]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: ext4_inode_info diet
[thirdparty/kernel/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
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);
ba39ebb6 434 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 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
0390131b
FM		 864 if (sbi->s_journal) {
865 err = jbd2_journal_destroy(sbi->s_journal);
866 sbi->s_journal = NULL;
867 if (err < 0)
c67d859e 868 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 869 }
d4edac31 870
a1c6c569 871 del_timer(&sbi->s_err_report);
d4edac31
JB		 872 ext4_release_system_zone(sb);
873 ext4_mb_release(sb);
874 ext4_ext_release(sb);
875 ext4_xattr_put_super(sb);
876
ac27a0ec 877 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 878 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 879 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 880 }
58c5873a 881 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 882 ext4_commit_super(sb, 1);
883
240799cd 884 if (sbi->s_proc) {
66acdcf4 885 remove_proc_entry("options", sbi->s_proc);
9f6200bb 886 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 887 }
3197ebdb 888 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 889
890 for (i = 0; i < sbi->s_gdb_count; i++)
891 brelse(sbi->s_group_desc[i]);
f18a5f21 892 ext4_kvfree(sbi->s_group_desc);
9933fc0a 893 ext4_kvfree(sbi->s_flex_groups);
57042651 894 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 895 percpu_counter_destroy(&sbi->s_freeinodes_counter);
896 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 897 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 898 brelse(sbi->s_sbh);
899#ifdef CONFIG_QUOTA
900 for (i = 0; i < MAXQUOTAS; i++)
901 kfree(sbi->s_qf_names[i]);
902#endif
903
904 /* Debugging code just in case the in-memory inode orphan list
905 * isn't empty. The on-disk one can be non-empty if we've
906 * detected an error and taken the fs readonly, but the
907 * in-memory list had better be clean by this point. */
908 if (!list_empty(&sbi->s_orphan))
909 dump_orphan_list(sb, sbi);
910 J_ASSERT(list_empty(&sbi->s_orphan));
911
f98393a6 912 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 913 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
914 /*
915 * Invalidate the journal device's buffers. We don't want them
916 * floating about in memory - the physical journal device may
917 * hotswapped, and it breaks the `ro-after' testing code.
918 */
919 sync_blockdev(sbi->journal_bdev);
f98393a6 920 invalidate_bdev(sbi->journal_bdev);
617ba13b 921 ext4_blkdev_remove(sbi);
ac27a0ec 922 }
c5e06d10
JL		 923 if (sbi->s_mmp_tsk)
924 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 925 sb->s_fs_info = NULL;
3197ebdb
TT		 926 /*
927 * Now that we are completely done shutting down the
928 * superblock, we need to actually destroy the kobject.
929 */
3197ebdb
TT		 930 kobject_put(&sbi->s_kobj);
931 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 932 if (sbi->s_chksum_driver)
933 crypto_free_shash(sbi->s_chksum_driver);
705895b6 934 kfree(sbi->s_blockgroup_lock);
ac27a0ec 935 kfree(sbi);
ac27a0ec
DK		 936}
937
e18b890b 938static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 939
940/*
941 * Called inside transaction, so use GFP_NOFS
942 */
617ba13b 943static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 944{
617ba13b 945 struct ext4_inode_info *ei;
ac27a0ec 946
e6b4f8da 947 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 948 if (!ei)
949 return NULL;
0b8e58a1 950
ac27a0ec 951 ei->vfs_inode.i_version = 1;
91246c00 952 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 953 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 954 INIT_LIST_HEAD(&ei->i_prealloc_list);
955 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 956 ei->i_reserved_data_blocks = 0;
957 ei->i_reserved_meta_blocks = 0;
958 ei->i_allocated_meta_blocks = 0;
9d0be502 959 ei->i_da_metadata_calc_len = 0;
7e731bc9 960 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 961 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 962#ifdef CONFIG_QUOTA
963 ei->i_reserved_quota = 0;
964#endif
8aefcd55 965 ei->jinode = NULL;
c7064ef1 966 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 967 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 968 ei->i_sync_tid = 0;
969 ei->i_datasync_tid = 0;
f7ad6d2e 970 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 971 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 972
ac27a0ec
DK		 973 return &ei->vfs_inode;
974}
975
7ff9c073
TT		 976static int ext4_drop_inode(struct inode *inode)
977{
978 int drop = generic_drop_inode(inode);
979
980 trace_ext4_drop_inode(inode, drop);
981 return drop;
982}
983
fa0d7e3d
NP		 984static void ext4_i_callback(struct rcu_head *head)
985{
986 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 987 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
988}
989
617ba13b 990static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 991{
9f7dd93d 992 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 993 ext4_msg(inode->i_sb, KERN_ERR,
994 "Inode %lu (%p): orphan list check failed!",
995 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 996 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
997 EXT4_I(inode), sizeof(struct ext4_inode_info),
998 true);
999 dump_stack();
1000 }
fa0d7e3d 1001 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1002}
1003
51cc5068 1004static void init_once(void *foo)
ac27a0ec 1005{
617ba13b 1006 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1007
a35afb83 1008 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 1009#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 1010 init_rwsem(&ei->xattr_sem);
ac27a0ec 1011#endif
0e855ac8 1012 init_rwsem(&ei->i_data_sem);
a35afb83 1013 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1014}
1015
1016static int init_inodecache(void)
1017{
617ba13b
MC		 1018 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1019 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 1020 0, (SLAB_RECLAIM_ACCOUNT|
1021 SLAB_MEM_SPREAD),
20c2df83 1022 init_once);
617ba13b 1023 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1024 return -ENOMEM;
1025 return 0;
1026}
1027
1028static void destroy_inodecache(void)
1029{
617ba13b 1030 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1031}
1032
0930fcc1 1033void ext4_clear_inode(struct inode *inode)
ac27a0ec 1034{
0930fcc1 1035 invalidate_inode_buffers(inode);
dbd5768f 1036 clear_inode(inode);
9f754758 1037 dquot_drop(inode);
c2ea3fde 1038 ext4_discard_preallocations(inode);
8aefcd55
TT		 1039 if (EXT4_I(inode)->jinode) {
1040 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1041 EXT4_I(inode)->jinode);
1042 jbd2_free_inode(EXT4_I(inode)->jinode);
1043 EXT4_I(inode)->jinode = NULL;
1044 }
ac27a0ec
DK		 1045}
1046
1b961ac0 1047static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1048 u64 ino, u32 generation)
ac27a0ec 1049{
ac27a0ec 1050 struct inode *inode;
ac27a0ec 1051
617ba13b 1052 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1053 return ERR_PTR(-ESTALE);
617ba13b 1054 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1055 return ERR_PTR(-ESTALE);
1056
1057 /* iget isn't really right if the inode is currently unallocated!!
1058 *
617ba13b 1059 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1060 * deleted, so we should be safe.
1061 *
1062 * Currently we don't know the generation for parent directory, so
1063 * a generation of 0 means "accept any"
1064 */
1d1fe1ee
DH		 1065 inode = ext4_iget(sb, ino);
1066 if (IS_ERR(inode))
1067 return ERR_CAST(inode);
1068 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1069 iput(inode);
1070 return ERR_PTR(-ESTALE);
1071 }
1b961ac0
CH		 1072
1073 return inode;
1074}
1075
1076static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1077 int fh_len, int fh_type)
1b961ac0
CH		 1078{
1079 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1080 ext4_nfs_get_inode);
1081}
1082
1083static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1084 int fh_len, int fh_type)
1b961ac0
CH		 1085{
1086 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1087 ext4_nfs_get_inode);
ac27a0ec
DK		 1088}
1089
c39a7f84
TO		 1090/*
1091 * Try to release metadata pages (indirect blocks, directories) which are
1092 * mapped via the block device. Since these pages could have journal heads
1093 * which would prevent try_to_free_buffers() from freeing them, we must use
1094 * jbd2 layer's try_to_free_buffers() function to release them.
1095 */
0b8e58a1
AD		 1096static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1097 gfp_t wait)
c39a7f84
TO		 1098{
1099 journal_t *journal = EXT4_SB(sb)->s_journal;
1100
1101 WARN_ON(PageChecked(page));
1102 if (!page_has_buffers(page))
1103 return 0;
1104 if (journal)
1105 return jbd2_journal_try_to_free_buffers(journal, page,
1106 wait & ~__GFP_WAIT);
1107 return try_to_free_buffers(page);
1108}
1109
ac27a0ec 1110#ifdef CONFIG_QUOTA
af5bc92d 1111#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1112#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1113
617ba13b
MC		 1114static int ext4_write_dquot(struct dquot *dquot);
1115static int ext4_acquire_dquot(struct dquot *dquot);
1116static int ext4_release_dquot(struct dquot *dquot);
1117static int ext4_mark_dquot_dirty(struct dquot *dquot);
1118static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1119static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1120 struct path *path);
7c319d32
AK		 1121static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1122 int format_id);
ca0e05e4 1123static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1124static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1125static int ext4_quota_on_mount(struct super_block *sb, int type);
1126static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1127 size_t len, loff_t off);
617ba13b 1128static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1129 const char *data, size_t len, loff_t off);
7c319d32
AK		 1130static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1131 unsigned int flags);
1132static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1133
61e225dc 1134static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1135 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1136 .write_dquot = ext4_write_dquot,
1137 .acquire_dquot = ext4_acquire_dquot,
1138 .release_dquot = ext4_release_dquot,
1139 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1140 .write_info = ext4_write_info,
1141 .alloc_dquot = dquot_alloc,
1142 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1143};
1144
0d54b217 1145static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1146 .quota_on = ext4_quota_on,
ca0e05e4 1147 .quota_off = ext4_quota_off,
287a8095
CH		 1148 .quota_sync = dquot_quota_sync,
1149 .get_info = dquot_get_dqinfo,
1150 .set_info = dquot_set_dqinfo,
1151 .get_dqblk = dquot_get_dqblk,
1152 .set_dqblk = dquot_set_dqblk
ac27a0ec 1153};
7c319d32
AK		 1154
1155static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1156 .quota_on_meta = ext4_quota_on_sysfile,
1157 .quota_off = ext4_quota_off_sysfile,
1158 .quota_sync = dquot_quota_sync,
1159 .get_info = dquot_get_dqinfo,
1160 .set_info = dquot_set_dqinfo,
1161 .get_dqblk = dquot_get_dqblk,
1162 .set_dqblk = dquot_set_dqblk
1163};
ac27a0ec
DK		 1164#endif
1165
ee9b6d61 1166static const struct super_operations ext4_sops = {
617ba13b
MC		 1167 .alloc_inode = ext4_alloc_inode,
1168 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1169 .write_inode = ext4_write_inode,
1170 .dirty_inode = ext4_dirty_inode,
7ff9c073 1171 .drop_inode = ext4_drop_inode,
0930fcc1 1172 .evict_inode = ext4_evict_inode,
617ba13b 1173 .put_super = ext4_put_super,
617ba13b 1174 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1175 .freeze_fs = ext4_freeze,
1176 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1177 .statfs = ext4_statfs,
1178 .remount_fs = ext4_remount,
617ba13b 1179 .show_options = ext4_show_options,
ac27a0ec 1180#ifdef CONFIG_QUOTA
617ba13b
MC		 1181 .quota_read = ext4_quota_read,
1182 .quota_write = ext4_quota_write,
ac27a0ec 1183#endif
c39a7f84 1184 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1185};
1186
9ca92389
TT		 1187static const struct super_operations ext4_nojournal_sops = {
1188 .alloc_inode = ext4_alloc_inode,
1189 .destroy_inode = ext4_destroy_inode,
1190 .write_inode = ext4_write_inode,
1191 .dirty_inode = ext4_dirty_inode,
7ff9c073 1192 .drop_inode = ext4_drop_inode,
0930fcc1 1193 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1194 .put_super = ext4_put_super,
1195 .statfs = ext4_statfs,
1196 .remount_fs = ext4_remount,
9ca92389
TT		 1197 .show_options = ext4_show_options,
1198#ifdef CONFIG_QUOTA
1199 .quota_read = ext4_quota_read,
1200 .quota_write = ext4_quota_write,
1201#endif
1202 .bdev_try_to_free_page = bdev_try_to_free_page,
1203};
1204
39655164 1205static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1206 .fh_to_dentry = ext4_fh_to_dentry,
1207 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1208 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1209};
1210
1211enum {
1212 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1213 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1214 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1215 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1216 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1217 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1218 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1219 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1220 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1221 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1222 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1223 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1224 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1225 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1226 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1227 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1228 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1229 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1230 Opt_max_dir_size_kb,
ac27a0ec
DK		 1231};
1232
a447c093 1233static const match_table_t tokens = {
ac27a0ec
DK		 1234 {Opt_bsd_df, "bsddf"},
1235 {Opt_minix_df, "minixdf"},
1236 {Opt_grpid, "grpid"},
1237 {Opt_grpid, "bsdgroups"},
1238 {Opt_nogrpid, "nogrpid"},
1239 {Opt_nogrpid, "sysvgroups"},
1240 {Opt_resgid, "resgid=%u"},
1241 {Opt_resuid, "resuid=%u"},
1242 {Opt_sb, "sb=%u"},
1243 {Opt_err_cont, "errors=continue"},
1244 {Opt_err_panic, "errors=panic"},
1245 {Opt_err_ro, "errors=remount-ro"},
1246 {Opt_nouid32, "nouid32"},
ac27a0ec 1247 {Opt_debug, "debug"},
72578c33
TT		 1248 {Opt_removed, "oldalloc"},
1249 {Opt_removed, "orlov"},
ac27a0ec
DK		 1250 {Opt_user_xattr, "user_xattr"},
1251 {Opt_nouser_xattr, "nouser_xattr"},
1252 {Opt_acl, "acl"},
1253 {Opt_noacl, "noacl"},
e3bb52ae 1254 {Opt_noload, "norecovery"},
5a916be1 1255 {Opt_noload, "noload"},
72578c33
TT		 1256 {Opt_removed, "nobh"},
1257 {Opt_removed, "bh"},
ac27a0ec 1258 {Opt_commit, "commit=%u"},
30773840
TT		 1259 {Opt_min_batch_time, "min_batch_time=%u"},
1260 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1261 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1262 {Opt_journal_checksum, "journal_checksum"},
1263 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1264 {Opt_abort, "abort"},
1265 {Opt_data_journal, "data=journal"},
1266 {Opt_data_ordered, "data=ordered"},
1267 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1268 {Opt_data_err_abort, "data_err=abort"},
1269 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1270 {Opt_offusrjquota, "usrjquota="},
1271 {Opt_usrjquota, "usrjquota=%s"},
1272 {Opt_offgrpjquota, "grpjquota="},
1273 {Opt_grpjquota, "grpjquota=%s"},
1274 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1275 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1276 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1277 {Opt_grpquota, "grpquota"},
1278 {Opt_noquota, "noquota"},
1279 {Opt_quota, "quota"},
1280 {Opt_usrquota, "usrquota"},
1281 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1282 {Opt_barrier, "barrier"},
1283 {Opt_nobarrier, "nobarrier"},
25ec56b5 1284 {Opt_i_version, "i_version"},
c9de560d 1285 {Opt_stripe, "stripe=%u"},
64769240 1286 {Opt_delalloc, "delalloc"},
dd919b98 1287 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1288 {Opt_mblk_io_submit, "mblk_io_submit"},
1289 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1290 {Opt_block_validity, "block_validity"},
1291 {Opt_noblock_validity, "noblock_validity"},
240799cd 1292 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1293 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1294 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1295 {Opt_auto_da_alloc, "auto_da_alloc"},
1296 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1297 {Opt_dioread_nolock, "dioread_nolock"},
1298 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1299 {Opt_discard, "discard"},
1300 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1301 {Opt_init_itable, "init_itable=%u"},
1302 {Opt_init_itable, "init_itable"},
1303 {Opt_noinit_itable, "noinit_itable"},
df981d03 1304 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1305 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1306 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1307 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1308 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1309 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1310 {Opt_err, NULL},
ac27a0ec
DK		 1311};
1312
617ba13b 1313static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1314{
617ba13b 1315 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1316 char *options = (char *) *data;
1317
1318 if (!options || strncmp(options, "sb=", 3) != 0)
1319 return 1; /* Default location */
0b8e58a1 1320
ac27a0ec 1321 options += 3;
0b8e58a1 1322 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1323 sb_block = simple_strtoul(options, &options, 0);
1324 if (*options && *options != ',') {
4776004f 1325 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1326 (char *) *data);
1327 return 1;
1328 }
1329 if (*options == ',')
1330 options++;
1331 *data = (void *) options;
0b8e58a1 1332
ac27a0ec
DK		 1333 return sb_block;
1334}
1335
b3881f74 1336#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1337static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1338 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1339
56c50f11
DM		 1340#ifdef CONFIG_QUOTA
1341static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1342{
1343 struct ext4_sb_info *sbi = EXT4_SB(sb);
1344 char *qname;
1345
1346 if (sb_any_quota_loaded(sb) &&
1347 !sbi->s_qf_names[qtype]) {
1348 ext4_msg(sb, KERN_ERR,
1349 "Cannot change journaled "
1350 "quota options when quota turned on");
57f73c2c 1351 return -1;
56c50f11
DM		 1352 }
1353 qname = match_strdup(args);
1354 if (!qname) {
1355 ext4_msg(sb, KERN_ERR,
1356 "Not enough memory for storing quotafile name");
57f73c2c 1357 return -1;
56c50f11
DM		 1358 }
1359 if (sbi->s_qf_names[qtype] &&
1360 strcmp(sbi->s_qf_names[qtype], qname)) {
1361 ext4_msg(sb, KERN_ERR,
1362 "%s quota file already specified", QTYPE2NAME(qtype));
1363 kfree(qname);
57f73c2c 1364 return -1;
56c50f11
DM		 1365 }
1366 sbi->s_qf_names[qtype] = qname;
1367 if (strchr(sbi->s_qf_names[qtype], '/')) {
1368 ext4_msg(sb, KERN_ERR,
1369 "quotafile must be on filesystem root");
1370 kfree(sbi->s_qf_names[qtype]);
1371 sbi->s_qf_names[qtype] = NULL;
57f73c2c 1372 return -1;
56c50f11 1373 }
fd8c37ec 1374 set_opt(sb, QUOTA);
56c50f11
DM		 1375 return 1;
1376}
1377
1378static int clear_qf_name(struct super_block *sb, int qtype)
1379{
1380
1381 struct ext4_sb_info *sbi = EXT4_SB(sb);
1382
1383 if (sb_any_quota_loaded(sb) &&
1384 sbi->s_qf_names[qtype]) {
1385 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1386 " when quota turned on");
57f73c2c 1387 return -1;
56c50f11
DM		 1388 }
1389 /*
1390 * The space will be released later when all options are confirmed
1391 * to be correct
1392 */
1393 sbi->s_qf_names[qtype] = NULL;
1394 return 1;
1395}
1396#endif
1397
26092bf5
TT		 1398#define MOPT_SET 0x0001
1399#define MOPT_CLEAR 0x0002
1400#define MOPT_NOSUPPORT 0x0004
1401#define MOPT_EXPLICIT 0x0008
1402#define MOPT_CLEAR_ERR 0x0010
1403#define MOPT_GTE0 0x0020
ac27a0ec 1404#ifdef CONFIG_QUOTA
26092bf5
TT		 1405#define MOPT_Q 0
1406#define MOPT_QFMT 0x0040
1407#else
1408#define MOPT_Q MOPT_NOSUPPORT
1409#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1410#endif
26092bf5
TT		 1411#define MOPT_DATAJ 0x0080
1412
1413static const struct mount_opts {
1414 int token;
1415 int mount_opt;
1416 int flags;
1417} ext4_mount_opts[] = {
1418 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1419 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1420 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1421 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1422 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1423 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1424 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1425 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1426 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1427 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1428 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1429 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1430 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1431 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1432 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1433 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1434 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1435 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1436 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1437 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1438 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1439 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1440 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1441 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1442 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1443 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1444 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1445 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1446 {Opt_commit, 0, MOPT_GTE0},
1447 {Opt_max_batch_time, 0, MOPT_GTE0},
1448 {Opt_min_batch_time, 0, MOPT_GTE0},
1449 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1450 {Opt_init_itable, 0, MOPT_GTE0},
1451 {Opt_stripe, 0, MOPT_GTE0},
1452 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1453 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1454 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
03010a33 1455#ifdef CONFIG_EXT4_FS_XATTR
26092bf5
TT		 1456 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1457 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
ac27a0ec 1458#else
26092bf5
TT		 1459 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1460 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
ac27a0ec 1461#endif
03010a33 1462#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1463 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1464 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1465#else
26092bf5
TT		 1466 {Opt_acl, 0, MOPT_NOSUPPORT},
1467 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1468#endif
26092bf5
TT		 1469 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1470 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1471 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1472 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1473 MOPT_SET | MOPT_Q},
1474 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1475 MOPT_SET | MOPT_Q},
1476 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1477 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1478 {Opt_usrjquota, 0, MOPT_Q},
1479 {Opt_grpjquota, 0, MOPT_Q},
1480 {Opt_offusrjquota, 0, MOPT_Q},
1481 {Opt_offgrpjquota, 0, MOPT_Q},
1482 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1483 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1484 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1485 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1486 {Opt_err, 0, 0}
1487};
1488
1489static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1490 substring_t *args, unsigned long *journal_devnum,
1491 unsigned int *journal_ioprio, int is_remount)
1492{
1493 struct ext4_sb_info *sbi = EXT4_SB(sb);
1494 const struct mount_opts *m;
08cefc7a
EB		 1495 kuid_t uid;
1496 kgid_t gid;
26092bf5
TT		 1497 int arg = 0;
1498
57f73c2c
TT		 1499#ifdef CONFIG_QUOTA
1500 if (token == Opt_usrjquota)
1501 return set_qf_name(sb, USRQUOTA, &args[0]);
1502 else if (token == Opt_grpjquota)
1503 return set_qf_name(sb, GRPQUOTA, &args[0]);
1504 else if (token == Opt_offusrjquota)
1505 return clear_qf_name(sb, USRQUOTA);
1506 else if (token == Opt_offgrpjquota)
1507 return clear_qf_name(sb, GRPQUOTA);
1508#endif
26092bf5
TT		 1509 if (args->from && match_int(args, &arg))
1510 return -1;
1511 switch (token) {
f7048605
TT		 1512 case Opt_noacl:
1513 case Opt_nouser_xattr:
1514 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1515 break;
26092bf5
TT		 1516 case Opt_sb:
1517 return 1; /* handled by get_sb_block() */
1518 case Opt_removed:
1519 ext4_msg(sb, KERN_WARNING,
1520 "Ignoring removed %s option", opt);
1521 return 1;
1522 case Opt_resuid:
08cefc7a
EB		 1523 uid = make_kuid(current_user_ns(), arg);
1524 if (!uid_valid(uid)) {
1525 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1526 return -1;
1527 }
1528 sbi->s_resuid = uid;
26092bf5
TT		 1529 return 1;
1530 case Opt_resgid:
08cefc7a
EB		 1531 gid = make_kgid(current_user_ns(), arg);
1532 if (!gid_valid(gid)) {
1533 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1534 return -1;
1535 }
1536 sbi->s_resgid = gid;
26092bf5
TT		 1537 return 1;
1538 case Opt_abort:
1539 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1540 return 1;
1541 case Opt_i_version:
1542 sb->s_flags |= MS_I_VERSION;
1543 return 1;
1544 case Opt_journal_dev:
1545 if (is_remount) {
1546 ext4_msg(sb, KERN_ERR,
1547 "Cannot specify journal on remount");
1548 return -1;
1549 }
1550 *journal_devnum = arg;
1551 return 1;
1552 case Opt_journal_ioprio:
1553 if (arg < 0 || arg > 7)
1554 return -1;
1555 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1556 return 1;
1557 }
1558
1559 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1560 if (token != m->token)
1561 continue;
1562 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1563 return -1;
1564 if (m->flags & MOPT_EXPLICIT)
1565 set_opt2(sb, EXPLICIT_DELALLOC);
1566 if (m->flags & MOPT_CLEAR_ERR)
1567 clear_opt(sb, ERRORS_MASK);
1568 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1569 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1570 "options when quota turned on");
1571 return -1;
1572 }
1573
1574 if (m->flags & MOPT_NOSUPPORT) {
1575 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1576 } else if (token == Opt_commit) {
1577 if (arg == 0)
1578 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1579 sbi->s_commit_interval = HZ * arg;
1580 } else if (token == Opt_max_batch_time) {
1581 if (arg == 0)
1582 arg = EXT4_DEF_MAX_BATCH_TIME;
1583 sbi->s_max_batch_time = arg;
1584 } else if (token == Opt_min_batch_time) {
1585 sbi->s_min_batch_time = arg;
1586 } else if (token == Opt_inode_readahead_blks) {
1587 if (arg > (1 << 30))
1588 return -1;
1589 if (arg && !is_power_of_2(arg)) {
b31e1552 1590 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1591 "EXT4-fs: inode_readahead_blks"
1592 " must be a power of 2");
1593 return -1;
ac27a0ec 1594 }
26092bf5
TT		 1595 sbi->s_inode_readahead_blks = arg;
1596 } else if (token == Opt_init_itable) {
1597 set_opt(sb, INIT_INODE_TABLE);
1598 if (!args->from)
1599 arg = EXT4_DEF_LI_WAIT_MULT;
1600 sbi->s_li_wait_mult = arg;
df981d03
TT		 1601 } else if (token == Opt_max_dir_size_kb) {
1602 sbi->s_max_dir_size_kb = arg;
26092bf5
TT		 1603 } else if (token == Opt_stripe) {
1604 sbi->s_stripe = arg;
1605 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1606 if (is_remount) {
eb513689
TT		 1607 if (!sbi->s_journal)
1608 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1609 else if (test_opt(sb, DATA_FLAGS) !=
1610 m->mount_opt) {
b31e1552 1611 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1612 "Cannot change data mode on remount");
1613 return -1;
ac27a0ec
DK		 1614 }
1615 } else {
fd8c37ec 1616 clear_opt(sb, DATA_FLAGS);
26092bf5 1617 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1618 }
ac27a0ec 1619#ifdef CONFIG_QUOTA
26092bf5 1620 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1621 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1622 sbi->s_jquota_fmt != m->mount_opt) {
1623 ext4_msg(sb, KERN_ERR, "Cannot "
1624 "change journaled quota options "
1625 "when quota turned on");
1626 return -1;
ac27a0ec 1627 }
26092bf5 1628 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1629#endif
26092bf5
TT		 1630 } else {
1631 if (!args->from)
1632 arg = 1;
1633 if (m->flags & MOPT_CLEAR)
1634 arg = !arg;
1635 else if (unlikely(!(m->flags & MOPT_SET))) {
1636 ext4_msg(sb, KERN_WARNING,
1637 "buggy handling of option %s", opt);
1638 WARN_ON(1);
1639 return -1;
3197ebdb 1640 }
26092bf5
TT		 1641 if (arg != 0)
1642 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1643 else
26092bf5 1644 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1645 }
26092bf5
TT		 1646 return 1;
1647 }
1648 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1649 "or missing value", opt);
1650 return -1;
1651}
1652
1653static int parse_options(char *options, struct super_block *sb,
1654 unsigned long *journal_devnum,
1655 unsigned int *journal_ioprio,
1656 int is_remount)
1657{
db7e5c66 1658#ifdef CONFIG_QUOTA
26092bf5 1659 struct ext4_sb_info *sbi = EXT4_SB(sb);
db7e5c66 1660#endif
26092bf5
TT		 1661 char *p;
1662 substring_t args[MAX_OPT_ARGS];
1663 int token;
1664
1665 if (!options)
1666 return 1;
1667
1668 while ((p = strsep(&options, ",")) != NULL) {
1669 if (!*p)
1670 continue;
1671 /*
1672 * Initialize args struct so we know whether arg was
1673 * found; some options take optional arguments.
1674 */
caecd0af 1675 args[0].to = args[0].from = NULL;
26092bf5
TT		 1676 token = match_token(p, tokens, args);
1677 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1678 journal_ioprio, is_remount) < 0)
1679 return 0;
ac27a0ec
DK		 1680 }
1681#ifdef CONFIG_QUOTA
1682 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1683 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1684 clear_opt(sb, USRQUOTA);
ac27a0ec 1685
482a7425 1686 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1687 clear_opt(sb, GRPQUOTA);
ac27a0ec 1688
56c50f11 1689 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1690 ext4_msg(sb, KERN_ERR, "old and new quota "
1691 "format mixing");
ac27a0ec
DK		 1692 return 0;
1693 }
1694
1695 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1696 ext4_msg(sb, KERN_ERR, "journaled quota format "
1697 "not specified");
ac27a0ec
DK		 1698 return 0;
1699 }
1700 } else {
1701 if (sbi->s_jquota_fmt) {
b31e1552 1702 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1703 "specified with no journaling "
b31e1552 1704 "enabled");
ac27a0ec
DK		 1705 return 0;
1706 }
1707 }
1708#endif
1709 return 1;
1710}
1711
2adf6da8
TT		 1712static inline void ext4_show_quota_options(struct seq_file *seq,
1713 struct super_block *sb)
1714{
1715#if defined(CONFIG_QUOTA)
1716 struct ext4_sb_info *sbi = EXT4_SB(sb);
1717
1718 if (sbi->s_jquota_fmt) {
1719 char *fmtname = "";
1720
1721 switch (sbi->s_jquota_fmt) {
1722 case QFMT_VFS_OLD:
1723 fmtname = "vfsold";
1724 break;
1725 case QFMT_VFS_V0:
1726 fmtname = "vfsv0";
1727 break;
1728 case QFMT_VFS_V1:
1729 fmtname = "vfsv1";
1730 break;
1731 }
1732 seq_printf(seq, ",jqfmt=%s", fmtname);
1733 }
1734
1735 if (sbi->s_qf_names[USRQUOTA])
1736 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1737
1738 if (sbi->s_qf_names[GRPQUOTA])
1739 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1740
1741 if (test_opt(sb, USRQUOTA))
1742 seq_puts(seq, ",usrquota");
1743
1744 if (test_opt(sb, GRPQUOTA))
1745 seq_puts(seq, ",grpquota");
1746#endif
1747}
1748
5a916be1
TT		 1749static const char *token2str(int token)
1750{
50df9fd5 1751 const struct match_token *t;
5a916be1
TT		 1752
1753 for (t = tokens; t->token != Opt_err; t++)
1754 if (t->token == token && !strchr(t->pattern, '='))
1755 break;
1756 return t->pattern;
1757}
1758
2adf6da8
TT		 1759/*
1760 * Show an option if
1761 * - it's set to a non-default value OR
1762 * - if the per-sb default is different from the global default
1763 */
66acdcf4
TT		 1764static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1765 int nodefs)
2adf6da8 1766{
2adf6da8
TT		 1767 struct ext4_sb_info *sbi = EXT4_SB(sb);
1768 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1769 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1770 const struct mount_opts *m;
66acdcf4 1771 char sep = nodefs ? '\n' : ',';
2adf6da8 1772
66acdcf4
TT		 1773#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1774#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1775
1776 if (sbi->s_sb_block != 1)
5a916be1
TT		 1777 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1778
1779 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1780 int want_set = m->flags & MOPT_SET;
1781 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1782 (m->flags & MOPT_CLEAR_ERR))
1783 continue;
66acdcf4 1784 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1785 continue; /* skip if same as the default */
1786 if ((want_set &&
1787 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1788 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1789 continue; /* select Opt_noFoo vs Opt_Foo */
1790 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1791 }
5a916be1 1792
08cefc7a 1793 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1794 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1795 SEQ_OPTS_PRINT("resuid=%u",
1796 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1797 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1798 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1799 SEQ_OPTS_PRINT("resgid=%u",
1800 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1801 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1802 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1803 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1804 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1805 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1806 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1807 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1808 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1809 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1810 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1811 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1812 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1813 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1814 if (sb->s_flags & MS_I_VERSION)
5a916be1 1815 SEQ_OPTS_PUTS("i_version");
66acdcf4 1816 if (nodefs || sbi->s_stripe)
5a916be1 1817 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1818 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1819 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1820 SEQ_OPTS_PUTS("data=journal");
1821 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1822 SEQ_OPTS_PUTS("data=ordered");
1823 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1824 SEQ_OPTS_PUTS("data=writeback");
1825 }
66acdcf4
TT		 1826 if (nodefs ||
1827 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1828 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1829 sbi->s_inode_readahead_blks);
2adf6da8 1830
66acdcf4
TT		 1831 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1832 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1833 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1834 if (nodefs || sbi->s_max_dir_size_kb)
1835 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1836
1837 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1838 return 0;
1839}
1840
66acdcf4
TT		 1841static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1842{
1843 return _ext4_show_options(seq, root->d_sb, 0);
1844}
1845
1846static int options_seq_show(struct seq_file *seq, void *offset)
1847{
1848 struct super_block *sb = seq->private;
1849 int rc;
1850
1851 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1852 rc = _ext4_show_options(seq, sb, 1);
1853 seq_puts(seq, "\n");
1854 return rc;
1855}
1856
1857static int options_open_fs(struct inode *inode, struct file *file)
1858{
1859 return single_open(file, options_seq_show, PDE(inode)->data);
1860}
1861
1862static const struct file_operations ext4_seq_options_fops = {
1863 .owner = THIS_MODULE,
1864 .open = options_open_fs,
1865 .read = seq_read,
1866 .llseek = seq_lseek,
1867 .release = single_release,
1868};
1869
617ba13b 1870static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1871 int read_only)
1872{
617ba13b 1873 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1874 int res = 0;
1875
617ba13b 1876 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1877 ext4_msg(sb, KERN_ERR, "revision level too high, "
1878 "forcing read-only mode");
ac27a0ec
DK		 1879 res = MS_RDONLY;
1880 }
1881 if (read_only)
281b5995 1882 goto done;
617ba13b 1883 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1884 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1885 "running e2fsck is recommended");
617ba13b 1886 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1887 ext4_msg(sb, KERN_WARNING,
1888 "warning: mounting fs with errors, "
1889 "running e2fsck is recommended");
ed3ce80a 1890 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1891 le16_to_cpu(es->s_mnt_count) >=
1892 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1893 ext4_msg(sb, KERN_WARNING,
1894 "warning: maximal mount count reached, "
1895 "running e2fsck is recommended");
ac27a0ec
DK		 1896 else if (le32_to_cpu(es->s_checkinterval) &&
1897 (le32_to_cpu(es->s_lastcheck) +
1898 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1899 ext4_msg(sb, KERN_WARNING,
1900 "warning: checktime reached, "
1901 "running e2fsck is recommended");
0b8e58a1 1902 if (!sbi->s_journal)
0390131b 1903 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1904 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1905 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1906 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1907 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1908 ext4_update_dynamic_rev(sb);
0390131b
FM		 1909 if (sbi->s_journal)
1910 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1911
e2d67052 1912 ext4_commit_super(sb, 1);
281b5995 1913done:
ac27a0ec 1914 if (test_opt(sb, DEBUG))
a9df9a49 1915 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1916 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1917 sb->s_blocksize,
1918 sbi->s_groups_count,
617ba13b
MC		 1919 EXT4_BLOCKS_PER_GROUP(sb),
1920 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1921 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1922
7abc52c2 1923 cleancache_init_fs(sb);
ac27a0ec
DK		 1924 return res;
1925}
1926
117fff10
TT		 1927int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1928{
1929 struct ext4_sb_info *sbi = EXT4_SB(sb);
1930 struct flex_groups *new_groups;
1931 int size;
1932
1933 if (!sbi->s_log_groups_per_flex)
1934 return 0;
1935
1936 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1937 if (size <= sbi->s_flex_groups_allocated)
1938 return 0;
1939
1940 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1941 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1942 if (!new_groups) {
1943 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1944 size / (int) sizeof(struct flex_groups));
1945 return -ENOMEM;
1946 }
1947
1948 if (sbi->s_flex_groups) {
1949 memcpy(new_groups, sbi->s_flex_groups,
1950 (sbi->s_flex_groups_allocated *
1951 sizeof(struct flex_groups)));
1952 ext4_kvfree(sbi->s_flex_groups);
1953 }
1954 sbi->s_flex_groups = new_groups;
1955 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1956 return 0;
1957}
1958
772cb7c8
JS		 1959static int ext4_fill_flex_info(struct super_block *sb)
1960{
1961 struct ext4_sb_info *sbi = EXT4_SB(sb);
1962 struct ext4_group_desc *gdp = NULL;
772cb7c8 1963 ext4_group_t flex_group;
d50f2ab6 1964 unsigned int groups_per_flex = 0;
117fff10 1965 int i, err;
772cb7c8 1966
503358ae 1967 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1968 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1969 sbi->s_log_groups_per_flex = 0;
1970 return 1;
1971 }
d50f2ab6 1972 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
772cb7c8 1973
117fff10
TT		 1974 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1975 if (err)
9933fc0a 1976 goto failed;
772cb7c8 1977
772cb7c8 1978 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1979 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1980
1981 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1982 atomic_add(ext4_free_inodes_count(sb, gdp),
1983 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1984 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1985 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1986 atomic_add(ext4_used_dirs_count(sb, gdp),
1987 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1988 }
1989
1990 return 1;
1991failed:
1992 return 0;
1993}
1994
feb0ab32
DW		 1995static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1996 struct ext4_group_desc *gdp)
717d50e4 1997{
feb0ab32 1998 int offset;
717d50e4 1999 __u16 crc = 0;
feb0ab32 2000 __le32 le_group = cpu_to_le32(block_group);
717d50e4 2001
feb0ab32
DW		 2002 if ((sbi->s_es->s_feature_ro_compat &
2003 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2004 /* Use new metadata_csum algorithm */
2005 __u16 old_csum;
2006 __u32 csum32;
2007
2008 old_csum = gdp->bg_checksum;
2009 gdp->bg_checksum = 0;
2010 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2011 sizeof(le_group));
2012 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2013 sbi->s_desc_size);
2014 gdp->bg_checksum = old_csum;
2015
2016 crc = csum32 & 0xFFFF;
2017 goto out;
717d50e4
AD		 2018 }
2019
feb0ab32
DW		 2020 /* old crc16 code */
2021 offset = offsetof(struct ext4_group_desc, bg_checksum);
2022
2023 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2024 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2025 crc = crc16(crc, (__u8 *)gdp, offset);
2026 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2027 /* for checksum of struct ext4_group_desc do the rest...*/
2028 if ((sbi->s_es->s_feature_incompat &
2029 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2030 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2031 crc = crc16(crc, (__u8 *)gdp + offset,
2032 le16_to_cpu(sbi->s_es->s_desc_size) -
2033 offset);
2034
2035out:
717d50e4
AD		 2036 return cpu_to_le16(crc);
2037}
2038
feb0ab32 2039int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2040 struct ext4_group_desc *gdp)
2041{
feb0ab32
DW		 2042 if (ext4_has_group_desc_csum(sb) &&
2043 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2044 block_group, gdp)))
717d50e4
AD		 2045 return 0;
2046
2047 return 1;
2048}
2049
feb0ab32
DW		 2050void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2051 struct ext4_group_desc *gdp)
2052{
2053 if (!ext4_has_group_desc_csum(sb))
2054 return;
2055 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2056}
2057
ac27a0ec 2058/* Called at mount-time, super-block is locked */
bfff6873
LC		 2059static int ext4_check_descriptors(struct super_block *sb,
2060 ext4_group_t *first_not_zeroed)
ac27a0ec 2061{
617ba13b
MC		 2062 struct ext4_sb_info *sbi = EXT4_SB(sb);
2063 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2064 ext4_fsblk_t last_block;
bd81d8ee
LV		 2065 ext4_fsblk_t block_bitmap;
2066 ext4_fsblk_t inode_bitmap;
2067 ext4_fsblk_t inode_table;
ce421581 2068 int flexbg_flag = 0;
bfff6873 2069 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2070
ce421581
JS		 2071 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2072 flexbg_flag = 1;
2073
af5bc92d 2074 ext4_debug("Checking group descriptors");
ac27a0ec 2075
197cd65a
AM		 2076 for (i = 0; i < sbi->s_groups_count; i++) {
2077 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2078
ce421581 2079 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2080 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2081 else
2082 last_block = first_block +
617ba13b 2083 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2084
bfff6873
LC		 2085 if ((grp == sbi->s_groups_count) &&
2086 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2087 grp = i;
2088
8fadc143 2089 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2090 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2091 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2092 "Block bitmap for group %u not in group "
b31e1552 2093 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2094 return 0;
2095 }
8fadc143 2096 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2097 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2098 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2099 "Inode bitmap for group %u not in group "
b31e1552 2100 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2101 return 0;
2102 }
8fadc143 2103 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2104 if (inode_table < first_block ||
2b2d6d01 2105 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2106 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2107 "Inode table for group %u not in group "
b31e1552 2108 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2109 return 0;
2110 }
955ce5f5 2111 ext4_lock_group(sb, i);
feb0ab32 2112 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2113 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2114 "Checksum for group %u failed (%u!=%u)",
2115 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2116 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2117 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2118 ext4_unlock_group(sb, i);
8a266467 2119 return 0;
7ee1ec4c 2120 }
717d50e4 2121 }
955ce5f5 2122 ext4_unlock_group(sb, i);
ce421581
JS		 2123 if (!flexbg_flag)
2124 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2125 }
bfff6873
LC		 2126 if (NULL != first_not_zeroed)
2127 *first_not_zeroed = grp;
ac27a0ec 2128
5dee5437
TT		 2129 ext4_free_blocks_count_set(sbi->s_es,
2130 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2131 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2132 return 1;
2133}
2134
617ba13b 2135/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2136 * the superblock) which were deleted from all directories, but held open by
2137 * a process at the time of a crash. We walk the list and try to delete these
2138 * inodes at recovery time (only with a read-write filesystem).
2139 *
2140 * In order to keep the orphan inode chain consistent during traversal (in
2141 * case of crash during recovery), we link each inode into the superblock
2142 * orphan list_head and handle it the same way as an inode deletion during
2143 * normal operation (which journals the operations for us).
2144 *
2145 * We only do an iget() and an iput() on each inode, which is very safe if we
2146 * accidentally point at an in-use or already deleted inode. The worst that
2147 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2148 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2149 * e2fsck was run on this filesystem, and it must have already done the orphan
2150 * inode cleanup for us, so we can safely abort without any further action.
2151 */
2b2d6d01
TT		 2152static void ext4_orphan_cleanup(struct super_block *sb,
2153 struct ext4_super_block *es)
ac27a0ec
DK		 2154{
2155 unsigned int s_flags = sb->s_flags;
2156 int nr_orphans = 0, nr_truncates = 0;
2157#ifdef CONFIG_QUOTA
2158 int i;
2159#endif
2160 if (!es->s_last_orphan) {
2161 jbd_debug(4, "no orphan inodes to clean up\n");
2162 return;
2163 }
2164
a8f48a95 2165 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2166 ext4_msg(sb, KERN_ERR, "write access "
2167 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2168 return;
2169 }
2170
d39195c3
AG		 2171 /* Check if feature set would not allow a r/w mount */
2172 if (!ext4_feature_set_ok(sb, 0)) {
2173 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2174 "unknown ROCOMPAT features");
2175 return;
2176 }
2177
617ba13b 2178 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2179 /* don't clear list on RO mount w/ errors */
2180 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2181 jbd_debug(1, "Errors on filesystem, "
2182 "clearing orphan list.\n");
c25f9bc6
ES		 2183 es->s_last_orphan = 0;
2184 }
ac27a0ec
DK		 2185 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2186 return;
2187 }
2188
2189 if (s_flags & MS_RDONLY) {
b31e1552 2190 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2191 sb->s_flags &= ~MS_RDONLY;
2192 }
2193#ifdef CONFIG_QUOTA
2194 /* Needed for iput() to work correctly and not trash data */
2195 sb->s_flags |= MS_ACTIVE;
2196 /* Turn on quotas so that they are updated correctly */
2197 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2198 if (EXT4_SB(sb)->s_qf_names[i]) {
2199 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2200 if (ret < 0)
b31e1552
ES		 2201 ext4_msg(sb, KERN_ERR,
2202 "Cannot turn on journaled "
2203 "quota: error %d", ret);
ac27a0ec
DK		 2204 }
2205 }
2206#endif
2207
2208 while (es->s_last_orphan) {
2209 struct inode *inode;
2210
97bd42b9
JB		 2211 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2212 if (IS_ERR(inode)) {
ac27a0ec
DK		 2213 es->s_last_orphan = 0;
2214 break;
2215 }
2216
617ba13b 2217 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2218 dquot_initialize(inode);
ac27a0ec 2219 if (inode->i_nlink) {
b31e1552
ES		 2220 ext4_msg(sb, KERN_DEBUG,
2221 "%s: truncating inode %lu to %lld bytes",
46e665e9 2222 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2223 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2224 inode->i_ino, inode->i_size);
617ba13b 2225 ext4_truncate(inode);
ac27a0ec
DK		 2226 nr_truncates++;
2227 } else {
b31e1552
ES		 2228 ext4_msg(sb, KERN_DEBUG,
2229 "%s: deleting unreferenced inode %lu",
46e665e9 2230 __func__, inode->i_ino);
ac27a0ec
DK		 2231 jbd_debug(2, "deleting unreferenced inode %lu\n",
2232 inode->i_ino);
2233 nr_orphans++;
2234 }
2235 iput(inode); /* The delete magic happens here! */
2236 }
2237
2b2d6d01 2238#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2239
2240 if (nr_orphans)
b31e1552
ES		 2241 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2242 PLURAL(nr_orphans));
ac27a0ec 2243 if (nr_truncates)
b31e1552
ES		 2244 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2245 PLURAL(nr_truncates));
ac27a0ec
DK		 2246#ifdef CONFIG_QUOTA
2247 /* Turn quotas off */
2248 for (i = 0; i < MAXQUOTAS; i++) {
2249 if (sb_dqopt(sb)->files[i])
287a8095 2250 dquot_quota_off(sb, i);
ac27a0ec
DK		 2251 }
2252#endif
2253 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2254}
0b8e58a1 2255
cd2291a4
ES		 2256/*
2257 * Maximal extent format file size.
2258 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2259 * extent format containers, within a sector_t, and within i_blocks
2260 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2261 * so that won't be a limiting factor.
2262 *
f17722f9
LC		 2263 * However there is other limiting factor. We do store extents in the form
2264 * of starting block and length, hence the resulting length of the extent
2265 * covering maximum file size must fit into on-disk format containers as
2266 * well. Given that length is always by 1 unit bigger than max unit (because
2267 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2268 *
cd2291a4
ES		 2269 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2270 */
f287a1a5 2271static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2272{
2273 loff_t res;
2274 loff_t upper_limit = MAX_LFS_FILESIZE;
2275
2276 /* small i_blocks in vfs inode? */
f287a1a5 2277 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2278 /*
90c699a9 2279 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2280 * i_block represent total blocks in 512 bytes
2281 * 32 == size of vfs inode i_blocks * 8
2282 */
2283 upper_limit = (1LL << 32) - 1;
2284
2285 /* total blocks in file system block size */
2286 upper_limit >>= (blkbits - 9);
2287 upper_limit <<= blkbits;
2288 }
2289
f17722f9
LC		 2290 /*
2291 * 32-bit extent-start container, ee_block. We lower the maxbytes
2292 * by one fs block, so ee_len can cover the extent of maximum file
2293 * size
2294 */
2295 res = (1LL << 32) - 1;
cd2291a4 2296 res <<= blkbits;
cd2291a4
ES		 2297
2298 /* Sanity check against vm- & vfs- imposed limits */
2299 if (res > upper_limit)
2300 res = upper_limit;
2301
2302 return res;
2303}
ac27a0ec 2304
ac27a0ec 2305/*
cd2291a4 2306 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2307 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2308 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2309 */
f287a1a5 2310static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2311{
617ba13b 2312 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2313 int meta_blocks;
2314 loff_t upper_limit;
0b8e58a1
AD		 2315 /* This is calculated to be the largest file size for a dense, block
2316 * mapped file such that the file's total number of 512-byte sectors,
2317 * including data and all indirect blocks, does not exceed (2^48 - 1).
2318 *
2319 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2320 * number of 512-byte sectors of the file.
0fc1b451
AK		 2321 */
2322
f287a1a5 2323 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2324 /*
90c699a9 2325 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2326 * the inode i_block field represents total file blocks in
2327 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2328 */
2329 upper_limit = (1LL << 32) - 1;
2330
2331 /* total blocks in file system block size */
2332 upper_limit >>= (bits - 9);
2333
2334 } else {
8180a562
AK		 2335 /*
2336 * We use 48 bit ext4_inode i_blocks
2337 * With EXT4_HUGE_FILE_FL set the i_blocks
2338 * represent total number of blocks in
2339 * file system block size
2340 */
0fc1b451
AK		 2341 upper_limit = (1LL << 48) - 1;
2342
0fc1b451
AK		 2343 }
2344
2345 /* indirect blocks */
2346 meta_blocks = 1;
2347 /* double indirect blocks */
2348 meta_blocks += 1 + (1LL << (bits-2));
2349 /* tripple indirect blocks */
2350 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2351
2352 upper_limit -= meta_blocks;
2353 upper_limit <<= bits;
ac27a0ec
DK		 2354
2355 res += 1LL << (bits-2);
2356 res += 1LL << (2*(bits-2));
2357 res += 1LL << (3*(bits-2));
2358 res <<= bits;
2359 if (res > upper_limit)
2360 res = upper_limit;
0fc1b451
AK		 2361
2362 if (res > MAX_LFS_FILESIZE)
2363 res = MAX_LFS_FILESIZE;
2364
ac27a0ec
DK		 2365 return res;
2366}
2367
617ba13b 2368static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2369 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2370{
617ba13b 2371 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2372 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2373 int has_super = 0;
2374
2375 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2376
617ba13b 2377 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2378 nr < first_meta_bg)
70bbb3e0 2379 return logical_sb_block + nr + 1;
ac27a0ec 2380 bg = sbi->s_desc_per_block * nr;
617ba13b 2381 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2382 has_super = 1;
0b8e58a1 2383
617ba13b 2384 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2385}
2386
c9de560d
AT		 2387/**
2388 * ext4_get_stripe_size: Get the stripe size.
2389 * @sbi: In memory super block info
2390 *
2391 * If we have specified it via mount option, then
2392 * use the mount option value. If the value specified at mount time is
2393 * greater than the blocks per group use the super block value.
2394 * If the super block value is greater than blocks per group return 0.
2395 * Allocator needs it be less than blocks per group.
2396 *
2397 */
2398static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2399{
2400 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2401 unsigned long stripe_width =
2402 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2403 int ret;
c9de560d
AT		 2404
2405 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2406 ret = sbi->s_stripe;
2407 else if (stripe_width <= sbi->s_blocks_per_group)
2408 ret = stripe_width;
2409 else if (stride <= sbi->s_blocks_per_group)
2410 ret = stride;
2411 else
2412 ret = 0;
c9de560d 2413
3eb08658
DE		 2414 /*
2415 * If the stripe width is 1, this makes no sense and
2416 * we set it to 0 to turn off stripe handling code.
2417 */
2418 if (ret <= 1)
2419 ret = 0;
c9de560d 2420
3eb08658 2421 return ret;
c9de560d 2422}
ac27a0ec 2423
3197ebdb
TT		 2424/* sysfs supprt */
2425
2426struct ext4_attr {
2427 struct attribute attr;
2428 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2429 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2430 const char *, size_t);
2431 int offset;
2432};
2433
2434static int parse_strtoul(const char *buf,
2435 unsigned long max, unsigned long *value)
2436{
2437 char *endp;
2438
e7d2860b
AGR		 2439 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2440 endp = skip_spaces(endp);
3197ebdb
TT		 2441 if (*endp || *value > max)
2442 return -EINVAL;
2443
2444 return 0;
2445}
2446
2447static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2448 struct ext4_sb_info *sbi,
2449 char *buf)
2450{
2451 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2452 (s64) EXT4_C2B(sbi,
2453 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2454}
2455
2456static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2457 struct ext4_sb_info *sbi, char *buf)
2458{
2459 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2460
f613dfcb
TT		 2461 if (!sb->s_bdev->bd_part)
2462 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2463 return snprintf(buf, PAGE_SIZE, "%lu\n",
2464 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2465 sbi->s_sectors_written_start) >> 1);
2466}
2467
2468static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2469 struct ext4_sb_info *sbi, char *buf)
2470{
2471 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2472
f613dfcb
TT		 2473 if (!sb->s_bdev->bd_part)
2474 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2475 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2476 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2477 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2478 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2479}
2480
2481static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2482 struct ext4_sb_info *sbi,
2483 const char *buf, size_t count)
2484{
2485 unsigned long t;
2486
2487 if (parse_strtoul(buf, 0x40000000, &t))
2488 return -EINVAL;
2489
5dbd571d 2490 if (t && !is_power_of_2(t))
3197ebdb
TT		 2491 return -EINVAL;
2492
2493 sbi->s_inode_readahead_blks = t;
2494 return count;
2495}
2496
2497static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2498 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2499{
2500 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2501
2502 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2503}
2504
2505static ssize_t sbi_ui_store(struct ext4_attr *a,
2506 struct ext4_sb_info *sbi,
2507 const char *buf, size_t count)
2508{
2509 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2510 unsigned long t;
2511
2512 if (parse_strtoul(buf, 0xffffffff, &t))
2513 return -EINVAL;
2514 *ui = t;
2515 return count;
2516}
2517
2c0544b2
TT		 2518static ssize_t trigger_test_error(struct ext4_attr *a,
2519 struct ext4_sb_info *sbi,
2520 const char *buf, size_t count)
2521{
2522 int len = count;
2523
2524 if (!capable(CAP_SYS_ADMIN))
2525 return -EPERM;
2526
2527 if (len && buf[len-1] == '\n')
2528 len--;
2529
2530 if (len)
2531 ext4_error(sbi->s_sb, "%.*s", len, buf);
2532 return count;
2533}
2534
3197ebdb
TT		 2535#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2536static struct ext4_attr ext4_attr_##_name = { \
2537 .attr = {.name = __stringify(_name), .mode = _mode }, \
2538 .show = _show, \
2539 .store = _store, \
2540 .offset = offsetof(struct ext4_sb_info, _elname), \
2541}
2542#define EXT4_ATTR(name, mode, show, store) \
2543static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2544
857ac889 2545#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2546#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2547#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2548#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2549 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2550#define ATTR_LIST(name) &ext4_attr_##name.attr
2551
2552EXT4_RO_ATTR(delayed_allocation_blocks);
2553EXT4_RO_ATTR(session_write_kbytes);
2554EXT4_RO_ATTR(lifetime_write_kbytes);
2555EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2556 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2557EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2558EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2559EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2560EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2561EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2562EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2563EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2564EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
67a5da56 2565EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2566EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2567
2568static struct attribute *ext4_attrs[] = {
2569 ATTR_LIST(delayed_allocation_blocks),
2570 ATTR_LIST(session_write_kbytes),
2571 ATTR_LIST(lifetime_write_kbytes),
2572 ATTR_LIST(inode_readahead_blks),
11013911 2573 ATTR_LIST(inode_goal),
3197ebdb
TT		 2574 ATTR_LIST(mb_stats),
2575 ATTR_LIST(mb_max_to_scan),
2576 ATTR_LIST(mb_min_to_scan),
2577 ATTR_LIST(mb_order2_req),
2578 ATTR_LIST(mb_stream_req),
2579 ATTR_LIST(mb_group_prealloc),
55138e0b 2580 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2581 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2582 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2583 NULL,
2584};
2585
857ac889
LC		 2586/* Features this copy of ext4 supports */
2587EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2588EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2589EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2590
2591static struct attribute *ext4_feat_attrs[] = {
2592 ATTR_LIST(lazy_itable_init),
27ee40df 2593 ATTR_LIST(batched_discard),
5e7bbef1 2594 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2595 NULL,
2596};
2597
3197ebdb
TT		 2598static ssize_t ext4_attr_show(struct kobject *kobj,
2599 struct attribute *attr, char *buf)
2600{
2601 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2602 s_kobj);
2603 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2604
2605 return a->show ? a->show(a, sbi, buf) : 0;
2606}
2607
2608static ssize_t ext4_attr_store(struct kobject *kobj,
2609 struct attribute *attr,
2610 const char *buf, size_t len)
2611{
2612 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2613 s_kobj);
2614 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2615
2616 return a->store ? a->store(a, sbi, buf, len) : 0;
2617}
2618
2619static void ext4_sb_release(struct kobject *kobj)
2620{
2621 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2622 s_kobj);
2623 complete(&sbi->s_kobj_unregister);
2624}
2625
52cf25d0 2626static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2627 .show = ext4_attr_show,
2628 .store = ext4_attr_store,
2629};
2630
2631static struct kobj_type ext4_ktype = {
2632 .default_attrs = ext4_attrs,
2633 .sysfs_ops = &ext4_attr_ops,
2634 .release = ext4_sb_release,
2635};
2636
857ac889
LC		 2637static void ext4_feat_release(struct kobject *kobj)
2638{
2639 complete(&ext4_feat->f_kobj_unregister);
2640}
2641
2642static struct kobj_type ext4_feat_ktype = {
2643 .default_attrs = ext4_feat_attrs,
2644 .sysfs_ops = &ext4_attr_ops,
2645 .release = ext4_feat_release,
2646};
2647
a13fb1a4
ES		 2648/*
2649 * Check whether this filesystem can be mounted based on
2650 * the features present and the RDONLY/RDWR mount requested.
2651 * Returns 1 if this filesystem can be mounted as requested,
2652 * 0 if it cannot be.
2653 */
2654static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2655{
2656 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2657 ext4_msg(sb, KERN_ERR,
2658 "Couldn't mount because of "
2659 "unsupported optional features (%x)",
2660 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2661 ~EXT4_FEATURE_INCOMPAT_SUPP));
2662 return 0;
2663 }
2664
2665 if (readonly)
2666 return 1;
2667
2668 /* Check that feature set is OK for a read-write mount */
2669 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2670 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2671 "unsupported optional features (%x)",
2672 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2673 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2674 return 0;
2675 }
2676 /*
2677 * Large file size enabled file system can only be mounted
2678 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2679 */
2680 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2681 if (sizeof(blkcnt_t) < sizeof(u64)) {
2682 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2683 "cannot be mounted RDWR without "
2684 "CONFIG_LBDAF");
2685 return 0;
2686 }
2687 }
bab08ab9
TT		 2688 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2689 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2690 ext4_msg(sb, KERN_ERR,
2691 "Can't support bigalloc feature without "
2692 "extents feature\n");
2693 return 0;
2694 }
7c319d32
AK		 2695
2696#ifndef CONFIG_QUOTA
2697 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2698 !readonly) {
2699 ext4_msg(sb, KERN_ERR,
2700 "Filesystem with quota feature cannot be mounted RDWR "
2701 "without CONFIG_QUOTA");
2702 return 0;
2703 }
2704#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2705 return 1;
2706}
2707
66e61a9e
TT		 2708/*
2709 * This function is called once a day if we have errors logged
2710 * on the file system
2711 */
2712static void print_daily_error_info(unsigned long arg)
2713{
2714 struct super_block *sb = (struct super_block *) arg;
2715 struct ext4_sb_info *sbi;
2716 struct ext4_super_block *es;
2717
2718 sbi = EXT4_SB(sb);
2719 es = sbi->s_es;
2720
2721 if (es->s_error_count)
2722 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2723 le32_to_cpu(es->s_error_count));
2724 if (es->s_first_error_time) {
2725 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2726 sb->s_id, le32_to_cpu(es->s_first_error_time),
2727 (int) sizeof(es->s_first_error_func),
2728 es->s_first_error_func,
2729 le32_to_cpu(es->s_first_error_line));
2730 if (es->s_first_error_ino)
2731 printk(": inode %u",
2732 le32_to_cpu(es->s_first_error_ino));
2733 if (es->s_first_error_block)
2734 printk(": block %llu", (unsigned long long)
2735 le64_to_cpu(es->s_first_error_block));
2736 printk("\n");
2737 }
2738 if (es->s_last_error_time) {
2739 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2740 sb->s_id, le32_to_cpu(es->s_last_error_time),
2741 (int) sizeof(es->s_last_error_func),
2742 es->s_last_error_func,
2743 le32_to_cpu(es->s_last_error_line));
2744 if (es->s_last_error_ino)
2745 printk(": inode %u",
2746 le32_to_cpu(es->s_last_error_ino));
2747 if (es->s_last_error_block)
2748 printk(": block %llu", (unsigned long long)
2749 le64_to_cpu(es->s_last_error_block));
2750 printk("\n");
2751 }
2752 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2753}
2754
bfff6873
LC		 2755/* Find next suitable group and run ext4_init_inode_table */
2756static int ext4_run_li_request(struct ext4_li_request *elr)
2757{
2758 struct ext4_group_desc *gdp = NULL;
2759 ext4_group_t group, ngroups;
2760 struct super_block *sb;
2761 unsigned long timeout = 0;
2762 int ret = 0;
2763
2764 sb = elr->lr_super;
2765 ngroups = EXT4_SB(sb)->s_groups_count;
2766
8e8ad8a5 2767 sb_start_write(sb);
bfff6873
LC		 2768 for (group = elr->lr_next_group; group < ngroups; group++) {
2769 gdp = ext4_get_group_desc(sb, group, NULL);
2770 if (!gdp) {
2771 ret = 1;
2772 break;
2773 }
2774
2775 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2776 break;
2777 }
2778
2779 if (group == ngroups)
2780 ret = 1;
2781
2782 if (!ret) {
2783 timeout = jiffies;
2784 ret = ext4_init_inode_table(sb, group,
2785 elr->lr_timeout ? 0 : 1);
2786 if (elr->lr_timeout == 0) {
51ce6511
LC		 2787 timeout = (jiffies - timeout) *
2788 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2789 elr->lr_timeout = timeout;
2790 }
2791 elr->lr_next_sched = jiffies + elr->lr_timeout;
2792 elr->lr_next_group = group + 1;
2793 }
8e8ad8a5 2794 sb_end_write(sb);
bfff6873
LC		 2795
2796 return ret;
2797}
2798
2799/*
2800 * Remove lr_request from the list_request and free the
4ed5c033 2801 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2802 */
2803static void ext4_remove_li_request(struct ext4_li_request *elr)
2804{
2805 struct ext4_sb_info *sbi;
2806
2807 if (!elr)
2808 return;
2809
2810 sbi = elr->lr_sbi;
2811
2812 list_del(&elr->lr_request);
2813 sbi->s_li_request = NULL;
2814 kfree(elr);
2815}
2816
2817static void ext4_unregister_li_request(struct super_block *sb)
2818{
1bb933fb
LC		 2819 mutex_lock(&ext4_li_mtx);
2820 if (!ext4_li_info) {
2821 mutex_unlock(&ext4_li_mtx);
bfff6873 2822 return;
1bb933fb 2823 }
bfff6873
LC		 2824
2825 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2826 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2827 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2828 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2829}
2830
8f1f7453
ES		 2831static struct task_struct *ext4_lazyinit_task;
2832
bfff6873
LC		 2833/*
2834 * This is the function where ext4lazyinit thread lives. It walks
2835 * through the request list searching for next scheduled filesystem.
2836 * When such a fs is found, run the lazy initialization request
2837 * (ext4_rn_li_request) and keep track of the time spend in this
2838 * function. Based on that time we compute next schedule time of
2839 * the request. When walking through the list is complete, compute
2840 * next waking time and put itself into sleep.
2841 */
2842static int ext4_lazyinit_thread(void *arg)
2843{
2844 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2845 struct list_head *pos, *n;
2846 struct ext4_li_request *elr;
4ed5c033 2847 unsigned long next_wakeup, cur;
bfff6873
LC		 2848
2849 BUG_ON(NULL == eli);
2850
bfff6873
LC		 2851cont_thread:
2852 while (true) {
2853 next_wakeup = MAX_JIFFY_OFFSET;
2854
2855 mutex_lock(&eli->li_list_mtx);
2856 if (list_empty(&eli->li_request_list)) {
2857 mutex_unlock(&eli->li_list_mtx);
2858 goto exit_thread;
2859 }
2860
2861 list_for_each_safe(pos, n, &eli->li_request_list) {
2862 elr = list_entry(pos, struct ext4_li_request,
2863 lr_request);
2864
b2c78cd0
TT		 2865 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2866 if (ext4_run_li_request(elr) != 0) {
2867 /* error, remove the lazy_init job */
2868 ext4_remove_li_request(elr);
2869 continue;
2870 }
bfff6873
LC		 2871 }
2872
2873 if (time_before(elr->lr_next_sched, next_wakeup))
2874 next_wakeup = elr->lr_next_sched;
2875 }
2876 mutex_unlock(&eli->li_list_mtx);
2877
a0acae0e 2878 try_to_freeze();
bfff6873 2879
4ed5c033
LC		 2880 cur = jiffies;
2881 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2882 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2883 cond_resched();
2884 continue;
2885 }
2886
4ed5c033
LC		 2887 schedule_timeout_interruptible(next_wakeup - cur);
2888
8f1f7453
ES		 2889 if (kthread_should_stop()) {
2890 ext4_clear_request_list();
2891 goto exit_thread;
2892 }
bfff6873
LC		 2893 }
2894
2895exit_thread:
2896 /*
2897 * It looks like the request list is empty, but we need
2898 * to check it under the li_list_mtx lock, to prevent any
2899 * additions into it, and of course we should lock ext4_li_mtx
2900 * to atomically free the list and ext4_li_info, because at
2901 * this point another ext4 filesystem could be registering
2902 * new one.
2903 */
2904 mutex_lock(&ext4_li_mtx);
2905 mutex_lock(&eli->li_list_mtx);
2906 if (!list_empty(&eli->li_request_list)) {
2907 mutex_unlock(&eli->li_list_mtx);
2908 mutex_unlock(&ext4_li_mtx);
2909 goto cont_thread;
2910 }
2911 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2912 kfree(ext4_li_info);
2913 ext4_li_info = NULL;
2914 mutex_unlock(&ext4_li_mtx);
2915
2916 return 0;
2917}
2918
2919static void ext4_clear_request_list(void)
2920{
2921 struct list_head *pos, *n;
2922 struct ext4_li_request *elr;
2923
2924 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2925 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2926 elr = list_entry(pos, struct ext4_li_request,
2927 lr_request);
2928 ext4_remove_li_request(elr);
2929 }
2930 mutex_unlock(&ext4_li_info->li_list_mtx);
2931}
2932
2933static int ext4_run_lazyinit_thread(void)
2934{
8f1f7453
ES		 2935 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2936 ext4_li_info, "ext4lazyinit");
2937 if (IS_ERR(ext4_lazyinit_task)) {
2938 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2939 ext4_clear_request_list();
bfff6873
LC		 2940 kfree(ext4_li_info);
2941 ext4_li_info = NULL;
92b97816 2942 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2943 "initialization thread\n",
2944 err);
2945 return err;
2946 }
2947 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2948 return 0;
2949}
2950
2951/*
2952 * Check whether it make sense to run itable init. thread or not.
2953 * If there is at least one uninitialized inode table, return
2954 * corresponding group number, else the loop goes through all
2955 * groups and return total number of groups.
2956 */
2957static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2958{
2959 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2960 struct ext4_group_desc *gdp = NULL;
2961
2962 for (group = 0; group < ngroups; group++) {
2963 gdp = ext4_get_group_desc(sb, group, NULL);
2964 if (!gdp)
2965 continue;
2966
2967 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2968 break;
2969 }
2970
2971 return group;
2972}
2973
2974static int ext4_li_info_new(void)
2975{
2976 struct ext4_lazy_init *eli = NULL;
2977
2978 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2979 if (!eli)
2980 return -ENOMEM;
2981
bfff6873
LC		 2982 INIT_LIST_HEAD(&eli->li_request_list);
2983 mutex_init(&eli->li_list_mtx);
2984
bfff6873
LC		 2985 eli->li_state |= EXT4_LAZYINIT_QUIT;
2986
2987 ext4_li_info = eli;
2988
2989 return 0;
2990}
2991
2992static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2993 ext4_group_t start)
2994{
2995 struct ext4_sb_info *sbi = EXT4_SB(sb);
2996 struct ext4_li_request *elr;
2997 unsigned long rnd;
2998
2999 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3000 if (!elr)
3001 return NULL;
3002
3003 elr->lr_super = sb;
3004 elr->lr_sbi = sbi;
3005 elr->lr_next_group = start;
3006
3007 /*
3008 * Randomize first schedule time of the request to
3009 * spread the inode table initialization requests
3010 * better.
3011 */
3012 get_random_bytes(&rnd, sizeof(rnd));
3013 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3014 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3015
3016 return elr;
3017}
3018
3019static int ext4_register_li_request(struct super_block *sb,
3020 ext4_group_t first_not_zeroed)
3021{
3022 struct ext4_sb_info *sbi = EXT4_SB(sb);
3023 struct ext4_li_request *elr;
3024 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3025 int ret = 0;
bfff6873 3026
51ce6511
LC		 3027 if (sbi->s_li_request != NULL) {
3028 /*
3029 * Reset timeout so it can be computed again, because
3030 * s_li_wait_mult might have changed.
3031 */
3032 sbi->s_li_request->lr_timeout = 0;
beed5ecb 3033 return 0;
51ce6511 3034 }
bfff6873
LC		 3035
3036 if (first_not_zeroed == ngroups ||
3037 (sb->s_flags & MS_RDONLY) ||
55ff3840 3038 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 3039 return 0;
bfff6873
LC		 3040
3041 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 3042 if (!elr)
3043 return -ENOMEM;
bfff6873
LC		 3044
3045 mutex_lock(&ext4_li_mtx);
3046
3047 if (NULL == ext4_li_info) {
3048 ret = ext4_li_info_new();
3049 if (ret)
3050 goto out;
3051 }
3052
3053 mutex_lock(&ext4_li_info->li_list_mtx);
3054 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3055 mutex_unlock(&ext4_li_info->li_list_mtx);
3056
3057 sbi->s_li_request = elr;
46e4690b
TM		 3058 /*
3059 * set elr to NULL here since it has been inserted to
3060 * the request_list and the removal and free of it is
3061 * handled by ext4_clear_request_list from now on.
3062 */
3063 elr = NULL;
bfff6873
LC		 3064
3065 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3066 ret = ext4_run_lazyinit_thread();
3067 if (ret)
3068 goto out;
3069 }
bfff6873 3070out:
beed5ecb
NK		 3071 mutex_unlock(&ext4_li_mtx);
3072 if (ret)
bfff6873 3073 kfree(elr);
bfff6873
LC		 3074 return ret;
3075}
3076
3077/*
3078 * We do not need to lock anything since this is called on
3079 * module unload.
3080 */
3081static void ext4_destroy_lazyinit_thread(void)
3082{
3083 /*
3084 * If thread exited earlier
3085 * there's nothing to be done.
3086 */
8f1f7453 3087 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3088 return;
3089
8f1f7453 3090 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3091}
3092
25ed6e8a
DW		 3093static int set_journal_csum_feature_set(struct super_block *sb)
3094{
3095 int ret = 1;
3096 int compat, incompat;
3097 struct ext4_sb_info *sbi = EXT4_SB(sb);
3098
3099 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3100 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3101 /* journal checksum v2 */
3102 compat = 0;
3103 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3104 } else {
3105 /* journal checksum v1 */
3106 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3107 incompat = 0;
3108 }
3109
3110 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3111 ret = jbd2_journal_set_features(sbi->s_journal,
3112 compat, 0,
3113 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3114 incompat);
3115 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3116 ret = jbd2_journal_set_features(sbi->s_journal,
3117 compat, 0,
3118 incompat);
3119 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3120 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3121 } else {
3122 jbd2_journal_clear_features(sbi->s_journal,
3123 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3124 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3125 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3126 }
3127
3128 return ret;
3129}
3130
952fc18e
TT		 3131/*
3132 * Note: calculating the overhead so we can be compatible with
3133 * historical BSD practice is quite difficult in the face of
3134 * clusters/bigalloc. This is because multiple metadata blocks from
3135 * different block group can end up in the same allocation cluster.
3136 * Calculating the exact overhead in the face of clustered allocation
3137 * requires either O(all block bitmaps) in memory or O(number of block
3138 * groups**2) in time. We will still calculate the superblock for
3139 * older file systems --- and if we come across with a bigalloc file
3140 * system with zero in s_overhead_clusters the estimate will be close to
3141 * correct especially for very large cluster sizes --- but for newer
3142 * file systems, it's better to calculate this figure once at mkfs
3143 * time, and store it in the superblock. If the superblock value is
3144 * present (even for non-bigalloc file systems), we will use it.
3145 */
3146static int count_overhead(struct super_block *sb, ext4_group_t grp,
3147 char *buf)
3148{
3149 struct ext4_sb_info *sbi = EXT4_SB(sb);
3150 struct ext4_group_desc *gdp;
3151 ext4_fsblk_t first_block, last_block, b;
3152 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3153 int s, j, count = 0;
3154
0548bbb8
TT		 3155 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3156 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3157 sbi->s_itb_per_group + 2);
3158
952fc18e
TT		 3159 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3160 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3161 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3162 for (i = 0; i < ngroups; i++) {
3163 gdp = ext4_get_group_desc(sb, i, NULL);
3164 b = ext4_block_bitmap(sb, gdp);
3165 if (b >= first_block && b <= last_block) {
3166 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3167 count++;
3168 }
3169 b = ext4_inode_bitmap(sb, gdp);
3170 if (b >= first_block && b <= last_block) {
3171 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3172 count++;
3173 }
3174 b = ext4_inode_table(sb, gdp);
3175 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3176 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3177 int c = EXT4_B2C(sbi, b - first_block);
3178 ext4_set_bit(c, buf);
3179 count++;
3180 }
3181 if (i != grp)
3182 continue;
3183 s = 0;
3184 if (ext4_bg_has_super(sb, grp)) {
3185 ext4_set_bit(s++, buf);
3186 count++;
3187 }
3188 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3189 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3190 count++;
3191 }
3192 }
3193 if (!count)
3194 return 0;
3195 return EXT4_CLUSTERS_PER_GROUP(sb) -
3196 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3197}
3198
3199/*
3200 * Compute the overhead and stash it in sbi->s_overhead
3201 */
3202int ext4_calculate_overhead(struct super_block *sb)
3203{
3204 struct ext4_sb_info *sbi = EXT4_SB(sb);
3205 struct ext4_super_block *es = sbi->s_es;
3206 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3207 ext4_fsblk_t overhead = 0;
3208 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3209
3210 memset(buf, 0, PAGE_SIZE);
3211 if (!buf)
3212 return -ENOMEM;
3213
3214 /*
3215 * Compute the overhead (FS structures). This is constant
3216 * for a given filesystem unless the number of block groups
3217 * changes so we cache the previous value until it does.
3218 */
3219
3220 /*
3221 * All of the blocks before first_data_block are overhead
3222 */
3223 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3224
3225 /*
3226 * Add the overhead found in each block group
3227 */
3228 for (i = 0; i < ngroups; i++) {
3229 int blks;
3230
3231 blks = count_overhead(sb, i, buf);
3232 overhead += blks;
3233 if (blks)
3234 memset(buf, 0, PAGE_SIZE);
3235 cond_resched();
3236 }
3237 sbi->s_overhead = overhead;
3238 smp_wmb();
3239 free_page((unsigned long) buf);
3240 return 0;
3241}
3242
2b2d6d01 3243static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3244{
d4c402d9 3245 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3246 struct buffer_head *bh;
617ba13b
MC		 3247 struct ext4_super_block *es = NULL;
3248 struct ext4_sb_info *sbi;
3249 ext4_fsblk_t block;
3250 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3251 ext4_fsblk_t logical_sb_block;
ac27a0ec 3252 unsigned long offset = 0;
ac27a0ec
DK		 3253 unsigned long journal_devnum = 0;
3254 unsigned long def_mount_opts;
3255 struct inode *root;
9f6200bb 3256 char *cp;
0390131b 3257 const char *descr;
dcc7dae3 3258 int ret = -ENOMEM;
281b5995 3259 int blocksize, clustersize;
4ec11028
TT		 3260 unsigned int db_count;
3261 unsigned int i;
281b5995 3262 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3263 __u64 blocks_count;
833f4077 3264 int err;
b3881f74 3265 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3266 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3267
3268 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3269 if (!sbi)
dcc7dae3 3270 goto out_free_orig;
705895b6
PE		 3271
3272 sbi->s_blockgroup_lock =
3273 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3274 if (!sbi->s_blockgroup_lock) {
3275 kfree(sbi);
dcc7dae3 3276 goto out_free_orig;
705895b6 3277 }
ac27a0ec 3278 sb->s_fs_info = sbi;
2c0544b2 3279 sbi->s_sb = sb;
ac27a0ec 3280 sbi->s_mount_opt = 0;
08cefc7a
EB		 3281 sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3282 sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
240799cd 3283 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3284 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3285 if (sb->s_bdev->bd_part)
3286 sbi->s_sectors_written_start =
3287 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3288
9f6200bb
TT		 3289 /* Cleanup superblock name */
3290 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3291 *cp = '!';
3292
dcc7dae3 3293 ret = -EINVAL;
617ba13b 3294 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3295 if (!blocksize) {
b31e1552 3296 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3297 goto out_fail;
3298 }
3299
3300 /*
617ba13b 3301 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3302 * block sizes. We need to calculate the offset from buffer start.
3303 */
617ba13b 3304 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3305 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3306 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3307 } else {
70bbb3e0 3308 logical_sb_block = sb_block;
ac27a0ec
DK		 3309 }
3310
70bbb3e0 3311 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3312 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3313 goto out_fail;
3314 }
3315 /*
3316 * Note: s_es must be initialized as soon as possible because
617ba13b 3317 * some ext4 macro-instructions depend on its value
ac27a0ec 3318 */
2716b802 3319 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3320 sbi->s_es = es;
3321 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3322 if (sb->s_magic != EXT4_SUPER_MAGIC)
3323 goto cantfind_ext4;
afc32f7e 3324 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3325
feb0ab32
DW		 3326 /* Warn if metadata_csum and gdt_csum are both set. */
3327 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3328 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3329 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3330 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3331 "redundant flags; please run fsck.");
3332
d25425f8
DW		 3333 /* Check for a known checksum algorithm */
3334 if (!ext4_verify_csum_type(sb, es)) {
3335 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3336 "unknown checksum algorithm.");
3337 silent = 1;
3338 goto cantfind_ext4;
3339 }
3340
0441984a
DW		 3341 /* Load the checksum driver */
3342 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3343 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3344 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3345 if (IS_ERR(sbi->s_chksum_driver)) {
3346 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3347 ret = PTR_ERR(sbi->s_chksum_driver);
3348 sbi->s_chksum_driver = NULL;
3349 goto failed_mount;
3350 }
3351 }
3352
a9c47317
DW		 3353 /* Check superblock checksum */
3354 if (!ext4_superblock_csum_verify(sb, es)) {
3355 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3356 "invalid superblock checksum. Run e2fsck?");
3357 silent = 1;
3358 goto cantfind_ext4;
3359 }
3360
3361 /* Precompute checksum seed for all metadata */
3362 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3363 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3364 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3365 sizeof(es->s_uuid));
3366
ac27a0ec
DK		 3367 /* Set defaults before we parse the mount options */
3368 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3369 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3370 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3371 set_opt(sb, DEBUG);
87f26807 3372 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3373 set_opt(sb, GRPID);
617ba13b 3374 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3375 set_opt(sb, NO_UID32);
ea663336 3376 /* xattr user namespace & acls are now defaulted on */
03010a33 3377#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3378 set_opt(sb, XATTR_USER);
2e7842b8 3379#endif
03010a33 3380#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3381 set_opt(sb, POSIX_ACL);
2e7842b8 3382#endif
6fd7a467 3383 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3384 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3385 set_opt(sb, JOURNAL_DATA);
617ba13b 3386 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3387 set_opt(sb, ORDERED_DATA);
617ba13b 3388 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3389 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3390
3391 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3392 set_opt(sb, ERRORS_PANIC);
bb4f397a 3393 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3394 set_opt(sb, ERRORS_CONT);
bb4f397a 3395 else
fd8c37ec 3396 set_opt(sb, ERRORS_RO);
8b67f04a 3397 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3398 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3399 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3400 set_opt(sb, DISCARD);
ac27a0ec 3401
08cefc7a
EB		 3402 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3403 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3404 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3405 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3406 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3407
8b67f04a 3408 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3409 set_opt(sb, BARRIER);
ac27a0ec 3410
dd919b98
AK		 3411 /*
3412 * enable delayed allocation by default
3413 * Use -o nodelalloc to turn it off
3414 */
bc0b75f7 3415 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3416 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3417 set_opt(sb, DELALLOC);
dd919b98 3418
51ce6511
LC		 3419 /*
3420 * set default s_li_wait_mult for lazyinit, for the case there is
3421 * no mount option specified.
3422 */
3423 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3424
8b67f04a 3425 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3426 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3427 ext4_msg(sb, KERN_WARNING,
3428 "failed to parse options in superblock: %s",
3429 sbi->s_es->s_mount_opts);
3430 }
5a916be1 3431 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3432 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3433 &journal_ioprio, 0))
ac27a0ec
DK		 3434 goto failed_mount;
3435
56889787
TT		 3436 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3437 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3438 "with data=journal disables delayed "
3439 "allocation and O_DIRECT support!\n");
3440 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3441 ext4_msg(sb, KERN_ERR, "can't mount with "
3442 "both data=journal and delalloc");
3443 goto failed_mount;
3444 }
3445 if (test_opt(sb, DIOREAD_NOLOCK)) {
3446 ext4_msg(sb, KERN_ERR, "can't mount with "
3447 "both data=journal and delalloc");
3448 goto failed_mount;
3449 }
3450 if (test_opt(sb, DELALLOC))
3451 clear_opt(sb, DELALLOC);
3452 }
3453
3454 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3455 if (test_opt(sb, DIOREAD_NOLOCK)) {
3456 if (blocksize < PAGE_SIZE) {
3457 ext4_msg(sb, KERN_ERR, "can't mount with "
3458 "dioread_nolock if block size != PAGE_SIZE");
3459 goto failed_mount;
3460 }
3461 }
3462
ac27a0ec 3463 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3464 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3465
617ba13b
MC		 3466 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3467 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3468 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3469 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3470 ext4_msg(sb, KERN_WARNING,
3471 "feature flags set on rev 0 fs, "
3472 "running e2fsck is recommended");
469108ff 3473
2035e776
TT		 3474 if (IS_EXT2_SB(sb)) {
3475 if (ext2_feature_set_ok(sb))
3476 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3477 "using the ext4 subsystem");
3478 else {
3479 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3480 "to feature incompatibilities");
3481 goto failed_mount;
3482 }
3483 }
3484
3485 if (IS_EXT3_SB(sb)) {
3486 if (ext3_feature_set_ok(sb))
3487 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3488 "using the ext4 subsystem");
3489 else {
3490 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3491 "to feature incompatibilities");
3492 goto failed_mount;
3493 }
3494 }
3495
ac27a0ec
DK		 3496 /*
3497 * Check feature flags regardless of the revision level, since we
3498 * previously didn't change the revision level when setting the flags,
3499 * so there is a chance incompat flags are set on a rev 0 filesystem.
3500 */
a13fb1a4 3501 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3502 goto failed_mount;
a13fb1a4 3503
617ba13b
MC		 3504 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3505 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3506 ext4_msg(sb, KERN_ERR,
3507 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3508 goto failed_mount;
3509 }
3510
ac27a0ec 3511 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3512 /* Validate the filesystem blocksize */
3513 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3514 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3515 blocksize);
ac27a0ec
DK		 3516 goto failed_mount;
3517 }
3518
2b2d6d01 3519 brelse(bh);
70bbb3e0
AM		 3520 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3521 offset = do_div(logical_sb_block, blocksize);
3522 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3523 if (!bh) {
b31e1552
ES		 3524 ext4_msg(sb, KERN_ERR,
3525 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3526 goto failed_mount;
3527 }
2716b802 3528 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3529 sbi->s_es = es;
617ba13b 3530 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3531 ext4_msg(sb, KERN_ERR,
3532 "Magic mismatch, very weird!");
ac27a0ec
DK		 3533 goto failed_mount;
3534 }
3535 }
3536
a13fb1a4
ES		 3537 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3538 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3539 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3540 has_huge_files);
3541 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3542
617ba13b
MC		 3543 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3544 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3545 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3546 } else {
3547 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3548 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3549 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3550 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3551 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3552 ext4_msg(sb, KERN_ERR,
3553 "unsupported inode size: %d",
2b2d6d01 3554 sbi->s_inode_size);
ac27a0ec
DK		 3555 goto failed_mount;
3556 }
ef7f3835
KS		 3557 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3558 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3559 }
0b8e58a1 3560
0d1ee42f
AR		 3561 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3562 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3563 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3564 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3565 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3566 ext4_msg(sb, KERN_ERR,
3567 "unsupported descriptor size %lu",
0d1ee42f
AR		 3568 sbi->s_desc_size);
3569 goto failed_mount;
3570 }
3571 } else
3572 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3573
ac27a0ec 3574 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3575 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3576 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3577 goto cantfind_ext4;
0b8e58a1 3578
617ba13b 3579 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3580 if (sbi->s_inodes_per_block == 0)
617ba13b 3581 goto cantfind_ext4;
ac27a0ec
DK		 3582 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3583 sbi->s_inodes_per_block;
0d1ee42f 3584 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3585 sbi->s_sbh = bh;
3586 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3587 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3588 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3589
2b2d6d01 3590 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3591 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3592 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3593 i = le32_to_cpu(es->s_flags);
3594 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3595 sbi->s_hash_unsigned = 3;
3596 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3597#ifdef __CHAR_UNSIGNED__
3598 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3599 sbi->s_hash_unsigned = 3;
3600#else
3601 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3602#endif
f99b2589 3603 }
ac27a0ec 3604
281b5995
TT		 3605 /* Handle clustersize */
3606 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3607 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3608 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3609 if (has_bigalloc) {
3610 if (clustersize < blocksize) {
3611 ext4_msg(sb, KERN_ERR,
3612 "cluster size (%d) smaller than "
3613 "block size (%d)", clustersize, blocksize);
3614 goto failed_mount;
3615 }
3616 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3617 le32_to_cpu(es->s_log_block_size);
3618 sbi->s_clusters_per_group =
3619 le32_to_cpu(es->s_clusters_per_group);
3620 if (sbi->s_clusters_per_group > blocksize * 8) {
3621 ext4_msg(sb, KERN_ERR,
3622 "#clusters per group too big: %lu",
3623 sbi->s_clusters_per_group);
3624 goto failed_mount;
3625 }
3626 if (sbi->s_blocks_per_group !=
3627 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3628 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3629 "clusters per group (%lu) inconsistent",
3630 sbi->s_blocks_per_group,
3631 sbi->s_clusters_per_group);
3632 goto failed_mount;
3633 }
3634 } else {
3635 if (clustersize != blocksize) {
3636 ext4_warning(sb, "fragment/cluster size (%d) != "
3637 "block size (%d)", clustersize,
3638 blocksize);
3639 clustersize = blocksize;
3640 }
3641 if (sbi->s_blocks_per_group > blocksize * 8) {
3642 ext4_msg(sb, KERN_ERR,
3643 "#blocks per group too big: %lu",
3644 sbi->s_blocks_per_group);
3645 goto failed_mount;
3646 }
3647 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3648 sbi->s_cluster_bits = 0;
ac27a0ec 3649 }
281b5995
TT		 3650 sbi->s_cluster_ratio = clustersize / blocksize;
3651
ac27a0ec 3652 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3653 ext4_msg(sb, KERN_ERR,
3654 "#inodes per group too big: %lu",
2b2d6d01 3655 sbi->s_inodes_per_group);
ac27a0ec
DK		 3656 goto failed_mount;
3657 }
3658
bf43d84b
ES		 3659 /*
3660 * Test whether we have more sectors than will fit in sector_t,
3661 * and whether the max offset is addressable by the page cache.
3662 */
5a9ae68a 3663 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3664 ext4_blocks_count(es));
5a9ae68a 3665 if (err) {
b31e1552 3666 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3667 " too large to mount safely on this system");
ac27a0ec 3668 if (sizeof(sector_t) < 8)
90c699a9 3669 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3670 ret = err;
ac27a0ec
DK		 3671 goto failed_mount;
3672 }
3673
617ba13b
MC		 3674 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3675 goto cantfind_ext4;
e7c95593 3676
0f2ddca6
FTN		 3677 /* check blocks count against device size */
3678 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3679 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3680 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3681 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3682 ext4_blocks_count(es), blocks_count);
3683 goto failed_mount;
3684 }
3685
0b8e58a1
AD		 3686 /*
3687 * It makes no sense for the first data block to be beyond the end
3688 * of the filesystem.
3689 */
3690 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3691 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3692 "block %u is beyond end of filesystem (%llu)",
3693 le32_to_cpu(es->s_first_data_block),
3694 ext4_blocks_count(es));
e7c95593
ES		 3695 goto failed_mount;
3696 }
bd81d8ee
LV		 3697 blocks_count = (ext4_blocks_count(es) -
3698 le32_to_cpu(es->s_first_data_block) +
3699 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3700 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3701 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3702 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3703 "(block count %llu, first data block %u, "
b31e1552 3704 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3705 ext4_blocks_count(es),
3706 le32_to_cpu(es->s_first_data_block),
3707 EXT4_BLOCKS_PER_GROUP(sb));
3708 goto failed_mount;
3709 }
bd81d8ee 3710 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3711 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3712 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3713 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3714 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3715 sbi->s_group_desc = ext4_kvmalloc(db_count *
3716 sizeof(struct buffer_head *),
3717 GFP_KERNEL);
ac27a0ec 3718 if (sbi->s_group_desc == NULL) {
b31e1552 3719 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3720 ret = -ENOMEM;
ac27a0ec
DK		 3721 goto failed_mount;
3722 }
3723
9f6200bb
TT		 3724 if (ext4_proc_root)
3725 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3726
66acdcf4
TT		 3727 if (sbi->s_proc)
3728 proc_create_data("options", S_IRUGO, sbi->s_proc,
3729 &ext4_seq_options_fops, sb);
3730
705895b6 3731 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3732
3733 for (i = 0; i < db_count; i++) {
70bbb3e0 3734 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3735 sbi->s_group_desc[i] = sb_bread(sb, block);
3736 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3737 ext4_msg(sb, KERN_ERR,
3738 "can't read group descriptor %d", i);
ac27a0ec
DK		 3739 db_count = i;
3740 goto failed_mount2;
3741 }
3742 }
bfff6873 3743 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3744 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3745 goto failed_mount2;
3746 }
772cb7c8
JS		 3747 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3748 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3749 ext4_msg(sb, KERN_ERR,
3750 "unable to initialize "
3751 "flex_bg meta info!");
772cb7c8
JS		 3752 goto failed_mount2;
3753 }
3754
ac27a0ec
DK		 3755 sbi->s_gdb_count = db_count;
3756 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3757 spin_lock_init(&sbi->s_next_gen_lock);
3758
04496411
TM		 3759 init_timer(&sbi->s_err_report);
3760 sbi->s_err_report.function = print_daily_error_info;
3761 sbi->s_err_report.data = (unsigned long) sb;
3762
57042651 3763 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3764 ext4_count_free_clusters(sb));
ce7e010a
TT		 3765 if (!err) {
3766 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3767 ext4_count_free_inodes(sb));
3768 }
3769 if (!err) {
3770 err = percpu_counter_init(&sbi->s_dirs_counter,
3771 ext4_count_dirs(sb));
3772 }
3773 if (!err) {
57042651 3774 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3775 }
3776 if (err) {
3777 ext4_msg(sb, KERN_ERR, "insufficient memory");
2cde417d 3778 ret = err;
ce7e010a
TT		 3779 goto failed_mount3;
3780 }
3781
c9de560d 3782 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3783 sbi->s_max_writeback_mb_bump = 128;
67a5da56 3784 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3785
ac27a0ec
DK		 3786 /*
3787 * set up enough so that it can read an inode
3788 */
9ca92389
TT		 3789 if (!test_opt(sb, NOLOAD) &&
3790 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3791 sb->s_op = &ext4_sops;
3792 else
3793 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3794 sb->s_export_op = &ext4_export_ops;
3795 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3796#ifdef CONFIG_QUOTA
617ba13b
MC		 3797 sb->s_qcop = &ext4_qctl_operations;
3798 sb->dq_op = &ext4_quota_operations;
7c319d32
AK		 3799
3800 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3801 /* Use qctl operations for hidden quota files. */
3802 sb->s_qcop = &ext4_qctl_sysfile_operations;
3803 }
ac27a0ec 3804#endif
f2fa2ffc
AK		 3805 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3806
ac27a0ec 3807 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3808 mutex_init(&sbi->s_orphan_lock);
8f82f840 3809 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3810
3811 sb->s_root = NULL;
3812
3813 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3814 EXT4_HAS_INCOMPAT_FEATURE(sb,
3815 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3816
c5e06d10
JL		 3817 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3818 !(sb->s_flags & MS_RDONLY))
3819 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3820 goto failed_mount3;
3821
ac27a0ec
DK		 3822 /*
3823 * The first inode we look at is the journal inode. Don't try
3824 * root first: it may be modified in the journal!
3825 */
3826 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3827 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3828 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3829 goto failed_mount3;
0390131b
FM		 3830 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3831 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3832 ext4_msg(sb, KERN_ERR, "required journal recovery "
3833 "suppressed and not mounted read-only");
744692dc 3834 goto failed_mount_wq;
ac27a0ec 3835 } else {
fd8c37ec 3836 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3837 sbi->s_journal = NULL;
3838 needs_recovery = 0;
3839 goto no_journal;
ac27a0ec
DK		 3840 }
3841
f32aaf2d 3842 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3843 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3844 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3845 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3846 goto failed_mount_wq;
eb40a09c
JS		 3847 }
3848
25ed6e8a
DW		 3849 if (!set_journal_csum_feature_set(sb)) {
3850 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3851 "feature set");
3852 goto failed_mount_wq;
d4da6c9c 3853 }
818d276c 3854
ac27a0ec
DK		 3855 /* We have now updated the journal if required, so we can
3856 * validate the data journaling mode. */
3857 switch (test_opt(sb, DATA_FLAGS)) {
3858 case 0:
3859 /* No mode set, assume a default based on the journal
63f57933
AM		 3860 * capabilities: ORDERED_DATA if the journal can
3861 * cope, else JOURNAL_DATA
3862 */
dab291af
MC		 3863 if (jbd2_journal_check_available_features
3864 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3865 set_opt(sb, ORDERED_DATA);
ac27a0ec 3866 else
fd8c37ec 3867 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3868 break;
3869
617ba13b
MC		 3870 case EXT4_MOUNT_ORDERED_DATA:
3871 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3872 if (!jbd2_journal_check_available_features
3873 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3874 ext4_msg(sb, KERN_ERR, "Journal does not support "
3875 "requested data journaling mode");
744692dc 3876 goto failed_mount_wq;
ac27a0ec
DK		 3877 }
3878 default:
3879 break;
3880 }
b3881f74 3881 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3882
18aadd47
BJ		 3883 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3884
ce7e010a
TT		 3885 /*
3886 * The journal may have updated the bg summary counts, so we
3887 * need to update the global counters.
3888 */
57042651 3889 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3890 ext4_count_free_clusters(sb));
ce7e010a
TT		 3891 percpu_counter_set(&sbi->s_freeinodes_counter,
3892 ext4_count_free_inodes(sb));
3893 percpu_counter_set(&sbi->s_dirs_counter,
3894 ext4_count_dirs(sb));
57042651 3895 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3896
ce7e010a 3897no_journal:
952fc18e
TT		 3898 /*
3899 * Get the # of file system overhead blocks from the
3900 * superblock if present.
3901 */
3902 if (es->s_overhead_clusters)
3903 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3904 else {
3905 ret = ext4_calculate_overhead(sb);
3906 if (ret)
3907 goto failed_mount_wq;
3908 }
3909
fd89d5f2
TH		 3910 /*
3911 * The maximum number of concurrent works can be high and
3912 * concurrency isn't really necessary. Limit it to 1.
3913 */
3914 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3915 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3916 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3917 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3918 goto failed_mount_wq;
3919 }
3920
ac27a0ec 3921 /*
dab291af 3922 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3923 * so we can safely mount the rest of the filesystem now.
3924 */
3925
1d1fe1ee
DH		 3926 root = ext4_iget(sb, EXT4_ROOT_INO);
3927 if (IS_ERR(root)) {
b31e1552 3928 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3929 ret = PTR_ERR(root);
32a9bb57 3930 root = NULL;
ac27a0ec
DK		 3931 goto failed_mount4;
3932 }
3933 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3934 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3935 iput(root);
ac27a0ec
DK		 3936 goto failed_mount4;
3937 }
48fde701 3938 sb->s_root = d_make_root(root);
1d1fe1ee 3939 if (!sb->s_root) {
b31e1552 3940 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3941 ret = -ENOMEM;
3942 goto failed_mount4;
3943 }
ac27a0ec 3944
7e84b621
ES		 3945 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3946 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3947
3948 /* determine the minimum size of new large inodes, if present */
3949 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3950 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3951 EXT4_GOOD_OLD_INODE_SIZE;
3952 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3953 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3954 if (sbi->s_want_extra_isize <
3955 le16_to_cpu(es->s_want_extra_isize))
3956 sbi->s_want_extra_isize =
3957 le16_to_cpu(es->s_want_extra_isize);
3958 if (sbi->s_want_extra_isize <
3959 le16_to_cpu(es->s_min_extra_isize))
3960 sbi->s_want_extra_isize =
3961 le16_to_cpu(es->s_min_extra_isize);
3962 }
3963 }
3964 /* Check if enough inode space is available */
3965 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3966 sbi->s_inode_size) {
3967 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3968 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3969 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3970 "available");
ef7f3835
KS		 3971 }
3972
6fd058f7
TT		 3973 err = ext4_setup_system_zone(sb);
3974 if (err) {
b31e1552 3975 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3976 "zone (%d)", err);
94bf608a 3977 goto failed_mount4a;
6fd058f7
TT		 3978 }
3979
c2774d84 3980 ext4_ext_init(sb);
9d99012f 3981 err = ext4_mb_init(sb);
c2774d84 3982 if (err) {
421f91d2 3983 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3984 err);
dcf2d804 3985 goto failed_mount5;
c2774d84
AK		 3986 }
3987
bfff6873
LC		 3988 err = ext4_register_li_request(sb, first_not_zeroed);
3989 if (err)
dcf2d804 3990 goto failed_mount6;
bfff6873 3991
3197ebdb
TT		 3992 sbi->s_kobj.kset = ext4_kset;
3993 init_completion(&sbi->s_kobj_unregister);
3994 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3995 "%s", sb->s_id);
dcf2d804
TM		 3996 if (err)
3997 goto failed_mount7;
3197ebdb 3998
617ba13b
MC		 3999 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4000 ext4_orphan_cleanup(sb, es);
4001 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4002 if (needs_recovery) {
b31e1552 4003 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4004 ext4_mark_recovery_complete(sb, es);
4005 }
4006 if (EXT4_SB(sb)->s_journal) {
4007 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4008 descr = " journalled data mode";
4009 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4010 descr = " ordered data mode";
4011 else
4012 descr = " writeback data mode";
4013 } else
4014 descr = "out journal";
4015
7c319d32
AK		 4016#ifdef CONFIG_QUOTA
4017 /* Enable quota usage during mount. */
4018 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4019 !(sb->s_flags & MS_RDONLY)) {
4020 ret = ext4_enable_quotas(sb);
4021 if (ret)
4022 goto failed_mount7;
4023 }
4024#endif /* CONFIG_QUOTA */
4025
d4c402d9 4026 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4027 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4028 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4029
66e61a9e
TT		 4030 if (es->s_error_count)
4031 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4032
d4c402d9 4033 kfree(orig_data);
ac27a0ec
DK		 4034 return 0;
4035
617ba13b 4036cantfind_ext4:
ac27a0ec 4037 if (!silent)
b31e1552 4038 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4039 goto failed_mount;
4040
dcf2d804
TM		 4041failed_mount7:
4042 ext4_unregister_li_request(sb);
4043failed_mount6:
dcf2d804 4044 ext4_mb_release(sb);
94bf608a
AV		 4045failed_mount5:
4046 ext4_ext_release(sb);
dcf2d804 4047 ext4_release_system_zone(sb);
94bf608a
AV		 4048failed_mount4a:
4049 dput(sb->s_root);
32a9bb57 4050 sb->s_root = NULL;
94bf608a 4051failed_mount4:
b31e1552 4052 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 4053 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4054failed_mount_wq:
0390131b
FM		 4055 if (sbi->s_journal) {
4056 jbd2_journal_destroy(sbi->s_journal);
4057 sbi->s_journal = NULL;
4058 }
ac27a0ec 4059failed_mount3:
04496411 4060 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4061 if (sbi->s_flex_groups)
4062 ext4_kvfree(sbi->s_flex_groups);
57042651 4063 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4064 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4065 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4066 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 4067 if (sbi->s_mmp_tsk)
4068 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4069failed_mount2:
4070 for (i = 0; i < db_count; i++)
4071 brelse(sbi->s_group_desc[i]);
f18a5f21 4072 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4073failed_mount:
0441984a
DW		 4074 if (sbi->s_chksum_driver)
4075 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4076 if (sbi->s_proc) {
66acdcf4 4077 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4078 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4079 }
ac27a0ec
DK		 4080#ifdef CONFIG_QUOTA
4081 for (i = 0; i < MAXQUOTAS; i++)
4082 kfree(sbi->s_qf_names[i]);
4083#endif
617ba13b 4084 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4085 brelse(bh);
4086out_fail:
4087 sb->s_fs_info = NULL;
f6830165 4088 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4089 kfree(sbi);
dcc7dae3 4090out_free_orig:
d4c402d9 4091 kfree(orig_data);
1d1fe1ee 4092 return ret;
ac27a0ec
DK		 4093}
4094
4095/*
4096 * Setup any per-fs journal parameters now. We'll do this both on
4097 * initial mount, once the journal has been initialised but before we've
4098 * done any recovery; and again on any subsequent remount.
4099 */
617ba13b 4100static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4101{
617ba13b 4102 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4103
30773840
TT		 4104 journal->j_commit_interval = sbi->s_commit_interval;
4105 journal->j_min_batch_time = sbi->s_min_batch_time;
4106 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4107
a931da6a 4108 write_lock(&journal->j_state_lock);
ac27a0ec 4109 if (test_opt(sb, BARRIER))
dab291af 4110 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4111 else
dab291af 4112 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4113 if (test_opt(sb, DATA_ERR_ABORT))
4114 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4115 else
4116 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4117 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4118}
4119
617ba13b 4120static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4121 unsigned int journal_inum)
4122{
4123 struct inode *journal_inode;
4124 journal_t *journal;
4125
0390131b
FM		 4126 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4127
ac27a0ec
DK		 4128 /* First, test for the existence of a valid inode on disk. Bad
4129 * things happen if we iget() an unused inode, as the subsequent
4130 * iput() will try to delete it. */
4131
1d1fe1ee
DH		 4132 journal_inode = ext4_iget(sb, journal_inum);
4133 if (IS_ERR(journal_inode)) {
b31e1552 4134 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4135 return NULL;
4136 }
4137 if (!journal_inode->i_nlink) {
4138 make_bad_inode(journal_inode);
4139 iput(journal_inode);
b31e1552 4140 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4141 return NULL;
4142 }
4143
e5f8eab8 4144 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4145 journal_inode, journal_inode->i_size);
1d1fe1ee 4146 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4147 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4148 iput(journal_inode);
4149 return NULL;
4150 }
4151
dab291af 4152 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4153 if (!journal) {
b31e1552 4154 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4155 iput(journal_inode);
4156 return NULL;
4157 }
4158 journal->j_private = sb;
617ba13b 4159 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4160 return journal;
4161}
4162
617ba13b 4163static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4164 dev_t j_dev)
4165{
2b2d6d01 4166 struct buffer_head *bh;
ac27a0ec 4167 journal_t *journal;
617ba13b
MC		 4168 ext4_fsblk_t start;
4169 ext4_fsblk_t len;
ac27a0ec 4170 int hblock, blocksize;
617ba13b 4171 ext4_fsblk_t sb_block;
ac27a0ec 4172 unsigned long offset;
2b2d6d01 4173 struct ext4_super_block *es;
ac27a0ec
DK		 4174 struct block_device *bdev;
4175
0390131b
FM		 4176 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4177
b31e1552 4178 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4179 if (bdev == NULL)
4180 return NULL;
4181
ac27a0ec 4182 blocksize = sb->s_blocksize;
e1defc4f 4183 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4184 if (blocksize < hblock) {
b31e1552
ES		 4185 ext4_msg(sb, KERN_ERR,
4186 "blocksize too small for journal device");
ac27a0ec
DK		 4187 goto out_bdev;
4188 }
4189
617ba13b
MC		 4190 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4191 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4192 set_blocksize(bdev, blocksize);
4193 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4194 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4195 "external journal");
ac27a0ec
DK		 4196 goto out_bdev;
4197 }
4198
2716b802 4199 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4200 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4201 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4202 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4203 ext4_msg(sb, KERN_ERR, "external journal has "
4204 "bad superblock");
ac27a0ec
DK		 4205 brelse(bh);
4206 goto out_bdev;
4207 }
4208
617ba13b 4209 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4210 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4211 brelse(bh);
4212 goto out_bdev;
4213 }
4214
bd81d8ee 4215 len = ext4_blocks_count(es);
ac27a0ec
DK		 4216 start = sb_block + 1;
4217 brelse(bh); /* we're done with the superblock */
4218
dab291af 4219 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4220 start, len, blocksize);
4221 if (!journal) {
b31e1552 4222 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4223 goto out_bdev;
4224 }
4225 journal->j_private = sb;
4226 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4227 wait_on_buffer(journal->j_sb_buffer);
4228 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4229 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4230 goto out_journal;
4231 }
4232 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4233 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4234 "user (unsupported) - %d",
ac27a0ec
DK		 4235 be32_to_cpu(journal->j_superblock->s_nr_users));
4236 goto out_journal;
4237 }
617ba13b
MC		 4238 EXT4_SB(sb)->journal_bdev = bdev;
4239 ext4_init_journal_params(sb, journal);
ac27a0ec 4240 return journal;
0b8e58a1 4241
ac27a0ec 4242out_journal:
dab291af 4243 jbd2_journal_destroy(journal);
ac27a0ec 4244out_bdev:
617ba13b 4245 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4246 return NULL;
4247}
4248
617ba13b
MC		 4249static int ext4_load_journal(struct super_block *sb,
4250 struct ext4_super_block *es,
ac27a0ec
DK		 4251 unsigned long journal_devnum)
4252{
4253 journal_t *journal;
4254 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4255 dev_t journal_dev;
4256 int err = 0;
4257 int really_read_only;
4258
0390131b
FM		 4259 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4260
ac27a0ec
DK		 4261 if (journal_devnum &&
4262 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4263 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4264 "numbers have changed");
ac27a0ec
DK		 4265 journal_dev = new_decode_dev(journal_devnum);
4266 } else
4267 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4268
4269 really_read_only = bdev_read_only(sb->s_bdev);
4270
4271 /*
4272 * Are we loading a blank journal or performing recovery after a
4273 * crash? For recovery, we need to check in advance whether we
4274 * can get read-write access to the device.
4275 */
617ba13b 4276 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4277 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4278 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4279 "required on readonly filesystem");
ac27a0ec 4280 if (really_read_only) {
b31e1552
ES		 4281 ext4_msg(sb, KERN_ERR, "write access "
4282 "unavailable, cannot proceed");
ac27a0ec
DK		 4283 return -EROFS;
4284 }
b31e1552
ES		 4285 ext4_msg(sb, KERN_INFO, "write access will "
4286 "be enabled during recovery");
ac27a0ec
DK		 4287 }
4288 }
4289
4290 if (journal_inum && journal_dev) {
b31e1552
ES		 4291 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4292 "and inode journals!");
ac27a0ec
DK		 4293 return -EINVAL;
4294 }
4295
4296 if (journal_inum) {
617ba13b 4297 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4298 return -EINVAL;
4299 } else {
617ba13b 4300 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4301 return -EINVAL;
4302 }
4303
90576c0b 4304 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4305 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4306
617ba13b 4307 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4308 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4309 if (!err) {
4310 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4311 if (save)
4312 memcpy(save, ((char *) es) +
4313 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4314 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4315 if (save)
4316 memcpy(((char *) es) + EXT4_S_ERR_START,
4317 save, EXT4_S_ERR_LEN);
4318 kfree(save);
4319 }
ac27a0ec
DK		 4320
4321 if (err) {
b31e1552 4322 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4323 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4324 return err;
4325 }
4326
617ba13b
MC		 4327 EXT4_SB(sb)->s_journal = journal;
4328 ext4_clear_journal_err(sb, es);
ac27a0ec 4329
c41303ce 4330 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4331 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4332 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4333
4334 /* Make sure we flush the recovery flag to disk. */
e2d67052 4335 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4336 }
4337
4338 return 0;
4339}
4340
e2d67052 4341static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4342{
e2d67052 4343 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4344 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4345 int error = 0;
ac27a0ec 4346
7c2e7087 4347 if (!sbh || block_device_ejected(sb))
c4be0c1d 4348 return error;
914258bf
TT		 4349 if (buffer_write_io_error(sbh)) {
4350 /*
4351 * Oh, dear. A previous attempt to write the
4352 * superblock failed. This could happen because the
4353 * USB device was yanked out. Or it could happen to
4354 * be a transient write error and maybe the block will
4355 * be remapped. Nothing we can do but to retry the
4356 * write and hope for the best.
4357 */
b31e1552
ES		 4358 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4359 "superblock detected");
914258bf
TT		 4360 clear_buffer_write_io_error(sbh);
4361 set_buffer_uptodate(sbh);
4362 }
71290b36
TT		 4363 /*
4364 * If the file system is mounted read-only, don't update the
4365 * superblock write time. This avoids updating the superblock
4366 * write time when we are mounting the root file system
4367 * read/only but we need to replay the journal; at that point,
4368 * for people who are east of GMT and who make their clock
4369 * tick in localtime for Windows bug-for-bug compatibility,
4370 * the clock is set in the future, and this will cause e2fsck
4371 * to complain and force a full file system check.
4372 */
4373 if (!(sb->s_flags & MS_RDONLY))
4374 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4375 if (sb->s_bdev->bd_part)
4376 es->s_kbytes_written =
4377 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4378 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4379 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4380 else
4381 es->s_kbytes_written =
4382 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4383 ext4_free_blocks_count_set(es,
4384 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4385 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4386 es->s_free_inodes_count =
4387 cpu_to_le32(percpu_counter_sum_positive(
4388 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4389 BUFFER_TRACE(sbh, "marking dirty");
a9c47317 4390 ext4_superblock_csum_set(sb, es);
ac27a0ec 4391 mark_buffer_dirty(sbh);
914258bf 4392 if (sync) {
c4be0c1d
TS		 4393 error = sync_dirty_buffer(sbh);
4394 if (error)
4395 return error;
4396
4397 error = buffer_write_io_error(sbh);
4398 if (error) {
b31e1552
ES		 4399 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4400 "superblock");
914258bf
TT		 4401 clear_buffer_write_io_error(sbh);
4402 set_buffer_uptodate(sbh);
4403 }
4404 }
c4be0c1d 4405 return error;
ac27a0ec
DK		 4406}
4407
ac27a0ec
DK		 4408/*
4409 * Have we just finished recovery? If so, and if we are mounting (or
4410 * remounting) the filesystem readonly, then we will end up with a
4411 * consistent fs on disk. Record that fact.
4412 */
2b2d6d01
TT		 4413static void ext4_mark_recovery_complete(struct super_block *sb,
4414 struct ext4_super_block *es)
ac27a0ec 4415{
617ba13b 4416 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4417
0390131b
FM		 4418 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4419 BUG_ON(journal != NULL);
4420 return;
4421 }
dab291af 4422 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4423 if (jbd2_journal_flush(journal) < 0)
4424 goto out;
4425
617ba13b 4426 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4427 sb->s_flags & MS_RDONLY) {
617ba13b 4428 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4429 ext4_commit_super(sb, 1);
ac27a0ec 4430 }
7ffe1ea8
HK		 4431
4432out:
dab291af 4433 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4434}
4435
4436/*
4437 * If we are mounting (or read-write remounting) a filesystem whose journal
4438 * has recorded an error from a previous lifetime, move that error to the
4439 * main filesystem now.
4440 */
2b2d6d01
TT		 4441static void ext4_clear_journal_err(struct super_block *sb,
4442 struct ext4_super_block *es)
ac27a0ec
DK		 4443{
4444 journal_t *journal;
4445 int j_errno;
4446 const char *errstr;
4447
0390131b
FM		 4448 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4449
617ba13b 4450 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4451
4452 /*
4453 * Now check for any error status which may have been recorded in the
617ba13b 4454 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4455 */
4456
dab291af 4457 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4458 if (j_errno) {
4459 char nbuf[16];
4460
617ba13b 4461 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4462 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4463 "from previous mount: %s", errstr);
12062ddd 4464 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4465
617ba13b
MC		 4466 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4467 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4468 ext4_commit_super(sb, 1);
ac27a0ec 4469
dab291af 4470 jbd2_journal_clear_err(journal);
d796c52e 4471 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4472 }
4473}
4474
4475/*
4476 * Force the running and committing transactions to commit,
4477 * and wait on the commit.
4478 */
617ba13b 4479int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4480{
4481 journal_t *journal;
0390131b 4482 int ret = 0;
ac27a0ec
DK		 4483
4484 if (sb->s_flags & MS_RDONLY)
4485 return 0;
4486
617ba13b 4487 journal = EXT4_SB(sb)->s_journal;
8e8ad8a5 4488 if (journal)
0390131b 4489 ret = ext4_journal_force_commit(journal);
0390131b 4490
ac27a0ec
DK		 4491 return ret;
4492}
4493
617ba13b 4494static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4495{
14ce0cb4 4496 int ret = 0;
9eddacf9 4497 tid_t target;
8d5d02e6 4498 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4499
9bffad1e 4500 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4501 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4502 /*
4503 * Writeback quota in non-journalled quota case - journalled quota has
4504 * no dirty dquots
4505 */
4506 dquot_writeback_dquots(sb, -1);
8d5d02e6 4507 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4508 if (wait)
8d5d02e6 4509 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4510 }
14ce0cb4 4511 return ret;
ac27a0ec
DK		 4512}
4513
4514/*
4515 * LVM calls this function before a (read-only) snapshot is created. This
4516 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4517 *
4518 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4519 * state independently. It relies on upper layer to stop all data & metadata
4520 * modifications.
ac27a0ec 4521 */
c4be0c1d 4522static int ext4_freeze(struct super_block *sb)
ac27a0ec 4523{
c4be0c1d
TS		 4524 int error = 0;
4525 journal_t *journal;
ac27a0ec 4526
9ca92389
TT		 4527 if (sb->s_flags & MS_RDONLY)
4528 return 0;
ac27a0ec 4529
9ca92389 4530 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4531
9ca92389
TT		 4532 /* Now we set up the journal barrier. */
4533 jbd2_journal_lock_updates(journal);
ac27a0ec 4534
9ca92389
TT		 4535 /*
4536 * Don't clear the needs_recovery flag if we failed to flush
4537 * the journal.
4538 */
4539 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4540 if (error < 0)
4541 goto out;
9ca92389
TT		 4542
4543 /* Journal blocked and flushed, clear needs_recovery flag. */
4544 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4545 error = ext4_commit_super(sb, 1);
6b0310fb 4546out:
8e8ad8a5 4547 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4548 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4549 return error;
ac27a0ec
DK		 4550}
4551
4552/*
4553 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4554 * flag here, even though the filesystem is not technically dirty yet.
4555 */
c4be0c1d 4556static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4557{
9ca92389
TT		 4558 if (sb->s_flags & MS_RDONLY)
4559 return 0;
4560
9ca92389
TT		 4561 /* Reset the needs_recovery flag before the fs is unlocked. */
4562 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4563 ext4_commit_super(sb, 1);
c4be0c1d 4564 return 0;
ac27a0ec
DK		 4565}
4566
673c6100
TT		 4567/*
4568 * Structure to save mount options for ext4_remount's benefit
4569 */
4570struct ext4_mount_options {
4571 unsigned long s_mount_opt;
a2595b8a 4572 unsigned long s_mount_opt2;
08cefc7a
EB		 4573 kuid_t s_resuid;
4574 kgid_t s_resgid;
673c6100
TT		 4575 unsigned long s_commit_interval;
4576 u32 s_min_batch_time, s_max_batch_time;
4577#ifdef CONFIG_QUOTA
4578 int s_jquota_fmt;
4579 char *s_qf_names[MAXQUOTAS];
4580#endif
4581};
4582
2b2d6d01 4583static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4584{
2b2d6d01 4585 struct ext4_super_block *es;
617ba13b 4586 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4587 unsigned long old_sb_flags;
617ba13b 4588 struct ext4_mount_options old_opts;
c79d967d 4589 int enable_quota = 0;
8a266467 4590 ext4_group_t g;
b3881f74 4591 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4592 int err = 0;
ac27a0ec
DK		 4593#ifdef CONFIG_QUOTA
4594 int i;
4595#endif
d4c402d9 4596 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4597
4598 /* Store the original options */
4599 old_sb_flags = sb->s_flags;
4600 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4601 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4602 old_opts.s_resuid = sbi->s_resuid;
4603 old_opts.s_resgid = sbi->s_resgid;
4604 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4605 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4606 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4607#ifdef CONFIG_QUOTA
4608 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4609 for (i = 0; i < MAXQUOTAS; i++)
4610 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4611#endif
b3881f74
TT		 4612 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4613 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4614
4615 /*
4616 * Allow the "check" option to be passed as a remount option.
4617 */
661aa520 4618 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4619 err = -EINVAL;
4620 goto restore_opts;
4621 }
4622
4ab2f15b 4623 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4624 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4625
4626 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4627 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4628
4629 es = sbi->s_es;
4630
b3881f74 4631 if (sbi->s_journal) {
0390131b 4632 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4633 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4634 }
ac27a0ec 4635
661aa520 4636 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4637 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4638 err = -EROFS;
4639 goto restore_opts;
4640 }
4641
4642 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4643 err = dquot_suspend(sb, -1);
4644 if (err < 0)
c79d967d 4645 goto restore_opts;
c79d967d 4646
ac27a0ec
DK		 4647 /*
4648 * First of all, the unconditional stuff we have to do
4649 * to disable replay of the journal when we next remount
4650 */
4651 sb->s_flags |= MS_RDONLY;
4652
4653 /*
4654 * OK, test if we are remounting a valid rw partition
4655 * readonly, and if so set the rdonly flag and then
4656 * mark the partition as valid again.
4657 */
617ba13b
MC		 4658 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4659 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4660 es->s_state = cpu_to_le16(sbi->s_mount_state);
4661
a63c9eb2 4662 if (sbi->s_journal)
0390131b 4663 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4664 } else {
a13fb1a4
ES		 4665 /* Make sure we can mount this feature set readwrite */
4666 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4667 err = -EROFS;
4668 goto restore_opts;
4669 }
8a266467
TT		 4670 /*
4671 * Make sure the group descriptor checksums
0b8e58a1 4672 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4673 */
4674 for (g = 0; g < sbi->s_groups_count; g++) {
4675 struct ext4_group_desc *gdp =
4676 ext4_get_group_desc(sb, g, NULL);
4677
feb0ab32 4678 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4679 ext4_msg(sb, KERN_ERR,
4680 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4681 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4682 le16_to_cpu(gdp->bg_checksum));
4683 err = -EINVAL;
4684 goto restore_opts;
4685 }
4686 }
4687
ead6596b
ES		 4688 /*
4689 * If we have an unprocessed orphan list hanging
4690 * around from a previously readonly bdev mount,
4691 * require a full umount/remount for now.
4692 */
4693 if (es->s_last_orphan) {
b31e1552 4694 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4695 "remount RDWR because of unprocessed "
4696 "orphan inode list. Please "
b31e1552 4697 "umount/remount instead");
ead6596b
ES		 4698 err = -EINVAL;
4699 goto restore_opts;
4700 }
4701
ac27a0ec
DK		 4702 /*
4703 * Mounting a RDONLY partition read-write, so reread
4704 * and store the current valid flag. (It may have
4705 * been changed by e2fsck since we originally mounted
4706 * the partition.)
4707 */
0390131b
FM		 4708 if (sbi->s_journal)
4709 ext4_clear_journal_err(sb, es);
ac27a0ec 4710 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4711 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4712 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4713 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4714 EXT4_FEATURE_INCOMPAT_MMP))
4715 if (ext4_multi_mount_protect(sb,
4716 le64_to_cpu(es->s_mmp_block))) {
4717 err = -EROFS;
4718 goto restore_opts;
4719 }
c79d967d 4720 enable_quota = 1;
ac27a0ec
DK		 4721 }
4722 }
bfff6873
LC		 4723
4724 /*
4725 * Reinitialize lazy itable initialization thread based on
4726 * current settings
4727 */
4728 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4729 ext4_unregister_li_request(sb);
4730 else {
4731 ext4_group_t first_not_zeroed;
4732 first_not_zeroed = ext4_has_uninit_itable(sb);
4733 ext4_register_li_request(sb, first_not_zeroed);
4734 }
4735
6fd058f7 4736 ext4_setup_system_zone(sb);
0390131b 4737 if (sbi->s_journal == NULL)
e2d67052 4738 ext4_commit_super(sb, 1);
0390131b 4739
ac27a0ec
DK		 4740#ifdef CONFIG_QUOTA
4741 /* Release old quota file names */
4742 for (i = 0; i < MAXQUOTAS; i++)
4743 if (old_opts.s_qf_names[i] &&
4744 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4745 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4746 if (enable_quota) {
4747 if (sb_any_quota_suspended(sb))
4748 dquot_resume(sb, -1);
4749 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4750 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4751 err = ext4_enable_quotas(sb);
07724f98 4752 if (err)
7c319d32 4753 goto restore_opts;
7c319d32
AK		 4754 }
4755 }
ac27a0ec 4756#endif
d4c402d9
CW		 4757
4758 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4759 kfree(orig_data);
ac27a0ec 4760 return 0;
0b8e58a1 4761
ac27a0ec
DK		 4762restore_opts:
4763 sb->s_flags = old_sb_flags;
4764 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4765 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4766 sbi->s_resuid = old_opts.s_resuid;
4767 sbi->s_resgid = old_opts.s_resgid;
4768 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4769 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4770 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4771#ifdef CONFIG_QUOTA
4772 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4773 for (i = 0; i < MAXQUOTAS; i++) {
4774 if (sbi->s_qf_names[i] &&
4775 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4776 kfree(sbi->s_qf_names[i]);
4777 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4778 }
4779#endif
d4c402d9 4780 kfree(orig_data);
ac27a0ec
DK		 4781 return err;
4782}
4783
2b2d6d01 4784static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4785{
4786 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4787 struct ext4_sb_info *sbi = EXT4_SB(sb);
4788 struct ext4_super_block *es = sbi->s_es;
952fc18e 4789 ext4_fsblk_t overhead = 0;
960cc398 4790 u64 fsid;
d02a9391 4791 s64 bfree;
ac27a0ec 4792
952fc18e
TT		 4793 if (!test_opt(sb, MINIX_DF))
4794 overhead = sbi->s_overhead;
ac27a0ec 4795
617ba13b 4796 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4797 buf->f_bsize = sb->s_blocksize;
952fc18e 4798 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
57042651
TT		 4799 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4800 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4801 /* prevent underflow in case that few free space is available */
57042651 4802 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4803 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4804 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4805 buf->f_bavail = 0;
4806 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4807 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4808 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4809 fsid = le64_to_cpup((void *)es->s_uuid) ^
4810 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4811 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4812 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4813
ac27a0ec
DK		 4814 return 0;
4815}
4816
0b8e58a1
AD		 4817/* Helper function for writing quotas on sync - we need to start transaction
4818 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4819 * Process 1 Process 2
617ba13b 4820 * ext4_create() quota_sync()
a269eb18 4821 * jbd2_journal_start() write_dquot()
871a2931 4822 * dquot_initialize() down(dqio_mutex)
dab291af 4823 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4824 *
4825 */
4826
4827#ifdef CONFIG_QUOTA
4828
4829static inline struct inode *dquot_to_inode(struct dquot *dquot)
4830{
4831 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4832}
4833
617ba13b 4834static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4835{
4836 int ret, err;
4837 handle_t *handle;
4838 struct inode *inode;
4839
4840 inode = dquot_to_inode(dquot);
617ba13b 4841 handle = ext4_journal_start(inode,
0b8e58a1 4842 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4843 if (IS_ERR(handle))
4844 return PTR_ERR(handle);
4845 ret = dquot_commit(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_acquire_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_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4859 if (IS_ERR(handle))
4860 return PTR_ERR(handle);
4861 ret = dquot_acquire(dquot);
617ba13b 4862 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4863 if (!ret)
4864 ret = err;
4865 return ret;
4866}
4867
617ba13b 4868static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4869{
4870 int ret, err;
4871 handle_t *handle;
4872
617ba13b 4873 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4874 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4875 if (IS_ERR(handle)) {
4876 /* Release dquot anyway to avoid endless cycle in dqput() */
4877 dquot_release(dquot);
ac27a0ec 4878 return PTR_ERR(handle);
9c3013e9 4879 }
ac27a0ec 4880 ret = dquot_release(dquot);
617ba13b 4881 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4882 if (!ret)
4883 ret = err;
4884 return ret;
4885}
4886
617ba13b 4887static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4888{
2c8be6b2 4889 /* Are we journaling quotas? */
617ba13b
MC		 4890 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4891 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4892 dquot_mark_dquot_dirty(dquot);
617ba13b 4893 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4894 } else {
4895 return dquot_mark_dquot_dirty(dquot);
4896 }
4897}
4898
617ba13b 4899static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4900{
4901 int ret, err;
4902 handle_t *handle;
4903
4904 /* Data block + inode block */
617ba13b 4905 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4906 if (IS_ERR(handle))
4907 return PTR_ERR(handle);
4908 ret = dquot_commit_info(sb, type);
617ba13b 4909 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4910 if (!ret)
4911 ret = err;
4912 return ret;
4913}
4914
4915/*
4916 * Turn on quotas during mount time - we need to find
4917 * the quota file and such...
4918 */
617ba13b 4919static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4920{
287a8095
CH		 4921 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4922 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4923}
4924
4925/*
4926 * Standard function to be called on quota_on
4927 */
617ba13b 4928static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4929 struct path *path)
ac27a0ec
DK		 4930{
4931 int err;
ac27a0ec
DK		 4932
4933 if (!test_opt(sb, QUOTA))
4934 return -EINVAL;
0623543b 4935
ac27a0ec 4936 /* Quotafile not on the same filesystem? */
d8c9584e 4937 if (path->dentry->d_sb != sb)
ac27a0ec 4938 return -EXDEV;
0623543b
JK		 4939 /* Journaling quota? */
4940 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4941 /* Quotafile not in fs root? */
f00c9e44 4942 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4943 ext4_msg(sb, KERN_WARNING,
4944 "Quota file not on filesystem root. "
4945 "Journaled quota will not work");
2b2d6d01 4946 }
0623543b
JK		 4947
4948 /*
4949 * When we journal data on quota file, we have to flush journal to see
4950 * all updates to the file when we bypass pagecache...
4951 */
0390131b 4952 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4953 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4954 /*
4955 * We don't need to lock updates but journal_flush() could
4956 * otherwise be livelocked...
4957 */
4958 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4959 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4960 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4961 if (err)
7ffe1ea8 4962 return err;
0623543b
JK		 4963 }
4964
f00c9e44 4965 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4966}
4967
7c319d32
AK		 4968static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4969 unsigned int flags)
4970{
4971 int err;
4972 struct inode *qf_inode;
4973 unsigned long qf_inums[MAXQUOTAS] = {
4974 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4975 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4976 };
4977
4978 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4979
4980 if (!qf_inums[type])
4981 return -EPERM;
4982
4983 qf_inode = ext4_iget(sb, qf_inums[type]);
4984 if (IS_ERR(qf_inode)) {
4985 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4986 return PTR_ERR(qf_inode);
4987 }
4988
4989 err = dquot_enable(qf_inode, type, format_id, flags);
4990 iput(qf_inode);
4991
4992 return err;
4993}
4994
4995/* Enable usage tracking for all quota types. */
4996static int ext4_enable_quotas(struct super_block *sb)
4997{
4998 int type, err = 0;
4999 unsigned long qf_inums[MAXQUOTAS] = {
5000 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5001 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5002 };
5003
5004 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5005 for (type = 0; type < MAXQUOTAS; type++) {
5006 if (qf_inums[type]) {
5007 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5008 DQUOT_USAGE_ENABLED);
5009 if (err) {
5010 ext4_warning(sb,
5011 "Failed to enable quota (type=%d) "
5012 "tracking. Please run e2fsck to fix.",
5013 type);
5014 return err;
5015 }
5016 }
5017 }
5018 return 0;
5019}
5020
5021/*
5022 * quota_on function that is used when QUOTA feature is set.
5023 */
5024static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5025 int format_id)
5026{
5027 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5028 return -EINVAL;
5029
5030 /*
5031 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5032 */
5033 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5034}
5035
ca0e05e4
DM		 5036static int ext4_quota_off(struct super_block *sb, int type)
5037{
21f97697
JK		 5038 struct inode *inode = sb_dqopt(sb)->files[type];
5039 handle_t *handle;
5040
87009d86
DM		 5041 /* Force all delayed allocation blocks to be allocated.
5042 * Caller already holds s_umount sem */
5043 if (test_opt(sb, DELALLOC))
ca0e05e4 5044 sync_filesystem(sb);
ca0e05e4 5045
0b268590
AG		 5046 if (!inode)
5047 goto out;
5048
21f97697
JK		 5049 /* Update modification times of quota files when userspace can
5050 * start looking at them */
5051 handle = ext4_journal_start(inode, 1);
5052 if (IS_ERR(handle))
5053 goto out;
5054 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5055 ext4_mark_inode_dirty(handle, inode);
5056 ext4_journal_stop(handle);
5057
5058out:
ca0e05e4
DM		 5059 return dquot_quota_off(sb, type);
5060}
5061
7c319d32
AK		 5062/*
5063 * quota_off function that is used when QUOTA feature is set.
5064 */
5065static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5066{
5067 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5068 return -EINVAL;
5069
5070 /* Disable only the limits. */
5071 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5072}
5073
ac27a0ec
DK		 5074/* Read data from quotafile - avoid pagecache and such because we cannot afford
5075 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5076 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5077 * we don't have to be afraid of races */
617ba13b 5078static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5079 size_t len, loff_t off)
5080{
5081 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5082 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5083 int err = 0;
5084 int offset = off & (sb->s_blocksize - 1);
5085 int tocopy;
5086 size_t toread;
5087 struct buffer_head *bh;
5088 loff_t i_size = i_size_read(inode);
5089
5090 if (off > i_size)
5091 return 0;
5092 if (off+len > i_size)
5093 len = i_size-off;
5094 toread = len;
5095 while (toread > 0) {
5096 tocopy = sb->s_blocksize - offset < toread ?
5097 sb->s_blocksize - offset : toread;
617ba13b 5098 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5099 if (err)
5100 return err;
5101 if (!bh) /* A hole? */
5102 memset(data, 0, tocopy);
5103 else
5104 memcpy(data, bh->b_data+offset, tocopy);
5105 brelse(bh);
5106 offset = 0;
5107 toread -= tocopy;
5108 data += tocopy;
5109 blk++;
5110 }
5111 return len;
5112}
5113
5114/* Write to quotafile (we know the transaction is already started and has
5115 * enough credits) */
617ba13b 5116static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5117 const char *data, size_t len, loff_t off)
5118{
5119 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5120 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5121 int err = 0;
5122 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5123 struct buffer_head *bh;
5124 handle_t *handle = journal_current_handle();
5125
0390131b 5126 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5127 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5128 " cancelled because transaction is not started",
9c3013e9
JK		 5129 (unsigned long long)off, (unsigned long long)len);
5130 return -EIO;
5131 }
67eeb568
DM		 5132 /*
5133 * Since we account only one data block in transaction credits,
5134 * then it is impossible to cross a block boundary.
5135 */
5136 if (sb->s_blocksize - offset < len) {
5137 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5138 " cancelled because not block aligned",
5139 (unsigned long long)off, (unsigned long long)len);
5140 return -EIO;
5141 }
5142
67eeb568
DM		 5143 bh = ext4_bread(handle, inode, blk, 1, &err);
5144 if (!bh)
5145 goto out;
62d2b5f2
JK		 5146 err = ext4_journal_get_write_access(handle, bh);
5147 if (err) {
5148 brelse(bh);
5149 goto out;
ac27a0ec 5150 }
67eeb568
DM		 5151 lock_buffer(bh);
5152 memcpy(bh->b_data+offset, data, len);
5153 flush_dcache_page(bh->b_page);
5154 unlock_buffer(bh);
62d2b5f2 5155 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5156 brelse(bh);
ac27a0ec 5157out:
0b7f7cef 5158 if (err)
ac27a0ec 5159 return err;
67eeb568
DM		 5160 if (inode->i_size < off + len) {
5161 i_size_write(inode, off + len);
617ba13b 5162 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5163 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5164 }
67eeb568 5165 return len;
ac27a0ec
DK		 5166}
5167
5168#endif
5169
152a0836
AV		 5170static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5171 const char *dev_name, void *data)
ac27a0ec 5172{
152a0836 5173 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5174}
5175
37f328eb 5176#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5177static inline void register_as_ext2(void)
5178{
5179 int err = register_filesystem(&ext2_fs_type);
5180 if (err)
5181 printk(KERN_WARNING
5182 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5183}
5184
5185static inline void unregister_as_ext2(void)
5186{
5187 unregister_filesystem(&ext2_fs_type);
5188}
2035e776
TT		 5189
5190static inline int ext2_feature_set_ok(struct super_block *sb)
5191{
5192 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5193 return 0;
5194 if (sb->s_flags & MS_RDONLY)
5195 return 1;
5196 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5197 return 0;
5198 return 1;
5199}
51b7e3c9 5200MODULE_ALIAS("ext2");
24b58424
TT		 5201#else
5202static inline void register_as_ext2(void) { }
5203static inline void unregister_as_ext2(void) { }
2035e776 5204static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5205#endif
5206
37f328eb 5207#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5208static inline void register_as_ext3(void)
5209{
5210 int err = register_filesystem(&ext3_fs_type);
5211 if (err)
5212 printk(KERN_WARNING
5213 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5214}
5215
5216static inline void unregister_as_ext3(void)
5217{
5218 unregister_filesystem(&ext3_fs_type);
5219}
2035e776
TT		 5220
5221static inline int ext3_feature_set_ok(struct super_block *sb)
5222{
5223 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5224 return 0;
5225 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5226 return 0;
5227 if (sb->s_flags & MS_RDONLY)
5228 return 1;
5229 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5230 return 0;
5231 return 1;
5232}
51b7e3c9 5233MODULE_ALIAS("ext3");
24b58424
TT		 5234#else
5235static inline void register_as_ext3(void) { }
5236static inline void unregister_as_ext3(void) { }
2035e776 5237static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5238#endif
5239
03010a33
TT		 5240static struct file_system_type ext4_fs_type = {
5241 .owner = THIS_MODULE,
5242 .name = "ext4",
152a0836 5243 .mount = ext4_mount,
03010a33
TT		 5244 .kill_sb = kill_block_super,
5245 .fs_flags = FS_REQUIRES_DEV,
5246};
5247
8f021222 5248static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5249{
5250 struct ext4_features *ef;
5251 int ret = -ENOMEM;
5252
5253 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5254 if (!ef)
5255 goto out;
5256
5257 ef->f_kobj.kset = ext4_kset;
5258 init_completion(&ef->f_kobj_unregister);
5259 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5260 "features");
5261 if (ret) {
5262 kfree(ef);
5263 goto out;
5264 }
5265
5266 ext4_feat = ef;
5267 ret = 0;
5268out:
5269 return ret;
5270}
5271
8f021222
LC		 5272static void ext4_exit_feat_adverts(void)
5273{
5274 kobject_put(&ext4_feat->f_kobj);
5275 wait_for_completion(&ext4_feat->f_kobj_unregister);
5276 kfree(ext4_feat);
5277}
5278
e9e3bcec
ES		 5279/* Shared across all ext4 file systems */
5280wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5281struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5282
5dabfc78 5283static int __init ext4_init_fs(void)
ac27a0ec 5284{
e9e3bcec 5285 int i, err;
c9de560d 5286
07c0c5d8
AV		 5287 ext4_li_info = NULL;
5288 mutex_init(&ext4_li_mtx);
5289
12e9b892 5290 ext4_check_flag_values();
e9e3bcec
ES		 5291
5292 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5293 mutex_init(&ext4__aio_mutex[i]);
5294 init_waitqueue_head(&ext4__ioend_wq[i]);
5295 }
5296
5dabfc78 5297 err = ext4_init_pageio();
6fd058f7
TT		 5298 if (err)
5299 return err;
5dabfc78 5300 err = ext4_init_system_zone();
bd2d0210 5301 if (err)
d44651d0 5302 goto out6;
3197ebdb 5303 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5304 if (!ext4_kset) {
5305 err = -ENOMEM;
dd68314c 5306 goto out5;
0e376b1e 5307 }
d44651d0 5308 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5309
5310 err = ext4_init_feat_adverts();
dd68314c
TT		 5311 if (err)
5312 goto out4;
857ac889 5313
5dabfc78 5314 err = ext4_init_mballoc();
ac27a0ec 5315 if (err)
6fd058f7 5316 goto out3;
c9de560d 5317
5dabfc78 5318 err = ext4_init_xattr();
c9de560d
AT		 5319 if (err)
5320 goto out2;
ac27a0ec
DK		 5321 err = init_inodecache();
5322 if (err)
5323 goto out1;
24b58424 5324 register_as_ext3();
2035e776 5325 register_as_ext2();
03010a33 5326 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5327 if (err)
5328 goto out;
bfff6873 5329
ac27a0ec
DK		 5330 return 0;
5331out:
24b58424
TT		 5332 unregister_as_ext2();
5333 unregister_as_ext3();
ac27a0ec
DK		 5334 destroy_inodecache();
5335out1:
5dabfc78 5336 ext4_exit_xattr();
c9de560d 5337out2:
5dabfc78 5338 ext4_exit_mballoc();
6fd058f7 5339out3:
8f021222 5340 ext4_exit_feat_adverts();
dd68314c 5341out4:
d44651d0
FJ		 5342 if (ext4_proc_root)
5343 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5344 kset_unregister(ext4_kset);
d44651d0 5345out5:
5dabfc78 5346 ext4_exit_system_zone();
d44651d0 5347out6:
5dabfc78 5348 ext4_exit_pageio();
ac27a0ec
DK		 5349 return err;
5350}
5351
5dabfc78 5352static void __exit ext4_exit_fs(void)
ac27a0ec 5353{
bfff6873 5354 ext4_destroy_lazyinit_thread();
24b58424
TT		 5355 unregister_as_ext2();
5356 unregister_as_ext3();
03010a33 5357 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5358 destroy_inodecache();
5dabfc78
TT		 5359 ext4_exit_xattr();
5360 ext4_exit_mballoc();
8f021222 5361 ext4_exit_feat_adverts();
9f6200bb 5362 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5363 kset_unregister(ext4_kset);
5dabfc78
TT		 5364 ext4_exit_system_zone();
5365 ext4_exit_pageio();
ac27a0ec
DK		 5366}
5367
5368MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5369MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5370MODULE_LICENSE("GPL");
5dabfc78
TT		 5371module_init(ext4_init_fs)
5372module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository