projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: remove unused function argument 'order' in mb_find_extent()
[GitHub/mt8127/android_kernel_alcatel_ttab.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);
c4be0c1d 77static int ext4_freeze(struct super_block *sb);
152a0836
AV		 78static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
2035e776
TT		 80static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 82static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 83static void ext4_destroy_lazyinit_thread(void);
84static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 85static void ext4_clear_request_list(void);
ac27a0ec 86
2035e776
TT		 87#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94};
95#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96#else
97#define IS_EXT2_SB(sb) (0)
98#endif
99
100
ba69f9ab
JK		 101#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
104 .name = "ext3",
152a0836 105 .mount = ext4_mount,
ba69f9ab
JK		 106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
108};
109#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110#else
111#define IS_EXT3_SB(sb) (0)
112#endif
bd81d8ee 113
d25425f8
DW		 114static int ext4_verify_csum_type(struct super_block *sb,
115 struct ext4_super_block *es)
116{
117 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
118 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
119 return 1;
120
121 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
122}
123
a9c47317
DW		 124static __le32 ext4_superblock_csum(struct super_block *sb,
125 struct ext4_super_block *es)
126{
127 struct ext4_sb_info *sbi = EXT4_SB(sb);
128 int offset = offsetof(struct ext4_super_block, s_checksum);
129 __u32 csum;
130
131 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
132
133 return cpu_to_le32(csum);
134}
135
136int ext4_superblock_csum_verify(struct super_block *sb,
137 struct ext4_super_block *es)
138{
139 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
140 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
141 return 1;
142
143 return es->s_checksum == ext4_superblock_csum(sb, es);
144}
145
146void ext4_superblock_csum_set(struct super_block *sb,
147 struct ext4_super_block *es)
148{
149 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
150 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
151 return;
152
153 es->s_checksum = ext4_superblock_csum(sb, es);
154}
155
9933fc0a
TT		 156void *ext4_kvmalloc(size_t size, gfp_t flags)
157{
158 void *ret;
159
160 ret = kmalloc(size, flags);
161 if (!ret)
162 ret = __vmalloc(size, flags, PAGE_KERNEL);
163 return ret;
164}
165
166void *ext4_kvzalloc(size_t size, gfp_t flags)
167{
168 void *ret;
169
db9481c0 170 ret = kzalloc(size, flags);
9933fc0a
TT		 171 if (!ret)
172 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
173 return ret;
174}
175
176void ext4_kvfree(void *ptr)
177{
178 if (is_vmalloc_addr(ptr))
179 vfree(ptr);
180 else
181 kfree(ptr);
182
183}
184
8fadc143
AR		 185ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
186 struct ext4_group_desc *bg)
bd81d8ee 187{
3a14589c 188 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 189 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 190 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 191}
192
8fadc143
AR		 193ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
194 struct ext4_group_desc *bg)
bd81d8ee 195{
5272f837 196 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 197 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 198 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 199}
200
8fadc143
AR		 201ext4_fsblk_t ext4_inode_table(struct super_block *sb,
202 struct ext4_group_desc *bg)
bd81d8ee 203{
5272f837 204 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 206 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 207}
208
021b65bb
TT		 209__u32 ext4_free_group_clusters(struct super_block *sb,
210 struct ext4_group_desc *bg)
560671a0
AK		 211{
212 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 214 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 215}
216
217__u32 ext4_free_inodes_count(struct super_block *sb,
218 struct ext4_group_desc *bg)
219{
220 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 222 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 223}
224
225__u32 ext4_used_dirs_count(struct super_block *sb,
226 struct ext4_group_desc *bg)
227{
228 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 230 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 231}
232
233__u32 ext4_itable_unused_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
235{
236 return le16_to_cpu(bg->bg_itable_unused_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 238 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 239}
240
8fadc143
AR		 241void ext4_block_bitmap_set(struct super_block *sb,
242 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 243{
3a14589c 244 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 245 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
246 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 247}
248
8fadc143
AR		 249void ext4_inode_bitmap_set(struct super_block *sb,
250 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 251{
5272f837 252 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 253 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
254 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 255}
256
8fadc143
AR		 257void ext4_inode_table_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 259{
5272f837 260 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 263}
264
021b65bb
TT		 265void ext4_free_group_clusters_set(struct super_block *sb,
266 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 267{
268 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
271}
272
273void ext4_free_inodes_set(struct super_block *sb,
274 struct ext4_group_desc *bg, __u32 count)
275{
276 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
279}
280
281void ext4_used_dirs_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
283{
284 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
287}
288
289void ext4_itable_unused_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
291{
292 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
295}
296
d3d1faf6
CW		 297
298/* Just increment the non-pointer handle value */
299static handle_t *ext4_get_nojournal(void)
300{
301 handle_t *handle = current->journal_info;
302 unsigned long ref_cnt = (unsigned long)handle;
303
304 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
305
306 ref_cnt++;
307 handle = (handle_t *)ref_cnt;
308
309 current->journal_info = handle;
310 return handle;
311}
312
313
314/* Decrement the non-pointer handle value */
315static void ext4_put_nojournal(handle_t *handle)
316{
317 unsigned long ref_cnt = (unsigned long)handle;
318
319 BUG_ON(ref_cnt == 0);
320
321 ref_cnt--;
322 handle = (handle_t *)ref_cnt;
323
324 current->journal_info = handle;
325}
326
ac27a0ec 327/*
dab291af 328 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 329 *
330 * The only special thing we need to do here is to make sure that all
331 * journal_end calls result in the superblock being marked dirty, so
332 * that sync() will call the filesystem's write_super callback if
333 * appropriate.
334 */
617ba13b 335handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 336{
337 journal_t *journal;
338
12706394 339 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 340 if (sb->s_flags & MS_RDONLY)
341 return ERR_PTR(-EROFS);
342
8e8ad8a5 343 WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE);
617ba13b 344 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 345 if (!journal)
346 return ext4_get_nojournal();
347 /*
348 * Special case here: if the journal has aborted behind our
349 * backs (eg. EIO in the commit thread), then we still need to
350 * take the FS itself readonly cleanly.
351 */
352 if (is_journal_aborted(journal)) {
353 ext4_abort(sb, "Detected aborted journal");
354 return ERR_PTR(-EROFS);
ac27a0ec 355 }
be4f27d3 356 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 357}
358
359/*
360 * The only special thing we need to do here is to make sure that all
dab291af 361 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 362 * that sync() will call the filesystem's write_super callback if
363 * appropriate.
364 */
c398eda0 365int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 366{
367 struct super_block *sb;
368 int err;
369 int rc;
370
0390131b 371 if (!ext4_handle_valid(handle)) {
d3d1faf6 372 ext4_put_nojournal(handle);
0390131b
FM		 373 return 0;
374 }
ac27a0ec
DK		 375 sb = handle->h_transaction->t_journal->j_private;
376 err = handle->h_err;
dab291af 377 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 378
379 if (!err)
380 err = rc;
381 if (err)
c398eda0 382 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 383 return err;
384}
385
90c7201b
TT		 386void ext4_journal_abort_handle(const char *caller, unsigned int line,
387 const char *err_fn, struct buffer_head *bh,
388 handle_t *handle, int err)
ac27a0ec
DK		 389{
390 char nbuf[16];
617ba13b 391 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 392
0390131b
FM		 393 BUG_ON(!ext4_handle_valid(handle));
394
ac27a0ec
DK		 395 if (bh)
396 BUFFER_TRACE(bh, "abort");
397
398 if (!handle->h_err)
399 handle->h_err = err;
400
401 if (is_handle_aborted(handle))
402 return;
403
92b97816 404 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
90c7201b 405 caller, line, errstr, err_fn);
ac27a0ec 406
dab291af 407 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 408}
409
1c13d5c0
TT		 410static void __save_error_info(struct super_block *sb, const char *func,
411 unsigned int line)
412{
413 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
414
415 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
416 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
417 es->s_last_error_time = cpu_to_le32(get_seconds());
418 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
419 es->s_last_error_line = cpu_to_le32(line);
420 if (!es->s_first_error_time) {
421 es->s_first_error_time = es->s_last_error_time;
422 strncpy(es->s_first_error_func, func,
423 sizeof(es->s_first_error_func));
424 es->s_first_error_line = cpu_to_le32(line);
425 es->s_first_error_ino = es->s_last_error_ino;
426 es->s_first_error_block = es->s_last_error_block;
427 }
66e61a9e
TT		 428 /*
429 * Start the daily error reporting function if it hasn't been
430 * started already
431 */
432 if (!es->s_error_count)
433 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 434 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
435}
436
437static void save_error_info(struct super_block *sb, const char *func,
438 unsigned int line)
439{
440 __save_error_info(sb, func, line);
441 ext4_commit_super(sb, 1);
442}
443
7c2e7087
TT		 444/*
445 * The del_gendisk() function uninitializes the disk-specific data
446 * structures, including the bdi structure, without telling anyone
447 * else. Once this happens, any attempt to call mark_buffer_dirty()
448 * (for example, by ext4_commit_super), will cause a kernel OOPS.
449 * This is a kludge to prevent these oops until we can put in a proper
450 * hook in del_gendisk() to inform the VFS and file system layers.
451 */
452static int block_device_ejected(struct super_block *sb)
453{
454 struct inode *bd_inode = sb->s_bdev->bd_inode;
455 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
456
457 return bdi->dev == NULL;
458}
459
18aadd47
BJ		 460static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
461{
462 struct super_block *sb = journal->j_private;
463 struct ext4_sb_info *sbi = EXT4_SB(sb);
464 int error = is_journal_aborted(journal);
465 struct ext4_journal_cb_entry *jce, *tmp;
466
467 spin_lock(&sbi->s_md_lock);
468 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
469 list_del_init(&jce->jce_list);
470 spin_unlock(&sbi->s_md_lock);
471 jce->jce_func(sb, jce, error);
472 spin_lock(&sbi->s_md_lock);
473 }
474 spin_unlock(&sbi->s_md_lock);
475}
1c13d5c0 476
ac27a0ec
DK		 477/* Deal with the reporting of failure conditions on a filesystem such as
478 * inconsistencies detected or read IO failures.
479 *
480 * On ext2, we can store the error state of the filesystem in the
617ba13b 481 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 482 * write ordering constraints on the superblock which prevent us from
483 * writing it out straight away; and given that the journal is about to
484 * be aborted, we can't rely on the current, or future, transactions to
485 * write out the superblock safely.
486 *
dab291af 487 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 488 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 489 * that error until we've noted it down and cleared it.
490 */
491
617ba13b 492static void ext4_handle_error(struct super_block *sb)
ac27a0ec 493{
ac27a0ec
DK		 494 if (sb->s_flags & MS_RDONLY)
495 return;
496
2b2d6d01 497 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 498 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 499
4ab2f15b 500 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 501 if (journal)
dab291af 502 jbd2_journal_abort(journal, -EIO);
ac27a0ec 503 }
2b2d6d01 504 if (test_opt(sb, ERRORS_RO)) {
b31e1552 505 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 506 sb->s_flags |= MS_RDONLY;
507 }
ac27a0ec 508 if (test_opt(sb, ERRORS_PANIC))
617ba13b 509 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 510 sb->s_id);
511}
512
12062ddd 513void __ext4_error(struct super_block *sb, const char *function,
c398eda0 514 unsigned int line, const char *fmt, ...)
ac27a0ec 515{
0ff2ea7d 516 struct va_format vaf;
ac27a0ec
DK		 517 va_list args;
518
519 va_start(args, fmt);
0ff2ea7d
JP		 520 vaf.fmt = fmt;
521 vaf.va = &args;
522 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
523 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec 524 va_end(args);
f3fc0210 525 save_error_info(sb, function, line);
ac27a0ec 526
617ba13b 527 ext4_handle_error(sb);
ac27a0ec
DK		 528}
529
c398eda0
TT		 530void ext4_error_inode(struct inode *inode, const char *function,
531 unsigned int line, ext4_fsblk_t block,
273df556
FM		 532 const char *fmt, ...)
533{
534 va_list args;
f7c21177 535 struct va_format vaf;
1c13d5c0 536 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 537
1c13d5c0
TT		 538 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
539 es->s_last_error_block = cpu_to_le64(block);
540 save_error_info(inode->i_sb, function, line);
273df556 541 va_start(args, fmt);
f7c21177
TT		 542 vaf.fmt = fmt;
543 vaf.va = &args;
c398eda0 544 if (block)
d9ee81da
JP		 545 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
546 "inode #%lu: block %llu: comm %s: %pV\n",
547 inode->i_sb->s_id, function, line, inode->i_ino,
548 block, current->comm, &vaf);
549 else
550 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
551 "inode #%lu: comm %s: %pV\n",
552 inode->i_sb->s_id, function, line, inode->i_ino,
553 current->comm, &vaf);
273df556
FM		 554 va_end(args);
555
556 ext4_handle_error(inode->i_sb);
557}
558
c398eda0 559void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 560 unsigned int line, ext4_fsblk_t block,
561 const char *fmt, ...)
273df556
FM		 562{
563 va_list args;
f7c21177 564 struct va_format vaf;
1c13d5c0 565 struct ext4_super_block *es;
273df556
FM		 566 struct inode *inode = file->f_dentry->d_inode;
567 char pathname[80], *path;
568
1c13d5c0
TT		 569 es = EXT4_SB(inode->i_sb)->s_es;
570 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
571 save_error_info(inode->i_sb, function, line);
273df556 572 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 573 if (IS_ERR(path))
273df556 574 path = "(unknown)";
f7c21177
TT		 575 va_start(args, fmt);
576 vaf.fmt = fmt;
577 vaf.va = &args;
d9ee81da
JP		 578 if (block)
579 printk(KERN_CRIT
580 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
581 "block %llu: comm %s: path %s: %pV\n",
582 inode->i_sb->s_id, function, line, inode->i_ino,
583 block, current->comm, path, &vaf);
584 else
585 printk(KERN_CRIT
586 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
587 "comm %s: path %s: %pV\n",
588 inode->i_sb->s_id, function, line, inode->i_ino,
589 current->comm, path, &vaf);
273df556
FM		 590 va_end(args);
591
592 ext4_handle_error(inode->i_sb);
593}
594
2b2d6d01 595static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 596 char nbuf[16])
597{
598 char *errstr = NULL;
599
600 switch (errno) {
601 case -EIO:
602 errstr = "IO failure";
603 break;
604 case -ENOMEM:
605 errstr = "Out of memory";
606 break;
607 case -EROFS:
78f1ddbb
TT		 608 if (!sb || (EXT4_SB(sb)->s_journal &&
609 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 610 errstr = "Journal has aborted";
611 else
612 errstr = "Readonly filesystem";
613 break;
614 default:
615 /* If the caller passed in an extra buffer for unknown
616 * errors, textualise them now. Else we just return
617 * NULL. */
618 if (nbuf) {
619 /* Check for truncated error codes... */
620 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
621 errstr = nbuf;
622 }
623 break;
624 }
625
626 return errstr;
627}
628
617ba13b 629/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 630 * automatically and invokes the appropriate error response. */
631
c398eda0
TT		 632void __ext4_std_error(struct super_block *sb, const char *function,
633 unsigned int line, int errno)
ac27a0ec
DK		 634{
635 char nbuf[16];
636 const char *errstr;
637
638 /* Special case: if the error is EROFS, and we're not already
639 * inside a transaction, then there's really no point in logging
640 * an error. */
641 if (errno == -EROFS && journal_current_handle() == NULL &&
642 (sb->s_flags & MS_RDONLY))
643 return;
644
617ba13b 645 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 646 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
647 sb->s_id, function, line, errstr);
1c13d5c0 648 save_error_info(sb, function, line);
ac27a0ec 649
617ba13b 650 ext4_handle_error(sb);
ac27a0ec
DK		 651}
652
653/*
617ba13b 654 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 655 * abort function may be used to deal with unrecoverable failures such
656 * as journal IO errors or ENOMEM at a critical moment in log management.
657 *
658 * We unconditionally force the filesystem into an ABORT|READONLY state,
659 * unless the error response on the fs has been set to panic in which
660 * case we take the easy way out and panic immediately.
661 */
662
c67d859e 663void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 664 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 665{
666 va_list args;
667
1c13d5c0 668 save_error_info(sb, function, line);
ac27a0ec 669 va_start(args, fmt);
c398eda0
TT		 670 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
671 function, line);
ac27a0ec
DK		 672 vprintk(fmt, args);
673 printk("\n");
674 va_end(args);
675
1c13d5c0
TT		 676 if ((sb->s_flags & MS_RDONLY) == 0) {
677 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
678 sb->s_flags |= MS_RDONLY;
679 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
680 if (EXT4_SB(sb)->s_journal)
681 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
682 save_error_info(sb, function, line);
683 }
ac27a0ec 684 if (test_opt(sb, ERRORS_PANIC))
617ba13b 685 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 686}
687
0ff2ea7d 688void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 689{
0ff2ea7d 690 struct va_format vaf;
b31e1552
ES		 691 va_list args;
692
693 va_start(args, fmt);
0ff2ea7d
JP		 694 vaf.fmt = fmt;
695 vaf.va = &args;
696 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 697 va_end(args);
698}
699
12062ddd 700void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 701 unsigned int line, const char *fmt, ...)
ac27a0ec 702{
0ff2ea7d 703 struct va_format vaf;
ac27a0ec
DK		 704 va_list args;
705
706 va_start(args, fmt);
0ff2ea7d
JP		 707 vaf.fmt = fmt;
708 vaf.va = &args;
709 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
710 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 711 va_end(args);
712}
713
e29136f8
TT		 714void __ext4_grp_locked_error(const char *function, unsigned int line,
715 struct super_block *sb, ext4_group_t grp,
716 unsigned long ino, ext4_fsblk_t block,
717 const char *fmt, ...)
5d1b1b3f
AK		 718__releases(bitlock)
719__acquires(bitlock)
720{
0ff2ea7d 721 struct va_format vaf;
5d1b1b3f
AK		 722 va_list args;
723 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
724
1c13d5c0
TT		 725 es->s_last_error_ino = cpu_to_le32(ino);
726 es->s_last_error_block = cpu_to_le64(block);
727 __save_error_info(sb, function, line);
0ff2ea7d 728
5d1b1b3f 729 va_start(args, fmt);
0ff2ea7d
JP		 730
731 vaf.fmt = fmt;
732 vaf.va = &args;
21149d61 733 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 734 sb->s_id, function, line, grp);
735 if (ino)
0ff2ea7d 736 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 737 if (block)
0ff2ea7d
JP		 738 printk(KERN_CONT "block %llu:", (unsigned long long) block);
739 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 740 va_end(args);
741
742 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 743 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 744 return;
745 }
1c13d5c0 746
5d1b1b3f
AK		 747 ext4_unlock_group(sb, grp);
748 ext4_handle_error(sb);
749 /*
750 * We only get here in the ERRORS_RO case; relocking the group
751 * may be dangerous, but nothing bad will happen since the
752 * filesystem will have already been marked read/only and the
753 * journal has been aborted. We return 1 as a hint to callers
754 * who might what to use the return value from
25985edc 755 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 756 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
757 * aggressively from the ext4 function in question, with a
758 * more appropriate error code.
759 */
760 ext4_lock_group(sb, grp);
761 return;
762}
763
617ba13b 764void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 765{
617ba13b 766 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 767
617ba13b 768 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 769 return;
770
12062ddd 771 ext4_warning(sb,
ac27a0ec
DK		 772 "updating to rev %d because of new feature flag, "
773 "running e2fsck is recommended",
617ba13b 774 EXT4_DYNAMIC_REV);
ac27a0ec 775
617ba13b
MC		 776 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
777 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
778 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 779 /* leave es->s_feature_*compat flags alone */
780 /* es->s_uuid will be set by e2fsck if empty */
781
782 /*
783 * The rest of the superblock fields should be zero, and if not it
784 * means they are likely already in use, so leave them alone. We
785 * can leave it up to e2fsck to clean up any inconsistencies there.
786 */
787}
788
789/*
790 * Open the external journal device
791 */
b31e1552 792static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 793{
794 struct block_device *bdev;
795 char b[BDEVNAME_SIZE];
796
d4d77629 797 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 798 if (IS_ERR(bdev))
799 goto fail;
800 return bdev;
801
802fail:
b31e1552 803 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 804 __bdevname(dev, b), PTR_ERR(bdev));
805 return NULL;
806}
807
808/*
809 * Release the journal device
810 */
617ba13b 811static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 812{
e525fd89 813 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 814}
815
617ba13b 816static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 817{
818 struct block_device *bdev;
819 int ret = -ENODEV;
820
821 bdev = sbi->journal_bdev;
822 if (bdev) {
617ba13b 823 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 824 sbi->journal_bdev = NULL;
825 }
826 return ret;
827}
828
829static inline struct inode *orphan_list_entry(struct list_head *l)
830{
617ba13b 831 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 832}
833
617ba13b 834static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 835{
836 struct list_head *l;
837
b31e1552
ES		 838 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
839 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 840
841 printk(KERN_ERR "sb_info orphan list:\n");
842 list_for_each(l, &sbi->s_orphan) {
843 struct inode *inode = orphan_list_entry(l);
844 printk(KERN_ERR " "
845 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
846 inode->i_sb->s_id, inode->i_ino, inode,
847 inode->i_mode, inode->i_nlink,
848 NEXT_ORPHAN(inode));
849 }
850}
851
2b2d6d01 852static void ext4_put_super(struct super_block *sb)
ac27a0ec 853{
617ba13b
MC		 854 struct ext4_sb_info *sbi = EXT4_SB(sb);
855 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 856 int i, err;
ac27a0ec 857
857ac889 858 ext4_unregister_li_request(sb);
e0ccfd95
CH		 859 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
860
4c0425ff
MC		 861 flush_workqueue(sbi->dio_unwritten_wq);
862 destroy_workqueue(sbi->dio_unwritten_wq);
863
a9e220f8 864 lock_super(sb);
0390131b
FM		 865 if (sbi->s_journal) {
866 err = jbd2_journal_destroy(sbi->s_journal);
867 sbi->s_journal = NULL;
868 if (err < 0)
c67d859e 869 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 870 }
d4edac31 871
a1c6c569 872 del_timer(&sbi->s_err_report);
d4edac31
JB		 873 ext4_release_system_zone(sb);
874 ext4_mb_release(sb);
875 ext4_ext_release(sb);
876 ext4_xattr_put_super(sb);
877
ac27a0ec 878 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 879 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 880 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 881 }
58c5873a 882 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 883 ext4_commit_super(sb, 1);
884
240799cd 885 if (sbi->s_proc) {
66acdcf4 886 remove_proc_entry("options", sbi->s_proc);
9f6200bb 887 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 888 }
3197ebdb 889 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 890
891 for (i = 0; i < sbi->s_gdb_count; i++)
892 brelse(sbi->s_group_desc[i]);
f18a5f21 893 ext4_kvfree(sbi->s_group_desc);
9933fc0a 894 ext4_kvfree(sbi->s_flex_groups);
57042651 895 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 896 percpu_counter_destroy(&sbi->s_freeinodes_counter);
897 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 898 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 899 brelse(sbi->s_sbh);
900#ifdef CONFIG_QUOTA
901 for (i = 0; i < MAXQUOTAS; i++)
902 kfree(sbi->s_qf_names[i]);
903#endif
904
905 /* Debugging code just in case the in-memory inode orphan list
906 * isn't empty. The on-disk one can be non-empty if we've
907 * detected an error and taken the fs readonly, but the
908 * in-memory list had better be clean by this point. */
909 if (!list_empty(&sbi->s_orphan))
910 dump_orphan_list(sb, sbi);
911 J_ASSERT(list_empty(&sbi->s_orphan));
912
f98393a6 913 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 914 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
915 /*
916 * Invalidate the journal device's buffers. We don't want them
917 * floating about in memory - the physical journal device may
918 * hotswapped, and it breaks the `ro-after' testing code.
919 */
920 sync_blockdev(sbi->journal_bdev);
f98393a6 921 invalidate_bdev(sbi->journal_bdev);
617ba13b 922 ext4_blkdev_remove(sbi);
ac27a0ec 923 }
c5e06d10
JL		 924 if (sbi->s_mmp_tsk)
925 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 926 sb->s_fs_info = NULL;
3197ebdb
TT		 927 /*
928 * Now that we are completely done shutting down the
929 * superblock, we need to actually destroy the kobject.
930 */
3197ebdb
TT		 931 unlock_super(sb);
932 kobject_put(&sbi->s_kobj);
933 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 934 if (sbi->s_chksum_driver)
935 crypto_free_shash(sbi->s_chksum_driver);
705895b6 936 kfree(sbi->s_blockgroup_lock);
ac27a0ec 937 kfree(sbi);
ac27a0ec
DK		 938}
939
e18b890b 940static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 941
942/*
943 * Called inside transaction, so use GFP_NOFS
944 */
617ba13b 945static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 946{
617ba13b 947 struct ext4_inode_info *ei;
ac27a0ec 948
e6b4f8da 949 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 950 if (!ei)
951 return NULL;
0b8e58a1 952
ac27a0ec 953 ei->vfs_inode.i_version = 1;
91246c00 954 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 955 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 956 INIT_LIST_HEAD(&ei->i_prealloc_list);
957 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 958 ei->i_reserved_data_blocks = 0;
959 ei->i_reserved_meta_blocks = 0;
960 ei->i_allocated_meta_blocks = 0;
9d0be502 961 ei->i_da_metadata_calc_len = 0;
7e731bc9 962 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 963 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 964#ifdef CONFIG_QUOTA
965 ei->i_reserved_quota = 0;
966#endif
8aefcd55 967 ei->jinode = NULL;
c7064ef1 968 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 969 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 970 ei->cur_aio_dio = NULL;
b436b9be
JK		 971 ei->i_sync_tid = 0;
972 ei->i_datasync_tid = 0;
f7ad6d2e 973 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 974 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 975
ac27a0ec
DK		 976 return &ei->vfs_inode;
977}
978
7ff9c073
TT		 979static int ext4_drop_inode(struct inode *inode)
980{
981 int drop = generic_drop_inode(inode);
982
983 trace_ext4_drop_inode(inode, drop);
984 return drop;
985}
986
fa0d7e3d
NP		 987static void ext4_i_callback(struct rcu_head *head)
988{
989 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 990 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
991}
992
617ba13b 993static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 994{
9f7dd93d 995 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 996 ext4_msg(inode->i_sb, KERN_ERR,
997 "Inode %lu (%p): orphan list check failed!",
998 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 999 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1000 EXT4_I(inode), sizeof(struct ext4_inode_info),
1001 true);
1002 dump_stack();
1003 }
fa0d7e3d 1004 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1005}
1006
51cc5068 1007static void init_once(void *foo)
ac27a0ec 1008{
617ba13b 1009 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1010
a35afb83 1011 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 1012#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 1013 init_rwsem(&ei->xattr_sem);
ac27a0ec 1014#endif
0e855ac8 1015 init_rwsem(&ei->i_data_sem);
a35afb83 1016 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1017}
1018
1019static int init_inodecache(void)
1020{
617ba13b
MC		 1021 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1022 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 1023 0, (SLAB_RECLAIM_ACCOUNT|
1024 SLAB_MEM_SPREAD),
20c2df83 1025 init_once);
617ba13b 1026 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1027 return -ENOMEM;
1028 return 0;
1029}
1030
1031static void destroy_inodecache(void)
1032{
617ba13b 1033 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1034}
1035
0930fcc1 1036void ext4_clear_inode(struct inode *inode)
ac27a0ec 1037{
0930fcc1 1038 invalidate_inode_buffers(inode);
dbd5768f 1039 clear_inode(inode);
9f754758 1040 dquot_drop(inode);
c2ea3fde 1041 ext4_discard_preallocations(inode);
8aefcd55
TT		 1042 if (EXT4_I(inode)->jinode) {
1043 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1044 EXT4_I(inode)->jinode);
1045 jbd2_free_inode(EXT4_I(inode)->jinode);
1046 EXT4_I(inode)->jinode = NULL;
1047 }
ac27a0ec
DK		 1048}
1049
1b961ac0 1050static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1051 u64 ino, u32 generation)
ac27a0ec 1052{
ac27a0ec 1053 struct inode *inode;
ac27a0ec 1054
617ba13b 1055 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1056 return ERR_PTR(-ESTALE);
617ba13b 1057 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1058 return ERR_PTR(-ESTALE);
1059
1060 /* iget isn't really right if the inode is currently unallocated!!
1061 *
617ba13b 1062 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1063 * deleted, so we should be safe.
1064 *
1065 * Currently we don't know the generation for parent directory, so
1066 * a generation of 0 means "accept any"
1067 */
1d1fe1ee
DH		 1068 inode = ext4_iget(sb, ino);
1069 if (IS_ERR(inode))
1070 return ERR_CAST(inode);
1071 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1072 iput(inode);
1073 return ERR_PTR(-ESTALE);
1074 }
1b961ac0
CH		 1075
1076 return inode;
1077}
1078
1079static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1080 int fh_len, int fh_type)
1b961ac0
CH		 1081{
1082 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1083 ext4_nfs_get_inode);
1084}
1085
1086static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1087 int fh_len, int fh_type)
1b961ac0
CH		 1088{
1089 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1090 ext4_nfs_get_inode);
ac27a0ec
DK		 1091}
1092
c39a7f84
TO		 1093/*
1094 * Try to release metadata pages (indirect blocks, directories) which are
1095 * mapped via the block device. Since these pages could have journal heads
1096 * which would prevent try_to_free_buffers() from freeing them, we must use
1097 * jbd2 layer's try_to_free_buffers() function to release them.
1098 */
0b8e58a1
AD		 1099static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1100 gfp_t wait)
c39a7f84
TO		 1101{
1102 journal_t *journal = EXT4_SB(sb)->s_journal;
1103
1104 WARN_ON(PageChecked(page));
1105 if (!page_has_buffers(page))
1106 return 0;
1107 if (journal)
1108 return jbd2_journal_try_to_free_buffers(journal, page,
1109 wait & ~__GFP_WAIT);
1110 return try_to_free_buffers(page);
1111}
1112
ac27a0ec 1113#ifdef CONFIG_QUOTA
af5bc92d 1114#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1115#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1116
617ba13b
MC		 1117static int ext4_write_dquot(struct dquot *dquot);
1118static int ext4_acquire_dquot(struct dquot *dquot);
1119static int ext4_release_dquot(struct dquot *dquot);
1120static int ext4_mark_dquot_dirty(struct dquot *dquot);
1121static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1122static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1123 struct path *path);
7c319d32
AK		 1124static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1125 int format_id);
ca0e05e4 1126static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1127static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1128static int ext4_quota_on_mount(struct super_block *sb, int type);
1129static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1130 size_t len, loff_t off);
617ba13b 1131static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1132 const char *data, size_t len, loff_t off);
7c319d32
AK		 1133static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1134 unsigned int flags);
1135static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1136
61e225dc 1137static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1138 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1139 .write_dquot = ext4_write_dquot,
1140 .acquire_dquot = ext4_acquire_dquot,
1141 .release_dquot = ext4_release_dquot,
1142 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1143 .write_info = ext4_write_info,
1144 .alloc_dquot = dquot_alloc,
1145 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1146};
1147
0d54b217 1148static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1149 .quota_on = ext4_quota_on,
ca0e05e4 1150 .quota_off = ext4_quota_off,
287a8095
CH		 1151 .quota_sync = dquot_quota_sync,
1152 .get_info = dquot_get_dqinfo,
1153 .set_info = dquot_set_dqinfo,
1154 .get_dqblk = dquot_get_dqblk,
1155 .set_dqblk = dquot_set_dqblk
ac27a0ec 1156};
7c319d32
AK		 1157
1158static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1159 .quota_on_meta = ext4_quota_on_sysfile,
1160 .quota_off = ext4_quota_off_sysfile,
1161 .quota_sync = dquot_quota_sync,
1162 .get_info = dquot_get_dqinfo,
1163 .set_info = dquot_set_dqinfo,
1164 .get_dqblk = dquot_get_dqblk,
1165 .set_dqblk = dquot_set_dqblk
1166};
ac27a0ec
DK		 1167#endif
1168
ee9b6d61 1169static const struct super_operations ext4_sops = {
617ba13b
MC		 1170 .alloc_inode = ext4_alloc_inode,
1171 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1172 .write_inode = ext4_write_inode,
1173 .dirty_inode = ext4_dirty_inode,
7ff9c073 1174 .drop_inode = ext4_drop_inode,
0930fcc1 1175 .evict_inode = ext4_evict_inode,
617ba13b 1176 .put_super = ext4_put_super,
617ba13b 1177 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1178 .freeze_fs = ext4_freeze,
1179 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1180 .statfs = ext4_statfs,
1181 .remount_fs = ext4_remount,
617ba13b 1182 .show_options = ext4_show_options,
ac27a0ec 1183#ifdef CONFIG_QUOTA
617ba13b
MC		 1184 .quota_read = ext4_quota_read,
1185 .quota_write = ext4_quota_write,
ac27a0ec 1186#endif
c39a7f84 1187 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1188};
1189
9ca92389
TT		 1190static const struct super_operations ext4_nojournal_sops = {
1191 .alloc_inode = ext4_alloc_inode,
1192 .destroy_inode = ext4_destroy_inode,
1193 .write_inode = ext4_write_inode,
1194 .dirty_inode = ext4_dirty_inode,
7ff9c073 1195 .drop_inode = ext4_drop_inode,
0930fcc1 1196 .evict_inode = ext4_evict_inode,
9ca92389
TT		 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,
df981d03 1233 Opt_max_dir_size_kb,
ac27a0ec
DK		 1234};
1235
a447c093 1236static const match_table_t tokens = {
ac27a0ec
DK		 1237 {Opt_bsd_df, "bsddf"},
1238 {Opt_minix_df, "minixdf"},
1239 {Opt_grpid, "grpid"},
1240 {Opt_grpid, "bsdgroups"},
1241 {Opt_nogrpid, "nogrpid"},
1242 {Opt_nogrpid, "sysvgroups"},
1243 {Opt_resgid, "resgid=%u"},
1244 {Opt_resuid, "resuid=%u"},
1245 {Opt_sb, "sb=%u"},
1246 {Opt_err_cont, "errors=continue"},
1247 {Opt_err_panic, "errors=panic"},
1248 {Opt_err_ro, "errors=remount-ro"},
1249 {Opt_nouid32, "nouid32"},
ac27a0ec 1250 {Opt_debug, "debug"},
72578c33
TT		 1251 {Opt_removed, "oldalloc"},
1252 {Opt_removed, "orlov"},
ac27a0ec
DK		 1253 {Opt_user_xattr, "user_xattr"},
1254 {Opt_nouser_xattr, "nouser_xattr"},
1255 {Opt_acl, "acl"},
1256 {Opt_noacl, "noacl"},
e3bb52ae 1257 {Opt_noload, "norecovery"},
5a916be1 1258 {Opt_noload, "noload"},
72578c33
TT		 1259 {Opt_removed, "nobh"},
1260 {Opt_removed, "bh"},
ac27a0ec 1261 {Opt_commit, "commit=%u"},
30773840
TT		 1262 {Opt_min_batch_time, "min_batch_time=%u"},
1263 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1264 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1265 {Opt_journal_checksum, "journal_checksum"},
1266 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1267 {Opt_abort, "abort"},
1268 {Opt_data_journal, "data=journal"},
1269 {Opt_data_ordered, "data=ordered"},
1270 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1271 {Opt_data_err_abort, "data_err=abort"},
1272 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1273 {Opt_offusrjquota, "usrjquota="},
1274 {Opt_usrjquota, "usrjquota=%s"},
1275 {Opt_offgrpjquota, "grpjquota="},
1276 {Opt_grpjquota, "grpjquota=%s"},
1277 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1278 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1279 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1280 {Opt_grpquota, "grpquota"},
1281 {Opt_noquota, "noquota"},
1282 {Opt_quota, "quota"},
1283 {Opt_usrquota, "usrquota"},
1284 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1285 {Opt_barrier, "barrier"},
1286 {Opt_nobarrier, "nobarrier"},
25ec56b5 1287 {Opt_i_version, "i_version"},
c9de560d 1288 {Opt_stripe, "stripe=%u"},
64769240 1289 {Opt_delalloc, "delalloc"},
dd919b98 1290 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1291 {Opt_mblk_io_submit, "mblk_io_submit"},
1292 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1293 {Opt_block_validity, "block_validity"},
1294 {Opt_noblock_validity, "noblock_validity"},
240799cd 1295 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1296 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1297 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1298 {Opt_auto_da_alloc, "auto_da_alloc"},
1299 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1300 {Opt_dioread_nolock, "dioread_nolock"},
1301 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1302 {Opt_discard, "discard"},
1303 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1304 {Opt_init_itable, "init_itable=%u"},
1305 {Opt_init_itable, "init_itable"},
1306 {Opt_noinit_itable, "noinit_itable"},
df981d03 1307 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1308 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1309 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1310 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1311 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1312 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1313 {Opt_err, NULL},
ac27a0ec
DK		 1314};
1315
617ba13b 1316static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1317{
617ba13b 1318 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1319 char *options = (char *) *data;
1320
1321 if (!options || strncmp(options, "sb=", 3) != 0)
1322 return 1; /* Default location */
0b8e58a1 1323
ac27a0ec 1324 options += 3;
0b8e58a1 1325 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1326 sb_block = simple_strtoul(options, &options, 0);
1327 if (*options && *options != ',') {
4776004f 1328 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1329 (char *) *data);
1330 return 1;
1331 }
1332 if (*options == ',')
1333 options++;
1334 *data = (void *) options;
0b8e58a1 1335
ac27a0ec
DK		 1336 return sb_block;
1337}
1338
b3881f74 1339#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1340static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1341 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1342
56c50f11
DM		 1343#ifdef CONFIG_QUOTA
1344static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1345{
1346 struct ext4_sb_info *sbi = EXT4_SB(sb);
1347 char *qname;
1348
1349 if (sb_any_quota_loaded(sb) &&
1350 !sbi->s_qf_names[qtype]) {
1351 ext4_msg(sb, KERN_ERR,
1352 "Cannot change journaled "
1353 "quota options when quota turned on");
57f73c2c 1354 return -1;
56c50f11
DM		 1355 }
1356 qname = match_strdup(args);
1357 if (!qname) {
1358 ext4_msg(sb, KERN_ERR,
1359 "Not enough memory for storing quotafile name");
57f73c2c 1360 return -1;
56c50f11
DM		 1361 }
1362 if (sbi->s_qf_names[qtype] &&
1363 strcmp(sbi->s_qf_names[qtype], qname)) {
1364 ext4_msg(sb, KERN_ERR,
1365 "%s quota file already specified", QTYPE2NAME(qtype));
1366 kfree(qname);
57f73c2c 1367 return -1;
56c50f11
DM		 1368 }
1369 sbi->s_qf_names[qtype] = qname;
1370 if (strchr(sbi->s_qf_names[qtype], '/')) {
1371 ext4_msg(sb, KERN_ERR,
1372 "quotafile must be on filesystem root");
1373 kfree(sbi->s_qf_names[qtype]);
1374 sbi->s_qf_names[qtype] = NULL;
57f73c2c 1375 return -1;
56c50f11 1376 }
fd8c37ec 1377 set_opt(sb, QUOTA);
56c50f11
DM		 1378 return 1;
1379}
1380
1381static int clear_qf_name(struct super_block *sb, int qtype)
1382{
1383
1384 struct ext4_sb_info *sbi = EXT4_SB(sb);
1385
1386 if (sb_any_quota_loaded(sb) &&
1387 sbi->s_qf_names[qtype]) {
1388 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1389 " when quota turned on");
57f73c2c 1390 return -1;
56c50f11
DM		 1391 }
1392 /*
1393 * The space will be released later when all options are confirmed
1394 * to be correct
1395 */
1396 sbi->s_qf_names[qtype] = NULL;
1397 return 1;
1398}
1399#endif
1400
26092bf5
TT		 1401#define MOPT_SET 0x0001
1402#define MOPT_CLEAR 0x0002
1403#define MOPT_NOSUPPORT 0x0004
1404#define MOPT_EXPLICIT 0x0008
1405#define MOPT_CLEAR_ERR 0x0010
1406#define MOPT_GTE0 0x0020
ac27a0ec 1407#ifdef CONFIG_QUOTA
26092bf5
TT		 1408#define MOPT_Q 0
1409#define MOPT_QFMT 0x0040
1410#else
1411#define MOPT_Q MOPT_NOSUPPORT
1412#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1413#endif
26092bf5
TT		 1414#define MOPT_DATAJ 0x0080
1415
1416static const struct mount_opts {
1417 int token;
1418 int mount_opt;
1419 int flags;
1420} ext4_mount_opts[] = {
1421 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1422 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1423 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1424 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1425 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1426 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1427 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1428 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1429 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1430 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1431 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1432 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1433 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1434 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1435 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1436 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1437 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1438 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1439 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1440 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1441 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1442 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1443 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1444 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1445 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1446 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1447 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1448 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1449 {Opt_commit, 0, MOPT_GTE0},
1450 {Opt_max_batch_time, 0, MOPT_GTE0},
1451 {Opt_min_batch_time, 0, MOPT_GTE0},
1452 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1453 {Opt_init_itable, 0, MOPT_GTE0},
1454 {Opt_stripe, 0, MOPT_GTE0},
1455 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1456 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1457 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
03010a33 1458#ifdef CONFIG_EXT4_FS_XATTR
26092bf5
TT		 1459 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1460 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
ac27a0ec 1461#else
26092bf5
TT		 1462 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1463 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
ac27a0ec 1464#endif
03010a33 1465#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1466 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1467 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1468#else
26092bf5
TT		 1469 {Opt_acl, 0, MOPT_NOSUPPORT},
1470 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1471#endif
26092bf5
TT		 1472 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1473 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1474 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1475 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1476 MOPT_SET | MOPT_Q},
1477 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1478 MOPT_SET | MOPT_Q},
1479 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1480 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1481 {Opt_usrjquota, 0, MOPT_Q},
1482 {Opt_grpjquota, 0, MOPT_Q},
1483 {Opt_offusrjquota, 0, MOPT_Q},
1484 {Opt_offgrpjquota, 0, MOPT_Q},
1485 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1486 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1487 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1488 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1489 {Opt_err, 0, 0}
1490};
1491
1492static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1493 substring_t *args, unsigned long *journal_devnum,
1494 unsigned int *journal_ioprio, int is_remount)
1495{
1496 struct ext4_sb_info *sbi = EXT4_SB(sb);
1497 const struct mount_opts *m;
08cefc7a
EB		 1498 kuid_t uid;
1499 kgid_t gid;
26092bf5
TT		 1500 int arg = 0;
1501
57f73c2c
TT		 1502#ifdef CONFIG_QUOTA
1503 if (token == Opt_usrjquota)
1504 return set_qf_name(sb, USRQUOTA, &args[0]);
1505 else if (token == Opt_grpjquota)
1506 return set_qf_name(sb, GRPQUOTA, &args[0]);
1507 else if (token == Opt_offusrjquota)
1508 return clear_qf_name(sb, USRQUOTA);
1509 else if (token == Opt_offgrpjquota)
1510 return clear_qf_name(sb, GRPQUOTA);
1511#endif
26092bf5
TT		 1512 if (args->from && match_int(args, &arg))
1513 return -1;
1514 switch (token) {
f7048605
TT		 1515 case Opt_noacl:
1516 case Opt_nouser_xattr:
1517 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1518 break;
26092bf5
TT		 1519 case Opt_sb:
1520 return 1; /* handled by get_sb_block() */
1521 case Opt_removed:
1522 ext4_msg(sb, KERN_WARNING,
1523 "Ignoring removed %s option", opt);
1524 return 1;
1525 case Opt_resuid:
08cefc7a
EB		 1526 uid = make_kuid(current_user_ns(), arg);
1527 if (!uid_valid(uid)) {
1528 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1529 return -1;
1530 }
1531 sbi->s_resuid = uid;
26092bf5
TT		 1532 return 1;
1533 case Opt_resgid:
08cefc7a
EB		 1534 gid = make_kgid(current_user_ns(), arg);
1535 if (!gid_valid(gid)) {
1536 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1537 return -1;
1538 }
1539 sbi->s_resgid = gid;
26092bf5
TT		 1540 return 1;
1541 case Opt_abort:
1542 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1543 return 1;
1544 case Opt_i_version:
1545 sb->s_flags |= MS_I_VERSION;
1546 return 1;
1547 case Opt_journal_dev:
1548 if (is_remount) {
1549 ext4_msg(sb, KERN_ERR,
1550 "Cannot specify journal on remount");
1551 return -1;
1552 }
1553 *journal_devnum = arg;
1554 return 1;
1555 case Opt_journal_ioprio:
1556 if (arg < 0 || arg > 7)
1557 return -1;
1558 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1559 return 1;
1560 }
1561
1562 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1563 if (token != m->token)
1564 continue;
1565 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1566 return -1;
1567 if (m->flags & MOPT_EXPLICIT)
1568 set_opt2(sb, EXPLICIT_DELALLOC);
1569 if (m->flags & MOPT_CLEAR_ERR)
1570 clear_opt(sb, ERRORS_MASK);
1571 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1572 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1573 "options when quota turned on");
1574 return -1;
1575 }
1576
1577 if (m->flags & MOPT_NOSUPPORT) {
1578 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1579 } else if (token == Opt_commit) {
1580 if (arg == 0)
1581 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1582 sbi->s_commit_interval = HZ * arg;
1583 } else if (token == Opt_max_batch_time) {
1584 if (arg == 0)
1585 arg = EXT4_DEF_MAX_BATCH_TIME;
1586 sbi->s_max_batch_time = arg;
1587 } else if (token == Opt_min_batch_time) {
1588 sbi->s_min_batch_time = arg;
1589 } else if (token == Opt_inode_readahead_blks) {
1590 if (arg > (1 << 30))
1591 return -1;
1592 if (arg && !is_power_of_2(arg)) {
b31e1552 1593 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1594 "EXT4-fs: inode_readahead_blks"
1595 " must be a power of 2");
1596 return -1;
ac27a0ec 1597 }
26092bf5
TT		 1598 sbi->s_inode_readahead_blks = arg;
1599 } else if (token == Opt_init_itable) {
1600 set_opt(sb, INIT_INODE_TABLE);
1601 if (!args->from)
1602 arg = EXT4_DEF_LI_WAIT_MULT;
1603 sbi->s_li_wait_mult = arg;
df981d03
TT		 1604 } else if (token == Opt_max_dir_size_kb) {
1605 sbi->s_max_dir_size_kb = arg;
26092bf5
TT		 1606 } else if (token == Opt_stripe) {
1607 sbi->s_stripe = arg;
1608 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1609 if (is_remount) {
eb513689
TT		 1610 if (!sbi->s_journal)
1611 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1612 else if (test_opt(sb, DATA_FLAGS) !=
1613 m->mount_opt) {
b31e1552 1614 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1615 "Cannot change data mode on remount");
1616 return -1;
ac27a0ec
DK		 1617 }
1618 } else {
fd8c37ec 1619 clear_opt(sb, DATA_FLAGS);
26092bf5 1620 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1621 }
ac27a0ec 1622#ifdef CONFIG_QUOTA
26092bf5 1623 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1624 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1625 sbi->s_jquota_fmt != m->mount_opt) {
1626 ext4_msg(sb, KERN_ERR, "Cannot "
1627 "change journaled quota options "
1628 "when quota turned on");
1629 return -1;
ac27a0ec 1630 }
26092bf5 1631 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1632#endif
26092bf5
TT		 1633 } else {
1634 if (!args->from)
1635 arg = 1;
1636 if (m->flags & MOPT_CLEAR)
1637 arg = !arg;
1638 else if (unlikely(!(m->flags & MOPT_SET))) {
1639 ext4_msg(sb, KERN_WARNING,
1640 "buggy handling of option %s", opt);
1641 WARN_ON(1);
1642 return -1;
3197ebdb 1643 }
26092bf5
TT		 1644 if (arg != 0)
1645 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1646 else
26092bf5 1647 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1648 }
26092bf5
TT		 1649 return 1;
1650 }
1651 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1652 "or missing value", opt);
1653 return -1;
1654}
1655
1656static int parse_options(char *options, struct super_block *sb,
1657 unsigned long *journal_devnum,
1658 unsigned int *journal_ioprio,
1659 int is_remount)
1660{
db7e5c66 1661#ifdef CONFIG_QUOTA
26092bf5 1662 struct ext4_sb_info *sbi = EXT4_SB(sb);
db7e5c66 1663#endif
26092bf5
TT		 1664 char *p;
1665 substring_t args[MAX_OPT_ARGS];
1666 int token;
1667
1668 if (!options)
1669 return 1;
1670
1671 while ((p = strsep(&options, ",")) != NULL) {
1672 if (!*p)
1673 continue;
1674 /*
1675 * Initialize args struct so we know whether arg was
1676 * found; some options take optional arguments.
1677 */
1678 args[0].to = args[0].from = 0;
1679 token = match_token(p, tokens, args);
1680 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1681 journal_ioprio, is_remount) < 0)
1682 return 0;
ac27a0ec
DK		 1683 }
1684#ifdef CONFIG_QUOTA
1685 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1686 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1687 clear_opt(sb, USRQUOTA);
ac27a0ec 1688
482a7425 1689 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1690 clear_opt(sb, GRPQUOTA);
ac27a0ec 1691
56c50f11 1692 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1693 ext4_msg(sb, KERN_ERR, "old and new quota "
1694 "format mixing");
ac27a0ec
DK		 1695 return 0;
1696 }
1697
1698 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1699 ext4_msg(sb, KERN_ERR, "journaled quota format "
1700 "not specified");
ac27a0ec
DK		 1701 return 0;
1702 }
1703 } else {
1704 if (sbi->s_jquota_fmt) {
b31e1552 1705 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1706 "specified with no journaling "
b31e1552 1707 "enabled");
ac27a0ec
DK		 1708 return 0;
1709 }
1710 }
1711#endif
1712 return 1;
1713}
1714
2adf6da8
TT		 1715static inline void ext4_show_quota_options(struct seq_file *seq,
1716 struct super_block *sb)
1717{
1718#if defined(CONFIG_QUOTA)
1719 struct ext4_sb_info *sbi = EXT4_SB(sb);
1720
1721 if (sbi->s_jquota_fmt) {
1722 char *fmtname = "";
1723
1724 switch (sbi->s_jquota_fmt) {
1725 case QFMT_VFS_OLD:
1726 fmtname = "vfsold";
1727 break;
1728 case QFMT_VFS_V0:
1729 fmtname = "vfsv0";
1730 break;
1731 case QFMT_VFS_V1:
1732 fmtname = "vfsv1";
1733 break;
1734 }
1735 seq_printf(seq, ",jqfmt=%s", fmtname);
1736 }
1737
1738 if (sbi->s_qf_names[USRQUOTA])
1739 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1740
1741 if (sbi->s_qf_names[GRPQUOTA])
1742 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1743
1744 if (test_opt(sb, USRQUOTA))
1745 seq_puts(seq, ",usrquota");
1746
1747 if (test_opt(sb, GRPQUOTA))
1748 seq_puts(seq, ",grpquota");
1749#endif
1750}
1751
5a916be1
TT		 1752static const char *token2str(int token)
1753{
1754 static const struct match_token *t;
1755
1756 for (t = tokens; t->token != Opt_err; t++)
1757 if (t->token == token && !strchr(t->pattern, '='))
1758 break;
1759 return t->pattern;
1760}
1761
2adf6da8
TT		 1762/*
1763 * Show an option if
1764 * - it's set to a non-default value OR
1765 * - if the per-sb default is different from the global default
1766 */
66acdcf4
TT		 1767static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1768 int nodefs)
2adf6da8 1769{
2adf6da8
TT		 1770 struct ext4_sb_info *sbi = EXT4_SB(sb);
1771 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1772 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1773 const struct mount_opts *m;
66acdcf4 1774 char sep = nodefs ? '\n' : ',';
2adf6da8 1775
66acdcf4
TT		 1776#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1777#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1778
1779 if (sbi->s_sb_block != 1)
5a916be1
TT		 1780 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1781
1782 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1783 int want_set = m->flags & MOPT_SET;
1784 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1785 (m->flags & MOPT_CLEAR_ERR))
1786 continue;
66acdcf4 1787 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1788 continue; /* skip if same as the default */
1789 if ((want_set &&
1790 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1791 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1792 continue; /* select Opt_noFoo vs Opt_Foo */
1793 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1794 }
5a916be1 1795
08cefc7a 1796 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1797 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1798 SEQ_OPTS_PRINT("resuid=%u",
1799 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1800 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1801 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1802 SEQ_OPTS_PRINT("resgid=%u",
1803 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1804 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1805 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1806 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1807 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1808 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1809 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1810 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1811 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1812 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1813 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1814 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1815 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1816 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1817 if (sb->s_flags & MS_I_VERSION)
5a916be1 1818 SEQ_OPTS_PUTS("i_version");
66acdcf4 1819 if (nodefs || sbi->s_stripe)
5a916be1 1820 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1821 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1822 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1823 SEQ_OPTS_PUTS("data=journal");
1824 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1825 SEQ_OPTS_PUTS("data=ordered");
1826 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1827 SEQ_OPTS_PUTS("data=writeback");
1828 }
66acdcf4
TT		 1829 if (nodefs ||
1830 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1831 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1832 sbi->s_inode_readahead_blks);
2adf6da8 1833
66acdcf4
TT		 1834 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1835 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1836 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1837 if (nodefs || sbi->s_max_dir_size_kb)
1838 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1839
1840 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1841 return 0;
1842}
1843
66acdcf4
TT		 1844static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1845{
1846 return _ext4_show_options(seq, root->d_sb, 0);
1847}
1848
1849static int options_seq_show(struct seq_file *seq, void *offset)
1850{
1851 struct super_block *sb = seq->private;
1852 int rc;
1853
1854 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1855 rc = _ext4_show_options(seq, sb, 1);
1856 seq_puts(seq, "\n");
1857 return rc;
1858}
1859
1860static int options_open_fs(struct inode *inode, struct file *file)
1861{
1862 return single_open(file, options_seq_show, PDE(inode)->data);
1863}
1864
1865static const struct file_operations ext4_seq_options_fops = {
1866 .owner = THIS_MODULE,
1867 .open = options_open_fs,
1868 .read = seq_read,
1869 .llseek = seq_lseek,
1870 .release = single_release,
1871};
1872
617ba13b 1873static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1874 int read_only)
1875{
617ba13b 1876 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1877 int res = 0;
1878
617ba13b 1879 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1880 ext4_msg(sb, KERN_ERR, "revision level too high, "
1881 "forcing read-only mode");
ac27a0ec
DK		 1882 res = MS_RDONLY;
1883 }
1884 if (read_only)
281b5995 1885 goto done;
617ba13b 1886 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1887 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1888 "running e2fsck is recommended");
617ba13b 1889 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1890 ext4_msg(sb, KERN_WARNING,
1891 "warning: mounting fs with errors, "
1892 "running e2fsck is recommended");
ed3ce80a 1893 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1894 le16_to_cpu(es->s_mnt_count) >=
1895 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1896 ext4_msg(sb, KERN_WARNING,
1897 "warning: maximal mount count reached, "
1898 "running e2fsck is recommended");
ac27a0ec
DK		 1899 else if (le32_to_cpu(es->s_checkinterval) &&
1900 (le32_to_cpu(es->s_lastcheck) +
1901 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1902 ext4_msg(sb, KERN_WARNING,
1903 "warning: checktime reached, "
1904 "running e2fsck is recommended");
0b8e58a1 1905 if (!sbi->s_journal)
0390131b 1906 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1907 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1908 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1909 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1910 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1911 ext4_update_dynamic_rev(sb);
0390131b
FM		 1912 if (sbi->s_journal)
1913 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1914
e2d67052 1915 ext4_commit_super(sb, 1);
281b5995 1916done:
ac27a0ec 1917 if (test_opt(sb, DEBUG))
a9df9a49 1918 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1919 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1920 sb->s_blocksize,
1921 sbi->s_groups_count,
617ba13b
MC		 1922 EXT4_BLOCKS_PER_GROUP(sb),
1923 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1924 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1925
7abc52c2 1926 cleancache_init_fs(sb);
ac27a0ec
DK		 1927 return res;
1928}
1929
772cb7c8
JS		 1930static int ext4_fill_flex_info(struct super_block *sb)
1931{
1932 struct ext4_sb_info *sbi = EXT4_SB(sb);
1933 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1934 ext4_group_t flex_group_count;
1935 ext4_group_t flex_group;
d50f2ab6 1936 unsigned int groups_per_flex = 0;
c5ca7c76 1937 size_t size;
772cb7c8
JS		 1938 int i;
1939
503358ae 1940 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1941 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1942 sbi->s_log_groups_per_flex = 0;
1943 return 1;
1944 }
d50f2ab6 1945 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
772cb7c8 1946
c62a11fd
FB		 1947 /* We allocate both existing and potentially added groups */
1948 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1949 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1950 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 1951 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 1952 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 1953 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 1954 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1955 flex_group_count);
1956 goto failed;
772cb7c8 1957 }
772cb7c8 1958
772cb7c8 1959 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1960 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1961
1962 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1963 atomic_add(ext4_free_inodes_count(sb, gdp),
1964 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1965 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1966 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1967 atomic_add(ext4_used_dirs_count(sb, gdp),
1968 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1969 }
1970
1971 return 1;
1972failed:
1973 return 0;
1974}
1975
feb0ab32
DW		 1976static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1977 struct ext4_group_desc *gdp)
717d50e4 1978{
feb0ab32 1979 int offset;
717d50e4 1980 __u16 crc = 0;
feb0ab32 1981 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1982
feb0ab32
DW		 1983 if ((sbi->s_es->s_feature_ro_compat &
1984 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1985 /* Use new metadata_csum algorithm */
1986 __u16 old_csum;
1987 __u32 csum32;
1988
1989 old_csum = gdp->bg_checksum;
1990 gdp->bg_checksum = 0;
1991 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1992 sizeof(le_group));
1993 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1994 sbi->s_desc_size);
1995 gdp->bg_checksum = old_csum;
1996
1997 crc = csum32 & 0xFFFF;
1998 goto out;
717d50e4
AD		 1999 }
2000
feb0ab32
DW		 2001 /* old crc16 code */
2002 offset = offsetof(struct ext4_group_desc, bg_checksum);
2003
2004 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2005 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2006 crc = crc16(crc, (__u8 *)gdp, offset);
2007 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2008 /* for checksum of struct ext4_group_desc do the rest...*/
2009 if ((sbi->s_es->s_feature_incompat &
2010 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2011 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2012 crc = crc16(crc, (__u8 *)gdp + offset,
2013 le16_to_cpu(sbi->s_es->s_desc_size) -
2014 offset);
2015
2016out:
717d50e4
AD		 2017 return cpu_to_le16(crc);
2018}
2019
feb0ab32 2020int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2021 struct ext4_group_desc *gdp)
2022{
feb0ab32
DW		 2023 if (ext4_has_group_desc_csum(sb) &&
2024 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2025 block_group, gdp)))
717d50e4
AD		 2026 return 0;
2027
2028 return 1;
2029}
2030
feb0ab32
DW		 2031void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2032 struct ext4_group_desc *gdp)
2033{
2034 if (!ext4_has_group_desc_csum(sb))
2035 return;
2036 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2037}
2038
ac27a0ec 2039/* Called at mount-time, super-block is locked */
bfff6873
LC		 2040static int ext4_check_descriptors(struct super_block *sb,
2041 ext4_group_t *first_not_zeroed)
ac27a0ec 2042{
617ba13b
MC		 2043 struct ext4_sb_info *sbi = EXT4_SB(sb);
2044 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2045 ext4_fsblk_t last_block;
bd81d8ee
LV		 2046 ext4_fsblk_t block_bitmap;
2047 ext4_fsblk_t inode_bitmap;
2048 ext4_fsblk_t inode_table;
ce421581 2049 int flexbg_flag = 0;
bfff6873 2050 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2051
ce421581
JS		 2052 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2053 flexbg_flag = 1;
2054
af5bc92d 2055 ext4_debug("Checking group descriptors");
ac27a0ec 2056
197cd65a
AM		 2057 for (i = 0; i < sbi->s_groups_count; i++) {
2058 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2059
ce421581 2060 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2061 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2062 else
2063 last_block = first_block +
617ba13b 2064 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2065
bfff6873
LC		 2066 if ((grp == sbi->s_groups_count) &&
2067 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2068 grp = i;
2069
8fadc143 2070 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2071 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2072 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2073 "Block bitmap for group %u not in group "
b31e1552 2074 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2075 return 0;
2076 }
8fadc143 2077 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2078 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2079 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2080 "Inode bitmap for group %u not in group "
b31e1552 2081 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2082 return 0;
2083 }
8fadc143 2084 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2085 if (inode_table < first_block ||
2b2d6d01 2086 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2087 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2088 "Inode table for group %u not in group "
b31e1552 2089 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2090 return 0;
2091 }
955ce5f5 2092 ext4_lock_group(sb, i);
feb0ab32 2093 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2094 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2095 "Checksum for group %u failed (%u!=%u)",
2096 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2097 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2098 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2099 ext4_unlock_group(sb, i);
8a266467 2100 return 0;
7ee1ec4c 2101 }
717d50e4 2102 }
955ce5f5 2103 ext4_unlock_group(sb, i);
ce421581
JS		 2104 if (!flexbg_flag)
2105 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2106 }
bfff6873
LC		 2107 if (NULL != first_not_zeroed)
2108 *first_not_zeroed = grp;
ac27a0ec 2109
5dee5437
TT		 2110 ext4_free_blocks_count_set(sbi->s_es,
2111 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2112 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2113 return 1;
2114}
2115
617ba13b 2116/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2117 * the superblock) which were deleted from all directories, but held open by
2118 * a process at the time of a crash. We walk the list and try to delete these
2119 * inodes at recovery time (only with a read-write filesystem).
2120 *
2121 * In order to keep the orphan inode chain consistent during traversal (in
2122 * case of crash during recovery), we link each inode into the superblock
2123 * orphan list_head and handle it the same way as an inode deletion during
2124 * normal operation (which journals the operations for us).
2125 *
2126 * We only do an iget() and an iput() on each inode, which is very safe if we
2127 * accidentally point at an in-use or already deleted inode. The worst that
2128 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2129 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2130 * e2fsck was run on this filesystem, and it must have already done the orphan
2131 * inode cleanup for us, so we can safely abort without any further action.
2132 */
2b2d6d01
TT		 2133static void ext4_orphan_cleanup(struct super_block *sb,
2134 struct ext4_super_block *es)
ac27a0ec
DK		 2135{
2136 unsigned int s_flags = sb->s_flags;
2137 int nr_orphans = 0, nr_truncates = 0;
2138#ifdef CONFIG_QUOTA
2139 int i;
2140#endif
2141 if (!es->s_last_orphan) {
2142 jbd_debug(4, "no orphan inodes to clean up\n");
2143 return;
2144 }
2145
a8f48a95 2146 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2147 ext4_msg(sb, KERN_ERR, "write access "
2148 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2149 return;
2150 }
2151
d39195c3
AG		 2152 /* Check if feature set would not allow a r/w mount */
2153 if (!ext4_feature_set_ok(sb, 0)) {
2154 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2155 "unknown ROCOMPAT features");
2156 return;
2157 }
2158
617ba13b 2159 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2160 if (es->s_last_orphan)
2161 jbd_debug(1, "Errors on filesystem, "
2162 "clearing orphan list.\n");
2163 es->s_last_orphan = 0;
2164 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2165 return;
2166 }
2167
2168 if (s_flags & MS_RDONLY) {
b31e1552 2169 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2170 sb->s_flags &= ~MS_RDONLY;
2171 }
2172#ifdef CONFIG_QUOTA
2173 /* Needed for iput() to work correctly and not trash data */
2174 sb->s_flags |= MS_ACTIVE;
2175 /* Turn on quotas so that they are updated correctly */
2176 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2177 if (EXT4_SB(sb)->s_qf_names[i]) {
2178 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2179 if (ret < 0)
b31e1552
ES		 2180 ext4_msg(sb, KERN_ERR,
2181 "Cannot turn on journaled "
2182 "quota: error %d", ret);
ac27a0ec
DK		 2183 }
2184 }
2185#endif
2186
2187 while (es->s_last_orphan) {
2188 struct inode *inode;
2189
97bd42b9
JB		 2190 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2191 if (IS_ERR(inode)) {
ac27a0ec
DK		 2192 es->s_last_orphan = 0;
2193 break;
2194 }
2195
617ba13b 2196 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2197 dquot_initialize(inode);
ac27a0ec 2198 if (inode->i_nlink) {
b31e1552
ES		 2199 ext4_msg(sb, KERN_DEBUG,
2200 "%s: truncating inode %lu to %lld bytes",
46e665e9 2201 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2202 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2203 inode->i_ino, inode->i_size);
617ba13b 2204 ext4_truncate(inode);
ac27a0ec
DK		 2205 nr_truncates++;
2206 } else {
b31e1552
ES		 2207 ext4_msg(sb, KERN_DEBUG,
2208 "%s: deleting unreferenced inode %lu",
46e665e9 2209 __func__, inode->i_ino);
ac27a0ec
DK		 2210 jbd_debug(2, "deleting unreferenced inode %lu\n",
2211 inode->i_ino);
2212 nr_orphans++;
2213 }
2214 iput(inode); /* The delete magic happens here! */
2215 }
2216
2b2d6d01 2217#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2218
2219 if (nr_orphans)
b31e1552
ES		 2220 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2221 PLURAL(nr_orphans));
ac27a0ec 2222 if (nr_truncates)
b31e1552
ES		 2223 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2224 PLURAL(nr_truncates));
ac27a0ec
DK		 2225#ifdef CONFIG_QUOTA
2226 /* Turn quotas off */
2227 for (i = 0; i < MAXQUOTAS; i++) {
2228 if (sb_dqopt(sb)->files[i])
287a8095 2229 dquot_quota_off(sb, i);
ac27a0ec
DK		 2230 }
2231#endif
2232 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2233}
0b8e58a1 2234
cd2291a4
ES		 2235/*
2236 * Maximal extent format file size.
2237 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2238 * extent format containers, within a sector_t, and within i_blocks
2239 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2240 * so that won't be a limiting factor.
2241 *
f17722f9
LC		 2242 * However there is other limiting factor. We do store extents in the form
2243 * of starting block and length, hence the resulting length of the extent
2244 * covering maximum file size must fit into on-disk format containers as
2245 * well. Given that length is always by 1 unit bigger than max unit (because
2246 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2247 *
cd2291a4
ES		 2248 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2249 */
f287a1a5 2250static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2251{
2252 loff_t res;
2253 loff_t upper_limit = MAX_LFS_FILESIZE;
2254
2255 /* small i_blocks in vfs inode? */
f287a1a5 2256 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2257 /*
90c699a9 2258 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2259 * i_block represent total blocks in 512 bytes
2260 * 32 == size of vfs inode i_blocks * 8
2261 */
2262 upper_limit = (1LL << 32) - 1;
2263
2264 /* total blocks in file system block size */
2265 upper_limit >>= (blkbits - 9);
2266 upper_limit <<= blkbits;
2267 }
2268
f17722f9
LC		 2269 /*
2270 * 32-bit extent-start container, ee_block. We lower the maxbytes
2271 * by one fs block, so ee_len can cover the extent of maximum file
2272 * size
2273 */
2274 res = (1LL << 32) - 1;
cd2291a4 2275 res <<= blkbits;
cd2291a4
ES		 2276
2277 /* Sanity check against vm- & vfs- imposed limits */
2278 if (res > upper_limit)
2279 res = upper_limit;
2280
2281 return res;
2282}
ac27a0ec 2283
ac27a0ec 2284/*
cd2291a4 2285 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2286 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2287 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2288 */
f287a1a5 2289static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2290{
617ba13b 2291 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2292 int meta_blocks;
2293 loff_t upper_limit;
0b8e58a1
AD		 2294 /* This is calculated to be the largest file size for a dense, block
2295 * mapped file such that the file's total number of 512-byte sectors,
2296 * including data and all indirect blocks, does not exceed (2^48 - 1).
2297 *
2298 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2299 * number of 512-byte sectors of the file.
0fc1b451
AK		 2300 */
2301
f287a1a5 2302 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2303 /*
90c699a9 2304 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2305 * the inode i_block field represents total file blocks in
2306 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2307 */
2308 upper_limit = (1LL << 32) - 1;
2309
2310 /* total blocks in file system block size */
2311 upper_limit >>= (bits - 9);
2312
2313 } else {
8180a562
AK		 2314 /*
2315 * We use 48 bit ext4_inode i_blocks
2316 * With EXT4_HUGE_FILE_FL set the i_blocks
2317 * represent total number of blocks in
2318 * file system block size
2319 */
0fc1b451
AK		 2320 upper_limit = (1LL << 48) - 1;
2321
0fc1b451
AK		 2322 }
2323
2324 /* indirect blocks */
2325 meta_blocks = 1;
2326 /* double indirect blocks */
2327 meta_blocks += 1 + (1LL << (bits-2));
2328 /* tripple indirect blocks */
2329 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2330
2331 upper_limit -= meta_blocks;
2332 upper_limit <<= bits;
ac27a0ec
DK		 2333
2334 res += 1LL << (bits-2);
2335 res += 1LL << (2*(bits-2));
2336 res += 1LL << (3*(bits-2));
2337 res <<= bits;
2338 if (res > upper_limit)
2339 res = upper_limit;
0fc1b451
AK		 2340
2341 if (res > MAX_LFS_FILESIZE)
2342 res = MAX_LFS_FILESIZE;
2343
ac27a0ec
DK		 2344 return res;
2345}
2346
617ba13b 2347static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2348 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2349{
617ba13b 2350 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2351 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2352 int has_super = 0;
2353
2354 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2355
617ba13b 2356 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2357 nr < first_meta_bg)
70bbb3e0 2358 return logical_sb_block + nr + 1;
ac27a0ec 2359 bg = sbi->s_desc_per_block * nr;
617ba13b 2360 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2361 has_super = 1;
0b8e58a1 2362
617ba13b 2363 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2364}
2365
c9de560d
AT		 2366/**
2367 * ext4_get_stripe_size: Get the stripe size.
2368 * @sbi: In memory super block info
2369 *
2370 * If we have specified it via mount option, then
2371 * use the mount option value. If the value specified at mount time is
2372 * greater than the blocks per group use the super block value.
2373 * If the super block value is greater than blocks per group return 0.
2374 * Allocator needs it be less than blocks per group.
2375 *
2376 */
2377static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2378{
2379 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2380 unsigned long stripe_width =
2381 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2382 int ret;
c9de560d
AT		 2383
2384 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2385 ret = sbi->s_stripe;
2386 else if (stripe_width <= sbi->s_blocks_per_group)
2387 ret = stripe_width;
2388 else if (stride <= sbi->s_blocks_per_group)
2389 ret = stride;
2390 else
2391 ret = 0;
c9de560d 2392
3eb08658
DE		 2393 /*
2394 * If the stripe width is 1, this makes no sense and
2395 * we set it to 0 to turn off stripe handling code.
2396 */
2397 if (ret <= 1)
2398 ret = 0;
c9de560d 2399
3eb08658 2400 return ret;
c9de560d 2401}
ac27a0ec 2402
3197ebdb
TT		 2403/* sysfs supprt */
2404
2405struct ext4_attr {
2406 struct attribute attr;
2407 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2408 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2409 const char *, size_t);
2410 int offset;
2411};
2412
2413static int parse_strtoul(const char *buf,
2414 unsigned long max, unsigned long *value)
2415{
2416 char *endp;
2417
e7d2860b
AGR		 2418 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2419 endp = skip_spaces(endp);
3197ebdb
TT		 2420 if (*endp || *value > max)
2421 return -EINVAL;
2422
2423 return 0;
2424}
2425
2426static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2427 struct ext4_sb_info *sbi,
2428 char *buf)
2429{
2430 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2431 (s64) EXT4_C2B(sbi,
2432 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2433}
2434
2435static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2436 struct ext4_sb_info *sbi, char *buf)
2437{
2438 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2439
f613dfcb
TT		 2440 if (!sb->s_bdev->bd_part)
2441 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2442 return snprintf(buf, PAGE_SIZE, "%lu\n",
2443 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2444 sbi->s_sectors_written_start) >> 1);
2445}
2446
2447static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2448 struct ext4_sb_info *sbi, char *buf)
2449{
2450 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2451
f613dfcb
TT		 2452 if (!sb->s_bdev->bd_part)
2453 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2454 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2455 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2456 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2457 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2458}
2459
2460static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2461 struct ext4_sb_info *sbi,
2462 const char *buf, size_t count)
2463{
2464 unsigned long t;
2465
2466 if (parse_strtoul(buf, 0x40000000, &t))
2467 return -EINVAL;
2468
5dbd571d 2469 if (t && !is_power_of_2(t))
3197ebdb
TT		 2470 return -EINVAL;
2471
2472 sbi->s_inode_readahead_blks = t;
2473 return count;
2474}
2475
2476static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2477 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2478{
2479 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2480
2481 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2482}
2483
2484static ssize_t sbi_ui_store(struct ext4_attr *a,
2485 struct ext4_sb_info *sbi,
2486 const char *buf, size_t count)
2487{
2488 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2489 unsigned long t;
2490
2491 if (parse_strtoul(buf, 0xffffffff, &t))
2492 return -EINVAL;
2493 *ui = t;
2494 return count;
2495}
2496
2c0544b2
TT		 2497static ssize_t trigger_test_error(struct ext4_attr *a,
2498 struct ext4_sb_info *sbi,
2499 const char *buf, size_t count)
2500{
2501 int len = count;
2502
2503 if (!capable(CAP_SYS_ADMIN))
2504 return -EPERM;
2505
2506 if (len && buf[len-1] == '\n')
2507 len--;
2508
2509 if (len)
2510 ext4_error(sbi->s_sb, "%.*s", len, buf);
2511 return count;
2512}
2513
3197ebdb
TT		 2514#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2515static struct ext4_attr ext4_attr_##_name = { \
2516 .attr = {.name = __stringify(_name), .mode = _mode }, \
2517 .show = _show, \
2518 .store = _store, \
2519 .offset = offsetof(struct ext4_sb_info, _elname), \
2520}
2521#define EXT4_ATTR(name, mode, show, store) \
2522static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2523
857ac889 2524#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2525#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2526#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2527#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2528 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2529#define ATTR_LIST(name) &ext4_attr_##name.attr
2530
2531EXT4_RO_ATTR(delayed_allocation_blocks);
2532EXT4_RO_ATTR(session_write_kbytes);
2533EXT4_RO_ATTR(lifetime_write_kbytes);
2534EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2535 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2536EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2537EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2538EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2539EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2540EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2541EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2542EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2543EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
67a5da56 2544EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2545EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2546
2547static struct attribute *ext4_attrs[] = {
2548 ATTR_LIST(delayed_allocation_blocks),
2549 ATTR_LIST(session_write_kbytes),
2550 ATTR_LIST(lifetime_write_kbytes),
2551 ATTR_LIST(inode_readahead_blks),
11013911 2552 ATTR_LIST(inode_goal),
3197ebdb
TT		 2553 ATTR_LIST(mb_stats),
2554 ATTR_LIST(mb_max_to_scan),
2555 ATTR_LIST(mb_min_to_scan),
2556 ATTR_LIST(mb_order2_req),
2557 ATTR_LIST(mb_stream_req),
2558 ATTR_LIST(mb_group_prealloc),
55138e0b 2559 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2560 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2561 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2562 NULL,
2563};
2564
857ac889
LC		 2565/* Features this copy of ext4 supports */
2566EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2567EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2568
2569static struct attribute *ext4_feat_attrs[] = {
2570 ATTR_LIST(lazy_itable_init),
27ee40df 2571 ATTR_LIST(batched_discard),
857ac889
LC		 2572 NULL,
2573};
2574
3197ebdb
TT		 2575static ssize_t ext4_attr_show(struct kobject *kobj,
2576 struct attribute *attr, char *buf)
2577{
2578 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2579 s_kobj);
2580 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2581
2582 return a->show ? a->show(a, sbi, buf) : 0;
2583}
2584
2585static ssize_t ext4_attr_store(struct kobject *kobj,
2586 struct attribute *attr,
2587 const char *buf, size_t len)
2588{
2589 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2590 s_kobj);
2591 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2592
2593 return a->store ? a->store(a, sbi, buf, len) : 0;
2594}
2595
2596static void ext4_sb_release(struct kobject *kobj)
2597{
2598 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2599 s_kobj);
2600 complete(&sbi->s_kobj_unregister);
2601}
2602
52cf25d0 2603static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2604 .show = ext4_attr_show,
2605 .store = ext4_attr_store,
2606};
2607
2608static struct kobj_type ext4_ktype = {
2609 .default_attrs = ext4_attrs,
2610 .sysfs_ops = &ext4_attr_ops,
2611 .release = ext4_sb_release,
2612};
2613
857ac889
LC		 2614static void ext4_feat_release(struct kobject *kobj)
2615{
2616 complete(&ext4_feat->f_kobj_unregister);
2617}
2618
2619static struct kobj_type ext4_feat_ktype = {
2620 .default_attrs = ext4_feat_attrs,
2621 .sysfs_ops = &ext4_attr_ops,
2622 .release = ext4_feat_release,
2623};
2624
a13fb1a4
ES		 2625/*
2626 * Check whether this filesystem can be mounted based on
2627 * the features present and the RDONLY/RDWR mount requested.
2628 * Returns 1 if this filesystem can be mounted as requested,
2629 * 0 if it cannot be.
2630 */
2631static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2632{
2633 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2634 ext4_msg(sb, KERN_ERR,
2635 "Couldn't mount because of "
2636 "unsupported optional features (%x)",
2637 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2638 ~EXT4_FEATURE_INCOMPAT_SUPP));
2639 return 0;
2640 }
2641
2642 if (readonly)
2643 return 1;
2644
2645 /* Check that feature set is OK for a read-write mount */
2646 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2647 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2648 "unsupported optional features (%x)",
2649 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2650 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2651 return 0;
2652 }
2653 /*
2654 * Large file size enabled file system can only be mounted
2655 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2656 */
2657 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2658 if (sizeof(blkcnt_t) < sizeof(u64)) {
2659 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2660 "cannot be mounted RDWR without "
2661 "CONFIG_LBDAF");
2662 return 0;
2663 }
2664 }
bab08ab9
TT		 2665 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2666 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2667 ext4_msg(sb, KERN_ERR,
2668 "Can't support bigalloc feature without "
2669 "extents feature\n");
2670 return 0;
2671 }
7c319d32
AK		 2672
2673#ifndef CONFIG_QUOTA
2674 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2675 !readonly) {
2676 ext4_msg(sb, KERN_ERR,
2677 "Filesystem with quota feature cannot be mounted RDWR "
2678 "without CONFIG_QUOTA");
2679 return 0;
2680 }
2681#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2682 return 1;
2683}
2684
66e61a9e
TT		 2685/*
2686 * This function is called once a day if we have errors logged
2687 * on the file system
2688 */
2689static void print_daily_error_info(unsigned long arg)
2690{
2691 struct super_block *sb = (struct super_block *) arg;
2692 struct ext4_sb_info *sbi;
2693 struct ext4_super_block *es;
2694
2695 sbi = EXT4_SB(sb);
2696 es = sbi->s_es;
2697
2698 if (es->s_error_count)
2699 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2700 le32_to_cpu(es->s_error_count));
2701 if (es->s_first_error_time) {
2702 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2703 sb->s_id, le32_to_cpu(es->s_first_error_time),
2704 (int) sizeof(es->s_first_error_func),
2705 es->s_first_error_func,
2706 le32_to_cpu(es->s_first_error_line));
2707 if (es->s_first_error_ino)
2708 printk(": inode %u",
2709 le32_to_cpu(es->s_first_error_ino));
2710 if (es->s_first_error_block)
2711 printk(": block %llu", (unsigned long long)
2712 le64_to_cpu(es->s_first_error_block));
2713 printk("\n");
2714 }
2715 if (es->s_last_error_time) {
2716 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2717 sb->s_id, le32_to_cpu(es->s_last_error_time),
2718 (int) sizeof(es->s_last_error_func),
2719 es->s_last_error_func,
2720 le32_to_cpu(es->s_last_error_line));
2721 if (es->s_last_error_ino)
2722 printk(": inode %u",
2723 le32_to_cpu(es->s_last_error_ino));
2724 if (es->s_last_error_block)
2725 printk(": block %llu", (unsigned long long)
2726 le64_to_cpu(es->s_last_error_block));
2727 printk("\n");
2728 }
2729 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2730}
2731
bfff6873
LC		 2732/* Find next suitable group and run ext4_init_inode_table */
2733static int ext4_run_li_request(struct ext4_li_request *elr)
2734{
2735 struct ext4_group_desc *gdp = NULL;
2736 ext4_group_t group, ngroups;
2737 struct super_block *sb;
2738 unsigned long timeout = 0;
2739 int ret = 0;
2740
2741 sb = elr->lr_super;
2742 ngroups = EXT4_SB(sb)->s_groups_count;
2743
8e8ad8a5 2744 sb_start_write(sb);
bfff6873
LC		 2745 for (group = elr->lr_next_group; group < ngroups; group++) {
2746 gdp = ext4_get_group_desc(sb, group, NULL);
2747 if (!gdp) {
2748 ret = 1;
2749 break;
2750 }
2751
2752 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2753 break;
2754 }
2755
2756 if (group == ngroups)
2757 ret = 1;
2758
2759 if (!ret) {
2760 timeout = jiffies;
2761 ret = ext4_init_inode_table(sb, group,
2762 elr->lr_timeout ? 0 : 1);
2763 if (elr->lr_timeout == 0) {
51ce6511
LC		 2764 timeout = (jiffies - timeout) *
2765 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2766 elr->lr_timeout = timeout;
2767 }
2768 elr->lr_next_sched = jiffies + elr->lr_timeout;
2769 elr->lr_next_group = group + 1;
2770 }
8e8ad8a5 2771 sb_end_write(sb);
bfff6873
LC		 2772
2773 return ret;
2774}
2775
2776/*
2777 * Remove lr_request from the list_request and free the
4ed5c033 2778 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2779 */
2780static void ext4_remove_li_request(struct ext4_li_request *elr)
2781{
2782 struct ext4_sb_info *sbi;
2783
2784 if (!elr)
2785 return;
2786
2787 sbi = elr->lr_sbi;
2788
2789 list_del(&elr->lr_request);
2790 sbi->s_li_request = NULL;
2791 kfree(elr);
2792}
2793
2794static void ext4_unregister_li_request(struct super_block *sb)
2795{
1bb933fb
LC		 2796 mutex_lock(&ext4_li_mtx);
2797 if (!ext4_li_info) {
2798 mutex_unlock(&ext4_li_mtx);
bfff6873 2799 return;
1bb933fb 2800 }
bfff6873
LC		 2801
2802 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2803 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2804 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2805 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2806}
2807
8f1f7453
ES		 2808static struct task_struct *ext4_lazyinit_task;
2809
bfff6873
LC		 2810/*
2811 * This is the function where ext4lazyinit thread lives. It walks
2812 * through the request list searching for next scheduled filesystem.
2813 * When such a fs is found, run the lazy initialization request
2814 * (ext4_rn_li_request) and keep track of the time spend in this
2815 * function. Based on that time we compute next schedule time of
2816 * the request. When walking through the list is complete, compute
2817 * next waking time and put itself into sleep.
2818 */
2819static int ext4_lazyinit_thread(void *arg)
2820{
2821 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2822 struct list_head *pos, *n;
2823 struct ext4_li_request *elr;
4ed5c033 2824 unsigned long next_wakeup, cur;
bfff6873
LC		 2825
2826 BUG_ON(NULL == eli);
2827
bfff6873
LC		 2828cont_thread:
2829 while (true) {
2830 next_wakeup = MAX_JIFFY_OFFSET;
2831
2832 mutex_lock(&eli->li_list_mtx);
2833 if (list_empty(&eli->li_request_list)) {
2834 mutex_unlock(&eli->li_list_mtx);
2835 goto exit_thread;
2836 }
2837
2838 list_for_each_safe(pos, n, &eli->li_request_list) {
2839 elr = list_entry(pos, struct ext4_li_request,
2840 lr_request);
2841
b2c78cd0
TT		 2842 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2843 if (ext4_run_li_request(elr) != 0) {
2844 /* error, remove the lazy_init job */
2845 ext4_remove_li_request(elr);
2846 continue;
2847 }
bfff6873
LC		 2848 }
2849
2850 if (time_before(elr->lr_next_sched, next_wakeup))
2851 next_wakeup = elr->lr_next_sched;
2852 }
2853 mutex_unlock(&eli->li_list_mtx);
2854
a0acae0e 2855 try_to_freeze();
bfff6873 2856
4ed5c033
LC		 2857 cur = jiffies;
2858 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2859 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2860 cond_resched();
2861 continue;
2862 }
2863
4ed5c033
LC		 2864 schedule_timeout_interruptible(next_wakeup - cur);
2865
8f1f7453
ES		 2866 if (kthread_should_stop()) {
2867 ext4_clear_request_list();
2868 goto exit_thread;
2869 }
bfff6873
LC		 2870 }
2871
2872exit_thread:
2873 /*
2874 * It looks like the request list is empty, but we need
2875 * to check it under the li_list_mtx lock, to prevent any
2876 * additions into it, and of course we should lock ext4_li_mtx
2877 * to atomically free the list and ext4_li_info, because at
2878 * this point another ext4 filesystem could be registering
2879 * new one.
2880 */
2881 mutex_lock(&ext4_li_mtx);
2882 mutex_lock(&eli->li_list_mtx);
2883 if (!list_empty(&eli->li_request_list)) {
2884 mutex_unlock(&eli->li_list_mtx);
2885 mutex_unlock(&ext4_li_mtx);
2886 goto cont_thread;
2887 }
2888 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2889 kfree(ext4_li_info);
2890 ext4_li_info = NULL;
2891 mutex_unlock(&ext4_li_mtx);
2892
2893 return 0;
2894}
2895
2896static void ext4_clear_request_list(void)
2897{
2898 struct list_head *pos, *n;
2899 struct ext4_li_request *elr;
2900
2901 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2902 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2903 elr = list_entry(pos, struct ext4_li_request,
2904 lr_request);
2905 ext4_remove_li_request(elr);
2906 }
2907 mutex_unlock(&ext4_li_info->li_list_mtx);
2908}
2909
2910static int ext4_run_lazyinit_thread(void)
2911{
8f1f7453
ES		 2912 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2913 ext4_li_info, "ext4lazyinit");
2914 if (IS_ERR(ext4_lazyinit_task)) {
2915 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2916 ext4_clear_request_list();
bfff6873
LC		 2917 kfree(ext4_li_info);
2918 ext4_li_info = NULL;
92b97816 2919 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2920 "initialization thread\n",
2921 err);
2922 return err;
2923 }
2924 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2925 return 0;
2926}
2927
2928/*
2929 * Check whether it make sense to run itable init. thread or not.
2930 * If there is at least one uninitialized inode table, return
2931 * corresponding group number, else the loop goes through all
2932 * groups and return total number of groups.
2933 */
2934static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2935{
2936 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2937 struct ext4_group_desc *gdp = NULL;
2938
2939 for (group = 0; group < ngroups; group++) {
2940 gdp = ext4_get_group_desc(sb, group, NULL);
2941 if (!gdp)
2942 continue;
2943
2944 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2945 break;
2946 }
2947
2948 return group;
2949}
2950
2951static int ext4_li_info_new(void)
2952{
2953 struct ext4_lazy_init *eli = NULL;
2954
2955 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2956 if (!eli)
2957 return -ENOMEM;
2958
bfff6873
LC		 2959 INIT_LIST_HEAD(&eli->li_request_list);
2960 mutex_init(&eli->li_list_mtx);
2961
bfff6873
LC		 2962 eli->li_state |= EXT4_LAZYINIT_QUIT;
2963
2964 ext4_li_info = eli;
2965
2966 return 0;
2967}
2968
2969static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2970 ext4_group_t start)
2971{
2972 struct ext4_sb_info *sbi = EXT4_SB(sb);
2973 struct ext4_li_request *elr;
2974 unsigned long rnd;
2975
2976 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2977 if (!elr)
2978 return NULL;
2979
2980 elr->lr_super = sb;
2981 elr->lr_sbi = sbi;
2982 elr->lr_next_group = start;
2983
2984 /*
2985 * Randomize first schedule time of the request to
2986 * spread the inode table initialization requests
2987 * better.
2988 */
2989 get_random_bytes(&rnd, sizeof(rnd));
2990 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2991 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2992
2993 return elr;
2994}
2995
2996static int ext4_register_li_request(struct super_block *sb,
2997 ext4_group_t first_not_zeroed)
2998{
2999 struct ext4_sb_info *sbi = EXT4_SB(sb);
3000 struct ext4_li_request *elr;
3001 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3002 int ret = 0;
bfff6873 3003
51ce6511
LC		 3004 if (sbi->s_li_request != NULL) {
3005 /*
3006 * Reset timeout so it can be computed again, because
3007 * s_li_wait_mult might have changed.
3008 */
3009 sbi->s_li_request->lr_timeout = 0;
beed5ecb 3010 return 0;
51ce6511 3011 }
bfff6873
LC		 3012
3013 if (first_not_zeroed == ngroups ||
3014 (sb->s_flags & MS_RDONLY) ||
55ff3840 3015 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 3016 return 0;
bfff6873
LC		 3017
3018 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 3019 if (!elr)
3020 return -ENOMEM;
bfff6873
LC		 3021
3022 mutex_lock(&ext4_li_mtx);
3023
3024 if (NULL == ext4_li_info) {
3025 ret = ext4_li_info_new();
3026 if (ret)
3027 goto out;
3028 }
3029
3030 mutex_lock(&ext4_li_info->li_list_mtx);
3031 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3032 mutex_unlock(&ext4_li_info->li_list_mtx);
3033
3034 sbi->s_li_request = elr;
46e4690b
TM		 3035 /*
3036 * set elr to NULL here since it has been inserted to
3037 * the request_list and the removal and free of it is
3038 * handled by ext4_clear_request_list from now on.
3039 */
3040 elr = NULL;
bfff6873
LC		 3041
3042 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3043 ret = ext4_run_lazyinit_thread();
3044 if (ret)
3045 goto out;
3046 }
bfff6873 3047out:
beed5ecb
NK		 3048 mutex_unlock(&ext4_li_mtx);
3049 if (ret)
bfff6873 3050 kfree(elr);
bfff6873
LC		 3051 return ret;
3052}
3053
3054/*
3055 * We do not need to lock anything since this is called on
3056 * module unload.
3057 */
3058static void ext4_destroy_lazyinit_thread(void)
3059{
3060 /*
3061 * If thread exited earlier
3062 * there's nothing to be done.
3063 */
8f1f7453 3064 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3065 return;
3066
8f1f7453 3067 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3068}
3069
25ed6e8a
DW		 3070static int set_journal_csum_feature_set(struct super_block *sb)
3071{
3072 int ret = 1;
3073 int compat, incompat;
3074 struct ext4_sb_info *sbi = EXT4_SB(sb);
3075
3076 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3077 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3078 /* journal checksum v2 */
3079 compat = 0;
3080 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3081 } else {
3082 /* journal checksum v1 */
3083 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3084 incompat = 0;
3085 }
3086
3087 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3088 ret = jbd2_journal_set_features(sbi->s_journal,
3089 compat, 0,
3090 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3091 incompat);
3092 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3093 ret = jbd2_journal_set_features(sbi->s_journal,
3094 compat, 0,
3095 incompat);
3096 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3097 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3098 } else {
3099 jbd2_journal_clear_features(sbi->s_journal,
3100 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3101 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3102 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3103 }
3104
3105 return ret;
3106}
3107
952fc18e
TT		 3108/*
3109 * Note: calculating the overhead so we can be compatible with
3110 * historical BSD practice is quite difficult in the face of
3111 * clusters/bigalloc. This is because multiple metadata blocks from
3112 * different block group can end up in the same allocation cluster.
3113 * Calculating the exact overhead in the face of clustered allocation
3114 * requires either O(all block bitmaps) in memory or O(number of block
3115 * groups**2) in time. We will still calculate the superblock for
3116 * older file systems --- and if we come across with a bigalloc file
3117 * system with zero in s_overhead_clusters the estimate will be close to
3118 * correct especially for very large cluster sizes --- but for newer
3119 * file systems, it's better to calculate this figure once at mkfs
3120 * time, and store it in the superblock. If the superblock value is
3121 * present (even for non-bigalloc file systems), we will use it.
3122 */
3123static int count_overhead(struct super_block *sb, ext4_group_t grp,
3124 char *buf)
3125{
3126 struct ext4_sb_info *sbi = EXT4_SB(sb);
3127 struct ext4_group_desc *gdp;
3128 ext4_fsblk_t first_block, last_block, b;
3129 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3130 int s, j, count = 0;
3131
0548bbb8
TT		 3132 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3133 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3134 sbi->s_itb_per_group + 2);
3135
952fc18e
TT		 3136 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3137 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3138 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3139 for (i = 0; i < ngroups; i++) {
3140 gdp = ext4_get_group_desc(sb, i, NULL);
3141 b = ext4_block_bitmap(sb, gdp);
3142 if (b >= first_block && b <= last_block) {
3143 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3144 count++;
3145 }
3146 b = ext4_inode_bitmap(sb, gdp);
3147 if (b >= first_block && b <= last_block) {
3148 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3149 count++;
3150 }
3151 b = ext4_inode_table(sb, gdp);
3152 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3153 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3154 int c = EXT4_B2C(sbi, b - first_block);
3155 ext4_set_bit(c, buf);
3156 count++;
3157 }
3158 if (i != grp)
3159 continue;
3160 s = 0;
3161 if (ext4_bg_has_super(sb, grp)) {
3162 ext4_set_bit(s++, buf);
3163 count++;
3164 }
3165 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3166 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3167 count++;
3168 }
3169 }
3170 if (!count)
3171 return 0;
3172 return EXT4_CLUSTERS_PER_GROUP(sb) -
3173 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3174}
3175
3176/*
3177 * Compute the overhead and stash it in sbi->s_overhead
3178 */
3179int ext4_calculate_overhead(struct super_block *sb)
3180{
3181 struct ext4_sb_info *sbi = EXT4_SB(sb);
3182 struct ext4_super_block *es = sbi->s_es;
3183 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3184 ext4_fsblk_t overhead = 0;
3185 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3186
3187 memset(buf, 0, PAGE_SIZE);
3188 if (!buf)
3189 return -ENOMEM;
3190
3191 /*
3192 * Compute the overhead (FS structures). This is constant
3193 * for a given filesystem unless the number of block groups
3194 * changes so we cache the previous value until it does.
3195 */
3196
3197 /*
3198 * All of the blocks before first_data_block are overhead
3199 */
3200 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3201
3202 /*
3203 * Add the overhead found in each block group
3204 */
3205 for (i = 0; i < ngroups; i++) {
3206 int blks;
3207
3208 blks = count_overhead(sb, i, buf);
3209 overhead += blks;
3210 if (blks)
3211 memset(buf, 0, PAGE_SIZE);
3212 cond_resched();
3213 }
3214 sbi->s_overhead = overhead;
3215 smp_wmb();
3216 free_page((unsigned long) buf);
3217 return 0;
3218}
3219
2b2d6d01 3220static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3221{
d4c402d9 3222 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3223 struct buffer_head *bh;
617ba13b
MC		 3224 struct ext4_super_block *es = NULL;
3225 struct ext4_sb_info *sbi;
3226 ext4_fsblk_t block;
3227 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3228 ext4_fsblk_t logical_sb_block;
ac27a0ec 3229 unsigned long offset = 0;
ac27a0ec
DK		 3230 unsigned long journal_devnum = 0;
3231 unsigned long def_mount_opts;
3232 struct inode *root;
9f6200bb 3233 char *cp;
0390131b 3234 const char *descr;
dcc7dae3 3235 int ret = -ENOMEM;
281b5995 3236 int blocksize, clustersize;
4ec11028
TT		 3237 unsigned int db_count;
3238 unsigned int i;
281b5995 3239 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3240 __u64 blocks_count;
833f4077 3241 int err;
b3881f74 3242 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3243 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3244
3245 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3246 if (!sbi)
dcc7dae3 3247 goto out_free_orig;
705895b6
PE		 3248
3249 sbi->s_blockgroup_lock =
3250 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3251 if (!sbi->s_blockgroup_lock) {
3252 kfree(sbi);
dcc7dae3 3253 goto out_free_orig;
705895b6 3254 }
ac27a0ec 3255 sb->s_fs_info = sbi;
2c0544b2 3256 sbi->s_sb = sb;
ac27a0ec 3257 sbi->s_mount_opt = 0;
08cefc7a
EB		 3258 sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3259 sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
240799cd 3260 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3261 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3262 if (sb->s_bdev->bd_part)
3263 sbi->s_sectors_written_start =
3264 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3265
9f6200bb
TT		 3266 /* Cleanup superblock name */
3267 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3268 *cp = '!';
3269
dcc7dae3 3270 ret = -EINVAL;
617ba13b 3271 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3272 if (!blocksize) {
b31e1552 3273 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3274 goto out_fail;
3275 }
3276
3277 /*
617ba13b 3278 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3279 * block sizes. We need to calculate the offset from buffer start.
3280 */
617ba13b 3281 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3282 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3283 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3284 } else {
70bbb3e0 3285 logical_sb_block = sb_block;
ac27a0ec
DK		 3286 }
3287
70bbb3e0 3288 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3289 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3290 goto out_fail;
3291 }
3292 /*
3293 * Note: s_es must be initialized as soon as possible because
617ba13b 3294 * some ext4 macro-instructions depend on its value
ac27a0ec 3295 */
2716b802 3296 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3297 sbi->s_es = es;
3298 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3299 if (sb->s_magic != EXT4_SUPER_MAGIC)
3300 goto cantfind_ext4;
afc32f7e 3301 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3302
feb0ab32
DW		 3303 /* Warn if metadata_csum and gdt_csum are both set. */
3304 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3305 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3306 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3307 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3308 "redundant flags; please run fsck.");
3309
d25425f8
DW		 3310 /* Check for a known checksum algorithm */
3311 if (!ext4_verify_csum_type(sb, es)) {
3312 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3313 "unknown checksum algorithm.");
3314 silent = 1;
3315 goto cantfind_ext4;
3316 }
3317
0441984a
DW		 3318 /* Load the checksum driver */
3319 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3320 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3321 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3322 if (IS_ERR(sbi->s_chksum_driver)) {
3323 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3324 ret = PTR_ERR(sbi->s_chksum_driver);
3325 sbi->s_chksum_driver = NULL;
3326 goto failed_mount;
3327 }
3328 }
3329
a9c47317
DW		 3330 /* Check superblock checksum */
3331 if (!ext4_superblock_csum_verify(sb, es)) {
3332 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3333 "invalid superblock checksum. Run e2fsck?");
3334 silent = 1;
3335 goto cantfind_ext4;
3336 }
3337
3338 /* Precompute checksum seed for all metadata */
3339 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3340 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3341 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3342 sizeof(es->s_uuid));
3343
ac27a0ec
DK		 3344 /* Set defaults before we parse the mount options */
3345 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3346 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3347 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3348 set_opt(sb, DEBUG);
87f26807 3349 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3350 set_opt(sb, GRPID);
617ba13b 3351 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3352 set_opt(sb, NO_UID32);
ea663336 3353 /* xattr user namespace & acls are now defaulted on */
03010a33 3354#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3355 set_opt(sb, XATTR_USER);
2e7842b8 3356#endif
03010a33 3357#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3358 set_opt(sb, POSIX_ACL);
2e7842b8 3359#endif
6fd7a467 3360 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3361 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3362 set_opt(sb, JOURNAL_DATA);
617ba13b 3363 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3364 set_opt(sb, ORDERED_DATA);
617ba13b 3365 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3366 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3367
3368 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3369 set_opt(sb, ERRORS_PANIC);
bb4f397a 3370 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3371 set_opt(sb, ERRORS_CONT);
bb4f397a 3372 else
fd8c37ec 3373 set_opt(sb, ERRORS_RO);
8b67f04a 3374 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3375 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3376 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3377 set_opt(sb, DISCARD);
ac27a0ec 3378
08cefc7a
EB		 3379 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3380 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3381 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3382 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3383 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3384
8b67f04a 3385 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3386 set_opt(sb, BARRIER);
ac27a0ec 3387
dd919b98
AK		 3388 /*
3389 * enable delayed allocation by default
3390 * Use -o nodelalloc to turn it off
3391 */
8b67f04a
TT		 3392 if (!IS_EXT3_SB(sb) &&
3393 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3394 set_opt(sb, DELALLOC);
dd919b98 3395
51ce6511
LC		 3396 /*
3397 * set default s_li_wait_mult for lazyinit, for the case there is
3398 * no mount option specified.
3399 */
3400 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3401
8b67f04a 3402 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3403 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3404 ext4_msg(sb, KERN_WARNING,
3405 "failed to parse options in superblock: %s",
3406 sbi->s_es->s_mount_opts);
3407 }
5a916be1 3408 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3409 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3410 &journal_ioprio, 0))
ac27a0ec
DK		 3411 goto failed_mount;
3412
56889787
TT		 3413 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3414 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3415 "with data=journal disables delayed "
3416 "allocation and O_DIRECT support!\n");
3417 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3418 ext4_msg(sb, KERN_ERR, "can't mount with "
3419 "both data=journal and delalloc");
3420 goto failed_mount;
3421 }
3422 if (test_opt(sb, DIOREAD_NOLOCK)) {
3423 ext4_msg(sb, KERN_ERR, "can't mount with "
3424 "both data=journal and delalloc");
3425 goto failed_mount;
3426 }
3427 if (test_opt(sb, DELALLOC))
3428 clear_opt(sb, DELALLOC);
3429 }
3430
3431 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3432 if (test_opt(sb, DIOREAD_NOLOCK)) {
3433 if (blocksize < PAGE_SIZE) {
3434 ext4_msg(sb, KERN_ERR, "can't mount with "
3435 "dioread_nolock if block size != PAGE_SIZE");
3436 goto failed_mount;
3437 }
3438 }
3439
ac27a0ec 3440 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3441 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3442
617ba13b
MC		 3443 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3444 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3445 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3446 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3447 ext4_msg(sb, KERN_WARNING,
3448 "feature flags set on rev 0 fs, "
3449 "running e2fsck is recommended");
469108ff 3450
2035e776
TT		 3451 if (IS_EXT2_SB(sb)) {
3452 if (ext2_feature_set_ok(sb))
3453 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3454 "using the ext4 subsystem");
3455 else {
3456 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3457 "to feature incompatibilities");
3458 goto failed_mount;
3459 }
3460 }
3461
3462 if (IS_EXT3_SB(sb)) {
3463 if (ext3_feature_set_ok(sb))
3464 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3465 "using the ext4 subsystem");
3466 else {
3467 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3468 "to feature incompatibilities");
3469 goto failed_mount;
3470 }
3471 }
3472
ac27a0ec
DK		 3473 /*
3474 * Check feature flags regardless of the revision level, since we
3475 * previously didn't change the revision level when setting the flags,
3476 * so there is a chance incompat flags are set on a rev 0 filesystem.
3477 */
a13fb1a4 3478 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3479 goto failed_mount;
a13fb1a4 3480
617ba13b
MC		 3481 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3482 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3483 ext4_msg(sb, KERN_ERR,
3484 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3485 goto failed_mount;
3486 }
3487
ac27a0ec 3488 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3489 /* Validate the filesystem blocksize */
3490 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3491 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3492 blocksize);
ac27a0ec
DK		 3493 goto failed_mount;
3494 }
3495
2b2d6d01 3496 brelse(bh);
70bbb3e0
AM		 3497 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3498 offset = do_div(logical_sb_block, blocksize);
3499 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3500 if (!bh) {
b31e1552
ES		 3501 ext4_msg(sb, KERN_ERR,
3502 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3503 goto failed_mount;
3504 }
2716b802 3505 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3506 sbi->s_es = es;
617ba13b 3507 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3508 ext4_msg(sb, KERN_ERR,
3509 "Magic mismatch, very weird!");
ac27a0ec
DK		 3510 goto failed_mount;
3511 }
3512 }
3513
a13fb1a4
ES		 3514 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3515 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3516 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3517 has_huge_files);
3518 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3519
617ba13b
MC		 3520 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3521 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3522 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3523 } else {
3524 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3525 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3526 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3527 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3528 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3529 ext4_msg(sb, KERN_ERR,
3530 "unsupported inode size: %d",
2b2d6d01 3531 sbi->s_inode_size);
ac27a0ec
DK		 3532 goto failed_mount;
3533 }
ef7f3835
KS		 3534 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3535 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3536 }
0b8e58a1 3537
0d1ee42f
AR		 3538 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3539 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3540 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3541 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3542 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3543 ext4_msg(sb, KERN_ERR,
3544 "unsupported descriptor size %lu",
0d1ee42f
AR		 3545 sbi->s_desc_size);
3546 goto failed_mount;
3547 }
3548 } else
3549 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3550
ac27a0ec 3551 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3552 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3553 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3554 goto cantfind_ext4;
0b8e58a1 3555
617ba13b 3556 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3557 if (sbi->s_inodes_per_block == 0)
617ba13b 3558 goto cantfind_ext4;
ac27a0ec
DK		 3559 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3560 sbi->s_inodes_per_block;
0d1ee42f 3561 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3562 sbi->s_sbh = bh;
3563 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3564 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3565 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3566
2b2d6d01 3567 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3568 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3569 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3570 i = le32_to_cpu(es->s_flags);
3571 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3572 sbi->s_hash_unsigned = 3;
3573 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3574#ifdef __CHAR_UNSIGNED__
3575 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3576 sbi->s_hash_unsigned = 3;
3577#else
3578 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3579#endif
f99b2589 3580 }
ac27a0ec 3581
281b5995
TT		 3582 /* Handle clustersize */
3583 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3584 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3585 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3586 if (has_bigalloc) {
3587 if (clustersize < blocksize) {
3588 ext4_msg(sb, KERN_ERR,
3589 "cluster size (%d) smaller than "
3590 "block size (%d)", clustersize, blocksize);
3591 goto failed_mount;
3592 }
3593 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3594 le32_to_cpu(es->s_log_block_size);
3595 sbi->s_clusters_per_group =
3596 le32_to_cpu(es->s_clusters_per_group);
3597 if (sbi->s_clusters_per_group > blocksize * 8) {
3598 ext4_msg(sb, KERN_ERR,
3599 "#clusters per group too big: %lu",
3600 sbi->s_clusters_per_group);
3601 goto failed_mount;
3602 }
3603 if (sbi->s_blocks_per_group !=
3604 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3605 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3606 "clusters per group (%lu) inconsistent",
3607 sbi->s_blocks_per_group,
3608 sbi->s_clusters_per_group);
3609 goto failed_mount;
3610 }
3611 } else {
3612 if (clustersize != blocksize) {
3613 ext4_warning(sb, "fragment/cluster size (%d) != "
3614 "block size (%d)", clustersize,
3615 blocksize);
3616 clustersize = blocksize;
3617 }
3618 if (sbi->s_blocks_per_group > blocksize * 8) {
3619 ext4_msg(sb, KERN_ERR,
3620 "#blocks per group too big: %lu",
3621 sbi->s_blocks_per_group);
3622 goto failed_mount;
3623 }
3624 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3625 sbi->s_cluster_bits = 0;
ac27a0ec 3626 }
281b5995
TT		 3627 sbi->s_cluster_ratio = clustersize / blocksize;
3628
ac27a0ec 3629 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3630 ext4_msg(sb, KERN_ERR,
3631 "#inodes per group too big: %lu",
2b2d6d01 3632 sbi->s_inodes_per_group);
ac27a0ec
DK		 3633 goto failed_mount;
3634 }
3635
bf43d84b
ES		 3636 /*
3637 * Test whether we have more sectors than will fit in sector_t,
3638 * and whether the max offset is addressable by the page cache.
3639 */
5a9ae68a 3640 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3641 ext4_blocks_count(es));
5a9ae68a 3642 if (err) {
b31e1552 3643 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3644 " too large to mount safely on this system");
ac27a0ec 3645 if (sizeof(sector_t) < 8)
90c699a9 3646 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3647 ret = err;
ac27a0ec
DK		 3648 goto failed_mount;
3649 }
3650
617ba13b
MC		 3651 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3652 goto cantfind_ext4;
e7c95593 3653
0f2ddca6
FTN		 3654 /* check blocks count against device size */
3655 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3656 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3657 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3658 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3659 ext4_blocks_count(es), blocks_count);
3660 goto failed_mount;
3661 }
3662
0b8e58a1
AD		 3663 /*
3664 * It makes no sense for the first data block to be beyond the end
3665 * of the filesystem.
3666 */
3667 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3668 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3669 "block %u is beyond end of filesystem (%llu)",
3670 le32_to_cpu(es->s_first_data_block),
3671 ext4_blocks_count(es));
e7c95593
ES		 3672 goto failed_mount;
3673 }
bd81d8ee
LV		 3674 blocks_count = (ext4_blocks_count(es) -
3675 le32_to_cpu(es->s_first_data_block) +
3676 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3677 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3678 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3679 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3680 "(block count %llu, first data block %u, "
b31e1552 3681 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3682 ext4_blocks_count(es),
3683 le32_to_cpu(es->s_first_data_block),
3684 EXT4_BLOCKS_PER_GROUP(sb));
3685 goto failed_mount;
3686 }
bd81d8ee 3687 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3688 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3689 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3690 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3691 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3692 sbi->s_group_desc = ext4_kvmalloc(db_count *
3693 sizeof(struct buffer_head *),
3694 GFP_KERNEL);
ac27a0ec 3695 if (sbi->s_group_desc == NULL) {
b31e1552 3696 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3697 ret = -ENOMEM;
ac27a0ec
DK		 3698 goto failed_mount;
3699 }
3700
9f6200bb
TT		 3701 if (ext4_proc_root)
3702 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3703
66acdcf4
TT		 3704 if (sbi->s_proc)
3705 proc_create_data("options", S_IRUGO, sbi->s_proc,
3706 &ext4_seq_options_fops, sb);
3707
705895b6 3708 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3709
3710 for (i = 0; i < db_count; i++) {
70bbb3e0 3711 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3712 sbi->s_group_desc[i] = sb_bread(sb, block);
3713 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3714 ext4_msg(sb, KERN_ERR,
3715 "can't read group descriptor %d", i);
ac27a0ec
DK		 3716 db_count = i;
3717 goto failed_mount2;
3718 }
3719 }
bfff6873 3720 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3721 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3722 goto failed_mount2;
3723 }
772cb7c8
JS		 3724 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3725 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3726 ext4_msg(sb, KERN_ERR,
3727 "unable to initialize "
3728 "flex_bg meta info!");
772cb7c8
JS		 3729 goto failed_mount2;
3730 }
3731
ac27a0ec
DK		 3732 sbi->s_gdb_count = db_count;
3733 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3734 spin_lock_init(&sbi->s_next_gen_lock);
3735
04496411
TM		 3736 init_timer(&sbi->s_err_report);
3737 sbi->s_err_report.function = print_daily_error_info;
3738 sbi->s_err_report.data = (unsigned long) sb;
3739
57042651 3740 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3741 ext4_count_free_clusters(sb));
ce7e010a
TT		 3742 if (!err) {
3743 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3744 ext4_count_free_inodes(sb));
3745 }
3746 if (!err) {
3747 err = percpu_counter_init(&sbi->s_dirs_counter,
3748 ext4_count_dirs(sb));
3749 }
3750 if (!err) {
57042651 3751 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3752 }
3753 if (err) {
3754 ext4_msg(sb, KERN_ERR, "insufficient memory");
2cde417d 3755 ret = err;
ce7e010a
TT		 3756 goto failed_mount3;
3757 }
3758
c9de560d 3759 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3760 sbi->s_max_writeback_mb_bump = 128;
67a5da56 3761 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3762
ac27a0ec
DK		 3763 /*
3764 * set up enough so that it can read an inode
3765 */
9ca92389
TT		 3766 if (!test_opt(sb, NOLOAD) &&
3767 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3768 sb->s_op = &ext4_sops;
3769 else
3770 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3771 sb->s_export_op = &ext4_export_ops;
3772 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3773#ifdef CONFIG_QUOTA
617ba13b
MC		 3774 sb->s_qcop = &ext4_qctl_operations;
3775 sb->dq_op = &ext4_quota_operations;
7c319d32
AK		 3776
3777 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3778 /* Use qctl operations for hidden quota files. */
3779 sb->s_qcop = &ext4_qctl_sysfile_operations;
3780 }
ac27a0ec 3781#endif
f2fa2ffc
AK		 3782 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3783
ac27a0ec 3784 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3785 mutex_init(&sbi->s_orphan_lock);
8f82f840 3786 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3787
3788 sb->s_root = NULL;
3789
3790 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3791 EXT4_HAS_INCOMPAT_FEATURE(sb,
3792 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3793
c5e06d10
JL		 3794 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3795 !(sb->s_flags & MS_RDONLY))
3796 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3797 goto failed_mount3;
3798
ac27a0ec
DK		 3799 /*
3800 * The first inode we look at is the journal inode. Don't try
3801 * root first: it may be modified in the journal!
3802 */
3803 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3804 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3805 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3806 goto failed_mount3;
0390131b
FM		 3807 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3808 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3809 ext4_msg(sb, KERN_ERR, "required journal recovery "
3810 "suppressed and not mounted read-only");
744692dc 3811 goto failed_mount_wq;
ac27a0ec 3812 } else {
fd8c37ec 3813 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3814 sbi->s_journal = NULL;
3815 needs_recovery = 0;
3816 goto no_journal;
ac27a0ec
DK		 3817 }
3818
f32aaf2d 3819 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3820 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3821 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3822 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3823 goto failed_mount_wq;
eb40a09c
JS		 3824 }
3825
25ed6e8a
DW		 3826 if (!set_journal_csum_feature_set(sb)) {
3827 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3828 "feature set");
3829 goto failed_mount_wq;
d4da6c9c 3830 }
818d276c 3831
ac27a0ec
DK		 3832 /* We have now updated the journal if required, so we can
3833 * validate the data journaling mode. */
3834 switch (test_opt(sb, DATA_FLAGS)) {
3835 case 0:
3836 /* No mode set, assume a default based on the journal
63f57933
AM		 3837 * capabilities: ORDERED_DATA if the journal can
3838 * cope, else JOURNAL_DATA
3839 */
dab291af
MC		 3840 if (jbd2_journal_check_available_features
3841 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3842 set_opt(sb, ORDERED_DATA);
ac27a0ec 3843 else
fd8c37ec 3844 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3845 break;
3846
617ba13b
MC		 3847 case EXT4_MOUNT_ORDERED_DATA:
3848 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3849 if (!jbd2_journal_check_available_features
3850 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3851 ext4_msg(sb, KERN_ERR, "Journal does not support "
3852 "requested data journaling mode");
744692dc 3853 goto failed_mount_wq;
ac27a0ec
DK		 3854 }
3855 default:
3856 break;
3857 }
b3881f74 3858 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3859
18aadd47
BJ		 3860 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3861
ce7e010a
TT		 3862 /*
3863 * The journal may have updated the bg summary counts, so we
3864 * need to update the global counters.
3865 */
57042651 3866 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3867 ext4_count_free_clusters(sb));
ce7e010a
TT		 3868 percpu_counter_set(&sbi->s_freeinodes_counter,
3869 ext4_count_free_inodes(sb));
3870 percpu_counter_set(&sbi->s_dirs_counter,
3871 ext4_count_dirs(sb));
57042651 3872 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3873
ce7e010a 3874no_journal:
952fc18e
TT		 3875 /*
3876 * Get the # of file system overhead blocks from the
3877 * superblock if present.
3878 */
3879 if (es->s_overhead_clusters)
3880 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3881 else {
3882 ret = ext4_calculate_overhead(sb);
3883 if (ret)
3884 goto failed_mount_wq;
3885 }
3886
fd89d5f2
TH		 3887 /*
3888 * The maximum number of concurrent works can be high and
3889 * concurrency isn't really necessary. Limit it to 1.
3890 */
3891 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3892 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3893 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3894 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3895 goto failed_mount_wq;
3896 }
3897
ac27a0ec 3898 /*
dab291af 3899 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3900 * so we can safely mount the rest of the filesystem now.
3901 */
3902
1d1fe1ee
DH		 3903 root = ext4_iget(sb, EXT4_ROOT_INO);
3904 if (IS_ERR(root)) {
b31e1552 3905 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3906 ret = PTR_ERR(root);
32a9bb57 3907 root = NULL;
ac27a0ec
DK		 3908 goto failed_mount4;
3909 }
3910 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3911 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3912 iput(root);
ac27a0ec
DK		 3913 goto failed_mount4;
3914 }
48fde701 3915 sb->s_root = d_make_root(root);
1d1fe1ee 3916 if (!sb->s_root) {
b31e1552 3917 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3918 ret = -ENOMEM;
3919 goto failed_mount4;
3920 }
ac27a0ec 3921
7e84b621
ES		 3922 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3923 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3924
3925 /* determine the minimum size of new large inodes, if present */
3926 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3927 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3928 EXT4_GOOD_OLD_INODE_SIZE;
3929 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3930 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3931 if (sbi->s_want_extra_isize <
3932 le16_to_cpu(es->s_want_extra_isize))
3933 sbi->s_want_extra_isize =
3934 le16_to_cpu(es->s_want_extra_isize);
3935 if (sbi->s_want_extra_isize <
3936 le16_to_cpu(es->s_min_extra_isize))
3937 sbi->s_want_extra_isize =
3938 le16_to_cpu(es->s_min_extra_isize);
3939 }
3940 }
3941 /* Check if enough inode space is available */
3942 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3943 sbi->s_inode_size) {
3944 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3945 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3946 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3947 "available");
ef7f3835
KS		 3948 }
3949
6fd058f7
TT		 3950 err = ext4_setup_system_zone(sb);
3951 if (err) {
b31e1552 3952 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3953 "zone (%d)", err);
94bf608a 3954 goto failed_mount4a;
6fd058f7
TT		 3955 }
3956
c2774d84 3957 ext4_ext_init(sb);
9d99012f 3958 err = ext4_mb_init(sb);
c2774d84 3959 if (err) {
421f91d2 3960 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3961 err);
dcf2d804 3962 goto failed_mount5;
c2774d84
AK		 3963 }
3964
bfff6873
LC		 3965 err = ext4_register_li_request(sb, first_not_zeroed);
3966 if (err)
dcf2d804 3967 goto failed_mount6;
bfff6873 3968
3197ebdb
TT		 3969 sbi->s_kobj.kset = ext4_kset;
3970 init_completion(&sbi->s_kobj_unregister);
3971 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3972 "%s", sb->s_id);
dcf2d804
TM		 3973 if (err)
3974 goto failed_mount7;
3197ebdb 3975
617ba13b
MC		 3976 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3977 ext4_orphan_cleanup(sb, es);
3978 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3979 if (needs_recovery) {
b31e1552 3980 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3981 ext4_mark_recovery_complete(sb, es);
3982 }
3983 if (EXT4_SB(sb)->s_journal) {
3984 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3985 descr = " journalled data mode";
3986 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3987 descr = " ordered data mode";
3988 else
3989 descr = " writeback data mode";
3990 } else
3991 descr = "out journal";
3992
7c319d32
AK		 3993#ifdef CONFIG_QUOTA
3994 /* Enable quota usage during mount. */
3995 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3996 !(sb->s_flags & MS_RDONLY)) {
3997 ret = ext4_enable_quotas(sb);
3998 if (ret)
3999 goto failed_mount7;
4000 }
4001#endif /* CONFIG_QUOTA */
4002
d4c402d9 4003 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4004 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4005 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4006
66e61a9e
TT		 4007 if (es->s_error_count)
4008 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4009
d4c402d9 4010 kfree(orig_data);
ac27a0ec
DK		 4011 return 0;
4012
617ba13b 4013cantfind_ext4:
ac27a0ec 4014 if (!silent)
b31e1552 4015 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4016 goto failed_mount;
4017
dcf2d804
TM		 4018failed_mount7:
4019 ext4_unregister_li_request(sb);
4020failed_mount6:
dcf2d804 4021 ext4_mb_release(sb);
94bf608a
AV		 4022failed_mount5:
4023 ext4_ext_release(sb);
dcf2d804 4024 ext4_release_system_zone(sb);
94bf608a
AV		 4025failed_mount4a:
4026 dput(sb->s_root);
32a9bb57 4027 sb->s_root = NULL;
94bf608a 4028failed_mount4:
b31e1552 4029 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 4030 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4031failed_mount_wq:
0390131b
FM		 4032 if (sbi->s_journal) {
4033 jbd2_journal_destroy(sbi->s_journal);
4034 sbi->s_journal = NULL;
4035 }
ac27a0ec 4036failed_mount3:
04496411 4037 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4038 if (sbi->s_flex_groups)
4039 ext4_kvfree(sbi->s_flex_groups);
57042651 4040 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4041 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4042 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4043 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 4044 if (sbi->s_mmp_tsk)
4045 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4046failed_mount2:
4047 for (i = 0; i < db_count; i++)
4048 brelse(sbi->s_group_desc[i]);
f18a5f21 4049 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4050failed_mount:
0441984a
DW		 4051 if (sbi->s_chksum_driver)
4052 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4053 if (sbi->s_proc) {
66acdcf4 4054 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4055 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4056 }
ac27a0ec
DK		 4057#ifdef CONFIG_QUOTA
4058 for (i = 0; i < MAXQUOTAS; i++)
4059 kfree(sbi->s_qf_names[i]);
4060#endif
617ba13b 4061 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4062 brelse(bh);
4063out_fail:
4064 sb->s_fs_info = NULL;
f6830165 4065 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4066 kfree(sbi);
dcc7dae3 4067out_free_orig:
d4c402d9 4068 kfree(orig_data);
1d1fe1ee 4069 return ret;
ac27a0ec
DK		 4070}
4071
4072/*
4073 * Setup any per-fs journal parameters now. We'll do this both on
4074 * initial mount, once the journal has been initialised but before we've
4075 * done any recovery; and again on any subsequent remount.
4076 */
617ba13b 4077static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4078{
617ba13b 4079 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4080
30773840
TT		 4081 journal->j_commit_interval = sbi->s_commit_interval;
4082 journal->j_min_batch_time = sbi->s_min_batch_time;
4083 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4084
a931da6a 4085 write_lock(&journal->j_state_lock);
ac27a0ec 4086 if (test_opt(sb, BARRIER))
dab291af 4087 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4088 else
dab291af 4089 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4090 if (test_opt(sb, DATA_ERR_ABORT))
4091 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4092 else
4093 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4094 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4095}
4096
617ba13b 4097static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4098 unsigned int journal_inum)
4099{
4100 struct inode *journal_inode;
4101 journal_t *journal;
4102
0390131b
FM		 4103 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4104
ac27a0ec
DK		 4105 /* First, test for the existence of a valid inode on disk. Bad
4106 * things happen if we iget() an unused inode, as the subsequent
4107 * iput() will try to delete it. */
4108
1d1fe1ee
DH		 4109 journal_inode = ext4_iget(sb, journal_inum);
4110 if (IS_ERR(journal_inode)) {
b31e1552 4111 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4112 return NULL;
4113 }
4114 if (!journal_inode->i_nlink) {
4115 make_bad_inode(journal_inode);
4116 iput(journal_inode);
b31e1552 4117 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4118 return NULL;
4119 }
4120
e5f8eab8 4121 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4122 journal_inode, journal_inode->i_size);
1d1fe1ee 4123 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4124 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4125 iput(journal_inode);
4126 return NULL;
4127 }
4128
dab291af 4129 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4130 if (!journal) {
b31e1552 4131 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4132 iput(journal_inode);
4133 return NULL;
4134 }
4135 journal->j_private = sb;
617ba13b 4136 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4137 return journal;
4138}
4139
617ba13b 4140static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4141 dev_t j_dev)
4142{
2b2d6d01 4143 struct buffer_head *bh;
ac27a0ec 4144 journal_t *journal;
617ba13b
MC		 4145 ext4_fsblk_t start;
4146 ext4_fsblk_t len;
ac27a0ec 4147 int hblock, blocksize;
617ba13b 4148 ext4_fsblk_t sb_block;
ac27a0ec 4149 unsigned long offset;
2b2d6d01 4150 struct ext4_super_block *es;
ac27a0ec
DK		 4151 struct block_device *bdev;
4152
0390131b
FM		 4153 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4154
b31e1552 4155 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4156 if (bdev == NULL)
4157 return NULL;
4158
ac27a0ec 4159 blocksize = sb->s_blocksize;
e1defc4f 4160 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4161 if (blocksize < hblock) {
b31e1552
ES		 4162 ext4_msg(sb, KERN_ERR,
4163 "blocksize too small for journal device");
ac27a0ec
DK		 4164 goto out_bdev;
4165 }
4166
617ba13b
MC		 4167 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4168 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4169 set_blocksize(bdev, blocksize);
4170 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4171 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4172 "external journal");
ac27a0ec
DK		 4173 goto out_bdev;
4174 }
4175
2716b802 4176 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4177 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4178 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4179 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4180 ext4_msg(sb, KERN_ERR, "external journal has "
4181 "bad superblock");
ac27a0ec
DK		 4182 brelse(bh);
4183 goto out_bdev;
4184 }
4185
617ba13b 4186 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4187 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4188 brelse(bh);
4189 goto out_bdev;
4190 }
4191
bd81d8ee 4192 len = ext4_blocks_count(es);
ac27a0ec
DK		 4193 start = sb_block + 1;
4194 brelse(bh); /* we're done with the superblock */
4195
dab291af 4196 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4197 start, len, blocksize);
4198 if (!journal) {
b31e1552 4199 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4200 goto out_bdev;
4201 }
4202 journal->j_private = sb;
4203 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4204 wait_on_buffer(journal->j_sb_buffer);
4205 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4206 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4207 goto out_journal;
4208 }
4209 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4210 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4211 "user (unsupported) - %d",
ac27a0ec
DK		 4212 be32_to_cpu(journal->j_superblock->s_nr_users));
4213 goto out_journal;
4214 }
617ba13b
MC		 4215 EXT4_SB(sb)->journal_bdev = bdev;
4216 ext4_init_journal_params(sb, journal);
ac27a0ec 4217 return journal;
0b8e58a1 4218
ac27a0ec 4219out_journal:
dab291af 4220 jbd2_journal_destroy(journal);
ac27a0ec 4221out_bdev:
617ba13b 4222 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4223 return NULL;
4224}
4225
617ba13b
MC		 4226static int ext4_load_journal(struct super_block *sb,
4227 struct ext4_super_block *es,
ac27a0ec
DK		 4228 unsigned long journal_devnum)
4229{
4230 journal_t *journal;
4231 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4232 dev_t journal_dev;
4233 int err = 0;
4234 int really_read_only;
4235
0390131b
FM		 4236 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4237
ac27a0ec
DK		 4238 if (journal_devnum &&
4239 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4240 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4241 "numbers have changed");
ac27a0ec
DK		 4242 journal_dev = new_decode_dev(journal_devnum);
4243 } else
4244 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4245
4246 really_read_only = bdev_read_only(sb->s_bdev);
4247
4248 /*
4249 * Are we loading a blank journal or performing recovery after a
4250 * crash? For recovery, we need to check in advance whether we
4251 * can get read-write access to the device.
4252 */
617ba13b 4253 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4254 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4255 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4256 "required on readonly filesystem");
ac27a0ec 4257 if (really_read_only) {
b31e1552
ES		 4258 ext4_msg(sb, KERN_ERR, "write access "
4259 "unavailable, cannot proceed");
ac27a0ec
DK		 4260 return -EROFS;
4261 }
b31e1552
ES		 4262 ext4_msg(sb, KERN_INFO, "write access will "
4263 "be enabled during recovery");
ac27a0ec
DK		 4264 }
4265 }
4266
4267 if (journal_inum && journal_dev) {
b31e1552
ES		 4268 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4269 "and inode journals!");
ac27a0ec
DK		 4270 return -EINVAL;
4271 }
4272
4273 if (journal_inum) {
617ba13b 4274 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4275 return -EINVAL;
4276 } else {
617ba13b 4277 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4278 return -EINVAL;
4279 }
4280
90576c0b 4281 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4282 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4283
617ba13b 4284 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4285 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4286 if (!err) {
4287 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4288 if (save)
4289 memcpy(save, ((char *) es) +
4290 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4291 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4292 if (save)
4293 memcpy(((char *) es) + EXT4_S_ERR_START,
4294 save, EXT4_S_ERR_LEN);
4295 kfree(save);
4296 }
ac27a0ec
DK		 4297
4298 if (err) {
b31e1552 4299 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4300 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4301 return err;
4302 }
4303
617ba13b
MC		 4304 EXT4_SB(sb)->s_journal = journal;
4305 ext4_clear_journal_err(sb, es);
ac27a0ec 4306
c41303ce 4307 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4308 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4309 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4310
4311 /* Make sure we flush the recovery flag to disk. */
e2d67052 4312 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4313 }
4314
4315 return 0;
4316}
4317
e2d67052 4318static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4319{
e2d67052 4320 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4321 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4322 int error = 0;
ac27a0ec 4323
7c2e7087 4324 if (!sbh || block_device_ejected(sb))
c4be0c1d 4325 return error;
914258bf
TT		 4326 if (buffer_write_io_error(sbh)) {
4327 /*
4328 * Oh, dear. A previous attempt to write the
4329 * superblock failed. This could happen because the
4330 * USB device was yanked out. Or it could happen to
4331 * be a transient write error and maybe the block will
4332 * be remapped. Nothing we can do but to retry the
4333 * write and hope for the best.
4334 */
b31e1552
ES		 4335 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4336 "superblock detected");
914258bf
TT		 4337 clear_buffer_write_io_error(sbh);
4338 set_buffer_uptodate(sbh);
4339 }
71290b36
TT		 4340 /*
4341 * If the file system is mounted read-only, don't update the
4342 * superblock write time. This avoids updating the superblock
4343 * write time when we are mounting the root file system
4344 * read/only but we need to replay the journal; at that point,
4345 * for people who are east of GMT and who make their clock
4346 * tick in localtime for Windows bug-for-bug compatibility,
4347 * the clock is set in the future, and this will cause e2fsck
4348 * to complain and force a full file system check.
4349 */
4350 if (!(sb->s_flags & MS_RDONLY))
4351 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4352 if (sb->s_bdev->bd_part)
4353 es->s_kbytes_written =
4354 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4355 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4356 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4357 else
4358 es->s_kbytes_written =
4359 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4360 ext4_free_blocks_count_set(es,
4361 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4362 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4363 es->s_free_inodes_count =
4364 cpu_to_le32(percpu_counter_sum_positive(
4365 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4366 BUFFER_TRACE(sbh, "marking dirty");
a9c47317 4367 ext4_superblock_csum_set(sb, es);
ac27a0ec 4368 mark_buffer_dirty(sbh);
914258bf 4369 if (sync) {
c4be0c1d
TS		 4370 error = sync_dirty_buffer(sbh);
4371 if (error)
4372 return error;
4373
4374 error = buffer_write_io_error(sbh);
4375 if (error) {
b31e1552
ES		 4376 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4377 "superblock");
914258bf
TT		 4378 clear_buffer_write_io_error(sbh);
4379 set_buffer_uptodate(sbh);
4380 }
4381 }
c4be0c1d 4382 return error;
ac27a0ec
DK		 4383}
4384
ac27a0ec
DK		 4385/*
4386 * Have we just finished recovery? If so, and if we are mounting (or
4387 * remounting) the filesystem readonly, then we will end up with a
4388 * consistent fs on disk. Record that fact.
4389 */
2b2d6d01
TT		 4390static void ext4_mark_recovery_complete(struct super_block *sb,
4391 struct ext4_super_block *es)
ac27a0ec 4392{
617ba13b 4393 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4394
0390131b
FM		 4395 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4396 BUG_ON(journal != NULL);
4397 return;
4398 }
dab291af 4399 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4400 if (jbd2_journal_flush(journal) < 0)
4401 goto out;
4402
617ba13b 4403 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4404 sb->s_flags & MS_RDONLY) {
617ba13b 4405 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4406 ext4_commit_super(sb, 1);
ac27a0ec 4407 }
7ffe1ea8
HK		 4408
4409out:
dab291af 4410 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4411}
4412
4413/*
4414 * If we are mounting (or read-write remounting) a filesystem whose journal
4415 * has recorded an error from a previous lifetime, move that error to the
4416 * main filesystem now.
4417 */
2b2d6d01
TT		 4418static void ext4_clear_journal_err(struct super_block *sb,
4419 struct ext4_super_block *es)
ac27a0ec
DK		 4420{
4421 journal_t *journal;
4422 int j_errno;
4423 const char *errstr;
4424
0390131b
FM		 4425 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4426
617ba13b 4427 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4428
4429 /*
4430 * Now check for any error status which may have been recorded in the
617ba13b 4431 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4432 */
4433
dab291af 4434 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4435 if (j_errno) {
4436 char nbuf[16];
4437
617ba13b 4438 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4439 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4440 "from previous mount: %s", errstr);
12062ddd 4441 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4442
617ba13b
MC		 4443 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4444 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4445 ext4_commit_super(sb, 1);
ac27a0ec 4446
dab291af 4447 jbd2_journal_clear_err(journal);
d796c52e 4448 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4449 }
4450}
4451
4452/*
4453 * Force the running and committing transactions to commit,
4454 * and wait on the commit.
4455 */
617ba13b 4456int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4457{
4458 journal_t *journal;
0390131b 4459 int ret = 0;
ac27a0ec
DK		 4460
4461 if (sb->s_flags & MS_RDONLY)
4462 return 0;
4463
617ba13b 4464 journal = EXT4_SB(sb)->s_journal;
8e8ad8a5 4465 if (journal)
0390131b 4466 ret = ext4_journal_force_commit(journal);
0390131b 4467
ac27a0ec
DK		 4468 return ret;
4469}
4470
617ba13b 4471static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4472{
14ce0cb4 4473 int ret = 0;
9eddacf9 4474 tid_t target;
8d5d02e6 4475 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4476
9bffad1e 4477 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4478 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4479 /*
4480 * Writeback quota in non-journalled quota case - journalled quota has
4481 * no dirty dquots
4482 */
4483 dquot_writeback_dquots(sb, -1);
8d5d02e6 4484 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4485 if (wait)
8d5d02e6 4486 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4487 }
14ce0cb4 4488 return ret;
ac27a0ec
DK		 4489}
4490
4491/*
4492 * LVM calls this function before a (read-only) snapshot is created. This
4493 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4494 *
4495 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4496 * state independently. It relies on upper layer to stop all data & metadata
4497 * modifications.
ac27a0ec 4498 */
c4be0c1d 4499static int ext4_freeze(struct super_block *sb)
ac27a0ec 4500{
c4be0c1d
TS		 4501 int error = 0;
4502 journal_t *journal;
ac27a0ec 4503
9ca92389
TT		 4504 if (sb->s_flags & MS_RDONLY)
4505 return 0;
ac27a0ec 4506
9ca92389 4507 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4508
9ca92389
TT		 4509 /* Now we set up the journal barrier. */
4510 jbd2_journal_lock_updates(journal);
ac27a0ec 4511
9ca92389
TT		 4512 /*
4513 * Don't clear the needs_recovery flag if we failed to flush
4514 * the journal.
4515 */
4516 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4517 if (error < 0)
4518 goto out;
9ca92389
TT		 4519
4520 /* Journal blocked and flushed, clear needs_recovery flag. */
4521 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4522 error = ext4_commit_super(sb, 1);
6b0310fb 4523out:
8e8ad8a5 4524 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4525 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4526 return error;
ac27a0ec
DK		 4527}
4528
4529/*
4530 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4531 * flag here, even though the filesystem is not technically dirty yet.
4532 */
c4be0c1d 4533static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4534{
9ca92389
TT		 4535 if (sb->s_flags & MS_RDONLY)
4536 return 0;
4537
4538 lock_super(sb);
4539 /* Reset the needs_recovery flag before the fs is unlocked. */
4540 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4541 ext4_commit_super(sb, 1);
4542 unlock_super(sb);
c4be0c1d 4543 return 0;
ac27a0ec
DK		 4544}
4545
673c6100
TT		 4546/*
4547 * Structure to save mount options for ext4_remount's benefit
4548 */
4549struct ext4_mount_options {
4550 unsigned long s_mount_opt;
a2595b8a 4551 unsigned long s_mount_opt2;
08cefc7a
EB		 4552 kuid_t s_resuid;
4553 kgid_t s_resgid;
673c6100
TT		 4554 unsigned long s_commit_interval;
4555 u32 s_min_batch_time, s_max_batch_time;
4556#ifdef CONFIG_QUOTA
4557 int s_jquota_fmt;
4558 char *s_qf_names[MAXQUOTAS];
4559#endif
4560};
4561
2b2d6d01 4562static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4563{
2b2d6d01 4564 struct ext4_super_block *es;
617ba13b 4565 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4566 unsigned long old_sb_flags;
617ba13b 4567 struct ext4_mount_options old_opts;
c79d967d 4568 int enable_quota = 0;
8a266467 4569 ext4_group_t g;
b3881f74 4570 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4571 int err = 0;
ac27a0ec
DK		 4572#ifdef CONFIG_QUOTA
4573 int i;
4574#endif
d4c402d9 4575 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4576
4577 /* Store the original options */
bbd6851a 4578 lock_super(sb);
ac27a0ec
DK		 4579 old_sb_flags = sb->s_flags;
4580 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4581 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4582 old_opts.s_resuid = sbi->s_resuid;
4583 old_opts.s_resgid = sbi->s_resgid;
4584 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4585 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4586 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4587#ifdef CONFIG_QUOTA
4588 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4589 for (i = 0; i < MAXQUOTAS; i++)
4590 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4591#endif
b3881f74
TT		 4592 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4593 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4594
4595 /*
4596 * Allow the "check" option to be passed as a remount option.
4597 */
661aa520 4598 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4599 err = -EINVAL;
4600 goto restore_opts;
4601 }
4602
4ab2f15b 4603 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4604 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4605
4606 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4607 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4608
4609 es = sbi->s_es;
4610
b3881f74 4611 if (sbi->s_journal) {
0390131b 4612 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4613 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4614 }
ac27a0ec 4615
661aa520 4616 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4617 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4618 err = -EROFS;
4619 goto restore_opts;
4620 }
4621
4622 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4623 err = dquot_suspend(sb, -1);
4624 if (err < 0)
c79d967d 4625 goto restore_opts;
c79d967d 4626
ac27a0ec
DK		 4627 /*
4628 * First of all, the unconditional stuff we have to do
4629 * to disable replay of the journal when we next remount
4630 */
4631 sb->s_flags |= MS_RDONLY;
4632
4633 /*
4634 * OK, test if we are remounting a valid rw partition
4635 * readonly, and if so set the rdonly flag and then
4636 * mark the partition as valid again.
4637 */
617ba13b
MC		 4638 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4639 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4640 es->s_state = cpu_to_le16(sbi->s_mount_state);
4641
a63c9eb2 4642 if (sbi->s_journal)
0390131b 4643 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4644 } else {
a13fb1a4
ES		 4645 /* Make sure we can mount this feature set readwrite */
4646 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4647 err = -EROFS;
4648 goto restore_opts;
4649 }
8a266467
TT		 4650 /*
4651 * Make sure the group descriptor checksums
0b8e58a1 4652 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4653 */
4654 for (g = 0; g < sbi->s_groups_count; g++) {
4655 struct ext4_group_desc *gdp =
4656 ext4_get_group_desc(sb, g, NULL);
4657
feb0ab32 4658 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4659 ext4_msg(sb, KERN_ERR,
4660 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4661 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4662 le16_to_cpu(gdp->bg_checksum));
4663 err = -EINVAL;
4664 goto restore_opts;
4665 }
4666 }
4667
ead6596b
ES		 4668 /*
4669 * If we have an unprocessed orphan list hanging
4670 * around from a previously readonly bdev mount,
4671 * require a full umount/remount for now.
4672 */
4673 if (es->s_last_orphan) {
b31e1552 4674 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4675 "remount RDWR because of unprocessed "
4676 "orphan inode list. Please "
b31e1552 4677 "umount/remount instead");
ead6596b
ES		 4678 err = -EINVAL;
4679 goto restore_opts;
4680 }
4681
ac27a0ec
DK		 4682 /*
4683 * Mounting a RDONLY partition read-write, so reread
4684 * and store the current valid flag. (It may have
4685 * been changed by e2fsck since we originally mounted
4686 * the partition.)
4687 */
0390131b
FM		 4688 if (sbi->s_journal)
4689 ext4_clear_journal_err(sb, es);
ac27a0ec 4690 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4691 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4692 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4693 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4694 EXT4_FEATURE_INCOMPAT_MMP))
4695 if (ext4_multi_mount_protect(sb,
4696 le64_to_cpu(es->s_mmp_block))) {
4697 err = -EROFS;
4698 goto restore_opts;
4699 }
c79d967d 4700 enable_quota = 1;
ac27a0ec
DK		 4701 }
4702 }
bfff6873
LC		 4703
4704 /*
4705 * Reinitialize lazy itable initialization thread based on
4706 * current settings
4707 */
4708 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4709 ext4_unregister_li_request(sb);
4710 else {
4711 ext4_group_t first_not_zeroed;
4712 first_not_zeroed = ext4_has_uninit_itable(sb);
4713 ext4_register_li_request(sb, first_not_zeroed);
4714 }
4715
6fd058f7 4716 ext4_setup_system_zone(sb);
0390131b 4717 if (sbi->s_journal == NULL)
e2d67052 4718 ext4_commit_super(sb, 1);
0390131b 4719
7c319d32 4720 unlock_super(sb);
ac27a0ec
DK		 4721#ifdef CONFIG_QUOTA
4722 /* Release old quota file names */
4723 for (i = 0; i < MAXQUOTAS; i++)
4724 if (old_opts.s_qf_names[i] &&
4725 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4726 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4727 if (enable_quota) {
4728 if (sb_any_quota_suspended(sb))
4729 dquot_resume(sb, -1);
4730 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4731 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4732 err = ext4_enable_quotas(sb);
4733 if (err) {
4734 lock_super(sb);
4735 goto restore_opts;
4736 }
4737 }
4738 }
ac27a0ec 4739#endif
d4c402d9
CW		 4740
4741 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4742 kfree(orig_data);
ac27a0ec 4743 return 0;
0b8e58a1 4744
ac27a0ec
DK		 4745restore_opts:
4746 sb->s_flags = old_sb_flags;
4747 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4748 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4749 sbi->s_resuid = old_opts.s_resuid;
4750 sbi->s_resgid = old_opts.s_resgid;
4751 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4752 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4753 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4754#ifdef CONFIG_QUOTA
4755 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4756 for (i = 0; i < MAXQUOTAS; i++) {
4757 if (sbi->s_qf_names[i] &&
4758 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4759 kfree(sbi->s_qf_names[i]);
4760 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4761 }
4762#endif
bbd6851a 4763 unlock_super(sb);
d4c402d9 4764 kfree(orig_data);
ac27a0ec
DK		 4765 return err;
4766}
4767
2b2d6d01 4768static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4769{
4770 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4771 struct ext4_sb_info *sbi = EXT4_SB(sb);
4772 struct ext4_super_block *es = sbi->s_es;
952fc18e 4773 ext4_fsblk_t overhead = 0;
960cc398 4774 u64 fsid;
d02a9391 4775 s64 bfree;
ac27a0ec 4776
952fc18e
TT		 4777 if (!test_opt(sb, MINIX_DF))
4778 overhead = sbi->s_overhead;
ac27a0ec 4779
617ba13b 4780 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4781 buf->f_bsize = sb->s_blocksize;
952fc18e 4782 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
57042651
TT		 4783 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4784 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4785 /* prevent underflow in case that few free space is available */
57042651 4786 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4787 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4788 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4789 buf->f_bavail = 0;
4790 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4791 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4792 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4793 fsid = le64_to_cpup((void *)es->s_uuid) ^
4794 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4795 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4796 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4797
ac27a0ec
DK		 4798 return 0;
4799}
4800
0b8e58a1
AD		 4801/* Helper function for writing quotas on sync - we need to start transaction
4802 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4803 * Process 1 Process 2
617ba13b 4804 * ext4_create() quota_sync()
a269eb18 4805 * jbd2_journal_start() write_dquot()
871a2931 4806 * dquot_initialize() down(dqio_mutex)
dab291af 4807 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4808 *
4809 */
4810
4811#ifdef CONFIG_QUOTA
4812
4813static inline struct inode *dquot_to_inode(struct dquot *dquot)
4814{
4815 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4816}
4817
617ba13b 4818static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4819{
4820 int ret, err;
4821 handle_t *handle;
4822 struct inode *inode;
4823
4824 inode = dquot_to_inode(dquot);
617ba13b 4825 handle = ext4_journal_start(inode,
0b8e58a1 4826 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4827 if (IS_ERR(handle))
4828 return PTR_ERR(handle);
4829 ret = dquot_commit(dquot);
617ba13b 4830 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4831 if (!ret)
4832 ret = err;
4833 return ret;
4834}
4835
617ba13b 4836static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4837{
4838 int ret, err;
4839 handle_t *handle;
4840
617ba13b 4841 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4842 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4843 if (IS_ERR(handle))
4844 return PTR_ERR(handle);
4845 ret = dquot_acquire(dquot);
617ba13b 4846 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4847 if (!ret)
4848 ret = err;
4849 return ret;
4850}
4851
617ba13b 4852static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4853{
4854 int ret, err;
4855 handle_t *handle;
4856
617ba13b 4857 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4858 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4859 if (IS_ERR(handle)) {
4860 /* Release dquot anyway to avoid endless cycle in dqput() */
4861 dquot_release(dquot);
ac27a0ec 4862 return PTR_ERR(handle);
9c3013e9 4863 }
ac27a0ec 4864 ret = dquot_release(dquot);
617ba13b 4865 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4866 if (!ret)
4867 ret = err;
4868 return ret;
4869}
4870
617ba13b 4871static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4872{
2c8be6b2 4873 /* Are we journaling quotas? */
617ba13b
MC		 4874 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4875 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4876 dquot_mark_dquot_dirty(dquot);
617ba13b 4877 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4878 } else {
4879 return dquot_mark_dquot_dirty(dquot);
4880 }
4881}
4882
617ba13b 4883static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4884{
4885 int ret, err;
4886 handle_t *handle;
4887
4888 /* Data block + inode block */
617ba13b 4889 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4890 if (IS_ERR(handle))
4891 return PTR_ERR(handle);
4892 ret = dquot_commit_info(sb, type);
617ba13b 4893 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4894 if (!ret)
4895 ret = err;
4896 return ret;
4897}
4898
4899/*
4900 * Turn on quotas during mount time - we need to find
4901 * the quota file and such...
4902 */
617ba13b 4903static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4904{
287a8095
CH		 4905 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4906 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4907}
4908
4909/*
4910 * Standard function to be called on quota_on
4911 */
617ba13b 4912static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4913 struct path *path)
ac27a0ec
DK		 4914{
4915 int err;
ac27a0ec
DK		 4916
4917 if (!test_opt(sb, QUOTA))
4918 return -EINVAL;
0623543b 4919
ac27a0ec 4920 /* Quotafile not on the same filesystem? */
d8c9584e 4921 if (path->dentry->d_sb != sb)
ac27a0ec 4922 return -EXDEV;
0623543b
JK		 4923 /* Journaling quota? */
4924 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4925 /* Quotafile not in fs root? */
f00c9e44 4926 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4927 ext4_msg(sb, KERN_WARNING,
4928 "Quota file not on filesystem root. "
4929 "Journaled quota will not work");
2b2d6d01 4930 }
0623543b
JK		 4931
4932 /*
4933 * When we journal data on quota file, we have to flush journal to see
4934 * all updates to the file when we bypass pagecache...
4935 */
0390131b 4936 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4937 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4938 /*
4939 * We don't need to lock updates but journal_flush() could
4940 * otherwise be livelocked...
4941 */
4942 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4943 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4944 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4945 if (err)
7ffe1ea8 4946 return err;
0623543b
JK		 4947 }
4948
f00c9e44 4949 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4950}
4951
7c319d32
AK		 4952static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4953 unsigned int flags)
4954{
4955 int err;
4956 struct inode *qf_inode;
4957 unsigned long qf_inums[MAXQUOTAS] = {
4958 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4959 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4960 };
4961
4962 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4963
4964 if (!qf_inums[type])
4965 return -EPERM;
4966
4967 qf_inode = ext4_iget(sb, qf_inums[type]);
4968 if (IS_ERR(qf_inode)) {
4969 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4970 return PTR_ERR(qf_inode);
4971 }
4972
4973 err = dquot_enable(qf_inode, type, format_id, flags);
4974 iput(qf_inode);
4975
4976 return err;
4977}
4978
4979/* Enable usage tracking for all quota types. */
4980static int ext4_enable_quotas(struct super_block *sb)
4981{
4982 int type, err = 0;
4983 unsigned long qf_inums[MAXQUOTAS] = {
4984 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4985 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4986 };
4987
4988 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4989 for (type = 0; type < MAXQUOTAS; type++) {
4990 if (qf_inums[type]) {
4991 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4992 DQUOT_USAGE_ENABLED);
4993 if (err) {
4994 ext4_warning(sb,
4995 "Failed to enable quota (type=%d) "
4996 "tracking. Please run e2fsck to fix.",
4997 type);
4998 return err;
4999 }
5000 }
5001 }
5002 return 0;
5003}
5004
5005/*
5006 * quota_on function that is used when QUOTA feature is set.
5007 */
5008static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5009 int format_id)
5010{
5011 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5012 return -EINVAL;
5013
5014 /*
5015 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5016 */
5017 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5018}
5019
ca0e05e4
DM		 5020static int ext4_quota_off(struct super_block *sb, int type)
5021{
21f97697
JK		 5022 struct inode *inode = sb_dqopt(sb)->files[type];
5023 handle_t *handle;
5024
87009d86
DM		 5025 /* Force all delayed allocation blocks to be allocated.
5026 * Caller already holds s_umount sem */
5027 if (test_opt(sb, DELALLOC))
ca0e05e4 5028 sync_filesystem(sb);
ca0e05e4 5029
0b268590
AG		 5030 if (!inode)
5031 goto out;
5032
21f97697
JK		 5033 /* Update modification times of quota files when userspace can
5034 * start looking at them */
5035 handle = ext4_journal_start(inode, 1);
5036 if (IS_ERR(handle))
5037 goto out;
5038 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5039 ext4_mark_inode_dirty(handle, inode);
5040 ext4_journal_stop(handle);
5041
5042out:
ca0e05e4
DM		 5043 return dquot_quota_off(sb, type);
5044}
5045
7c319d32
AK		 5046/*
5047 * quota_off function that is used when QUOTA feature is set.
5048 */
5049static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5050{
5051 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5052 return -EINVAL;
5053
5054 /* Disable only the limits. */
5055 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5056}
5057
ac27a0ec
DK		 5058/* Read data from quotafile - avoid pagecache and such because we cannot afford
5059 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5060 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5061 * we don't have to be afraid of races */
617ba13b 5062static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5063 size_t len, loff_t off)
5064{
5065 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5066 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5067 int err = 0;
5068 int offset = off & (sb->s_blocksize - 1);
5069 int tocopy;
5070 size_t toread;
5071 struct buffer_head *bh;
5072 loff_t i_size = i_size_read(inode);
5073
5074 if (off > i_size)
5075 return 0;
5076 if (off+len > i_size)
5077 len = i_size-off;
5078 toread = len;
5079 while (toread > 0) {
5080 tocopy = sb->s_blocksize - offset < toread ?
5081 sb->s_blocksize - offset : toread;
617ba13b 5082 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5083 if (err)
5084 return err;
5085 if (!bh) /* A hole? */
5086 memset(data, 0, tocopy);
5087 else
5088 memcpy(data, bh->b_data+offset, tocopy);
5089 brelse(bh);
5090 offset = 0;
5091 toread -= tocopy;
5092 data += tocopy;
5093 blk++;
5094 }
5095 return len;
5096}
5097
5098/* Write to quotafile (we know the transaction is already started and has
5099 * enough credits) */
617ba13b 5100static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5101 const char *data, size_t len, loff_t off)
5102{
5103 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5104 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5105 int err = 0;
5106 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5107 struct buffer_head *bh;
5108 handle_t *handle = journal_current_handle();
5109
0390131b 5110 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5111 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5112 " cancelled because transaction is not started",
9c3013e9
JK		 5113 (unsigned long long)off, (unsigned long long)len);
5114 return -EIO;
5115 }
67eeb568
DM		 5116 /*
5117 * Since we account only one data block in transaction credits,
5118 * then it is impossible to cross a block boundary.
5119 */
5120 if (sb->s_blocksize - offset < len) {
5121 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5122 " cancelled because not block aligned",
5123 (unsigned long long)off, (unsigned long long)len);
5124 return -EIO;
5125 }
5126
67eeb568
DM		 5127 bh = ext4_bread(handle, inode, blk, 1, &err);
5128 if (!bh)
5129 goto out;
62d2b5f2
JK		 5130 err = ext4_journal_get_write_access(handle, bh);
5131 if (err) {
5132 brelse(bh);
5133 goto out;
ac27a0ec 5134 }
67eeb568
DM		 5135 lock_buffer(bh);
5136 memcpy(bh->b_data+offset, data, len);
5137 flush_dcache_page(bh->b_page);
5138 unlock_buffer(bh);
62d2b5f2 5139 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5140 brelse(bh);
ac27a0ec 5141out:
0b7f7cef 5142 if (err)
ac27a0ec 5143 return err;
67eeb568
DM		 5144 if (inode->i_size < off + len) {
5145 i_size_write(inode, off + len);
617ba13b 5146 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5147 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5148 }
67eeb568 5149 return len;
ac27a0ec
DK		 5150}
5151
5152#endif
5153
152a0836
AV		 5154static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5155 const char *dev_name, void *data)
ac27a0ec 5156{
152a0836 5157 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5158}
5159
37f328eb 5160#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5161static inline void register_as_ext2(void)
5162{
5163 int err = register_filesystem(&ext2_fs_type);
5164 if (err)
5165 printk(KERN_WARNING
5166 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5167}
5168
5169static inline void unregister_as_ext2(void)
5170{
5171 unregister_filesystem(&ext2_fs_type);
5172}
2035e776
TT		 5173
5174static inline int ext2_feature_set_ok(struct super_block *sb)
5175{
5176 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5177 return 0;
5178 if (sb->s_flags & MS_RDONLY)
5179 return 1;
5180 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5181 return 0;
5182 return 1;
5183}
51b7e3c9 5184MODULE_ALIAS("ext2");
24b58424
TT		 5185#else
5186static inline void register_as_ext2(void) { }
5187static inline void unregister_as_ext2(void) { }
2035e776 5188static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5189#endif
5190
37f328eb 5191#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5192static inline void register_as_ext3(void)
5193{
5194 int err = register_filesystem(&ext3_fs_type);
5195 if (err)
5196 printk(KERN_WARNING
5197 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5198}
5199
5200static inline void unregister_as_ext3(void)
5201{
5202 unregister_filesystem(&ext3_fs_type);
5203}
2035e776
TT		 5204
5205static inline int ext3_feature_set_ok(struct super_block *sb)
5206{
5207 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5208 return 0;
5209 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5210 return 0;
5211 if (sb->s_flags & MS_RDONLY)
5212 return 1;
5213 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5214 return 0;
5215 return 1;
5216}
51b7e3c9 5217MODULE_ALIAS("ext3");
24b58424
TT		 5218#else
5219static inline void register_as_ext3(void) { }
5220static inline void unregister_as_ext3(void) { }
2035e776 5221static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5222#endif
5223
03010a33
TT		 5224static struct file_system_type ext4_fs_type = {
5225 .owner = THIS_MODULE,
5226 .name = "ext4",
152a0836 5227 .mount = ext4_mount,
03010a33
TT		 5228 .kill_sb = kill_block_super,
5229 .fs_flags = FS_REQUIRES_DEV,
5230};
5231
8f021222 5232static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5233{
5234 struct ext4_features *ef;
5235 int ret = -ENOMEM;
5236
5237 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5238 if (!ef)
5239 goto out;
5240
5241 ef->f_kobj.kset = ext4_kset;
5242 init_completion(&ef->f_kobj_unregister);
5243 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5244 "features");
5245 if (ret) {
5246 kfree(ef);
5247 goto out;
5248 }
5249
5250 ext4_feat = ef;
5251 ret = 0;
5252out:
5253 return ret;
5254}
5255
8f021222
LC		 5256static void ext4_exit_feat_adverts(void)
5257{
5258 kobject_put(&ext4_feat->f_kobj);
5259 wait_for_completion(&ext4_feat->f_kobj_unregister);
5260 kfree(ext4_feat);
5261}
5262
e9e3bcec
ES		 5263/* Shared across all ext4 file systems */
5264wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5265struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5266
5dabfc78 5267static int __init ext4_init_fs(void)
ac27a0ec 5268{
e9e3bcec 5269 int i, err;
c9de560d 5270
07c0c5d8
AV		 5271 ext4_li_info = NULL;
5272 mutex_init(&ext4_li_mtx);
5273
12e9b892 5274 ext4_check_flag_values();
e9e3bcec
ES		 5275
5276 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5277 mutex_init(&ext4__aio_mutex[i]);
5278 init_waitqueue_head(&ext4__ioend_wq[i]);
5279 }
5280
5dabfc78 5281 err = ext4_init_pageio();
6fd058f7
TT		 5282 if (err)
5283 return err;
5dabfc78 5284 err = ext4_init_system_zone();
bd2d0210 5285 if (err)
d44651d0 5286 goto out6;
3197ebdb
TT		 5287 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5288 if (!ext4_kset)
dd68314c 5289 goto out5;
d44651d0 5290 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5291
5292 err = ext4_init_feat_adverts();
dd68314c
TT		 5293 if (err)
5294 goto out4;
857ac889 5295
5dabfc78 5296 err = ext4_init_mballoc();
ac27a0ec 5297 if (err)
6fd058f7 5298 goto out3;
c9de560d 5299
5dabfc78 5300 err = ext4_init_xattr();
c9de560d
AT		 5301 if (err)
5302 goto out2;
ac27a0ec
DK		 5303 err = init_inodecache();
5304 if (err)
5305 goto out1;
24b58424 5306 register_as_ext3();
2035e776 5307 register_as_ext2();
03010a33 5308 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5309 if (err)
5310 goto out;
bfff6873 5311
ac27a0ec
DK		 5312 return 0;
5313out:
24b58424
TT		 5314 unregister_as_ext2();
5315 unregister_as_ext3();
ac27a0ec
DK		 5316 destroy_inodecache();
5317out1:
5dabfc78 5318 ext4_exit_xattr();
c9de560d 5319out2:
5dabfc78 5320 ext4_exit_mballoc();
6fd058f7 5321out3:
8f021222 5322 ext4_exit_feat_adverts();
dd68314c 5323out4:
d44651d0
FJ		 5324 if (ext4_proc_root)
5325 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5326 kset_unregister(ext4_kset);
d44651d0 5327out5:
5dabfc78 5328 ext4_exit_system_zone();
d44651d0 5329out6:
5dabfc78 5330 ext4_exit_pageio();
ac27a0ec
DK		 5331 return err;
5332}
5333
5dabfc78 5334static void __exit ext4_exit_fs(void)
ac27a0ec 5335{
bfff6873 5336 ext4_destroy_lazyinit_thread();
24b58424
TT		 5337 unregister_as_ext2();
5338 unregister_as_ext3();
03010a33 5339 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5340 destroy_inodecache();
5dabfc78
TT		 5341 ext4_exit_xattr();
5342 ext4_exit_mballoc();
8f021222 5343 ext4_exit_feat_adverts();
9f6200bb 5344 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5345 kset_unregister(ext4_kset);
5dabfc78
TT		 5346 ext4_exit_system_zone();
5347 ext4_exit_pageio();
ac27a0ec
DK		 5348}
5349
5350MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5351MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5352MODULE_LICENSE("GPL");
5dabfc78
TT		 5353module_init(ext4_init_fs)
5354module_exit(ext4_exit_fs)
kernel source for telekom puls (alcatel/mediatek)