1).
2393 *
2394 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2395 * number of 512-byte sectors of the file.
0fc1b451
AK		 2396 */
2397
f287a1a5 2398 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2399 /*
90c699a9 2400 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2401 * the inode i_block field represents total file blocks in
2402 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2403 */
2404 upper_limit = (1LL << 32) - 1;
2405
2406 /* total blocks in file system block size */
2407 upper_limit >>= (bits - 9);
2408
2409 } else {
8180a562
AK		 2410 /*
2411 * We use 48 bit ext4_inode i_blocks
2412 * With EXT4_HUGE_FILE_FL set the i_blocks
2413 * represent total number of blocks in
2414 * file system block size
2415 */
0fc1b451
AK		 2416 upper_limit = (1LL << 48) - 1;
2417
0fc1b451
AK		 2418 }
2419
2420 /* indirect blocks */
2421 meta_blocks = 1;
2422 /* double indirect blocks */
2423 meta_blocks += 1 + (1LL << (bits-2));
2424 /* tripple indirect blocks */
2425 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2426
2427 upper_limit -= meta_blocks;
2428 upper_limit <<= bits;
ac27a0ec
DK		 2429
2430 res += 1LL << (bits-2);
2431 res += 1LL << (2*(bits-2));
2432 res += 1LL << (3*(bits-2));
2433 res <<= bits;
2434 if (res > upper_limit)
2435 res = upper_limit;
0fc1b451
AK		 2436
2437 if (res > MAX_LFS_FILESIZE)
2438 res = MAX_LFS_FILESIZE;
2439
ac27a0ec
DK		 2440 return res;
2441}
2442
617ba13b 2443static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2444 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2445{
617ba13b 2446 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2447 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2448 int has_super = 0;
2449
2450 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2451
e2b911c5 2452 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2453 return logical_sb_block + nr + 1;
ac27a0ec 2454 bg = sbi->s_desc_per_block * nr;
617ba13b 2455 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2456 has_super = 1;
0b8e58a1 2457
bd63f6b0
DW		 2458 /*
2459 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2460 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2461 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2462 * compensate.
2463 */
2464 if (sb->s_blocksize == 1024 && nr == 0 &&
2465 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2466 has_super++;
2467
617ba13b 2468 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2469}
2470
c9de560d
AT		 2471/**
2472 * ext4_get_stripe_size: Get the stripe size.
2473 * @sbi: In memory super block info
2474 *
2475 * If we have specified it via mount option, then
2476 * use the mount option value. If the value specified at mount time is
2477 * greater than the blocks per group use the super block value.
2478 * If the super block value is greater than blocks per group return 0.
2479 * Allocator needs it be less than blocks per group.
2480 *
2481 */
2482static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2483{
2484 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2485 unsigned long stripe_width =
2486 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2487 int ret;
c9de560d
AT		 2488
2489 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2490 ret = sbi->s_stripe;
2491 else if (stripe_width <= sbi->s_blocks_per_group)
2492 ret = stripe_width;
2493 else if (stride <= sbi->s_blocks_per_group)
2494 ret = stride;
2495 else
2496 ret = 0;
c9de560d 2497
3eb08658
DE		 2498 /*
2499 * If the stripe width is 1, this makes no sense and
2500 * we set it to 0 to turn off stripe handling code.
2501 */
2502 if (ret <= 1)
2503 ret = 0;
c9de560d 2504
3eb08658 2505 return ret;
c9de560d 2506}
ac27a0ec 2507
a13fb1a4
ES		 2508/*
2509 * Check whether this filesystem can be mounted based on
2510 * the features present and the RDONLY/RDWR mount requested.
2511 * Returns 1 if this filesystem can be mounted as requested,
2512 * 0 if it cannot be.
2513 */
2514static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2515{
e2b911c5 2516 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2517 ext4_msg(sb, KERN_ERR,
2518 "Couldn't mount because of "
2519 "unsupported optional features (%x)",
2520 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2521 ~EXT4_FEATURE_INCOMPAT_SUPP));
2522 return 0;
2523 }
2524
2525 if (readonly)
2526 return 1;
2527
e2b911c5 2528 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b
DW		 2529 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2530 sb->s_flags |= MS_RDONLY;
2531 return 1;
2532 }
2533
a13fb1a4 2534 /* Check that feature set is OK for a read-write mount */
e2b911c5 2535 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2536 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2537 "unsupported optional features (%x)",
2538 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2539 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2540 return 0;
2541 }
2542 /*
2543 * Large file size enabled file system can only be mounted
2544 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2545 */
e2b911c5 2546 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2547 if (sizeof(blkcnt_t) < sizeof(u64)) {
2548 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2549 "cannot be mounted RDWR without "
2550 "CONFIG_LBDAF");
2551 return 0;
2552 }
2553 }
e2b911c5 2554 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2555 ext4_msg(sb, KERN_ERR,
2556 "Can't support bigalloc feature without "
2557 "extents feature\n");
2558 return 0;
2559 }
7c319d32
AK		 2560
2561#ifndef CONFIG_QUOTA
e2b911c5 2562 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2563 ext4_msg(sb, KERN_ERR,
2564 "Filesystem with quota feature cannot be mounted RDWR "
2565 "without CONFIG_QUOTA");
2566 return 0;
2567 }
689c958c
LX		 2568 if (ext4_has_feature_project(sb) && !readonly) {
2569 ext4_msg(sb, KERN_ERR,
2570 "Filesystem with project quota feature cannot be mounted RDWR "
2571 "without CONFIG_QUOTA");
2572 return 0;
2573 }
7c319d32 2574#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2575 return 1;
2576}
2577
66e61a9e
TT		 2578/*
2579 * This function is called once a day if we have errors logged
2580 * on the file system
2581 */
2582static void print_daily_error_info(unsigned long arg)
2583{
2584 struct super_block *sb = (struct super_block *) arg;
2585 struct ext4_sb_info *sbi;
2586 struct ext4_super_block *es;
2587
2588 sbi = EXT4_SB(sb);
2589 es = sbi->s_es;
2590
2591 if (es->s_error_count)
ae0f78de
TT		 2592 /* fsck newer than v1.41.13 is needed to clean this condition. */
2593 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2594 le32_to_cpu(es->s_error_count));
2595 if (es->s_first_error_time) {
ae0f78de 2596 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2597 sb->s_id, le32_to_cpu(es->s_first_error_time),
2598 (int) sizeof(es->s_first_error_func),
2599 es->s_first_error_func,
2600 le32_to_cpu(es->s_first_error_line));
2601 if (es->s_first_error_ino)
2602 printk(": inode %u",
2603 le32_to_cpu(es->s_first_error_ino));
2604 if (es->s_first_error_block)
2605 printk(": block %llu", (unsigned long long)
2606 le64_to_cpu(es->s_first_error_block));
2607 printk("\n");
2608 }
2609 if (es->s_last_error_time) {
ae0f78de 2610 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2611 sb->s_id, le32_to_cpu(es->s_last_error_time),
2612 (int) sizeof(es->s_last_error_func),
2613 es->s_last_error_func,
2614 le32_to_cpu(es->s_last_error_line));
2615 if (es->s_last_error_ino)
2616 printk(": inode %u",
2617 le32_to_cpu(es->s_last_error_ino));
2618 if (es->s_last_error_block)
2619 printk(": block %llu", (unsigned long long)
2620 le64_to_cpu(es->s_last_error_block));
2621 printk("\n");
2622 }
2623 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2624}
2625
bfff6873
LC		 2626/* Find next suitable group and run ext4_init_inode_table */
2627static int ext4_run_li_request(struct ext4_li_request *elr)
2628{
2629 struct ext4_group_desc *gdp = NULL;
2630 ext4_group_t group, ngroups;
2631 struct super_block *sb;
2632 unsigned long timeout = 0;
2633 int ret = 0;
2634
2635 sb = elr->lr_super;
2636 ngroups = EXT4_SB(sb)->s_groups_count;
2637
8e8ad8a5 2638 sb_start_write(sb);
bfff6873
LC		 2639 for (group = elr->lr_next_group; group < ngroups; group++) {
2640 gdp = ext4_get_group_desc(sb, group, NULL);
2641 if (!gdp) {
2642 ret = 1;
2643 break;
2644 }
2645
2646 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2647 break;
2648 }
2649
7f511862 2650 if (group >= ngroups)
bfff6873
LC		 2651 ret = 1;
2652
2653 if (!ret) {
2654 timeout = jiffies;
2655 ret = ext4_init_inode_table(sb, group,
2656 elr->lr_timeout ? 0 : 1);
2657 if (elr->lr_timeout == 0) {
51ce6511
LC		 2658 timeout = (jiffies - timeout) *
2659 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2660 elr->lr_timeout = timeout;
2661 }
2662 elr->lr_next_sched = jiffies + elr->lr_timeout;
2663 elr->lr_next_group = group + 1;
2664 }
8e8ad8a5 2665 sb_end_write(sb);
bfff6873
LC		 2666
2667 return ret;
2668}
2669
2670/*
2671 * Remove lr_request from the list_request and free the
4ed5c033 2672 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2673 */
2674static void ext4_remove_li_request(struct ext4_li_request *elr)
2675{
2676 struct ext4_sb_info *sbi;
2677
2678 if (!elr)
2679 return;
2680
2681 sbi = elr->lr_sbi;
2682
2683 list_del(&elr->lr_request);
2684 sbi->s_li_request = NULL;
2685 kfree(elr);
2686}
2687
2688static void ext4_unregister_li_request(struct super_block *sb)
2689{
1bb933fb
LC		 2690 mutex_lock(&ext4_li_mtx);
2691 if (!ext4_li_info) {
2692 mutex_unlock(&ext4_li_mtx);
bfff6873 2693 return;
1bb933fb 2694 }
bfff6873
LC		 2695
2696 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2697 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2698 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2699 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2700}
2701
8f1f7453
ES		 2702static struct task_struct *ext4_lazyinit_task;
2703
bfff6873
LC		 2704/*
2705 * This is the function where ext4lazyinit thread lives. It walks
2706 * through the request list searching for next scheduled filesystem.
2707 * When such a fs is found, run the lazy initialization request
2708 * (ext4_rn_li_request) and keep track of the time spend in this
2709 * function. Based on that time we compute next schedule time of
2710 * the request. When walking through the list is complete, compute
2711 * next waking time and put itself into sleep.
2712 */
2713static int ext4_lazyinit_thread(void *arg)
2714{
2715 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2716 struct list_head *pos, *n;
2717 struct ext4_li_request *elr;
4ed5c033 2718 unsigned long next_wakeup, cur;
bfff6873
LC		 2719
2720 BUG_ON(NULL == eli);
2721
bfff6873
LC		 2722cont_thread:
2723 while (true) {
2724 next_wakeup = MAX_JIFFY_OFFSET;
2725
2726 mutex_lock(&eli->li_list_mtx);
2727 if (list_empty(&eli->li_request_list)) {
2728 mutex_unlock(&eli->li_list_mtx);
2729 goto exit_thread;
2730 }
2731
2732 list_for_each_safe(pos, n, &eli->li_request_list) {
2733 elr = list_entry(pos, struct ext4_li_request,
2734 lr_request);
2735
b2c78cd0
TT		 2736 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2737 if (ext4_run_li_request(elr) != 0) {
2738 /* error, remove the lazy_init job */
2739 ext4_remove_li_request(elr);
2740 continue;
2741 }
bfff6873
LC		 2742 }
2743
2744 if (time_before(elr->lr_next_sched, next_wakeup))
2745 next_wakeup = elr->lr_next_sched;
2746 }
2747 mutex_unlock(&eli->li_list_mtx);
2748
a0acae0e 2749 try_to_freeze();
bfff6873 2750
4ed5c033
LC		 2751 cur = jiffies;
2752 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2753 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2754 cond_resched();
2755 continue;
2756 }
2757
4ed5c033
LC		 2758 schedule_timeout_interruptible(next_wakeup - cur);
2759
8f1f7453
ES		 2760 if (kthread_should_stop()) {
2761 ext4_clear_request_list();
2762 goto exit_thread;
2763 }
bfff6873
LC		 2764 }
2765
2766exit_thread:
2767 /*
2768 * It looks like the request list is empty, but we need
2769 * to check it under the li_list_mtx lock, to prevent any
2770 * additions into it, and of course we should lock ext4_li_mtx
2771 * to atomically free the list and ext4_li_info, because at
2772 * this point another ext4 filesystem could be registering
2773 * new one.
2774 */
2775 mutex_lock(&ext4_li_mtx);
2776 mutex_lock(&eli->li_list_mtx);
2777 if (!list_empty(&eli->li_request_list)) {
2778 mutex_unlock(&eli->li_list_mtx);
2779 mutex_unlock(&ext4_li_mtx);
2780 goto cont_thread;
2781 }
2782 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2783 kfree(ext4_li_info);
2784 ext4_li_info = NULL;
2785 mutex_unlock(&ext4_li_mtx);
2786
2787 return 0;
2788}
2789
2790static void ext4_clear_request_list(void)
2791{
2792 struct list_head *pos, *n;
2793 struct ext4_li_request *elr;
2794
2795 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2796 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2797 elr = list_entry(pos, struct ext4_li_request,
2798 lr_request);
2799 ext4_remove_li_request(elr);
2800 }
2801 mutex_unlock(&ext4_li_info->li_list_mtx);
2802}
2803
2804static int ext4_run_lazyinit_thread(void)
2805{
8f1f7453
ES		 2806 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2807 ext4_li_info, "ext4lazyinit");
2808 if (IS_ERR(ext4_lazyinit_task)) {
2809 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2810 ext4_clear_request_list();
bfff6873
LC		 2811 kfree(ext4_li_info);
2812 ext4_li_info = NULL;
92b97816 2813 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2814 "initialization thread\n",
2815 err);
2816 return err;
2817 }
2818 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2819 return 0;
2820}
2821
2822/*
2823 * Check whether it make sense to run itable init. thread or not.
2824 * If there is at least one uninitialized inode table, return
2825 * corresponding group number, else the loop goes through all
2826 * groups and return total number of groups.
2827 */
2828static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2829{
2830 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2831 struct ext4_group_desc *gdp = NULL;
2832
2833 for (group = 0; group < ngroups; group++) {
2834 gdp = ext4_get_group_desc(sb, group, NULL);
2835 if (!gdp)
2836 continue;
2837
2838 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2839 break;
2840 }
2841
2842 return group;
2843}
2844
2845static int ext4_li_info_new(void)
2846{
2847 struct ext4_lazy_init *eli = NULL;
2848
2849 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2850 if (!eli)
2851 return -ENOMEM;
2852
bfff6873
LC		 2853 INIT_LIST_HEAD(&eli->li_request_list);
2854 mutex_init(&eli->li_list_mtx);
2855
bfff6873
LC		 2856 eli->li_state |= EXT4_LAZYINIT_QUIT;
2857
2858 ext4_li_info = eli;
2859
2860 return 0;
2861}
2862
2863static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2864 ext4_group_t start)
2865{
2866 struct ext4_sb_info *sbi = EXT4_SB(sb);
2867 struct ext4_li_request *elr;
bfff6873
LC		 2868
2869 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2870 if (!elr)
2871 return NULL;
2872
2873 elr->lr_super = sb;
2874 elr->lr_sbi = sbi;
2875 elr->lr_next_group = start;
2876
2877 /*
2878 * Randomize first schedule time of the request to
2879 * spread the inode table initialization requests
2880 * better.
2881 */
dd1f723b
TT		 2882 elr->lr_next_sched = jiffies + (prandom_u32() %
2883 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 2884 return elr;
2885}
2886
7f511862
TT		 2887int ext4_register_li_request(struct super_block *sb,
2888 ext4_group_t first_not_zeroed)
bfff6873
LC		 2889{
2890 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 2891 struct ext4_li_request *elr = NULL;
bfff6873 2892 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2893 int ret = 0;
bfff6873 2894
7f511862 2895 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 2896 if (sbi->s_li_request != NULL) {
2897 /*
2898 * Reset timeout so it can be computed again, because
2899 * s_li_wait_mult might have changed.
2900 */
2901 sbi->s_li_request->lr_timeout = 0;
7f511862 2902 goto out;
51ce6511 2903 }
bfff6873
LC		 2904
2905 if (first_not_zeroed == ngroups ||
2906 (sb->s_flags & MS_RDONLY) ||
55ff3840 2907 !test_opt(sb, INIT_INODE_TABLE))
7f511862 2908 goto out;
bfff6873
LC		 2909
2910 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 2911 if (!elr) {
2912 ret = -ENOMEM;
2913 goto out;
2914 }
bfff6873
LC		 2915
2916 if (NULL == ext4_li_info) {
2917 ret = ext4_li_info_new();
2918 if (ret)
2919 goto out;
2920 }
2921
2922 mutex_lock(&ext4_li_info->li_list_mtx);
2923 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2924 mutex_unlock(&ext4_li_info->li_list_mtx);
2925
2926 sbi->s_li_request = elr;
46e4690b
TM		 2927 /*
2928 * set elr to NULL here since it has been inserted to
2929 * the request_list and the removal and free of it is
2930 * handled by ext4_clear_request_list from now on.
2931 */
2932 elr = NULL;
bfff6873
LC		 2933
2934 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2935 ret = ext4_run_lazyinit_thread();
2936 if (ret)
2937 goto out;
2938 }
bfff6873 2939out:
beed5ecb
NK		 2940 mutex_unlock(&ext4_li_mtx);
2941 if (ret)
bfff6873 2942 kfree(elr);
bfff6873
LC		 2943 return ret;
2944}
2945
2946/*
2947 * We do not need to lock anything since this is called on
2948 * module unload.
2949 */
2950static void ext4_destroy_lazyinit_thread(void)
2951{
2952 /*
2953 * If thread exited earlier
2954 * there's nothing to be done.
2955 */
8f1f7453 2956 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2957 return;
2958
8f1f7453 2959 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2960}
2961
25ed6e8a
DW		 2962static int set_journal_csum_feature_set(struct super_block *sb)
2963{
2964 int ret = 1;
2965 int compat, incompat;
2966 struct ext4_sb_info *sbi = EXT4_SB(sb);
2967
9aa5d32b 2968 if (ext4_has_metadata_csum(sb)) {
db9ee220 2969 /* journal checksum v3 */
25ed6e8a 2970 compat = 0;
db9ee220 2971 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 2972 } else {
2973 /* journal checksum v1 */
2974 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2975 incompat = 0;
2976 }
2977
feb8c6d3
DW		 2978 jbd2_journal_clear_features(sbi->s_journal,
2979 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2980 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2981 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 2982 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2983 ret = jbd2_journal_set_features(sbi->s_journal,
2984 compat, 0,
2985 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2986 incompat);
2987 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2988 ret = jbd2_journal_set_features(sbi->s_journal,
2989 compat, 0,
2990 incompat);
2991 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2992 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2993 } else {
feb8c6d3
DW		 2994 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2995 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 2996 }
2997
2998 return ret;
2999}
3000
952fc18e
TT		 3001/*
3002 * Note: calculating the overhead so we can be compatible with
3003 * historical BSD practice is quite difficult in the face of
3004 * clusters/bigalloc. This is because multiple metadata blocks from
3005 * different block group can end up in the same allocation cluster.
3006 * Calculating the exact overhead in the face of clustered allocation
3007 * requires either O(all block bitmaps) in memory or O(number of block
3008 * groups**2) in time. We will still calculate the superblock for
3009 * older file systems --- and if we come across with a bigalloc file
3010 * system with zero in s_overhead_clusters the estimate will be close to
3011 * correct especially for very large cluster sizes --- but for newer
3012 * file systems, it's better to calculate this figure once at mkfs
3013 * time, and store it in the superblock. If the superblock value is
3014 * present (even for non-bigalloc file systems), we will use it.
3015 */
3016static int count_overhead(struct super_block *sb, ext4_group_t grp,
3017 char *buf)
3018{
3019 struct ext4_sb_info *sbi = EXT4_SB(sb);
3020 struct ext4_group_desc *gdp;
3021 ext4_fsblk_t first_block, last_block, b;
3022 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3023 int s, j, count = 0;
3024
e2b911c5 3025 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3026 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3027 sbi->s_itb_per_group + 2);
3028
952fc18e
TT		 3029 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3030 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3031 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3032 for (i = 0; i < ngroups; i++) {
3033 gdp = ext4_get_group_desc(sb, i, NULL);
3034 b = ext4_block_bitmap(sb, gdp);
3035 if (b >= first_block && b <= last_block) {
3036 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3037 count++;
3038 }
3039 b = ext4_inode_bitmap(sb, gdp);
3040 if (b >= first_block && b <= last_block) {
3041 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3042 count++;
3043 }
3044 b = ext4_inode_table(sb, gdp);
3045 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3046 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3047 int c = EXT4_B2C(sbi, b - first_block);
3048 ext4_set_bit(c, buf);
3049 count++;
3050 }
3051 if (i != grp)
3052 continue;
3053 s = 0;
3054 if (ext4_bg_has_super(sb, grp)) {
3055 ext4_set_bit(s++, buf);
3056 count++;
3057 }
3058 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3059 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3060 count++;
3061 }
3062 }
3063 if (!count)
3064 return 0;
3065 return EXT4_CLUSTERS_PER_GROUP(sb) -
3066 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3067}
3068
3069/*
3070 * Compute the overhead and stash it in sbi->s_overhead
3071 */
3072int ext4_calculate_overhead(struct super_block *sb)
3073{
3074 struct ext4_sb_info *sbi = EXT4_SB(sb);
3075 struct ext4_super_block *es = sbi->s_es;
3076 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3077 ext4_fsblk_t overhead = 0;
4fdb5543 3078 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3079
952fc18e
TT		 3080 if (!buf)
3081 return -ENOMEM;
3082
3083 /*
3084 * Compute the overhead (FS structures). This is constant
3085 * for a given filesystem unless the number of block groups
3086 * changes so we cache the previous value until it does.
3087 */
3088
3089 /*
3090 * All of the blocks before first_data_block are overhead
3091 */
3092 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3093
3094 /*
3095 * Add the overhead found in each block group
3096 */
3097 for (i = 0; i < ngroups; i++) {
3098 int blks;
3099
3100 blks = count_overhead(sb, i, buf);
3101 overhead += blks;
3102 if (blks)
3103 memset(buf, 0, PAGE_SIZE);
3104 cond_resched();
3105 }
b003b524
ES		 3106 /* Add the internal journal blocks as well */
3107 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3108 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3109
952fc18e
TT		 3110 sbi->s_overhead = overhead;
3111 smp_wmb();
3112 free_page((unsigned long) buf);
3113 return 0;
3114}
3115
b5799018 3116static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3117{
3118 ext4_fsblk_t resv_clusters;
b5799018 3119 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3120
30fac0f7
JK		 3121 /*
3122 * There's no need to reserve anything when we aren't using extents.
3123 * The space estimates are exact, there are no unwritten extents,
3124 * hole punching doesn't need new metadata... This is needed especially
3125 * to keep ext2/3 backward compatibility.
3126 */
e2b911c5 3127 if (!ext4_has_feature_extents(sb))
b5799018 3128 return;
27dd4385
LC		 3129 /*
3130 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3131 * This should cover the situations where we can not afford to run
3132 * out of space like for example punch hole, or converting
556615dc 3133 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3134 * allocation would require 1, or 2 blocks, higher numbers are
3135 * very rare.
3136 */
b5799018
TT		 3137 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3138 sbi->s_cluster_bits);
27dd4385
LC		 3139
3140 do_div(resv_clusters, 50);
3141 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3142
b5799018 3143 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3144}
3145
2b2d6d01 3146static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3147{
d4c402d9 3148 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3149 struct buffer_head *bh;
617ba13b
MC		 3150 struct ext4_super_block *es = NULL;
3151 struct ext4_sb_info *sbi;
3152 ext4_fsblk_t block;
3153 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3154 ext4_fsblk_t logical_sb_block;
ac27a0ec 3155 unsigned long offset = 0;
ac27a0ec
DK		 3156 unsigned long journal_devnum = 0;
3157 unsigned long def_mount_opts;
3158 struct inode *root;
0390131b 3159 const char *descr;
dcc7dae3 3160 int ret = -ENOMEM;
281b5995 3161 int blocksize, clustersize;
4ec11028
TT		 3162 unsigned int db_count;
3163 unsigned int i;
281b5995 3164 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3165 __u64 blocks_count;
07aa2ea1 3166 int err = 0;
b3881f74 3167 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3168 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3169
3170 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3171 if (!sbi)
dcc7dae3 3172 goto out_free_orig;
705895b6
PE		 3173
3174 sbi->s_blockgroup_lock =
3175 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3176 if (!sbi->s_blockgroup_lock) {
3177 kfree(sbi);
dcc7dae3 3178 goto out_free_orig;
705895b6 3179 }
ac27a0ec 3180 sb->s_fs_info = sbi;
2c0544b2 3181 sbi->s_sb = sb;
240799cd 3182 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3183 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3184 if (sb->s_bdev->bd_part)
3185 sbi->s_sectors_written_start =
3186 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3187
9f6200bb 3188 /* Cleanup superblock name */
ec3904dc 3189 strreplace(sb->s_id, '/', '!');
9f6200bb 3190
07aa2ea1 3191 /* -EINVAL is default */
dcc7dae3 3192 ret = -EINVAL;
617ba13b 3193 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3194 if (!blocksize) {
b31e1552 3195 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3196 goto out_fail;
3197 }
3198
3199 /*
617ba13b 3200 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3201 * block sizes. We need to calculate the offset from buffer start.
3202 */
617ba13b 3203 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3204 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3205 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3206 } else {
70bbb3e0 3207 logical_sb_block = sb_block;
ac27a0ec
DK		 3208 }
3209
a8ac900b 3210 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3211 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3212 goto out_fail;
3213 }
3214 /*
3215 * Note: s_es must be initialized as soon as possible because
617ba13b 3216 * some ext4 macro-instructions depend on its value
ac27a0ec 3217 */
2716b802 3218 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3219 sbi->s_es = es;
3220 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3221 if (sb->s_magic != EXT4_SUPER_MAGIC)
3222 goto cantfind_ext4;
afc32f7e 3223 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3224
feb0ab32 3225 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3226 if (ext4_has_feature_metadata_csum(sb) &&
3227 ext4_has_feature_gdt_csum(sb))
363307e6 3228 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3229 "redundant flags; please run fsck.");
3230
d25425f8
DW		 3231 /* Check for a known checksum algorithm */
3232 if (!ext4_verify_csum_type(sb, es)) {
3233 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3234 "unknown checksum algorithm.");
3235 silent = 1;
3236 goto cantfind_ext4;
3237 }
3238
0441984a 3239 /* Load the checksum driver */
e2b911c5 3240 if (ext4_has_feature_metadata_csum(sb)) {
0441984a
DW		 3241 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3242 if (IS_ERR(sbi->s_chksum_driver)) {
3243 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3244 ret = PTR_ERR(sbi->s_chksum_driver);
3245 sbi->s_chksum_driver = NULL;
3246 goto failed_mount;
3247 }
3248 }
3249
a9c47317
DW		 3250 /* Check superblock checksum */
3251 if (!ext4_superblock_csum_verify(sb, es)) {
3252 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3253 "invalid superblock checksum. Run e2fsck?");
3254 silent = 1;
6a797d27 3255 ret = -EFSBADCRC;
a9c47317
DW		 3256 goto cantfind_ext4;
3257 }
3258
3259 /* Precompute checksum seed for all metadata */
e2b911c5 3260 if (ext4_has_feature_csum_seed(sb))
8c81bd8f
DW		 3261 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3262 else if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3263 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3264 sizeof(es->s_uuid));
3265
ac27a0ec
DK		 3266 /* Set defaults before we parse the mount options */
3267 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3268 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3269 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3270 set_opt(sb, DEBUG);
87f26807 3271 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3272 set_opt(sb, GRPID);
617ba13b 3273 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3274 set_opt(sb, NO_UID32);
ea663336 3275 /* xattr user namespace & acls are now defaulted on */
ea663336 3276 set_opt(sb, XATTR_USER);
03010a33 3277#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3278 set_opt(sb, POSIX_ACL);
2e7842b8 3279#endif
98c1a759
DW		 3280 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3281 if (ext4_has_metadata_csum(sb))
3282 set_opt(sb, JOURNAL_CHECKSUM);
3283
617ba13b 3284 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3285 set_opt(sb, JOURNAL_DATA);
617ba13b 3286 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3287 set_opt(sb, ORDERED_DATA);
617ba13b 3288 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3289 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3290
3291 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3292 set_opt(sb, ERRORS_PANIC);
bb4f397a 3293 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3294 set_opt(sb, ERRORS_CONT);
bb4f397a 3295 else
fd8c37ec 3296 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3297 /* block_validity enabled by default; disable with noblock_validity */
3298 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3299 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3300 set_opt(sb, DISCARD);
ac27a0ec 3301
08cefc7a
EB		 3302 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3303 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3304 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3305 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3306 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3307
8b67f04a 3308 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3309 set_opt(sb, BARRIER);
ac27a0ec 3310
dd919b98
AK		 3311 /*
3312 * enable delayed allocation by default
3313 * Use -o nodelalloc to turn it off
3314 */
bc0b75f7 3315 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3316 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3317 set_opt(sb, DELALLOC);
dd919b98 3318
51ce6511
LC		 3319 /*
3320 * set default s_li_wait_mult for lazyinit, for the case there is
3321 * no mount option specified.
3322 */
3323 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3324
8b67f04a 3325 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3326 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3327 ext4_msg(sb, KERN_WARNING,
3328 "failed to parse options in superblock: %s",
3329 sbi->s_es->s_mount_opts);
3330 }
5a916be1 3331 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3332 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3333 &journal_ioprio, 0))
ac27a0ec
DK		 3334 goto failed_mount;
3335
56889787
TT		 3336 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3337 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3338 "with data=journal disables delayed "
3339 "allocation and O_DIRECT support!\n");
3340 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3341 ext4_msg(sb, KERN_ERR, "can't mount with "
3342 "both data=journal and delalloc");
3343 goto failed_mount;
3344 }
3345 if (test_opt(sb, DIOREAD_NOLOCK)) {
3346 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3347 "both data=journal and dioread_nolock");
56889787
TT		 3348 goto failed_mount;
3349 }
923ae0ff
RZ		 3350 if (test_opt(sb, DAX)) {
3351 ext4_msg(sb, KERN_ERR, "can't mount with "
3352 "both data=journal and dax");
3353 goto failed_mount;
3354 }
56889787
TT		 3355 if (test_opt(sb, DELALLOC))
3356 clear_opt(sb, DELALLOC);
001e4a87
TH		 3357 } else {
3358 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3359 }
3360
ac27a0ec 3361 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3362 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3363
617ba13b 3364 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3365 (ext4_has_compat_features(sb) ||
3366 ext4_has_ro_compat_features(sb) ||
3367 ext4_has_incompat_features(sb)))
b31e1552
ES		 3368 ext4_msg(sb, KERN_WARNING,
3369 "feature flags set on rev 0 fs, "
3370 "running e2fsck is recommended");
469108ff 3371
ed3654eb
TT		 3372 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3373 set_opt2(sb, HURD_COMPAT);
e2b911c5 3374 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3375 ext4_msg(sb, KERN_ERR,
3376 "The Hurd can't support 64-bit file systems");
3377 goto failed_mount;
3378 }
3379 }
3380
2035e776
TT		 3381 if (IS_EXT2_SB(sb)) {
3382 if (ext2_feature_set_ok(sb))
3383 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3384 "using the ext4 subsystem");
3385 else {
3386 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3387 "to feature incompatibilities");
3388 goto failed_mount;
3389 }
3390 }
3391
3392 if (IS_EXT3_SB(sb)) {
3393 if (ext3_feature_set_ok(sb))
3394 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3395 "using the ext4 subsystem");
3396 else {
3397 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3398 "to feature incompatibilities");
3399 goto failed_mount;
3400 }
3401 }
3402
ac27a0ec
DK		 3403 /*
3404 * Check feature flags regardless of the revision level, since we
3405 * previously didn't change the revision level when setting the flags,
3406 * so there is a chance incompat flags are set on a rev 0 filesystem.
3407 */
a13fb1a4 3408 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3409 goto failed_mount;
a13fb1a4 3410
261cb20c 3411 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3412 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3413 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3414 ext4_msg(sb, KERN_ERR,
3415 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3416 goto failed_mount;
3417 }
3418
923ae0ff 3419 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
87eefeb4
TK		 3420 err = bdev_dax_supported(sb, blocksize);
3421 if (err)
923ae0ff 3422 goto failed_mount;
923ae0ff
RZ		 3423 }
3424
e2b911c5 3425 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3426 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3427 es->s_encryption_level);
3428 goto failed_mount;
3429 }
3430
ac27a0ec 3431 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3432 /* Validate the filesystem blocksize */
3433 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3434 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3435 blocksize);
ac27a0ec
DK		 3436 goto failed_mount;
3437 }
3438
2b2d6d01 3439 brelse(bh);
70bbb3e0
AM		 3440 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3441 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3442 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3443 if (!bh) {
b31e1552
ES		 3444 ext4_msg(sb, KERN_ERR,
3445 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3446 goto failed_mount;
3447 }
2716b802 3448 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3449 sbi->s_es = es;
617ba13b 3450 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3451 ext4_msg(sb, KERN_ERR,
3452 "Magic mismatch, very weird!");
ac27a0ec
DK		 3453 goto failed_mount;
3454 }
3455 }
3456
e2b911c5 3457 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3458 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3459 has_huge_files);
3460 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3461
617ba13b
MC		 3462 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3463 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3464 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3465 } else {
3466 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3467 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3468 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3469 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3470 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3471 ext4_msg(sb, KERN_ERR,
3472 "unsupported inode size: %d",
2b2d6d01 3473 sbi->s_inode_size);
ac27a0ec
DK		 3474 goto failed_mount;
3475 }
ef7f3835
KS		 3476 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3477 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3478 }
0b8e58a1 3479
0d1ee42f 3480 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3481 if (ext4_has_feature_64bit(sb)) {
8fadc143 3482 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3483 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3484 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3485 ext4_msg(sb, KERN_ERR,
3486 "unsupported descriptor size %lu",
0d1ee42f
AR		 3487 sbi->s_desc_size);
3488 goto failed_mount;
3489 }
3490 } else
3491 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3492
ac27a0ec 3493 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3494 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3495 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3496 goto cantfind_ext4;
0b8e58a1 3497
617ba13b 3498 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3499 if (sbi->s_inodes_per_block == 0)
617ba13b 3500 goto cantfind_ext4;
ac27a0ec
DK		 3501 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3502 sbi->s_inodes_per_block;
0d1ee42f 3503 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3504 sbi->s_sbh = bh;
3505 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3506 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3507 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3508
2b2d6d01 3509 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3510 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3511 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3512 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3513 i = le32_to_cpu(es->s_flags);
3514 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3515 sbi->s_hash_unsigned = 3;
3516 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3517#ifdef __CHAR_UNSIGNED__
23301410
TT		 3518 if (!(sb->s_flags & MS_RDONLY))
3519 es->s_flags |=
3520 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3521 sbi->s_hash_unsigned = 3;
f99b2589 3522#else
23301410
TT		 3523 if (!(sb->s_flags & MS_RDONLY))
3524 es->s_flags |=
3525 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3526#endif
23301410 3527 }
f99b2589 3528 }
ac27a0ec 3529
281b5995
TT		 3530 /* Handle clustersize */
3531 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3532 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3533 if (has_bigalloc) {
3534 if (clustersize < blocksize) {
3535 ext4_msg(sb, KERN_ERR,
3536 "cluster size (%d) smaller than "
3537 "block size (%d)", clustersize, blocksize);
3538 goto failed_mount;
3539 }
3540 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3541 le32_to_cpu(es->s_log_block_size);
3542 sbi->s_clusters_per_group =
3543 le32_to_cpu(es->s_clusters_per_group);
3544 if (sbi->s_clusters_per_group > blocksize * 8) {
3545 ext4_msg(sb, KERN_ERR,
3546 "#clusters per group too big: %lu",
3547 sbi->s_clusters_per_group);
3548 goto failed_mount;
3549 }
3550 if (sbi->s_blocks_per_group !=
3551 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3552 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3553 "clusters per group (%lu) inconsistent",
3554 sbi->s_blocks_per_group,
3555 sbi->s_clusters_per_group);
3556 goto failed_mount;
3557 }
3558 } else {
3559 if (clustersize != blocksize) {
3560 ext4_warning(sb, "fragment/cluster size (%d) != "
3561 "block size (%d)", clustersize,
3562 blocksize);
3563 clustersize = blocksize;
3564 }
3565 if (sbi->s_blocks_per_group > blocksize * 8) {
3566 ext4_msg(sb, KERN_ERR,
3567 "#blocks per group too big: %lu",
3568 sbi->s_blocks_per_group);
3569 goto failed_mount;
3570 }
3571 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3572 sbi->s_cluster_bits = 0;
ac27a0ec 3573 }
281b5995
TT		 3574 sbi->s_cluster_ratio = clustersize / blocksize;
3575
ac27a0ec 3576 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3577 ext4_msg(sb, KERN_ERR,
3578 "#inodes per group too big: %lu",
2b2d6d01 3579 sbi->s_inodes_per_group);
ac27a0ec
DK		 3580 goto failed_mount;
3581 }
3582
960fd856
TT		 3583 /* Do we have standard group size of clustersize * 8 blocks ? */
3584 if (sbi->s_blocks_per_group == clustersize << 3)
3585 set_opt2(sb, STD_GROUP_SIZE);
3586
bf43d84b
ES		 3587 /*
3588 * Test whether we have more sectors than will fit in sector_t,
3589 * and whether the max offset is addressable by the page cache.
3590 */
5a9ae68a 3591 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3592 ext4_blocks_count(es));
5a9ae68a 3593 if (err) {
b31e1552 3594 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3595 " too large to mount safely on this system");
ac27a0ec 3596 if (sizeof(sector_t) < 8)
90c699a9 3597 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3598 goto failed_mount;
3599 }
3600
617ba13b
MC		 3601 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3602 goto cantfind_ext4;
e7c95593 3603
0f2ddca6
FTN		 3604 /* check blocks count against device size */
3605 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3606 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3607 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3608 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3609 ext4_blocks_count(es), blocks_count);
3610 goto failed_mount;
3611 }
3612
0b8e58a1
AD		 3613 /*
3614 * It makes no sense for the first data block to be beyond the end
3615 * of the filesystem.
3616 */
3617 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3618 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3619 "block %u is beyond end of filesystem (%llu)",
3620 le32_to_cpu(es->s_first_data_block),
3621 ext4_blocks_count(es));
e7c95593
ES		 3622 goto failed_mount;
3623 }
bd81d8ee
LV		 3624 blocks_count = (ext4_blocks_count(es) -
3625 le32_to_cpu(es->s_first_data_block) +
3626 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3627 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3628 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3629 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3630 "(block count %llu, first data block %u, "
b31e1552 3631 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3632 ext4_blocks_count(es),
3633 le32_to_cpu(es->s_first_data_block),
3634 EXT4_BLOCKS_PER_GROUP(sb));
3635 goto failed_mount;
3636 }
bd81d8ee 3637 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3638 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3639 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3640 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3641 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3642 sbi->s_group_desc = ext4_kvmalloc(db_count *
3643 sizeof(struct buffer_head *),
3644 GFP_KERNEL);
ac27a0ec 3645 if (sbi->s_group_desc == NULL) {
b31e1552 3646 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3647 ret = -ENOMEM;
ac27a0ec
DK		 3648 goto failed_mount;
3649 }
3650
705895b6 3651 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3652
3653 for (i = 0; i < db_count; i++) {
70bbb3e0 3654 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3655 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3656 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3657 ext4_msg(sb, KERN_ERR,
3658 "can't read group descriptor %d", i);
ac27a0ec
DK		 3659 db_count = i;
3660 goto failed_mount2;
3661 }
3662 }
bfff6873 3663 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3664 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 3665 ret = -EFSCORRUPTED;
f9ae9cf5 3666 goto failed_mount2;
ac27a0ec 3667 }
772cb7c8 3668
f9ae9cf5 3669 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3670 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3671 spin_lock_init(&sbi->s_next_gen_lock);
3672
04ecddb7
JM		 3673 setup_timer(&sbi->s_err_report, print_daily_error_info,
3674 (unsigned long) sb);
04496411 3675
a75ae78f 3676 /* Register extent status tree shrinker */
eb68d0e2 3677 if (ext4_es_register_shrinker(sbi))
ce7e010a 3678 goto failed_mount3;
ce7e010a 3679
c9de560d 3680 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3681 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3682
f9ae9cf5
TT		 3683 /*
3684 * set up enough so that it can read an inode
3685 */
f6e63f90 3686 sb->s_op = &ext4_sops;
617ba13b
MC		 3687 sb->s_export_op = &ext4_export_ops;
3688 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3689#ifdef CONFIG_QUOTA
617ba13b 3690 sb->dq_op = &ext4_quota_operations;
e2b911c5 3691 if (ext4_has_feature_quota(sb))
1fa5efe3 3692 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3693 else
3694 sb->s_qcop = &ext4_qctl_operations;
689c958c 3695 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 3696#endif
f2fa2ffc
AK		 3697 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3698
ac27a0ec 3699 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3700 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3701
3702 sb->s_root = NULL;
3703
3704 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 3705 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 3706
e2b911c5 3707 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
c5e06d10 3708 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3709 goto failed_mount3a;
c5e06d10 3710
ac27a0ec
DK		 3711 /*
3712 * The first inode we look at is the journal inode. Don't try
3713 * root first: it may be modified in the journal!
3714 */
e2b911c5 3715 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
617ba13b 3716 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3717 goto failed_mount3a;
0390131b 3718 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
e2b911c5 3719 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 3720 ext4_msg(sb, KERN_ERR, "required journal recovery "
3721 "suppressed and not mounted read-only");
744692dc 3722 goto failed_mount_wq;
ac27a0ec 3723 } else {
1e381f60
DM		 3724 /* Nojournal mode, all journal mount options are illegal */
3725 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3726 ext4_msg(sb, KERN_ERR, "can't mount with "
3727 "journal_checksum, fs mounted w/o journal");
3728 goto failed_mount_wq;
3729 }
3730 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3731 ext4_msg(sb, KERN_ERR, "can't mount with "
3732 "journal_async_commit, fs mounted w/o journal");
3733 goto failed_mount_wq;
3734 }
3735 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3736 ext4_msg(sb, KERN_ERR, "can't mount with "
3737 "commit=%lu, fs mounted w/o journal",
3738 sbi->s_commit_interval / HZ);
3739 goto failed_mount_wq;
3740 }
3741 if (EXT4_MOUNT_DATA_FLAGS &
3742 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3743 ext4_msg(sb, KERN_ERR, "can't mount with "
3744 "data=, fs mounted w/o journal");
3745 goto failed_mount_wq;
3746 }
3747 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3748 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 3749 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3750 sbi->s_journal = NULL;
3751 needs_recovery = 0;
3752 goto no_journal;
ac27a0ec
DK		 3753 }
3754
e2b911c5 3755 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 3756 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3757 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3758 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3759 goto failed_mount_wq;
eb40a09c
JS		 3760 }
3761
25ed6e8a
DW		 3762 if (!set_journal_csum_feature_set(sb)) {
3763 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3764 "feature set");
3765 goto failed_mount_wq;
d4da6c9c 3766 }
818d276c 3767
ac27a0ec
DK		 3768 /* We have now updated the journal if required, so we can
3769 * validate the data journaling mode. */
3770 switch (test_opt(sb, DATA_FLAGS)) {
3771 case 0:
3772 /* No mode set, assume a default based on the journal
63f57933
AM		 3773 * capabilities: ORDERED_DATA if the journal can
3774 * cope, else JOURNAL_DATA
3775 */
dab291af
MC		 3776 if (jbd2_journal_check_available_features
3777 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3778 set_opt(sb, ORDERED_DATA);
ac27a0ec 3779 else
fd8c37ec 3780 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3781 break;
3782
617ba13b
MC		 3783 case EXT4_MOUNT_ORDERED_DATA:
3784 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3785 if (!jbd2_journal_check_available_features
3786 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3787 ext4_msg(sb, KERN_ERR, "Journal does not support "
3788 "requested data journaling mode");
744692dc 3789 goto failed_mount_wq;
ac27a0ec
DK		 3790 }
3791 default:
3792 break;
3793 }
b3881f74 3794 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3795
18aadd47
BJ		 3796 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3797
ce7e010a 3798no_journal:
2335d05f
AG		 3799 sbi->s_mb_cache = ext4_xattr_create_cache();
3800 if (!sbi->s_mb_cache) {
3801 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3802 goto failed_mount_wq;
9c191f70
M		 3803 }
3804
e2b911c5 3805 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 3806 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 3807 ext4_msg(sb, KERN_ERR,
3808 "Unsupported blocksize for fs encryption");
3809 goto failed_mount_wq;
3810 }
3811
e2b911c5
DW		 3812 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3813 !ext4_has_feature_encrypt(sb)) {
3814 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 3815 ext4_commit_super(sb, 1);
3816 }
3817
952fc18e
TT		 3818 /*
3819 * Get the # of file system overhead blocks from the
3820 * superblock if present.
3821 */
3822 if (es->s_overhead_clusters)
3823 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3824 else {
07aa2ea1
LC		 3825 err = ext4_calculate_overhead(sb);
3826 if (err)
952fc18e
TT		 3827 goto failed_mount_wq;
3828 }
3829
fd89d5f2
TH		 3830 /*
3831 * The maximum number of concurrent works can be high and
3832 * concurrency isn't really necessary. Limit it to 1.
3833 */
2e8fa54e
JK		 3834 EXT4_SB(sb)->rsv_conversion_wq =
3835 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3836 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3837 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 3838 ret = -ENOMEM;
2e8fa54e
JK		 3839 goto failed_mount4;
3840 }
3841
ac27a0ec 3842 /*
dab291af 3843 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3844 * so we can safely mount the rest of the filesystem now.
3845 */
3846
1d1fe1ee
DH		 3847 root = ext4_iget(sb, EXT4_ROOT_INO);
3848 if (IS_ERR(root)) {
b31e1552 3849 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3850 ret = PTR_ERR(root);
32a9bb57 3851 root = NULL;
ac27a0ec
DK		 3852 goto failed_mount4;
3853 }
3854 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3855 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3856 iput(root);
ac27a0ec
DK		 3857 goto failed_mount4;
3858 }
48fde701 3859 sb->s_root = d_make_root(root);
1d1fe1ee 3860 if (!sb->s_root) {
b31e1552 3861 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3862 ret = -ENOMEM;
3863 goto failed_mount4;
3864 }
ac27a0ec 3865
7e84b621
ES		 3866 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3867 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3868
3869 /* determine the minimum size of new large inodes, if present */
3870 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3871 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3872 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 3873 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 3874 if (sbi->s_want_extra_isize <
3875 le16_to_cpu(es->s_want_extra_isize))
3876 sbi->s_want_extra_isize =
3877 le16_to_cpu(es->s_want_extra_isize);
3878 if (sbi->s_want_extra_isize <
3879 le16_to_cpu(es->s_min_extra_isize))
3880 sbi->s_want_extra_isize =
3881 le16_to_cpu(es->s_min_extra_isize);
3882 }
3883 }
3884 /* Check if enough inode space is available */
3885 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3886 sbi->s_inode_size) {
3887 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3888 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3889 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3890 "available");
ef7f3835
KS		 3891 }
3892
b5799018 3893 ext4_set_resv_clusters(sb);
27dd4385 3894
6fd058f7
TT		 3895 err = ext4_setup_system_zone(sb);
3896 if (err) {
b31e1552 3897 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3898 "zone (%d)", err);
f9ae9cf5
TT		 3899 goto failed_mount4a;
3900 }
3901
3902 ext4_ext_init(sb);
3903 err = ext4_mb_init(sb);
3904 if (err) {
3905 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3906 err);
dcf2d804 3907 goto failed_mount5;
c2774d84
AK		 3908 }
3909
d5e03cbb
TT		 3910 block = ext4_count_free_clusters(sb);
3911 ext4_free_blocks_count_set(sbi->s_es,
3912 EXT4_C2B(sbi, block));
908c7f19
TH		 3913 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3914 GFP_KERNEL);
d5e03cbb
TT		 3915 if (!err) {
3916 unsigned long freei = ext4_count_free_inodes(sb);
3917 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 3918 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3919 GFP_KERNEL);
d5e03cbb
TT		 3920 }
3921 if (!err)
3922 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 3923 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 3924 if (!err)
908c7f19
TH		 3925 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3926 GFP_KERNEL);
c8585c6f
DJ		 3927 if (!err)
3928 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
3929
d5e03cbb
TT		 3930 if (err) {
3931 ext4_msg(sb, KERN_ERR, "insufficient memory");
3932 goto failed_mount6;
3933 }
3934
e2b911c5 3935 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 3936 if (!ext4_fill_flex_info(sb)) {
3937 ext4_msg(sb, KERN_ERR,
3938 "unable to initialize "
3939 "flex_bg meta info!");
3940 goto failed_mount6;
3941 }
3942
bfff6873
LC		 3943 err = ext4_register_li_request(sb, first_not_zeroed);
3944 if (err)
dcf2d804 3945 goto failed_mount6;
bfff6873 3946
b5799018 3947 err = ext4_register_sysfs(sb);
dcf2d804
TM		 3948 if (err)
3949 goto failed_mount7;
3197ebdb 3950
9b2ff357
JK		 3951#ifdef CONFIG_QUOTA
3952 /* Enable quota usage during mount. */
e2b911c5 3953 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
9b2ff357
JK		 3954 err = ext4_enable_quotas(sb);
3955 if (err)
3956 goto failed_mount8;
3957 }
3958#endif /* CONFIG_QUOTA */
3959
617ba13b
MC		 3960 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3961 ext4_orphan_cleanup(sb, es);
3962 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3963 if (needs_recovery) {
b31e1552 3964 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3965 ext4_mark_recovery_complete(sb, es);
3966 }
3967 if (EXT4_SB(sb)->s_journal) {
3968 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3969 descr = " journalled data mode";
3970 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3971 descr = " ordered data mode";
3972 else
3973 descr = " writeback data mode";
3974 } else
3975 descr = "out journal";
3976
79add3a3
LC		 3977 if (test_opt(sb, DISCARD)) {
3978 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3979 if (!blk_queue_discard(q))
3980 ext4_msg(sb, KERN_WARNING,
3981 "mounting with \"discard\" option, but "
3982 "the device does not support discard");
3983 }
3984
e294a537
TT		 3985 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3986 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3987 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3988 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3989
66e61a9e
TT		 3990 if (es->s_error_count)
3991 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3992
efbed4dc
TT		 3993 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3994 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
3995 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
3996 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
3997
d4c402d9 3998 kfree(orig_data);
ac27a0ec
DK		 3999 return 0;
4000
617ba13b 4001cantfind_ext4:
ac27a0ec 4002 if (!silent)
b31e1552 4003 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4004 goto failed_mount;
4005
72ba7450
TT		 4006#ifdef CONFIG_QUOTA
4007failed_mount8:
ebd173be 4008 ext4_unregister_sysfs(sb);
72ba7450 4009#endif
dcf2d804
TM		 4010failed_mount7:
4011 ext4_unregister_li_request(sb);
4012failed_mount6:
f9ae9cf5 4013 ext4_mb_release(sb);
d5e03cbb 4014 if (sbi->s_flex_groups)
b93b41d4 4015 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4016 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4017 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4018 percpu_counter_destroy(&sbi->s_dirs_counter);
4019 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4020failed_mount5:
f9ae9cf5
TT		 4021 ext4_ext_release(sb);
4022 ext4_release_system_zone(sb);
4023failed_mount4a:
94bf608a 4024 dput(sb->s_root);
32a9bb57 4025 sb->s_root = NULL;
94bf608a 4026failed_mount4:
b31e1552 4027 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4028 if (EXT4_SB(sb)->rsv_conversion_wq)
4029 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4030failed_mount_wq:
82939d79
JK		 4031 if (sbi->s_mb_cache) {
4032 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4033 sbi->s_mb_cache = NULL;
4034 }
0390131b
FM		 4035 if (sbi->s_journal) {
4036 jbd2_journal_destroy(sbi->s_journal);
4037 sbi->s_journal = NULL;
4038 }
50460fe8 4039failed_mount3a:
d3922a77 4040 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4041failed_mount3:
9105bb14 4042 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4043 if (sbi->s_mmp_tsk)
4044 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4045failed_mount2:
4046 for (i = 0; i < db_count; i++)
4047 brelse(sbi->s_group_desc[i]);
b93b41d4 4048 kvfree(sbi->s_group_desc);
ac27a0ec 4049failed_mount:
0441984a
DW		 4050 if (sbi->s_chksum_driver)
4051 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4052#ifdef CONFIG_QUOTA
a2d4a646 4053 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4054 kfree(sbi->s_qf_names[i]);
4055#endif
617ba13b 4056 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4057 brelse(bh);
4058out_fail:
4059 sb->s_fs_info = NULL;
f6830165 4060 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4061 kfree(sbi);
dcc7dae3 4062out_free_orig:
d4c402d9 4063 kfree(orig_data);
07aa2ea1 4064 return err ? err : ret;
ac27a0ec
DK		 4065}
4066
4067/*
4068 * Setup any per-fs journal parameters now. We'll do this both on
4069 * initial mount, once the journal has been initialised but before we've
4070 * done any recovery; and again on any subsequent remount.
4071 */
617ba13b 4072static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4073{
617ba13b 4074 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4075
30773840
TT		 4076 journal->j_commit_interval = sbi->s_commit_interval;
4077 journal->j_min_batch_time = sbi->s_min_batch_time;
4078 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4079
a931da6a 4080 write_lock(&journal->j_state_lock);
ac27a0ec 4081 if (test_opt(sb, BARRIER))
dab291af 4082 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4083 else
dab291af 4084 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4085 if (test_opt(sb, DATA_ERR_ABORT))
4086 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4087 else
4088 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4089 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4090}
4091
617ba13b 4092static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4093 unsigned int journal_inum)
4094{
4095 struct inode *journal_inode;
4096 journal_t *journal;
4097
e2b911c5 4098 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4099
ac27a0ec
DK		 4100 /* First, test for the existence of a valid inode on disk. Bad
4101 * things happen if we iget() an unused inode, as the subsequent
4102 * iput() will try to delete it. */
4103
1d1fe1ee
DH		 4104 journal_inode = ext4_iget(sb, journal_inum);
4105 if (IS_ERR(journal_inode)) {
b31e1552 4106 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4107 return NULL;
4108 }
4109 if (!journal_inode->i_nlink) {
4110 make_bad_inode(journal_inode);
4111 iput(journal_inode);
b31e1552 4112 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4113 return NULL;
4114 }
4115
e5f8eab8 4116 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4117 journal_inode, journal_inode->i_size);
1d1fe1ee 4118 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4119 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4120 iput(journal_inode);
4121 return NULL;
4122 }
4123
dab291af 4124 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4125 if (!journal) {
b31e1552 4126 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4127 iput(journal_inode);
4128 return NULL;
4129 }
4130 journal->j_private = sb;
617ba13b 4131 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4132 return journal;
4133}
4134
617ba13b 4135static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4136 dev_t j_dev)
4137{
2b2d6d01 4138 struct buffer_head *bh;
ac27a0ec 4139 journal_t *journal;
617ba13b
MC		 4140 ext4_fsblk_t start;
4141 ext4_fsblk_t len;
ac27a0ec 4142 int hblock, blocksize;
617ba13b 4143 ext4_fsblk_t sb_block;
ac27a0ec 4144 unsigned long offset;
2b2d6d01 4145 struct ext4_super_block *es;
ac27a0ec
DK		 4146 struct block_device *bdev;
4147
e2b911c5 4148 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4149
b31e1552 4150 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4151 if (bdev == NULL)
4152 return NULL;
4153
ac27a0ec 4154 blocksize = sb->s_blocksize;
e1defc4f 4155 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4156 if (blocksize < hblock) {
b31e1552
ES		 4157 ext4_msg(sb, KERN_ERR,
4158 "blocksize too small for journal device");
ac27a0ec
DK		 4159 goto out_bdev;
4160 }
4161
617ba13b
MC		 4162 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4163 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4164 set_blocksize(bdev, blocksize);
4165 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4166 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4167 "external journal");
ac27a0ec
DK		 4168 goto out_bdev;
4169 }
4170
2716b802 4171 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4172 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4173 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4174 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4175 ext4_msg(sb, KERN_ERR, "external journal has "
4176 "bad superblock");
ac27a0ec
DK		 4177 brelse(bh);
4178 goto out_bdev;
4179 }
4180
df4763be
DW		 4181 if ((le32_to_cpu(es->s_feature_ro_compat) &
4182 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4183 es->s_checksum != ext4_superblock_csum(sb, es)) {
4184 ext4_msg(sb, KERN_ERR, "external journal has "
4185 "corrupt superblock");
4186 brelse(bh);
4187 goto out_bdev;
4188 }
4189
617ba13b 4190 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4191 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4192 brelse(bh);
4193 goto out_bdev;
4194 }
4195
bd81d8ee 4196 len = ext4_blocks_count(es);
ac27a0ec
DK		 4197 start = sb_block + 1;
4198 brelse(bh); /* we're done with the superblock */
4199
dab291af 4200 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4201 start, len, blocksize);
4202 if (!journal) {
b31e1552 4203 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4204 goto out_bdev;
4205 }
4206 journal->j_private = sb;
9f203507 4207 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4208 wait_on_buffer(journal->j_sb_buffer);
4209 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4210 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4211 goto out_journal;
4212 }
4213 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4214 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4215 "user (unsupported) - %d",
ac27a0ec
DK		 4216 be32_to_cpu(journal->j_superblock->s_nr_users));
4217 goto out_journal;
4218 }
617ba13b
MC		 4219 EXT4_SB(sb)->journal_bdev = bdev;
4220 ext4_init_journal_params(sb, journal);
ac27a0ec 4221 return journal;
0b8e58a1 4222
ac27a0ec 4223out_journal:
dab291af 4224 jbd2_journal_destroy(journal);
ac27a0ec 4225out_bdev:
617ba13b 4226 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4227 return NULL;
4228}
4229
617ba13b
MC		 4230static int ext4_load_journal(struct super_block *sb,
4231 struct ext4_super_block *es,
ac27a0ec
DK		 4232 unsigned long journal_devnum)
4233{
4234 journal_t *journal;
4235 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4236 dev_t journal_dev;
4237 int err = 0;
4238 int really_read_only;
4239
e2b911c5 4240 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4241
ac27a0ec
DK		 4242 if (journal_devnum &&
4243 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4244 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4245 "numbers have changed");
ac27a0ec
DK		 4246 journal_dev = new_decode_dev(journal_devnum);
4247 } else
4248 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4249
4250 really_read_only = bdev_read_only(sb->s_bdev);
4251
4252 /*
4253 * Are we loading a blank journal or performing recovery after a
4254 * crash? For recovery, we need to check in advance whether we
4255 * can get read-write access to the device.
4256 */
e2b911c5 4257 if (ext4_has_feature_journal_needs_recovery(sb)) {
ac27a0ec 4258 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4259 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4260 "required on readonly filesystem");
ac27a0ec 4261 if (really_read_only) {
b31e1552
ES		 4262 ext4_msg(sb, KERN_ERR, "write access "
4263 "unavailable, cannot proceed");
ac27a0ec
DK		 4264 return -EROFS;
4265 }
b31e1552
ES		 4266 ext4_msg(sb, KERN_INFO, "write access will "
4267 "be enabled during recovery");
ac27a0ec
DK		 4268 }
4269 }
4270
4271 if (journal_inum && journal_dev) {
b31e1552
ES		 4272 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4273 "and inode journals!");
ac27a0ec
DK		 4274 return -EINVAL;
4275 }
4276
4277 if (journal_inum) {
617ba13b 4278 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4279 return -EINVAL;
4280 } else {
617ba13b 4281 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4282 return -EINVAL;
4283 }
4284
90576c0b 4285 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4286 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4287
e2b911c5 4288 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4289 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4290 if (!err) {
4291 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4292 if (save)
4293 memcpy(save, ((char *) es) +
4294 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4295 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4296 if (save)
4297 memcpy(((char *) es) + EXT4_S_ERR_START,
4298 save, EXT4_S_ERR_LEN);
4299 kfree(save);
4300 }
ac27a0ec
DK		 4301
4302 if (err) {
b31e1552 4303 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4304 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4305 return err;
4306 }
4307
617ba13b
MC		 4308 EXT4_SB(sb)->s_journal = journal;
4309 ext4_clear_journal_err(sb, es);
ac27a0ec 4310
c41303ce 4311 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4312 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4313 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4314
4315 /* Make sure we flush the recovery flag to disk. */
e2d67052 4316 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4317 }
4318
4319 return 0;
4320}
4321
e2d67052 4322static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4323{
e2d67052 4324 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4325 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4326 int error = 0;
ac27a0ec 4327
bdfe0cbd 4328 if (!sbh || block_device_ejected(sb))
c4be0c1d 4329 return error;
914258bf
TT		 4330 if (buffer_write_io_error(sbh)) {
4331 /*
4332 * Oh, dear. A previous attempt to write the
4333 * superblock failed. This could happen because the
4334 * USB device was yanked out. Or it could happen to
4335 * be a transient write error and maybe the block will
4336 * be remapped. Nothing we can do but to retry the
4337 * write and hope for the best.
4338 */
b31e1552
ES		 4339 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4340 "superblock detected");
914258bf
TT		 4341 clear_buffer_write_io_error(sbh);
4342 set_buffer_uptodate(sbh);
4343 }
71290b36
TT		 4344 /*
4345 * If the file system is mounted read-only, don't update the
4346 * superblock write time. This avoids updating the superblock
4347 * write time when we are mounting the root file system
4348 * read/only but we need to replay the journal; at that point,
4349 * for people who are east of GMT and who make their clock
4350 * tick in localtime for Windows bug-for-bug compatibility,
4351 * the clock is set in the future, and this will cause e2fsck
4352 * to complain and force a full file system check.
4353 */
4354 if (!(sb->s_flags & MS_RDONLY))
4355 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4356 if (sb->s_bdev->bd_part)
4357 es->s_kbytes_written =
4358 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4359 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4360 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4361 else
4362 es->s_kbytes_written =
4363 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4364 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4365 ext4_free_blocks_count_set(es,
57042651
TT		 4366 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4367 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4368 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4369 es->s_free_inodes_count =
4370 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4371 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4372 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4373 ext4_superblock_csum_set(sb);
ac27a0ec 4374 mark_buffer_dirty(sbh);
914258bf 4375 if (sync) {
564bc402
DJ		 4376 error = __sync_dirty_buffer(sbh,
4377 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
c4be0c1d
TS		 4378 if (error)
4379 return error;
4380
4381 error = buffer_write_io_error(sbh);
4382 if (error) {
b31e1552
ES		 4383 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4384 "superblock");
914258bf
TT		 4385 clear_buffer_write_io_error(sbh);
4386 set_buffer_uptodate(sbh);
4387 }
4388 }
c4be0c1d 4389 return error;
ac27a0ec
DK		 4390}
4391
ac27a0ec
DK		 4392/*
4393 * Have we just finished recovery? If so, and if we are mounting (or
4394 * remounting) the filesystem readonly, then we will end up with a
4395 * consistent fs on disk. Record that fact.
4396 */
2b2d6d01
TT		 4397static void ext4_mark_recovery_complete(struct super_block *sb,
4398 struct ext4_super_block *es)
ac27a0ec 4399{
617ba13b 4400 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4401
e2b911c5 4402 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4403 BUG_ON(journal != NULL);
4404 return;
4405 }
dab291af 4406 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4407 if (jbd2_journal_flush(journal) < 0)
4408 goto out;
4409
e2b911c5 4410 if (ext4_has_feature_journal_needs_recovery(sb) &&
ac27a0ec 4411 sb->s_flags & MS_RDONLY) {
e2b911c5 4412 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4413 ext4_commit_super(sb, 1);
ac27a0ec 4414 }
7ffe1ea8
HK		 4415
4416out:
dab291af 4417 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4418}
4419
4420/*
4421 * If we are mounting (or read-write remounting) a filesystem whose journal
4422 * has recorded an error from a previous lifetime, move that error to the
4423 * main filesystem now.
4424 */
2b2d6d01
TT		 4425static void ext4_clear_journal_err(struct super_block *sb,
4426 struct ext4_super_block *es)
ac27a0ec
DK		 4427{
4428 journal_t *journal;
4429 int j_errno;
4430 const char *errstr;
4431
e2b911c5 4432 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4433
617ba13b 4434 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4435
4436 /*
4437 * Now check for any error status which may have been recorded in the
617ba13b 4438 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4439 */
4440
dab291af 4441 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4442 if (j_errno) {
4443 char nbuf[16];
4444
617ba13b 4445 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4446 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4447 "from previous mount: %s", errstr);
12062ddd 4448 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4449
617ba13b
MC		 4450 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4451 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4452 ext4_commit_super(sb, 1);
ac27a0ec 4453
dab291af 4454 jbd2_journal_clear_err(journal);
d796c52e 4455 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4456 }
4457}
4458
4459/*
4460 * Force the running and committing transactions to commit,
4461 * and wait on the commit.
4462 */
617ba13b 4463int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4464{
4465 journal_t *journal;
ac27a0ec
DK		 4466
4467 if (sb->s_flags & MS_RDONLY)
4468 return 0;
4469
617ba13b 4470 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4471 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4472}
4473
617ba13b 4474static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4475{
14ce0cb4 4476 int ret = 0;
9eddacf9 4477 tid_t target;
06a407f1 4478 bool needs_barrier = false;
8d5d02e6 4479 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4480
9bffad1e 4481 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4482 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4483 /*
4484 * Writeback quota in non-journalled quota case - journalled quota has
4485 * no dirty dquots
4486 */
4487 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4488 /*
4489 * Data writeback is possible w/o journal transaction, so barrier must
4490 * being sent at the end of the function. But we can skip it if
4491 * transaction_commit will do it for us.
4492 */
bda32530
TT		 4493 if (sbi->s_journal) {
4494 target = jbd2_get_latest_transaction(sbi->s_journal);
4495 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4496 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4497 needs_barrier = true;
4498
4499 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4500 if (wait)
4501 ret = jbd2_log_wait_commit(sbi->s_journal,
4502 target);
4503 }
4504 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4505 needs_barrier = true;
06a407f1
DM		 4506 if (needs_barrier) {
4507 int err;
4508 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4509 if (!ret)
4510 ret = err;
0390131b 4511 }
06a407f1
DM		 4512
4513 return ret;
4514}
4515
ac27a0ec
DK		 4516/*
4517 * LVM calls this function before a (read-only) snapshot is created. This
4518 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4519 *
4520 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4521 * state independently. It relies on upper layer to stop all data & metadata
4522 * modifications.
ac27a0ec 4523 */
c4be0c1d 4524static int ext4_freeze(struct super_block *sb)
ac27a0ec 4525{
c4be0c1d
TS		 4526 int error = 0;
4527 journal_t *journal;
ac27a0ec 4528
9ca92389
TT		 4529 if (sb->s_flags & MS_RDONLY)
4530 return 0;
ac27a0ec 4531
9ca92389 4532 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4533
bb044576
TT		 4534 if (journal) {
4535 /* Now we set up the journal barrier. */
4536 jbd2_journal_lock_updates(journal);
ac27a0ec 4537
bb044576
TT		 4538 /*
4539 * Don't clear the needs_recovery flag if we failed to
4540 * flush the journal.
4541 */
4542 error = jbd2_journal_flush(journal);
4543 if (error < 0)
4544 goto out;
c642dc9e
ES		 4545
4546 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4547 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4548 }
9ca92389 4549
9ca92389 4550 error = ext4_commit_super(sb, 1);
6b0310fb 4551out:
bb044576
TT		 4552 if (journal)
4553 /* we rely on upper layer to stop further updates */
4554 jbd2_journal_unlock_updates(journal);
6b0310fb 4555 return error;
ac27a0ec
DK		 4556}
4557
4558/*
4559 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4560 * flag here, even though the filesystem is not technically dirty yet.
4561 */
c4be0c1d 4562static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4563{
9ca92389
TT		 4564 if (sb->s_flags & MS_RDONLY)
4565 return 0;
4566
c642dc9e
ES		 4567 if (EXT4_SB(sb)->s_journal) {
4568 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4569 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4570 }
4571
9ca92389 4572 ext4_commit_super(sb, 1);
c4be0c1d 4573 return 0;
ac27a0ec
DK		 4574}
4575
673c6100
TT		 4576/*
4577 * Structure to save mount options for ext4_remount's benefit
4578 */
4579struct ext4_mount_options {
4580 unsigned long s_mount_opt;
a2595b8a 4581 unsigned long s_mount_opt2;
08cefc7a
EB		 4582 kuid_t s_resuid;
4583 kgid_t s_resgid;
673c6100
TT		 4584 unsigned long s_commit_interval;
4585 u32 s_min_batch_time, s_max_batch_time;
4586#ifdef CONFIG_QUOTA
4587 int s_jquota_fmt;
a2d4a646 4588 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4589#endif
4590};
4591
2b2d6d01 4592static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4593{
2b2d6d01 4594 struct ext4_super_block *es;
617ba13b 4595 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4596 unsigned long old_sb_flags;
617ba13b 4597 struct ext4_mount_options old_opts;
c79d967d 4598 int enable_quota = 0;
8a266467 4599 ext4_group_t g;
b3881f74 4600 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4601 int err = 0;
ac27a0ec 4602#ifdef CONFIG_QUOTA
03dafb5f 4603 int i, j;
ac27a0ec 4604#endif
d4c402d9 4605 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4606
4607 /* Store the original options */
4608 old_sb_flags = sb->s_flags;
4609 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4610 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4611 old_opts.s_resuid = sbi->s_resuid;
4612 old_opts.s_resgid = sbi->s_resgid;
4613 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4614 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4615 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4616#ifdef CONFIG_QUOTA
4617 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4618 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4619 if (sbi->s_qf_names[i]) {
4620 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4621 GFP_KERNEL);
4622 if (!old_opts.s_qf_names[i]) {
4623 for (j = 0; j < i; j++)
4624 kfree(old_opts.s_qf_names[j]);
3e36a163 4625 kfree(orig_data);
03dafb5f
CG		 4626 return -ENOMEM;
4627 }
4628 } else
4629 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4630#endif
b3881f74
TT		 4631 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4632 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4633
661aa520 4634 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4635 err = -EINVAL;
4636 goto restore_opts;
4637 }
4638
6b992ff2 4639 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4640 test_opt(sb, JOURNAL_CHECKSUM)) {
4641 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4642 "during remount not supported; ignoring");
4643 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4644 }
4645
6ae6514b
PS		 4646 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4647 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4648 ext4_msg(sb, KERN_ERR, "can't mount with "
4649 "both data=journal and delalloc");
4650 err = -EINVAL;
4651 goto restore_opts;
4652 }
4653 if (test_opt(sb, DIOREAD_NOLOCK)) {
4654 ext4_msg(sb, KERN_ERR, "can't mount with "
4655 "both data=journal and dioread_nolock");
4656 err = -EINVAL;
4657 goto restore_opts;
4658 }
923ae0ff
RZ		 4659 if (test_opt(sb, DAX)) {
4660 ext4_msg(sb, KERN_ERR, "can't mount with "
4661 "both data=journal and dax");
4662 err = -EINVAL;
4663 goto restore_opts;
4664 }
4665 }
4666
4667 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4668 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4669 "dax flag with busy inodes while remounting");
4670 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4671 }
4672
4ab2f15b 4673 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4674 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4675
4676 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4677 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4678
4679 es = sbi->s_es;
4680
b3881f74 4681 if (sbi->s_journal) {
0390131b 4682 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4683 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4684 }
ac27a0ec 4685
a2fd66d0
TT		 4686 if (*flags & MS_LAZYTIME)
4687 sb->s_flags |= MS_LAZYTIME;
4688
661aa520 4689 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4690 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4691 err = -EROFS;
4692 goto restore_opts;
4693 }
4694
4695 if (*flags & MS_RDONLY) {
38c03b34
TT		 4696 err = sync_filesystem(sb);
4697 if (err < 0)
4698 goto restore_opts;
0f0dd62f
CH		 4699 err = dquot_suspend(sb, -1);
4700 if (err < 0)
c79d967d 4701 goto restore_opts;
c79d967d 4702
ac27a0ec
DK		 4703 /*
4704 * First of all, the unconditional stuff we have to do
4705 * to disable replay of the journal when we next remount
4706 */
4707 sb->s_flags |= MS_RDONLY;
4708
4709 /*
4710 * OK, test if we are remounting a valid rw partition
4711 * readonly, and if so set the rdonly flag and then
4712 * mark the partition as valid again.
4713 */
617ba13b
MC		 4714 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4715 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4716 es->s_state = cpu_to_le16(sbi->s_mount_state);
4717
a63c9eb2 4718 if (sbi->s_journal)
0390131b 4719 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4720 } else {
a13fb1a4 4721 /* Make sure we can mount this feature set readwrite */
e2b911c5 4722 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 4723 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4724 err = -EROFS;
4725 goto restore_opts;
4726 }
8a266467
TT		 4727 /*
4728 * Make sure the group descriptor checksums
0b8e58a1 4729 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4730 */
4731 for (g = 0; g < sbi->s_groups_count; g++) {
4732 struct ext4_group_desc *gdp =
4733 ext4_get_group_desc(sb, g, NULL);
4734
feb0ab32 4735 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4736 ext4_msg(sb, KERN_ERR,
4737 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 4738 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 4739 le16_to_cpu(gdp->bg_checksum));
6a797d27 4740 err = -EFSBADCRC;
8a266467
TT		 4741 goto restore_opts;
4742 }
4743 }
4744
ead6596b
ES		 4745 /*
4746 * If we have an unprocessed orphan list hanging
4747 * around from a previously readonly bdev mount,
4748 * require a full umount/remount for now.
4749 */
4750 if (es->s_last_orphan) {
b31e1552 4751 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4752 "remount RDWR because of unprocessed "
4753 "orphan inode list. Please "
b31e1552 4754 "umount/remount instead");
ead6596b
ES		 4755 err = -EINVAL;
4756 goto restore_opts;
4757 }
4758
ac27a0ec
DK		 4759 /*
4760 * Mounting a RDONLY partition read-write, so reread
4761 * and store the current valid flag. (It may have
4762 * been changed by e2fsck since we originally mounted
4763 * the partition.)
4764 */
0390131b
FM		 4765 if (sbi->s_journal)
4766 ext4_clear_journal_err(sb, es);
ac27a0ec 4767 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4768 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4769 sb->s_flags &= ~MS_RDONLY;
e2b911c5 4770 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 4771 if (ext4_multi_mount_protect(sb,
4772 le64_to_cpu(es->s_mmp_block))) {
4773 err = -EROFS;
4774 goto restore_opts;
4775 }
c79d967d 4776 enable_quota = 1;
ac27a0ec
DK		 4777 }
4778 }
bfff6873
LC		 4779
4780 /*
4781 * Reinitialize lazy itable initialization thread based on
4782 * current settings
4783 */
4784 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4785 ext4_unregister_li_request(sb);
4786 else {
4787 ext4_group_t first_not_zeroed;
4788 first_not_zeroed = ext4_has_uninit_itable(sb);
4789 ext4_register_li_request(sb, first_not_zeroed);
4790 }
4791
6fd058f7 4792 ext4_setup_system_zone(sb);
d096ad0f 4793 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4794 ext4_commit_super(sb, 1);
0390131b 4795
ac27a0ec
DK		 4796#ifdef CONFIG_QUOTA
4797 /* Release old quota file names */
a2d4a646 4798 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 4799 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4800 if (enable_quota) {
4801 if (sb_any_quota_suspended(sb))
4802 dquot_resume(sb, -1);
e2b911c5 4803 else if (ext4_has_feature_quota(sb)) {
7c319d32 4804 err = ext4_enable_quotas(sb);
07724f98 4805 if (err)
7c319d32 4806 goto restore_opts;
7c319d32
AK		 4807 }
4808 }
ac27a0ec 4809#endif
d4c402d9 4810
a26f4992 4811 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 4812 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4813 kfree(orig_data);
ac27a0ec 4814 return 0;
0b8e58a1 4815
ac27a0ec
DK		 4816restore_opts:
4817 sb->s_flags = old_sb_flags;
4818 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4819 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4820 sbi->s_resuid = old_opts.s_resuid;
4821 sbi->s_resgid = old_opts.s_resgid;
4822 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4823 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4824 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4825#ifdef CONFIG_QUOTA
4826 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 4827 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 4828 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4829 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4830 }
4831#endif
d4c402d9 4832 kfree(orig_data);
ac27a0ec
DK		 4833 return err;
4834}
4835
689c958c
LX		 4836#ifdef CONFIG_QUOTA
4837static int ext4_statfs_project(struct super_block *sb,
4838 kprojid_t projid, struct kstatfs *buf)
4839{
4840 struct kqid qid;
4841 struct dquot *dquot;
4842 u64 limit;
4843 u64 curblock;
4844
4845 qid = make_kqid_projid(projid);
4846 dquot = dqget(sb, qid);
4847 if (IS_ERR(dquot))
4848 return PTR_ERR(dquot);
4849 spin_lock(&dq_data_lock);
4850
4851 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
4852 dquot->dq_dqb.dqb_bsoftlimit :
4853 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
4854 if (limit && buf->f_blocks > limit) {
4855 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
4856 buf->f_blocks = limit;
4857 buf->f_bfree = buf->f_bavail =
4858 (buf->f_blocks > curblock) ?
4859 (buf->f_blocks - curblock) : 0;
4860 }
4861
4862 limit = dquot->dq_dqb.dqb_isoftlimit ?
4863 dquot->dq_dqb.dqb_isoftlimit :
4864 dquot->dq_dqb.dqb_ihardlimit;
4865 if (limit && buf->f_files > limit) {
4866 buf->f_files = limit;
4867 buf->f_ffree =
4868 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
4869 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
4870 }
4871
4872 spin_unlock(&dq_data_lock);
4873 dqput(dquot);
4874 return 0;
4875}
4876#endif
4877
2b2d6d01 4878static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4879{
4880 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4881 struct ext4_sb_info *sbi = EXT4_SB(sb);
4882 struct ext4_super_block *es = sbi->s_es;
27dd4385 4883 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 4884 u64 fsid;
d02a9391 4885 s64 bfree;
27dd4385 4886 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 4887
952fc18e
TT		 4888 if (!test_opt(sb, MINIX_DF))
4889 overhead = sbi->s_overhead;
ac27a0ec 4890
617ba13b 4891 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4892 buf->f_bsize = sb->s_blocksize;
b72f78cb 4893 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4894 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4895 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4896 /* prevent underflow in case that few free space is available */
57042651 4897 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 4898 buf->f_bavail = buf->f_bfree -
4899 (ext4_r_blocks_count(es) + resv_blocks);
4900 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 4901 buf->f_bavail = 0;
4902 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4903 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4904 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4905 fsid = le64_to_cpup((void *)es->s_uuid) ^
4906 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4907 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4908 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4909
689c958c
LX		 4910#ifdef CONFIG_QUOTA
4911 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
4912 sb_has_quota_limits_enabled(sb, PRJQUOTA))
4913 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
4914#endif
ac27a0ec
DK		 4915 return 0;
4916}
4917
0b8e58a1
AD		 4918/* Helper function for writing quotas on sync - we need to start transaction
4919 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4920 * Process 1 Process 2
617ba13b 4921 * ext4_create() quota_sync()
a269eb18 4922 * jbd2_journal_start() write_dquot()
871a2931 4923 * dquot_initialize() down(dqio_mutex)
dab291af 4924 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4925 *
4926 */
4927
4928#ifdef CONFIG_QUOTA
4929
4930static inline struct inode *dquot_to_inode(struct dquot *dquot)
4931{
4c376dca 4932 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4933}
4934
617ba13b 4935static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4936{
4937 int ret, err;
4938 handle_t *handle;
4939 struct inode *inode;
4940
4941 inode = dquot_to_inode(dquot);
9924a92a 4942 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 4943 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4944 if (IS_ERR(handle))
4945 return PTR_ERR(handle);
4946 ret = dquot_commit(dquot);
617ba13b 4947 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4948 if (!ret)
4949 ret = err;
4950 return ret;
4951}
4952
617ba13b 4953static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4954{
4955 int ret, err;
4956 handle_t *handle;
4957
9924a92a 4958 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4959 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4960 if (IS_ERR(handle))
4961 return PTR_ERR(handle);
4962 ret = dquot_acquire(dquot);
617ba13b 4963 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4964 if (!ret)
4965 ret = err;
4966 return ret;
4967}
4968
617ba13b 4969static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4970{
4971 int ret, err;
4972 handle_t *handle;
4973
9924a92a 4974 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4975 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4976 if (IS_ERR(handle)) {
4977 /* Release dquot anyway to avoid endless cycle in dqput() */
4978 dquot_release(dquot);
ac27a0ec 4979 return PTR_ERR(handle);
9c3013e9 4980 }
ac27a0ec 4981 ret = dquot_release(dquot);
617ba13b 4982 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4983 if (!ret)
4984 ret = err;
4985 return ret;
4986}
4987
617ba13b 4988static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4989{
262b4662
JK		 4990 struct super_block *sb = dquot->dq_sb;
4991 struct ext4_sb_info *sbi = EXT4_SB(sb);
4992
2c8be6b2 4993 /* Are we journaling quotas? */
e2b911c5 4994 if (ext4_has_feature_quota(sb) ||
262b4662 4995 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 4996 dquot_mark_dquot_dirty(dquot);
617ba13b 4997 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4998 } else {
4999 return dquot_mark_dquot_dirty(dquot);
5000 }
5001}
5002
617ba13b 5003static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5004{
5005 int ret, err;
5006 handle_t *handle;
5007
5008 /* Data block + inode block */
2b0143b5 5009 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5010 if (IS_ERR(handle))
5011 return PTR_ERR(handle);
5012 ret = dquot_commit_info(sb, type);
617ba13b 5013 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5014 if (!ret)
5015 ret = err;
5016 return ret;
5017}
5018
5019/*
5020 * Turn on quotas during mount time - we need to find
5021 * the quota file and such...
5022 */
617ba13b 5023static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5024{
287a8095
CH		 5025 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5026 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5027}
5028
daf647d2
TT		 5029static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5030{
5031 struct ext4_inode_info *ei = EXT4_I(inode);
5032
5033 /* The first argument of lockdep_set_subclass has to be
5034 * *exactly* the same as the argument to init_rwsem() --- in
5035 * this case, in init_once() --- or lockdep gets unhappy
5036 * because the name of the lock is set using the
5037 * stringification of the argument to init_rwsem().
5038 */
5039 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5040 lockdep_set_subclass(&ei->i_data_sem, subclass);
5041}
5042
ac27a0ec
DK		 5043/*
5044 * Standard function to be called on quota_on
5045 */
617ba13b 5046static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5047 struct path *path)
ac27a0ec
DK		 5048{
5049 int err;
ac27a0ec
DK		 5050
5051 if (!test_opt(sb, QUOTA))
5052 return -EINVAL;
0623543b 5053
ac27a0ec 5054 /* Quotafile not on the same filesystem? */
d8c9584e 5055 if (path->dentry->d_sb != sb)
ac27a0ec 5056 return -EXDEV;
0623543b
JK		 5057 /* Journaling quota? */
5058 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5059 /* Quotafile not in fs root? */
f00c9e44 5060 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5061 ext4_msg(sb, KERN_WARNING,
5062 "Quota file not on filesystem root. "
5063 "Journaled quota will not work");
2b2d6d01 5064 }
0623543b
JK		 5065
5066 /*
5067 * When we journal data on quota file, we have to flush journal to see
5068 * all updates to the file when we bypass pagecache...
5069 */
0390131b 5070 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5071 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5072 /*
5073 * We don't need to lock updates but journal_flush() could
5074 * otherwise be livelocked...
5075 */
5076 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5077 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5078 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5079 if (err)
7ffe1ea8 5080 return err;
0623543b 5081 }
daf647d2
TT		 5082 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5083 err = dquot_quota_on(sb, type, format_id, path);
5084 if (err)
5085 lockdep_set_quota_inode(path->dentry->d_inode,
5086 I_DATA_SEM_NORMAL);
5087 return err;
ac27a0ec
DK		 5088}
5089
7c319d32
AK		 5090static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5091 unsigned int flags)
5092{
5093 int err;
5094 struct inode *qf_inode;
a2d4a646 5095 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5096 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5097 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5098 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5099 };
5100
e2b911c5 5101 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5102
5103 if (!qf_inums[type])
5104 return -EPERM;
5105
5106 qf_inode = ext4_iget(sb, qf_inums[type]);
5107 if (IS_ERR(qf_inode)) {
5108 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5109 return PTR_ERR(qf_inode);
5110 }
5111
bcb13850
JK		 5112 /* Don't account quota for quota files to avoid recursion */
5113 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5114 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5115 err = dquot_enable(qf_inode, type, format_id, flags);
5116 iput(qf_inode);
daf647d2
TT		 5117 if (err)
5118 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5119
5120 return err;
5121}
5122
5123/* Enable usage tracking for all quota types. */
5124static int ext4_enable_quotas(struct super_block *sb)
5125{
5126 int type, err = 0;
a2d4a646 5127 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5128 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5129 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5130 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5131 };
5132
5133 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5134 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5135 if (qf_inums[type]) {
5136 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5137 DQUOT_USAGE_ENABLED);
5138 if (err) {
5139 ext4_warning(sb,
72ba7450
TT		 5140 "Failed to enable quota tracking "
5141 "(type=%d, err=%d). Please run "
5142 "e2fsck to fix.", type, err);
7c319d32
AK		 5143 return err;
5144 }
5145 }
5146 }
5147 return 0;
5148}
5149
ca0e05e4
DM		 5150static int ext4_quota_off(struct super_block *sb, int type)
5151{
21f97697
JK		 5152 struct inode *inode = sb_dqopt(sb)->files[type];
5153 handle_t *handle;
5154
87009d86
DM		 5155 /* Force all delayed allocation blocks to be allocated.
5156 * Caller already holds s_umount sem */
5157 if (test_opt(sb, DELALLOC))
ca0e05e4 5158 sync_filesystem(sb);
ca0e05e4 5159
0b268590
AG		 5160 if (!inode)
5161 goto out;
5162
21f97697
JK		 5163 /* Update modification times of quota files when userspace can
5164 * start looking at them */
9924a92a 5165 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5166 if (IS_ERR(handle))
5167 goto out;
5168 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5169 ext4_mark_inode_dirty(handle, inode);
5170 ext4_journal_stop(handle);
5171
5172out:
ca0e05e4
DM		 5173 return dquot_quota_off(sb, type);
5174}
5175
ac27a0ec
DK		 5176/* Read data from quotafile - avoid pagecache and such because we cannot afford
5177 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5178 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5179 * we don't have to be afraid of races */
617ba13b 5180static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5181 size_t len, loff_t off)
5182{
5183 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5184 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5185 int offset = off & (sb->s_blocksize - 1);
5186 int tocopy;
5187 size_t toread;
5188 struct buffer_head *bh;
5189 loff_t i_size = i_size_read(inode);
5190
5191 if (off > i_size)
5192 return 0;
5193 if (off+len > i_size)
5194 len = i_size-off;
5195 toread = len;
5196 while (toread > 0) {
5197 tocopy = sb->s_blocksize - offset < toread ?
5198 sb->s_blocksize - offset : toread;
1c215028
TT		 5199 bh = ext4_bread(NULL, inode, blk, 0);
5200 if (IS_ERR(bh))
5201 return PTR_ERR(bh);
ac27a0ec
DK		 5202 if (!bh) /* A hole? */
5203 memset(data, 0, tocopy);
5204 else
5205 memcpy(data, bh->b_data+offset, tocopy);
5206 brelse(bh);
5207 offset = 0;
5208 toread -= tocopy;
5209 data += tocopy;
5210 blk++;
5211 }
5212 return len;
5213}
5214
5215/* Write to quotafile (we know the transaction is already started and has
5216 * enough credits) */
617ba13b 5217static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5218 const char *data, size_t len, loff_t off)
5219{
5220 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5221 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5222 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5223 int retries = 0;
ac27a0ec
DK		 5224 struct buffer_head *bh;
5225 handle_t *handle = journal_current_handle();
5226
0390131b 5227 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5228 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5229 " cancelled because transaction is not started",
9c3013e9
JK		 5230 (unsigned long long)off, (unsigned long long)len);
5231 return -EIO;
5232 }
67eeb568
DM		 5233 /*
5234 * Since we account only one data block in transaction credits,
5235 * then it is impossible to cross a block boundary.
5236 */
5237 if (sb->s_blocksize - offset < len) {
5238 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5239 " cancelled because not block aligned",
5240 (unsigned long long)off, (unsigned long long)len);
5241 return -EIO;
5242 }
5243
c5e298ae
TT		 5244 do {
5245 bh = ext4_bread(handle, inode, blk,
5246 EXT4_GET_BLOCKS_CREATE |
5247 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5248 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5249 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5250 if (IS_ERR(bh))
5251 return PTR_ERR(bh);
67eeb568
DM		 5252 if (!bh)
5253 goto out;
5d601255 5254 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5255 err = ext4_journal_get_write_access(handle, bh);
5256 if (err) {
5257 brelse(bh);
1c215028 5258 return err;
ac27a0ec 5259 }
67eeb568
DM		 5260 lock_buffer(bh);
5261 memcpy(bh->b_data+offset, data, len);
5262 flush_dcache_page(bh->b_page);
5263 unlock_buffer(bh);
62d2b5f2 5264 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5265 brelse(bh);
ac27a0ec 5266out:
67eeb568
DM		 5267 if (inode->i_size < off + len) {
5268 i_size_write(inode, off + len);
617ba13b 5269 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5270 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5271 }
67eeb568 5272 return len;
ac27a0ec
DK		 5273}
5274
8f0e8746
TT		 5275static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5276{
5277 const struct quota_format_ops *ops;
5278
5279 if (!sb_has_quota_loaded(sb, qid->type))
5280 return -ESRCH;
5281 ops = sb_dqopt(sb)->ops[qid->type];
5282 if (!ops || !ops->get_next_id)
5283 return -ENOSYS;
5284 return dquot_get_next_id(sb, qid);
5285}
ac27a0ec
DK		 5286#endif
5287
152a0836
AV		 5288static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5289 const char *dev_name, void *data)
ac27a0ec 5290{
152a0836 5291 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5292}
5293
c290ea01 5294#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5295static inline void register_as_ext2(void)
5296{
5297 int err = register_filesystem(&ext2_fs_type);
5298 if (err)
5299 printk(KERN_WARNING
5300 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5301}
5302
5303static inline void unregister_as_ext2(void)
5304{
5305 unregister_filesystem(&ext2_fs_type);
5306}
2035e776
TT		 5307
5308static inline int ext2_feature_set_ok(struct super_block *sb)
5309{
e2b911c5 5310 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776
TT		 5311 return 0;
5312 if (sb->s_flags & MS_RDONLY)
5313 return 1;
e2b911c5 5314 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5315 return 0;
5316 return 1;
5317}
24b58424
TT		 5318#else
5319static inline void register_as_ext2(void) { }
5320static inline void unregister_as_ext2(void) { }
2035e776 5321static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5322#endif
5323
24b58424
TT		 5324static inline void register_as_ext3(void)
5325{
5326 int err = register_filesystem(&ext3_fs_type);
5327 if (err)
5328 printk(KERN_WARNING
5329 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5330}
5331
5332static inline void unregister_as_ext3(void)
5333{
5334 unregister_filesystem(&ext3_fs_type);
5335}
2035e776
TT		 5336
5337static inline int ext3_feature_set_ok(struct super_block *sb)
5338{
e2b911c5 5339 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5340 return 0;
e2b911c5 5341 if (!ext4_has_feature_journal(sb))
2035e776
TT		 5342 return 0;
5343 if (sb->s_flags & MS_RDONLY)
5344 return 1;
e2b911c5 5345 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5346 return 0;
5347 return 1;
5348}
24b58424 5349
03010a33
TT		 5350static struct file_system_type ext4_fs_type = {
5351 .owner = THIS_MODULE,
5352 .name = "ext4",
152a0836 5353 .mount = ext4_mount,
03010a33
TT		 5354 .kill_sb = kill_block_super,
5355 .fs_flags = FS_REQUIRES_DEV,
5356};
7f78e035 5357MODULE_ALIAS_FS("ext4");
03010a33 5358
e9e3bcec
ES		 5359/* Shared across all ext4 file systems */
5360wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5361
5dabfc78 5362static int __init ext4_init_fs(void)
ac27a0ec 5363{
e9e3bcec 5364 int i, err;
c9de560d 5365
e294a537 5366 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5367 ext4_li_info = NULL;
5368 mutex_init(&ext4_li_mtx);
5369
9a4c8019 5370 /* Build-time check for flags consistency */
12e9b892 5371 ext4_check_flag_values();
e9e3bcec 5372
e142d052 5373 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5374 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5375
51865fda 5376 err = ext4_init_es();
6fd058f7
TT		 5377 if (err)
5378 return err;
51865fda
ZL		 5379
5380 err = ext4_init_pageio();
5381 if (err)
b5799018 5382 goto out5;
51865fda 5383
5dabfc78 5384 err = ext4_init_system_zone();
bd2d0210 5385 if (err)
b5799018 5386 goto out4;
857ac889 5387
b5799018 5388 err = ext4_init_sysfs();
dd68314c 5389 if (err)
b5799018 5390 goto out3;
857ac889 5391
5dabfc78 5392 err = ext4_init_mballoc();
c9de560d
AT		 5393 if (err)
5394 goto out2;
ac27a0ec
DK		 5395 err = init_inodecache();
5396 if (err)
5397 goto out1;
24b58424 5398 register_as_ext3();
2035e776 5399 register_as_ext2();
03010a33 5400 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5401 if (err)
5402 goto out;
bfff6873 5403
ac27a0ec
DK		 5404 return 0;
5405out:
24b58424
TT		 5406 unregister_as_ext2();
5407 unregister_as_ext3();
ac27a0ec
DK		 5408 destroy_inodecache();
5409out1:
5dabfc78 5410 ext4_exit_mballoc();
9c191f70 5411out2:
b5799018
TT		 5412 ext4_exit_sysfs();
5413out3:
5dabfc78 5414 ext4_exit_system_zone();
b5799018 5415out4:
5dabfc78 5416 ext4_exit_pageio();
b5799018 5417out5:
51865fda
ZL		 5418 ext4_exit_es();
5419
ac27a0ec
DK		 5420 return err;
5421}
5422
5dabfc78 5423static void __exit ext4_exit_fs(void)
ac27a0ec 5424{
e298e73b 5425 ext4_exit_crypto();
bfff6873 5426 ext4_destroy_lazyinit_thread();
24b58424
TT		 5427 unregister_as_ext2();
5428 unregister_as_ext3();
03010a33 5429 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5430 destroy_inodecache();
5dabfc78 5431 ext4_exit_mballoc();
b5799018 5432 ext4_exit_sysfs();
5dabfc78
TT		 5433 ext4_exit_system_zone();
5434 ext4_exit_pageio();
dd12ed14 5435 ext4_exit_es();
ac27a0ec
DK		 5436}
5437
5438MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5439MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5440MODULE_LICENSE("GPL");
5dabfc78
TT		 5441module_init(ext4_init_fs)
5442module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository