projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
userns: Modify dqget to take struct kqid
[deliverable/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
6f91bc5f 48#include "ext4_extents.h"
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 72static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 73 char nbuf[16]);
2b2d6d01
TT		 74static int ext4_remount(struct super_block *sb, int *flags, char *data);
75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 76static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 77static int ext4_freeze(struct super_block *sb);
152a0836
AV		 78static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
2035e776
TT		 80static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 82static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 83static void ext4_destroy_lazyinit_thread(void);
84static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 85static void ext4_clear_request_list(void);
ac27a0ec 86
2035e776
TT		 87#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94};
95#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96#else
97#define IS_EXT2_SB(sb) (0)
98#endif
99
100
ba69f9ab
JK		 101#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
104 .name = "ext3",
152a0836 105 .mount = ext4_mount,
ba69f9ab
JK		 106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
108};
109#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110#else
111#define IS_EXT3_SB(sb) (0)
112#endif
bd81d8ee 113
d25425f8
DW		 114static int ext4_verify_csum_type(struct super_block *sb,
115 struct ext4_super_block *es)
116{
117 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
118 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
119 return 1;
120
121 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
122}
123
a9c47317
DW		 124static __le32 ext4_superblock_csum(struct super_block *sb,
125 struct ext4_super_block *es)
126{
127 struct ext4_sb_info *sbi = EXT4_SB(sb);
128 int offset = offsetof(struct ext4_super_block, s_checksum);
129 __u32 csum;
130
131 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
132
133 return cpu_to_le32(csum);
134}
135
136int ext4_superblock_csum_verify(struct super_block *sb,
137 struct ext4_super_block *es)
138{
139 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
140 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
141 return 1;
142
143 return es->s_checksum == ext4_superblock_csum(sb, es);
144}
145
146void ext4_superblock_csum_set(struct super_block *sb,
147 struct ext4_super_block *es)
148{
149 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
150 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
151 return;
152
153 es->s_checksum = ext4_superblock_csum(sb, es);
154}
155
9933fc0a
TT		 156void *ext4_kvmalloc(size_t size, gfp_t flags)
157{
158 void *ret;
159
160 ret = kmalloc(size, flags);
161 if (!ret)
162 ret = __vmalloc(size, flags, PAGE_KERNEL);
163 return ret;
164}
165
166void *ext4_kvzalloc(size_t size, gfp_t flags)
167{
168 void *ret;
169
db9481c0 170 ret = kzalloc(size, flags);
9933fc0a
TT		 171 if (!ret)
172 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
173 return ret;
174}
175
176void ext4_kvfree(void *ptr)
177{
178 if (is_vmalloc_addr(ptr))
179 vfree(ptr);
180 else
181 kfree(ptr);
182
183}
184
8fadc143
AR		 185ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
186 struct ext4_group_desc *bg)
bd81d8ee 187{
3a14589c 188 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 189 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 190 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 191}
192
8fadc143
AR		 193ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
194 struct ext4_group_desc *bg)
bd81d8ee 195{
5272f837 196 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 197 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 198 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 199}
200
8fadc143
AR		 201ext4_fsblk_t ext4_inode_table(struct super_block *sb,
202 struct ext4_group_desc *bg)
bd81d8ee 203{
5272f837 204 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 206 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 207}
208
021b65bb
TT		 209__u32 ext4_free_group_clusters(struct super_block *sb,
210 struct ext4_group_desc *bg)
560671a0
AK		 211{
212 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 214 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 215}
216
217__u32 ext4_free_inodes_count(struct super_block *sb,
218 struct ext4_group_desc *bg)
219{
220 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 222 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 223}
224
225__u32 ext4_used_dirs_count(struct super_block *sb,
226 struct ext4_group_desc *bg)
227{
228 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 230 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 231}
232
233__u32 ext4_itable_unused_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
235{
236 return le16_to_cpu(bg->bg_itable_unused_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 238 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 239}
240
8fadc143
AR		 241void ext4_block_bitmap_set(struct super_block *sb,
242 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 243{
3a14589c 244 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 245 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
246 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 247}
248
8fadc143
AR		 249void ext4_inode_bitmap_set(struct super_block *sb,
250 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 251{
5272f837 252 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 253 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
254 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 255}
256
8fadc143
AR		 257void ext4_inode_table_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 259{
5272f837 260 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 263}
264
021b65bb
TT		 265void ext4_free_group_clusters_set(struct super_block *sb,
266 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 267{
268 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
271}
272
273void ext4_free_inodes_set(struct super_block *sb,
274 struct ext4_group_desc *bg, __u32 count)
275{
276 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
279}
280
281void ext4_used_dirs_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
283{
284 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
287}
288
289void ext4_itable_unused_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
291{
292 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
295}
296
d3d1faf6
CW		 297
298/* Just increment the non-pointer handle value */
299static handle_t *ext4_get_nojournal(void)
300{
301 handle_t *handle = current->journal_info;
302 unsigned long ref_cnt = (unsigned long)handle;
303
304 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
305
306 ref_cnt++;
307 handle = (handle_t *)ref_cnt;
308
309 current->journal_info = handle;
310 return handle;
311}
312
313
314/* Decrement the non-pointer handle value */
315static void ext4_put_nojournal(handle_t *handle)
316{
317 unsigned long ref_cnt = (unsigned long)handle;
318
319 BUG_ON(ref_cnt == 0);
320
321 ref_cnt--;
322 handle = (handle_t *)ref_cnt;
323
324 current->journal_info = handle;
325}
326
ac27a0ec 327/*
dab291af 328 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 329 *
330 * The only special thing we need to do here is to make sure that all
331 * journal_end calls result in the superblock being marked dirty, so
332 * that sync() will call the filesystem's write_super callback if
333 * appropriate.
334 */
617ba13b 335handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 336{
337 journal_t *journal;
338
12706394 339 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 340 if (sb->s_flags & MS_RDONLY)
341 return ERR_PTR(-EROFS);
342
8e8ad8a5 343 WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE);
617ba13b 344 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 345 if (!journal)
346 return ext4_get_nojournal();
347 /*
348 * Special case here: if the journal has aborted behind our
349 * backs (eg. EIO in the commit thread), then we still need to
350 * take the FS itself readonly cleanly.
351 */
352 if (is_journal_aborted(journal)) {
353 ext4_abort(sb, "Detected aborted journal");
354 return ERR_PTR(-EROFS);
ac27a0ec 355 }
be4f27d3 356 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 357}
358
359/*
360 * The only special thing we need to do here is to make sure that all
dab291af 361 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 362 * that sync() will call the filesystem's write_super callback if
363 * appropriate.
364 */
c398eda0 365int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 366{
367 struct super_block *sb;
368 int err;
369 int rc;
370
0390131b 371 if (!ext4_handle_valid(handle)) {
d3d1faf6 372 ext4_put_nojournal(handle);
0390131b
FM		 373 return 0;
374 }
ac27a0ec
DK		 375 sb = handle->h_transaction->t_journal->j_private;
376 err = handle->h_err;
dab291af 377 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 378
379 if (!err)
380 err = rc;
381 if (err)
c398eda0 382 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 383 return err;
384}
385
90c7201b
TT		 386void ext4_journal_abort_handle(const char *caller, unsigned int line,
387 const char *err_fn, struct buffer_head *bh,
388 handle_t *handle, int err)
ac27a0ec
DK		 389{
390 char nbuf[16];
617ba13b 391 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 392
0390131b
FM		 393 BUG_ON(!ext4_handle_valid(handle));
394
ac27a0ec
DK		 395 if (bh)
396 BUFFER_TRACE(bh, "abort");
397
398 if (!handle->h_err)
399 handle->h_err = err;
400
401 if (is_handle_aborted(handle))
402 return;
403
92b97816 404 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
90c7201b 405 caller, line, errstr, err_fn);
ac27a0ec 406
dab291af 407 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 408}
409
1c13d5c0
TT		 410static void __save_error_info(struct super_block *sb, const char *func,
411 unsigned int line)
412{
413 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
414
415 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
416 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
417 es->s_last_error_time = cpu_to_le32(get_seconds());
418 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
419 es->s_last_error_line = cpu_to_le32(line);
420 if (!es->s_first_error_time) {
421 es->s_first_error_time = es->s_last_error_time;
422 strncpy(es->s_first_error_func, func,
423 sizeof(es->s_first_error_func));
424 es->s_first_error_line = cpu_to_le32(line);
425 es->s_first_error_ino = es->s_last_error_ino;
426 es->s_first_error_block = es->s_last_error_block;
427 }
66e61a9e
TT		 428 /*
429 * Start the daily error reporting function if it hasn't been
430 * started already
431 */
432 if (!es->s_error_count)
433 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 434 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
435}
436
437static void save_error_info(struct super_block *sb, const char *func,
438 unsigned int line)
439{
440 __save_error_info(sb, func, line);
441 ext4_commit_super(sb, 1);
442}
443
7c2e7087
TT		 444/*
445 * The del_gendisk() function uninitializes the disk-specific data
446 * structures, including the bdi structure, without telling anyone
447 * else. Once this happens, any attempt to call mark_buffer_dirty()
448 * (for example, by ext4_commit_super), will cause a kernel OOPS.
449 * This is a kludge to prevent these oops until we can put in a proper
450 * hook in del_gendisk() to inform the VFS and file system layers.
451 */
452static int block_device_ejected(struct super_block *sb)
453{
454 struct inode *bd_inode = sb->s_bdev->bd_inode;
455 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
456
457 return bdi->dev == NULL;
458}
459
18aadd47
BJ		 460static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
461{
462 struct super_block *sb = journal->j_private;
463 struct ext4_sb_info *sbi = EXT4_SB(sb);
464 int error = is_journal_aborted(journal);
465 struct ext4_journal_cb_entry *jce, *tmp;
466
467 spin_lock(&sbi->s_md_lock);
468 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
469 list_del_init(&jce->jce_list);
470 spin_unlock(&sbi->s_md_lock);
471 jce->jce_func(sb, jce, error);
472 spin_lock(&sbi->s_md_lock);
473 }
474 spin_unlock(&sbi->s_md_lock);
475}
1c13d5c0 476
ac27a0ec
DK		 477/* Deal with the reporting of failure conditions on a filesystem such as
478 * inconsistencies detected or read IO failures.
479 *
480 * On ext2, we can store the error state of the filesystem in the
617ba13b 481 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 482 * write ordering constraints on the superblock which prevent us from
483 * writing it out straight away; and given that the journal is about to
484 * be aborted, we can't rely on the current, or future, transactions to
485 * write out the superblock safely.
486 *
dab291af 487 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 488 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 489 * that error until we've noted it down and cleared it.
490 */
491
617ba13b 492static void ext4_handle_error(struct super_block *sb)
ac27a0ec 493{
ac27a0ec
DK		 494 if (sb->s_flags & MS_RDONLY)
495 return;
496
2b2d6d01 497 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 498 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 499
4ab2f15b 500 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 501 if (journal)
dab291af 502 jbd2_journal_abort(journal, -EIO);
ac27a0ec 503 }
2b2d6d01 504 if (test_opt(sb, ERRORS_RO)) {
b31e1552 505 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 506 sb->s_flags |= MS_RDONLY;
507 }
ac27a0ec 508 if (test_opt(sb, ERRORS_PANIC))
617ba13b 509 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 510 sb->s_id);
511}
512
12062ddd 513void __ext4_error(struct super_block *sb, const char *function,
c398eda0 514 unsigned int line, const char *fmt, ...)
ac27a0ec 515{
0ff2ea7d 516 struct va_format vaf;
ac27a0ec
DK		 517 va_list args;
518
519 va_start(args, fmt);
0ff2ea7d
JP		 520 vaf.fmt = fmt;
521 vaf.va = &args;
522 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
523 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec 524 va_end(args);
f3fc0210 525 save_error_info(sb, function, line);
ac27a0ec 526
617ba13b 527 ext4_handle_error(sb);
ac27a0ec
DK		 528}
529
c398eda0
TT		 530void ext4_error_inode(struct inode *inode, const char *function,
531 unsigned int line, ext4_fsblk_t block,
273df556
FM		 532 const char *fmt, ...)
533{
534 va_list args;
f7c21177 535 struct va_format vaf;
1c13d5c0 536 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 537
1c13d5c0
TT		 538 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
539 es->s_last_error_block = cpu_to_le64(block);
540 save_error_info(inode->i_sb, function, line);
273df556 541 va_start(args, fmt);
f7c21177
TT		 542 vaf.fmt = fmt;
543 vaf.va = &args;
c398eda0 544 if (block)
d9ee81da
JP		 545 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
546 "inode #%lu: block %llu: comm %s: %pV\n",
547 inode->i_sb->s_id, function, line, inode->i_ino,
548 block, current->comm, &vaf);
549 else
550 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
551 "inode #%lu: comm %s: %pV\n",
552 inode->i_sb->s_id, function, line, inode->i_ino,
553 current->comm, &vaf);
273df556
FM		 554 va_end(args);
555
556 ext4_handle_error(inode->i_sb);
557}
558
c398eda0 559void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 560 unsigned int line, ext4_fsblk_t block,
561 const char *fmt, ...)
273df556
FM		 562{
563 va_list args;
f7c21177 564 struct va_format vaf;
1c13d5c0 565 struct ext4_super_block *es;
273df556
FM		 566 struct inode *inode = file->f_dentry->d_inode;
567 char pathname[80], *path;
568
1c13d5c0
TT		 569 es = EXT4_SB(inode->i_sb)->s_es;
570 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
571 save_error_info(inode->i_sb, function, line);
273df556 572 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 573 if (IS_ERR(path))
273df556 574 path = "(unknown)";
f7c21177
TT		 575 va_start(args, fmt);
576 vaf.fmt = fmt;
577 vaf.va = &args;
d9ee81da
JP		 578 if (block)
579 printk(KERN_CRIT
580 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
581 "block %llu: comm %s: path %s: %pV\n",
582 inode->i_sb->s_id, function, line, inode->i_ino,
583 block, current->comm, path, &vaf);
584 else
585 printk(KERN_CRIT
586 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
587 "comm %s: path %s: %pV\n",
588 inode->i_sb->s_id, function, line, inode->i_ino,
589 current->comm, path, &vaf);
273df556
FM		 590 va_end(args);
591
592 ext4_handle_error(inode->i_sb);
593}
594
2b2d6d01 595static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 596 char nbuf[16])
597{
598 char *errstr = NULL;
599
600 switch (errno) {
601 case -EIO:
602 errstr = "IO failure";
603 break;
604 case -ENOMEM:
605 errstr = "Out of memory";
606 break;
607 case -EROFS:
78f1ddbb
TT		 608 if (!sb || (EXT4_SB(sb)->s_journal &&
609 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 610 errstr = "Journal has aborted";
611 else
612 errstr = "Readonly filesystem";
613 break;
614 default:
615 /* If the caller passed in an extra buffer for unknown
616 * errors, textualise them now. Else we just return
617 * NULL. */
618 if (nbuf) {
619 /* Check for truncated error codes... */
620 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
621 errstr = nbuf;
622 }
623 break;
624 }
625
626 return errstr;
627}
628
617ba13b 629/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 630 * automatically and invokes the appropriate error response. */
631
c398eda0
TT		 632void __ext4_std_error(struct super_block *sb, const char *function,
633 unsigned int line, int errno)
ac27a0ec
DK		 634{
635 char nbuf[16];
636 const char *errstr;
637
638 /* Special case: if the error is EROFS, and we're not already
639 * inside a transaction, then there's really no point in logging
640 * an error. */
641 if (errno == -EROFS && journal_current_handle() == NULL &&
642 (sb->s_flags & MS_RDONLY))
643 return;
644
617ba13b 645 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 646 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
647 sb->s_id, function, line, errstr);
1c13d5c0 648 save_error_info(sb, function, line);
ac27a0ec 649
617ba13b 650 ext4_handle_error(sb);
ac27a0ec
DK		 651}
652
653/*
617ba13b 654 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 655 * abort function may be used to deal with unrecoverable failures such
656 * as journal IO errors or ENOMEM at a critical moment in log management.
657 *
658 * We unconditionally force the filesystem into an ABORT|READONLY state,
659 * unless the error response on the fs has been set to panic in which
660 * case we take the easy way out and panic immediately.
661 */
662
c67d859e 663void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 664 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 665{
666 va_list args;
667
1c13d5c0 668 save_error_info(sb, function, line);
ac27a0ec 669 va_start(args, fmt);
c398eda0
TT		 670 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
671 function, line);
ac27a0ec
DK		 672 vprintk(fmt, args);
673 printk("\n");
674 va_end(args);
675
1c13d5c0
TT		 676 if ((sb->s_flags & MS_RDONLY) == 0) {
677 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
678 sb->s_flags |= MS_RDONLY;
679 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
680 if (EXT4_SB(sb)->s_journal)
681 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
682 save_error_info(sb, function, line);
683 }
ac27a0ec 684 if (test_opt(sb, ERRORS_PANIC))
617ba13b 685 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 686}
687
0ff2ea7d 688void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 689{
0ff2ea7d 690 struct va_format vaf;
b31e1552
ES		 691 va_list args;
692
693 va_start(args, fmt);
0ff2ea7d
JP		 694 vaf.fmt = fmt;
695 vaf.va = &args;
696 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 697 va_end(args);
698}
699
12062ddd 700void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 701 unsigned int line, const char *fmt, ...)
ac27a0ec 702{
0ff2ea7d 703 struct va_format vaf;
ac27a0ec
DK		 704 va_list args;
705
706 va_start(args, fmt);
0ff2ea7d
JP		 707 vaf.fmt = fmt;
708 vaf.va = &args;
709 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
710 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 711 va_end(args);
712}
713
e29136f8
TT		 714void __ext4_grp_locked_error(const char *function, unsigned int line,
715 struct super_block *sb, ext4_group_t grp,
716 unsigned long ino, ext4_fsblk_t block,
717 const char *fmt, ...)
5d1b1b3f
AK		 718__releases(bitlock)
719__acquires(bitlock)
720{
0ff2ea7d 721 struct va_format vaf;
5d1b1b3f
AK		 722 va_list args;
723 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
724
1c13d5c0
TT		 725 es->s_last_error_ino = cpu_to_le32(ino);
726 es->s_last_error_block = cpu_to_le64(block);
727 __save_error_info(sb, function, line);
0ff2ea7d 728
5d1b1b3f 729 va_start(args, fmt);
0ff2ea7d
JP		 730
731 vaf.fmt = fmt;
732 vaf.va = &args;
21149d61 733 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 734 sb->s_id, function, line, grp);
735 if (ino)
0ff2ea7d 736 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 737 if (block)
0ff2ea7d
JP		 738 printk(KERN_CONT "block %llu:", (unsigned long long) block);
739 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 740 va_end(args);
741
742 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 743 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 744 return;
745 }
1c13d5c0 746
5d1b1b3f
AK		 747 ext4_unlock_group(sb, grp);
748 ext4_handle_error(sb);
749 /*
750 * We only get here in the ERRORS_RO case; relocking the group
751 * may be dangerous, but nothing bad will happen since the
752 * filesystem will have already been marked read/only and the
753 * journal has been aborted. We return 1 as a hint to callers
754 * who might what to use the return value from
25985edc 755 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 756 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
757 * aggressively from the ext4 function in question, with a
758 * more appropriate error code.
759 */
760 ext4_lock_group(sb, grp);
761 return;
762}
763
617ba13b 764void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 765{
617ba13b 766 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 767
617ba13b 768 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 769 return;
770
12062ddd 771 ext4_warning(sb,
ac27a0ec
DK		 772 "updating to rev %d because of new feature flag, "
773 "running e2fsck is recommended",
617ba13b 774 EXT4_DYNAMIC_REV);
ac27a0ec 775
617ba13b
MC		 776 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
777 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
778 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 779 /* leave es->s_feature_*compat flags alone */
780 /* es->s_uuid will be set by e2fsck if empty */
781
782 /*
783 * The rest of the superblock fields should be zero, and if not it
784 * means they are likely already in use, so leave them alone. We
785 * can leave it up to e2fsck to clean up any inconsistencies there.
786 */
787}
788
789/*
790 * Open the external journal device
791 */
b31e1552 792static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 793{
794 struct block_device *bdev;
795 char b[BDEVNAME_SIZE];
796
d4d77629 797 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 798 if (IS_ERR(bdev))
799 goto fail;
800 return bdev;
801
802fail:
b31e1552 803 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 804 __bdevname(dev, b), PTR_ERR(bdev));
805 return NULL;
806}
807
808/*
809 * Release the journal device
810 */
617ba13b 811static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 812{
e525fd89 813 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 814}
815
617ba13b 816static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 817{
818 struct block_device *bdev;
819 int ret = -ENODEV;
820
821 bdev = sbi->journal_bdev;
822 if (bdev) {
617ba13b 823 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 824 sbi->journal_bdev = NULL;
825 }
826 return ret;
827}
828
829static inline struct inode *orphan_list_entry(struct list_head *l)
830{
617ba13b 831 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 832}
833
617ba13b 834static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 835{
836 struct list_head *l;
837
b31e1552
ES		 838 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
839 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 840
841 printk(KERN_ERR "sb_info orphan list:\n");
842 list_for_each(l, &sbi->s_orphan) {
843 struct inode *inode = orphan_list_entry(l);
844 printk(KERN_ERR " "
845 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
846 inode->i_sb->s_id, inode->i_ino, inode,
847 inode->i_mode, inode->i_nlink,
848 NEXT_ORPHAN(inode));
849 }
850}
851
2b2d6d01 852static void ext4_put_super(struct super_block *sb)
ac27a0ec 853{
617ba13b
MC		 854 struct ext4_sb_info *sbi = EXT4_SB(sb);
855 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 856 int i, err;
ac27a0ec 857
857ac889 858 ext4_unregister_li_request(sb);
e0ccfd95
CH		 859 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
860
4c0425ff
MC		 861 flush_workqueue(sbi->dio_unwritten_wq);
862 destroy_workqueue(sbi->dio_unwritten_wq);
863
a9e220f8 864 lock_super(sb);
0390131b
FM		 865 if (sbi->s_journal) {
866 err = jbd2_journal_destroy(sbi->s_journal);
867 sbi->s_journal = NULL;
868 if (err < 0)
c67d859e 869 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 870 }
d4edac31 871
a1c6c569 872 del_timer(&sbi->s_err_report);
d4edac31
JB		 873 ext4_release_system_zone(sb);
874 ext4_mb_release(sb);
875 ext4_ext_release(sb);
876 ext4_xattr_put_super(sb);
877
ac27a0ec 878 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 879 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 880 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 881 }
58c5873a 882 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 883 ext4_commit_super(sb, 1);
884
240799cd 885 if (sbi->s_proc) {
66acdcf4 886 remove_proc_entry("options", sbi->s_proc);
9f6200bb 887 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 888 }
3197ebdb 889 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 890
891 for (i = 0; i < sbi->s_gdb_count; i++)
892 brelse(sbi->s_group_desc[i]);
f18a5f21 893 ext4_kvfree(sbi->s_group_desc);
9933fc0a 894 ext4_kvfree(sbi->s_flex_groups);
57042651 895 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 896 percpu_counter_destroy(&sbi->s_freeinodes_counter);
897 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 898 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 899 brelse(sbi->s_sbh);
900#ifdef CONFIG_QUOTA
901 for (i = 0; i < MAXQUOTAS; i++)
902 kfree(sbi->s_qf_names[i]);
903#endif
904
905 /* Debugging code just in case the in-memory inode orphan list
906 * isn't empty. The on-disk one can be non-empty if we've
907 * detected an error and taken the fs readonly, but the
908 * in-memory list had better be clean by this point. */
909 if (!list_empty(&sbi->s_orphan))
910 dump_orphan_list(sb, sbi);
911 J_ASSERT(list_empty(&sbi->s_orphan));
912
f98393a6 913 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 914 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
915 /*
916 * Invalidate the journal device's buffers. We don't want them
917 * floating about in memory - the physical journal device may
918 * hotswapped, and it breaks the `ro-after' testing code.
919 */
920 sync_blockdev(sbi->journal_bdev);
f98393a6 921 invalidate_bdev(sbi->journal_bdev);
617ba13b 922 ext4_blkdev_remove(sbi);
ac27a0ec 923 }
c5e06d10
JL		 924 if (sbi->s_mmp_tsk)
925 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 926 sb->s_fs_info = NULL;
3197ebdb
TT		 927 /*
928 * Now that we are completely done shutting down the
929 * superblock, we need to actually destroy the kobject.
930 */
3197ebdb
TT		 931 unlock_super(sb);
932 kobject_put(&sbi->s_kobj);
933 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 934 if (sbi->s_chksum_driver)
935 crypto_free_shash(sbi->s_chksum_driver);
705895b6 936 kfree(sbi->s_blockgroup_lock);
ac27a0ec 937 kfree(sbi);
ac27a0ec
DK		 938}
939
e18b890b 940static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 941
942/*
943 * Called inside transaction, so use GFP_NOFS
944 */
617ba13b 945static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 946{
617ba13b 947 struct ext4_inode_info *ei;
ac27a0ec 948
e6b4f8da 949 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 950 if (!ei)
951 return NULL;
0b8e58a1 952
ac27a0ec 953 ei->vfs_inode.i_version = 1;
91246c00 954 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 955 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 956 INIT_LIST_HEAD(&ei->i_prealloc_list);
957 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 958 ei->i_reserved_data_blocks = 0;
959 ei->i_reserved_meta_blocks = 0;
960 ei->i_allocated_meta_blocks = 0;
9d0be502 961 ei->i_da_metadata_calc_len = 0;
d2a17637 962 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 963#ifdef CONFIG_QUOTA
964 ei->i_reserved_quota = 0;
965#endif
8aefcd55 966 ei->jinode = NULL;
c7064ef1 967 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 968 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 969 ei->cur_aio_dio = NULL;
b436b9be
JK		 970 ei->i_sync_tid = 0;
971 ei->i_datasync_tid = 0;
f7ad6d2e 972 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 973 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 974
ac27a0ec
DK		 975 return &ei->vfs_inode;
976}
977
7ff9c073
TT		 978static int ext4_drop_inode(struct inode *inode)
979{
980 int drop = generic_drop_inode(inode);
981
982 trace_ext4_drop_inode(inode, drop);
983 return drop;
984}
985
fa0d7e3d
NP		 986static void ext4_i_callback(struct rcu_head *head)
987{
988 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 989 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
990}
991
617ba13b 992static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 993{
9f7dd93d 994 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 995 ext4_msg(inode->i_sb, KERN_ERR,
996 "Inode %lu (%p): orphan list check failed!",
997 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 998 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
999 EXT4_I(inode), sizeof(struct ext4_inode_info),
1000 true);
1001 dump_stack();
1002 }
fa0d7e3d 1003 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1004}
1005
51cc5068 1006static void init_once(void *foo)
ac27a0ec 1007{
617ba13b 1008 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1009
a35afb83 1010 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 1011#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 1012 init_rwsem(&ei->xattr_sem);
ac27a0ec 1013#endif
0e855ac8 1014 init_rwsem(&ei->i_data_sem);
a35afb83 1015 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1016}
1017
1018static int init_inodecache(void)
1019{
617ba13b
MC		 1020 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1021 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 1022 0, (SLAB_RECLAIM_ACCOUNT|
1023 SLAB_MEM_SPREAD),
20c2df83 1024 init_once);
617ba13b 1025 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1026 return -ENOMEM;
1027 return 0;
1028}
1029
1030static void destroy_inodecache(void)
1031{
617ba13b 1032 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1033}
1034
0930fcc1 1035void ext4_clear_inode(struct inode *inode)
ac27a0ec 1036{
0930fcc1 1037 invalidate_inode_buffers(inode);
dbd5768f 1038 clear_inode(inode);
9f754758 1039 dquot_drop(inode);
c2ea3fde 1040 ext4_discard_preallocations(inode);
8aefcd55
TT		 1041 if (EXT4_I(inode)->jinode) {
1042 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1043 EXT4_I(inode)->jinode);
1044 jbd2_free_inode(EXT4_I(inode)->jinode);
1045 EXT4_I(inode)->jinode = NULL;
1046 }
ac27a0ec
DK		 1047}
1048
1b961ac0 1049static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1050 u64 ino, u32 generation)
ac27a0ec 1051{
ac27a0ec 1052 struct inode *inode;
ac27a0ec 1053
617ba13b 1054 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1055 return ERR_PTR(-ESTALE);
617ba13b 1056 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1057 return ERR_PTR(-ESTALE);
1058
1059 /* iget isn't really right if the inode is currently unallocated!!
1060 *
617ba13b 1061 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1062 * deleted, so we should be safe.
1063 *
1064 * Currently we don't know the generation for parent directory, so
1065 * a generation of 0 means "accept any"
1066 */
1d1fe1ee
DH		 1067 inode = ext4_iget(sb, ino);
1068 if (IS_ERR(inode))
1069 return ERR_CAST(inode);
1070 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1071 iput(inode);
1072 return ERR_PTR(-ESTALE);
1073 }
1b961ac0
CH		 1074
1075 return inode;
1076}
1077
1078static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1079 int fh_len, int fh_type)
1b961ac0
CH		 1080{
1081 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1082 ext4_nfs_get_inode);
1083}
1084
1085static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1086 int fh_len, int fh_type)
1b961ac0
CH		 1087{
1088 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1089 ext4_nfs_get_inode);
ac27a0ec
DK		 1090}
1091
c39a7f84
TO		 1092/*
1093 * Try to release metadata pages (indirect blocks, directories) which are
1094 * mapped via the block device. Since these pages could have journal heads
1095 * which would prevent try_to_free_buffers() from freeing them, we must use
1096 * jbd2 layer's try_to_free_buffers() function to release them.
1097 */
0b8e58a1
AD		 1098static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1099 gfp_t wait)
c39a7f84
TO		 1100{
1101 journal_t *journal = EXT4_SB(sb)->s_journal;
1102
1103 WARN_ON(PageChecked(page));
1104 if (!page_has_buffers(page))
1105 return 0;
1106 if (journal)
1107 return jbd2_journal_try_to_free_buffers(journal, page,
1108 wait & ~__GFP_WAIT);
1109 return try_to_free_buffers(page);
1110}
1111
ac27a0ec 1112#ifdef CONFIG_QUOTA
af5bc92d 1113#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1114#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1115
617ba13b
MC		 1116static int ext4_write_dquot(struct dquot *dquot);
1117static int ext4_acquire_dquot(struct dquot *dquot);
1118static int ext4_release_dquot(struct dquot *dquot);
1119static int ext4_mark_dquot_dirty(struct dquot *dquot);
1120static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1121static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1122 struct path *path);
7c319d32
AK		 1123static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1124 int format_id);
ca0e05e4 1125static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1126static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1127static int ext4_quota_on_mount(struct super_block *sb, int type);
1128static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1129 size_t len, loff_t off);
617ba13b 1130static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1131 const char *data, size_t len, loff_t off);
7c319d32
AK		 1132static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1133 unsigned int flags);
1134static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1135
61e225dc 1136static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1137 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1138 .write_dquot = ext4_write_dquot,
1139 .acquire_dquot = ext4_acquire_dquot,
1140 .release_dquot = ext4_release_dquot,
1141 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1142 .write_info = ext4_write_info,
1143 .alloc_dquot = dquot_alloc,
1144 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1145};
1146
0d54b217 1147static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1148 .quota_on = ext4_quota_on,
ca0e05e4 1149 .quota_off = ext4_quota_off,
287a8095
CH		 1150 .quota_sync = dquot_quota_sync,
1151 .get_info = dquot_get_dqinfo,
1152 .set_info = dquot_set_dqinfo,
1153 .get_dqblk = dquot_get_dqblk,
1154 .set_dqblk = dquot_set_dqblk
ac27a0ec 1155};
7c319d32
AK		 1156
1157static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1158 .quota_on_meta = ext4_quota_on_sysfile,
1159 .quota_off = ext4_quota_off_sysfile,
1160 .quota_sync = dquot_quota_sync,
1161 .get_info = dquot_get_dqinfo,
1162 .set_info = dquot_set_dqinfo,
1163 .get_dqblk = dquot_get_dqblk,
1164 .set_dqblk = dquot_set_dqblk
1165};
ac27a0ec
DK		 1166#endif
1167
ee9b6d61 1168static const struct super_operations ext4_sops = {
617ba13b
MC		 1169 .alloc_inode = ext4_alloc_inode,
1170 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1171 .write_inode = ext4_write_inode,
1172 .dirty_inode = ext4_dirty_inode,
7ff9c073 1173 .drop_inode = ext4_drop_inode,
0930fcc1 1174 .evict_inode = ext4_evict_inode,
617ba13b 1175 .put_super = ext4_put_super,
617ba13b 1176 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1177 .freeze_fs = ext4_freeze,
1178 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1179 .statfs = ext4_statfs,
1180 .remount_fs = ext4_remount,
617ba13b 1181 .show_options = ext4_show_options,
ac27a0ec 1182#ifdef CONFIG_QUOTA
617ba13b
MC		 1183 .quota_read = ext4_quota_read,
1184 .quota_write = ext4_quota_write,
ac27a0ec 1185#endif
c39a7f84 1186 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1187};
1188
9ca92389
TT		 1189static const struct super_operations ext4_nojournal_sops = {
1190 .alloc_inode = ext4_alloc_inode,
1191 .destroy_inode = ext4_destroy_inode,
1192 .write_inode = ext4_write_inode,
1193 .dirty_inode = ext4_dirty_inode,
7ff9c073 1194 .drop_inode = ext4_drop_inode,
0930fcc1 1195 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1196 .put_super = ext4_put_super,
1197 .statfs = ext4_statfs,
1198 .remount_fs = ext4_remount,
9ca92389
TT		 1199 .show_options = ext4_show_options,
1200#ifdef CONFIG_QUOTA
1201 .quota_read = ext4_quota_read,
1202 .quota_write = ext4_quota_write,
1203#endif
1204 .bdev_try_to_free_page = bdev_try_to_free_page,
1205};
1206
39655164 1207static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1208 .fh_to_dentry = ext4_fh_to_dentry,
1209 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1210 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1211};
1212
1213enum {
1214 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1215 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1216 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1217 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1218 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1219 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1220 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1221 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1222 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1223 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1224 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1225 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1226 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1227 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1228 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1229 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1230 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1231 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
ac27a0ec
DK		 1232};
1233
a447c093 1234static const match_table_t tokens = {
ac27a0ec
DK		 1235 {Opt_bsd_df, "bsddf"},
1236 {Opt_minix_df, "minixdf"},
1237 {Opt_grpid, "grpid"},
1238 {Opt_grpid, "bsdgroups"},
1239 {Opt_nogrpid, "nogrpid"},
1240 {Opt_nogrpid, "sysvgroups"},
1241 {Opt_resgid, "resgid=%u"},
1242 {Opt_resuid, "resuid=%u"},
1243 {Opt_sb, "sb=%u"},
1244 {Opt_err_cont, "errors=continue"},
1245 {Opt_err_panic, "errors=panic"},
1246 {Opt_err_ro, "errors=remount-ro"},
1247 {Opt_nouid32, "nouid32"},
ac27a0ec 1248 {Opt_debug, "debug"},
72578c33
TT		 1249 {Opt_removed, "oldalloc"},
1250 {Opt_removed, "orlov"},
ac27a0ec
DK		 1251 {Opt_user_xattr, "user_xattr"},
1252 {Opt_nouser_xattr, "nouser_xattr"},
1253 {Opt_acl, "acl"},
1254 {Opt_noacl, "noacl"},
e3bb52ae 1255 {Opt_noload, "norecovery"},
5a916be1 1256 {Opt_noload, "noload"},
72578c33
TT		 1257 {Opt_removed, "nobh"},
1258 {Opt_removed, "bh"},
ac27a0ec 1259 {Opt_commit, "commit=%u"},
30773840
TT		 1260 {Opt_min_batch_time, "min_batch_time=%u"},
1261 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1262 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1263 {Opt_journal_checksum, "journal_checksum"},
1264 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1265 {Opt_abort, "abort"},
1266 {Opt_data_journal, "data=journal"},
1267 {Opt_data_ordered, "data=ordered"},
1268 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1269 {Opt_data_err_abort, "data_err=abort"},
1270 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1271 {Opt_offusrjquota, "usrjquota="},
1272 {Opt_usrjquota, "usrjquota=%s"},
1273 {Opt_offgrpjquota, "grpjquota="},
1274 {Opt_grpjquota, "grpjquota=%s"},
1275 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1276 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1277 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1278 {Opt_grpquota, "grpquota"},
1279 {Opt_noquota, "noquota"},
1280 {Opt_quota, "quota"},
1281 {Opt_usrquota, "usrquota"},
1282 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1283 {Opt_barrier, "barrier"},
1284 {Opt_nobarrier, "nobarrier"},
25ec56b5 1285 {Opt_i_version, "i_version"},
c9de560d 1286 {Opt_stripe, "stripe=%u"},
64769240 1287 {Opt_delalloc, "delalloc"},
dd919b98 1288 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1289 {Opt_mblk_io_submit, "mblk_io_submit"},
1290 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1291 {Opt_block_validity, "block_validity"},
1292 {Opt_noblock_validity, "noblock_validity"},
240799cd 1293 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1294 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1295 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1296 {Opt_auto_da_alloc, "auto_da_alloc"},
1297 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1298 {Opt_dioread_nolock, "dioread_nolock"},
1299 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1300 {Opt_discard, "discard"},
1301 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1302 {Opt_init_itable, "init_itable=%u"},
1303 {Opt_init_itable, "init_itable"},
1304 {Opt_noinit_itable, "noinit_itable"},
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},
1485 {Opt_err, 0, 0}
1486};
1487
1488static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1489 substring_t *args, unsigned long *journal_devnum,
1490 unsigned int *journal_ioprio, int is_remount)
1491{
1492 struct ext4_sb_info *sbi = EXT4_SB(sb);
1493 const struct mount_opts *m;
08cefc7a
EB		 1494 kuid_t uid;
1495 kgid_t gid;
26092bf5
TT		 1496 int arg = 0;
1497
57f73c2c
TT		 1498#ifdef CONFIG_QUOTA
1499 if (token == Opt_usrjquota)
1500 return set_qf_name(sb, USRQUOTA, &args[0]);
1501 else if (token == Opt_grpjquota)
1502 return set_qf_name(sb, GRPQUOTA, &args[0]);
1503 else if (token == Opt_offusrjquota)
1504 return clear_qf_name(sb, USRQUOTA);
1505 else if (token == Opt_offgrpjquota)
1506 return clear_qf_name(sb, GRPQUOTA);
1507#endif
26092bf5
TT		 1508 if (args->from && match_int(args, &arg))
1509 return -1;
1510 switch (token) {
f7048605
TT		 1511 case Opt_noacl:
1512 case Opt_nouser_xattr:
1513 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1514 break;
26092bf5
TT		 1515 case Opt_sb:
1516 return 1; /* handled by get_sb_block() */
1517 case Opt_removed:
1518 ext4_msg(sb, KERN_WARNING,
1519 "Ignoring removed %s option", opt);
1520 return 1;
1521 case Opt_resuid:
08cefc7a
EB		 1522 uid = make_kuid(current_user_ns(), arg);
1523 if (!uid_valid(uid)) {
1524 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1525 return -1;
1526 }
1527 sbi->s_resuid = uid;
26092bf5
TT		 1528 return 1;
1529 case Opt_resgid:
08cefc7a
EB		 1530 gid = make_kgid(current_user_ns(), arg);
1531 if (!gid_valid(gid)) {
1532 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1533 return -1;
1534 }
1535 sbi->s_resgid = gid;
26092bf5
TT		 1536 return 1;
1537 case Opt_abort:
1538 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1539 return 1;
1540 case Opt_i_version:
1541 sb->s_flags |= MS_I_VERSION;
1542 return 1;
1543 case Opt_journal_dev:
1544 if (is_remount) {
1545 ext4_msg(sb, KERN_ERR,
1546 "Cannot specify journal on remount");
1547 return -1;
1548 }
1549 *journal_devnum = arg;
1550 return 1;
1551 case Opt_journal_ioprio:
1552 if (arg < 0 || arg > 7)
1553 return -1;
1554 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1555 return 1;
1556 }
1557
1558 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1559 if (token != m->token)
1560 continue;
1561 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1562 return -1;
1563 if (m->flags & MOPT_EXPLICIT)
1564 set_opt2(sb, EXPLICIT_DELALLOC);
1565 if (m->flags & MOPT_CLEAR_ERR)
1566 clear_opt(sb, ERRORS_MASK);
1567 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1568 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1569 "options when quota turned on");
1570 return -1;
1571 }
1572
1573 if (m->flags & MOPT_NOSUPPORT) {
1574 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1575 } else if (token == Opt_commit) {
1576 if (arg == 0)
1577 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1578 sbi->s_commit_interval = HZ * arg;
1579 } else if (token == Opt_max_batch_time) {
1580 if (arg == 0)
1581 arg = EXT4_DEF_MAX_BATCH_TIME;
1582 sbi->s_max_batch_time = arg;
1583 } else if (token == Opt_min_batch_time) {
1584 sbi->s_min_batch_time = arg;
1585 } else if (token == Opt_inode_readahead_blks) {
1586 if (arg > (1 << 30))
1587 return -1;
1588 if (arg && !is_power_of_2(arg)) {
b31e1552 1589 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1590 "EXT4-fs: inode_readahead_blks"
1591 " must be a power of 2");
1592 return -1;
ac27a0ec 1593 }
26092bf5
TT		 1594 sbi->s_inode_readahead_blks = arg;
1595 } else if (token == Opt_init_itable) {
1596 set_opt(sb, INIT_INODE_TABLE);
1597 if (!args->from)
1598 arg = EXT4_DEF_LI_WAIT_MULT;
1599 sbi->s_li_wait_mult = arg;
1600 } else if (token == Opt_stripe) {
1601 sbi->s_stripe = arg;
1602 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1603 if (is_remount) {
eb513689
TT		 1604 if (!sbi->s_journal)
1605 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1606 else if (test_opt(sb, DATA_FLAGS) !=
1607 m->mount_opt) {
b31e1552 1608 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1609 "Cannot change data mode on remount");
1610 return -1;
ac27a0ec
DK		 1611 }
1612 } else {
fd8c37ec 1613 clear_opt(sb, DATA_FLAGS);
26092bf5 1614 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1615 }
ac27a0ec 1616#ifdef CONFIG_QUOTA
26092bf5 1617 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1618 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1619 sbi->s_jquota_fmt != m->mount_opt) {
1620 ext4_msg(sb, KERN_ERR, "Cannot "
1621 "change journaled quota options "
1622 "when quota turned on");
1623 return -1;
ac27a0ec 1624 }
26092bf5 1625 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1626#endif
26092bf5
TT		 1627 } else {
1628 if (!args->from)
1629 arg = 1;
1630 if (m->flags & MOPT_CLEAR)
1631 arg = !arg;
1632 else if (unlikely(!(m->flags & MOPT_SET))) {
1633 ext4_msg(sb, KERN_WARNING,
1634 "buggy handling of option %s", opt);
1635 WARN_ON(1);
1636 return -1;
3197ebdb 1637 }
26092bf5
TT		 1638 if (arg != 0)
1639 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1640 else
26092bf5 1641 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1642 }
26092bf5
TT		 1643 return 1;
1644 }
1645 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1646 "or missing value", opt);
1647 return -1;
1648}
1649
1650static int parse_options(char *options, struct super_block *sb,
1651 unsigned long *journal_devnum,
1652 unsigned int *journal_ioprio,
1653 int is_remount)
1654{
db7e5c66 1655#ifdef CONFIG_QUOTA
26092bf5 1656 struct ext4_sb_info *sbi = EXT4_SB(sb);
db7e5c66 1657#endif
26092bf5
TT		 1658 char *p;
1659 substring_t args[MAX_OPT_ARGS];
1660 int token;
1661
1662 if (!options)
1663 return 1;
1664
1665 while ((p = strsep(&options, ",")) != NULL) {
1666 if (!*p)
1667 continue;
1668 /*
1669 * Initialize args struct so we know whether arg was
1670 * found; some options take optional arguments.
1671 */
1672 args[0].to = args[0].from = 0;
1673 token = match_token(p, tokens, args);
1674 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1675 journal_ioprio, is_remount) < 0)
1676 return 0;
ac27a0ec
DK		 1677 }
1678#ifdef CONFIG_QUOTA
1679 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1680 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1681 clear_opt(sb, USRQUOTA);
ac27a0ec 1682
482a7425 1683 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1684 clear_opt(sb, GRPQUOTA);
ac27a0ec 1685
56c50f11 1686 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1687 ext4_msg(sb, KERN_ERR, "old and new quota "
1688 "format mixing");
ac27a0ec
DK		 1689 return 0;
1690 }
1691
1692 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1693 ext4_msg(sb, KERN_ERR, "journaled quota format "
1694 "not specified");
ac27a0ec
DK		 1695 return 0;
1696 }
1697 } else {
1698 if (sbi->s_jquota_fmt) {
b31e1552 1699 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1700 "specified with no journaling "
b31e1552 1701 "enabled");
ac27a0ec
DK		 1702 return 0;
1703 }
1704 }
1705#endif
1706 return 1;
1707}
1708
2adf6da8
TT		 1709static inline void ext4_show_quota_options(struct seq_file *seq,
1710 struct super_block *sb)
1711{
1712#if defined(CONFIG_QUOTA)
1713 struct ext4_sb_info *sbi = EXT4_SB(sb);
1714
1715 if (sbi->s_jquota_fmt) {
1716 char *fmtname = "";
1717
1718 switch (sbi->s_jquota_fmt) {
1719 case QFMT_VFS_OLD:
1720 fmtname = "vfsold";
1721 break;
1722 case QFMT_VFS_V0:
1723 fmtname = "vfsv0";
1724 break;
1725 case QFMT_VFS_V1:
1726 fmtname = "vfsv1";
1727 break;
1728 }
1729 seq_printf(seq, ",jqfmt=%s", fmtname);
1730 }
1731
1732 if (sbi->s_qf_names[USRQUOTA])
1733 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1734
1735 if (sbi->s_qf_names[GRPQUOTA])
1736 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1737
1738 if (test_opt(sb, USRQUOTA))
1739 seq_puts(seq, ",usrquota");
1740
1741 if (test_opt(sb, GRPQUOTA))
1742 seq_puts(seq, ",grpquota");
1743#endif
1744}
1745
5a916be1
TT		 1746static const char *token2str(int token)
1747{
1748 static const struct match_token *t;
1749
1750 for (t = tokens; t->token != Opt_err; t++)
1751 if (t->token == token && !strchr(t->pattern, '='))
1752 break;
1753 return t->pattern;
1754}
1755
2adf6da8
TT		 1756/*
1757 * Show an option if
1758 * - it's set to a non-default value OR
1759 * - if the per-sb default is different from the global default
1760 */
66acdcf4
TT		 1761static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1762 int nodefs)
2adf6da8 1763{
2adf6da8
TT		 1764 struct ext4_sb_info *sbi = EXT4_SB(sb);
1765 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1766 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1767 const struct mount_opts *m;
66acdcf4 1768 char sep = nodefs ? '\n' : ',';
2adf6da8 1769
66acdcf4
TT		 1770#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1771#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1772
1773 if (sbi->s_sb_block != 1)
5a916be1
TT		 1774 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1775
1776 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1777 int want_set = m->flags & MOPT_SET;
1778 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1779 (m->flags & MOPT_CLEAR_ERR))
1780 continue;
66acdcf4 1781 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1782 continue; /* skip if same as the default */
1783 if ((want_set &&
1784 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1785 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1786 continue; /* select Opt_noFoo vs Opt_Foo */
1787 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1788 }
5a916be1 1789
08cefc7a 1790 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1791 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1792 SEQ_OPTS_PRINT("resuid=%u",
1793 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1794 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1795 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1796 SEQ_OPTS_PRINT("resgid=%u",
1797 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1798 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1799 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1800 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1801 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1802 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1803 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1804 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1805 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1806 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1807 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1808 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1809 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1810 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1811 if (sb->s_flags & MS_I_VERSION)
5a916be1 1812 SEQ_OPTS_PUTS("i_version");
66acdcf4 1813 if (nodefs || sbi->s_stripe)
5a916be1 1814 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1815 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1816 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1817 SEQ_OPTS_PUTS("data=journal");
1818 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1819 SEQ_OPTS_PUTS("data=ordered");
1820 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1821 SEQ_OPTS_PUTS("data=writeback");
1822 }
66acdcf4
TT		 1823 if (nodefs ||
1824 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1825 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1826 sbi->s_inode_readahead_blks);
2adf6da8 1827
66acdcf4
TT		 1828 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1829 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1830 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2adf6da8
TT		 1831
1832 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1833 return 0;
1834}
1835
66acdcf4
TT		 1836static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1837{
1838 return _ext4_show_options(seq, root->d_sb, 0);
1839}
1840
1841static int options_seq_show(struct seq_file *seq, void *offset)
1842{
1843 struct super_block *sb = seq->private;
1844 int rc;
1845
1846 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1847 rc = _ext4_show_options(seq, sb, 1);
1848 seq_puts(seq, "\n");
1849 return rc;
1850}
1851
1852static int options_open_fs(struct inode *inode, struct file *file)
1853{
1854 return single_open(file, options_seq_show, PDE(inode)->data);
1855}
1856
1857static const struct file_operations ext4_seq_options_fops = {
1858 .owner = THIS_MODULE,
1859 .open = options_open_fs,
1860 .read = seq_read,
1861 .llseek = seq_lseek,
1862 .release = single_release,
1863};
1864
617ba13b 1865static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1866 int read_only)
1867{
617ba13b 1868 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1869 int res = 0;
1870
617ba13b 1871 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1872 ext4_msg(sb, KERN_ERR, "revision level too high, "
1873 "forcing read-only mode");
ac27a0ec
DK		 1874 res = MS_RDONLY;
1875 }
1876 if (read_only)
281b5995 1877 goto done;
617ba13b 1878 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1879 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1880 "running e2fsck is recommended");
617ba13b 1881 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1882 ext4_msg(sb, KERN_WARNING,
1883 "warning: mounting fs with errors, "
1884 "running e2fsck is recommended");
ed3ce80a 1885 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1886 le16_to_cpu(es->s_mnt_count) >=
1887 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1888 ext4_msg(sb, KERN_WARNING,
1889 "warning: maximal mount count reached, "
1890 "running e2fsck is recommended");
ac27a0ec
DK		 1891 else if (le32_to_cpu(es->s_checkinterval) &&
1892 (le32_to_cpu(es->s_lastcheck) +
1893 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1894 ext4_msg(sb, KERN_WARNING,
1895 "warning: checktime reached, "
1896 "running e2fsck is recommended");
0b8e58a1 1897 if (!sbi->s_journal)
0390131b 1898 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1899 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1900 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1901 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1902 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1903 ext4_update_dynamic_rev(sb);
0390131b
FM		 1904 if (sbi->s_journal)
1905 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1906
e2d67052 1907 ext4_commit_super(sb, 1);
281b5995 1908done:
ac27a0ec 1909 if (test_opt(sb, DEBUG))
a9df9a49 1910 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1911 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1912 sb->s_blocksize,
1913 sbi->s_groups_count,
617ba13b
MC		 1914 EXT4_BLOCKS_PER_GROUP(sb),
1915 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1916 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1917
7abc52c2 1918 cleancache_init_fs(sb);
ac27a0ec
DK		 1919 return res;
1920}
1921
772cb7c8
JS		 1922static int ext4_fill_flex_info(struct super_block *sb)
1923{
1924 struct ext4_sb_info *sbi = EXT4_SB(sb);
1925 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1926 ext4_group_t flex_group_count;
1927 ext4_group_t flex_group;
d50f2ab6 1928 unsigned int groups_per_flex = 0;
c5ca7c76 1929 size_t size;
772cb7c8
JS		 1930 int i;
1931
503358ae 1932 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1933 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1934 sbi->s_log_groups_per_flex = 0;
1935 return 1;
1936 }
d50f2ab6 1937 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
772cb7c8 1938
c62a11fd
FB		 1939 /* We allocate both existing and potentially added groups */
1940 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1941 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1942 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 1943 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 1944 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 1945 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 1946 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1947 flex_group_count);
1948 goto failed;
772cb7c8 1949 }
772cb7c8 1950
772cb7c8 1951 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1952 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1953
1954 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1955 atomic_add(ext4_free_inodes_count(sb, gdp),
1956 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1957 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1958 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1959 atomic_add(ext4_used_dirs_count(sb, gdp),
1960 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1961 }
1962
1963 return 1;
1964failed:
1965 return 0;
1966}
1967
feb0ab32
DW		 1968static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1969 struct ext4_group_desc *gdp)
717d50e4 1970{
feb0ab32 1971 int offset;
717d50e4 1972 __u16 crc = 0;
feb0ab32 1973 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1974
feb0ab32
DW		 1975 if ((sbi->s_es->s_feature_ro_compat &
1976 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1977 /* Use new metadata_csum algorithm */
1978 __u16 old_csum;
1979 __u32 csum32;
1980
1981 old_csum = gdp->bg_checksum;
1982 gdp->bg_checksum = 0;
1983 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1984 sizeof(le_group));
1985 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1986 sbi->s_desc_size);
1987 gdp->bg_checksum = old_csum;
1988
1989 crc = csum32 & 0xFFFF;
1990 goto out;
717d50e4
AD		 1991 }
1992
feb0ab32
DW		 1993 /* old crc16 code */
1994 offset = offsetof(struct ext4_group_desc, bg_checksum);
1995
1996 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1997 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1998 crc = crc16(crc, (__u8 *)gdp, offset);
1999 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2000 /* for checksum of struct ext4_group_desc do the rest...*/
2001 if ((sbi->s_es->s_feature_incompat &
2002 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2003 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2004 crc = crc16(crc, (__u8 *)gdp + offset,
2005 le16_to_cpu(sbi->s_es->s_desc_size) -
2006 offset);
2007
2008out:
717d50e4
AD		 2009 return cpu_to_le16(crc);
2010}
2011
feb0ab32 2012int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2013 struct ext4_group_desc *gdp)
2014{
feb0ab32
DW		 2015 if (ext4_has_group_desc_csum(sb) &&
2016 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2017 block_group, gdp)))
717d50e4
AD		 2018 return 0;
2019
2020 return 1;
2021}
2022
feb0ab32
DW		 2023void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2024 struct ext4_group_desc *gdp)
2025{
2026 if (!ext4_has_group_desc_csum(sb))
2027 return;
2028 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2029}
2030
ac27a0ec 2031/* Called at mount-time, super-block is locked */
bfff6873
LC		 2032static int ext4_check_descriptors(struct super_block *sb,
2033 ext4_group_t *first_not_zeroed)
ac27a0ec 2034{
617ba13b
MC		 2035 struct ext4_sb_info *sbi = EXT4_SB(sb);
2036 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2037 ext4_fsblk_t last_block;
bd81d8ee
LV		 2038 ext4_fsblk_t block_bitmap;
2039 ext4_fsblk_t inode_bitmap;
2040 ext4_fsblk_t inode_table;
ce421581 2041 int flexbg_flag = 0;
bfff6873 2042 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2043
ce421581
JS		 2044 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2045 flexbg_flag = 1;
2046
af5bc92d 2047 ext4_debug("Checking group descriptors");
ac27a0ec 2048
197cd65a
AM		 2049 for (i = 0; i < sbi->s_groups_count; i++) {
2050 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2051
ce421581 2052 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2053 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2054 else
2055 last_block = first_block +
617ba13b 2056 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2057
bfff6873
LC		 2058 if ((grp == sbi->s_groups_count) &&
2059 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2060 grp = i;
2061
8fadc143 2062 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2063 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2064 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2065 "Block bitmap for group %u not in group "
b31e1552 2066 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2067 return 0;
2068 }
8fadc143 2069 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2070 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2071 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2072 "Inode bitmap for group %u not in group "
b31e1552 2073 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2074 return 0;
2075 }
8fadc143 2076 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2077 if (inode_table < first_block ||
2b2d6d01 2078 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2079 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2080 "Inode table for group %u not in group "
b31e1552 2081 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2082 return 0;
2083 }
955ce5f5 2084 ext4_lock_group(sb, i);
feb0ab32 2085 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2086 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2087 "Checksum for group %u failed (%u!=%u)",
2088 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2089 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2090 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2091 ext4_unlock_group(sb, i);
8a266467 2092 return 0;
7ee1ec4c 2093 }
717d50e4 2094 }
955ce5f5 2095 ext4_unlock_group(sb, i);
ce421581
JS		 2096 if (!flexbg_flag)
2097 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2098 }
bfff6873
LC		 2099 if (NULL != first_not_zeroed)
2100 *first_not_zeroed = grp;
ac27a0ec 2101
5dee5437
TT		 2102 ext4_free_blocks_count_set(sbi->s_es,
2103 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2104 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2105 return 1;
2106}
2107
617ba13b 2108/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2109 * the superblock) which were deleted from all directories, but held open by
2110 * a process at the time of a crash. We walk the list and try to delete these
2111 * inodes at recovery time (only with a read-write filesystem).
2112 *
2113 * In order to keep the orphan inode chain consistent during traversal (in
2114 * case of crash during recovery), we link each inode into the superblock
2115 * orphan list_head and handle it the same way as an inode deletion during
2116 * normal operation (which journals the operations for us).
2117 *
2118 * We only do an iget() and an iput() on each inode, which is very safe if we
2119 * accidentally point at an in-use or already deleted inode. The worst that
2120 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2121 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2122 * e2fsck was run on this filesystem, and it must have already done the orphan
2123 * inode cleanup for us, so we can safely abort without any further action.
2124 */
2b2d6d01
TT		 2125static void ext4_orphan_cleanup(struct super_block *sb,
2126 struct ext4_super_block *es)
ac27a0ec
DK		 2127{
2128 unsigned int s_flags = sb->s_flags;
2129 int nr_orphans = 0, nr_truncates = 0;
2130#ifdef CONFIG_QUOTA
2131 int i;
2132#endif
2133 if (!es->s_last_orphan) {
2134 jbd_debug(4, "no orphan inodes to clean up\n");
2135 return;
2136 }
2137
a8f48a95 2138 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2139 ext4_msg(sb, KERN_ERR, "write access "
2140 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2141 return;
2142 }
2143
d39195c3
AG		 2144 /* Check if feature set would not allow a r/w mount */
2145 if (!ext4_feature_set_ok(sb, 0)) {
2146 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2147 "unknown ROCOMPAT features");
2148 return;
2149 }
2150
617ba13b 2151 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2152 if (es->s_last_orphan)
2153 jbd_debug(1, "Errors on filesystem, "
2154 "clearing orphan list.\n");
2155 es->s_last_orphan = 0;
2156 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2157 return;
2158 }
2159
2160 if (s_flags & MS_RDONLY) {
b31e1552 2161 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2162 sb->s_flags &= ~MS_RDONLY;
2163 }
2164#ifdef CONFIG_QUOTA
2165 /* Needed for iput() to work correctly and not trash data */
2166 sb->s_flags |= MS_ACTIVE;
2167 /* Turn on quotas so that they are updated correctly */
2168 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2169 if (EXT4_SB(sb)->s_qf_names[i]) {
2170 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2171 if (ret < 0)
b31e1552
ES		 2172 ext4_msg(sb, KERN_ERR,
2173 "Cannot turn on journaled "
2174 "quota: error %d", ret);
ac27a0ec
DK		 2175 }
2176 }
2177#endif
2178
2179 while (es->s_last_orphan) {
2180 struct inode *inode;
2181
97bd42b9
JB		 2182 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2183 if (IS_ERR(inode)) {
ac27a0ec
DK		 2184 es->s_last_orphan = 0;
2185 break;
2186 }
2187
617ba13b 2188 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2189 dquot_initialize(inode);
ac27a0ec 2190 if (inode->i_nlink) {
b31e1552
ES		 2191 ext4_msg(sb, KERN_DEBUG,
2192 "%s: truncating inode %lu to %lld bytes",
46e665e9 2193 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2194 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2195 inode->i_ino, inode->i_size);
617ba13b 2196 ext4_truncate(inode);
ac27a0ec
DK		 2197 nr_truncates++;
2198 } else {
b31e1552
ES		 2199 ext4_msg(sb, KERN_DEBUG,
2200 "%s: deleting unreferenced inode %lu",
46e665e9 2201 __func__, inode->i_ino);
ac27a0ec
DK		 2202 jbd_debug(2, "deleting unreferenced inode %lu\n",
2203 inode->i_ino);
2204 nr_orphans++;
2205 }
2206 iput(inode); /* The delete magic happens here! */
2207 }
2208
2b2d6d01 2209#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2210
2211 if (nr_orphans)
b31e1552
ES		 2212 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2213 PLURAL(nr_orphans));
ac27a0ec 2214 if (nr_truncates)
b31e1552
ES		 2215 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2216 PLURAL(nr_truncates));
ac27a0ec
DK		 2217#ifdef CONFIG_QUOTA
2218 /* Turn quotas off */
2219 for (i = 0; i < MAXQUOTAS; i++) {
2220 if (sb_dqopt(sb)->files[i])
287a8095 2221 dquot_quota_off(sb, i);
ac27a0ec
DK		 2222 }
2223#endif
2224 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2225}
0b8e58a1 2226
cd2291a4
ES		 2227/*
2228 * Maximal extent format file size.
2229 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2230 * extent format containers, within a sector_t, and within i_blocks
2231 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2232 * so that won't be a limiting factor.
2233 *
f17722f9
LC		 2234 * However there is other limiting factor. We do store extents in the form
2235 * of starting block and length, hence the resulting length of the extent
2236 * covering maximum file size must fit into on-disk format containers as
2237 * well. Given that length is always by 1 unit bigger than max unit (because
2238 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2239 *
cd2291a4
ES		 2240 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2241 */
f287a1a5 2242static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2243{
2244 loff_t res;
2245 loff_t upper_limit = MAX_LFS_FILESIZE;
2246
2247 /* small i_blocks in vfs inode? */
f287a1a5 2248 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2249 /*
90c699a9 2250 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2251 * i_block represent total blocks in 512 bytes
2252 * 32 == size of vfs inode i_blocks * 8
2253 */
2254 upper_limit = (1LL << 32) - 1;
2255
2256 /* total blocks in file system block size */
2257 upper_limit >>= (blkbits - 9);
2258 upper_limit <<= blkbits;
2259 }
2260
f17722f9
LC		 2261 /*
2262 * 32-bit extent-start container, ee_block. We lower the maxbytes
2263 * by one fs block, so ee_len can cover the extent of maximum file
2264 * size
2265 */
2266 res = (1LL << 32) - 1;
cd2291a4 2267 res <<= blkbits;
cd2291a4
ES		 2268
2269 /* Sanity check against vm- & vfs- imposed limits */
2270 if (res > upper_limit)
2271 res = upper_limit;
2272
2273 return res;
2274}
ac27a0ec 2275
ac27a0ec 2276/*
cd2291a4 2277 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2278 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2279 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2280 */
f287a1a5 2281static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2282{
617ba13b 2283 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2284 int meta_blocks;
2285 loff_t upper_limit;
0b8e58a1
AD		 2286 /* This is calculated to be the largest file size for a dense, block
2287 * mapped file such that the file's total number of 512-byte sectors,
2288 * including data and all indirect blocks, does not exceed (2^48 - 1).
2289 *
2290 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2291 * number of 512-byte sectors of the file.
0fc1b451
AK		 2292 */
2293
f287a1a5 2294 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2295 /*
90c699a9 2296 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2297 * the inode i_block field represents total file blocks in
2298 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2299 */
2300 upper_limit = (1LL << 32) - 1;
2301
2302 /* total blocks in file system block size */
2303 upper_limit >>= (bits - 9);
2304
2305 } else {
8180a562
AK		 2306 /*
2307 * We use 48 bit ext4_inode i_blocks
2308 * With EXT4_HUGE_FILE_FL set the i_blocks
2309 * represent total number of blocks in
2310 * file system block size
2311 */
0fc1b451
AK		 2312 upper_limit = (1LL << 48) - 1;
2313
0fc1b451
AK		 2314 }
2315
2316 /* indirect blocks */
2317 meta_blocks = 1;
2318 /* double indirect blocks */
2319 meta_blocks += 1 + (1LL << (bits-2));
2320 /* tripple indirect blocks */
2321 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2322
2323 upper_limit -= meta_blocks;
2324 upper_limit <<= bits;
ac27a0ec
DK		 2325
2326 res += 1LL << (bits-2);
2327 res += 1LL << (2*(bits-2));
2328 res += 1LL << (3*(bits-2));
2329 res <<= bits;
2330 if (res > upper_limit)
2331 res = upper_limit;
0fc1b451
AK		 2332
2333 if (res > MAX_LFS_FILESIZE)
2334 res = MAX_LFS_FILESIZE;
2335
ac27a0ec
DK		 2336 return res;
2337}
2338
617ba13b 2339static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2340 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2341{
617ba13b 2342 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2343 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2344 int has_super = 0;
2345
2346 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2347
617ba13b 2348 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2349 nr < first_meta_bg)
70bbb3e0 2350 return logical_sb_block + nr + 1;
ac27a0ec 2351 bg = sbi->s_desc_per_block * nr;
617ba13b 2352 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2353 has_super = 1;
0b8e58a1 2354
617ba13b 2355 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2356}
2357
c9de560d
AT		 2358/**
2359 * ext4_get_stripe_size: Get the stripe size.
2360 * @sbi: In memory super block info
2361 *
2362 * If we have specified it via mount option, then
2363 * use the mount option value. If the value specified at mount time is
2364 * greater than the blocks per group use the super block value.
2365 * If the super block value is greater than blocks per group return 0.
2366 * Allocator needs it be less than blocks per group.
2367 *
2368 */
2369static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2370{
2371 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2372 unsigned long stripe_width =
2373 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2374 int ret;
c9de560d
AT		 2375
2376 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2377 ret = sbi->s_stripe;
2378 else if (stripe_width <= sbi->s_blocks_per_group)
2379 ret = stripe_width;
2380 else if (stride <= sbi->s_blocks_per_group)
2381 ret = stride;
2382 else
2383 ret = 0;
c9de560d 2384
3eb08658
DE		 2385 /*
2386 * If the stripe width is 1, this makes no sense and
2387 * we set it to 0 to turn off stripe handling code.
2388 */
2389 if (ret <= 1)
2390 ret = 0;
c9de560d 2391
3eb08658 2392 return ret;
c9de560d 2393}
ac27a0ec 2394
3197ebdb
TT		 2395/* sysfs supprt */
2396
2397struct ext4_attr {
2398 struct attribute attr;
2399 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2400 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2401 const char *, size_t);
2402 int offset;
2403};
2404
2405static int parse_strtoul(const char *buf,
2406 unsigned long max, unsigned long *value)
2407{
2408 char *endp;
2409
e7d2860b
AGR		 2410 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2411 endp = skip_spaces(endp);
3197ebdb
TT		 2412 if (*endp || *value > max)
2413 return -EINVAL;
2414
2415 return 0;
2416}
2417
2418static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2419 struct ext4_sb_info *sbi,
2420 char *buf)
2421{
2422 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2423 (s64) EXT4_C2B(sbi,
2424 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2425}
2426
2427static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2428 struct ext4_sb_info *sbi, char *buf)
2429{
2430 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2431
f613dfcb
TT		 2432 if (!sb->s_bdev->bd_part)
2433 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2434 return snprintf(buf, PAGE_SIZE, "%lu\n",
2435 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2436 sbi->s_sectors_written_start) >> 1);
2437}
2438
2439static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2440 struct ext4_sb_info *sbi, char *buf)
2441{
2442 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2443
f613dfcb
TT		 2444 if (!sb->s_bdev->bd_part)
2445 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2446 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2447 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2448 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2449 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2450}
2451
2452static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2453 struct ext4_sb_info *sbi,
2454 const char *buf, size_t count)
2455{
2456 unsigned long t;
2457
2458 if (parse_strtoul(buf, 0x40000000, &t))
2459 return -EINVAL;
2460
5dbd571d 2461 if (t && !is_power_of_2(t))
3197ebdb
TT		 2462 return -EINVAL;
2463
2464 sbi->s_inode_readahead_blks = t;
2465 return count;
2466}
2467
2468static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2469 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2470{
2471 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2472
2473 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2474}
2475
2476static ssize_t sbi_ui_store(struct ext4_attr *a,
2477 struct ext4_sb_info *sbi,
2478 const char *buf, size_t count)
2479{
2480 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2481 unsigned long t;
2482
2483 if (parse_strtoul(buf, 0xffffffff, &t))
2484 return -EINVAL;
2485 *ui = t;
2486 return count;
2487}
2488
2c0544b2
TT		 2489static ssize_t trigger_test_error(struct ext4_attr *a,
2490 struct ext4_sb_info *sbi,
2491 const char *buf, size_t count)
2492{
2493 int len = count;
2494
2495 if (!capable(CAP_SYS_ADMIN))
2496 return -EPERM;
2497
2498 if (len && buf[len-1] == '\n')
2499 len--;
2500
2501 if (len)
2502 ext4_error(sbi->s_sb, "%.*s", len, buf);
2503 return count;
2504}
2505
3197ebdb
TT		 2506#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2507static struct ext4_attr ext4_attr_##_name = { \
2508 .attr = {.name = __stringify(_name), .mode = _mode }, \
2509 .show = _show, \
2510 .store = _store, \
2511 .offset = offsetof(struct ext4_sb_info, _elname), \
2512}
2513#define EXT4_ATTR(name, mode, show, store) \
2514static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2515
857ac889 2516#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2517#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2518#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2519#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2520 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2521#define ATTR_LIST(name) &ext4_attr_##name.attr
2522
2523EXT4_RO_ATTR(delayed_allocation_blocks);
2524EXT4_RO_ATTR(session_write_kbytes);
2525EXT4_RO_ATTR(lifetime_write_kbytes);
2526EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2527 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2528EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2529EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2530EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2531EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2532EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2533EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2534EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2535EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2c0544b2 2536EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2537
2538static struct attribute *ext4_attrs[] = {
2539 ATTR_LIST(delayed_allocation_blocks),
2540 ATTR_LIST(session_write_kbytes),
2541 ATTR_LIST(lifetime_write_kbytes),
2542 ATTR_LIST(inode_readahead_blks),
11013911 2543 ATTR_LIST(inode_goal),
3197ebdb
TT		 2544 ATTR_LIST(mb_stats),
2545 ATTR_LIST(mb_max_to_scan),
2546 ATTR_LIST(mb_min_to_scan),
2547 ATTR_LIST(mb_order2_req),
2548 ATTR_LIST(mb_stream_req),
2549 ATTR_LIST(mb_group_prealloc),
55138e0b 2550 ATTR_LIST(max_writeback_mb_bump),
2c0544b2 2551 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2552 NULL,
2553};
2554
857ac889
LC		 2555/* Features this copy of ext4 supports */
2556EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2557EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2558
2559static struct attribute *ext4_feat_attrs[] = {
2560 ATTR_LIST(lazy_itable_init),
27ee40df 2561 ATTR_LIST(batched_discard),
857ac889
LC		 2562 NULL,
2563};
2564
3197ebdb
TT		 2565static ssize_t ext4_attr_show(struct kobject *kobj,
2566 struct attribute *attr, char *buf)
2567{
2568 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2569 s_kobj);
2570 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2571
2572 return a->show ? a->show(a, sbi, buf) : 0;
2573}
2574
2575static ssize_t ext4_attr_store(struct kobject *kobj,
2576 struct attribute *attr,
2577 const char *buf, size_t len)
2578{
2579 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2580 s_kobj);
2581 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2582
2583 return a->store ? a->store(a, sbi, buf, len) : 0;
2584}
2585
2586static void ext4_sb_release(struct kobject *kobj)
2587{
2588 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2589 s_kobj);
2590 complete(&sbi->s_kobj_unregister);
2591}
2592
52cf25d0 2593static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2594 .show = ext4_attr_show,
2595 .store = ext4_attr_store,
2596};
2597
2598static struct kobj_type ext4_ktype = {
2599 .default_attrs = ext4_attrs,
2600 .sysfs_ops = &ext4_attr_ops,
2601 .release = ext4_sb_release,
2602};
2603
857ac889
LC		 2604static void ext4_feat_release(struct kobject *kobj)
2605{
2606 complete(&ext4_feat->f_kobj_unregister);
2607}
2608
2609static struct kobj_type ext4_feat_ktype = {
2610 .default_attrs = ext4_feat_attrs,
2611 .sysfs_ops = &ext4_attr_ops,
2612 .release = ext4_feat_release,
2613};
2614
a13fb1a4
ES		 2615/*
2616 * Check whether this filesystem can be mounted based on
2617 * the features present and the RDONLY/RDWR mount requested.
2618 * Returns 1 if this filesystem can be mounted as requested,
2619 * 0 if it cannot be.
2620 */
2621static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2622{
2623 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2624 ext4_msg(sb, KERN_ERR,
2625 "Couldn't mount because of "
2626 "unsupported optional features (%x)",
2627 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2628 ~EXT4_FEATURE_INCOMPAT_SUPP));
2629 return 0;
2630 }
2631
2632 if (readonly)
2633 return 1;
2634
2635 /* Check that feature set is OK for a read-write mount */
2636 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2637 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2638 "unsupported optional features (%x)",
2639 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2640 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2641 return 0;
2642 }
2643 /*
2644 * Large file size enabled file system can only be mounted
2645 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2646 */
2647 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2648 if (sizeof(blkcnt_t) < sizeof(u64)) {
2649 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2650 "cannot be mounted RDWR without "
2651 "CONFIG_LBDAF");
2652 return 0;
2653 }
2654 }
bab08ab9
TT		 2655 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2656 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2657 ext4_msg(sb, KERN_ERR,
2658 "Can't support bigalloc feature without "
2659 "extents feature\n");
2660 return 0;
2661 }
7c319d32
AK		 2662
2663#ifndef CONFIG_QUOTA
2664 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2665 !readonly) {
2666 ext4_msg(sb, KERN_ERR,
2667 "Filesystem with quota feature cannot be mounted RDWR "
2668 "without CONFIG_QUOTA");
2669 return 0;
2670 }
2671#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2672 return 1;
2673}
2674
66e61a9e
TT		 2675/*
2676 * This function is called once a day if we have errors logged
2677 * on the file system
2678 */
2679static void print_daily_error_info(unsigned long arg)
2680{
2681 struct super_block *sb = (struct super_block *) arg;
2682 struct ext4_sb_info *sbi;
2683 struct ext4_super_block *es;
2684
2685 sbi = EXT4_SB(sb);
2686 es = sbi->s_es;
2687
2688 if (es->s_error_count)
2689 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2690 le32_to_cpu(es->s_error_count));
2691 if (es->s_first_error_time) {
2692 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2693 sb->s_id, le32_to_cpu(es->s_first_error_time),
2694 (int) sizeof(es->s_first_error_func),
2695 es->s_first_error_func,
2696 le32_to_cpu(es->s_first_error_line));
2697 if (es->s_first_error_ino)
2698 printk(": inode %u",
2699 le32_to_cpu(es->s_first_error_ino));
2700 if (es->s_first_error_block)
2701 printk(": block %llu", (unsigned long long)
2702 le64_to_cpu(es->s_first_error_block));
2703 printk("\n");
2704 }
2705 if (es->s_last_error_time) {
2706 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2707 sb->s_id, le32_to_cpu(es->s_last_error_time),
2708 (int) sizeof(es->s_last_error_func),
2709 es->s_last_error_func,
2710 le32_to_cpu(es->s_last_error_line));
2711 if (es->s_last_error_ino)
2712 printk(": inode %u",
2713 le32_to_cpu(es->s_last_error_ino));
2714 if (es->s_last_error_block)
2715 printk(": block %llu", (unsigned long long)
2716 le64_to_cpu(es->s_last_error_block));
2717 printk("\n");
2718 }
2719 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2720}
2721
bfff6873
LC		 2722/* Find next suitable group and run ext4_init_inode_table */
2723static int ext4_run_li_request(struct ext4_li_request *elr)
2724{
2725 struct ext4_group_desc *gdp = NULL;
2726 ext4_group_t group, ngroups;
2727 struct super_block *sb;
2728 unsigned long timeout = 0;
2729 int ret = 0;
2730
2731 sb = elr->lr_super;
2732 ngroups = EXT4_SB(sb)->s_groups_count;
2733
8e8ad8a5 2734 sb_start_write(sb);
bfff6873
LC		 2735 for (group = elr->lr_next_group; group < ngroups; group++) {
2736 gdp = ext4_get_group_desc(sb, group, NULL);
2737 if (!gdp) {
2738 ret = 1;
2739 break;
2740 }
2741
2742 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2743 break;
2744 }
2745
2746 if (group == ngroups)
2747 ret = 1;
2748
2749 if (!ret) {
2750 timeout = jiffies;
2751 ret = ext4_init_inode_table(sb, group,
2752 elr->lr_timeout ? 0 : 1);
2753 if (elr->lr_timeout == 0) {
51ce6511
LC		 2754 timeout = (jiffies - timeout) *
2755 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2756 elr->lr_timeout = timeout;
2757 }
2758 elr->lr_next_sched = jiffies + elr->lr_timeout;
2759 elr->lr_next_group = group + 1;
2760 }
8e8ad8a5 2761 sb_end_write(sb);
bfff6873
LC		 2762
2763 return ret;
2764}
2765
2766/*
2767 * Remove lr_request from the list_request and free the
4ed5c033 2768 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2769 */
2770static void ext4_remove_li_request(struct ext4_li_request *elr)
2771{
2772 struct ext4_sb_info *sbi;
2773
2774 if (!elr)
2775 return;
2776
2777 sbi = elr->lr_sbi;
2778
2779 list_del(&elr->lr_request);
2780 sbi->s_li_request = NULL;
2781 kfree(elr);
2782}
2783
2784static void ext4_unregister_li_request(struct super_block *sb)
2785{
1bb933fb
LC		 2786 mutex_lock(&ext4_li_mtx);
2787 if (!ext4_li_info) {
2788 mutex_unlock(&ext4_li_mtx);
bfff6873 2789 return;
1bb933fb 2790 }
bfff6873
LC		 2791
2792 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2793 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2794 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2795 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2796}
2797
8f1f7453
ES		 2798static struct task_struct *ext4_lazyinit_task;
2799
bfff6873
LC		 2800/*
2801 * This is the function where ext4lazyinit thread lives. It walks
2802 * through the request list searching for next scheduled filesystem.
2803 * When such a fs is found, run the lazy initialization request
2804 * (ext4_rn_li_request) and keep track of the time spend in this
2805 * function. Based on that time we compute next schedule time of
2806 * the request. When walking through the list is complete, compute
2807 * next waking time and put itself into sleep.
2808 */
2809static int ext4_lazyinit_thread(void *arg)
2810{
2811 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2812 struct list_head *pos, *n;
2813 struct ext4_li_request *elr;
4ed5c033 2814 unsigned long next_wakeup, cur;
bfff6873
LC		 2815
2816 BUG_ON(NULL == eli);
2817
bfff6873
LC		 2818cont_thread:
2819 while (true) {
2820 next_wakeup = MAX_JIFFY_OFFSET;
2821
2822 mutex_lock(&eli->li_list_mtx);
2823 if (list_empty(&eli->li_request_list)) {
2824 mutex_unlock(&eli->li_list_mtx);
2825 goto exit_thread;
2826 }
2827
2828 list_for_each_safe(pos, n, &eli->li_request_list) {
2829 elr = list_entry(pos, struct ext4_li_request,
2830 lr_request);
2831
b2c78cd0
TT		 2832 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2833 if (ext4_run_li_request(elr) != 0) {
2834 /* error, remove the lazy_init job */
2835 ext4_remove_li_request(elr);
2836 continue;
2837 }
bfff6873
LC		 2838 }
2839
2840 if (time_before(elr->lr_next_sched, next_wakeup))
2841 next_wakeup = elr->lr_next_sched;
2842 }
2843 mutex_unlock(&eli->li_list_mtx);
2844
a0acae0e 2845 try_to_freeze();
bfff6873 2846
4ed5c033
LC		 2847 cur = jiffies;
2848 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2849 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2850 cond_resched();
2851 continue;
2852 }
2853
4ed5c033
LC		 2854 schedule_timeout_interruptible(next_wakeup - cur);
2855
8f1f7453
ES		 2856 if (kthread_should_stop()) {
2857 ext4_clear_request_list();
2858 goto exit_thread;
2859 }
bfff6873
LC		 2860 }
2861
2862exit_thread:
2863 /*
2864 * It looks like the request list is empty, but we need
2865 * to check it under the li_list_mtx lock, to prevent any
2866 * additions into it, and of course we should lock ext4_li_mtx
2867 * to atomically free the list and ext4_li_info, because at
2868 * this point another ext4 filesystem could be registering
2869 * new one.
2870 */
2871 mutex_lock(&ext4_li_mtx);
2872 mutex_lock(&eli->li_list_mtx);
2873 if (!list_empty(&eli->li_request_list)) {
2874 mutex_unlock(&eli->li_list_mtx);
2875 mutex_unlock(&ext4_li_mtx);
2876 goto cont_thread;
2877 }
2878 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2879 kfree(ext4_li_info);
2880 ext4_li_info = NULL;
2881 mutex_unlock(&ext4_li_mtx);
2882
2883 return 0;
2884}
2885
2886static void ext4_clear_request_list(void)
2887{
2888 struct list_head *pos, *n;
2889 struct ext4_li_request *elr;
2890
2891 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2892 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2893 elr = list_entry(pos, struct ext4_li_request,
2894 lr_request);
2895 ext4_remove_li_request(elr);
2896 }
2897 mutex_unlock(&ext4_li_info->li_list_mtx);
2898}
2899
2900static int ext4_run_lazyinit_thread(void)
2901{
8f1f7453
ES		 2902 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2903 ext4_li_info, "ext4lazyinit");
2904 if (IS_ERR(ext4_lazyinit_task)) {
2905 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2906 ext4_clear_request_list();
bfff6873
LC		 2907 kfree(ext4_li_info);
2908 ext4_li_info = NULL;
92b97816 2909 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2910 "initialization thread\n",
2911 err);
2912 return err;
2913 }
2914 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2915 return 0;
2916}
2917
2918/*
2919 * Check whether it make sense to run itable init. thread or not.
2920 * If there is at least one uninitialized inode table, return
2921 * corresponding group number, else the loop goes through all
2922 * groups and return total number of groups.
2923 */
2924static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2925{
2926 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2927 struct ext4_group_desc *gdp = NULL;
2928
2929 for (group = 0; group < ngroups; group++) {
2930 gdp = ext4_get_group_desc(sb, group, NULL);
2931 if (!gdp)
2932 continue;
2933
2934 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2935 break;
2936 }
2937
2938 return group;
2939}
2940
2941static int ext4_li_info_new(void)
2942{
2943 struct ext4_lazy_init *eli = NULL;
2944
2945 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2946 if (!eli)
2947 return -ENOMEM;
2948
bfff6873
LC		 2949 INIT_LIST_HEAD(&eli->li_request_list);
2950 mutex_init(&eli->li_list_mtx);
2951
bfff6873
LC		 2952 eli->li_state |= EXT4_LAZYINIT_QUIT;
2953
2954 ext4_li_info = eli;
2955
2956 return 0;
2957}
2958
2959static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2960 ext4_group_t start)
2961{
2962 struct ext4_sb_info *sbi = EXT4_SB(sb);
2963 struct ext4_li_request *elr;
2964 unsigned long rnd;
2965
2966 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2967 if (!elr)
2968 return NULL;
2969
2970 elr->lr_super = sb;
2971 elr->lr_sbi = sbi;
2972 elr->lr_next_group = start;
2973
2974 /*
2975 * Randomize first schedule time of the request to
2976 * spread the inode table initialization requests
2977 * better.
2978 */
2979 get_random_bytes(&rnd, sizeof(rnd));
2980 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2981 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2982
2983 return elr;
2984}
2985
2986static int ext4_register_li_request(struct super_block *sb,
2987 ext4_group_t first_not_zeroed)
2988{
2989 struct ext4_sb_info *sbi = EXT4_SB(sb);
2990 struct ext4_li_request *elr;
2991 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2992 int ret = 0;
bfff6873 2993
51ce6511
LC		 2994 if (sbi->s_li_request != NULL) {
2995 /*
2996 * Reset timeout so it can be computed again, because
2997 * s_li_wait_mult might have changed.
2998 */
2999 sbi->s_li_request->lr_timeout = 0;
beed5ecb 3000 return 0;
51ce6511 3001 }
bfff6873
LC		 3002
3003 if (first_not_zeroed == ngroups ||
3004 (sb->s_flags & MS_RDONLY) ||
55ff3840 3005 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 3006 return 0;
bfff6873
LC		 3007
3008 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 3009 if (!elr)
3010 return -ENOMEM;
bfff6873
LC		 3011
3012 mutex_lock(&ext4_li_mtx);
3013
3014 if (NULL == ext4_li_info) {
3015 ret = ext4_li_info_new();
3016 if (ret)
3017 goto out;
3018 }
3019
3020 mutex_lock(&ext4_li_info->li_list_mtx);
3021 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3022 mutex_unlock(&ext4_li_info->li_list_mtx);
3023
3024 sbi->s_li_request = elr;
46e4690b
TM		 3025 /*
3026 * set elr to NULL here since it has been inserted to
3027 * the request_list and the removal and free of it is
3028 * handled by ext4_clear_request_list from now on.
3029 */
3030 elr = NULL;
bfff6873
LC		 3031
3032 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3033 ret = ext4_run_lazyinit_thread();
3034 if (ret)
3035 goto out;
3036 }
bfff6873 3037out:
beed5ecb
NK		 3038 mutex_unlock(&ext4_li_mtx);
3039 if (ret)
bfff6873 3040 kfree(elr);
bfff6873
LC		 3041 return ret;
3042}
3043
3044/*
3045 * We do not need to lock anything since this is called on
3046 * module unload.
3047 */
3048static void ext4_destroy_lazyinit_thread(void)
3049{
3050 /*
3051 * If thread exited earlier
3052 * there's nothing to be done.
3053 */
8f1f7453 3054 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3055 return;
3056
8f1f7453 3057 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3058}
3059
25ed6e8a
DW		 3060static int set_journal_csum_feature_set(struct super_block *sb)
3061{
3062 int ret = 1;
3063 int compat, incompat;
3064 struct ext4_sb_info *sbi = EXT4_SB(sb);
3065
3066 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3067 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3068 /* journal checksum v2 */
3069 compat = 0;
3070 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3071 } else {
3072 /* journal checksum v1 */
3073 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3074 incompat = 0;
3075 }
3076
3077 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3078 ret = jbd2_journal_set_features(sbi->s_journal,
3079 compat, 0,
3080 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3081 incompat);
3082 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3083 ret = jbd2_journal_set_features(sbi->s_journal,
3084 compat, 0,
3085 incompat);
3086 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3087 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3088 } else {
3089 jbd2_journal_clear_features(sbi->s_journal,
3090 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3091 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3092 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3093 }
3094
3095 return ret;
3096}
3097
952fc18e
TT		 3098/*
3099 * Note: calculating the overhead so we can be compatible with
3100 * historical BSD practice is quite difficult in the face of
3101 * clusters/bigalloc. This is because multiple metadata blocks from
3102 * different block group can end up in the same allocation cluster.
3103 * Calculating the exact overhead in the face of clustered allocation
3104 * requires either O(all block bitmaps) in memory or O(number of block
3105 * groups**2) in time. We will still calculate the superblock for
3106 * older file systems --- and if we come across with a bigalloc file
3107 * system with zero in s_overhead_clusters the estimate will be close to
3108 * correct especially for very large cluster sizes --- but for newer
3109 * file systems, it's better to calculate this figure once at mkfs
3110 * time, and store it in the superblock. If the superblock value is
3111 * present (even for non-bigalloc file systems), we will use it.
3112 */
3113static int count_overhead(struct super_block *sb, ext4_group_t grp,
3114 char *buf)
3115{
3116 struct ext4_sb_info *sbi = EXT4_SB(sb);
3117 struct ext4_group_desc *gdp;
3118 ext4_fsblk_t first_block, last_block, b;
3119 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3120 int s, j, count = 0;
3121
3122 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3123 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3124 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3125 for (i = 0; i < ngroups; i++) {
3126 gdp = ext4_get_group_desc(sb, i, NULL);
3127 b = ext4_block_bitmap(sb, gdp);
3128 if (b >= first_block && b <= last_block) {
3129 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3130 count++;
3131 }
3132 b = ext4_inode_bitmap(sb, gdp);
3133 if (b >= first_block && b <= last_block) {
3134 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3135 count++;
3136 }
3137 b = ext4_inode_table(sb, gdp);
3138 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3139 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3140 int c = EXT4_B2C(sbi, b - first_block);
3141 ext4_set_bit(c, buf);
3142 count++;
3143 }
3144 if (i != grp)
3145 continue;
3146 s = 0;
3147 if (ext4_bg_has_super(sb, grp)) {
3148 ext4_set_bit(s++, buf);
3149 count++;
3150 }
3151 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3152 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3153 count++;
3154 }
3155 }
3156 if (!count)
3157 return 0;
3158 return EXT4_CLUSTERS_PER_GROUP(sb) -
3159 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3160}
3161
3162/*
3163 * Compute the overhead and stash it in sbi->s_overhead
3164 */
3165int ext4_calculate_overhead(struct super_block *sb)
3166{
3167 struct ext4_sb_info *sbi = EXT4_SB(sb);
3168 struct ext4_super_block *es = sbi->s_es;
3169 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3170 ext4_fsblk_t overhead = 0;
3171 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3172
3173 memset(buf, 0, PAGE_SIZE);
3174 if (!buf)
3175 return -ENOMEM;
3176
3177 /*
3178 * Compute the overhead (FS structures). This is constant
3179 * for a given filesystem unless the number of block groups
3180 * changes so we cache the previous value until it does.
3181 */
3182
3183 /*
3184 * All of the blocks before first_data_block are overhead
3185 */
3186 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3187
3188 /*
3189 * Add the overhead found in each block group
3190 */
3191 for (i = 0; i < ngroups; i++) {
3192 int blks;
3193
3194 blks = count_overhead(sb, i, buf);
3195 overhead += blks;
3196 if (blks)
3197 memset(buf, 0, PAGE_SIZE);
3198 cond_resched();
3199 }
3200 sbi->s_overhead = overhead;
3201 smp_wmb();
3202 free_page((unsigned long) buf);
3203 return 0;
3204}
3205
2b2d6d01 3206static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3207{
d4c402d9 3208 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3209 struct buffer_head *bh;
617ba13b
MC		 3210 struct ext4_super_block *es = NULL;
3211 struct ext4_sb_info *sbi;
3212 ext4_fsblk_t block;
3213 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3214 ext4_fsblk_t logical_sb_block;
ac27a0ec 3215 unsigned long offset = 0;
ac27a0ec
DK		 3216 unsigned long journal_devnum = 0;
3217 unsigned long def_mount_opts;
3218 struct inode *root;
9f6200bb 3219 char *cp;
0390131b 3220 const char *descr;
dcc7dae3 3221 int ret = -ENOMEM;
281b5995 3222 int blocksize, clustersize;
4ec11028
TT		 3223 unsigned int db_count;
3224 unsigned int i;
281b5995 3225 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3226 __u64 blocks_count;
833f4077 3227 int err;
b3881f74 3228 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3229 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3230
3231 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3232 if (!sbi)
dcc7dae3 3233 goto out_free_orig;
705895b6
PE		 3234
3235 sbi->s_blockgroup_lock =
3236 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3237 if (!sbi->s_blockgroup_lock) {
3238 kfree(sbi);
dcc7dae3 3239 goto out_free_orig;
705895b6 3240 }
ac27a0ec 3241 sb->s_fs_info = sbi;
2c0544b2 3242 sbi->s_sb = sb;
ac27a0ec 3243 sbi->s_mount_opt = 0;
08cefc7a
EB		 3244 sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3245 sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
240799cd 3246 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3247 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3248 if (sb->s_bdev->bd_part)
3249 sbi->s_sectors_written_start =
3250 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3251
9f6200bb
TT		 3252 /* Cleanup superblock name */
3253 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3254 *cp = '!';
3255
dcc7dae3 3256 ret = -EINVAL;
617ba13b 3257 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3258 if (!blocksize) {
b31e1552 3259 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3260 goto out_fail;
3261 }
3262
3263 /*
617ba13b 3264 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3265 * block sizes. We need to calculate the offset from buffer start.
3266 */
617ba13b 3267 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3268 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3269 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3270 } else {
70bbb3e0 3271 logical_sb_block = sb_block;
ac27a0ec
DK		 3272 }
3273
70bbb3e0 3274 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3275 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3276 goto out_fail;
3277 }
3278 /*
3279 * Note: s_es must be initialized as soon as possible because
617ba13b 3280 * some ext4 macro-instructions depend on its value
ac27a0ec 3281 */
2716b802 3282 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3283 sbi->s_es = es;
3284 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3285 if (sb->s_magic != EXT4_SUPER_MAGIC)
3286 goto cantfind_ext4;
afc32f7e 3287 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3288
feb0ab32
DW		 3289 /* Warn if metadata_csum and gdt_csum are both set. */
3290 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3291 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3292 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3293 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3294 "redundant flags; please run fsck.");
3295
d25425f8
DW		 3296 /* Check for a known checksum algorithm */
3297 if (!ext4_verify_csum_type(sb, es)) {
3298 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3299 "unknown checksum algorithm.");
3300 silent = 1;
3301 goto cantfind_ext4;
3302 }
3303
0441984a
DW		 3304 /* Load the checksum driver */
3305 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3306 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3307 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3308 if (IS_ERR(sbi->s_chksum_driver)) {
3309 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3310 ret = PTR_ERR(sbi->s_chksum_driver);
3311 sbi->s_chksum_driver = NULL;
3312 goto failed_mount;
3313 }
3314 }
3315
a9c47317
DW		 3316 /* Check superblock checksum */
3317 if (!ext4_superblock_csum_verify(sb, es)) {
3318 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3319 "invalid superblock checksum. Run e2fsck?");
3320 silent = 1;
3321 goto cantfind_ext4;
3322 }
3323
3324 /* Precompute checksum seed for all metadata */
3325 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3326 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3327 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3328 sizeof(es->s_uuid));
3329
ac27a0ec
DK		 3330 /* Set defaults before we parse the mount options */
3331 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3332 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3333 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3334 set_opt(sb, DEBUG);
87f26807 3335 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3336 set_opt(sb, GRPID);
617ba13b 3337 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3338 set_opt(sb, NO_UID32);
ea663336 3339 /* xattr user namespace & acls are now defaulted on */
03010a33 3340#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3341 set_opt(sb, XATTR_USER);
2e7842b8 3342#endif
03010a33 3343#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3344 set_opt(sb, POSIX_ACL);
2e7842b8 3345#endif
6fd7a467 3346 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3347 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3348 set_opt(sb, JOURNAL_DATA);
617ba13b 3349 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3350 set_opt(sb, ORDERED_DATA);
617ba13b 3351 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3352 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3353
3354 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3355 set_opt(sb, ERRORS_PANIC);
bb4f397a 3356 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3357 set_opt(sb, ERRORS_CONT);
bb4f397a 3358 else
fd8c37ec 3359 set_opt(sb, ERRORS_RO);
8b67f04a 3360 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3361 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3362 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3363 set_opt(sb, DISCARD);
ac27a0ec 3364
08cefc7a
EB		 3365 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3366 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3367 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3368 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3369 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3370
8b67f04a 3371 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3372 set_opt(sb, BARRIER);
ac27a0ec 3373
dd919b98
AK		 3374 /*
3375 * enable delayed allocation by default
3376 * Use -o nodelalloc to turn it off
3377 */
8b67f04a
TT		 3378 if (!IS_EXT3_SB(sb) &&
3379 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3380 set_opt(sb, DELALLOC);
dd919b98 3381
51ce6511
LC		 3382 /*
3383 * set default s_li_wait_mult for lazyinit, for the case there is
3384 * no mount option specified.
3385 */
3386 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3387
8b67f04a 3388 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3389 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3390 ext4_msg(sb, KERN_WARNING,
3391 "failed to parse options in superblock: %s",
3392 sbi->s_es->s_mount_opts);
3393 }
5a916be1 3394 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3395 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3396 &journal_ioprio, 0))
ac27a0ec
DK		 3397 goto failed_mount;
3398
56889787
TT		 3399 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3400 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3401 "with data=journal disables delayed "
3402 "allocation and O_DIRECT support!\n");
3403 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3404 ext4_msg(sb, KERN_ERR, "can't mount with "
3405 "both data=journal and delalloc");
3406 goto failed_mount;
3407 }
3408 if (test_opt(sb, DIOREAD_NOLOCK)) {
3409 ext4_msg(sb, KERN_ERR, "can't mount with "
3410 "both data=journal and delalloc");
3411 goto failed_mount;
3412 }
3413 if (test_opt(sb, DELALLOC))
3414 clear_opt(sb, DELALLOC);
3415 }
3416
3417 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3418 if (test_opt(sb, DIOREAD_NOLOCK)) {
3419 if (blocksize < PAGE_SIZE) {
3420 ext4_msg(sb, KERN_ERR, "can't mount with "
3421 "dioread_nolock if block size != PAGE_SIZE");
3422 goto failed_mount;
3423 }
3424 }
3425
ac27a0ec 3426 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3427 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3428
617ba13b
MC		 3429 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3430 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3431 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3432 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3433 ext4_msg(sb, KERN_WARNING,
3434 "feature flags set on rev 0 fs, "
3435 "running e2fsck is recommended");
469108ff 3436
2035e776
TT		 3437 if (IS_EXT2_SB(sb)) {
3438 if (ext2_feature_set_ok(sb))
3439 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3440 "using the ext4 subsystem");
3441 else {
3442 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3443 "to feature incompatibilities");
3444 goto failed_mount;
3445 }
3446 }
3447
3448 if (IS_EXT3_SB(sb)) {
3449 if (ext3_feature_set_ok(sb))
3450 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3451 "using the ext4 subsystem");
3452 else {
3453 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3454 "to feature incompatibilities");
3455 goto failed_mount;
3456 }
3457 }
3458
ac27a0ec
DK		 3459 /*
3460 * Check feature flags regardless of the revision level, since we
3461 * previously didn't change the revision level when setting the flags,
3462 * so there is a chance incompat flags are set on a rev 0 filesystem.
3463 */
a13fb1a4 3464 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3465 goto failed_mount;
a13fb1a4 3466
617ba13b
MC		 3467 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3468 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3469 ext4_msg(sb, KERN_ERR,
3470 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3471 goto failed_mount;
3472 }
3473
ac27a0ec 3474 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3475 /* Validate the filesystem blocksize */
3476 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3477 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3478 blocksize);
ac27a0ec
DK		 3479 goto failed_mount;
3480 }
3481
2b2d6d01 3482 brelse(bh);
70bbb3e0
AM		 3483 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3484 offset = do_div(logical_sb_block, blocksize);
3485 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3486 if (!bh) {
b31e1552
ES		 3487 ext4_msg(sb, KERN_ERR,
3488 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3489 goto failed_mount;
3490 }
2716b802 3491 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3492 sbi->s_es = es;
617ba13b 3493 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3494 ext4_msg(sb, KERN_ERR,
3495 "Magic mismatch, very weird!");
ac27a0ec
DK		 3496 goto failed_mount;
3497 }
3498 }
3499
a13fb1a4
ES		 3500 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3501 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3502 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3503 has_huge_files);
3504 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3505
617ba13b
MC		 3506 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3507 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3508 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3509 } else {
3510 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3511 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3512 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3513 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3514 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3515 ext4_msg(sb, KERN_ERR,
3516 "unsupported inode size: %d",
2b2d6d01 3517 sbi->s_inode_size);
ac27a0ec
DK		 3518 goto failed_mount;
3519 }
ef7f3835
KS		 3520 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3521 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3522 }
0b8e58a1 3523
0d1ee42f
AR		 3524 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3525 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3526 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3527 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3528 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3529 ext4_msg(sb, KERN_ERR,
3530 "unsupported descriptor size %lu",
0d1ee42f
AR		 3531 sbi->s_desc_size);
3532 goto failed_mount;
3533 }
3534 } else
3535 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3536
ac27a0ec 3537 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3538 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3539 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3540 goto cantfind_ext4;
0b8e58a1 3541
617ba13b 3542 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3543 if (sbi->s_inodes_per_block == 0)
617ba13b 3544 goto cantfind_ext4;
ac27a0ec
DK		 3545 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3546 sbi->s_inodes_per_block;
0d1ee42f 3547 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3548 sbi->s_sbh = bh;
3549 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3550 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3551 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3552
2b2d6d01 3553 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3554 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3555 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3556 i = le32_to_cpu(es->s_flags);
3557 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3558 sbi->s_hash_unsigned = 3;
3559 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3560#ifdef __CHAR_UNSIGNED__
3561 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3562 sbi->s_hash_unsigned = 3;
3563#else
3564 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3565#endif
f99b2589 3566 }
ac27a0ec 3567
281b5995
TT		 3568 /* Handle clustersize */
3569 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3570 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3571 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3572 if (has_bigalloc) {
3573 if (clustersize < blocksize) {
3574 ext4_msg(sb, KERN_ERR,
3575 "cluster size (%d) smaller than "
3576 "block size (%d)", clustersize, blocksize);
3577 goto failed_mount;
3578 }
3579 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3580 le32_to_cpu(es->s_log_block_size);
3581 sbi->s_clusters_per_group =
3582 le32_to_cpu(es->s_clusters_per_group);
3583 if (sbi->s_clusters_per_group > blocksize * 8) {
3584 ext4_msg(sb, KERN_ERR,
3585 "#clusters per group too big: %lu",
3586 sbi->s_clusters_per_group);
3587 goto failed_mount;
3588 }
3589 if (sbi->s_blocks_per_group !=
3590 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3591 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3592 "clusters per group (%lu) inconsistent",
3593 sbi->s_blocks_per_group,
3594 sbi->s_clusters_per_group);
3595 goto failed_mount;
3596 }
3597 } else {
3598 if (clustersize != blocksize) {
3599 ext4_warning(sb, "fragment/cluster size (%d) != "
3600 "block size (%d)", clustersize,
3601 blocksize);
3602 clustersize = blocksize;
3603 }
3604 if (sbi->s_blocks_per_group > blocksize * 8) {
3605 ext4_msg(sb, KERN_ERR,
3606 "#blocks per group too big: %lu",
3607 sbi->s_blocks_per_group);
3608 goto failed_mount;
3609 }
3610 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3611 sbi->s_cluster_bits = 0;
ac27a0ec 3612 }
281b5995
TT		 3613 sbi->s_cluster_ratio = clustersize / blocksize;
3614
ac27a0ec 3615 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3616 ext4_msg(sb, KERN_ERR,
3617 "#inodes per group too big: %lu",
2b2d6d01 3618 sbi->s_inodes_per_group);
ac27a0ec
DK		 3619 goto failed_mount;
3620 }
3621
bf43d84b
ES		 3622 /*
3623 * Test whether we have more sectors than will fit in sector_t,
3624 * and whether the max offset is addressable by the page cache.
3625 */
5a9ae68a 3626 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3627 ext4_blocks_count(es));
5a9ae68a 3628 if (err) {
b31e1552 3629 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3630 " too large to mount safely on this system");
ac27a0ec 3631 if (sizeof(sector_t) < 8)
90c699a9 3632 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3633 ret = err;
ac27a0ec
DK		 3634 goto failed_mount;
3635 }
3636
617ba13b
MC		 3637 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3638 goto cantfind_ext4;
e7c95593 3639
0f2ddca6
FTN		 3640 /* check blocks count against device size */
3641 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3642 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3643 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3644 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3645 ext4_blocks_count(es), blocks_count);
3646 goto failed_mount;
3647 }
3648
0b8e58a1
AD		 3649 /*
3650 * It makes no sense for the first data block to be beyond the end
3651 * of the filesystem.
3652 */
3653 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3654 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3655 "block %u is beyond end of filesystem (%llu)",
3656 le32_to_cpu(es->s_first_data_block),
3657 ext4_blocks_count(es));
e7c95593
ES		 3658 goto failed_mount;
3659 }
bd81d8ee
LV		 3660 blocks_count = (ext4_blocks_count(es) -
3661 le32_to_cpu(es->s_first_data_block) +
3662 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3663 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3664 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3665 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3666 "(block count %llu, first data block %u, "
b31e1552 3667 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3668 ext4_blocks_count(es),
3669 le32_to_cpu(es->s_first_data_block),
3670 EXT4_BLOCKS_PER_GROUP(sb));
3671 goto failed_mount;
3672 }
bd81d8ee 3673 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3674 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3675 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3676 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3677 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3678 sbi->s_group_desc = ext4_kvmalloc(db_count *
3679 sizeof(struct buffer_head *),
3680 GFP_KERNEL);
ac27a0ec 3681 if (sbi->s_group_desc == NULL) {
b31e1552 3682 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3683 ret = -ENOMEM;
ac27a0ec
DK		 3684 goto failed_mount;
3685 }
3686
9f6200bb
TT		 3687 if (ext4_proc_root)
3688 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3689
66acdcf4
TT		 3690 if (sbi->s_proc)
3691 proc_create_data("options", S_IRUGO, sbi->s_proc,
3692 &ext4_seq_options_fops, sb);
3693
705895b6 3694 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3695
3696 for (i = 0; i < db_count; i++) {
70bbb3e0 3697 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3698 sbi->s_group_desc[i] = sb_bread(sb, block);
3699 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3700 ext4_msg(sb, KERN_ERR,
3701 "can't read group descriptor %d", i);
ac27a0ec
DK		 3702 db_count = i;
3703 goto failed_mount2;
3704 }
3705 }
bfff6873 3706 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3707 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3708 goto failed_mount2;
3709 }
772cb7c8
JS		 3710 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3711 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3712 ext4_msg(sb, KERN_ERR,
3713 "unable to initialize "
3714 "flex_bg meta info!");
772cb7c8
JS		 3715 goto failed_mount2;
3716 }
3717
ac27a0ec
DK		 3718 sbi->s_gdb_count = db_count;
3719 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3720 spin_lock_init(&sbi->s_next_gen_lock);
3721
04496411
TM		 3722 init_timer(&sbi->s_err_report);
3723 sbi->s_err_report.function = print_daily_error_info;
3724 sbi->s_err_report.data = (unsigned long) sb;
3725
57042651 3726 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3727 ext4_count_free_clusters(sb));
ce7e010a
TT		 3728 if (!err) {
3729 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3730 ext4_count_free_inodes(sb));
3731 }
3732 if (!err) {
3733 err = percpu_counter_init(&sbi->s_dirs_counter,
3734 ext4_count_dirs(sb));
3735 }
3736 if (!err) {
57042651 3737 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3738 }
3739 if (err) {
3740 ext4_msg(sb, KERN_ERR, "insufficient memory");
2cde417d 3741 ret = err;
ce7e010a
TT		 3742 goto failed_mount3;
3743 }
3744
c9de560d 3745 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3746 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3747
ac27a0ec
DK		 3748 /*
3749 * set up enough so that it can read an inode
3750 */
9ca92389
TT		 3751 if (!test_opt(sb, NOLOAD) &&
3752 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3753 sb->s_op = &ext4_sops;
3754 else
3755 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3756 sb->s_export_op = &ext4_export_ops;
3757 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3758#ifdef CONFIG_QUOTA
617ba13b
MC		 3759 sb->s_qcop = &ext4_qctl_operations;
3760 sb->dq_op = &ext4_quota_operations;
7c319d32
AK		 3761
3762 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3763 /* Use qctl operations for hidden quota files. */
3764 sb->s_qcop = &ext4_qctl_sysfile_operations;
3765 }
ac27a0ec 3766#endif
f2fa2ffc
AK		 3767 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3768
ac27a0ec 3769 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3770 mutex_init(&sbi->s_orphan_lock);
8f82f840 3771 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3772
3773 sb->s_root = NULL;
3774
3775 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3776 EXT4_HAS_INCOMPAT_FEATURE(sb,
3777 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3778
c5e06d10
JL		 3779 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3780 !(sb->s_flags & MS_RDONLY))
3781 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3782 goto failed_mount3;
3783
ac27a0ec
DK		 3784 /*
3785 * The first inode we look at is the journal inode. Don't try
3786 * root first: it may be modified in the journal!
3787 */
3788 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3789 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3790 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3791 goto failed_mount3;
0390131b
FM		 3792 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3793 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3794 ext4_msg(sb, KERN_ERR, "required journal recovery "
3795 "suppressed and not mounted read-only");
744692dc 3796 goto failed_mount_wq;
ac27a0ec 3797 } else {
fd8c37ec 3798 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3799 sbi->s_journal = NULL;
3800 needs_recovery = 0;
3801 goto no_journal;
ac27a0ec
DK		 3802 }
3803
f32aaf2d 3804 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3805 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3806 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3807 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3808 goto failed_mount_wq;
eb40a09c
JS		 3809 }
3810
25ed6e8a
DW		 3811 if (!set_journal_csum_feature_set(sb)) {
3812 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3813 "feature set");
3814 goto failed_mount_wq;
d4da6c9c 3815 }
818d276c 3816
ac27a0ec
DK		 3817 /* We have now updated the journal if required, so we can
3818 * validate the data journaling mode. */
3819 switch (test_opt(sb, DATA_FLAGS)) {
3820 case 0:
3821 /* No mode set, assume a default based on the journal
63f57933
AM		 3822 * capabilities: ORDERED_DATA if the journal can
3823 * cope, else JOURNAL_DATA
3824 */
dab291af
MC		 3825 if (jbd2_journal_check_available_features
3826 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3827 set_opt(sb, ORDERED_DATA);
ac27a0ec 3828 else
fd8c37ec 3829 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3830 break;
3831
617ba13b
MC		 3832 case EXT4_MOUNT_ORDERED_DATA:
3833 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3834 if (!jbd2_journal_check_available_features
3835 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3836 ext4_msg(sb, KERN_ERR, "Journal does not support "
3837 "requested data journaling mode");
744692dc 3838 goto failed_mount_wq;
ac27a0ec
DK		 3839 }
3840 default:
3841 break;
3842 }
b3881f74 3843 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3844
18aadd47
BJ		 3845 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3846
ce7e010a
TT		 3847 /*
3848 * The journal may have updated the bg summary counts, so we
3849 * need to update the global counters.
3850 */
57042651 3851 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3852 ext4_count_free_clusters(sb));
ce7e010a
TT		 3853 percpu_counter_set(&sbi->s_freeinodes_counter,
3854 ext4_count_free_inodes(sb));
3855 percpu_counter_set(&sbi->s_dirs_counter,
3856 ext4_count_dirs(sb));
57042651 3857 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3858
ce7e010a 3859no_journal:
952fc18e
TT		 3860 /*
3861 * Get the # of file system overhead blocks from the
3862 * superblock if present.
3863 */
3864 if (es->s_overhead_clusters)
3865 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3866 else {
3867 ret = ext4_calculate_overhead(sb);
3868 if (ret)
3869 goto failed_mount_wq;
3870 }
3871
fd89d5f2
TH		 3872 /*
3873 * The maximum number of concurrent works can be high and
3874 * concurrency isn't really necessary. Limit it to 1.
3875 */
3876 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3877 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3878 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3879 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3880 goto failed_mount_wq;
3881 }
3882
ac27a0ec 3883 /*
dab291af 3884 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3885 * so we can safely mount the rest of the filesystem now.
3886 */
3887
1d1fe1ee
DH		 3888 root = ext4_iget(sb, EXT4_ROOT_INO);
3889 if (IS_ERR(root)) {
b31e1552 3890 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3891 ret = PTR_ERR(root);
32a9bb57 3892 root = NULL;
ac27a0ec
DK		 3893 goto failed_mount4;
3894 }
3895 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3896 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3897 iput(root);
ac27a0ec
DK		 3898 goto failed_mount4;
3899 }
48fde701 3900 sb->s_root = d_make_root(root);
1d1fe1ee 3901 if (!sb->s_root) {
b31e1552 3902 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3903 ret = -ENOMEM;
3904 goto failed_mount4;
3905 }
ac27a0ec 3906
7e84b621
ES		 3907 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3908 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3909
3910 /* determine the minimum size of new large inodes, if present */
3911 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3912 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3913 EXT4_GOOD_OLD_INODE_SIZE;
3914 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3915 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3916 if (sbi->s_want_extra_isize <
3917 le16_to_cpu(es->s_want_extra_isize))
3918 sbi->s_want_extra_isize =
3919 le16_to_cpu(es->s_want_extra_isize);
3920 if (sbi->s_want_extra_isize <
3921 le16_to_cpu(es->s_min_extra_isize))
3922 sbi->s_want_extra_isize =
3923 le16_to_cpu(es->s_min_extra_isize);
3924 }
3925 }
3926 /* Check if enough inode space is available */
3927 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3928 sbi->s_inode_size) {
3929 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3930 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3931 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3932 "available");
ef7f3835
KS		 3933 }
3934
6fd058f7
TT		 3935 err = ext4_setup_system_zone(sb);
3936 if (err) {
b31e1552 3937 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3938 "zone (%d)", err);
94bf608a 3939 goto failed_mount4a;
6fd058f7
TT		 3940 }
3941
c2774d84 3942 ext4_ext_init(sb);
9d99012f 3943 err = ext4_mb_init(sb);
c2774d84 3944 if (err) {
421f91d2 3945 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3946 err);
dcf2d804 3947 goto failed_mount5;
c2774d84
AK		 3948 }
3949
bfff6873
LC		 3950 err = ext4_register_li_request(sb, first_not_zeroed);
3951 if (err)
dcf2d804 3952 goto failed_mount6;
bfff6873 3953
3197ebdb
TT		 3954 sbi->s_kobj.kset = ext4_kset;
3955 init_completion(&sbi->s_kobj_unregister);
3956 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3957 "%s", sb->s_id);
dcf2d804
TM		 3958 if (err)
3959 goto failed_mount7;
3197ebdb 3960
617ba13b
MC		 3961 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3962 ext4_orphan_cleanup(sb, es);
3963 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3964 if (needs_recovery) {
b31e1552 3965 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3966 ext4_mark_recovery_complete(sb, es);
3967 }
3968 if (EXT4_SB(sb)->s_journal) {
3969 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3970 descr = " journalled data mode";
3971 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3972 descr = " ordered data mode";
3973 else
3974 descr = " writeback data mode";
3975 } else
3976 descr = "out journal";
3977
7c319d32
AK		 3978#ifdef CONFIG_QUOTA
3979 /* Enable quota usage during mount. */
3980 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3981 !(sb->s_flags & MS_RDONLY)) {
3982 ret = ext4_enable_quotas(sb);
3983 if (ret)
3984 goto failed_mount7;
3985 }
3986#endif /* CONFIG_QUOTA */
3987
d4c402d9 3988 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3989 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3990 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3991
66e61a9e
TT		 3992 if (es->s_error_count)
3993 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3994
d4c402d9 3995 kfree(orig_data);
ac27a0ec
DK		 3996 return 0;
3997
617ba13b 3998cantfind_ext4:
ac27a0ec 3999 if (!silent)
b31e1552 4000 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4001 goto failed_mount;
4002
dcf2d804
TM		 4003failed_mount7:
4004 ext4_unregister_li_request(sb);
4005failed_mount6:
dcf2d804 4006 ext4_mb_release(sb);
94bf608a
AV		 4007failed_mount5:
4008 ext4_ext_release(sb);
dcf2d804 4009 ext4_release_system_zone(sb);
94bf608a
AV		 4010failed_mount4a:
4011 dput(sb->s_root);
32a9bb57 4012 sb->s_root = NULL;
94bf608a 4013failed_mount4:
b31e1552 4014 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 4015 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4016failed_mount_wq:
0390131b
FM		 4017 if (sbi->s_journal) {
4018 jbd2_journal_destroy(sbi->s_journal);
4019 sbi->s_journal = NULL;
4020 }
ac27a0ec 4021failed_mount3:
04496411 4022 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4023 if (sbi->s_flex_groups)
4024 ext4_kvfree(sbi->s_flex_groups);
57042651 4025 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4026 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4027 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4028 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 4029 if (sbi->s_mmp_tsk)
4030 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4031failed_mount2:
4032 for (i = 0; i < db_count; i++)
4033 brelse(sbi->s_group_desc[i]);
f18a5f21 4034 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4035failed_mount:
0441984a
DW		 4036 if (sbi->s_chksum_driver)
4037 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4038 if (sbi->s_proc) {
66acdcf4 4039 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4040 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4041 }
ac27a0ec
DK		 4042#ifdef CONFIG_QUOTA
4043 for (i = 0; i < MAXQUOTAS; i++)
4044 kfree(sbi->s_qf_names[i]);
4045#endif
617ba13b 4046 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4047 brelse(bh);
4048out_fail:
4049 sb->s_fs_info = NULL;
f6830165 4050 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4051 kfree(sbi);
dcc7dae3 4052out_free_orig:
d4c402d9 4053 kfree(orig_data);
1d1fe1ee 4054 return ret;
ac27a0ec
DK		 4055}
4056
4057/*
4058 * Setup any per-fs journal parameters now. We'll do this both on
4059 * initial mount, once the journal has been initialised but before we've
4060 * done any recovery; and again on any subsequent remount.
4061 */
617ba13b 4062static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4063{
617ba13b 4064 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4065
30773840
TT		 4066 journal->j_commit_interval = sbi->s_commit_interval;
4067 journal->j_min_batch_time = sbi->s_min_batch_time;
4068 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4069
a931da6a 4070 write_lock(&journal->j_state_lock);
ac27a0ec 4071 if (test_opt(sb, BARRIER))
dab291af 4072 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4073 else
dab291af 4074 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4075 if (test_opt(sb, DATA_ERR_ABORT))
4076 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4077 else
4078 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4079 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4080}
4081
617ba13b 4082static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4083 unsigned int journal_inum)
4084{
4085 struct inode *journal_inode;
4086 journal_t *journal;
4087
0390131b
FM		 4088 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4089
ac27a0ec
DK		 4090 /* First, test for the existence of a valid inode on disk. Bad
4091 * things happen if we iget() an unused inode, as the subsequent
4092 * iput() will try to delete it. */
4093
1d1fe1ee
DH		 4094 journal_inode = ext4_iget(sb, journal_inum);
4095 if (IS_ERR(journal_inode)) {
b31e1552 4096 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4097 return NULL;
4098 }
4099 if (!journal_inode->i_nlink) {
4100 make_bad_inode(journal_inode);
4101 iput(journal_inode);
b31e1552 4102 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4103 return NULL;
4104 }
4105
e5f8eab8 4106 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4107 journal_inode, journal_inode->i_size);
1d1fe1ee 4108 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4109 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4110 iput(journal_inode);
4111 return NULL;
4112 }
4113
dab291af 4114 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4115 if (!journal) {
b31e1552 4116 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4117 iput(journal_inode);
4118 return NULL;
4119 }
4120 journal->j_private = sb;
617ba13b 4121 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4122 return journal;
4123}
4124
617ba13b 4125static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4126 dev_t j_dev)
4127{
2b2d6d01 4128 struct buffer_head *bh;
ac27a0ec 4129 journal_t *journal;
617ba13b
MC		 4130 ext4_fsblk_t start;
4131 ext4_fsblk_t len;
ac27a0ec 4132 int hblock, blocksize;
617ba13b 4133 ext4_fsblk_t sb_block;
ac27a0ec 4134 unsigned long offset;
2b2d6d01 4135 struct ext4_super_block *es;
ac27a0ec
DK		 4136 struct block_device *bdev;
4137
0390131b
FM		 4138 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4139
b31e1552 4140 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4141 if (bdev == NULL)
4142 return NULL;
4143
ac27a0ec 4144 blocksize = sb->s_blocksize;
e1defc4f 4145 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4146 if (blocksize < hblock) {
b31e1552
ES		 4147 ext4_msg(sb, KERN_ERR,
4148 "blocksize too small for journal device");
ac27a0ec
DK		 4149 goto out_bdev;
4150 }
4151
617ba13b
MC		 4152 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4153 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4154 set_blocksize(bdev, blocksize);
4155 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4156 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4157 "external journal");
ac27a0ec
DK		 4158 goto out_bdev;
4159 }
4160
2716b802 4161 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4162 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4163 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4164 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4165 ext4_msg(sb, KERN_ERR, "external journal has "
4166 "bad superblock");
ac27a0ec
DK		 4167 brelse(bh);
4168 goto out_bdev;
4169 }
4170
617ba13b 4171 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4172 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4173 brelse(bh);
4174 goto out_bdev;
4175 }
4176
bd81d8ee 4177 len = ext4_blocks_count(es);
ac27a0ec
DK		 4178 start = sb_block + 1;
4179 brelse(bh); /* we're done with the superblock */
4180
dab291af 4181 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4182 start, len, blocksize);
4183 if (!journal) {
b31e1552 4184 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4185 goto out_bdev;
4186 }
4187 journal->j_private = sb;
4188 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4189 wait_on_buffer(journal->j_sb_buffer);
4190 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4191 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4192 goto out_journal;
4193 }
4194 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4195 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4196 "user (unsupported) - %d",
ac27a0ec
DK		 4197 be32_to_cpu(journal->j_superblock->s_nr_users));
4198 goto out_journal;
4199 }
617ba13b
MC		 4200 EXT4_SB(sb)->journal_bdev = bdev;
4201 ext4_init_journal_params(sb, journal);
ac27a0ec 4202 return journal;
0b8e58a1 4203
ac27a0ec 4204out_journal:
dab291af 4205 jbd2_journal_destroy(journal);
ac27a0ec 4206out_bdev:
617ba13b 4207 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4208 return NULL;
4209}
4210
617ba13b
MC		 4211static int ext4_load_journal(struct super_block *sb,
4212 struct ext4_super_block *es,
ac27a0ec
DK		 4213 unsigned long journal_devnum)
4214{
4215 journal_t *journal;
4216 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4217 dev_t journal_dev;
4218 int err = 0;
4219 int really_read_only;
4220
0390131b
FM		 4221 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4222
ac27a0ec
DK		 4223 if (journal_devnum &&
4224 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4225 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4226 "numbers have changed");
ac27a0ec
DK		 4227 journal_dev = new_decode_dev(journal_devnum);
4228 } else
4229 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4230
4231 really_read_only = bdev_read_only(sb->s_bdev);
4232
4233 /*
4234 * Are we loading a blank journal or performing recovery after a
4235 * crash? For recovery, we need to check in advance whether we
4236 * can get read-write access to the device.
4237 */
617ba13b 4238 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4239 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4240 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4241 "required on readonly filesystem");
ac27a0ec 4242 if (really_read_only) {
b31e1552
ES		 4243 ext4_msg(sb, KERN_ERR, "write access "
4244 "unavailable, cannot proceed");
ac27a0ec
DK		 4245 return -EROFS;
4246 }
b31e1552
ES		 4247 ext4_msg(sb, KERN_INFO, "write access will "
4248 "be enabled during recovery");
ac27a0ec
DK		 4249 }
4250 }
4251
4252 if (journal_inum && journal_dev) {
b31e1552
ES		 4253 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4254 "and inode journals!");
ac27a0ec
DK		 4255 return -EINVAL;
4256 }
4257
4258 if (journal_inum) {
617ba13b 4259 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4260 return -EINVAL;
4261 } else {
617ba13b 4262 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4263 return -EINVAL;
4264 }
4265
90576c0b 4266 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4267 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4268
617ba13b 4269 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4270 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4271 if (!err) {
4272 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4273 if (save)
4274 memcpy(save, ((char *) es) +
4275 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4276 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4277 if (save)
4278 memcpy(((char *) es) + EXT4_S_ERR_START,
4279 save, EXT4_S_ERR_LEN);
4280 kfree(save);
4281 }
ac27a0ec
DK		 4282
4283 if (err) {
b31e1552 4284 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4285 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4286 return err;
4287 }
4288
617ba13b
MC		 4289 EXT4_SB(sb)->s_journal = journal;
4290 ext4_clear_journal_err(sb, es);
ac27a0ec 4291
c41303ce 4292 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4293 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4294 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4295
4296 /* Make sure we flush the recovery flag to disk. */
e2d67052 4297 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4298 }
4299
4300 return 0;
4301}
4302
e2d67052 4303static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4304{
e2d67052 4305 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4306 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4307 int error = 0;
ac27a0ec 4308
7c2e7087 4309 if (!sbh || block_device_ejected(sb))
c4be0c1d 4310 return error;
914258bf
TT		 4311 if (buffer_write_io_error(sbh)) {
4312 /*
4313 * Oh, dear. A previous attempt to write the
4314 * superblock failed. This could happen because the
4315 * USB device was yanked out. Or it could happen to
4316 * be a transient write error and maybe the block will
4317 * be remapped. Nothing we can do but to retry the
4318 * write and hope for the best.
4319 */
b31e1552
ES		 4320 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4321 "superblock detected");
914258bf
TT		 4322 clear_buffer_write_io_error(sbh);
4323 set_buffer_uptodate(sbh);
4324 }
71290b36
TT		 4325 /*
4326 * If the file system is mounted read-only, don't update the
4327 * superblock write time. This avoids updating the superblock
4328 * write time when we are mounting the root file system
4329 * read/only but we need to replay the journal; at that point,
4330 * for people who are east of GMT and who make their clock
4331 * tick in localtime for Windows bug-for-bug compatibility,
4332 * the clock is set in the future, and this will cause e2fsck
4333 * to complain and force a full file system check.
4334 */
4335 if (!(sb->s_flags & MS_RDONLY))
4336 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4337 if (sb->s_bdev->bd_part)
4338 es->s_kbytes_written =
4339 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4340 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4341 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4342 else
4343 es->s_kbytes_written =
4344 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4345 ext4_free_blocks_count_set(es,
4346 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4347 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4348 es->s_free_inodes_count =
4349 cpu_to_le32(percpu_counter_sum_positive(
4350 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4351 BUFFER_TRACE(sbh, "marking dirty");
a9c47317 4352 ext4_superblock_csum_set(sb, es);
ac27a0ec 4353 mark_buffer_dirty(sbh);
914258bf 4354 if (sync) {
c4be0c1d
TS		 4355 error = sync_dirty_buffer(sbh);
4356 if (error)
4357 return error;
4358
4359 error = buffer_write_io_error(sbh);
4360 if (error) {
b31e1552
ES		 4361 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4362 "superblock");
914258bf
TT		 4363 clear_buffer_write_io_error(sbh);
4364 set_buffer_uptodate(sbh);
4365 }
4366 }
c4be0c1d 4367 return error;
ac27a0ec
DK		 4368}
4369
ac27a0ec
DK		 4370/*
4371 * Have we just finished recovery? If so, and if we are mounting (or
4372 * remounting) the filesystem readonly, then we will end up with a
4373 * consistent fs on disk. Record that fact.
4374 */
2b2d6d01
TT		 4375static void ext4_mark_recovery_complete(struct super_block *sb,
4376 struct ext4_super_block *es)
ac27a0ec 4377{
617ba13b 4378 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4379
0390131b
FM		 4380 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4381 BUG_ON(journal != NULL);
4382 return;
4383 }
dab291af 4384 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4385 if (jbd2_journal_flush(journal) < 0)
4386 goto out;
4387
617ba13b 4388 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4389 sb->s_flags & MS_RDONLY) {
617ba13b 4390 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4391 ext4_commit_super(sb, 1);
ac27a0ec 4392 }
7ffe1ea8
HK		 4393
4394out:
dab291af 4395 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4396}
4397
4398/*
4399 * If we are mounting (or read-write remounting) a filesystem whose journal
4400 * has recorded an error from a previous lifetime, move that error to the
4401 * main filesystem now.
4402 */
2b2d6d01
TT		 4403static void ext4_clear_journal_err(struct super_block *sb,
4404 struct ext4_super_block *es)
ac27a0ec
DK		 4405{
4406 journal_t *journal;
4407 int j_errno;
4408 const char *errstr;
4409
0390131b
FM		 4410 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4411
617ba13b 4412 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4413
4414 /*
4415 * Now check for any error status which may have been recorded in the
617ba13b 4416 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4417 */
4418
dab291af 4419 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4420 if (j_errno) {
4421 char nbuf[16];
4422
617ba13b 4423 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4424 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4425 "from previous mount: %s", errstr);
12062ddd 4426 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4427
617ba13b
MC		 4428 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4429 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4430 ext4_commit_super(sb, 1);
ac27a0ec 4431
dab291af 4432 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4433 }
4434}
4435
4436/*
4437 * Force the running and committing transactions to commit,
4438 * and wait on the commit.
4439 */
617ba13b 4440int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4441{
4442 journal_t *journal;
0390131b 4443 int ret = 0;
ac27a0ec
DK		 4444
4445 if (sb->s_flags & MS_RDONLY)
4446 return 0;
4447
617ba13b 4448 journal = EXT4_SB(sb)->s_journal;
8e8ad8a5 4449 if (journal)
0390131b 4450 ret = ext4_journal_force_commit(journal);
0390131b 4451
ac27a0ec
DK		 4452 return ret;
4453}
4454
617ba13b 4455static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4456{
14ce0cb4 4457 int ret = 0;
9eddacf9 4458 tid_t target;
8d5d02e6 4459 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4460
9bffad1e 4461 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4462 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4463 /*
4464 * Writeback quota in non-journalled quota case - journalled quota has
4465 * no dirty dquots
4466 */
4467 dquot_writeback_dquots(sb, -1);
8d5d02e6 4468 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4469 if (wait)
8d5d02e6 4470 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4471 }
14ce0cb4 4472 return ret;
ac27a0ec
DK		 4473}
4474
4475/*
4476 * LVM calls this function before a (read-only) snapshot is created. This
4477 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4478 *
4479 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4480 * state independently. It relies on upper layer to stop all data & metadata
4481 * modifications.
ac27a0ec 4482 */
c4be0c1d 4483static int ext4_freeze(struct super_block *sb)
ac27a0ec 4484{
c4be0c1d
TS		 4485 int error = 0;
4486 journal_t *journal;
ac27a0ec 4487
9ca92389
TT		 4488 if (sb->s_flags & MS_RDONLY)
4489 return 0;
ac27a0ec 4490
9ca92389 4491 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4492
9ca92389
TT		 4493 /* Now we set up the journal barrier. */
4494 jbd2_journal_lock_updates(journal);
ac27a0ec 4495
9ca92389
TT		 4496 /*
4497 * Don't clear the needs_recovery flag if we failed to flush
4498 * the journal.
4499 */
4500 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4501 if (error < 0)
4502 goto out;
9ca92389
TT		 4503
4504 /* Journal blocked and flushed, clear needs_recovery flag. */
4505 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4506 error = ext4_commit_super(sb, 1);
6b0310fb 4507out:
8e8ad8a5 4508 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4509 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4510 return error;
ac27a0ec
DK		 4511}
4512
4513/*
4514 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4515 * flag here, even though the filesystem is not technically dirty yet.
4516 */
c4be0c1d 4517static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4518{
9ca92389
TT		 4519 if (sb->s_flags & MS_RDONLY)
4520 return 0;
4521
4522 lock_super(sb);
4523 /* Reset the needs_recovery flag before the fs is unlocked. */
4524 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4525 ext4_commit_super(sb, 1);
4526 unlock_super(sb);
c4be0c1d 4527 return 0;
ac27a0ec
DK		 4528}
4529
673c6100
TT		 4530/*
4531 * Structure to save mount options for ext4_remount's benefit
4532 */
4533struct ext4_mount_options {
4534 unsigned long s_mount_opt;
a2595b8a 4535 unsigned long s_mount_opt2;
08cefc7a
EB		 4536 kuid_t s_resuid;
4537 kgid_t s_resgid;
673c6100
TT		 4538 unsigned long s_commit_interval;
4539 u32 s_min_batch_time, s_max_batch_time;
4540#ifdef CONFIG_QUOTA
4541 int s_jquota_fmt;
4542 char *s_qf_names[MAXQUOTAS];
4543#endif
4544};
4545
2b2d6d01 4546static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4547{
2b2d6d01 4548 struct ext4_super_block *es;
617ba13b 4549 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4550 unsigned long old_sb_flags;
617ba13b 4551 struct ext4_mount_options old_opts;
c79d967d 4552 int enable_quota = 0;
8a266467 4553 ext4_group_t g;
b3881f74 4554 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4555 int err = 0;
ac27a0ec
DK		 4556#ifdef CONFIG_QUOTA
4557 int i;
4558#endif
d4c402d9 4559 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4560
4561 /* Store the original options */
bbd6851a 4562 lock_super(sb);
ac27a0ec
DK		 4563 old_sb_flags = sb->s_flags;
4564 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4565 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4566 old_opts.s_resuid = sbi->s_resuid;
4567 old_opts.s_resgid = sbi->s_resgid;
4568 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4569 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4570 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4571#ifdef CONFIG_QUOTA
4572 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4573 for (i = 0; i < MAXQUOTAS; i++)
4574 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4575#endif
b3881f74
TT		 4576 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4577 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4578
4579 /*
4580 * Allow the "check" option to be passed as a remount option.
4581 */
661aa520 4582 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4583 err = -EINVAL;
4584 goto restore_opts;
4585 }
4586
4ab2f15b 4587 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4588 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4589
4590 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4591 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4592
4593 es = sbi->s_es;
4594
b3881f74 4595 if (sbi->s_journal) {
0390131b 4596 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4597 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4598 }
ac27a0ec 4599
661aa520 4600 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4601 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4602 err = -EROFS;
4603 goto restore_opts;
4604 }
4605
4606 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4607 err = dquot_suspend(sb, -1);
4608 if (err < 0)
c79d967d 4609 goto restore_opts;
c79d967d 4610
ac27a0ec
DK		 4611 /*
4612 * First of all, the unconditional stuff we have to do
4613 * to disable replay of the journal when we next remount
4614 */
4615 sb->s_flags |= MS_RDONLY;
4616
4617 /*
4618 * OK, test if we are remounting a valid rw partition
4619 * readonly, and if so set the rdonly flag and then
4620 * mark the partition as valid again.
4621 */
617ba13b
MC		 4622 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4623 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4624 es->s_state = cpu_to_le16(sbi->s_mount_state);
4625
a63c9eb2 4626 if (sbi->s_journal)
0390131b 4627 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4628 } else {
a13fb1a4
ES		 4629 /* Make sure we can mount this feature set readwrite */
4630 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4631 err = -EROFS;
4632 goto restore_opts;
4633 }
8a266467
TT		 4634 /*
4635 * Make sure the group descriptor checksums
0b8e58a1 4636 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4637 */
4638 for (g = 0; g < sbi->s_groups_count; g++) {
4639 struct ext4_group_desc *gdp =
4640 ext4_get_group_desc(sb, g, NULL);
4641
feb0ab32 4642 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4643 ext4_msg(sb, KERN_ERR,
4644 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4645 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4646 le16_to_cpu(gdp->bg_checksum));
4647 err = -EINVAL;
4648 goto restore_opts;
4649 }
4650 }
4651
ead6596b
ES		 4652 /*
4653 * If we have an unprocessed orphan list hanging
4654 * around from a previously readonly bdev mount,
4655 * require a full umount/remount for now.
4656 */
4657 if (es->s_last_orphan) {
b31e1552 4658 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4659 "remount RDWR because of unprocessed "
4660 "orphan inode list. Please "
b31e1552 4661 "umount/remount instead");
ead6596b
ES		 4662 err = -EINVAL;
4663 goto restore_opts;
4664 }
4665
ac27a0ec
DK		 4666 /*
4667 * Mounting a RDONLY partition read-write, so reread
4668 * and store the current valid flag. (It may have
4669 * been changed by e2fsck since we originally mounted
4670 * the partition.)
4671 */
0390131b
FM		 4672 if (sbi->s_journal)
4673 ext4_clear_journal_err(sb, es);
ac27a0ec 4674 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4675 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4676 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4677 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4678 EXT4_FEATURE_INCOMPAT_MMP))
4679 if (ext4_multi_mount_protect(sb,
4680 le64_to_cpu(es->s_mmp_block))) {
4681 err = -EROFS;
4682 goto restore_opts;
4683 }
c79d967d 4684 enable_quota = 1;
ac27a0ec
DK		 4685 }
4686 }
bfff6873
LC		 4687
4688 /*
4689 * Reinitialize lazy itable initialization thread based on
4690 * current settings
4691 */
4692 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4693 ext4_unregister_li_request(sb);
4694 else {
4695 ext4_group_t first_not_zeroed;
4696 first_not_zeroed = ext4_has_uninit_itable(sb);
4697 ext4_register_li_request(sb, first_not_zeroed);
4698 }
4699
6fd058f7 4700 ext4_setup_system_zone(sb);
0390131b 4701 if (sbi->s_journal == NULL)
e2d67052 4702 ext4_commit_super(sb, 1);
0390131b 4703
7c319d32 4704 unlock_super(sb);
ac27a0ec
DK		 4705#ifdef CONFIG_QUOTA
4706 /* Release old quota file names */
4707 for (i = 0; i < MAXQUOTAS; i++)
4708 if (old_opts.s_qf_names[i] &&
4709 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4710 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4711 if (enable_quota) {
4712 if (sb_any_quota_suspended(sb))
4713 dquot_resume(sb, -1);
4714 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4715 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4716 err = ext4_enable_quotas(sb);
4717 if (err) {
4718 lock_super(sb);
4719 goto restore_opts;
4720 }
4721 }
4722 }
ac27a0ec 4723#endif
d4c402d9
CW		 4724
4725 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4726 kfree(orig_data);
ac27a0ec 4727 return 0;
0b8e58a1 4728
ac27a0ec
DK		 4729restore_opts:
4730 sb->s_flags = old_sb_flags;
4731 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4732 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4733 sbi->s_resuid = old_opts.s_resuid;
4734 sbi->s_resgid = old_opts.s_resgid;
4735 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4736 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4737 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4738#ifdef CONFIG_QUOTA
4739 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4740 for (i = 0; i < MAXQUOTAS; i++) {
4741 if (sbi->s_qf_names[i] &&
4742 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4743 kfree(sbi->s_qf_names[i]);
4744 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4745 }
4746#endif
bbd6851a 4747 unlock_super(sb);
d4c402d9 4748 kfree(orig_data);
ac27a0ec
DK		 4749 return err;
4750}
4751
2b2d6d01 4752static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4753{
4754 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4755 struct ext4_sb_info *sbi = EXT4_SB(sb);
4756 struct ext4_super_block *es = sbi->s_es;
952fc18e 4757 ext4_fsblk_t overhead = 0;
960cc398 4758 u64 fsid;
d02a9391 4759 s64 bfree;
ac27a0ec 4760
952fc18e
TT		 4761 if (!test_opt(sb, MINIX_DF))
4762 overhead = sbi->s_overhead;
ac27a0ec 4763
617ba13b 4764 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4765 buf->f_bsize = sb->s_blocksize;
952fc18e 4766 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
57042651
TT		 4767 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4768 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4769 /* prevent underflow in case that few free space is available */
57042651 4770 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4771 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4772 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4773 buf->f_bavail = 0;
4774 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4775 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4776 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4777 fsid = le64_to_cpup((void *)es->s_uuid) ^
4778 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4779 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4780 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4781
ac27a0ec
DK		 4782 return 0;
4783}
4784
0b8e58a1
AD		 4785/* Helper function for writing quotas on sync - we need to start transaction
4786 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4787 * Process 1 Process 2
617ba13b 4788 * ext4_create() quota_sync()
a269eb18 4789 * jbd2_journal_start() write_dquot()
871a2931 4790 * dquot_initialize() down(dqio_mutex)
dab291af 4791 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4792 *
4793 */
4794
4795#ifdef CONFIG_QUOTA
4796
4797static inline struct inode *dquot_to_inode(struct dquot *dquot)
4798{
4799 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4800}
4801
617ba13b 4802static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4803{
4804 int ret, err;
4805 handle_t *handle;
4806 struct inode *inode;
4807
4808 inode = dquot_to_inode(dquot);
617ba13b 4809 handle = ext4_journal_start(inode,
0b8e58a1 4810 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4811 if (IS_ERR(handle))
4812 return PTR_ERR(handle);
4813 ret = dquot_commit(dquot);
617ba13b 4814 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4815 if (!ret)
4816 ret = err;
4817 return ret;
4818}
4819
617ba13b 4820static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4821{
4822 int ret, err;
4823 handle_t *handle;
4824
617ba13b 4825 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4826 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4827 if (IS_ERR(handle))
4828 return PTR_ERR(handle);
4829 ret = dquot_acquire(dquot);
617ba13b 4830 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4831 if (!ret)
4832 ret = err;
4833 return ret;
4834}
4835
617ba13b 4836static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4837{
4838 int ret, err;
4839 handle_t *handle;
4840
617ba13b 4841 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4842 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4843 if (IS_ERR(handle)) {
4844 /* Release dquot anyway to avoid endless cycle in dqput() */
4845 dquot_release(dquot);
ac27a0ec 4846 return PTR_ERR(handle);
9c3013e9 4847 }
ac27a0ec 4848 ret = dquot_release(dquot);
617ba13b 4849 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4850 if (!ret)
4851 ret = err;
4852 return ret;
4853}
4854
617ba13b 4855static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4856{
2c8be6b2 4857 /* Are we journaling quotas? */
617ba13b
MC		 4858 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4859 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4860 dquot_mark_dquot_dirty(dquot);
617ba13b 4861 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4862 } else {
4863 return dquot_mark_dquot_dirty(dquot);
4864 }
4865}
4866
617ba13b 4867static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4868{
4869 int ret, err;
4870 handle_t *handle;
4871
4872 /* Data block + inode block */
617ba13b 4873 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4874 if (IS_ERR(handle))
4875 return PTR_ERR(handle);
4876 ret = dquot_commit_info(sb, type);
617ba13b 4877 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4878 if (!ret)
4879 ret = err;
4880 return ret;
4881}
4882
4883/*
4884 * Turn on quotas during mount time - we need to find
4885 * the quota file and such...
4886 */
617ba13b 4887static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4888{
287a8095
CH		 4889 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4890 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4891}
4892
4893/*
4894 * Standard function to be called on quota_on
4895 */
617ba13b 4896static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4897 struct path *path)
ac27a0ec
DK		 4898{
4899 int err;
ac27a0ec
DK		 4900
4901 if (!test_opt(sb, QUOTA))
4902 return -EINVAL;
0623543b 4903
ac27a0ec 4904 /* Quotafile not on the same filesystem? */
d8c9584e 4905 if (path->dentry->d_sb != sb)
ac27a0ec 4906 return -EXDEV;
0623543b
JK		 4907 /* Journaling quota? */
4908 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4909 /* Quotafile not in fs root? */
f00c9e44 4910 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4911 ext4_msg(sb, KERN_WARNING,
4912 "Quota file not on filesystem root. "
4913 "Journaled quota will not work");
2b2d6d01 4914 }
0623543b
JK		 4915
4916 /*
4917 * When we journal data on quota file, we have to flush journal to see
4918 * all updates to the file when we bypass pagecache...
4919 */
0390131b 4920 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4921 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4922 /*
4923 * We don't need to lock updates but journal_flush() could
4924 * otherwise be livelocked...
4925 */
4926 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4927 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4928 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4929 if (err)
7ffe1ea8 4930 return err;
0623543b
JK		 4931 }
4932
f00c9e44 4933 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4934}
4935
7c319d32
AK		 4936static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4937 unsigned int flags)
4938{
4939 int err;
4940 struct inode *qf_inode;
4941 unsigned long qf_inums[MAXQUOTAS] = {
4942 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4943 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4944 };
4945
4946 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4947
4948 if (!qf_inums[type])
4949 return -EPERM;
4950
4951 qf_inode = ext4_iget(sb, qf_inums[type]);
4952 if (IS_ERR(qf_inode)) {
4953 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4954 return PTR_ERR(qf_inode);
4955 }
4956
4957 err = dquot_enable(qf_inode, type, format_id, flags);
4958 iput(qf_inode);
4959
4960 return err;
4961}
4962
4963/* Enable usage tracking for all quota types. */
4964static int ext4_enable_quotas(struct super_block *sb)
4965{
4966 int type, err = 0;
4967 unsigned long qf_inums[MAXQUOTAS] = {
4968 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4969 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4970 };
4971
4972 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4973 for (type = 0; type < MAXQUOTAS; type++) {
4974 if (qf_inums[type]) {
4975 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4976 DQUOT_USAGE_ENABLED);
4977 if (err) {
4978 ext4_warning(sb,
4979 "Failed to enable quota (type=%d) "
4980 "tracking. Please run e2fsck to fix.",
4981 type);
4982 return err;
4983 }
4984 }
4985 }
4986 return 0;
4987}
4988
4989/*
4990 * quota_on function that is used when QUOTA feature is set.
4991 */
4992static int ext4_quota_on_sysfile(struct super_block *sb, int type,
4993 int format_id)
4994{
4995 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4996 return -EINVAL;
4997
4998 /*
4999 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5000 */
5001 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5002}
5003
ca0e05e4
DM		 5004static int ext4_quota_off(struct super_block *sb, int type)
5005{
21f97697
JK		 5006 struct inode *inode = sb_dqopt(sb)->files[type];
5007 handle_t *handle;
5008
87009d86
DM		 5009 /* Force all delayed allocation blocks to be allocated.
5010 * Caller already holds s_umount sem */
5011 if (test_opt(sb, DELALLOC))
ca0e05e4 5012 sync_filesystem(sb);
ca0e05e4 5013
0b268590
AG		 5014 if (!inode)
5015 goto out;
5016
21f97697
JK		 5017 /* Update modification times of quota files when userspace can
5018 * start looking at them */
5019 handle = ext4_journal_start(inode, 1);
5020 if (IS_ERR(handle))
5021 goto out;
5022 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5023 ext4_mark_inode_dirty(handle, inode);
5024 ext4_journal_stop(handle);
5025
5026out:
ca0e05e4
DM		 5027 return dquot_quota_off(sb, type);
5028}
5029
7c319d32
AK		 5030/*
5031 * quota_off function that is used when QUOTA feature is set.
5032 */
5033static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5034{
5035 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5036 return -EINVAL;
5037
5038 /* Disable only the limits. */
5039 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5040}
5041
ac27a0ec
DK		 5042/* Read data from quotafile - avoid pagecache and such because we cannot afford
5043 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5044 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5045 * we don't have to be afraid of races */
617ba13b 5046static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5047 size_t len, loff_t off)
5048{
5049 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5050 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5051 int err = 0;
5052 int offset = off & (sb->s_blocksize - 1);
5053 int tocopy;
5054 size_t toread;
5055 struct buffer_head *bh;
5056 loff_t i_size = i_size_read(inode);
5057
5058 if (off > i_size)
5059 return 0;
5060 if (off+len > i_size)
5061 len = i_size-off;
5062 toread = len;
5063 while (toread > 0) {
5064 tocopy = sb->s_blocksize - offset < toread ?
5065 sb->s_blocksize - offset : toread;
617ba13b 5066 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5067 if (err)
5068 return err;
5069 if (!bh) /* A hole? */
5070 memset(data, 0, tocopy);
5071 else
5072 memcpy(data, bh->b_data+offset, tocopy);
5073 brelse(bh);
5074 offset = 0;
5075 toread -= tocopy;
5076 data += tocopy;
5077 blk++;
5078 }
5079 return len;
5080}
5081
5082/* Write to quotafile (we know the transaction is already started and has
5083 * enough credits) */
617ba13b 5084static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5085 const char *data, size_t len, loff_t off)
5086{
5087 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5088 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5089 int err = 0;
5090 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5091 struct buffer_head *bh;
5092 handle_t *handle = journal_current_handle();
5093
0390131b 5094 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5095 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5096 " cancelled because transaction is not started",
9c3013e9
JK		 5097 (unsigned long long)off, (unsigned long long)len);
5098 return -EIO;
5099 }
67eeb568
DM		 5100 /*
5101 * Since we account only one data block in transaction credits,
5102 * then it is impossible to cross a block boundary.
5103 */
5104 if (sb->s_blocksize - offset < len) {
5105 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5106 " cancelled because not block aligned",
5107 (unsigned long long)off, (unsigned long long)len);
5108 return -EIO;
5109 }
5110
67eeb568
DM		 5111 bh = ext4_bread(handle, inode, blk, 1, &err);
5112 if (!bh)
5113 goto out;
62d2b5f2
JK		 5114 err = ext4_journal_get_write_access(handle, bh);
5115 if (err) {
5116 brelse(bh);
5117 goto out;
ac27a0ec 5118 }
67eeb568
DM		 5119 lock_buffer(bh);
5120 memcpy(bh->b_data+offset, data, len);
5121 flush_dcache_page(bh->b_page);
5122 unlock_buffer(bh);
62d2b5f2 5123 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5124 brelse(bh);
ac27a0ec 5125out:
0b7f7cef 5126 if (err)
ac27a0ec 5127 return err;
67eeb568
DM		 5128 if (inode->i_size < off + len) {
5129 i_size_write(inode, off + len);
617ba13b 5130 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5131 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5132 }
67eeb568 5133 return len;
ac27a0ec
DK		 5134}
5135
5136#endif
5137
152a0836
AV		 5138static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5139 const char *dev_name, void *data)
ac27a0ec 5140{
152a0836 5141 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5142}
5143
37f328eb 5144#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5145static inline void register_as_ext2(void)
5146{
5147 int err = register_filesystem(&ext2_fs_type);
5148 if (err)
5149 printk(KERN_WARNING
5150 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5151}
5152
5153static inline void unregister_as_ext2(void)
5154{
5155 unregister_filesystem(&ext2_fs_type);
5156}
2035e776
TT		 5157
5158static inline int ext2_feature_set_ok(struct super_block *sb)
5159{
5160 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5161 return 0;
5162 if (sb->s_flags & MS_RDONLY)
5163 return 1;
5164 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5165 return 0;
5166 return 1;
5167}
51b7e3c9 5168MODULE_ALIAS("ext2");
24b58424
TT		 5169#else
5170static inline void register_as_ext2(void) { }
5171static inline void unregister_as_ext2(void) { }
2035e776 5172static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5173#endif
5174
37f328eb 5175#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5176static inline void register_as_ext3(void)
5177{
5178 int err = register_filesystem(&ext3_fs_type);
5179 if (err)
5180 printk(KERN_WARNING
5181 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5182}
5183
5184static inline void unregister_as_ext3(void)
5185{
5186 unregister_filesystem(&ext3_fs_type);
5187}
2035e776
TT		 5188
5189static inline int ext3_feature_set_ok(struct super_block *sb)
5190{
5191 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5192 return 0;
5193 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5194 return 0;
5195 if (sb->s_flags & MS_RDONLY)
5196 return 1;
5197 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5198 return 0;
5199 return 1;
5200}
51b7e3c9 5201MODULE_ALIAS("ext3");
24b58424
TT		 5202#else
5203static inline void register_as_ext3(void) { }
5204static inline void unregister_as_ext3(void) { }
2035e776 5205static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5206#endif
5207
03010a33
TT		 5208static struct file_system_type ext4_fs_type = {
5209 .owner = THIS_MODULE,
5210 .name = "ext4",
152a0836 5211 .mount = ext4_mount,
03010a33
TT		 5212 .kill_sb = kill_block_super,
5213 .fs_flags = FS_REQUIRES_DEV,
5214};
5215
8f021222 5216static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5217{
5218 struct ext4_features *ef;
5219 int ret = -ENOMEM;
5220
5221 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5222 if (!ef)
5223 goto out;
5224
5225 ef->f_kobj.kset = ext4_kset;
5226 init_completion(&ef->f_kobj_unregister);
5227 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5228 "features");
5229 if (ret) {
5230 kfree(ef);
5231 goto out;
5232 }
5233
5234 ext4_feat = ef;
5235 ret = 0;
5236out:
5237 return ret;
5238}
5239
8f021222
LC		 5240static void ext4_exit_feat_adverts(void)
5241{
5242 kobject_put(&ext4_feat->f_kobj);
5243 wait_for_completion(&ext4_feat->f_kobj_unregister);
5244 kfree(ext4_feat);
5245}
5246
e9e3bcec
ES		 5247/* Shared across all ext4 file systems */
5248wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5249struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5250
5dabfc78 5251static int __init ext4_init_fs(void)
ac27a0ec 5252{
e9e3bcec 5253 int i, err;
c9de560d 5254
07c0c5d8
AV		 5255 ext4_li_info = NULL;
5256 mutex_init(&ext4_li_mtx);
5257
12e9b892 5258 ext4_check_flag_values();
e9e3bcec
ES		 5259
5260 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5261 mutex_init(&ext4__aio_mutex[i]);
5262 init_waitqueue_head(&ext4__ioend_wq[i]);
5263 }
5264
5dabfc78 5265 err = ext4_init_pageio();
6fd058f7
TT		 5266 if (err)
5267 return err;
5dabfc78 5268 err = ext4_init_system_zone();
bd2d0210 5269 if (err)
d44651d0 5270 goto out6;
3197ebdb
TT		 5271 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5272 if (!ext4_kset)
dd68314c 5273 goto out5;
d44651d0 5274 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5275
5276 err = ext4_init_feat_adverts();
dd68314c
TT		 5277 if (err)
5278 goto out4;
857ac889 5279
5dabfc78 5280 err = ext4_init_mballoc();
ac27a0ec 5281 if (err)
6fd058f7 5282 goto out3;
c9de560d 5283
5dabfc78 5284 err = ext4_init_xattr();
c9de560d
AT		 5285 if (err)
5286 goto out2;
ac27a0ec
DK		 5287 err = init_inodecache();
5288 if (err)
5289 goto out1;
24b58424 5290 register_as_ext3();
2035e776 5291 register_as_ext2();
03010a33 5292 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5293 if (err)
5294 goto out;
bfff6873 5295
ac27a0ec
DK		 5296 return 0;
5297out:
24b58424
TT		 5298 unregister_as_ext2();
5299 unregister_as_ext3();
ac27a0ec
DK		 5300 destroy_inodecache();
5301out1:
5dabfc78 5302 ext4_exit_xattr();
c9de560d 5303out2:
5dabfc78 5304 ext4_exit_mballoc();
6fd058f7 5305out3:
8f021222 5306 ext4_exit_feat_adverts();
dd68314c 5307out4:
d44651d0
FJ		 5308 if (ext4_proc_root)
5309 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5310 kset_unregister(ext4_kset);
d44651d0 5311out5:
5dabfc78 5312 ext4_exit_system_zone();
d44651d0 5313out6:
5dabfc78 5314 ext4_exit_pageio();
ac27a0ec
DK		 5315 return err;
5316}
5317
5dabfc78 5318static void __exit ext4_exit_fs(void)
ac27a0ec 5319{
bfff6873 5320 ext4_destroy_lazyinit_thread();
24b58424
TT		 5321 unregister_as_ext2();
5322 unregister_as_ext3();
03010a33 5323 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5324 destroy_inodecache();
5dabfc78
TT		 5325 ext4_exit_xattr();
5326 ext4_exit_mballoc();
8f021222 5327 ext4_exit_feat_adverts();
9f6200bb 5328 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5329 kset_unregister(ext4_kset);
5dabfc78
TT		 5330 ext4_exit_system_zone();
5331 ext4_exit_pageio();
ac27a0ec
DK		 5332}
5333
5334MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5335MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5336MODULE_LICENSE("GPL");
5dabfc78
TT		 5337module_init(ext4_init_fs)
5338module_exit(ext4_exit_fs)
Linux kernel - Deliverables
This page took 1.048073 seconds and 5 git commands to generate.