projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Calculate and verify checksums of extended attribute blocks
[deliverable/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
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>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
6f91bc5f 48#include "ext4_extents.h"
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 72static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 73 char nbuf[16]);
2b2d6d01
TT		 74static int ext4_remount(struct super_block *sb, int *flags, char *data);
75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 76static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 77static void ext4_write_super(struct super_block *sb);
c4be0c1d 78static int ext4_freeze(struct super_block *sb);
152a0836
AV		 79static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 const char *dev_name, void *data);
2035e776
TT		 81static inline int ext2_feature_set_ok(struct super_block *sb);
82static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 83static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 84static void ext4_destroy_lazyinit_thread(void);
85static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 86static void ext4_clear_request_list(void);
ac27a0ec 87
2035e776
TT		 88#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89static struct file_system_type ext2_fs_type = {
90 .owner = THIS_MODULE,
91 .name = "ext2",
92 .mount = ext4_mount,
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
95};
96#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97#else
98#define IS_EXT2_SB(sb) (0)
99#endif
100
101
ba69f9ab
JK		 102#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
152a0836 106 .mount = ext4_mount,
ba69f9ab
JK		 107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109};
110#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111#else
112#define IS_EXT3_SB(sb) (0)
113#endif
bd81d8ee 114
d25425f8
DW		 115static int ext4_verify_csum_type(struct super_block *sb,
116 struct ext4_super_block *es)
117{
118 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
119 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
120 return 1;
121
122 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
123}
124
a9c47317
DW		 125static __le32 ext4_superblock_csum(struct super_block *sb,
126 struct ext4_super_block *es)
127{
128 struct ext4_sb_info *sbi = EXT4_SB(sb);
129 int offset = offsetof(struct ext4_super_block, s_checksum);
130 __u32 csum;
131
132 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
133
134 return cpu_to_le32(csum);
135}
136
137int ext4_superblock_csum_verify(struct super_block *sb,
138 struct ext4_super_block *es)
139{
140 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
141 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
142 return 1;
143
144 return es->s_checksum == ext4_superblock_csum(sb, es);
145}
146
147void ext4_superblock_csum_set(struct super_block *sb,
148 struct ext4_super_block *es)
149{
150 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
151 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
152 return;
153
154 es->s_checksum = ext4_superblock_csum(sb, es);
155}
156
9933fc0a
TT		 157void *ext4_kvmalloc(size_t size, gfp_t flags)
158{
159 void *ret;
160
161 ret = kmalloc(size, flags);
162 if (!ret)
163 ret = __vmalloc(size, flags, PAGE_KERNEL);
164 return ret;
165}
166
167void *ext4_kvzalloc(size_t size, gfp_t flags)
168{
169 void *ret;
170
db9481c0 171 ret = kzalloc(size, flags);
9933fc0a
TT		 172 if (!ret)
173 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
174 return ret;
175}
176
177void ext4_kvfree(void *ptr)
178{
179 if (is_vmalloc_addr(ptr))
180 vfree(ptr);
181 else
182 kfree(ptr);
183
184}
185
8fadc143
AR		 186ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
bd81d8ee 188{
3a14589c 189 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 191 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 192}
193
8fadc143
AR		 194ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
195 struct ext4_group_desc *bg)
bd81d8ee 196{
5272f837 197 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 200}
201
8fadc143
AR		 202ext4_fsblk_t ext4_inode_table(struct super_block *sb,
203 struct ext4_group_desc *bg)
bd81d8ee 204{
5272f837 205 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 207 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 208}
209
021b65bb
TT		 210__u32 ext4_free_group_clusters(struct super_block *sb,
211 struct ext4_group_desc *bg)
560671a0
AK		 212{
213 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 215 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 216}
217
218__u32 ext4_free_inodes_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220{
221 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 223 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 224}
225
226__u32 ext4_used_dirs_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
228{
229 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 231 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 232}
233
234__u32 ext4_itable_unused_count(struct super_block *sb,
235 struct ext4_group_desc *bg)
236{
237 return le16_to_cpu(bg->bg_itable_unused_lo) |
238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 239 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 240}
241
8fadc143
AR		 242void ext4_block_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 244{
3a14589c 245 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 248}
249
8fadc143
AR		 250void ext4_inode_bitmap_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 252{
5272f837 253 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 256}
257
8fadc143
AR		 258void ext4_inode_table_set(struct super_block *sb,
259 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 260{
5272f837 261 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 264}
265
021b65bb
TT		 266void ext4_free_group_clusters_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 268{
269 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
272}
273
274void ext4_free_inodes_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276{
277 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
280}
281
282void ext4_used_dirs_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284{
285 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
288}
289
290void ext4_itable_unused_set(struct super_block *sb,
291 struct ext4_group_desc *bg, __u32 count)
292{
293 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
296}
297
d3d1faf6
CW		 298
299/* Just increment the non-pointer handle value */
300static handle_t *ext4_get_nojournal(void)
301{
302 handle_t *handle = current->journal_info;
303 unsigned long ref_cnt = (unsigned long)handle;
304
305 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
306
307 ref_cnt++;
308 handle = (handle_t *)ref_cnt;
309
310 current->journal_info = handle;
311 return handle;
312}
313
314
315/* Decrement the non-pointer handle value */
316static void ext4_put_nojournal(handle_t *handle)
317{
318 unsigned long ref_cnt = (unsigned long)handle;
319
320 BUG_ON(ref_cnt == 0);
321
322 ref_cnt--;
323 handle = (handle_t *)ref_cnt;
324
325 current->journal_info = handle;
326}
327
ac27a0ec 328/*
dab291af 329 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 330 *
331 * The only special thing we need to do here is to make sure that all
332 * journal_end calls result in the superblock being marked dirty, so
333 * that sync() will call the filesystem's write_super callback if
334 * appropriate.
be4f27d3
YY		 335 *
336 * To avoid j_barrier hold in userspace when a user calls freeze(),
337 * ext4 prevents a new handle from being started by s_frozen, which
338 * is in an upper layer.
ac27a0ec 339 */
617ba13b 340handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 341{
342 journal_t *journal;
be4f27d3 343 handle_t *handle;
ac27a0ec 344
12706394 345 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 346 if (sb->s_flags & MS_RDONLY)
347 return ERR_PTR(-EROFS);
348
617ba13b 349 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 350 handle = ext4_journal_current_handle();
351
352 /*
353 * If a handle has been started, it should be allowed to
354 * finish, otherwise deadlock could happen between freeze
355 * and others(e.g. truncate) due to the restart of the
356 * journal handle if the filesystem is forzen and active
357 * handles are not stopped.
358 */
359 if (!handle)
360 vfs_check_frozen(sb, SB_FREEZE_TRANS);
361
362 if (!journal)
363 return ext4_get_nojournal();
364 /*
365 * Special case here: if the journal has aborted behind our
366 * backs (eg. EIO in the commit thread), then we still need to
367 * take the FS itself readonly cleanly.
368 */
369 if (is_journal_aborted(journal)) {
370 ext4_abort(sb, "Detected aborted journal");
371 return ERR_PTR(-EROFS);
ac27a0ec 372 }
be4f27d3 373 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 374}
375
376/*
377 * The only special thing we need to do here is to make sure that all
dab291af 378 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 379 * that sync() will call the filesystem's write_super callback if
380 * appropriate.
381 */
c398eda0 382int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 383{
384 struct super_block *sb;
385 int err;
386 int rc;
387
0390131b 388 if (!ext4_handle_valid(handle)) {
d3d1faf6 389 ext4_put_nojournal(handle);
0390131b
FM		 390 return 0;
391 }
ac27a0ec
DK		 392 sb = handle->h_transaction->t_journal->j_private;
393 err = handle->h_err;
dab291af 394 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 395
396 if (!err)
397 err = rc;
398 if (err)
c398eda0 399 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 400 return err;
401}
402
90c7201b
TT		 403void ext4_journal_abort_handle(const char *caller, unsigned int line,
404 const char *err_fn, struct buffer_head *bh,
405 handle_t *handle, int err)
ac27a0ec
DK		 406{
407 char nbuf[16];
617ba13b 408 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 409
0390131b
FM		 410 BUG_ON(!ext4_handle_valid(handle));
411
ac27a0ec
DK		 412 if (bh)
413 BUFFER_TRACE(bh, "abort");
414
415 if (!handle->h_err)
416 handle->h_err = err;
417
418 if (is_handle_aborted(handle))
419 return;
420
92b97816 421 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
90c7201b 422 caller, line, errstr, err_fn);
ac27a0ec 423
dab291af 424 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 425}
426
1c13d5c0
TT		 427static void __save_error_info(struct super_block *sb, const char *func,
428 unsigned int line)
429{
430 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
431
432 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
433 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
434 es->s_last_error_time = cpu_to_le32(get_seconds());
435 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
436 es->s_last_error_line = cpu_to_le32(line);
437 if (!es->s_first_error_time) {
438 es->s_first_error_time = es->s_last_error_time;
439 strncpy(es->s_first_error_func, func,
440 sizeof(es->s_first_error_func));
441 es->s_first_error_line = cpu_to_le32(line);
442 es->s_first_error_ino = es->s_last_error_ino;
443 es->s_first_error_block = es->s_last_error_block;
444 }
66e61a9e
TT		 445 /*
446 * Start the daily error reporting function if it hasn't been
447 * started already
448 */
449 if (!es->s_error_count)
450 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 451 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
452}
453
454static void save_error_info(struct super_block *sb, const char *func,
455 unsigned int line)
456{
457 __save_error_info(sb, func, line);
458 ext4_commit_super(sb, 1);
459}
460
7c2e7087
TT		 461/*
462 * The del_gendisk() function uninitializes the disk-specific data
463 * structures, including the bdi structure, without telling anyone
464 * else. Once this happens, any attempt to call mark_buffer_dirty()
465 * (for example, by ext4_commit_super), will cause a kernel OOPS.
466 * This is a kludge to prevent these oops until we can put in a proper
467 * hook in del_gendisk() to inform the VFS and file system layers.
468 */
469static int block_device_ejected(struct super_block *sb)
470{
471 struct inode *bd_inode = sb->s_bdev->bd_inode;
472 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
473
474 return bdi->dev == NULL;
475}
476
18aadd47
BJ		 477static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
478{
479 struct super_block *sb = journal->j_private;
480 struct ext4_sb_info *sbi = EXT4_SB(sb);
481 int error = is_journal_aborted(journal);
482 struct ext4_journal_cb_entry *jce, *tmp;
483
484 spin_lock(&sbi->s_md_lock);
485 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
486 list_del_init(&jce->jce_list);
487 spin_unlock(&sbi->s_md_lock);
488 jce->jce_func(sb, jce, error);
489 spin_lock(&sbi->s_md_lock);
490 }
491 spin_unlock(&sbi->s_md_lock);
492}
1c13d5c0 493
ac27a0ec
DK		 494/* Deal with the reporting of failure conditions on a filesystem such as
495 * inconsistencies detected or read IO failures.
496 *
497 * On ext2, we can store the error state of the filesystem in the
617ba13b 498 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 499 * write ordering constraints on the superblock which prevent us from
500 * writing it out straight away; and given that the journal is about to
501 * be aborted, we can't rely on the current, or future, transactions to
502 * write out the superblock safely.
503 *
dab291af 504 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 505 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 506 * that error until we've noted it down and cleared it.
507 */
508
617ba13b 509static void ext4_handle_error(struct super_block *sb)
ac27a0ec 510{
ac27a0ec
DK		 511 if (sb->s_flags & MS_RDONLY)
512 return;
513
2b2d6d01 514 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 515 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 516
4ab2f15b 517 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 518 if (journal)
dab291af 519 jbd2_journal_abort(journal, -EIO);
ac27a0ec 520 }
2b2d6d01 521 if (test_opt(sb, ERRORS_RO)) {
b31e1552 522 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 523 sb->s_flags |= MS_RDONLY;
524 }
ac27a0ec 525 if (test_opt(sb, ERRORS_PANIC))
617ba13b 526 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 527 sb->s_id);
528}
529
12062ddd 530void __ext4_error(struct super_block *sb, const char *function,
c398eda0 531 unsigned int line, const char *fmt, ...)
ac27a0ec 532{
0ff2ea7d 533 struct va_format vaf;
ac27a0ec
DK		 534 va_list args;
535
536 va_start(args, fmt);
0ff2ea7d
JP		 537 vaf.fmt = fmt;
538 vaf.va = &args;
539 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
540 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec
DK		 541 va_end(args);
542
617ba13b 543 ext4_handle_error(sb);
ac27a0ec
DK		 544}
545
c398eda0
TT		 546void ext4_error_inode(struct inode *inode, const char *function,
547 unsigned int line, ext4_fsblk_t block,
273df556
FM		 548 const char *fmt, ...)
549{
550 va_list args;
f7c21177 551 struct va_format vaf;
1c13d5c0 552 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 553
1c13d5c0
TT		 554 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
555 es->s_last_error_block = cpu_to_le64(block);
556 save_error_info(inode->i_sb, function, line);
273df556 557 va_start(args, fmt);
f7c21177
TT		 558 vaf.fmt = fmt;
559 vaf.va = &args;
c398eda0 560 if (block)
d9ee81da
JP		 561 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
562 "inode #%lu: block %llu: comm %s: %pV\n",
563 inode->i_sb->s_id, function, line, inode->i_ino,
564 block, current->comm, &vaf);
565 else
566 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
567 "inode #%lu: comm %s: %pV\n",
568 inode->i_sb->s_id, function, line, inode->i_ino,
569 current->comm, &vaf);
273df556
FM		 570 va_end(args);
571
572 ext4_handle_error(inode->i_sb);
573}
574
c398eda0 575void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 576 unsigned int line, ext4_fsblk_t block,
577 const char *fmt, ...)
273df556
FM		 578{
579 va_list args;
f7c21177 580 struct va_format vaf;
1c13d5c0 581 struct ext4_super_block *es;
273df556
FM		 582 struct inode *inode = file->f_dentry->d_inode;
583 char pathname[80], *path;
584
1c13d5c0
TT		 585 es = EXT4_SB(inode->i_sb)->s_es;
586 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
587 save_error_info(inode->i_sb, function, line);
273df556 588 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 589 if (IS_ERR(path))
273df556 590 path = "(unknown)";
f7c21177
TT		 591 va_start(args, fmt);
592 vaf.fmt = fmt;
593 vaf.va = &args;
d9ee81da
JP		 594 if (block)
595 printk(KERN_CRIT
596 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
597 "block %llu: comm %s: path %s: %pV\n",
598 inode->i_sb->s_id, function, line, inode->i_ino,
599 block, current->comm, path, &vaf);
600 else
601 printk(KERN_CRIT
602 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
603 "comm %s: path %s: %pV\n",
604 inode->i_sb->s_id, function, line, inode->i_ino,
605 current->comm, path, &vaf);
273df556
FM		 606 va_end(args);
607
608 ext4_handle_error(inode->i_sb);
609}
610
2b2d6d01 611static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 612 char nbuf[16])
613{
614 char *errstr = NULL;
615
616 switch (errno) {
617 case -EIO:
618 errstr = "IO failure";
619 break;
620 case -ENOMEM:
621 errstr = "Out of memory";
622 break;
623 case -EROFS:
78f1ddbb
TT		 624 if (!sb || (EXT4_SB(sb)->s_journal &&
625 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 626 errstr = "Journal has aborted";
627 else
628 errstr = "Readonly filesystem";
629 break;
630 default:
631 /* If the caller passed in an extra buffer for unknown
632 * errors, textualise them now. Else we just return
633 * NULL. */
634 if (nbuf) {
635 /* Check for truncated error codes... */
636 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
637 errstr = nbuf;
638 }
639 break;
640 }
641
642 return errstr;
643}
644
617ba13b 645/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 646 * automatically and invokes the appropriate error response. */
647
c398eda0
TT		 648void __ext4_std_error(struct super_block *sb, const char *function,
649 unsigned int line, int errno)
ac27a0ec
DK		 650{
651 char nbuf[16];
652 const char *errstr;
653
654 /* Special case: if the error is EROFS, and we're not already
655 * inside a transaction, then there's really no point in logging
656 * an error. */
657 if (errno == -EROFS && journal_current_handle() == NULL &&
658 (sb->s_flags & MS_RDONLY))
659 return;
660
617ba13b 661 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 662 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
663 sb->s_id, function, line, errstr);
1c13d5c0 664 save_error_info(sb, function, line);
ac27a0ec 665
617ba13b 666 ext4_handle_error(sb);
ac27a0ec
DK		 667}
668
669/*
617ba13b 670 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 671 * abort function may be used to deal with unrecoverable failures such
672 * as journal IO errors or ENOMEM at a critical moment in log management.
673 *
674 * We unconditionally force the filesystem into an ABORT|READONLY state,
675 * unless the error response on the fs has been set to panic in which
676 * case we take the easy way out and panic immediately.
677 */
678
c67d859e 679void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 680 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 681{
682 va_list args;
683
1c13d5c0 684 save_error_info(sb, function, line);
ac27a0ec 685 va_start(args, fmt);
c398eda0
TT		 686 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
687 function, line);
ac27a0ec
DK		 688 vprintk(fmt, args);
689 printk("\n");
690 va_end(args);
691
1c13d5c0
TT		 692 if ((sb->s_flags & MS_RDONLY) == 0) {
693 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
694 sb->s_flags |= MS_RDONLY;
695 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
696 if (EXT4_SB(sb)->s_journal)
697 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
698 save_error_info(sb, function, line);
699 }
ac27a0ec 700 if (test_opt(sb, ERRORS_PANIC))
617ba13b 701 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 702}
703
0ff2ea7d 704void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 705{
0ff2ea7d 706 struct va_format vaf;
b31e1552
ES		 707 va_list args;
708
709 va_start(args, fmt);
0ff2ea7d
JP		 710 vaf.fmt = fmt;
711 vaf.va = &args;
712 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 713 va_end(args);
714}
715
12062ddd 716void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 717 unsigned int line, const char *fmt, ...)
ac27a0ec 718{
0ff2ea7d 719 struct va_format vaf;
ac27a0ec
DK		 720 va_list args;
721
722 va_start(args, fmt);
0ff2ea7d
JP		 723 vaf.fmt = fmt;
724 vaf.va = &args;
725 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
726 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 727 va_end(args);
728}
729
e29136f8
TT		 730void __ext4_grp_locked_error(const char *function, unsigned int line,
731 struct super_block *sb, ext4_group_t grp,
732 unsigned long ino, ext4_fsblk_t block,
733 const char *fmt, ...)
5d1b1b3f
AK		 734__releases(bitlock)
735__acquires(bitlock)
736{
0ff2ea7d 737 struct va_format vaf;
5d1b1b3f
AK		 738 va_list args;
739 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
740
1c13d5c0
TT		 741 es->s_last_error_ino = cpu_to_le32(ino);
742 es->s_last_error_block = cpu_to_le64(block);
743 __save_error_info(sb, function, line);
0ff2ea7d 744
5d1b1b3f 745 va_start(args, fmt);
0ff2ea7d
JP		 746
747 vaf.fmt = fmt;
748 vaf.va = &args;
21149d61 749 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 750 sb->s_id, function, line, grp);
751 if (ino)
0ff2ea7d 752 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 753 if (block)
0ff2ea7d
JP		 754 printk(KERN_CONT "block %llu:", (unsigned long long) block);
755 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 756 va_end(args);
757
758 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 759 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 760 return;
761 }
1c13d5c0 762
5d1b1b3f
AK		 763 ext4_unlock_group(sb, grp);
764 ext4_handle_error(sb);
765 /*
766 * We only get here in the ERRORS_RO case; relocking the group
767 * may be dangerous, but nothing bad will happen since the
768 * filesystem will have already been marked read/only and the
769 * journal has been aborted. We return 1 as a hint to callers
770 * who might what to use the return value from
25985edc 771 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 772 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
773 * aggressively from the ext4 function in question, with a
774 * more appropriate error code.
775 */
776 ext4_lock_group(sb, grp);
777 return;
778}
779
617ba13b 780void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 781{
617ba13b 782 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 783
617ba13b 784 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 785 return;
786
12062ddd 787 ext4_warning(sb,
ac27a0ec
DK		 788 "updating to rev %d because of new feature flag, "
789 "running e2fsck is recommended",
617ba13b 790 EXT4_DYNAMIC_REV);
ac27a0ec 791
617ba13b
MC		 792 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
793 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
794 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 795 /* leave es->s_feature_*compat flags alone */
796 /* es->s_uuid will be set by e2fsck if empty */
797
798 /*
799 * The rest of the superblock fields should be zero, and if not it
800 * means they are likely already in use, so leave them alone. We
801 * can leave it up to e2fsck to clean up any inconsistencies there.
802 */
803}
804
805/*
806 * Open the external journal device
807 */
b31e1552 808static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 809{
810 struct block_device *bdev;
811 char b[BDEVNAME_SIZE];
812
d4d77629 813 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 814 if (IS_ERR(bdev))
815 goto fail;
816 return bdev;
817
818fail:
b31e1552 819 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 820 __bdevname(dev, b), PTR_ERR(bdev));
821 return NULL;
822}
823
824/*
825 * Release the journal device
826 */
617ba13b 827static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 828{
e525fd89 829 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 830}
831
617ba13b 832static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 833{
834 struct block_device *bdev;
835 int ret = -ENODEV;
836
837 bdev = sbi->journal_bdev;
838 if (bdev) {
617ba13b 839 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 840 sbi->journal_bdev = NULL;
841 }
842 return ret;
843}
844
845static inline struct inode *orphan_list_entry(struct list_head *l)
846{
617ba13b 847 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 848}
849
617ba13b 850static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 851{
852 struct list_head *l;
853
b31e1552
ES		 854 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
855 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 856
857 printk(KERN_ERR "sb_info orphan list:\n");
858 list_for_each(l, &sbi->s_orphan) {
859 struct inode *inode = orphan_list_entry(l);
860 printk(KERN_ERR " "
861 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
862 inode->i_sb->s_id, inode->i_ino, inode,
863 inode->i_mode, inode->i_nlink,
864 NEXT_ORPHAN(inode));
865 }
866}
867
2b2d6d01 868static void ext4_put_super(struct super_block *sb)
ac27a0ec 869{
617ba13b
MC		 870 struct ext4_sb_info *sbi = EXT4_SB(sb);
871 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 872 int i, err;
ac27a0ec 873
857ac889 874 ext4_unregister_li_request(sb);
e0ccfd95
CH		 875 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
876
4c0425ff
MC		 877 flush_workqueue(sbi->dio_unwritten_wq);
878 destroy_workqueue(sbi->dio_unwritten_wq);
879
a9e220f8 880 lock_super(sb);
0390131b
FM		 881 if (sbi->s_journal) {
882 err = jbd2_journal_destroy(sbi->s_journal);
883 sbi->s_journal = NULL;
884 if (err < 0)
c67d859e 885 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 886 }
d4edac31 887
a1c6c569 888 del_timer(&sbi->s_err_report);
d4edac31
JB		 889 ext4_release_system_zone(sb);
890 ext4_mb_release(sb);
891 ext4_ext_release(sb);
892 ext4_xattr_put_super(sb);
893
ac27a0ec 894 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 895 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 896 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 897 }
a8e25a83
AB		 898 if (sb->s_dirt || !(sb->s_flags & MS_RDONLY))
899 ext4_commit_super(sb, 1);
900
240799cd 901 if (sbi->s_proc) {
66acdcf4 902 remove_proc_entry("options", sbi->s_proc);
9f6200bb 903 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 904 }
3197ebdb 905 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 906
907 for (i = 0; i < sbi->s_gdb_count; i++)
908 brelse(sbi->s_group_desc[i]);
f18a5f21 909 ext4_kvfree(sbi->s_group_desc);
9933fc0a 910 ext4_kvfree(sbi->s_flex_groups);
57042651 911 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 912 percpu_counter_destroy(&sbi->s_freeinodes_counter);
913 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 914 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 915 brelse(sbi->s_sbh);
916#ifdef CONFIG_QUOTA
917 for (i = 0; i < MAXQUOTAS; i++)
918 kfree(sbi->s_qf_names[i]);
919#endif
920
921 /* Debugging code just in case the in-memory inode orphan list
922 * isn't empty. The on-disk one can be non-empty if we've
923 * detected an error and taken the fs readonly, but the
924 * in-memory list had better be clean by this point. */
925 if (!list_empty(&sbi->s_orphan))
926 dump_orphan_list(sb, sbi);
927 J_ASSERT(list_empty(&sbi->s_orphan));
928
f98393a6 929 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 930 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
931 /*
932 * Invalidate the journal device's buffers. We don't want them
933 * floating about in memory - the physical journal device may
934 * hotswapped, and it breaks the `ro-after' testing code.
935 */
936 sync_blockdev(sbi->journal_bdev);
f98393a6 937 invalidate_bdev(sbi->journal_bdev);
617ba13b 938 ext4_blkdev_remove(sbi);
ac27a0ec 939 }
c5e06d10
JL		 940 if (sbi->s_mmp_tsk)
941 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 942 sb->s_fs_info = NULL;
3197ebdb
TT		 943 /*
944 * Now that we are completely done shutting down the
945 * superblock, we need to actually destroy the kobject.
946 */
3197ebdb
TT		 947 unlock_super(sb);
948 kobject_put(&sbi->s_kobj);
949 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 950 if (sbi->s_chksum_driver)
951 crypto_free_shash(sbi->s_chksum_driver);
705895b6 952 kfree(sbi->s_blockgroup_lock);
ac27a0ec 953 kfree(sbi);
ac27a0ec
DK		 954}
955
e18b890b 956static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 957
958/*
959 * Called inside transaction, so use GFP_NOFS
960 */
617ba13b 961static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 962{
617ba13b 963 struct ext4_inode_info *ei;
ac27a0ec 964
e6b4f8da 965 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 966 if (!ei)
967 return NULL;
0b8e58a1 968
ac27a0ec 969 ei->vfs_inode.i_version = 1;
91246c00 970 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 971 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 972 INIT_LIST_HEAD(&ei->i_prealloc_list);
973 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 974 ei->i_reserved_data_blocks = 0;
975 ei->i_reserved_meta_blocks = 0;
976 ei->i_allocated_meta_blocks = 0;
9d0be502 977 ei->i_da_metadata_calc_len = 0;
d2a17637 978 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 979#ifdef CONFIG_QUOTA
980 ei->i_reserved_quota = 0;
981#endif
8aefcd55 982 ei->jinode = NULL;
c7064ef1 983 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 984 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 985 ei->cur_aio_dio = NULL;
b436b9be
JK		 986 ei->i_sync_tid = 0;
987 ei->i_datasync_tid = 0;
f7ad6d2e 988 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 989 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 990
ac27a0ec
DK		 991 return &ei->vfs_inode;
992}
993
7ff9c073
TT		 994static int ext4_drop_inode(struct inode *inode)
995{
996 int drop = generic_drop_inode(inode);
997
998 trace_ext4_drop_inode(inode, drop);
999 return drop;
1000}
1001
fa0d7e3d
NP		 1002static void ext4_i_callback(struct rcu_head *head)
1003{
1004 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 1005 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1006}
1007
617ba13b 1008static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1009{
9f7dd93d 1010 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1011 ext4_msg(inode->i_sb, KERN_ERR,
1012 "Inode %lu (%p): orphan list check failed!",
1013 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1014 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1015 EXT4_I(inode), sizeof(struct ext4_inode_info),
1016 true);
1017 dump_stack();
1018 }
fa0d7e3d 1019 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1020}
1021
51cc5068 1022static void init_once(void *foo)
ac27a0ec 1023{
617ba13b 1024 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1025
a35afb83 1026 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 1027#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 1028 init_rwsem(&ei->xattr_sem);
ac27a0ec 1029#endif
0e855ac8 1030 init_rwsem(&ei->i_data_sem);
a35afb83 1031 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1032}
1033
1034static int init_inodecache(void)
1035{
617ba13b
MC		 1036 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1037 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 1038 0, (SLAB_RECLAIM_ACCOUNT|
1039 SLAB_MEM_SPREAD),
20c2df83 1040 init_once);
617ba13b 1041 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1042 return -ENOMEM;
1043 return 0;
1044}
1045
1046static void destroy_inodecache(void)
1047{
617ba13b 1048 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1049}
1050
0930fcc1 1051void ext4_clear_inode(struct inode *inode)
ac27a0ec 1052{
0930fcc1
AV		 1053 invalidate_inode_buffers(inode);
1054 end_writeback(inode);
9f754758 1055 dquot_drop(inode);
c2ea3fde 1056 ext4_discard_preallocations(inode);
8aefcd55
TT		 1057 if (EXT4_I(inode)->jinode) {
1058 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1059 EXT4_I(inode)->jinode);
1060 jbd2_free_inode(EXT4_I(inode)->jinode);
1061 EXT4_I(inode)->jinode = NULL;
1062 }
ac27a0ec
DK		 1063}
1064
1b961ac0 1065static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1066 u64 ino, u32 generation)
ac27a0ec 1067{
ac27a0ec 1068 struct inode *inode;
ac27a0ec 1069
617ba13b 1070 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1071 return ERR_PTR(-ESTALE);
617ba13b 1072 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1073 return ERR_PTR(-ESTALE);
1074
1075 /* iget isn't really right if the inode is currently unallocated!!
1076 *
617ba13b 1077 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1078 * deleted, so we should be safe.
1079 *
1080 * Currently we don't know the generation for parent directory, so
1081 * a generation of 0 means "accept any"
1082 */
1d1fe1ee
DH		 1083 inode = ext4_iget(sb, ino);
1084 if (IS_ERR(inode))
1085 return ERR_CAST(inode);
1086 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1087 iput(inode);
1088 return ERR_PTR(-ESTALE);
1089 }
1b961ac0
CH		 1090
1091 return inode;
1092}
1093
1094static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1095 int fh_len, int fh_type)
1b961ac0
CH		 1096{
1097 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1098 ext4_nfs_get_inode);
1099}
1100
1101static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1102 int fh_len, int fh_type)
1b961ac0
CH		 1103{
1104 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1105 ext4_nfs_get_inode);
ac27a0ec
DK		 1106}
1107
c39a7f84
TO		 1108/*
1109 * Try to release metadata pages (indirect blocks, directories) which are
1110 * mapped via the block device. Since these pages could have journal heads
1111 * which would prevent try_to_free_buffers() from freeing them, we must use
1112 * jbd2 layer's try_to_free_buffers() function to release them.
1113 */
0b8e58a1
AD		 1114static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1115 gfp_t wait)
c39a7f84
TO		 1116{
1117 journal_t *journal = EXT4_SB(sb)->s_journal;
1118
1119 WARN_ON(PageChecked(page));
1120 if (!page_has_buffers(page))
1121 return 0;
1122 if (journal)
1123 return jbd2_journal_try_to_free_buffers(journal, page,
1124 wait & ~__GFP_WAIT);
1125 return try_to_free_buffers(page);
1126}
1127
ac27a0ec 1128#ifdef CONFIG_QUOTA
af5bc92d 1129#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1130#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1131
617ba13b
MC		 1132static int ext4_write_dquot(struct dquot *dquot);
1133static int ext4_acquire_dquot(struct dquot *dquot);
1134static int ext4_release_dquot(struct dquot *dquot);
1135static int ext4_mark_dquot_dirty(struct dquot *dquot);
1136static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1137static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1138 struct path *path);
ca0e05e4 1139static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1140static int ext4_quota_on_mount(struct super_block *sb, int type);
1141static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1142 size_t len, loff_t off);
617ba13b 1143static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1144 const char *data, size_t len, loff_t off);
1145
61e225dc 1146static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1147 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1148 .write_dquot = ext4_write_dquot,
1149 .acquire_dquot = ext4_acquire_dquot,
1150 .release_dquot = ext4_release_dquot,
1151 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1152 .write_info = ext4_write_info,
1153 .alloc_dquot = dquot_alloc,
1154 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1155};
1156
0d54b217 1157static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1158 .quota_on = ext4_quota_on,
ca0e05e4 1159 .quota_off = ext4_quota_off,
287a8095
CH		 1160 .quota_sync = dquot_quota_sync,
1161 .get_info = dquot_get_dqinfo,
1162 .set_info = dquot_set_dqinfo,
1163 .get_dqblk = dquot_get_dqblk,
1164 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1165};
1166#endif
1167
ee9b6d61 1168static const struct super_operations ext4_sops = {
617ba13b
MC		 1169 .alloc_inode = ext4_alloc_inode,
1170 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1171 .write_inode = ext4_write_inode,
1172 .dirty_inode = ext4_dirty_inode,
7ff9c073 1173 .drop_inode = ext4_drop_inode,
0930fcc1 1174 .evict_inode = ext4_evict_inode,
617ba13b 1175 .put_super = ext4_put_super,
617ba13b 1176 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1177 .freeze_fs = ext4_freeze,
1178 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1179 .statfs = ext4_statfs,
1180 .remount_fs = ext4_remount,
617ba13b 1181 .show_options = ext4_show_options,
ac27a0ec 1182#ifdef CONFIG_QUOTA
617ba13b
MC		 1183 .quota_read = ext4_quota_read,
1184 .quota_write = ext4_quota_write,
ac27a0ec 1185#endif
c39a7f84 1186 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1187};
1188
9ca92389
TT		 1189static const struct super_operations ext4_nojournal_sops = {
1190 .alloc_inode = ext4_alloc_inode,
1191 .destroy_inode = ext4_destroy_inode,
1192 .write_inode = ext4_write_inode,
1193 .dirty_inode = ext4_dirty_inode,
7ff9c073 1194 .drop_inode = ext4_drop_inode,
0930fcc1 1195 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1196 .write_super = ext4_write_super,
1197 .put_super = ext4_put_super,
1198 .statfs = ext4_statfs,
1199 .remount_fs = ext4_remount,
9ca92389
TT		 1200 .show_options = ext4_show_options,
1201#ifdef CONFIG_QUOTA
1202 .quota_read = ext4_quota_read,
1203 .quota_write = ext4_quota_write,
1204#endif
1205 .bdev_try_to_free_page = bdev_try_to_free_page,
1206};
1207
39655164 1208static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1209 .fh_to_dentry = ext4_fh_to_dentry,
1210 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1211 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1212};
1213
1214enum {
1215 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1216 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1217 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1218 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1219 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1220 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1221 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1222 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1223 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1224 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1225 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1226 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1227 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1228 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1229 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1230 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1231 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1232 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
ac27a0ec
DK		 1233};
1234
a447c093 1235static const match_table_t tokens = {
ac27a0ec
DK		 1236 {Opt_bsd_df, "bsddf"},
1237 {Opt_minix_df, "minixdf"},
1238 {Opt_grpid, "grpid"},
1239 {Opt_grpid, "bsdgroups"},
1240 {Opt_nogrpid, "nogrpid"},
1241 {Opt_nogrpid, "sysvgroups"},
1242 {Opt_resgid, "resgid=%u"},
1243 {Opt_resuid, "resuid=%u"},
1244 {Opt_sb, "sb=%u"},
1245 {Opt_err_cont, "errors=continue"},
1246 {Opt_err_panic, "errors=panic"},
1247 {Opt_err_ro, "errors=remount-ro"},
1248 {Opt_nouid32, "nouid32"},
ac27a0ec 1249 {Opt_debug, "debug"},
72578c33
TT		 1250 {Opt_removed, "oldalloc"},
1251 {Opt_removed, "orlov"},
ac27a0ec
DK		 1252 {Opt_user_xattr, "user_xattr"},
1253 {Opt_nouser_xattr, "nouser_xattr"},
1254 {Opt_acl, "acl"},
1255 {Opt_noacl, "noacl"},
e3bb52ae 1256 {Opt_noload, "norecovery"},
5a916be1 1257 {Opt_noload, "noload"},
72578c33
TT		 1258 {Opt_removed, "nobh"},
1259 {Opt_removed, "bh"},
ac27a0ec 1260 {Opt_commit, "commit=%u"},
30773840
TT		 1261 {Opt_min_batch_time, "min_batch_time=%u"},
1262 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1263 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1264 {Opt_journal_checksum, "journal_checksum"},
1265 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1266 {Opt_abort, "abort"},
1267 {Opt_data_journal, "data=journal"},
1268 {Opt_data_ordered, "data=ordered"},
1269 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1270 {Opt_data_err_abort, "data_err=abort"},
1271 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1272 {Opt_offusrjquota, "usrjquota="},
1273 {Opt_usrjquota, "usrjquota=%s"},
1274 {Opt_offgrpjquota, "grpjquota="},
1275 {Opt_grpjquota, "grpjquota=%s"},
1276 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1277 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1278 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1279 {Opt_grpquota, "grpquota"},
1280 {Opt_noquota, "noquota"},
1281 {Opt_quota, "quota"},
1282 {Opt_usrquota, "usrquota"},
1283 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1284 {Opt_barrier, "barrier"},
1285 {Opt_nobarrier, "nobarrier"},
25ec56b5 1286 {Opt_i_version, "i_version"},
c9de560d 1287 {Opt_stripe, "stripe=%u"},
64769240 1288 {Opt_delalloc, "delalloc"},
dd919b98 1289 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1290 {Opt_mblk_io_submit, "mblk_io_submit"},
1291 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1292 {Opt_block_validity, "block_validity"},
1293 {Opt_noblock_validity, "noblock_validity"},
240799cd 1294 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1295 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1296 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1297 {Opt_auto_da_alloc, "auto_da_alloc"},
1298 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1299 {Opt_dioread_nolock, "dioread_nolock"},
1300 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1301 {Opt_discard, "discard"},
1302 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1303 {Opt_init_itable, "init_itable=%u"},
1304 {Opt_init_itable, "init_itable"},
1305 {Opt_noinit_itable, "noinit_itable"},
c7198b9c
TT		 1306 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1307 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1308 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1309 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1310 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1311 {Opt_err, NULL},
ac27a0ec
DK		 1312};
1313
617ba13b 1314static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1315{
617ba13b 1316 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1317 char *options = (char *) *data;
1318
1319 if (!options || strncmp(options, "sb=", 3) != 0)
1320 return 1; /* Default location */
0b8e58a1 1321
ac27a0ec 1322 options += 3;
0b8e58a1 1323 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1324 sb_block = simple_strtoul(options, &options, 0);
1325 if (*options && *options != ',') {
4776004f 1326 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1327 (char *) *data);
1328 return 1;
1329 }
1330 if (*options == ',')
1331 options++;
1332 *data = (void *) options;
0b8e58a1 1333
ac27a0ec
DK		 1334 return sb_block;
1335}
1336
b3881f74 1337#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1338static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1339 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1340
56c50f11
DM		 1341#ifdef CONFIG_QUOTA
1342static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1343{
1344 struct ext4_sb_info *sbi = EXT4_SB(sb);
1345 char *qname;
1346
1347 if (sb_any_quota_loaded(sb) &&
1348 !sbi->s_qf_names[qtype]) {
1349 ext4_msg(sb, KERN_ERR,
1350 "Cannot change journaled "
1351 "quota options when quota turned on");
57f73c2c 1352 return -1;
56c50f11
DM		 1353 }
1354 qname = match_strdup(args);
1355 if (!qname) {
1356 ext4_msg(sb, KERN_ERR,
1357 "Not enough memory for storing quotafile name");
57f73c2c 1358 return -1;
56c50f11
DM		 1359 }
1360 if (sbi->s_qf_names[qtype] &&
1361 strcmp(sbi->s_qf_names[qtype], qname)) {
1362 ext4_msg(sb, KERN_ERR,
1363 "%s quota file already specified", QTYPE2NAME(qtype));
1364 kfree(qname);
57f73c2c 1365 return -1;
56c50f11
DM		 1366 }
1367 sbi->s_qf_names[qtype] = qname;
1368 if (strchr(sbi->s_qf_names[qtype], '/')) {
1369 ext4_msg(sb, KERN_ERR,
1370 "quotafile must be on filesystem root");
1371 kfree(sbi->s_qf_names[qtype]);
1372 sbi->s_qf_names[qtype] = NULL;
57f73c2c 1373 return -1;
56c50f11 1374 }
fd8c37ec 1375 set_opt(sb, QUOTA);
56c50f11
DM		 1376 return 1;
1377}
1378
1379static int clear_qf_name(struct super_block *sb, int qtype)
1380{
1381
1382 struct ext4_sb_info *sbi = EXT4_SB(sb);
1383
1384 if (sb_any_quota_loaded(sb) &&
1385 sbi->s_qf_names[qtype]) {
1386 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1387 " when quota turned on");
57f73c2c 1388 return -1;
56c50f11
DM		 1389 }
1390 /*
1391 * The space will be released later when all options are confirmed
1392 * to be correct
1393 */
1394 sbi->s_qf_names[qtype] = NULL;
1395 return 1;
1396}
1397#endif
1398
26092bf5
TT		 1399#define MOPT_SET 0x0001
1400#define MOPT_CLEAR 0x0002
1401#define MOPT_NOSUPPORT 0x0004
1402#define MOPT_EXPLICIT 0x0008
1403#define MOPT_CLEAR_ERR 0x0010
1404#define MOPT_GTE0 0x0020
ac27a0ec 1405#ifdef CONFIG_QUOTA
26092bf5
TT		 1406#define MOPT_Q 0
1407#define MOPT_QFMT 0x0040
1408#else
1409#define MOPT_Q MOPT_NOSUPPORT
1410#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1411#endif
26092bf5
TT		 1412#define MOPT_DATAJ 0x0080
1413
1414static const struct mount_opts {
1415 int token;
1416 int mount_opt;
1417 int flags;
1418} ext4_mount_opts[] = {
1419 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1420 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1421 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1422 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1423 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1424 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1425 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1426 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1427 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1428 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1429 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1430 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1431 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1432 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1433 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1434 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1435 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1436 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1437 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1438 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1439 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1440 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1441 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1442 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1443 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1444 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1445 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1446 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1447 {Opt_commit, 0, MOPT_GTE0},
1448 {Opt_max_batch_time, 0, MOPT_GTE0},
1449 {Opt_min_batch_time, 0, MOPT_GTE0},
1450 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1451 {Opt_init_itable, 0, MOPT_GTE0},
1452 {Opt_stripe, 0, MOPT_GTE0},
1453 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1454 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1455 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
03010a33 1456#ifdef CONFIG_EXT4_FS_XATTR
26092bf5
TT		 1457 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1458 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
ac27a0ec 1459#else
26092bf5
TT		 1460 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1461 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
ac27a0ec 1462#endif
03010a33 1463#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1464 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1465 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1466#else
26092bf5
TT		 1467 {Opt_acl, 0, MOPT_NOSUPPORT},
1468 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1469#endif
26092bf5
TT		 1470 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1471 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1472 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1473 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1474 MOPT_SET | MOPT_Q},
1475 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1476 MOPT_SET | MOPT_Q},
1477 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1478 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1479 {Opt_usrjquota, 0, MOPT_Q},
1480 {Opt_grpjquota, 0, MOPT_Q},
1481 {Opt_offusrjquota, 0, MOPT_Q},
1482 {Opt_offgrpjquota, 0, MOPT_Q},
1483 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1484 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1485 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1486 {Opt_err, 0, 0}
1487};
1488
1489static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1490 substring_t *args, unsigned long *journal_devnum,
1491 unsigned int *journal_ioprio, int is_remount)
1492{
1493 struct ext4_sb_info *sbi = EXT4_SB(sb);
1494 const struct mount_opts *m;
1495 int arg = 0;
1496
57f73c2c
TT		 1497#ifdef CONFIG_QUOTA
1498 if (token == Opt_usrjquota)
1499 return set_qf_name(sb, USRQUOTA, &args[0]);
1500 else if (token == Opt_grpjquota)
1501 return set_qf_name(sb, GRPQUOTA, &args[0]);
1502 else if (token == Opt_offusrjquota)
1503 return clear_qf_name(sb, USRQUOTA);
1504 else if (token == Opt_offgrpjquota)
1505 return clear_qf_name(sb, GRPQUOTA);
1506#endif
26092bf5
TT		 1507 if (args->from && match_int(args, &arg))
1508 return -1;
1509 switch (token) {
f7048605
TT		 1510 case Opt_noacl:
1511 case Opt_nouser_xattr:
1512 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1513 break;
26092bf5
TT		 1514 case Opt_sb:
1515 return 1; /* handled by get_sb_block() */
1516 case Opt_removed:
1517 ext4_msg(sb, KERN_WARNING,
1518 "Ignoring removed %s option", opt);
1519 return 1;
1520 case Opt_resuid:
1521 sbi->s_resuid = arg;
1522 return 1;
1523 case Opt_resgid:
1524 sbi->s_resgid = arg;
1525 return 1;
1526 case Opt_abort:
1527 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1528 return 1;
1529 case Opt_i_version:
1530 sb->s_flags |= MS_I_VERSION;
1531 return 1;
1532 case Opt_journal_dev:
1533 if (is_remount) {
1534 ext4_msg(sb, KERN_ERR,
1535 "Cannot specify journal on remount");
1536 return -1;
1537 }
1538 *journal_devnum = arg;
1539 return 1;
1540 case Opt_journal_ioprio:
1541 if (arg < 0 || arg > 7)
1542 return -1;
1543 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1544 return 1;
1545 }
1546
1547 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1548 if (token != m->token)
1549 continue;
1550 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1551 return -1;
1552 if (m->flags & MOPT_EXPLICIT)
1553 set_opt2(sb, EXPLICIT_DELALLOC);
1554 if (m->flags & MOPT_CLEAR_ERR)
1555 clear_opt(sb, ERRORS_MASK);
1556 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1557 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1558 "options when quota turned on");
1559 return -1;
1560 }
1561
1562 if (m->flags & MOPT_NOSUPPORT) {
1563 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1564 } else if (token == Opt_commit) {
1565 if (arg == 0)
1566 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1567 sbi->s_commit_interval = HZ * arg;
1568 } else if (token == Opt_max_batch_time) {
1569 if (arg == 0)
1570 arg = EXT4_DEF_MAX_BATCH_TIME;
1571 sbi->s_max_batch_time = arg;
1572 } else if (token == Opt_min_batch_time) {
1573 sbi->s_min_batch_time = arg;
1574 } else if (token == Opt_inode_readahead_blks) {
1575 if (arg > (1 << 30))
1576 return -1;
1577 if (arg && !is_power_of_2(arg)) {
b31e1552 1578 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1579 "EXT4-fs: inode_readahead_blks"
1580 " must be a power of 2");
1581 return -1;
ac27a0ec 1582 }
26092bf5
TT		 1583 sbi->s_inode_readahead_blks = arg;
1584 } else if (token == Opt_init_itable) {
1585 set_opt(sb, INIT_INODE_TABLE);
1586 if (!args->from)
1587 arg = EXT4_DEF_LI_WAIT_MULT;
1588 sbi->s_li_wait_mult = arg;
1589 } else if (token == Opt_stripe) {
1590 sbi->s_stripe = arg;
1591 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1592 if (is_remount) {
eb513689
TT		 1593 if (!sbi->s_journal)
1594 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1595 else if (test_opt(sb, DATA_FLAGS) !=
1596 m->mount_opt) {
b31e1552 1597 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1598 "Cannot change data mode on remount");
1599 return -1;
ac27a0ec
DK		 1600 }
1601 } else {
fd8c37ec 1602 clear_opt(sb, DATA_FLAGS);
26092bf5 1603 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1604 }
ac27a0ec 1605#ifdef CONFIG_QUOTA
26092bf5 1606 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1607 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1608 sbi->s_jquota_fmt != m->mount_opt) {
1609 ext4_msg(sb, KERN_ERR, "Cannot "
1610 "change journaled quota options "
1611 "when quota turned on");
1612 return -1;
ac27a0ec 1613 }
26092bf5 1614 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1615#endif
26092bf5
TT		 1616 } else {
1617 if (!args->from)
1618 arg = 1;
1619 if (m->flags & MOPT_CLEAR)
1620 arg = !arg;
1621 else if (unlikely(!(m->flags & MOPT_SET))) {
1622 ext4_msg(sb, KERN_WARNING,
1623 "buggy handling of option %s", opt);
1624 WARN_ON(1);
1625 return -1;
3197ebdb 1626 }
26092bf5
TT		 1627 if (arg != 0)
1628 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1629 else
26092bf5 1630 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1631 }
26092bf5
TT		 1632 return 1;
1633 }
1634 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1635 "or missing value", opt);
1636 return -1;
1637}
1638
1639static int parse_options(char *options, struct super_block *sb,
1640 unsigned long *journal_devnum,
1641 unsigned int *journal_ioprio,
1642 int is_remount)
1643{
db7e5c66 1644#ifdef CONFIG_QUOTA
26092bf5 1645 struct ext4_sb_info *sbi = EXT4_SB(sb);
db7e5c66 1646#endif
26092bf5
TT		 1647 char *p;
1648 substring_t args[MAX_OPT_ARGS];
1649 int token;
1650
1651 if (!options)
1652 return 1;
1653
1654 while ((p = strsep(&options, ",")) != NULL) {
1655 if (!*p)
1656 continue;
1657 /*
1658 * Initialize args struct so we know whether arg was
1659 * found; some options take optional arguments.
1660 */
1661 args[0].to = args[0].from = 0;
1662 token = match_token(p, tokens, args);
1663 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1664 journal_ioprio, is_remount) < 0)
1665 return 0;
ac27a0ec
DK		 1666 }
1667#ifdef CONFIG_QUOTA
1668 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1669 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1670 clear_opt(sb, USRQUOTA);
ac27a0ec 1671
482a7425 1672 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1673 clear_opt(sb, GRPQUOTA);
ac27a0ec 1674
56c50f11 1675 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1676 ext4_msg(sb, KERN_ERR, "old and new quota "
1677 "format mixing");
ac27a0ec
DK		 1678 return 0;
1679 }
1680
1681 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1682 ext4_msg(sb, KERN_ERR, "journaled quota format "
1683 "not specified");
ac27a0ec
DK		 1684 return 0;
1685 }
1686 } else {
1687 if (sbi->s_jquota_fmt) {
b31e1552 1688 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1689 "specified with no journaling "
b31e1552 1690 "enabled");
ac27a0ec
DK		 1691 return 0;
1692 }
1693 }
1694#endif
1695 return 1;
1696}
1697
2adf6da8
TT		 1698static inline void ext4_show_quota_options(struct seq_file *seq,
1699 struct super_block *sb)
1700{
1701#if defined(CONFIG_QUOTA)
1702 struct ext4_sb_info *sbi = EXT4_SB(sb);
1703
1704 if (sbi->s_jquota_fmt) {
1705 char *fmtname = "";
1706
1707 switch (sbi->s_jquota_fmt) {
1708 case QFMT_VFS_OLD:
1709 fmtname = "vfsold";
1710 break;
1711 case QFMT_VFS_V0:
1712 fmtname = "vfsv0";
1713 break;
1714 case QFMT_VFS_V1:
1715 fmtname = "vfsv1";
1716 break;
1717 }
1718 seq_printf(seq, ",jqfmt=%s", fmtname);
1719 }
1720
1721 if (sbi->s_qf_names[USRQUOTA])
1722 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1723
1724 if (sbi->s_qf_names[GRPQUOTA])
1725 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1726
1727 if (test_opt(sb, USRQUOTA))
1728 seq_puts(seq, ",usrquota");
1729
1730 if (test_opt(sb, GRPQUOTA))
1731 seq_puts(seq, ",grpquota");
1732#endif
1733}
1734
5a916be1
TT		 1735static const char *token2str(int token)
1736{
1737 static const struct match_token *t;
1738
1739 for (t = tokens; t->token != Opt_err; t++)
1740 if (t->token == token && !strchr(t->pattern, '='))
1741 break;
1742 return t->pattern;
1743}
1744
2adf6da8
TT		 1745/*
1746 * Show an option if
1747 * - it's set to a non-default value OR
1748 * - if the per-sb default is different from the global default
1749 */
66acdcf4
TT		 1750static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1751 int nodefs)
2adf6da8 1752{
2adf6da8
TT		 1753 struct ext4_sb_info *sbi = EXT4_SB(sb);
1754 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1755 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1756 const struct mount_opts *m;
66acdcf4 1757 char sep = nodefs ? '\n' : ',';
2adf6da8 1758
66acdcf4
TT		 1759#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1760#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1761
1762 if (sbi->s_sb_block != 1)
5a916be1
TT		 1763 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1764
1765 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1766 int want_set = m->flags & MOPT_SET;
1767 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1768 (m->flags & MOPT_CLEAR_ERR))
1769 continue;
66acdcf4 1770 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1771 continue; /* skip if same as the default */
1772 if ((want_set &&
1773 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1774 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1775 continue; /* select Opt_noFoo vs Opt_Foo */
1776 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1777 }
5a916be1 1778
66acdcf4 1779 if (nodefs || sbi->s_resuid != EXT4_DEF_RESUID ||
5a916be1
TT		 1780 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1781 SEQ_OPTS_PRINT("resuid=%u", sbi->s_resuid);
66acdcf4 1782 if (nodefs || sbi->s_resgid != EXT4_DEF_RESGID ||
5a916be1
TT		 1783 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1784 SEQ_OPTS_PRINT("resgid=%u", sbi->s_resgid);
66acdcf4 1785 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1786 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1787 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1788 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1789 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1790 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1791 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1792 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1793 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1794 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1795 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1796 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1797 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1798 if (sb->s_flags & MS_I_VERSION)
5a916be1 1799 SEQ_OPTS_PUTS("i_version");
66acdcf4 1800 if (nodefs || sbi->s_stripe)
5a916be1 1801 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1802 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1803 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1804 SEQ_OPTS_PUTS("data=journal");
1805 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1806 SEQ_OPTS_PUTS("data=ordered");
1807 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1808 SEQ_OPTS_PUTS("data=writeback");
1809 }
66acdcf4
TT		 1810 if (nodefs ||
1811 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1812 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1813 sbi->s_inode_readahead_blks);
2adf6da8 1814
66acdcf4
TT		 1815 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1816 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1817 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2adf6da8
TT		 1818
1819 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1820 return 0;
1821}
1822
66acdcf4
TT		 1823static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1824{
1825 return _ext4_show_options(seq, root->d_sb, 0);
1826}
1827
1828static int options_seq_show(struct seq_file *seq, void *offset)
1829{
1830 struct super_block *sb = seq->private;
1831 int rc;
1832
1833 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1834 rc = _ext4_show_options(seq, sb, 1);
1835 seq_puts(seq, "\n");
1836 return rc;
1837}
1838
1839static int options_open_fs(struct inode *inode, struct file *file)
1840{
1841 return single_open(file, options_seq_show, PDE(inode)->data);
1842}
1843
1844static const struct file_operations ext4_seq_options_fops = {
1845 .owner = THIS_MODULE,
1846 .open = options_open_fs,
1847 .read = seq_read,
1848 .llseek = seq_lseek,
1849 .release = single_release,
1850};
1851
617ba13b 1852static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1853 int read_only)
1854{
617ba13b 1855 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1856 int res = 0;
1857
617ba13b 1858 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1859 ext4_msg(sb, KERN_ERR, "revision level too high, "
1860 "forcing read-only mode");
ac27a0ec
DK		 1861 res = MS_RDONLY;
1862 }
1863 if (read_only)
281b5995 1864 goto done;
617ba13b 1865 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1866 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1867 "running e2fsck is recommended");
617ba13b 1868 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1869 ext4_msg(sb, KERN_WARNING,
1870 "warning: mounting fs with errors, "
1871 "running e2fsck is recommended");
ed3ce80a 1872 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1873 le16_to_cpu(es->s_mnt_count) >=
1874 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1875 ext4_msg(sb, KERN_WARNING,
1876 "warning: maximal mount count reached, "
1877 "running e2fsck is recommended");
ac27a0ec
DK		 1878 else if (le32_to_cpu(es->s_checkinterval) &&
1879 (le32_to_cpu(es->s_lastcheck) +
1880 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1881 ext4_msg(sb, KERN_WARNING,
1882 "warning: checktime reached, "
1883 "running e2fsck is recommended");
0b8e58a1 1884 if (!sbi->s_journal)
0390131b 1885 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1886 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1887 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1888 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1889 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1890 ext4_update_dynamic_rev(sb);
0390131b
FM		 1891 if (sbi->s_journal)
1892 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1893
e2d67052 1894 ext4_commit_super(sb, 1);
281b5995 1895done:
ac27a0ec 1896 if (test_opt(sb, DEBUG))
a9df9a49 1897 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1898 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1899 sb->s_blocksize,
1900 sbi->s_groups_count,
617ba13b
MC		 1901 EXT4_BLOCKS_PER_GROUP(sb),
1902 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1903 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1904
7abc52c2 1905 cleancache_init_fs(sb);
ac27a0ec
DK		 1906 return res;
1907}
1908
772cb7c8
JS		 1909static int ext4_fill_flex_info(struct super_block *sb)
1910{
1911 struct ext4_sb_info *sbi = EXT4_SB(sb);
1912 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1913 ext4_group_t flex_group_count;
1914 ext4_group_t flex_group;
d50f2ab6 1915 unsigned int groups_per_flex = 0;
c5ca7c76 1916 size_t size;
772cb7c8
JS		 1917 int i;
1918
503358ae 1919 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1920 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1921 sbi->s_log_groups_per_flex = 0;
1922 return 1;
1923 }
d50f2ab6 1924 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
772cb7c8 1925
c62a11fd
FB		 1926 /* We allocate both existing and potentially added groups */
1927 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1928 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1929 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 1930 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 1931 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 1932 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 1933 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1934 flex_group_count);
1935 goto failed;
772cb7c8 1936 }
772cb7c8 1937
772cb7c8 1938 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1939 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1940
1941 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1942 atomic_add(ext4_free_inodes_count(sb, gdp),
1943 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1944 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1945 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1946 atomic_add(ext4_used_dirs_count(sb, gdp),
1947 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1948 }
1949
1950 return 1;
1951failed:
1952 return 0;
1953}
1954
717d50e4
AD		 1955__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1956 struct ext4_group_desc *gdp)
1957{
1958 __u16 crc = 0;
1959
1960 if (sbi->s_es->s_feature_ro_compat &
1961 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1962 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1963 __le32 le_group = cpu_to_le32(block_group);
1964
1965 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1966 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1967 crc = crc16(crc, (__u8 *)gdp, offset);
1968 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1969 /* for checksum of struct ext4_group_desc do the rest...*/
1970 if ((sbi->s_es->s_feature_incompat &
1971 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1972 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1973 crc = crc16(crc, (__u8 *)gdp + offset,
1974 le16_to_cpu(sbi->s_es->s_desc_size) -
1975 offset);
1976 }
1977
1978 return cpu_to_le16(crc);
1979}
1980
1981int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1982 struct ext4_group_desc *gdp)
1983{
1984 if ((sbi->s_es->s_feature_ro_compat &
1985 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1986 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1987 return 0;
1988
1989 return 1;
1990}
1991
ac27a0ec 1992/* Called at mount-time, super-block is locked */
bfff6873
LC		 1993static int ext4_check_descriptors(struct super_block *sb,
1994 ext4_group_t *first_not_zeroed)
ac27a0ec 1995{
617ba13b
MC		 1996 struct ext4_sb_info *sbi = EXT4_SB(sb);
1997 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1998 ext4_fsblk_t last_block;
bd81d8ee
LV		 1999 ext4_fsblk_t block_bitmap;
2000 ext4_fsblk_t inode_bitmap;
2001 ext4_fsblk_t inode_table;
ce421581 2002 int flexbg_flag = 0;
bfff6873 2003 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2004
ce421581
JS		 2005 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2006 flexbg_flag = 1;
2007
af5bc92d 2008 ext4_debug("Checking group descriptors");
ac27a0ec 2009
197cd65a
AM		 2010 for (i = 0; i < sbi->s_groups_count; i++) {
2011 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2012
ce421581 2013 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2014 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2015 else
2016 last_block = first_block +
617ba13b 2017 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2018
bfff6873
LC		 2019 if ((grp == sbi->s_groups_count) &&
2020 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2021 grp = i;
2022
8fadc143 2023 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2024 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2025 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2026 "Block bitmap for group %u not in group "
b31e1552 2027 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2028 return 0;
2029 }
8fadc143 2030 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2031 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2032 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2033 "Inode bitmap for group %u not in group "
b31e1552 2034 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2035 return 0;
2036 }
8fadc143 2037 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2038 if (inode_table < first_block ||
2b2d6d01 2039 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2040 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2041 "Inode table for group %u not in group "
b31e1552 2042 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2043 return 0;
2044 }
955ce5f5 2045 ext4_lock_group(sb, i);
717d50e4 2046 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2047 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2048 "Checksum for group %u failed (%u!=%u)",
2049 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2050 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2051 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2052 ext4_unlock_group(sb, i);
8a266467 2053 return 0;
7ee1ec4c 2054 }
717d50e4 2055 }
955ce5f5 2056 ext4_unlock_group(sb, i);
ce421581
JS		 2057 if (!flexbg_flag)
2058 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2059 }
bfff6873
LC		 2060 if (NULL != first_not_zeroed)
2061 *first_not_zeroed = grp;
ac27a0ec 2062
5dee5437
TT		 2063 ext4_free_blocks_count_set(sbi->s_es,
2064 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2065 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2066 return 1;
2067}
2068
617ba13b 2069/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2070 * the superblock) which were deleted from all directories, but held open by
2071 * a process at the time of a crash. We walk the list and try to delete these
2072 * inodes at recovery time (only with a read-write filesystem).
2073 *
2074 * In order to keep the orphan inode chain consistent during traversal (in
2075 * case of crash during recovery), we link each inode into the superblock
2076 * orphan list_head and handle it the same way as an inode deletion during
2077 * normal operation (which journals the operations for us).
2078 *
2079 * We only do an iget() and an iput() on each inode, which is very safe if we
2080 * accidentally point at an in-use or already deleted inode. The worst that
2081 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2082 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2083 * e2fsck was run on this filesystem, and it must have already done the orphan
2084 * inode cleanup for us, so we can safely abort without any further action.
2085 */
2b2d6d01
TT		 2086static void ext4_orphan_cleanup(struct super_block *sb,
2087 struct ext4_super_block *es)
ac27a0ec
DK		 2088{
2089 unsigned int s_flags = sb->s_flags;
2090 int nr_orphans = 0, nr_truncates = 0;
2091#ifdef CONFIG_QUOTA
2092 int i;
2093#endif
2094 if (!es->s_last_orphan) {
2095 jbd_debug(4, "no orphan inodes to clean up\n");
2096 return;
2097 }
2098
a8f48a95 2099 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2100 ext4_msg(sb, KERN_ERR, "write access "
2101 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2102 return;
2103 }
2104
d39195c3
AG		 2105 /* Check if feature set would not allow a r/w mount */
2106 if (!ext4_feature_set_ok(sb, 0)) {
2107 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2108 "unknown ROCOMPAT features");
2109 return;
2110 }
2111
617ba13b 2112 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2113 if (es->s_last_orphan)
2114 jbd_debug(1, "Errors on filesystem, "
2115 "clearing orphan list.\n");
2116 es->s_last_orphan = 0;
2117 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2118 return;
2119 }
2120
2121 if (s_flags & MS_RDONLY) {
b31e1552 2122 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2123 sb->s_flags &= ~MS_RDONLY;
2124 }
2125#ifdef CONFIG_QUOTA
2126 /* Needed for iput() to work correctly and not trash data */
2127 sb->s_flags |= MS_ACTIVE;
2128 /* Turn on quotas so that they are updated correctly */
2129 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2130 if (EXT4_SB(sb)->s_qf_names[i]) {
2131 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2132 if (ret < 0)
b31e1552
ES		 2133 ext4_msg(sb, KERN_ERR,
2134 "Cannot turn on journaled "
2135 "quota: error %d", ret);
ac27a0ec
DK		 2136 }
2137 }
2138#endif
2139
2140 while (es->s_last_orphan) {
2141 struct inode *inode;
2142
97bd42b9
JB		 2143 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2144 if (IS_ERR(inode)) {
ac27a0ec
DK		 2145 es->s_last_orphan = 0;
2146 break;
2147 }
2148
617ba13b 2149 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2150 dquot_initialize(inode);
ac27a0ec 2151 if (inode->i_nlink) {
b31e1552
ES		 2152 ext4_msg(sb, KERN_DEBUG,
2153 "%s: truncating inode %lu to %lld bytes",
46e665e9 2154 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2155 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2156 inode->i_ino, inode->i_size);
617ba13b 2157 ext4_truncate(inode);
ac27a0ec
DK		 2158 nr_truncates++;
2159 } else {
b31e1552
ES		 2160 ext4_msg(sb, KERN_DEBUG,
2161 "%s: deleting unreferenced inode %lu",
46e665e9 2162 __func__, inode->i_ino);
ac27a0ec
DK		 2163 jbd_debug(2, "deleting unreferenced inode %lu\n",
2164 inode->i_ino);
2165 nr_orphans++;
2166 }
2167 iput(inode); /* The delete magic happens here! */
2168 }
2169
2b2d6d01 2170#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2171
2172 if (nr_orphans)
b31e1552
ES		 2173 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2174 PLURAL(nr_orphans));
ac27a0ec 2175 if (nr_truncates)
b31e1552
ES		 2176 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2177 PLURAL(nr_truncates));
ac27a0ec
DK		 2178#ifdef CONFIG_QUOTA
2179 /* Turn quotas off */
2180 for (i = 0; i < MAXQUOTAS; i++) {
2181 if (sb_dqopt(sb)->files[i])
287a8095 2182 dquot_quota_off(sb, i);
ac27a0ec
DK		 2183 }
2184#endif
2185 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2186}
0b8e58a1 2187
cd2291a4
ES		 2188/*
2189 * Maximal extent format file size.
2190 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2191 * extent format containers, within a sector_t, and within i_blocks
2192 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2193 * so that won't be a limiting factor.
2194 *
f17722f9
LC		 2195 * However there is other limiting factor. We do store extents in the form
2196 * of starting block and length, hence the resulting length of the extent
2197 * covering maximum file size must fit into on-disk format containers as
2198 * well. Given that length is always by 1 unit bigger than max unit (because
2199 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2200 *
cd2291a4
ES		 2201 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2202 */
f287a1a5 2203static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2204{
2205 loff_t res;
2206 loff_t upper_limit = MAX_LFS_FILESIZE;
2207
2208 /* small i_blocks in vfs inode? */
f287a1a5 2209 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2210 /*
90c699a9 2211 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2212 * i_block represent total blocks in 512 bytes
2213 * 32 == size of vfs inode i_blocks * 8
2214 */
2215 upper_limit = (1LL << 32) - 1;
2216
2217 /* total blocks in file system block size */
2218 upper_limit >>= (blkbits - 9);
2219 upper_limit <<= blkbits;
2220 }
2221
f17722f9
LC		 2222 /*
2223 * 32-bit extent-start container, ee_block. We lower the maxbytes
2224 * by one fs block, so ee_len can cover the extent of maximum file
2225 * size
2226 */
2227 res = (1LL << 32) - 1;
cd2291a4 2228 res <<= blkbits;
cd2291a4
ES		 2229
2230 /* Sanity check against vm- & vfs- imposed limits */
2231 if (res > upper_limit)
2232 res = upper_limit;
2233
2234 return res;
2235}
ac27a0ec 2236
ac27a0ec 2237/*
cd2291a4 2238 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2239 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2240 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2241 */
f287a1a5 2242static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2243{
617ba13b 2244 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2245 int meta_blocks;
2246 loff_t upper_limit;
0b8e58a1
AD		 2247 /* This is calculated to be the largest file size for a dense, block
2248 * mapped file such that the file's total number of 512-byte sectors,
2249 * including data and all indirect blocks, does not exceed (2^48 - 1).
2250 *
2251 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2252 * number of 512-byte sectors of the file.
0fc1b451
AK		 2253 */
2254
f287a1a5 2255 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2256 /*
90c699a9 2257 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2258 * the inode i_block field represents total file blocks in
2259 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2260 */
2261 upper_limit = (1LL << 32) - 1;
2262
2263 /* total blocks in file system block size */
2264 upper_limit >>= (bits - 9);
2265
2266 } else {
8180a562
AK		 2267 /*
2268 * We use 48 bit ext4_inode i_blocks
2269 * With EXT4_HUGE_FILE_FL set the i_blocks
2270 * represent total number of blocks in
2271 * file system block size
2272 */
0fc1b451
AK		 2273 upper_limit = (1LL << 48) - 1;
2274
0fc1b451
AK		 2275 }
2276
2277 /* indirect blocks */
2278 meta_blocks = 1;
2279 /* double indirect blocks */
2280 meta_blocks += 1 + (1LL << (bits-2));
2281 /* tripple indirect blocks */
2282 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2283
2284 upper_limit -= meta_blocks;
2285 upper_limit <<= bits;
ac27a0ec
DK		 2286
2287 res += 1LL << (bits-2);
2288 res += 1LL << (2*(bits-2));
2289 res += 1LL << (3*(bits-2));
2290 res <<= bits;
2291 if (res > upper_limit)
2292 res = upper_limit;
0fc1b451
AK		 2293
2294 if (res > MAX_LFS_FILESIZE)
2295 res = MAX_LFS_FILESIZE;
2296
ac27a0ec
DK		 2297 return res;
2298}
2299
617ba13b 2300static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2301 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2302{
617ba13b 2303 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2304 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2305 int has_super = 0;
2306
2307 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2308
617ba13b 2309 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2310 nr < first_meta_bg)
70bbb3e0 2311 return logical_sb_block + nr + 1;
ac27a0ec 2312 bg = sbi->s_desc_per_block * nr;
617ba13b 2313 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2314 has_super = 1;
0b8e58a1 2315
617ba13b 2316 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2317}
2318
c9de560d
AT		 2319/**
2320 * ext4_get_stripe_size: Get the stripe size.
2321 * @sbi: In memory super block info
2322 *
2323 * If we have specified it via mount option, then
2324 * use the mount option value. If the value specified at mount time is
2325 * greater than the blocks per group use the super block value.
2326 * If the super block value is greater than blocks per group return 0.
2327 * Allocator needs it be less than blocks per group.
2328 *
2329 */
2330static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2331{
2332 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2333 unsigned long stripe_width =
2334 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2335 int ret;
c9de560d
AT		 2336
2337 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2338 ret = sbi->s_stripe;
2339 else if (stripe_width <= sbi->s_blocks_per_group)
2340 ret = stripe_width;
2341 else if (stride <= sbi->s_blocks_per_group)
2342 ret = stride;
2343 else
2344 ret = 0;
c9de560d 2345
3eb08658
DE		 2346 /*
2347 * If the stripe width is 1, this makes no sense and
2348 * we set it to 0 to turn off stripe handling code.
2349 */
2350 if (ret <= 1)
2351 ret = 0;
c9de560d 2352
3eb08658 2353 return ret;
c9de560d 2354}
ac27a0ec 2355
3197ebdb
TT		 2356/* sysfs supprt */
2357
2358struct ext4_attr {
2359 struct attribute attr;
2360 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2361 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2362 const char *, size_t);
2363 int offset;
2364};
2365
2366static int parse_strtoul(const char *buf,
2367 unsigned long max, unsigned long *value)
2368{
2369 char *endp;
2370
e7d2860b
AGR		 2371 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2372 endp = skip_spaces(endp);
3197ebdb
TT		 2373 if (*endp || *value > max)
2374 return -EINVAL;
2375
2376 return 0;
2377}
2378
2379static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2380 struct ext4_sb_info *sbi,
2381 char *buf)
2382{
2383 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2384 (s64) EXT4_C2B(sbi,
2385 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2386}
2387
2388static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2389 struct ext4_sb_info *sbi, char *buf)
2390{
2391 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2392
f613dfcb
TT		 2393 if (!sb->s_bdev->bd_part)
2394 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2395 return snprintf(buf, PAGE_SIZE, "%lu\n",
2396 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2397 sbi->s_sectors_written_start) >> 1);
2398}
2399
2400static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2401 struct ext4_sb_info *sbi, char *buf)
2402{
2403 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2404
f613dfcb
TT		 2405 if (!sb->s_bdev->bd_part)
2406 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2407 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2408 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2409 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2410 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2411}
2412
2413static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2414 struct ext4_sb_info *sbi,
2415 const char *buf, size_t count)
2416{
2417 unsigned long t;
2418
2419 if (parse_strtoul(buf, 0x40000000, &t))
2420 return -EINVAL;
2421
5dbd571d 2422 if (t && !is_power_of_2(t))
3197ebdb
TT		 2423 return -EINVAL;
2424
2425 sbi->s_inode_readahead_blks = t;
2426 return count;
2427}
2428
2429static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2430 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2431{
2432 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2433
2434 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2435}
2436
2437static ssize_t sbi_ui_store(struct ext4_attr *a,
2438 struct ext4_sb_info *sbi,
2439 const char *buf, size_t count)
2440{
2441 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2442 unsigned long t;
2443
2444 if (parse_strtoul(buf, 0xffffffff, &t))
2445 return -EINVAL;
2446 *ui = t;
2447 return count;
2448}
2449
2450#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2451static struct ext4_attr ext4_attr_##_name = { \
2452 .attr = {.name = __stringify(_name), .mode = _mode }, \
2453 .show = _show, \
2454 .store = _store, \
2455 .offset = offsetof(struct ext4_sb_info, _elname), \
2456}
2457#define EXT4_ATTR(name, mode, show, store) \
2458static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2459
857ac889 2460#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2461#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2462#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2463#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2464 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2465#define ATTR_LIST(name) &ext4_attr_##name.attr
2466
2467EXT4_RO_ATTR(delayed_allocation_blocks);
2468EXT4_RO_ATTR(session_write_kbytes);
2469EXT4_RO_ATTR(lifetime_write_kbytes);
2470EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2471 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2472EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2473EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2474EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2475EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2476EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2477EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2478EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2479EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2480
2481static struct attribute *ext4_attrs[] = {
2482 ATTR_LIST(delayed_allocation_blocks),
2483 ATTR_LIST(session_write_kbytes),
2484 ATTR_LIST(lifetime_write_kbytes),
2485 ATTR_LIST(inode_readahead_blks),
11013911 2486 ATTR_LIST(inode_goal),
3197ebdb
TT		 2487 ATTR_LIST(mb_stats),
2488 ATTR_LIST(mb_max_to_scan),
2489 ATTR_LIST(mb_min_to_scan),
2490 ATTR_LIST(mb_order2_req),
2491 ATTR_LIST(mb_stream_req),
2492 ATTR_LIST(mb_group_prealloc),
55138e0b 2493 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2494 NULL,
2495};
2496
857ac889
LC		 2497/* Features this copy of ext4 supports */
2498EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2499EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2500
2501static struct attribute *ext4_feat_attrs[] = {
2502 ATTR_LIST(lazy_itable_init),
27ee40df 2503 ATTR_LIST(batched_discard),
857ac889
LC		 2504 NULL,
2505};
2506
3197ebdb
TT		 2507static ssize_t ext4_attr_show(struct kobject *kobj,
2508 struct attribute *attr, char *buf)
2509{
2510 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2511 s_kobj);
2512 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2513
2514 return a->show ? a->show(a, sbi, buf) : 0;
2515}
2516
2517static ssize_t ext4_attr_store(struct kobject *kobj,
2518 struct attribute *attr,
2519 const char *buf, size_t len)
2520{
2521 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2522 s_kobj);
2523 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2524
2525 return a->store ? a->store(a, sbi, buf, len) : 0;
2526}
2527
2528static void ext4_sb_release(struct kobject *kobj)
2529{
2530 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2531 s_kobj);
2532 complete(&sbi->s_kobj_unregister);
2533}
2534
52cf25d0 2535static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2536 .show = ext4_attr_show,
2537 .store = ext4_attr_store,
2538};
2539
2540static struct kobj_type ext4_ktype = {
2541 .default_attrs = ext4_attrs,
2542 .sysfs_ops = &ext4_attr_ops,
2543 .release = ext4_sb_release,
2544};
2545
857ac889
LC		 2546static void ext4_feat_release(struct kobject *kobj)
2547{
2548 complete(&ext4_feat->f_kobj_unregister);
2549}
2550
2551static struct kobj_type ext4_feat_ktype = {
2552 .default_attrs = ext4_feat_attrs,
2553 .sysfs_ops = &ext4_attr_ops,
2554 .release = ext4_feat_release,
2555};
2556
a13fb1a4
ES		 2557/*
2558 * Check whether this filesystem can be mounted based on
2559 * the features present and the RDONLY/RDWR mount requested.
2560 * Returns 1 if this filesystem can be mounted as requested,
2561 * 0 if it cannot be.
2562 */
2563static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2564{
2565 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2566 ext4_msg(sb, KERN_ERR,
2567 "Couldn't mount because of "
2568 "unsupported optional features (%x)",
2569 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2570 ~EXT4_FEATURE_INCOMPAT_SUPP));
2571 return 0;
2572 }
2573
2574 if (readonly)
2575 return 1;
2576
2577 /* Check that feature set is OK for a read-write mount */
2578 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2579 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2580 "unsupported optional features (%x)",
2581 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2582 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2583 return 0;
2584 }
2585 /*
2586 * Large file size enabled file system can only be mounted
2587 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2588 */
2589 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2590 if (sizeof(blkcnt_t) < sizeof(u64)) {
2591 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2592 "cannot be mounted RDWR without "
2593 "CONFIG_LBDAF");
2594 return 0;
2595 }
2596 }
bab08ab9
TT		 2597 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2598 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2599 ext4_msg(sb, KERN_ERR,
2600 "Can't support bigalloc feature without "
2601 "extents feature\n");
2602 return 0;
2603 }
a13fb1a4
ES		 2604 return 1;
2605}
2606
66e61a9e
TT		 2607/*
2608 * This function is called once a day if we have errors logged
2609 * on the file system
2610 */
2611static void print_daily_error_info(unsigned long arg)
2612{
2613 struct super_block *sb = (struct super_block *) arg;
2614 struct ext4_sb_info *sbi;
2615 struct ext4_super_block *es;
2616
2617 sbi = EXT4_SB(sb);
2618 es = sbi->s_es;
2619
2620 if (es->s_error_count)
2621 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2622 le32_to_cpu(es->s_error_count));
2623 if (es->s_first_error_time) {
2624 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2625 sb->s_id, le32_to_cpu(es->s_first_error_time),
2626 (int) sizeof(es->s_first_error_func),
2627 es->s_first_error_func,
2628 le32_to_cpu(es->s_first_error_line));
2629 if (es->s_first_error_ino)
2630 printk(": inode %u",
2631 le32_to_cpu(es->s_first_error_ino));
2632 if (es->s_first_error_block)
2633 printk(": block %llu", (unsigned long long)
2634 le64_to_cpu(es->s_first_error_block));
2635 printk("\n");
2636 }
2637 if (es->s_last_error_time) {
2638 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2639 sb->s_id, le32_to_cpu(es->s_last_error_time),
2640 (int) sizeof(es->s_last_error_func),
2641 es->s_last_error_func,
2642 le32_to_cpu(es->s_last_error_line));
2643 if (es->s_last_error_ino)
2644 printk(": inode %u",
2645 le32_to_cpu(es->s_last_error_ino));
2646 if (es->s_last_error_block)
2647 printk(": block %llu", (unsigned long long)
2648 le64_to_cpu(es->s_last_error_block));
2649 printk("\n");
2650 }
2651 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2652}
2653
bfff6873
LC		 2654/* Find next suitable group and run ext4_init_inode_table */
2655static int ext4_run_li_request(struct ext4_li_request *elr)
2656{
2657 struct ext4_group_desc *gdp = NULL;
2658 ext4_group_t group, ngroups;
2659 struct super_block *sb;
2660 unsigned long timeout = 0;
2661 int ret = 0;
2662
2663 sb = elr->lr_super;
2664 ngroups = EXT4_SB(sb)->s_groups_count;
2665
2666 for (group = elr->lr_next_group; group < ngroups; group++) {
2667 gdp = ext4_get_group_desc(sb, group, NULL);
2668 if (!gdp) {
2669 ret = 1;
2670 break;
2671 }
2672
2673 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2674 break;
2675 }
2676
2677 if (group == ngroups)
2678 ret = 1;
2679
2680 if (!ret) {
2681 timeout = jiffies;
2682 ret = ext4_init_inode_table(sb, group,
2683 elr->lr_timeout ? 0 : 1);
2684 if (elr->lr_timeout == 0) {
51ce6511
LC		 2685 timeout = (jiffies - timeout) *
2686 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2687 elr->lr_timeout = timeout;
2688 }
2689 elr->lr_next_sched = jiffies + elr->lr_timeout;
2690 elr->lr_next_group = group + 1;
2691 }
2692
2693 return ret;
2694}
2695
2696/*
2697 * Remove lr_request from the list_request and free the
4ed5c033 2698 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2699 */
2700static void ext4_remove_li_request(struct ext4_li_request *elr)
2701{
2702 struct ext4_sb_info *sbi;
2703
2704 if (!elr)
2705 return;
2706
2707 sbi = elr->lr_sbi;
2708
2709 list_del(&elr->lr_request);
2710 sbi->s_li_request = NULL;
2711 kfree(elr);
2712}
2713
2714static void ext4_unregister_li_request(struct super_block *sb)
2715{
1bb933fb
LC		 2716 mutex_lock(&ext4_li_mtx);
2717 if (!ext4_li_info) {
2718 mutex_unlock(&ext4_li_mtx);
bfff6873 2719 return;
1bb933fb 2720 }
bfff6873
LC		 2721
2722 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2723 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2724 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2725 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2726}
2727
8f1f7453
ES		 2728static struct task_struct *ext4_lazyinit_task;
2729
bfff6873
LC		 2730/*
2731 * This is the function where ext4lazyinit thread lives. It walks
2732 * through the request list searching for next scheduled filesystem.
2733 * When such a fs is found, run the lazy initialization request
2734 * (ext4_rn_li_request) and keep track of the time spend in this
2735 * function. Based on that time we compute next schedule time of
2736 * the request. When walking through the list is complete, compute
2737 * next waking time and put itself into sleep.
2738 */
2739static int ext4_lazyinit_thread(void *arg)
2740{
2741 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2742 struct list_head *pos, *n;
2743 struct ext4_li_request *elr;
4ed5c033 2744 unsigned long next_wakeup, cur;
bfff6873
LC		 2745
2746 BUG_ON(NULL == eli);
2747
bfff6873
LC		 2748cont_thread:
2749 while (true) {
2750 next_wakeup = MAX_JIFFY_OFFSET;
2751
2752 mutex_lock(&eli->li_list_mtx);
2753 if (list_empty(&eli->li_request_list)) {
2754 mutex_unlock(&eli->li_list_mtx);
2755 goto exit_thread;
2756 }
2757
2758 list_for_each_safe(pos, n, &eli->li_request_list) {
2759 elr = list_entry(pos, struct ext4_li_request,
2760 lr_request);
2761
b2c78cd0
TT		 2762 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2763 if (ext4_run_li_request(elr) != 0) {
2764 /* error, remove the lazy_init job */
2765 ext4_remove_li_request(elr);
2766 continue;
2767 }
bfff6873
LC		 2768 }
2769
2770 if (time_before(elr->lr_next_sched, next_wakeup))
2771 next_wakeup = elr->lr_next_sched;
2772 }
2773 mutex_unlock(&eli->li_list_mtx);
2774
a0acae0e 2775 try_to_freeze();
bfff6873 2776
4ed5c033
LC		 2777 cur = jiffies;
2778 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2779 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2780 cond_resched();
2781 continue;
2782 }
2783
4ed5c033
LC		 2784 schedule_timeout_interruptible(next_wakeup - cur);
2785
8f1f7453
ES		 2786 if (kthread_should_stop()) {
2787 ext4_clear_request_list();
2788 goto exit_thread;
2789 }
bfff6873
LC		 2790 }
2791
2792exit_thread:
2793 /*
2794 * It looks like the request list is empty, but we need
2795 * to check it under the li_list_mtx lock, to prevent any
2796 * additions into it, and of course we should lock ext4_li_mtx
2797 * to atomically free the list and ext4_li_info, because at
2798 * this point another ext4 filesystem could be registering
2799 * new one.
2800 */
2801 mutex_lock(&ext4_li_mtx);
2802 mutex_lock(&eli->li_list_mtx);
2803 if (!list_empty(&eli->li_request_list)) {
2804 mutex_unlock(&eli->li_list_mtx);
2805 mutex_unlock(&ext4_li_mtx);
2806 goto cont_thread;
2807 }
2808 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2809 kfree(ext4_li_info);
2810 ext4_li_info = NULL;
2811 mutex_unlock(&ext4_li_mtx);
2812
2813 return 0;
2814}
2815
2816static void ext4_clear_request_list(void)
2817{
2818 struct list_head *pos, *n;
2819 struct ext4_li_request *elr;
2820
2821 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2822 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2823 elr = list_entry(pos, struct ext4_li_request,
2824 lr_request);
2825 ext4_remove_li_request(elr);
2826 }
2827 mutex_unlock(&ext4_li_info->li_list_mtx);
2828}
2829
2830static int ext4_run_lazyinit_thread(void)
2831{
8f1f7453
ES		 2832 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2833 ext4_li_info, "ext4lazyinit");
2834 if (IS_ERR(ext4_lazyinit_task)) {
2835 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2836 ext4_clear_request_list();
bfff6873
LC		 2837 kfree(ext4_li_info);
2838 ext4_li_info = NULL;
92b97816 2839 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2840 "initialization thread\n",
2841 err);
2842 return err;
2843 }
2844 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2845 return 0;
2846}
2847
2848/*
2849 * Check whether it make sense to run itable init. thread or not.
2850 * If there is at least one uninitialized inode table, return
2851 * corresponding group number, else the loop goes through all
2852 * groups and return total number of groups.
2853 */
2854static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2855{
2856 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2857 struct ext4_group_desc *gdp = NULL;
2858
2859 for (group = 0; group < ngroups; group++) {
2860 gdp = ext4_get_group_desc(sb, group, NULL);
2861 if (!gdp)
2862 continue;
2863
2864 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2865 break;
2866 }
2867
2868 return group;
2869}
2870
2871static int ext4_li_info_new(void)
2872{
2873 struct ext4_lazy_init *eli = NULL;
2874
2875 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2876 if (!eli)
2877 return -ENOMEM;
2878
bfff6873
LC		 2879 INIT_LIST_HEAD(&eli->li_request_list);
2880 mutex_init(&eli->li_list_mtx);
2881
bfff6873
LC		 2882 eli->li_state |= EXT4_LAZYINIT_QUIT;
2883
2884 ext4_li_info = eli;
2885
2886 return 0;
2887}
2888
2889static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2890 ext4_group_t start)
2891{
2892 struct ext4_sb_info *sbi = EXT4_SB(sb);
2893 struct ext4_li_request *elr;
2894 unsigned long rnd;
2895
2896 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2897 if (!elr)
2898 return NULL;
2899
2900 elr->lr_super = sb;
2901 elr->lr_sbi = sbi;
2902 elr->lr_next_group = start;
2903
2904 /*
2905 * Randomize first schedule time of the request to
2906 * spread the inode table initialization requests
2907 * better.
2908 */
2909 get_random_bytes(&rnd, sizeof(rnd));
2910 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2911 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2912
2913 return elr;
2914}
2915
2916static int ext4_register_li_request(struct super_block *sb,
2917 ext4_group_t first_not_zeroed)
2918{
2919 struct ext4_sb_info *sbi = EXT4_SB(sb);
2920 struct ext4_li_request *elr;
2921 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2922 int ret = 0;
bfff6873 2923
51ce6511
LC		 2924 if (sbi->s_li_request != NULL) {
2925 /*
2926 * Reset timeout so it can be computed again, because
2927 * s_li_wait_mult might have changed.
2928 */
2929 sbi->s_li_request->lr_timeout = 0;
beed5ecb 2930 return 0;
51ce6511 2931 }
bfff6873
LC		 2932
2933 if (first_not_zeroed == ngroups ||
2934 (sb->s_flags & MS_RDONLY) ||
55ff3840 2935 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 2936 return 0;
bfff6873
LC		 2937
2938 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 2939 if (!elr)
2940 return -ENOMEM;
bfff6873
LC		 2941
2942 mutex_lock(&ext4_li_mtx);
2943
2944 if (NULL == ext4_li_info) {
2945 ret = ext4_li_info_new();
2946 if (ret)
2947 goto out;
2948 }
2949
2950 mutex_lock(&ext4_li_info->li_list_mtx);
2951 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2952 mutex_unlock(&ext4_li_info->li_list_mtx);
2953
2954 sbi->s_li_request = elr;
46e4690b
TM		 2955 /*
2956 * set elr to NULL here since it has been inserted to
2957 * the request_list and the removal and free of it is
2958 * handled by ext4_clear_request_list from now on.
2959 */
2960 elr = NULL;
bfff6873
LC		 2961
2962 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2963 ret = ext4_run_lazyinit_thread();
2964 if (ret)
2965 goto out;
2966 }
bfff6873 2967out:
beed5ecb
NK		 2968 mutex_unlock(&ext4_li_mtx);
2969 if (ret)
bfff6873 2970 kfree(elr);
bfff6873
LC		 2971 return ret;
2972}
2973
2974/*
2975 * We do not need to lock anything since this is called on
2976 * module unload.
2977 */
2978static void ext4_destroy_lazyinit_thread(void)
2979{
2980 /*
2981 * If thread exited earlier
2982 * there's nothing to be done.
2983 */
8f1f7453 2984 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2985 return;
2986
8f1f7453 2987 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2988}
2989
2b2d6d01 2990static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 2991{
d4c402d9 2992 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 2993 struct buffer_head *bh;
617ba13b
MC		 2994 struct ext4_super_block *es = NULL;
2995 struct ext4_sb_info *sbi;
2996 ext4_fsblk_t block;
2997 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2998 ext4_fsblk_t logical_sb_block;
ac27a0ec 2999 unsigned long offset = 0;
ac27a0ec
DK		 3000 unsigned long journal_devnum = 0;
3001 unsigned long def_mount_opts;
3002 struct inode *root;
9f6200bb 3003 char *cp;
0390131b 3004 const char *descr;
dcc7dae3 3005 int ret = -ENOMEM;
281b5995 3006 int blocksize, clustersize;
4ec11028
TT		 3007 unsigned int db_count;
3008 unsigned int i;
281b5995 3009 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3010 __u64 blocks_count;
833f4077 3011 int err;
b3881f74 3012 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3013 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3014
3015 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3016 if (!sbi)
dcc7dae3 3017 goto out_free_orig;
705895b6
PE		 3018
3019 sbi->s_blockgroup_lock =
3020 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3021 if (!sbi->s_blockgroup_lock) {
3022 kfree(sbi);
dcc7dae3 3023 goto out_free_orig;
705895b6 3024 }
ac27a0ec
DK		 3025 sb->s_fs_info = sbi;
3026 sbi->s_mount_opt = 0;
617ba13b
MC		 3027 sbi->s_resuid = EXT4_DEF_RESUID;
3028 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 3029 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3030 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3031 if (sb->s_bdev->bd_part)
3032 sbi->s_sectors_written_start =
3033 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3034
9f6200bb
TT		 3035 /* Cleanup superblock name */
3036 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3037 *cp = '!';
3038
dcc7dae3 3039 ret = -EINVAL;
617ba13b 3040 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3041 if (!blocksize) {
b31e1552 3042 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3043 goto out_fail;
3044 }
3045
3046 /*
617ba13b 3047 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3048 * block sizes. We need to calculate the offset from buffer start.
3049 */
617ba13b 3050 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3051 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3052 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3053 } else {
70bbb3e0 3054 logical_sb_block = sb_block;
ac27a0ec
DK		 3055 }
3056
70bbb3e0 3057 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3058 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3059 goto out_fail;
3060 }
3061 /*
3062 * Note: s_es must be initialized as soon as possible because
617ba13b 3063 * some ext4 macro-instructions depend on its value
ac27a0ec 3064 */
617ba13b 3065 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3066 sbi->s_es = es;
3067 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3068 if (sb->s_magic != EXT4_SUPER_MAGIC)
3069 goto cantfind_ext4;
afc32f7e 3070 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3071
d25425f8
DW		 3072 /* Check for a known checksum algorithm */
3073 if (!ext4_verify_csum_type(sb, es)) {
3074 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3075 "unknown checksum algorithm.");
3076 silent = 1;
3077 goto cantfind_ext4;
3078 }
3079
0441984a
DW		 3080 /* Load the checksum driver */
3081 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3082 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3083 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3084 if (IS_ERR(sbi->s_chksum_driver)) {
3085 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3086 ret = PTR_ERR(sbi->s_chksum_driver);
3087 sbi->s_chksum_driver = NULL;
3088 goto failed_mount;
3089 }
3090 }
3091
a9c47317
DW		 3092 /* Check superblock checksum */
3093 if (!ext4_superblock_csum_verify(sb, es)) {
3094 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3095 "invalid superblock checksum. Run e2fsck?");
3096 silent = 1;
3097 goto cantfind_ext4;
3098 }
3099
3100 /* Precompute checksum seed for all metadata */
3101 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3102 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3103 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3104 sizeof(es->s_uuid));
3105
ac27a0ec
DK		 3106 /* Set defaults before we parse the mount options */
3107 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3108 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3109 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3110 set_opt(sb, DEBUG);
87f26807 3111 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3112 set_opt(sb, GRPID);
617ba13b 3113 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3114 set_opt(sb, NO_UID32);
ea663336 3115 /* xattr user namespace & acls are now defaulted on */
03010a33 3116#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3117 set_opt(sb, XATTR_USER);
2e7842b8 3118#endif
03010a33 3119#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3120 set_opt(sb, POSIX_ACL);
2e7842b8 3121#endif
6fd7a467 3122 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3123 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3124 set_opt(sb, JOURNAL_DATA);
617ba13b 3125 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3126 set_opt(sb, ORDERED_DATA);
617ba13b 3127 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3128 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3129
3130 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3131 set_opt(sb, ERRORS_PANIC);
bb4f397a 3132 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3133 set_opt(sb, ERRORS_CONT);
bb4f397a 3134 else
fd8c37ec 3135 set_opt(sb, ERRORS_RO);
8b67f04a 3136 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3137 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3138 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3139 set_opt(sb, DISCARD);
ac27a0ec
DK		 3140
3141 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3142 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3143 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3144 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3145 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3146
8b67f04a 3147 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3148 set_opt(sb, BARRIER);
ac27a0ec 3149
dd919b98
AK		 3150 /*
3151 * enable delayed allocation by default
3152 * Use -o nodelalloc to turn it off
3153 */
8b67f04a
TT		 3154 if (!IS_EXT3_SB(sb) &&
3155 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3156 set_opt(sb, DELALLOC);
dd919b98 3157
51ce6511
LC		 3158 /*
3159 * set default s_li_wait_mult for lazyinit, for the case there is
3160 * no mount option specified.
3161 */
3162 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3163
8b67f04a 3164 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3165 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3166 ext4_msg(sb, KERN_WARNING,
3167 "failed to parse options in superblock: %s",
3168 sbi->s_es->s_mount_opts);
3169 }
5a916be1 3170 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3171 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3172 &journal_ioprio, 0))
ac27a0ec
DK		 3173 goto failed_mount;
3174
56889787
TT		 3175 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3176 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3177 "with data=journal disables delayed "
3178 "allocation and O_DIRECT support!\n");
3179 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3180 ext4_msg(sb, KERN_ERR, "can't mount with "
3181 "both data=journal and delalloc");
3182 goto failed_mount;
3183 }
3184 if (test_opt(sb, DIOREAD_NOLOCK)) {
3185 ext4_msg(sb, KERN_ERR, "can't mount with "
3186 "both data=journal and delalloc");
3187 goto failed_mount;
3188 }
3189 if (test_opt(sb, DELALLOC))
3190 clear_opt(sb, DELALLOC);
3191 }
3192
3193 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3194 if (test_opt(sb, DIOREAD_NOLOCK)) {
3195 if (blocksize < PAGE_SIZE) {
3196 ext4_msg(sb, KERN_ERR, "can't mount with "
3197 "dioread_nolock if block size != PAGE_SIZE");
3198 goto failed_mount;
3199 }
3200 }
3201
ac27a0ec 3202 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3203 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3204
617ba13b
MC		 3205 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3206 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3207 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3208 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3209 ext4_msg(sb, KERN_WARNING,
3210 "feature flags set on rev 0 fs, "
3211 "running e2fsck is recommended");
469108ff 3212
2035e776
TT		 3213 if (IS_EXT2_SB(sb)) {
3214 if (ext2_feature_set_ok(sb))
3215 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3216 "using the ext4 subsystem");
3217 else {
3218 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3219 "to feature incompatibilities");
3220 goto failed_mount;
3221 }
3222 }
3223
3224 if (IS_EXT3_SB(sb)) {
3225 if (ext3_feature_set_ok(sb))
3226 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3227 "using the ext4 subsystem");
3228 else {
3229 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3230 "to feature incompatibilities");
3231 goto failed_mount;
3232 }
3233 }
3234
ac27a0ec
DK		 3235 /*
3236 * Check feature flags regardless of the revision level, since we
3237 * previously didn't change the revision level when setting the flags,
3238 * so there is a chance incompat flags are set on a rev 0 filesystem.
3239 */
a13fb1a4 3240 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3241 goto failed_mount;
a13fb1a4 3242
617ba13b
MC		 3243 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3244 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3245 ext4_msg(sb, KERN_ERR,
3246 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3247 goto failed_mount;
3248 }
3249
ac27a0ec 3250 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3251 /* Validate the filesystem blocksize */
3252 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3253 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3254 blocksize);
ac27a0ec
DK		 3255 goto failed_mount;
3256 }
3257
2b2d6d01 3258 brelse(bh);
70bbb3e0
AM		 3259 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3260 offset = do_div(logical_sb_block, blocksize);
3261 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3262 if (!bh) {
b31e1552
ES		 3263 ext4_msg(sb, KERN_ERR,
3264 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3265 goto failed_mount;
3266 }
617ba13b 3267 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3268 sbi->s_es = es;
617ba13b 3269 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3270 ext4_msg(sb, KERN_ERR,
3271 "Magic mismatch, very weird!");
ac27a0ec
DK		 3272 goto failed_mount;
3273 }
3274 }
3275
a13fb1a4
ES		 3276 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3277 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3278 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3279 has_huge_files);
3280 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3281
617ba13b
MC		 3282 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3283 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3284 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3285 } else {
3286 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3287 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3288 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3289 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3290 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3291 ext4_msg(sb, KERN_ERR,
3292 "unsupported inode size: %d",
2b2d6d01 3293 sbi->s_inode_size);
ac27a0ec
DK		 3294 goto failed_mount;
3295 }
ef7f3835
KS		 3296 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3297 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3298 }
0b8e58a1 3299
0d1ee42f
AR		 3300 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3301 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3302 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3303 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3304 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3305 ext4_msg(sb, KERN_ERR,
3306 "unsupported descriptor size %lu",
0d1ee42f
AR		 3307 sbi->s_desc_size);
3308 goto failed_mount;
3309 }
3310 } else
3311 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3312
ac27a0ec 3313 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3314 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3315 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3316 goto cantfind_ext4;
0b8e58a1 3317
617ba13b 3318 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3319 if (sbi->s_inodes_per_block == 0)
617ba13b 3320 goto cantfind_ext4;
ac27a0ec
DK		 3321 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3322 sbi->s_inodes_per_block;
0d1ee42f 3323 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3324 sbi->s_sbh = bh;
3325 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3326 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3327 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3328
2b2d6d01 3329 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3330 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3331 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3332 i = le32_to_cpu(es->s_flags);
3333 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3334 sbi->s_hash_unsigned = 3;
3335 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3336#ifdef __CHAR_UNSIGNED__
3337 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3338 sbi->s_hash_unsigned = 3;
3339#else
3340 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3341#endif
f99b2589 3342 }
ac27a0ec 3343
281b5995
TT		 3344 /* Handle clustersize */
3345 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3346 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3347 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3348 if (has_bigalloc) {
3349 if (clustersize < blocksize) {
3350 ext4_msg(sb, KERN_ERR,
3351 "cluster size (%d) smaller than "
3352 "block size (%d)", clustersize, blocksize);
3353 goto failed_mount;
3354 }
3355 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3356 le32_to_cpu(es->s_log_block_size);
3357 sbi->s_clusters_per_group =
3358 le32_to_cpu(es->s_clusters_per_group);
3359 if (sbi->s_clusters_per_group > blocksize * 8) {
3360 ext4_msg(sb, KERN_ERR,
3361 "#clusters per group too big: %lu",
3362 sbi->s_clusters_per_group);
3363 goto failed_mount;
3364 }
3365 if (sbi->s_blocks_per_group !=
3366 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3367 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3368 "clusters per group (%lu) inconsistent",
3369 sbi->s_blocks_per_group,
3370 sbi->s_clusters_per_group);
3371 goto failed_mount;
3372 }
3373 } else {
3374 if (clustersize != blocksize) {
3375 ext4_warning(sb, "fragment/cluster size (%d) != "
3376 "block size (%d)", clustersize,
3377 blocksize);
3378 clustersize = blocksize;
3379 }
3380 if (sbi->s_blocks_per_group > blocksize * 8) {
3381 ext4_msg(sb, KERN_ERR,
3382 "#blocks per group too big: %lu",
3383 sbi->s_blocks_per_group);
3384 goto failed_mount;
3385 }
3386 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3387 sbi->s_cluster_bits = 0;
ac27a0ec 3388 }
281b5995
TT		 3389 sbi->s_cluster_ratio = clustersize / blocksize;
3390
ac27a0ec 3391 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3392 ext4_msg(sb, KERN_ERR,
3393 "#inodes per group too big: %lu",
2b2d6d01 3394 sbi->s_inodes_per_group);
ac27a0ec
DK		 3395 goto failed_mount;
3396 }
3397
bf43d84b
ES		 3398 /*
3399 * Test whether we have more sectors than will fit in sector_t,
3400 * and whether the max offset is addressable by the page cache.
3401 */
5a9ae68a 3402 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3403 ext4_blocks_count(es));
5a9ae68a 3404 if (err) {
b31e1552 3405 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3406 " too large to mount safely on this system");
ac27a0ec 3407 if (sizeof(sector_t) < 8)
90c699a9 3408 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3409 ret = err;
ac27a0ec
DK		 3410 goto failed_mount;
3411 }
3412
617ba13b
MC		 3413 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3414 goto cantfind_ext4;
e7c95593 3415
0f2ddca6
FTN		 3416 /* check blocks count against device size */
3417 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3418 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3419 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3420 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3421 ext4_blocks_count(es), blocks_count);
3422 goto failed_mount;
3423 }
3424
0b8e58a1
AD		 3425 /*
3426 * It makes no sense for the first data block to be beyond the end
3427 * of the filesystem.
3428 */
3429 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3430 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3431 "block %u is beyond end of filesystem (%llu)",
3432 le32_to_cpu(es->s_first_data_block),
3433 ext4_blocks_count(es));
e7c95593
ES		 3434 goto failed_mount;
3435 }
bd81d8ee
LV		 3436 blocks_count = (ext4_blocks_count(es) -
3437 le32_to_cpu(es->s_first_data_block) +
3438 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3439 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3440 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3441 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3442 "(block count %llu, first data block %u, "
b31e1552 3443 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3444 ext4_blocks_count(es),
3445 le32_to_cpu(es->s_first_data_block),
3446 EXT4_BLOCKS_PER_GROUP(sb));
3447 goto failed_mount;
3448 }
bd81d8ee 3449 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3450 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3451 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3452 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3453 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3454 sbi->s_group_desc = ext4_kvmalloc(db_count *
3455 sizeof(struct buffer_head *),
3456 GFP_KERNEL);
ac27a0ec 3457 if (sbi->s_group_desc == NULL) {
b31e1552 3458 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3459 goto failed_mount;
3460 }
3461
9f6200bb
TT		 3462 if (ext4_proc_root)
3463 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3464
66acdcf4
TT		 3465 if (sbi->s_proc)
3466 proc_create_data("options", S_IRUGO, sbi->s_proc,
3467 &ext4_seq_options_fops, sb);
3468
705895b6 3469 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3470
3471 for (i = 0; i < db_count; i++) {
70bbb3e0 3472 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3473 sbi->s_group_desc[i] = sb_bread(sb, block);
3474 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3475 ext4_msg(sb, KERN_ERR,
3476 "can't read group descriptor %d", i);
ac27a0ec
DK		 3477 db_count = i;
3478 goto failed_mount2;
3479 }
3480 }
bfff6873 3481 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3482 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3483 goto failed_mount2;
3484 }
772cb7c8
JS		 3485 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3486 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3487 ext4_msg(sb, KERN_ERR,
3488 "unable to initialize "
3489 "flex_bg meta info!");
772cb7c8
JS		 3490 goto failed_mount2;
3491 }
3492
ac27a0ec
DK		 3493 sbi->s_gdb_count = db_count;
3494 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3495 spin_lock_init(&sbi->s_next_gen_lock);
3496
04496411
TM		 3497 init_timer(&sbi->s_err_report);
3498 sbi->s_err_report.function = print_daily_error_info;
3499 sbi->s_err_report.data = (unsigned long) sb;
3500
57042651 3501 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3502 ext4_count_free_clusters(sb));
ce7e010a
TT		 3503 if (!err) {
3504 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3505 ext4_count_free_inodes(sb));
3506 }
3507 if (!err) {
3508 err = percpu_counter_init(&sbi->s_dirs_counter,
3509 ext4_count_dirs(sb));
3510 }
3511 if (!err) {
57042651 3512 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3513 }
3514 if (err) {
3515 ext4_msg(sb, KERN_ERR, "insufficient memory");
3516 goto failed_mount3;
3517 }
3518
c9de560d 3519 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3520 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3521
ac27a0ec
DK		 3522 /*
3523 * set up enough so that it can read an inode
3524 */
9ca92389
TT		 3525 if (!test_opt(sb, NOLOAD) &&
3526 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3527 sb->s_op = &ext4_sops;
3528 else
3529 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3530 sb->s_export_op = &ext4_export_ops;
3531 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3532#ifdef CONFIG_QUOTA
617ba13b
MC		 3533 sb->s_qcop = &ext4_qctl_operations;
3534 sb->dq_op = &ext4_quota_operations;
ac27a0ec 3535#endif
f2fa2ffc
AK		 3536 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3537
ac27a0ec 3538 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3539 mutex_init(&sbi->s_orphan_lock);
8f82f840 3540 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3541
3542 sb->s_root = NULL;
3543
3544 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3545 EXT4_HAS_INCOMPAT_FEATURE(sb,
3546 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3547
c5e06d10
JL		 3548 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3549 !(sb->s_flags & MS_RDONLY))
3550 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3551 goto failed_mount3;
3552
ac27a0ec
DK		 3553 /*
3554 * The first inode we look at is the journal inode. Don't try
3555 * root first: it may be modified in the journal!
3556 */
3557 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3558 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3559 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3560 goto failed_mount3;
0390131b
FM		 3561 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3562 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3563 ext4_msg(sb, KERN_ERR, "required journal recovery "
3564 "suppressed and not mounted read-only");
744692dc 3565 goto failed_mount_wq;
ac27a0ec 3566 } else {
fd8c37ec 3567 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3568 sbi->s_journal = NULL;
3569 needs_recovery = 0;
3570 goto no_journal;
ac27a0ec
DK		 3571 }
3572
eb40a09c
JS		 3573 if (ext4_blocks_count(es) > 0xffffffffULL &&
3574 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3575 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3576 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3577 goto failed_mount_wq;
eb40a09c
JS		 3578 }
3579
d4da6c9c
LT		 3580 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3581 jbd2_journal_set_features(sbi->s_journal,
3582 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3583 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3584 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3585 jbd2_journal_set_features(sbi->s_journal,
3586 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3587 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3588 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3589 } else {
3590 jbd2_journal_clear_features(sbi->s_journal,
3591 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3592 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3593 }
818d276c 3594
ac27a0ec
DK		 3595 /* We have now updated the journal if required, so we can
3596 * validate the data journaling mode. */
3597 switch (test_opt(sb, DATA_FLAGS)) {
3598 case 0:
3599 /* No mode set, assume a default based on the journal
63f57933
AM		 3600 * capabilities: ORDERED_DATA if the journal can
3601 * cope, else JOURNAL_DATA
3602 */
dab291af
MC		 3603 if (jbd2_journal_check_available_features
3604 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3605 set_opt(sb, ORDERED_DATA);
ac27a0ec 3606 else
fd8c37ec 3607 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3608 break;
3609
617ba13b
MC		 3610 case EXT4_MOUNT_ORDERED_DATA:
3611 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3612 if (!jbd2_journal_check_available_features
3613 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3614 ext4_msg(sb, KERN_ERR, "Journal does not support "
3615 "requested data journaling mode");
744692dc 3616 goto failed_mount_wq;
ac27a0ec
DK		 3617 }
3618 default:
3619 break;
3620 }
b3881f74 3621 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3622
18aadd47
BJ		 3623 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3624
ce7e010a
TT		 3625 /*
3626 * The journal may have updated the bg summary counts, so we
3627 * need to update the global counters.
3628 */
57042651 3629 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3630 ext4_count_free_clusters(sb));
ce7e010a
TT		 3631 percpu_counter_set(&sbi->s_freeinodes_counter,
3632 ext4_count_free_inodes(sb));
3633 percpu_counter_set(&sbi->s_dirs_counter,
3634 ext4_count_dirs(sb));
57042651 3635 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3636
ce7e010a 3637no_journal:
fd89d5f2
TH		 3638 /*
3639 * The maximum number of concurrent works can be high and
3640 * concurrency isn't really necessary. Limit it to 1.
3641 */
3642 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3643 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3644 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3645 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3646 goto failed_mount_wq;
3647 }
3648
ac27a0ec 3649 /*
dab291af 3650 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3651 * so we can safely mount the rest of the filesystem now.
3652 */
3653
1d1fe1ee
DH		 3654 root = ext4_iget(sb, EXT4_ROOT_INO);
3655 if (IS_ERR(root)) {
b31e1552 3656 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3657 ret = PTR_ERR(root);
32a9bb57 3658 root = NULL;
ac27a0ec
DK		 3659 goto failed_mount4;
3660 }
3661 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3662 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3663 iput(root);
ac27a0ec
DK		 3664 goto failed_mount4;
3665 }
48fde701 3666 sb->s_root = d_make_root(root);
1d1fe1ee 3667 if (!sb->s_root) {
b31e1552 3668 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3669 ret = -ENOMEM;
3670 goto failed_mount4;
3671 }
ac27a0ec 3672
2b2d6d01 3673 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3674
3675 /* determine the minimum size of new large inodes, if present */
3676 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3677 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3678 EXT4_GOOD_OLD_INODE_SIZE;
3679 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3680 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3681 if (sbi->s_want_extra_isize <
3682 le16_to_cpu(es->s_want_extra_isize))
3683 sbi->s_want_extra_isize =
3684 le16_to_cpu(es->s_want_extra_isize);
3685 if (sbi->s_want_extra_isize <
3686 le16_to_cpu(es->s_min_extra_isize))
3687 sbi->s_want_extra_isize =
3688 le16_to_cpu(es->s_min_extra_isize);
3689 }
3690 }
3691 /* Check if enough inode space is available */
3692 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3693 sbi->s_inode_size) {
3694 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3695 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3696 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3697 "available");
ef7f3835
KS		 3698 }
3699
6fd058f7
TT		 3700 err = ext4_setup_system_zone(sb);
3701 if (err) {
b31e1552 3702 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3703 "zone (%d)", err);
94bf608a 3704 goto failed_mount4a;
6fd058f7
TT		 3705 }
3706
c2774d84
AK		 3707 ext4_ext_init(sb);
3708 err = ext4_mb_init(sb, needs_recovery);
3709 if (err) {
421f91d2 3710 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3711 err);
dcf2d804 3712 goto failed_mount5;
c2774d84
AK		 3713 }
3714
bfff6873
LC		 3715 err = ext4_register_li_request(sb, first_not_zeroed);
3716 if (err)
dcf2d804 3717 goto failed_mount6;
bfff6873 3718
3197ebdb
TT		 3719 sbi->s_kobj.kset = ext4_kset;
3720 init_completion(&sbi->s_kobj_unregister);
3721 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3722 "%s", sb->s_id);
dcf2d804
TM		 3723 if (err)
3724 goto failed_mount7;
3197ebdb 3725
617ba13b
MC		 3726 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3727 ext4_orphan_cleanup(sb, es);
3728 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3729 if (needs_recovery) {
b31e1552 3730 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3731 ext4_mark_recovery_complete(sb, es);
3732 }
3733 if (EXT4_SB(sb)->s_journal) {
3734 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3735 descr = " journalled data mode";
3736 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3737 descr = " ordered data mode";
3738 else
3739 descr = " writeback data mode";
3740 } else
3741 descr = "out journal";
3742
d4c402d9 3743 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3744 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3745 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3746
66e61a9e
TT		 3747 if (es->s_error_count)
3748 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3749
d4c402d9 3750 kfree(orig_data);
ac27a0ec
DK		 3751 return 0;
3752
617ba13b 3753cantfind_ext4:
ac27a0ec 3754 if (!silent)
b31e1552 3755 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3756 goto failed_mount;
3757
dcf2d804
TM		 3758failed_mount7:
3759 ext4_unregister_li_request(sb);
3760failed_mount6:
dcf2d804 3761 ext4_mb_release(sb);
94bf608a
AV		 3762failed_mount5:
3763 ext4_ext_release(sb);
dcf2d804 3764 ext4_release_system_zone(sb);
94bf608a
AV		 3765failed_mount4a:
3766 dput(sb->s_root);
32a9bb57 3767 sb->s_root = NULL;
94bf608a 3768failed_mount4:
b31e1552 3769 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3770 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3771failed_mount_wq:
0390131b
FM		 3772 if (sbi->s_journal) {
3773 jbd2_journal_destroy(sbi->s_journal);
3774 sbi->s_journal = NULL;
3775 }
ac27a0ec 3776failed_mount3:
04496411 3777 del_timer(&sbi->s_err_report);
9933fc0a
TT		 3778 if (sbi->s_flex_groups)
3779 ext4_kvfree(sbi->s_flex_groups);
57042651 3780 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 3781 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3782 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 3783 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 3784 if (sbi->s_mmp_tsk)
3785 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3786failed_mount2:
3787 for (i = 0; i < db_count; i++)
3788 brelse(sbi->s_group_desc[i]);
f18a5f21 3789 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 3790failed_mount:
0441984a
DW		 3791 if (sbi->s_chksum_driver)
3792 crypto_free_shash(sbi->s_chksum_driver);
240799cd 3793 if (sbi->s_proc) {
66acdcf4 3794 remove_proc_entry("options", sbi->s_proc);
9f6200bb 3795 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3796 }
ac27a0ec
DK		 3797#ifdef CONFIG_QUOTA
3798 for (i = 0; i < MAXQUOTAS; i++)
3799 kfree(sbi->s_qf_names[i]);
3800#endif
617ba13b 3801 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3802 brelse(bh);
3803out_fail:
3804 sb->s_fs_info = NULL;
f6830165 3805 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3806 kfree(sbi);
dcc7dae3 3807out_free_orig:
d4c402d9 3808 kfree(orig_data);
1d1fe1ee 3809 return ret;
ac27a0ec
DK		 3810}
3811
3812/*
3813 * Setup any per-fs journal parameters now. We'll do this both on
3814 * initial mount, once the journal has been initialised but before we've
3815 * done any recovery; and again on any subsequent remount.
3816 */
617ba13b 3817static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3818{
617ba13b 3819 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3820
30773840
TT		 3821 journal->j_commit_interval = sbi->s_commit_interval;
3822 journal->j_min_batch_time = sbi->s_min_batch_time;
3823 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3824
a931da6a 3825 write_lock(&journal->j_state_lock);
ac27a0ec 3826 if (test_opt(sb, BARRIER))
dab291af 3827 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3828 else
dab291af 3829 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3830 if (test_opt(sb, DATA_ERR_ABORT))
3831 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3832 else
3833 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3834 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3835}
3836
617ba13b 3837static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3838 unsigned int journal_inum)
3839{
3840 struct inode *journal_inode;
3841 journal_t *journal;
3842
0390131b
FM		 3843 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3844
ac27a0ec
DK		 3845 /* First, test for the existence of a valid inode on disk. Bad
3846 * things happen if we iget() an unused inode, as the subsequent
3847 * iput() will try to delete it. */
3848
1d1fe1ee
DH		 3849 journal_inode = ext4_iget(sb, journal_inum);
3850 if (IS_ERR(journal_inode)) {
b31e1552 3851 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3852 return NULL;
3853 }
3854 if (!journal_inode->i_nlink) {
3855 make_bad_inode(journal_inode);
3856 iput(journal_inode);
b31e1552 3857 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3858 return NULL;
3859 }
3860
e5f8eab8 3861 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3862 journal_inode, journal_inode->i_size);
1d1fe1ee 3863 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3864 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3865 iput(journal_inode);
3866 return NULL;
3867 }
3868
dab291af 3869 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3870 if (!journal) {
b31e1552 3871 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3872 iput(journal_inode);
3873 return NULL;
3874 }
3875 journal->j_private = sb;
617ba13b 3876 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3877 return journal;
3878}
3879
617ba13b 3880static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3881 dev_t j_dev)
3882{
2b2d6d01 3883 struct buffer_head *bh;
ac27a0ec 3884 journal_t *journal;
617ba13b
MC		 3885 ext4_fsblk_t start;
3886 ext4_fsblk_t len;
ac27a0ec 3887 int hblock, blocksize;
617ba13b 3888 ext4_fsblk_t sb_block;
ac27a0ec 3889 unsigned long offset;
2b2d6d01 3890 struct ext4_super_block *es;
ac27a0ec
DK		 3891 struct block_device *bdev;
3892
0390131b
FM		 3893 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3894
b31e1552 3895 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3896 if (bdev == NULL)
3897 return NULL;
3898
ac27a0ec 3899 blocksize = sb->s_blocksize;
e1defc4f 3900 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3901 if (blocksize < hblock) {
b31e1552
ES		 3902 ext4_msg(sb, KERN_ERR,
3903 "blocksize too small for journal device");
ac27a0ec
DK		 3904 goto out_bdev;
3905 }
3906
617ba13b
MC		 3907 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3908 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3909 set_blocksize(bdev, blocksize);
3910 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3911 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3912 "external journal");
ac27a0ec
DK		 3913 goto out_bdev;
3914 }
3915
617ba13b
MC		 3916 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3917 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3918 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3919 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3920 ext4_msg(sb, KERN_ERR, "external journal has "
3921 "bad superblock");
ac27a0ec
DK		 3922 brelse(bh);
3923 goto out_bdev;
3924 }
3925
617ba13b 3926 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3927 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3928 brelse(bh);
3929 goto out_bdev;
3930 }
3931
bd81d8ee 3932 len = ext4_blocks_count(es);
ac27a0ec
DK		 3933 start = sb_block + 1;
3934 brelse(bh); /* we're done with the superblock */
3935
dab291af 3936 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3937 start, len, blocksize);
3938 if (!journal) {
b31e1552 3939 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3940 goto out_bdev;
3941 }
3942 journal->j_private = sb;
3943 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3944 wait_on_buffer(journal->j_sb_buffer);
3945 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3946 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3947 goto out_journal;
3948 }
3949 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3950 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3951 "user (unsupported) - %d",
ac27a0ec
DK		 3952 be32_to_cpu(journal->j_superblock->s_nr_users));
3953 goto out_journal;
3954 }
617ba13b
MC		 3955 EXT4_SB(sb)->journal_bdev = bdev;
3956 ext4_init_journal_params(sb, journal);
ac27a0ec 3957 return journal;
0b8e58a1 3958
ac27a0ec 3959out_journal:
dab291af 3960 jbd2_journal_destroy(journal);
ac27a0ec 3961out_bdev:
617ba13b 3962 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3963 return NULL;
3964}
3965
617ba13b
MC		 3966static int ext4_load_journal(struct super_block *sb,
3967 struct ext4_super_block *es,
ac27a0ec
DK		 3968 unsigned long journal_devnum)
3969{
3970 journal_t *journal;
3971 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3972 dev_t journal_dev;
3973 int err = 0;
3974 int really_read_only;
3975
0390131b
FM		 3976 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3977
ac27a0ec
DK		 3978 if (journal_devnum &&
3979 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3980 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3981 "numbers have changed");
ac27a0ec
DK		 3982 journal_dev = new_decode_dev(journal_devnum);
3983 } else
3984 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3985
3986 really_read_only = bdev_read_only(sb->s_bdev);
3987
3988 /*
3989 * Are we loading a blank journal or performing recovery after a
3990 * crash? For recovery, we need to check in advance whether we
3991 * can get read-write access to the device.
3992 */
617ba13b 3993 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3994 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3995 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3996 "required on readonly filesystem");
ac27a0ec 3997 if (really_read_only) {
b31e1552
ES		 3998 ext4_msg(sb, KERN_ERR, "write access "
3999 "unavailable, cannot proceed");
ac27a0ec
DK		 4000 return -EROFS;
4001 }
b31e1552
ES		 4002 ext4_msg(sb, KERN_INFO, "write access will "
4003 "be enabled during recovery");
ac27a0ec
DK		 4004 }
4005 }
4006
4007 if (journal_inum && journal_dev) {
b31e1552
ES		 4008 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4009 "and inode journals!");
ac27a0ec
DK		 4010 return -EINVAL;
4011 }
4012
4013 if (journal_inum) {
617ba13b 4014 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4015 return -EINVAL;
4016 } else {
617ba13b 4017 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4018 return -EINVAL;
4019 }
4020
90576c0b 4021 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4022 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4023
617ba13b 4024 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4025 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4026 if (!err) {
4027 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4028 if (save)
4029 memcpy(save, ((char *) es) +
4030 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4031 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4032 if (save)
4033 memcpy(((char *) es) + EXT4_S_ERR_START,
4034 save, EXT4_S_ERR_LEN);
4035 kfree(save);
4036 }
ac27a0ec
DK		 4037
4038 if (err) {
b31e1552 4039 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4040 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4041 return err;
4042 }
4043
617ba13b
MC		 4044 EXT4_SB(sb)->s_journal = journal;
4045 ext4_clear_journal_err(sb, es);
ac27a0ec 4046
c41303ce 4047 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4048 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4049 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4050
4051 /* Make sure we flush the recovery flag to disk. */
e2d67052 4052 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4053 }
4054
4055 return 0;
4056}
4057
e2d67052 4058static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4059{
e2d67052 4060 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4061 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4062 int error = 0;
ac27a0ec 4063
7c2e7087 4064 if (!sbh || block_device_ejected(sb))
c4be0c1d 4065 return error;
914258bf
TT		 4066 if (buffer_write_io_error(sbh)) {
4067 /*
4068 * Oh, dear. A previous attempt to write the
4069 * superblock failed. This could happen because the
4070 * USB device was yanked out. Or it could happen to
4071 * be a transient write error and maybe the block will
4072 * be remapped. Nothing we can do but to retry the
4073 * write and hope for the best.
4074 */
b31e1552
ES		 4075 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4076 "superblock detected");
914258bf
TT		 4077 clear_buffer_write_io_error(sbh);
4078 set_buffer_uptodate(sbh);
4079 }
71290b36
TT		 4080 /*
4081 * If the file system is mounted read-only, don't update the
4082 * superblock write time. This avoids updating the superblock
4083 * write time when we are mounting the root file system
4084 * read/only but we need to replay the journal; at that point,
4085 * for people who are east of GMT and who make their clock
4086 * tick in localtime for Windows bug-for-bug compatibility,
4087 * the clock is set in the future, and this will cause e2fsck
4088 * to complain and force a full file system check.
4089 */
4090 if (!(sb->s_flags & MS_RDONLY))
4091 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4092 if (sb->s_bdev->bd_part)
4093 es->s_kbytes_written =
4094 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4095 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4096 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4097 else
4098 es->s_kbytes_written =
4099 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4100 ext4_free_blocks_count_set(es,
4101 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4102 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4103 es->s_free_inodes_count =
4104 cpu_to_le32(percpu_counter_sum_positive(
4105 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 4106 sb->s_dirt = 0;
ac27a0ec 4107 BUFFER_TRACE(sbh, "marking dirty");
a9c47317 4108 ext4_superblock_csum_set(sb, es);
ac27a0ec 4109 mark_buffer_dirty(sbh);
914258bf 4110 if (sync) {
c4be0c1d
TS		 4111 error = sync_dirty_buffer(sbh);
4112 if (error)
4113 return error;
4114
4115 error = buffer_write_io_error(sbh);
4116 if (error) {
b31e1552
ES		 4117 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4118 "superblock");
914258bf
TT		 4119 clear_buffer_write_io_error(sbh);
4120 set_buffer_uptodate(sbh);
4121 }
4122 }
c4be0c1d 4123 return error;
ac27a0ec
DK		 4124}
4125
ac27a0ec
DK		 4126/*
4127 * Have we just finished recovery? If so, and if we are mounting (or
4128 * remounting) the filesystem readonly, then we will end up with a
4129 * consistent fs on disk. Record that fact.
4130 */
2b2d6d01
TT		 4131static void ext4_mark_recovery_complete(struct super_block *sb,
4132 struct ext4_super_block *es)
ac27a0ec 4133{
617ba13b 4134 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4135
0390131b
FM		 4136 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4137 BUG_ON(journal != NULL);
4138 return;
4139 }
dab291af 4140 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4141 if (jbd2_journal_flush(journal) < 0)
4142 goto out;
4143
617ba13b 4144 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4145 sb->s_flags & MS_RDONLY) {
617ba13b 4146 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4147 ext4_commit_super(sb, 1);
ac27a0ec 4148 }
7ffe1ea8
HK		 4149
4150out:
dab291af 4151 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4152}
4153
4154/*
4155 * If we are mounting (or read-write remounting) a filesystem whose journal
4156 * has recorded an error from a previous lifetime, move that error to the
4157 * main filesystem now.
4158 */
2b2d6d01
TT		 4159static void ext4_clear_journal_err(struct super_block *sb,
4160 struct ext4_super_block *es)
ac27a0ec
DK		 4161{
4162 journal_t *journal;
4163 int j_errno;
4164 const char *errstr;
4165
0390131b
FM		 4166 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4167
617ba13b 4168 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4169
4170 /*
4171 * Now check for any error status which may have been recorded in the
617ba13b 4172 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4173 */
4174
dab291af 4175 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4176 if (j_errno) {
4177 char nbuf[16];
4178
617ba13b 4179 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4180 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4181 "from previous mount: %s", errstr);
12062ddd 4182 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4183
617ba13b
MC		 4184 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4185 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4186 ext4_commit_super(sb, 1);
ac27a0ec 4187
dab291af 4188 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4189 }
4190}
4191
4192/*
4193 * Force the running and committing transactions to commit,
4194 * and wait on the commit.
4195 */
617ba13b 4196int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4197{
4198 journal_t *journal;
0390131b 4199 int ret = 0;
ac27a0ec
DK		 4200
4201 if (sb->s_flags & MS_RDONLY)
4202 return 0;
4203
617ba13b 4204 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4205 if (journal) {
437f88cc 4206 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4207 ret = ext4_journal_force_commit(journal);
6b0310fb 4208 }
0390131b 4209
ac27a0ec
DK		 4210 return ret;
4211}
4212
2b2d6d01 4213static void ext4_write_super(struct super_block *sb)
ac27a0ec 4214{
ebc1ac16 4215 lock_super(sb);
9ca92389 4216 ext4_commit_super(sb, 1);
ebc1ac16 4217 unlock_super(sb);
ac27a0ec
DK		 4218}
4219
617ba13b 4220static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4221{
14ce0cb4 4222 int ret = 0;
9eddacf9 4223 tid_t target;
8d5d02e6 4224 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4225
9bffad1e 4226 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4227 flush_workqueue(sbi->dio_unwritten_wq);
4228 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4229 if (wait)
8d5d02e6 4230 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4231 }
14ce0cb4 4232 return ret;
ac27a0ec
DK		 4233}
4234
4235/*
4236 * LVM calls this function before a (read-only) snapshot is created. This
4237 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4238 *
4239 * Note that only this function cannot bring a filesystem to be in a clean
4240 * state independently, because ext4 prevents a new handle from being started
4241 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4242 * the upper layer.
ac27a0ec 4243 */
c4be0c1d 4244static int ext4_freeze(struct super_block *sb)
ac27a0ec 4245{
c4be0c1d
TS		 4246 int error = 0;
4247 journal_t *journal;
ac27a0ec 4248
9ca92389
TT		 4249 if (sb->s_flags & MS_RDONLY)
4250 return 0;
ac27a0ec 4251
9ca92389 4252 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4253
9ca92389
TT		 4254 /* Now we set up the journal barrier. */
4255 jbd2_journal_lock_updates(journal);
ac27a0ec 4256
9ca92389
TT		 4257 /*
4258 * Don't clear the needs_recovery flag if we failed to flush
4259 * the journal.
4260 */
4261 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4262 if (error < 0)
4263 goto out;
9ca92389
TT		 4264
4265 /* Journal blocked and flushed, clear needs_recovery flag. */
4266 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4267 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4268out:
4269 /* we rely on s_frozen to stop further updates */
4270 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4271 return error;
ac27a0ec
DK		 4272}
4273
4274/*
4275 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4276 * flag here, even though the filesystem is not technically dirty yet.
4277 */
c4be0c1d 4278static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4279{
9ca92389
TT		 4280 if (sb->s_flags & MS_RDONLY)
4281 return 0;
4282
4283 lock_super(sb);
4284 /* Reset the needs_recovery flag before the fs is unlocked. */
4285 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4286 ext4_commit_super(sb, 1);
4287 unlock_super(sb);
c4be0c1d 4288 return 0;
ac27a0ec
DK		 4289}
4290
673c6100
TT		 4291/*
4292 * Structure to save mount options for ext4_remount's benefit
4293 */
4294struct ext4_mount_options {
4295 unsigned long s_mount_opt;
a2595b8a 4296 unsigned long s_mount_opt2;
673c6100
TT		 4297 uid_t s_resuid;
4298 gid_t s_resgid;
4299 unsigned long s_commit_interval;
4300 u32 s_min_batch_time, s_max_batch_time;
4301#ifdef CONFIG_QUOTA
4302 int s_jquota_fmt;
4303 char *s_qf_names[MAXQUOTAS];
4304#endif
4305};
4306
2b2d6d01 4307static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4308{
2b2d6d01 4309 struct ext4_super_block *es;
617ba13b 4310 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4311 unsigned long old_sb_flags;
617ba13b 4312 struct ext4_mount_options old_opts;
c79d967d 4313 int enable_quota = 0;
8a266467 4314 ext4_group_t g;
b3881f74 4315 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4316 int err = 0;
ac27a0ec
DK		 4317#ifdef CONFIG_QUOTA
4318 int i;
4319#endif
d4c402d9 4320 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4321
4322 /* Store the original options */
bbd6851a 4323 lock_super(sb);
ac27a0ec
DK		 4324 old_sb_flags = sb->s_flags;
4325 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4326 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4327 old_opts.s_resuid = sbi->s_resuid;
4328 old_opts.s_resgid = sbi->s_resgid;
4329 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4330 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4331 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4332#ifdef CONFIG_QUOTA
4333 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4334 for (i = 0; i < MAXQUOTAS; i++)
4335 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4336#endif
b3881f74
TT		 4337 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4338 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4339
4340 /*
4341 * Allow the "check" option to be passed as a remount option.
4342 */
661aa520 4343 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4344 err = -EINVAL;
4345 goto restore_opts;
4346 }
4347
4ab2f15b 4348 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4349 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4350
4351 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4352 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4353
4354 es = sbi->s_es;
4355
b3881f74 4356 if (sbi->s_journal) {
0390131b 4357 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4358 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4359 }
ac27a0ec 4360
661aa520 4361 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4362 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4363 err = -EROFS;
4364 goto restore_opts;
4365 }
4366
4367 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4368 err = dquot_suspend(sb, -1);
4369 if (err < 0)
c79d967d 4370 goto restore_opts;
c79d967d 4371
ac27a0ec
DK		 4372 /*
4373 * First of all, the unconditional stuff we have to do
4374 * to disable replay of the journal when we next remount
4375 */
4376 sb->s_flags |= MS_RDONLY;
4377
4378 /*
4379 * OK, test if we are remounting a valid rw partition
4380 * readonly, and if so set the rdonly flag and then
4381 * mark the partition as valid again.
4382 */
617ba13b
MC		 4383 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4384 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4385 es->s_state = cpu_to_le16(sbi->s_mount_state);
4386
a63c9eb2 4387 if (sbi->s_journal)
0390131b 4388 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4389 } else {
a13fb1a4
ES		 4390 /* Make sure we can mount this feature set readwrite */
4391 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4392 err = -EROFS;
4393 goto restore_opts;
4394 }
8a266467
TT		 4395 /*
4396 * Make sure the group descriptor checksums
0b8e58a1 4397 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4398 */
4399 for (g = 0; g < sbi->s_groups_count; g++) {
4400 struct ext4_group_desc *gdp =
4401 ext4_get_group_desc(sb, g, NULL);
4402
4403 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4404 ext4_msg(sb, KERN_ERR,
4405 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4406 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4407 le16_to_cpu(gdp->bg_checksum));
4408 err = -EINVAL;
4409 goto restore_opts;
4410 }
4411 }
4412
ead6596b
ES		 4413 /*
4414 * If we have an unprocessed orphan list hanging
4415 * around from a previously readonly bdev mount,
4416 * require a full umount/remount for now.
4417 */
4418 if (es->s_last_orphan) {
b31e1552 4419 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4420 "remount RDWR because of unprocessed "
4421 "orphan inode list. Please "
b31e1552 4422 "umount/remount instead");
ead6596b
ES		 4423 err = -EINVAL;
4424 goto restore_opts;
4425 }
4426
ac27a0ec
DK		 4427 /*
4428 * Mounting a RDONLY partition read-write, so reread
4429 * and store the current valid flag. (It may have
4430 * been changed by e2fsck since we originally mounted
4431 * the partition.)
4432 */
0390131b
FM		 4433 if (sbi->s_journal)
4434 ext4_clear_journal_err(sb, es);
ac27a0ec 4435 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4436 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4437 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4438 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4439 EXT4_FEATURE_INCOMPAT_MMP))
4440 if (ext4_multi_mount_protect(sb,
4441 le64_to_cpu(es->s_mmp_block))) {
4442 err = -EROFS;
4443 goto restore_opts;
4444 }
c79d967d 4445 enable_quota = 1;
ac27a0ec
DK		 4446 }
4447 }
bfff6873
LC		 4448
4449 /*
4450 * Reinitialize lazy itable initialization thread based on
4451 * current settings
4452 */
4453 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4454 ext4_unregister_li_request(sb);
4455 else {
4456 ext4_group_t first_not_zeroed;
4457 first_not_zeroed = ext4_has_uninit_itable(sb);
4458 ext4_register_li_request(sb, first_not_zeroed);
4459 }
4460
6fd058f7 4461 ext4_setup_system_zone(sb);
0390131b 4462 if (sbi->s_journal == NULL)
e2d67052 4463 ext4_commit_super(sb, 1);
0390131b 4464
ac27a0ec
DK		 4465#ifdef CONFIG_QUOTA
4466 /* Release old quota file names */
4467 for (i = 0; i < MAXQUOTAS; i++)
4468 if (old_opts.s_qf_names[i] &&
4469 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4470 kfree(old_opts.s_qf_names[i]);
4471#endif
bbd6851a 4472 unlock_super(sb);
c79d967d 4473 if (enable_quota)
0f0dd62f 4474 dquot_resume(sb, -1);
d4c402d9
CW		 4475
4476 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4477 kfree(orig_data);
ac27a0ec 4478 return 0;
0b8e58a1 4479
ac27a0ec
DK		 4480restore_opts:
4481 sb->s_flags = old_sb_flags;
4482 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4483 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4484 sbi->s_resuid = old_opts.s_resuid;
4485 sbi->s_resgid = old_opts.s_resgid;
4486 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4487 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4488 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4489#ifdef CONFIG_QUOTA
4490 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4491 for (i = 0; i < MAXQUOTAS; i++) {
4492 if (sbi->s_qf_names[i] &&
4493 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4494 kfree(sbi->s_qf_names[i]);
4495 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4496 }
4497#endif
bbd6851a 4498 unlock_super(sb);
d4c402d9 4499 kfree(orig_data);
ac27a0ec
DK		 4500 return err;
4501}
4502
f975d6bc
TT		 4503/*
4504 * Note: calculating the overhead so we can be compatible with
4505 * historical BSD practice is quite difficult in the face of
4506 * clusters/bigalloc. This is because multiple metadata blocks from
4507 * different block group can end up in the same allocation cluster.
4508 * Calculating the exact overhead in the face of clustered allocation
4509 * requires either O(all block bitmaps) in memory or O(number of block
4510 * groups**2) in time. We will still calculate the superblock for
4511 * older file systems --- and if we come across with a bigalloc file
4512 * system with zero in s_overhead_clusters the estimate will be close to
4513 * correct especially for very large cluster sizes --- but for newer
4514 * file systems, it's better to calculate this figure once at mkfs
4515 * time, and store it in the superblock. If the superblock value is
4516 * present (even for non-bigalloc file systems), we will use it.
4517 */
2b2d6d01 4518static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4519{
4520 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4521 struct ext4_sb_info *sbi = EXT4_SB(sb);
4522 struct ext4_super_block *es = sbi->s_es;
f975d6bc 4523 struct ext4_group_desc *gdp;
960cc398 4524 u64 fsid;
d02a9391 4525 s64 bfree;
ac27a0ec 4526
5e70030d
BP		 4527 if (test_opt(sb, MINIX_DF)) {
4528 sbi->s_overhead_last = 0;
f975d6bc
TT		 4529 } else if (es->s_overhead_clusters) {
4530 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
6bc9feff 4531 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4532 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4533 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4534
4535 /*
5e70030d
BP		 4536 * Compute the overhead (FS structures). This is constant
4537 * for a given filesystem unless the number of block groups
4538 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4539 */
4540
4541 /*
4542 * All of the blocks before first_data_block are
4543 * overhead
4544 */
f975d6bc 4545 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
ac27a0ec
DK		 4546
4547 /*
f975d6bc 4548 * Add the overhead found in each block group
ac27a0ec
DK		 4549 */
4550 for (i = 0; i < ngroups; i++) {
f975d6bc
TT		 4551 gdp = ext4_get_group_desc(sb, i, NULL);
4552 overhead += ext4_num_overhead_clusters(sb, i, gdp);
ac27a0ec
DK		 4553 cond_resched();
4554 }
5e70030d
BP		 4555 sbi->s_overhead_last = overhead;
4556 smp_wmb();
6bc9feff 4557 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4558 }
4559
617ba13b 4560 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4561 buf->f_bsize = sb->s_blocksize;
f975d6bc
TT		 4562 buf->f_blocks = (ext4_blocks_count(es) -
4563 EXT4_C2B(sbi, sbi->s_overhead_last));
57042651
TT		 4564 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4565 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4566 /* prevent underflow in case that few free space is available */
57042651 4567 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4568 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4569 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4570 buf->f_bavail = 0;
4571 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4572 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4573 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4574 fsid = le64_to_cpup((void *)es->s_uuid) ^
4575 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4576 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4577 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4578
ac27a0ec
DK		 4579 return 0;
4580}
4581
0b8e58a1
AD		 4582/* Helper function for writing quotas on sync - we need to start transaction
4583 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4584 * Process 1 Process 2
617ba13b 4585 * ext4_create() quota_sync()
a269eb18 4586 * jbd2_journal_start() write_dquot()
871a2931 4587 * dquot_initialize() down(dqio_mutex)
dab291af 4588 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4589 *
4590 */
4591
4592#ifdef CONFIG_QUOTA
4593
4594static inline struct inode *dquot_to_inode(struct dquot *dquot)
4595{
4596 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4597}
4598
617ba13b 4599static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4600{
4601 int ret, err;
4602 handle_t *handle;
4603 struct inode *inode;
4604
4605 inode = dquot_to_inode(dquot);
617ba13b 4606 handle = ext4_journal_start(inode,
0b8e58a1 4607 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4608 if (IS_ERR(handle))
4609 return PTR_ERR(handle);
4610 ret = dquot_commit(dquot);
617ba13b 4611 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4612 if (!ret)
4613 ret = err;
4614 return ret;
4615}
4616
617ba13b 4617static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4618{
4619 int ret, err;
4620 handle_t *handle;
4621
617ba13b 4622 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4623 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4624 if (IS_ERR(handle))
4625 return PTR_ERR(handle);
4626 ret = dquot_acquire(dquot);
617ba13b 4627 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4628 if (!ret)
4629 ret = err;
4630 return ret;
4631}
4632
617ba13b 4633static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4634{
4635 int ret, err;
4636 handle_t *handle;
4637
617ba13b 4638 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4639 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4640 if (IS_ERR(handle)) {
4641 /* Release dquot anyway to avoid endless cycle in dqput() */
4642 dquot_release(dquot);
ac27a0ec 4643 return PTR_ERR(handle);
9c3013e9 4644 }
ac27a0ec 4645 ret = dquot_release(dquot);
617ba13b 4646 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4647 if (!ret)
4648 ret = err;
4649 return ret;
4650}
4651
617ba13b 4652static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4653{
2c8be6b2 4654 /* Are we journaling quotas? */
617ba13b
MC		 4655 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4656 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4657 dquot_mark_dquot_dirty(dquot);
617ba13b 4658 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4659 } else {
4660 return dquot_mark_dquot_dirty(dquot);
4661 }
4662}
4663
617ba13b 4664static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4665{
4666 int ret, err;
4667 handle_t *handle;
4668
4669 /* Data block + inode block */
617ba13b 4670 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4671 if (IS_ERR(handle))
4672 return PTR_ERR(handle);
4673 ret = dquot_commit_info(sb, type);
617ba13b 4674 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4675 if (!ret)
4676 ret = err;
4677 return ret;
4678}
4679
4680/*
4681 * Turn on quotas during mount time - we need to find
4682 * the quota file and such...
4683 */
617ba13b 4684static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4685{
287a8095
CH		 4686 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4687 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4688}
4689
4690/*
4691 * Standard function to be called on quota_on
4692 */
617ba13b 4693static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4694 struct path *path)
ac27a0ec
DK		 4695{
4696 int err;
ac27a0ec
DK		 4697
4698 if (!test_opt(sb, QUOTA))
4699 return -EINVAL;
0623543b 4700
ac27a0ec 4701 /* Quotafile not on the same filesystem? */
d8c9584e 4702 if (path->dentry->d_sb != sb)
ac27a0ec 4703 return -EXDEV;
0623543b
JK		 4704 /* Journaling quota? */
4705 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4706 /* Quotafile not in fs root? */
f00c9e44 4707 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4708 ext4_msg(sb, KERN_WARNING,
4709 "Quota file not on filesystem root. "
4710 "Journaled quota will not work");
2b2d6d01 4711 }
0623543b
JK		 4712
4713 /*
4714 * When we journal data on quota file, we have to flush journal to see
4715 * all updates to the file when we bypass pagecache...
4716 */
0390131b 4717 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4718 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4719 /*
4720 * We don't need to lock updates but journal_flush() could
4721 * otherwise be livelocked...
4722 */
4723 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4724 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4725 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4726 if (err)
7ffe1ea8 4727 return err;
0623543b
JK		 4728 }
4729
f00c9e44 4730 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4731}
4732
ca0e05e4
DM		 4733static int ext4_quota_off(struct super_block *sb, int type)
4734{
21f97697
JK		 4735 struct inode *inode = sb_dqopt(sb)->files[type];
4736 handle_t *handle;
4737
87009d86
DM		 4738 /* Force all delayed allocation blocks to be allocated.
4739 * Caller already holds s_umount sem */
4740 if (test_opt(sb, DELALLOC))
ca0e05e4 4741 sync_filesystem(sb);
ca0e05e4 4742
0b268590
AG		 4743 if (!inode)
4744 goto out;
4745
21f97697
JK		 4746 /* Update modification times of quota files when userspace can
4747 * start looking at them */
4748 handle = ext4_journal_start(inode, 1);
4749 if (IS_ERR(handle))
4750 goto out;
4751 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4752 ext4_mark_inode_dirty(handle, inode);
4753 ext4_journal_stop(handle);
4754
4755out:
ca0e05e4
DM		 4756 return dquot_quota_off(sb, type);
4757}
4758
ac27a0ec
DK		 4759/* Read data from quotafile - avoid pagecache and such because we cannot afford
4760 * acquiring the locks... As quota files are never truncated and quota code
25985edc 4761 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 4762 * we don't have to be afraid of races */
617ba13b 4763static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4764 size_t len, loff_t off)
4765{
4766 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4767 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4768 int err = 0;
4769 int offset = off & (sb->s_blocksize - 1);
4770 int tocopy;
4771 size_t toread;
4772 struct buffer_head *bh;
4773 loff_t i_size = i_size_read(inode);
4774
4775 if (off > i_size)
4776 return 0;
4777 if (off+len > i_size)
4778 len = i_size-off;
4779 toread = len;
4780 while (toread > 0) {
4781 tocopy = sb->s_blocksize - offset < toread ?
4782 sb->s_blocksize - offset : toread;
617ba13b 4783 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4784 if (err)
4785 return err;
4786 if (!bh) /* A hole? */
4787 memset(data, 0, tocopy);
4788 else
4789 memcpy(data, bh->b_data+offset, tocopy);
4790 brelse(bh);
4791 offset = 0;
4792 toread -= tocopy;
4793 data += tocopy;
4794 blk++;
4795 }
4796 return len;
4797}
4798
4799/* Write to quotafile (we know the transaction is already started and has
4800 * enough credits) */
617ba13b 4801static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4802 const char *data, size_t len, loff_t off)
4803{
4804 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4805 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4806 int err = 0;
4807 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4808 struct buffer_head *bh;
4809 handle_t *handle = journal_current_handle();
4810
0390131b 4811 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4812 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4813 " cancelled because transaction is not started",
9c3013e9
JK		 4814 (unsigned long long)off, (unsigned long long)len);
4815 return -EIO;
4816 }
67eeb568
DM		 4817 /*
4818 * Since we account only one data block in transaction credits,
4819 * then it is impossible to cross a block boundary.
4820 */
4821 if (sb->s_blocksize - offset < len) {
4822 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4823 " cancelled because not block aligned",
4824 (unsigned long long)off, (unsigned long long)len);
4825 return -EIO;
4826 }
4827
ac27a0ec 4828 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4829 bh = ext4_bread(handle, inode, blk, 1, &err);
4830 if (!bh)
4831 goto out;
62d2b5f2
JK		 4832 err = ext4_journal_get_write_access(handle, bh);
4833 if (err) {
4834 brelse(bh);
4835 goto out;
ac27a0ec 4836 }
67eeb568
DM		 4837 lock_buffer(bh);
4838 memcpy(bh->b_data+offset, data, len);
4839 flush_dcache_page(bh->b_page);
4840 unlock_buffer(bh);
62d2b5f2 4841 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4842 brelse(bh);
ac27a0ec 4843out:
67eeb568 4844 if (err) {
4d04e4fb 4845 mutex_unlock(&inode->i_mutex);
ac27a0ec 4846 return err;
4d04e4fb 4847 }
67eeb568
DM		 4848 if (inode->i_size < off + len) {
4849 i_size_write(inode, off + len);
617ba13b 4850 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 4851 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4852 }
ac27a0ec 4853 mutex_unlock(&inode->i_mutex);
67eeb568 4854 return len;
ac27a0ec
DK		 4855}
4856
4857#endif
4858
152a0836
AV		 4859static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4860 const char *dev_name, void *data)
ac27a0ec 4861{
152a0836 4862 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4863}
4864
37f328eb 4865#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4866static inline void register_as_ext2(void)
4867{
4868 int err = register_filesystem(&ext2_fs_type);
4869 if (err)
4870 printk(KERN_WARNING
4871 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4872}
4873
4874static inline void unregister_as_ext2(void)
4875{
4876 unregister_filesystem(&ext2_fs_type);
4877}
2035e776
TT		 4878
4879static inline int ext2_feature_set_ok(struct super_block *sb)
4880{
4881 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4882 return 0;
4883 if (sb->s_flags & MS_RDONLY)
4884 return 1;
4885 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4886 return 0;
4887 return 1;
4888}
51b7e3c9 4889MODULE_ALIAS("ext2");
24b58424
TT		 4890#else
4891static inline void register_as_ext2(void) { }
4892static inline void unregister_as_ext2(void) { }
2035e776 4893static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4894#endif
4895
37f328eb 4896#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4897static inline void register_as_ext3(void)
4898{
4899 int err = register_filesystem(&ext3_fs_type);
4900 if (err)
4901 printk(KERN_WARNING
4902 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4903}
4904
4905static inline void unregister_as_ext3(void)
4906{
4907 unregister_filesystem(&ext3_fs_type);
4908}
2035e776
TT		 4909
4910static inline int ext3_feature_set_ok(struct super_block *sb)
4911{
4912 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4913 return 0;
4914 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4915 return 0;
4916 if (sb->s_flags & MS_RDONLY)
4917 return 1;
4918 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4919 return 0;
4920 return 1;
4921}
51b7e3c9 4922MODULE_ALIAS("ext3");
24b58424
TT		 4923#else
4924static inline void register_as_ext3(void) { }
4925static inline void unregister_as_ext3(void) { }
2035e776 4926static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4927#endif
4928
03010a33
TT		 4929static struct file_system_type ext4_fs_type = {
4930 .owner = THIS_MODULE,
4931 .name = "ext4",
152a0836 4932 .mount = ext4_mount,
03010a33
TT		 4933 .kill_sb = kill_block_super,
4934 .fs_flags = FS_REQUIRES_DEV,
4935};
4936
8f021222 4937static int __init ext4_init_feat_adverts(void)
857ac889
LC		 4938{
4939 struct ext4_features *ef;
4940 int ret = -ENOMEM;
4941
4942 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4943 if (!ef)
4944 goto out;
4945
4946 ef->f_kobj.kset = ext4_kset;
4947 init_completion(&ef->f_kobj_unregister);
4948 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4949 "features");
4950 if (ret) {
4951 kfree(ef);
4952 goto out;
4953 }
4954
4955 ext4_feat = ef;
4956 ret = 0;
4957out:
4958 return ret;
4959}
4960
8f021222
LC		 4961static void ext4_exit_feat_adverts(void)
4962{
4963 kobject_put(&ext4_feat->f_kobj);
4964 wait_for_completion(&ext4_feat->f_kobj_unregister);
4965 kfree(ext4_feat);
4966}
4967
e9e3bcec
ES		 4968/* Shared across all ext4 file systems */
4969wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4970struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4971
5dabfc78 4972static int __init ext4_init_fs(void)
ac27a0ec 4973{
e9e3bcec 4974 int i, err;
c9de560d 4975
07c0c5d8
AV		 4976 ext4_li_info = NULL;
4977 mutex_init(&ext4_li_mtx);
4978
12e9b892 4979 ext4_check_flag_values();
e9e3bcec
ES		 4980
4981 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4982 mutex_init(&ext4__aio_mutex[i]);
4983 init_waitqueue_head(&ext4__ioend_wq[i]);
4984 }
4985
5dabfc78 4986 err = ext4_init_pageio();
6fd058f7
TT		 4987 if (err)
4988 return err;
5dabfc78 4989 err = ext4_init_system_zone();
bd2d0210 4990 if (err)
d44651d0 4991 goto out6;
3197ebdb
TT		 4992 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4993 if (!ext4_kset)
dd68314c 4994 goto out5;
d44651d0 4995 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 4996
4997 err = ext4_init_feat_adverts();
dd68314c
TT		 4998 if (err)
4999 goto out4;
857ac889 5000
5dabfc78 5001 err = ext4_init_mballoc();
ac27a0ec 5002 if (err)
6fd058f7 5003 goto out3;
c9de560d 5004
5dabfc78 5005 err = ext4_init_xattr();
c9de560d
AT		 5006 if (err)
5007 goto out2;
ac27a0ec
DK		 5008 err = init_inodecache();
5009 if (err)
5010 goto out1;
24b58424 5011 register_as_ext3();
2035e776 5012 register_as_ext2();
03010a33 5013 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5014 if (err)
5015 goto out;
bfff6873 5016
ac27a0ec
DK		 5017 return 0;
5018out:
24b58424
TT		 5019 unregister_as_ext2();
5020 unregister_as_ext3();
ac27a0ec
DK		 5021 destroy_inodecache();
5022out1:
5dabfc78 5023 ext4_exit_xattr();
c9de560d 5024out2:
5dabfc78 5025 ext4_exit_mballoc();
6fd058f7 5026out3:
8f021222 5027 ext4_exit_feat_adverts();
dd68314c 5028out4:
d44651d0
FJ		 5029 if (ext4_proc_root)
5030 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5031 kset_unregister(ext4_kset);
d44651d0 5032out5:
5dabfc78 5033 ext4_exit_system_zone();
d44651d0 5034out6:
5dabfc78 5035 ext4_exit_pageio();
ac27a0ec
DK		 5036 return err;
5037}
5038
5dabfc78 5039static void __exit ext4_exit_fs(void)
ac27a0ec 5040{
bfff6873 5041 ext4_destroy_lazyinit_thread();
24b58424
TT		 5042 unregister_as_ext2();
5043 unregister_as_ext3();
03010a33 5044 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5045 destroy_inodecache();
5dabfc78
TT		 5046 ext4_exit_xattr();
5047 ext4_exit_mballoc();
8f021222 5048 ext4_exit_feat_adverts();
9f6200bb 5049 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5050 kset_unregister(ext4_kset);
5dabfc78
TT		 5051 ext4_exit_system_zone();
5052 ext4_exit_pageio();
ac27a0ec
DK		 5053}
5054
5055MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5056MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5057MODULE_LICENSE("GPL");
5dabfc78
TT		 5058module_init(ext4_init_fs)
5059module_exit(ext4_exit_fs)
Linux kernel - Deliverables
This page took 1.233725 seconds and 5 git commands to generate.