projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: introduce ext4_kvmalloc(), ext4_kzalloc(), and ext4_kvfree()
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451
CH		 47#include "ext4.h"
48#include "ext4_jbd2.h"
ac27a0ec
DK		 49#include "xattr.h"
50#include "acl.h"
3661d286 51#include "mballoc.h"
ac27a0ec 52
9bffad1e
TT		 53#define CREATE_TRACE_POINTS
54#include <trace/events/ext4.h>
55
1f109d5a 56static struct proc_dir_entry *ext4_proc_root;
3197ebdb 57static struct kset *ext4_kset;
0b75a840
LC		 58static struct ext4_lazy_init *ext4_li_info;
59static struct mutex ext4_li_mtx;
60static struct ext4_features *ext4_feat;
9f6200bb 61
617ba13b 62static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 63 unsigned long journal_devnum);
e2d67052 64static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 65static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
617ba13b 69static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 70static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 71 char nbuf[16]);
2b2d6d01
TT		 72static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 74static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 75static void ext4_write_super(struct super_block *sb);
c4be0c1d 76static int ext4_freeze(struct super_block *sb);
152a0836
AV		 77static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
2035e776
TT		 79static inline int ext2_feature_set_ok(struct super_block *sb);
80static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 81static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 82static void ext4_destroy_lazyinit_thread(void);
83static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 84static void ext4_clear_request_list(void);
ac27a0ec 85
2035e776
TT		 86#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87static struct file_system_type ext2_fs_type = {
88 .owner = THIS_MODULE,
89 .name = "ext2",
90 .mount = ext4_mount,
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
93};
94#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
95#else
96#define IS_EXT2_SB(sb) (0)
97#endif
98
99
ba69f9ab
JK		 100#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
101static struct file_system_type ext3_fs_type = {
102 .owner = THIS_MODULE,
103 .name = "ext3",
152a0836 104 .mount = ext4_mount,
ba69f9ab
JK		 105 .kill_sb = kill_block_super,
106 .fs_flags = FS_REQUIRES_DEV,
107};
108#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
109#else
110#define IS_EXT3_SB(sb) (0)
111#endif
bd81d8ee 112
9933fc0a
TT		 113void *ext4_kvmalloc(size_t size, gfp_t flags)
114{
115 void *ret;
116
117 ret = kmalloc(size, flags);
118 if (!ret)
119 ret = __vmalloc(size, flags, PAGE_KERNEL);
120 return ret;
121}
122
123void *ext4_kvzalloc(size_t size, gfp_t flags)
124{
125 void *ret;
126
127 ret = kmalloc(size, flags);
128 if (!ret)
129 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
130 return ret;
131}
132
133void ext4_kvfree(void *ptr)
134{
135 if (is_vmalloc_addr(ptr))
136 vfree(ptr);
137 else
138 kfree(ptr);
139
140}
141
8fadc143
AR		 142ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
143 struct ext4_group_desc *bg)
bd81d8ee 144{
3a14589c 145 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 146 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 147 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 148}
149
8fadc143
AR		 150ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
151 struct ext4_group_desc *bg)
bd81d8ee 152{
5272f837 153 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 154 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 155 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 156}
157
8fadc143
AR		 158ext4_fsblk_t ext4_inode_table(struct super_block *sb,
159 struct ext4_group_desc *bg)
bd81d8ee 160{
5272f837 161 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 162 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 163 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 164}
165
560671a0
AK		 166__u32 ext4_free_blks_count(struct super_block *sb,
167 struct ext4_group_desc *bg)
168{
169 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
170 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 171 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 172}
173
174__u32 ext4_free_inodes_count(struct super_block *sb,
175 struct ext4_group_desc *bg)
176{
177 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
178 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 179 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 180}
181
182__u32 ext4_used_dirs_count(struct super_block *sb,
183 struct ext4_group_desc *bg)
184{
185 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
186 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 187 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 188}
189
190__u32 ext4_itable_unused_count(struct super_block *sb,
191 struct ext4_group_desc *bg)
192{
193 return le16_to_cpu(bg->bg_itable_unused_lo) |
194 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 195 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 196}
197
8fadc143
AR		 198void ext4_block_bitmap_set(struct super_block *sb,
199 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 200{
3a14589c 201 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 202 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
203 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 204}
205
8fadc143
AR		 206void ext4_inode_bitmap_set(struct super_block *sb,
207 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 208{
5272f837 209 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 210 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
211 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 212}
213
8fadc143
AR		 214void ext4_inode_table_set(struct super_block *sb,
215 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 216{
5272f837 217 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 218 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
219 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 220}
221
560671a0
AK		 222void ext4_free_blks_set(struct super_block *sb,
223 struct ext4_group_desc *bg, __u32 count)
224{
225 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
226 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
227 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
228}
229
230void ext4_free_inodes_set(struct super_block *sb,
231 struct ext4_group_desc *bg, __u32 count)
232{
233 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
234 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
235 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
236}
237
238void ext4_used_dirs_set(struct super_block *sb,
239 struct ext4_group_desc *bg, __u32 count)
240{
241 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
242 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
243 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
244}
245
246void ext4_itable_unused_set(struct super_block *sb,
247 struct ext4_group_desc *bg, __u32 count)
248{
249 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
250 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
251 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
252}
253
d3d1faf6
CW		 254
255/* Just increment the non-pointer handle value */
256static handle_t *ext4_get_nojournal(void)
257{
258 handle_t *handle = current->journal_info;
259 unsigned long ref_cnt = (unsigned long)handle;
260
261 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
262
263 ref_cnt++;
264 handle = (handle_t *)ref_cnt;
265
266 current->journal_info = handle;
267 return handle;
268}
269
270
271/* Decrement the non-pointer handle value */
272static void ext4_put_nojournal(handle_t *handle)
273{
274 unsigned long ref_cnt = (unsigned long)handle;
275
276 BUG_ON(ref_cnt == 0);
277
278 ref_cnt--;
279 handle = (handle_t *)ref_cnt;
280
281 current->journal_info = handle;
282}
283
ac27a0ec 284/*
dab291af 285 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 286 *
287 * The only special thing we need to do here is to make sure that all
288 * journal_end calls result in the superblock being marked dirty, so
289 * that sync() will call the filesystem's write_super callback if
290 * appropriate.
be4f27d3
YY		 291 *
292 * To avoid j_barrier hold in userspace when a user calls freeze(),
293 * ext4 prevents a new handle from being started by s_frozen, which
294 * is in an upper layer.
ac27a0ec 295 */
617ba13b 296handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 297{
298 journal_t *journal;
be4f27d3 299 handle_t *handle;
ac27a0ec 300
12706394 301 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 302 if (sb->s_flags & MS_RDONLY)
303 return ERR_PTR(-EROFS);
304
617ba13b 305 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 306 handle = ext4_journal_current_handle();
307
308 /*
309 * If a handle has been started, it should be allowed to
310 * finish, otherwise deadlock could happen between freeze
311 * and others(e.g. truncate) due to the restart of the
312 * journal handle if the filesystem is forzen and active
313 * handles are not stopped.
314 */
315 if (!handle)
316 vfs_check_frozen(sb, SB_FREEZE_TRANS);
317
318 if (!journal)
319 return ext4_get_nojournal();
320 /*
321 * Special case here: if the journal has aborted behind our
322 * backs (eg. EIO in the commit thread), then we still need to
323 * take the FS itself readonly cleanly.
324 */
325 if (is_journal_aborted(journal)) {
326 ext4_abort(sb, "Detected aborted journal");
327 return ERR_PTR(-EROFS);
ac27a0ec 328 }
be4f27d3 329 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 330}
331
332/*
333 * The only special thing we need to do here is to make sure that all
dab291af 334 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 335 * that sync() will call the filesystem's write_super callback if
336 * appropriate.
337 */
c398eda0 338int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 339{
340 struct super_block *sb;
341 int err;
342 int rc;
343
0390131b 344 if (!ext4_handle_valid(handle)) {
d3d1faf6 345 ext4_put_nojournal(handle);
0390131b
FM		 346 return 0;
347 }
ac27a0ec
DK		 348 sb = handle->h_transaction->t_journal->j_private;
349 err = handle->h_err;
dab291af 350 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 351
352 if (!err)
353 err = rc;
354 if (err)
c398eda0 355 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 356 return err;
357}
358
90c7201b
TT		 359void ext4_journal_abort_handle(const char *caller, unsigned int line,
360 const char *err_fn, struct buffer_head *bh,
361 handle_t *handle, int err)
ac27a0ec
DK		 362{
363 char nbuf[16];
617ba13b 364 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 365
0390131b
FM		 366 BUG_ON(!ext4_handle_valid(handle));
367
ac27a0ec
DK		 368 if (bh)
369 BUFFER_TRACE(bh, "abort");
370
371 if (!handle->h_err)
372 handle->h_err = err;
373
374 if (is_handle_aborted(handle))
375 return;
376
90c7201b
TT		 377 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
378 caller, line, errstr, err_fn);
ac27a0ec 379
dab291af 380 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 381}
382
1c13d5c0
TT		 383static void __save_error_info(struct super_block *sb, const char *func,
384 unsigned int line)
385{
386 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
387
388 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
389 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
390 es->s_last_error_time = cpu_to_le32(get_seconds());
391 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
392 es->s_last_error_line = cpu_to_le32(line);
393 if (!es->s_first_error_time) {
394 es->s_first_error_time = es->s_last_error_time;
395 strncpy(es->s_first_error_func, func,
396 sizeof(es->s_first_error_func));
397 es->s_first_error_line = cpu_to_le32(line);
398 es->s_first_error_ino = es->s_last_error_ino;
399 es->s_first_error_block = es->s_last_error_block;
400 }
66e61a9e
TT		 401 /*
402 * Start the daily error reporting function if it hasn't been
403 * started already
404 */
405 if (!es->s_error_count)
406 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 407 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
408}
409
410static void save_error_info(struct super_block *sb, const char *func,
411 unsigned int line)
412{
413 __save_error_info(sb, func, line);
414 ext4_commit_super(sb, 1);
415}
416
417
ac27a0ec
DK		 418/* Deal with the reporting of failure conditions on a filesystem such as
419 * inconsistencies detected or read IO failures.
420 *
421 * On ext2, we can store the error state of the filesystem in the
617ba13b 422 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 423 * write ordering constraints on the superblock which prevent us from
424 * writing it out straight away; and given that the journal is about to
425 * be aborted, we can't rely on the current, or future, transactions to
426 * write out the superblock safely.
427 *
dab291af 428 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 429 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 430 * that error until we've noted it down and cleared it.
431 */
432
617ba13b 433static void ext4_handle_error(struct super_block *sb)
ac27a0ec 434{
ac27a0ec
DK		 435 if (sb->s_flags & MS_RDONLY)
436 return;
437
2b2d6d01 438 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 439 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 440
4ab2f15b 441 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 442 if (journal)
dab291af 443 jbd2_journal_abort(journal, -EIO);
ac27a0ec 444 }
2b2d6d01 445 if (test_opt(sb, ERRORS_RO)) {
b31e1552 446 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 447 sb->s_flags |= MS_RDONLY;
448 }
ac27a0ec 449 if (test_opt(sb, ERRORS_PANIC))
617ba13b 450 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 451 sb->s_id);
452}
453
12062ddd 454void __ext4_error(struct super_block *sb, const char *function,
c398eda0 455 unsigned int line, const char *fmt, ...)
ac27a0ec 456{
0ff2ea7d 457 struct va_format vaf;
ac27a0ec
DK		 458 va_list args;
459
460 va_start(args, fmt);
0ff2ea7d
JP		 461 vaf.fmt = fmt;
462 vaf.va = &args;
463 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
464 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec
DK		 465 va_end(args);
466
617ba13b 467 ext4_handle_error(sb);
ac27a0ec
DK		 468}
469
c398eda0
TT		 470void ext4_error_inode(struct inode *inode, const char *function,
471 unsigned int line, ext4_fsblk_t block,
273df556
FM		 472 const char *fmt, ...)
473{
474 va_list args;
f7c21177 475 struct va_format vaf;
1c13d5c0 476 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 477
1c13d5c0
TT		 478 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
479 es->s_last_error_block = cpu_to_le64(block);
480 save_error_info(inode->i_sb, function, line);
273df556 481 va_start(args, fmt);
f7c21177
TT		 482 vaf.fmt = fmt;
483 vaf.va = &args;
c398eda0
TT		 484 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
485 inode->i_sb->s_id, function, line, inode->i_ino);
486 if (block)
f7c21177
TT		 487 printk(KERN_CONT "block %llu: ", block);
488 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
273df556
FM		 489 va_end(args);
490
491 ext4_handle_error(inode->i_sb);
492}
493
c398eda0 494void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 495 unsigned int line, ext4_fsblk_t block,
496 const char *fmt, ...)
273df556
FM		 497{
498 va_list args;
f7c21177 499 struct va_format vaf;
1c13d5c0 500 struct ext4_super_block *es;
273df556
FM		 501 struct inode *inode = file->f_dentry->d_inode;
502 char pathname[80], *path;
503
1c13d5c0
TT		 504 es = EXT4_SB(inode->i_sb)->s_es;
505 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
506 save_error_info(inode->i_sb, function, line);
273df556 507 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 508 if (IS_ERR(path))
273df556
FM		 509 path = "(unknown)";
510 printk(KERN_CRIT
f7c21177
TT		 511 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
512 inode->i_sb->s_id, function, line, inode->i_ino);
513 if (block)
514 printk(KERN_CONT "block %llu: ", block);
515 va_start(args, fmt);
516 vaf.fmt = fmt;
517 vaf.va = &args;
518 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
273df556
FM		 519 va_end(args);
520
521 ext4_handle_error(inode->i_sb);
522}
523
2b2d6d01 524static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 525 char nbuf[16])
526{
527 char *errstr = NULL;
528
529 switch (errno) {
530 case -EIO:
531 errstr = "IO failure";
532 break;
533 case -ENOMEM:
534 errstr = "Out of memory";
535 break;
536 case -EROFS:
78f1ddbb
TT		 537 if (!sb || (EXT4_SB(sb)->s_journal &&
538 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 539 errstr = "Journal has aborted";
540 else
541 errstr = "Readonly filesystem";
542 break;
543 default:
544 /* If the caller passed in an extra buffer for unknown
545 * errors, textualise them now. Else we just return
546 * NULL. */
547 if (nbuf) {
548 /* Check for truncated error codes... */
549 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
550 errstr = nbuf;
551 }
552 break;
553 }
554
555 return errstr;
556}
557
617ba13b 558/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 559 * automatically and invokes the appropriate error response. */
560
c398eda0
TT		 561void __ext4_std_error(struct super_block *sb, const char *function,
562 unsigned int line, int errno)
ac27a0ec
DK		 563{
564 char nbuf[16];
565 const char *errstr;
566
567 /* Special case: if the error is EROFS, and we're not already
568 * inside a transaction, then there's really no point in logging
569 * an error. */
570 if (errno == -EROFS && journal_current_handle() == NULL &&
571 (sb->s_flags & MS_RDONLY))
572 return;
573
617ba13b 574 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 575 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
576 sb->s_id, function, line, errstr);
1c13d5c0 577 save_error_info(sb, function, line);
ac27a0ec 578
617ba13b 579 ext4_handle_error(sb);
ac27a0ec
DK		 580}
581
582/*
617ba13b 583 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 584 * abort function may be used to deal with unrecoverable failures such
585 * as journal IO errors or ENOMEM at a critical moment in log management.
586 *
587 * We unconditionally force the filesystem into an ABORT|READONLY state,
588 * unless the error response on the fs has been set to panic in which
589 * case we take the easy way out and panic immediately.
590 */
591
c67d859e 592void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 593 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 594{
595 va_list args;
596
1c13d5c0 597 save_error_info(sb, function, line);
ac27a0ec 598 va_start(args, fmt);
c398eda0
TT		 599 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
600 function, line);
ac27a0ec
DK		 601 vprintk(fmt, args);
602 printk("\n");
603 va_end(args);
604
1c13d5c0
TT		 605 if ((sb->s_flags & MS_RDONLY) == 0) {
606 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
607 sb->s_flags |= MS_RDONLY;
608 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
609 if (EXT4_SB(sb)->s_journal)
610 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
611 save_error_info(sb, function, line);
612 }
ac27a0ec 613 if (test_opt(sb, ERRORS_PANIC))
617ba13b 614 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 615}
616
0ff2ea7d 617void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 618{
0ff2ea7d 619 struct va_format vaf;
b31e1552
ES		 620 va_list args;
621
622 va_start(args, fmt);
0ff2ea7d
JP		 623 vaf.fmt = fmt;
624 vaf.va = &args;
625 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 626 va_end(args);
627}
628
12062ddd 629void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 630 unsigned int line, const char *fmt, ...)
ac27a0ec 631{
0ff2ea7d 632 struct va_format vaf;
ac27a0ec
DK		 633 va_list args;
634
635 va_start(args, fmt);
0ff2ea7d
JP		 636 vaf.fmt = fmt;
637 vaf.va = &args;
638 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
639 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 640 va_end(args);
641}
642
e29136f8
TT		 643void __ext4_grp_locked_error(const char *function, unsigned int line,
644 struct super_block *sb, ext4_group_t grp,
645 unsigned long ino, ext4_fsblk_t block,
646 const char *fmt, ...)
5d1b1b3f
AK		 647__releases(bitlock)
648__acquires(bitlock)
649{
0ff2ea7d 650 struct va_format vaf;
5d1b1b3f
AK		 651 va_list args;
652 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
653
1c13d5c0
TT		 654 es->s_last_error_ino = cpu_to_le32(ino);
655 es->s_last_error_block = cpu_to_le64(block);
656 __save_error_info(sb, function, line);
0ff2ea7d 657
5d1b1b3f 658 va_start(args, fmt);
0ff2ea7d
JP		 659
660 vaf.fmt = fmt;
661 vaf.va = &args;
21149d61 662 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 663 sb->s_id, function, line, grp);
664 if (ino)
0ff2ea7d 665 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 666 if (block)
0ff2ea7d
JP		 667 printk(KERN_CONT "block %llu:", (unsigned long long) block);
668 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 669 va_end(args);
670
671 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 672 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 673 return;
674 }
1c13d5c0 675
5d1b1b3f
AK		 676 ext4_unlock_group(sb, grp);
677 ext4_handle_error(sb);
678 /*
679 * We only get here in the ERRORS_RO case; relocking the group
680 * may be dangerous, but nothing bad will happen since the
681 * filesystem will have already been marked read/only and the
682 * journal has been aborted. We return 1 as a hint to callers
683 * who might what to use the return value from
25985edc 684 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 685 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
686 * aggressively from the ext4 function in question, with a
687 * more appropriate error code.
688 */
689 ext4_lock_group(sb, grp);
690 return;
691}
692
617ba13b 693void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 694{
617ba13b 695 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 696
617ba13b 697 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 698 return;
699
12062ddd 700 ext4_warning(sb,
ac27a0ec
DK		 701 "updating to rev %d because of new feature flag, "
702 "running e2fsck is recommended",
617ba13b 703 EXT4_DYNAMIC_REV);
ac27a0ec 704
617ba13b
MC		 705 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
706 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
707 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 708 /* leave es->s_feature_*compat flags alone */
709 /* es->s_uuid will be set by e2fsck if empty */
710
711 /*
712 * The rest of the superblock fields should be zero, and if not it
713 * means they are likely already in use, so leave them alone. We
714 * can leave it up to e2fsck to clean up any inconsistencies there.
715 */
716}
717
718/*
719 * Open the external journal device
720 */
b31e1552 721static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 722{
723 struct block_device *bdev;
724 char b[BDEVNAME_SIZE];
725
d4d77629 726 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 727 if (IS_ERR(bdev))
728 goto fail;
729 return bdev;
730
731fail:
b31e1552 732 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 733 __bdevname(dev, b), PTR_ERR(bdev));
734 return NULL;
735}
736
737/*
738 * Release the journal device
739 */
617ba13b 740static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 741{
e525fd89 742 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 743}
744
617ba13b 745static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 746{
747 struct block_device *bdev;
748 int ret = -ENODEV;
749
750 bdev = sbi->journal_bdev;
751 if (bdev) {
617ba13b 752 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 753 sbi->journal_bdev = NULL;
754 }
755 return ret;
756}
757
758static inline struct inode *orphan_list_entry(struct list_head *l)
759{
617ba13b 760 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 761}
762
617ba13b 763static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 764{
765 struct list_head *l;
766
b31e1552
ES		 767 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
768 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 769
770 printk(KERN_ERR "sb_info orphan list:\n");
771 list_for_each(l, &sbi->s_orphan) {
772 struct inode *inode = orphan_list_entry(l);
773 printk(KERN_ERR " "
774 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
775 inode->i_sb->s_id, inode->i_ino, inode,
776 inode->i_mode, inode->i_nlink,
777 NEXT_ORPHAN(inode));
778 }
779}
780
2b2d6d01 781static void ext4_put_super(struct super_block *sb)
ac27a0ec 782{
617ba13b
MC		 783 struct ext4_sb_info *sbi = EXT4_SB(sb);
784 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 785 int i, err;
ac27a0ec 786
857ac889 787 ext4_unregister_li_request(sb);
e0ccfd95
CH		 788 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
789
4c0425ff
MC		 790 flush_workqueue(sbi->dio_unwritten_wq);
791 destroy_workqueue(sbi->dio_unwritten_wq);
792
a9e220f8 793 lock_super(sb);
8c85e125 794 if (sb->s_dirt)
ebc1ac16 795 ext4_commit_super(sb, 1);
8c85e125 796
0390131b
FM		 797 if (sbi->s_journal) {
798 err = jbd2_journal_destroy(sbi->s_journal);
799 sbi->s_journal = NULL;
800 if (err < 0)
c67d859e 801 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 802 }
d4edac31 803
a1c6c569 804 del_timer(&sbi->s_err_report);
d4edac31
JB		 805 ext4_release_system_zone(sb);
806 ext4_mb_release(sb);
807 ext4_ext_release(sb);
808 ext4_xattr_put_super(sb);
809
ac27a0ec 810 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 811 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 812 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 813 ext4_commit_super(sb, 1);
ac27a0ec 814 }
240799cd 815 if (sbi->s_proc) {
9f6200bb 816 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 817 }
3197ebdb 818 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 819
820 for (i = 0; i < sbi->s_gdb_count; i++)
821 brelse(sbi->s_group_desc[i]);
822 kfree(sbi->s_group_desc);
9933fc0a 823 ext4_kvfree(sbi->s_flex_groups);
ac27a0ec
DK		 824 percpu_counter_destroy(&sbi->s_freeblocks_counter);
825 percpu_counter_destroy(&sbi->s_freeinodes_counter);
826 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 827 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 828 brelse(sbi->s_sbh);
829#ifdef CONFIG_QUOTA
830 for (i = 0; i < MAXQUOTAS; i++)
831 kfree(sbi->s_qf_names[i]);
832#endif
833
834 /* Debugging code just in case the in-memory inode orphan list
835 * isn't empty. The on-disk one can be non-empty if we've
836 * detected an error and taken the fs readonly, but the
837 * in-memory list had better be clean by this point. */
838 if (!list_empty(&sbi->s_orphan))
839 dump_orphan_list(sb, sbi);
840 J_ASSERT(list_empty(&sbi->s_orphan));
841
f98393a6 842 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 843 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
844 /*
845 * Invalidate the journal device's buffers. We don't want them
846 * floating about in memory - the physical journal device may
847 * hotswapped, and it breaks the `ro-after' testing code.
848 */
849 sync_blockdev(sbi->journal_bdev);
f98393a6 850 invalidate_bdev(sbi->journal_bdev);
617ba13b 851 ext4_blkdev_remove(sbi);
ac27a0ec 852 }
c5e06d10
JL		 853 if (sbi->s_mmp_tsk)
854 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 855 sb->s_fs_info = NULL;
3197ebdb
TT		 856 /*
857 * Now that we are completely done shutting down the
858 * superblock, we need to actually destroy the kobject.
859 */
3197ebdb
TT		 860 unlock_super(sb);
861 kobject_put(&sbi->s_kobj);
862 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 863 kfree(sbi->s_blockgroup_lock);
ac27a0ec 864 kfree(sbi);
ac27a0ec
DK		 865}
866
e18b890b 867static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 868
869/*
870 * Called inside transaction, so use GFP_NOFS
871 */
617ba13b 872static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 873{
617ba13b 874 struct ext4_inode_info *ei;
ac27a0ec 875
e6b4f8da 876 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 877 if (!ei)
878 return NULL;
0b8e58a1 879
ac27a0ec 880 ei->vfs_inode.i_version = 1;
91246c00 881 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 882 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 883 INIT_LIST_HEAD(&ei->i_prealloc_list);
884 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 885 ei->i_reserved_data_blocks = 0;
886 ei->i_reserved_meta_blocks = 0;
887 ei->i_allocated_meta_blocks = 0;
9d0be502 888 ei->i_da_metadata_calc_len = 0;
d2a17637 889 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 890#ifdef CONFIG_QUOTA
891 ei->i_reserved_quota = 0;
892#endif
8aefcd55 893 ei->jinode = NULL;
c7064ef1 894 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 895 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 896 ei->cur_aio_dio = NULL;
b436b9be
JK		 897 ei->i_sync_tid = 0;
898 ei->i_datasync_tid = 0;
f7ad6d2e 899 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 900 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 901
ac27a0ec
DK		 902 return &ei->vfs_inode;
903}
904
7ff9c073
TT		 905static int ext4_drop_inode(struct inode *inode)
906{
907 int drop = generic_drop_inode(inode);
908
909 trace_ext4_drop_inode(inode, drop);
910 return drop;
911}
912
fa0d7e3d
NP		 913static void ext4_i_callback(struct rcu_head *head)
914{
915 struct inode *inode = container_of(head, struct inode, i_rcu);
916 INIT_LIST_HEAD(&inode->i_dentry);
917 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
918}
919
617ba13b 920static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 921{
f7ad6d2e 922 ext4_ioend_wait(inode);
9f7dd93d 923 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 924 ext4_msg(inode->i_sb, KERN_ERR,
925 "Inode %lu (%p): orphan list check failed!",
926 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 927 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
928 EXT4_I(inode), sizeof(struct ext4_inode_info),
929 true);
930 dump_stack();
931 }
fa0d7e3d 932 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 933}
934
51cc5068 935static void init_once(void *foo)
ac27a0ec 936{
617ba13b 937 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 938
a35afb83 939 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 940#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 941 init_rwsem(&ei->xattr_sem);
ac27a0ec 942#endif
0e855ac8 943 init_rwsem(&ei->i_data_sem);
a35afb83 944 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 945}
946
947static int init_inodecache(void)
948{
617ba13b
MC		 949 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
950 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 951 0, (SLAB_RECLAIM_ACCOUNT|
952 SLAB_MEM_SPREAD),
20c2df83 953 init_once);
617ba13b 954 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 955 return -ENOMEM;
956 return 0;
957}
958
959static void destroy_inodecache(void)
960{
617ba13b 961 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 962}
963
0930fcc1 964void ext4_clear_inode(struct inode *inode)
ac27a0ec 965{
0930fcc1
AV		 966 invalidate_inode_buffers(inode);
967 end_writeback(inode);
9f754758 968 dquot_drop(inode);
c2ea3fde 969 ext4_discard_preallocations(inode);
8aefcd55
TT		 970 if (EXT4_I(inode)->jinode) {
971 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
972 EXT4_I(inode)->jinode);
973 jbd2_free_inode(EXT4_I(inode)->jinode);
974 EXT4_I(inode)->jinode = NULL;
975 }
ac27a0ec
DK		 976}
977
2b2d6d01
TT		 978static inline void ext4_show_quota_options(struct seq_file *seq,
979 struct super_block *sb)
ac27a0ec
DK		 980{
981#if defined(CONFIG_QUOTA)
617ba13b 982 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 983
5a20bdfc
JK		 984 if (sbi->s_jquota_fmt) {
985 char *fmtname = "";
986
987 switch (sbi->s_jquota_fmt) {
988 case QFMT_VFS_OLD:
989 fmtname = "vfsold";
990 break;
991 case QFMT_VFS_V0:
992 fmtname = "vfsv0";
993 break;
994 case QFMT_VFS_V1:
995 fmtname = "vfsv1";
996 break;
997 }
998 seq_printf(seq, ",jqfmt=%s", fmtname);
999 }
ac27a0ec
DK		 1000
1001 if (sbi->s_qf_names[USRQUOTA])
1002 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1003
1004 if (sbi->s_qf_names[GRPQUOTA])
1005 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1006
482a7425 1007 if (test_opt(sb, USRQUOTA))
ac27a0ec
DK		 1008 seq_puts(seq, ",usrquota");
1009
482a7425 1010 if (test_opt(sb, GRPQUOTA))
ac27a0ec
DK		 1011 seq_puts(seq, ",grpquota");
1012#endif
1013}
1014
d9c9bef1
MS		 1015/*
1016 * Show an option if
1017 * - it's set to a non-default value OR
1018 * - if the per-sb default is different from the global default
1019 */
617ba13b 1020static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 1021{
aa22df2c
AK		 1022 int def_errors;
1023 unsigned long def_mount_opts;
ac27a0ec 1024 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 1025 struct ext4_sb_info *sbi = EXT4_SB(sb);
1026 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 1027
1028 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 1029 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 1030
1031 if (sbi->s_sb_block != 1)
1032 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1033 if (test_opt(sb, MINIX_DF))
1034 seq_puts(seq, ",minixdf");
aa22df2c 1035 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 1036 seq_puts(seq, ",grpid");
1037 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1038 seq_puts(seq, ",nogrpid");
1039 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1040 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1041 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1042 }
1043 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1044 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1045 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1046 }
bb4f397a 1047 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 1048 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 1049 def_errors == EXT4_ERRORS_CONTINUE) {
1050 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 1051 }
1052 }
aa22df2c 1053 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 1054 seq_puts(seq, ",errors=continue");
aa22df2c 1055 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 1056 seq_puts(seq, ",errors=panic");
aa22df2c 1057 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 1058 seq_puts(seq, ",nouid32");
aa22df2c 1059 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 1060 seq_puts(seq, ",debug");
1061 if (test_opt(sb, OLDALLOC))
1062 seq_puts(seq, ",oldalloc");
03010a33 1063#ifdef CONFIG_EXT4_FS_XATTR
ea663336 1064 if (test_opt(sb, XATTR_USER))
d9c9bef1 1065 seq_puts(seq, ",user_xattr");
ea663336 1066 if (!test_opt(sb, XATTR_USER))
d9c9bef1 1067 seq_puts(seq, ",nouser_xattr");
d9c9bef1 1068#endif
03010a33 1069#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 1070 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 1071 seq_puts(seq, ",acl");
1072 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1073 seq_puts(seq, ",noacl");
1074#endif
30773840 1075 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 1076 seq_printf(seq, ",commit=%u",
1077 (unsigned) (sbi->s_commit_interval / HZ));
1078 }
30773840
TT		 1079 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1080 seq_printf(seq, ",min_batch_time=%u",
1081 (unsigned) sbi->s_min_batch_time);
1082 }
1083 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1084 seq_printf(seq, ",max_batch_time=%u",
1085 (unsigned) sbi->s_min_batch_time);
1086 }
1087
571640ca
ES		 1088 /*
1089 * We're changing the default of barrier mount option, so
1090 * let's always display its mount state so it's clear what its
1091 * status is.
1092 */
1093 seq_puts(seq, ",barrier=");
1094 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 1095 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1096 seq_puts(seq, ",journal_async_commit");
39a4bade
JK		 1097 else if (test_opt(sb, JOURNAL_CHECKSUM))
1098 seq_puts(seq, ",journal_checksum");
25ec56b5
JNC		 1099 if (test_opt(sb, I_VERSION))
1100 seq_puts(seq, ",i_version");
8b67f04a
TT		 1101 if (!test_opt(sb, DELALLOC) &&
1102 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
dd919b98
AK		 1103 seq_puts(seq, ",nodelalloc");
1104
6fd7a467
TT		 1105 if (!test_opt(sb, MBLK_IO_SUBMIT))
1106 seq_puts(seq, ",nomblk_io_submit");
cb45bbe4
MS		 1107 if (sbi->s_stripe)
1108 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 1109 /*
1110 * journal mode get enabled in different ways
1111 * So just print the value even if we didn't specify it
1112 */
617ba13b 1113 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 1114 seq_puts(seq, ",data=journal");
617ba13b 1115 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 1116 seq_puts(seq, ",data=ordered");
617ba13b 1117 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 1118 seq_puts(seq, ",data=writeback");
1119
240799cd
TT		 1120 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1121 seq_printf(seq, ",inode_readahead_blks=%u",
1122 sbi->s_inode_readahead_blks);
1123
5bf5683a
HK		 1124 if (test_opt(sb, DATA_ERR_ABORT))
1125 seq_puts(seq, ",data_err=abort");
1126
afd4672d 1127 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 1128 seq_puts(seq, ",noauto_da_alloc");
afd4672d 1129
8b67f04a 1130 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
5328e635
ES		 1131 seq_puts(seq, ",discard");
1132
e3bb52ae
ES		 1133 if (test_opt(sb, NOLOAD))
1134 seq_puts(seq, ",norecovery");
1135
744692dc
JZ		 1136 if (test_opt(sb, DIOREAD_NOLOCK))
1137 seq_puts(seq, ",dioread_nolock");
1138
8b67f04a
TT		 1139 if (test_opt(sb, BLOCK_VALIDITY) &&
1140 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1141 seq_puts(seq, ",block_validity");
1142
bfff6873
LC		 1143 if (!test_opt(sb, INIT_INODE_TABLE))
1144 seq_puts(seq, ",noinit_inode_table");
51ce6511 1145 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
bfff6873
LC		 1146 seq_printf(seq, ",init_inode_table=%u",
1147 (unsigned) sbi->s_li_wait_mult);
1148
617ba13b 1149 ext4_show_quota_options(seq, sb);
0b8e58a1 1150
ac27a0ec
DK		 1151 return 0;
1152}
1153
1b961ac0 1154static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1155 u64 ino, u32 generation)
ac27a0ec 1156{
ac27a0ec 1157 struct inode *inode;
ac27a0ec 1158
617ba13b 1159 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1160 return ERR_PTR(-ESTALE);
617ba13b 1161 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1162 return ERR_PTR(-ESTALE);
1163
1164 /* iget isn't really right if the inode is currently unallocated!!
1165 *
617ba13b 1166 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1167 * deleted, so we should be safe.
1168 *
1169 * Currently we don't know the generation for parent directory, so
1170 * a generation of 0 means "accept any"
1171 */
1d1fe1ee
DH		 1172 inode = ext4_iget(sb, ino);
1173 if (IS_ERR(inode))
1174 return ERR_CAST(inode);
1175 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1176 iput(inode);
1177 return ERR_PTR(-ESTALE);
1178 }
1b961ac0
CH		 1179
1180 return inode;
1181}
1182
1183static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1184 int fh_len, int fh_type)
1b961ac0
CH		 1185{
1186 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1187 ext4_nfs_get_inode);
1188}
1189
1190static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1191 int fh_len, int fh_type)
1b961ac0
CH		 1192{
1193 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1194 ext4_nfs_get_inode);
ac27a0ec
DK		 1195}
1196
c39a7f84
TO		 1197/*
1198 * Try to release metadata pages (indirect blocks, directories) which are
1199 * mapped via the block device. Since these pages could have journal heads
1200 * which would prevent try_to_free_buffers() from freeing them, we must use
1201 * jbd2 layer's try_to_free_buffers() function to release them.
1202 */
0b8e58a1
AD		 1203static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1204 gfp_t wait)
c39a7f84
TO		 1205{
1206 journal_t *journal = EXT4_SB(sb)->s_journal;
1207
1208 WARN_ON(PageChecked(page));
1209 if (!page_has_buffers(page))
1210 return 0;
1211 if (journal)
1212 return jbd2_journal_try_to_free_buffers(journal, page,
1213 wait & ~__GFP_WAIT);
1214 return try_to_free_buffers(page);
1215}
1216
ac27a0ec 1217#ifdef CONFIG_QUOTA
af5bc92d 1218#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1219#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1220
617ba13b
MC		 1221static int ext4_write_dquot(struct dquot *dquot);
1222static int ext4_acquire_dquot(struct dquot *dquot);
1223static int ext4_release_dquot(struct dquot *dquot);
1224static int ext4_mark_dquot_dirty(struct dquot *dquot);
1225static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1226static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1227 struct path *path);
ca0e05e4 1228static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1229static int ext4_quota_on_mount(struct super_block *sb, int type);
1230static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1231 size_t len, loff_t off);
617ba13b 1232static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1233 const char *data, size_t len, loff_t off);
1234
61e225dc 1235static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1236 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1237 .write_dquot = ext4_write_dquot,
1238 .acquire_dquot = ext4_acquire_dquot,
1239 .release_dquot = ext4_release_dquot,
1240 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1241 .write_info = ext4_write_info,
1242 .alloc_dquot = dquot_alloc,
1243 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1244};
1245
0d54b217 1246static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1247 .quota_on = ext4_quota_on,
ca0e05e4 1248 .quota_off = ext4_quota_off,
287a8095
CH		 1249 .quota_sync = dquot_quota_sync,
1250 .get_info = dquot_get_dqinfo,
1251 .set_info = dquot_set_dqinfo,
1252 .get_dqblk = dquot_get_dqblk,
1253 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1254};
1255#endif
1256
ee9b6d61 1257static const struct super_operations ext4_sops = {
617ba13b
MC		 1258 .alloc_inode = ext4_alloc_inode,
1259 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1260 .write_inode = ext4_write_inode,
1261 .dirty_inode = ext4_dirty_inode,
7ff9c073 1262 .drop_inode = ext4_drop_inode,
0930fcc1 1263 .evict_inode = ext4_evict_inode,
617ba13b 1264 .put_super = ext4_put_super,
617ba13b 1265 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1266 .freeze_fs = ext4_freeze,
1267 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1268 .statfs = ext4_statfs,
1269 .remount_fs = ext4_remount,
617ba13b 1270 .show_options = ext4_show_options,
ac27a0ec 1271#ifdef CONFIG_QUOTA
617ba13b
MC		 1272 .quota_read = ext4_quota_read,
1273 .quota_write = ext4_quota_write,
ac27a0ec 1274#endif
c39a7f84 1275 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1276};
1277
9ca92389
TT		 1278static const struct super_operations ext4_nojournal_sops = {
1279 .alloc_inode = ext4_alloc_inode,
1280 .destroy_inode = ext4_destroy_inode,
1281 .write_inode = ext4_write_inode,
1282 .dirty_inode = ext4_dirty_inode,
7ff9c073 1283 .drop_inode = ext4_drop_inode,
0930fcc1 1284 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1285 .write_super = ext4_write_super,
1286 .put_super = ext4_put_super,
1287 .statfs = ext4_statfs,
1288 .remount_fs = ext4_remount,
9ca92389
TT		 1289 .show_options = ext4_show_options,
1290#ifdef CONFIG_QUOTA
1291 .quota_read = ext4_quota_read,
1292 .quota_write = ext4_quota_write,
1293#endif
1294 .bdev_try_to_free_page = bdev_try_to_free_page,
1295};
1296
39655164 1297static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1298 .fh_to_dentry = ext4_fh_to_dentry,
1299 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1300 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1301};
1302
1303enum {
1304 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1305 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1306 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1307 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1308 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1309 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1310 Opt_journal_update, Opt_journal_dev,
818d276c 1311 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1312 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1313 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1314 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1315 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1316 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1317 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1318 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1319 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1320 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1321 Opt_dioread_nolock, Opt_dioread_lock,
5328e635 1322 Opt_discard, Opt_nodiscard,
bfff6873 1323 Opt_init_inode_table, Opt_noinit_inode_table,
ac27a0ec
DK		 1324};
1325
a447c093 1326static const match_table_t tokens = {
ac27a0ec
DK		 1327 {Opt_bsd_df, "bsddf"},
1328 {Opt_minix_df, "minixdf"},
1329 {Opt_grpid, "grpid"},
1330 {Opt_grpid, "bsdgroups"},
1331 {Opt_nogrpid, "nogrpid"},
1332 {Opt_nogrpid, "sysvgroups"},
1333 {Opt_resgid, "resgid=%u"},
1334 {Opt_resuid, "resuid=%u"},
1335 {Opt_sb, "sb=%u"},
1336 {Opt_err_cont, "errors=continue"},
1337 {Opt_err_panic, "errors=panic"},
1338 {Opt_err_ro, "errors=remount-ro"},
1339 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1340 {Opt_debug, "debug"},
1341 {Opt_oldalloc, "oldalloc"},
1342 {Opt_orlov, "orlov"},
1343 {Opt_user_xattr, "user_xattr"},
1344 {Opt_nouser_xattr, "nouser_xattr"},
1345 {Opt_acl, "acl"},
1346 {Opt_noacl, "noacl"},
ac27a0ec 1347 {Opt_noload, "noload"},
e3bb52ae 1348 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1349 {Opt_nobh, "nobh"},
1350 {Opt_bh, "bh"},
1351 {Opt_commit, "commit=%u"},
30773840
TT		 1352 {Opt_min_batch_time, "min_batch_time=%u"},
1353 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1354 {Opt_journal_update, "journal=update"},
ac27a0ec 1355 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1356 {Opt_journal_checksum, "journal_checksum"},
1357 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1358 {Opt_abort, "abort"},
1359 {Opt_data_journal, "data=journal"},
1360 {Opt_data_ordered, "data=ordered"},
1361 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1362 {Opt_data_err_abort, "data_err=abort"},
1363 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1364 {Opt_offusrjquota, "usrjquota="},
1365 {Opt_usrjquota, "usrjquota=%s"},
1366 {Opt_offgrpjquota, "grpjquota="},
1367 {Opt_grpjquota, "grpjquota=%s"},
1368 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1369 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1370 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1371 {Opt_grpquota, "grpquota"},
1372 {Opt_noquota, "noquota"},
1373 {Opt_quota, "quota"},
1374 {Opt_usrquota, "usrquota"},
1375 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1376 {Opt_barrier, "barrier"},
1377 {Opt_nobarrier, "nobarrier"},
25ec56b5 1378 {Opt_i_version, "i_version"},
c9de560d 1379 {Opt_stripe, "stripe=%u"},
ac27a0ec 1380 {Opt_resize, "resize"},
64769240 1381 {Opt_delalloc, "delalloc"},
dd919b98 1382 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1383 {Opt_mblk_io_submit, "mblk_io_submit"},
1384 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1385 {Opt_block_validity, "block_validity"},
1386 {Opt_noblock_validity, "noblock_validity"},
240799cd 1387 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1388 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1389 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1390 {Opt_auto_da_alloc, "auto_da_alloc"},
1391 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1392 {Opt_dioread_nolock, "dioread_nolock"},
1393 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1394 {Opt_discard, "discard"},
1395 {Opt_nodiscard, "nodiscard"},
bfff6873
LC		 1396 {Opt_init_inode_table, "init_itable=%u"},
1397 {Opt_init_inode_table, "init_itable"},
1398 {Opt_noinit_inode_table, "noinit_itable"},
f3f12faa 1399 {Opt_err, NULL},
ac27a0ec
DK		 1400};
1401
617ba13b 1402static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1403{
617ba13b 1404 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1405 char *options = (char *) *data;
1406
1407 if (!options || strncmp(options, "sb=", 3) != 0)
1408 return 1; /* Default location */
0b8e58a1 1409
ac27a0ec 1410 options += 3;
0b8e58a1 1411 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1412 sb_block = simple_strtoul(options, &options, 0);
1413 if (*options && *options != ',') {
4776004f 1414 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1415 (char *) *data);
1416 return 1;
1417 }
1418 if (*options == ',')
1419 options++;
1420 *data = (void *) options;
0b8e58a1 1421
ac27a0ec
DK		 1422 return sb_block;
1423}
1424
b3881f74 1425#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1426static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1427 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1428
56c50f11
DM		 1429#ifdef CONFIG_QUOTA
1430static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1431{
1432 struct ext4_sb_info *sbi = EXT4_SB(sb);
1433 char *qname;
1434
1435 if (sb_any_quota_loaded(sb) &&
1436 !sbi->s_qf_names[qtype]) {
1437 ext4_msg(sb, KERN_ERR,
1438 "Cannot change journaled "
1439 "quota options when quota turned on");
1440 return 0;
1441 }
1442 qname = match_strdup(args);
1443 if (!qname) {
1444 ext4_msg(sb, KERN_ERR,
1445 "Not enough memory for storing quotafile name");
1446 return 0;
1447 }
1448 if (sbi->s_qf_names[qtype] &&
1449 strcmp(sbi->s_qf_names[qtype], qname)) {
1450 ext4_msg(sb, KERN_ERR,
1451 "%s quota file already specified", QTYPE2NAME(qtype));
1452 kfree(qname);
1453 return 0;
1454 }
1455 sbi->s_qf_names[qtype] = qname;
1456 if (strchr(sbi->s_qf_names[qtype], '/')) {
1457 ext4_msg(sb, KERN_ERR,
1458 "quotafile must be on filesystem root");
1459 kfree(sbi->s_qf_names[qtype]);
1460 sbi->s_qf_names[qtype] = NULL;
1461 return 0;
1462 }
fd8c37ec 1463 set_opt(sb, QUOTA);
56c50f11
DM		 1464 return 1;
1465}
1466
1467static int clear_qf_name(struct super_block *sb, int qtype)
1468{
1469
1470 struct ext4_sb_info *sbi = EXT4_SB(sb);
1471
1472 if (sb_any_quota_loaded(sb) &&
1473 sbi->s_qf_names[qtype]) {
1474 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1475 " when quota turned on");
1476 return 0;
1477 }
1478 /*
1479 * The space will be released later when all options are confirmed
1480 * to be correct
1481 */
1482 sbi->s_qf_names[qtype] = NULL;
1483 return 1;
1484}
1485#endif
1486
2b2d6d01 1487static int parse_options(char *options, struct super_block *sb,
c3191067 1488 unsigned long *journal_devnum,
b3881f74 1489 unsigned int *journal_ioprio,
2b2d6d01 1490 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1491{
617ba13b 1492 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1493 char *p;
ac27a0ec
DK		 1494 substring_t args[MAX_OPT_ARGS];
1495 int data_opt = 0;
1496 int option;
1497#ifdef CONFIG_QUOTA
56c50f11 1498 int qfmt;
ac27a0ec
DK		 1499#endif
1500
1501 if (!options)
1502 return 1;
1503
2b2d6d01 1504 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1505 int token;
1506 if (!*p)
1507 continue;
1508
15121c18
ES		 1509 /*
1510 * Initialize args struct so we know whether arg was
1511 * found; some options take optional arguments.
1512 */
7dc57615 1513 args[0].to = args[0].from = NULL;
ac27a0ec
DK		 1514 token = match_token(p, tokens, args);
1515 switch (token) {
1516 case Opt_bsd_df:
437ca0fd 1517 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1518 clear_opt(sb, MINIX_DF);
ac27a0ec
DK		 1519 break;
1520 case Opt_minix_df:
437ca0fd 1521 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1522 set_opt(sb, MINIX_DF);
437ca0fd 1523
ac27a0ec
DK		 1524 break;
1525 case Opt_grpid:
437ca0fd 1526 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1527 set_opt(sb, GRPID);
437ca0fd 1528
ac27a0ec
DK		 1529 break;
1530 case Opt_nogrpid:
437ca0fd 1531 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1532 clear_opt(sb, GRPID);
437ca0fd 1533
ac27a0ec
DK		 1534 break;
1535 case Opt_resuid:
1536 if (match_int(&args[0], &option))
1537 return 0;
1538 sbi->s_resuid = option;
1539 break;
1540 case Opt_resgid:
1541 if (match_int(&args[0], &option))
1542 return 0;
1543 sbi->s_resgid = option;
1544 break;
1545 case Opt_sb:
1546 /* handled by get_sb_block() instead of here */
1547 /* *sb_block = match_int(&args[0]); */
1548 break;
1549 case Opt_err_panic:
fd8c37ec
TT		 1550 clear_opt(sb, ERRORS_CONT);
1551 clear_opt(sb, ERRORS_RO);
1552 set_opt(sb, ERRORS_PANIC);
ac27a0ec
DK		 1553 break;
1554 case Opt_err_ro:
fd8c37ec
TT		 1555 clear_opt(sb, ERRORS_CONT);
1556 clear_opt(sb, ERRORS_PANIC);
1557 set_opt(sb, ERRORS_RO);
ac27a0ec
DK		 1558 break;
1559 case Opt_err_cont:
fd8c37ec
TT		 1560 clear_opt(sb, ERRORS_RO);
1561 clear_opt(sb, ERRORS_PANIC);
1562 set_opt(sb, ERRORS_CONT);
ac27a0ec
DK		 1563 break;
1564 case Opt_nouid32:
fd8c37ec 1565 set_opt(sb, NO_UID32);
ac27a0ec 1566 break;
ac27a0ec 1567 case Opt_debug:
fd8c37ec 1568 set_opt(sb, DEBUG);
ac27a0ec
DK		 1569 break;
1570 case Opt_oldalloc:
fd8c37ec 1571 set_opt(sb, OLDALLOC);
ac27a0ec
DK		 1572 break;
1573 case Opt_orlov:
fd8c37ec 1574 clear_opt(sb, OLDALLOC);
ac27a0ec 1575 break;
03010a33 1576#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1577 case Opt_user_xattr:
fd8c37ec 1578 set_opt(sb, XATTR_USER);
ac27a0ec
DK		 1579 break;
1580 case Opt_nouser_xattr:
fd8c37ec 1581 clear_opt(sb, XATTR_USER);
ac27a0ec
DK		 1582 break;
1583#else
1584 case Opt_user_xattr:
1585 case Opt_nouser_xattr:
b31e1552 1586 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1587 break;
1588#endif
03010a33 1589#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec 1590 case Opt_acl:
fd8c37ec 1591 set_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1592 break;
1593 case Opt_noacl:
fd8c37ec 1594 clear_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1595 break;
1596#else
1597 case Opt_acl:
1598 case Opt_noacl:
b31e1552 1599 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1600 break;
1601#endif
ac27a0ec
DK		 1602 case Opt_journal_update:
1603 /* @@@ FIXME */
1604 /* Eventually we will want to be able to create
1605 a journal file here. For now, only allow the
1606 user to specify an existing inode to be the
1607 journal file. */
1608 if (is_remount) {
b31e1552
ES		 1609 ext4_msg(sb, KERN_ERR,
1610 "Cannot specify journal on remount");
ac27a0ec
DK		 1611 return 0;
1612 }
fd8c37ec 1613 set_opt(sb, UPDATE_JOURNAL);
ac27a0ec 1614 break;
ac27a0ec
DK		 1615 case Opt_journal_dev:
1616 if (is_remount) {
b31e1552
ES		 1617 ext4_msg(sb, KERN_ERR,
1618 "Cannot specify journal on remount");
ac27a0ec
DK		 1619 return 0;
1620 }
1621 if (match_int(&args[0], &option))
1622 return 0;
1623 *journal_devnum = option;
1624 break;
818d276c 1625 case Opt_journal_checksum:
fd8c37ec 1626 set_opt(sb, JOURNAL_CHECKSUM);
d4da6c9c 1627 break;
818d276c 1628 case Opt_journal_async_commit:
fd8c37ec
TT		 1629 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1630 set_opt(sb, JOURNAL_CHECKSUM);
818d276c 1631 break;
ac27a0ec 1632 case Opt_noload:
fd8c37ec 1633 set_opt(sb, NOLOAD);
ac27a0ec
DK		 1634 break;
1635 case Opt_commit:
1636 if (match_int(&args[0], &option))
1637 return 0;
1638 if (option < 0)
1639 return 0;
1640 if (option == 0)
cd02ff0b 1641 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1642 sbi->s_commit_interval = HZ * option;
1643 break;
30773840
TT		 1644 case Opt_max_batch_time:
1645 if (match_int(&args[0], &option))
1646 return 0;
1647 if (option < 0)
1648 return 0;
1649 if (option == 0)
1650 option = EXT4_DEF_MAX_BATCH_TIME;
1651 sbi->s_max_batch_time = option;
1652 break;
1653 case Opt_min_batch_time:
1654 if (match_int(&args[0], &option))
1655 return 0;
1656 if (option < 0)
1657 return 0;
1658 sbi->s_min_batch_time = option;
1659 break;
ac27a0ec 1660 case Opt_data_journal:
617ba13b 1661 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1662 goto datacheck;
1663 case Opt_data_ordered:
617ba13b 1664 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1665 goto datacheck;
1666 case Opt_data_writeback:
617ba13b 1667 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1668 datacheck:
1669 if (is_remount) {
482a7425 1670 if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1671 ext4_msg(sb, KERN_ERR,
1672 "Cannot change data mode on remount");
ac27a0ec
DK		 1673 return 0;
1674 }
1675 } else {
fd8c37ec 1676 clear_opt(sb, DATA_FLAGS);
ac27a0ec
DK		 1677 sbi->s_mount_opt |= data_opt;
1678 }
1679 break;
5bf5683a 1680 case Opt_data_err_abort:
fd8c37ec 1681 set_opt(sb, DATA_ERR_ABORT);
5bf5683a
HK		 1682 break;
1683 case Opt_data_err_ignore:
fd8c37ec 1684 clear_opt(sb, DATA_ERR_ABORT);
5bf5683a 1685 break;
ac27a0ec
DK		 1686#ifdef CONFIG_QUOTA
1687 case Opt_usrjquota:
56c50f11 1688 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1689 return 0;
56c50f11
DM		 1690 break;
1691 case Opt_grpjquota:
1692 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1693 return 0;
ac27a0ec
DK		 1694 break;
1695 case Opt_offusrjquota:
56c50f11
DM		 1696 if (!clear_qf_name(sb, USRQUOTA))
1697 return 0;
1698 break;
ac27a0ec 1699 case Opt_offgrpjquota:
56c50f11 1700 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1701 return 0;
ac27a0ec 1702 break;
56c50f11 1703
ac27a0ec 1704 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1705 qfmt = QFMT_VFS_OLD;
1706 goto set_qf_format;
ac27a0ec 1707 case Opt_jqfmt_vfsv0:
dfc5d03f 1708 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1709 goto set_qf_format;
1710 case Opt_jqfmt_vfsv1:
1711 qfmt = QFMT_VFS_V1;
dfc5d03f 1712set_qf_format:
17bd13b3 1713 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1714 sbi->s_jquota_fmt != qfmt) {
b31e1552 1715 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1716 "journaled quota options when "
b31e1552 1717 "quota turned on");
dfc5d03f
JK		 1718 return 0;
1719 }
1720 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1721 break;
1722 case Opt_quota:
1723 case Opt_usrquota:
fd8c37ec
TT		 1724 set_opt(sb, QUOTA);
1725 set_opt(sb, USRQUOTA);
ac27a0ec
DK		 1726 break;
1727 case Opt_grpquota:
fd8c37ec
TT		 1728 set_opt(sb, QUOTA);
1729 set_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1730 break;
1731 case Opt_noquota:
17bd13b3 1732 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1733 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1734 "options when quota turned on");
ac27a0ec
DK		 1735 return 0;
1736 }
fd8c37ec
TT		 1737 clear_opt(sb, QUOTA);
1738 clear_opt(sb, USRQUOTA);
1739 clear_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1740 break;
1741#else
1742 case Opt_quota:
1743 case Opt_usrquota:
1744 case Opt_grpquota:
b31e1552
ES		 1745 ext4_msg(sb, KERN_ERR,
1746 "quota options not supported");
cd59e7b9 1747 break;
ac27a0ec
DK		 1748 case Opt_usrjquota:
1749 case Opt_grpjquota:
1750 case Opt_offusrjquota:
1751 case Opt_offgrpjquota:
1752 case Opt_jqfmt_vfsold:
1753 case Opt_jqfmt_vfsv0:
5a20bdfc 1754 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1755 ext4_msg(sb, KERN_ERR,
1756 "journaled quota options not supported");
ac27a0ec
DK		 1757 break;
1758 case Opt_noquota:
1759 break;
1760#endif
1761 case Opt_abort:
4ab2f15b 1762 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1763 break;
06705bff 1764 case Opt_nobarrier:
fd8c37ec 1765 clear_opt(sb, BARRIER);
06705bff 1766 break;
ac27a0ec 1767 case Opt_barrier:
15121c18
ES		 1768 if (args[0].from) {
1769 if (match_int(&args[0], &option))
1770 return 0;
1771 } else
1772 option = 1; /* No argument, default to 1 */
ac27a0ec 1773 if (option)
fd8c37ec 1774 set_opt(sb, BARRIER);
ac27a0ec 1775 else
fd8c37ec 1776 clear_opt(sb, BARRIER);
ac27a0ec
DK		 1777 break;
1778 case Opt_ignore:
1779 break;
1780 case Opt_resize:
1781 if (!is_remount) {
b31e1552
ES		 1782 ext4_msg(sb, KERN_ERR,
1783 "resize option only available "
1784 "for remount");
ac27a0ec
DK		 1785 return 0;
1786 }
1787 if (match_int(&args[0], &option) != 0)
1788 return 0;
1789 *n_blocks_count = option;
1790 break;
1791 case Opt_nobh:
206f7ab4
CH		 1792 ext4_msg(sb, KERN_WARNING,
1793 "Ignoring deprecated nobh option");
ac27a0ec
DK		 1794 break;
1795 case Opt_bh:
206f7ab4
CH		 1796 ext4_msg(sb, KERN_WARNING,
1797 "Ignoring deprecated bh option");
ac27a0ec 1798 break;
25ec56b5 1799 case Opt_i_version:
fd8c37ec 1800 set_opt(sb, I_VERSION);
25ec56b5
JNC		 1801 sb->s_flags |= MS_I_VERSION;
1802 break;
dd919b98 1803 case Opt_nodelalloc:
fd8c37ec 1804 clear_opt(sb, DELALLOC);
dd919b98 1805 break;
1449032b 1806 case Opt_mblk_io_submit:
fd8c37ec 1807 set_opt(sb, MBLK_IO_SUBMIT);
1449032b
TT		 1808 break;
1809 case Opt_nomblk_io_submit:
fd8c37ec 1810 clear_opt(sb, MBLK_IO_SUBMIT);
1449032b 1811 break;
c9de560d
AT		 1812 case Opt_stripe:
1813 if (match_int(&args[0], &option))
1814 return 0;
1815 if (option < 0)
1816 return 0;
1817 sbi->s_stripe = option;
1818 break;
64769240 1819 case Opt_delalloc:
fd8c37ec 1820 set_opt(sb, DELALLOC);
64769240 1821 break;
6fd058f7 1822 case Opt_block_validity:
fd8c37ec 1823 set_opt(sb, BLOCK_VALIDITY);
6fd058f7
TT		 1824 break;
1825 case Opt_noblock_validity:
fd8c37ec 1826 clear_opt(sb, BLOCK_VALIDITY);
6fd058f7 1827 break;
240799cd
TT		 1828 case Opt_inode_readahead_blks:
1829 if (match_int(&args[0], &option))
1830 return 0;
1831 if (option < 0 || option > (1 << 30))
1832 return 0;
5dbd571d 1833 if (option && !is_power_of_2(option)) {
b31e1552
ES		 1834 ext4_msg(sb, KERN_ERR,
1835 "EXT4-fs: inode_readahead_blks"
1836 " must be a power of 2");
3197ebdb
TT		 1837 return 0;
1838 }
240799cd
TT		 1839 sbi->s_inode_readahead_blks = option;
1840 break;
b3881f74
TT		 1841 case Opt_journal_ioprio:
1842 if (match_int(&args[0], &option))
1843 return 0;
1844 if (option < 0 || option > 7)
1845 break;
1846 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1847 option);
1848 break;
06705bff 1849 case Opt_noauto_da_alloc:
fd8c37ec 1850 set_opt(sb, NO_AUTO_DA_ALLOC);
06705bff 1851 break;
afd4672d 1852 case Opt_auto_da_alloc:
15121c18
ES		 1853 if (args[0].from) {
1854 if (match_int(&args[0], &option))
1855 return 0;
1856 } else
1857 option = 1; /* No argument, default to 1 */
afd4672d 1858 if (option)
fd8c37ec 1859 clear_opt(sb, NO_AUTO_DA_ALLOC);
afd4672d 1860 else
fd8c37ec 1861 set_opt(sb,NO_AUTO_DA_ALLOC);
afd4672d 1862 break;
5328e635 1863 case Opt_discard:
fd8c37ec 1864 set_opt(sb, DISCARD);
5328e635
ES		 1865 break;
1866 case Opt_nodiscard:
fd8c37ec 1867 clear_opt(sb, DISCARD);
5328e635 1868 break;
744692dc 1869 case Opt_dioread_nolock:
fd8c37ec 1870 set_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 1871 break;
1872 case Opt_dioread_lock:
fd8c37ec 1873 clear_opt(sb, DIOREAD_NOLOCK);
744692dc 1874 break;
bfff6873 1875 case Opt_init_inode_table:
fd8c37ec 1876 set_opt(sb, INIT_INODE_TABLE);
bfff6873
LC		 1877 if (args[0].from) {
1878 if (match_int(&args[0], &option))
1879 return 0;
1880 } else
1881 option = EXT4_DEF_LI_WAIT_MULT;
1882 if (option < 0)
1883 return 0;
1884 sbi->s_li_wait_mult = option;
1885 break;
1886 case Opt_noinit_inode_table:
fd8c37ec 1887 clear_opt(sb, INIT_INODE_TABLE);
bfff6873 1888 break;
ac27a0ec 1889 default:
b31e1552
ES		 1890 ext4_msg(sb, KERN_ERR,
1891 "Unrecognized mount option \"%s\" "
1892 "or missing value", p);
ac27a0ec
DK		 1893 return 0;
1894 }
1895 }
1896#ifdef CONFIG_QUOTA
1897 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1898 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1899 clear_opt(sb, USRQUOTA);
ac27a0ec 1900
482a7425 1901 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1902 clear_opt(sb, GRPQUOTA);
ac27a0ec 1903
56c50f11 1904 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1905 ext4_msg(sb, KERN_ERR, "old and new quota "
1906 "format mixing");
ac27a0ec
DK		 1907 return 0;
1908 }
1909
1910 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1911 ext4_msg(sb, KERN_ERR, "journaled quota format "
1912 "not specified");
ac27a0ec
DK		 1913 return 0;
1914 }
1915 } else {
1916 if (sbi->s_jquota_fmt) {
b31e1552 1917 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1918 "specified with no journaling "
b31e1552 1919 "enabled");
ac27a0ec
DK		 1920 return 0;
1921 }
1922 }
1923#endif
1924 return 1;
1925}
1926
617ba13b 1927static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1928 int read_only)
1929{
617ba13b 1930 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1931 int res = 0;
1932
617ba13b 1933 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1934 ext4_msg(sb, KERN_ERR, "revision level too high, "
1935 "forcing read-only mode");
ac27a0ec
DK		 1936 res = MS_RDONLY;
1937 }
1938 if (read_only)
1939 return res;
617ba13b 1940 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1941 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1942 "running e2fsck is recommended");
617ba13b 1943 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1944 ext4_msg(sb, KERN_WARNING,
1945 "warning: mounting fs with errors, "
1946 "running e2fsck is recommended");
ed3ce80a 1947 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1948 le16_to_cpu(es->s_mnt_count) >=
1949 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1950 ext4_msg(sb, KERN_WARNING,
1951 "warning: maximal mount count reached, "
1952 "running e2fsck is recommended");
ac27a0ec
DK		 1953 else if (le32_to_cpu(es->s_checkinterval) &&
1954 (le32_to_cpu(es->s_lastcheck) +
1955 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1956 ext4_msg(sb, KERN_WARNING,
1957 "warning: checktime reached, "
1958 "running e2fsck is recommended");
0b8e58a1 1959 if (!sbi->s_journal)
0390131b 1960 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1961 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1962 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1963 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1964 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1965 ext4_update_dynamic_rev(sb);
0390131b
FM		 1966 if (sbi->s_journal)
1967 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1968
e2d67052 1969 ext4_commit_super(sb, 1);
ac27a0ec 1970 if (test_opt(sb, DEBUG))
a9df9a49 1971 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1972 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1973 sb->s_blocksize,
1974 sbi->s_groups_count,
617ba13b
MC		 1975 EXT4_BLOCKS_PER_GROUP(sb),
1976 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1977 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1978
7abc52c2 1979 cleancache_init_fs(sb);
ac27a0ec
DK		 1980 return res;
1981}
1982
772cb7c8
JS		 1983static int ext4_fill_flex_info(struct super_block *sb)
1984{
1985 struct ext4_sb_info *sbi = EXT4_SB(sb);
1986 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1987 ext4_group_t flex_group_count;
1988 ext4_group_t flex_group;
1989 int groups_per_flex = 0;
c5ca7c76 1990 size_t size;
772cb7c8
JS		 1991 int i;
1992
503358ae
TT		 1993 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1994 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1995
1996 if (groups_per_flex < 2) {
772cb7c8
JS		 1997 sbi->s_log_groups_per_flex = 0;
1998 return 1;
1999 }
2000
c62a11fd
FB		 2001 /* We allocate both existing and potentially added groups */
2002 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 2003 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2004 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 2005 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 2006 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 2007 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 2008 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2009 flex_group_count);
2010 goto failed;
772cb7c8 2011 }
772cb7c8 2012
772cb7c8 2013 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2014 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2015
2016 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2017 atomic_add(ext4_free_inodes_count(sb, gdp),
2018 &sbi->s_flex_groups[flex_group].free_inodes);
2019 atomic_add(ext4_free_blks_count(sb, gdp),
2020 &sbi->s_flex_groups[flex_group].free_blocks);
2021 atomic_add(ext4_used_dirs_count(sb, gdp),
2022 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2023 }
2024
2025 return 1;
2026failed:
2027 return 0;
2028}
2029
717d50e4
AD		 2030__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2031 struct ext4_group_desc *gdp)
2032{
2033 __u16 crc = 0;
2034
2035 if (sbi->s_es->s_feature_ro_compat &
2036 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2037 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2038 __le32 le_group = cpu_to_le32(block_group);
2039
2040 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2041 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2042 crc = crc16(crc, (__u8 *)gdp, offset);
2043 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2044 /* for checksum of struct ext4_group_desc do the rest...*/
2045 if ((sbi->s_es->s_feature_incompat &
2046 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2047 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2048 crc = crc16(crc, (__u8 *)gdp + offset,
2049 le16_to_cpu(sbi->s_es->s_desc_size) -
2050 offset);
2051 }
2052
2053 return cpu_to_le16(crc);
2054}
2055
2056int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2057 struct ext4_group_desc *gdp)
2058{
2059 if ((sbi->s_es->s_feature_ro_compat &
2060 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2061 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2062 return 0;
2063
2064 return 1;
2065}
2066
ac27a0ec 2067/* Called at mount-time, super-block is locked */
bfff6873
LC		 2068static int ext4_check_descriptors(struct super_block *sb,
2069 ext4_group_t *first_not_zeroed)
ac27a0ec 2070{
617ba13b
MC		 2071 struct ext4_sb_info *sbi = EXT4_SB(sb);
2072 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2073 ext4_fsblk_t last_block;
bd81d8ee
LV		 2074 ext4_fsblk_t block_bitmap;
2075 ext4_fsblk_t inode_bitmap;
2076 ext4_fsblk_t inode_table;
ce421581 2077 int flexbg_flag = 0;
bfff6873 2078 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2079
ce421581
JS		 2080 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2081 flexbg_flag = 1;
2082
af5bc92d 2083 ext4_debug("Checking group descriptors");
ac27a0ec 2084
197cd65a
AM		 2085 for (i = 0; i < sbi->s_groups_count; i++) {
2086 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2087
ce421581 2088 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2089 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2090 else
2091 last_block = first_block +
617ba13b 2092 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2093
bfff6873
LC		 2094 if ((grp == sbi->s_groups_count) &&
2095 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2096 grp = i;
2097
8fadc143 2098 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2099 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2100 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2101 "Block bitmap for group %u not in group "
b31e1552 2102 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2103 return 0;
2104 }
8fadc143 2105 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2106 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2107 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2108 "Inode bitmap for group %u not in group "
b31e1552 2109 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2110 return 0;
2111 }
8fadc143 2112 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2113 if (inode_table < first_block ||
2b2d6d01 2114 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2115 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2116 "Inode table for group %u not in group "
b31e1552 2117 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2118 return 0;
2119 }
955ce5f5 2120 ext4_lock_group(sb, i);
717d50e4 2121 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2122 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2123 "Checksum for group %u failed (%u!=%u)",
2124 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2125 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2126 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2127 ext4_unlock_group(sb, i);
8a266467 2128 return 0;
7ee1ec4c 2129 }
717d50e4 2130 }
955ce5f5 2131 ext4_unlock_group(sb, i);
ce421581
JS		 2132 if (!flexbg_flag)
2133 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2134 }
bfff6873
LC		 2135 if (NULL != first_not_zeroed)
2136 *first_not_zeroed = grp;
ac27a0ec 2137
bd81d8ee 2138 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 2139 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2140 return 1;
2141}
2142
617ba13b 2143/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2144 * the superblock) which were deleted from all directories, but held open by
2145 * a process at the time of a crash. We walk the list and try to delete these
2146 * inodes at recovery time (only with a read-write filesystem).
2147 *
2148 * In order to keep the orphan inode chain consistent during traversal (in
2149 * case of crash during recovery), we link each inode into the superblock
2150 * orphan list_head and handle it the same way as an inode deletion during
2151 * normal operation (which journals the operations for us).
2152 *
2153 * We only do an iget() and an iput() on each inode, which is very safe if we
2154 * accidentally point at an in-use or already deleted inode. The worst that
2155 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2156 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2157 * e2fsck was run on this filesystem, and it must have already done the orphan
2158 * inode cleanup for us, so we can safely abort without any further action.
2159 */
2b2d6d01
TT		 2160static void ext4_orphan_cleanup(struct super_block *sb,
2161 struct ext4_super_block *es)
ac27a0ec
DK		 2162{
2163 unsigned int s_flags = sb->s_flags;
2164 int nr_orphans = 0, nr_truncates = 0;
2165#ifdef CONFIG_QUOTA
2166 int i;
2167#endif
2168 if (!es->s_last_orphan) {
2169 jbd_debug(4, "no orphan inodes to clean up\n");
2170 return;
2171 }
2172
a8f48a95 2173 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2174 ext4_msg(sb, KERN_ERR, "write access "
2175 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2176 return;
2177 }
2178
d39195c3
AG		 2179 /* Check if feature set would not allow a r/w mount */
2180 if (!ext4_feature_set_ok(sb, 0)) {
2181 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2182 "unknown ROCOMPAT features");
2183 return;
2184 }
2185
617ba13b 2186 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2187 if (es->s_last_orphan)
2188 jbd_debug(1, "Errors on filesystem, "
2189 "clearing orphan list.\n");
2190 es->s_last_orphan = 0;
2191 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2192 return;
2193 }
2194
2195 if (s_flags & MS_RDONLY) {
b31e1552 2196 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2197 sb->s_flags &= ~MS_RDONLY;
2198 }
2199#ifdef CONFIG_QUOTA
2200 /* Needed for iput() to work correctly and not trash data */
2201 sb->s_flags |= MS_ACTIVE;
2202 /* Turn on quotas so that they are updated correctly */
2203 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2204 if (EXT4_SB(sb)->s_qf_names[i]) {
2205 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2206 if (ret < 0)
b31e1552
ES		 2207 ext4_msg(sb, KERN_ERR,
2208 "Cannot turn on journaled "
2209 "quota: error %d", ret);
ac27a0ec
DK		 2210 }
2211 }
2212#endif
2213
2214 while (es->s_last_orphan) {
2215 struct inode *inode;
2216
97bd42b9
JB		 2217 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2218 if (IS_ERR(inode)) {
ac27a0ec
DK		 2219 es->s_last_orphan = 0;
2220 break;
2221 }
2222
617ba13b 2223 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2224 dquot_initialize(inode);
ac27a0ec 2225 if (inode->i_nlink) {
b31e1552
ES		 2226 ext4_msg(sb, KERN_DEBUG,
2227 "%s: truncating inode %lu to %lld bytes",
46e665e9 2228 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2229 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2230 inode->i_ino, inode->i_size);
617ba13b 2231 ext4_truncate(inode);
ac27a0ec
DK		 2232 nr_truncates++;
2233 } else {
b31e1552
ES		 2234 ext4_msg(sb, KERN_DEBUG,
2235 "%s: deleting unreferenced inode %lu",
46e665e9 2236 __func__, inode->i_ino);
ac27a0ec
DK		 2237 jbd_debug(2, "deleting unreferenced inode %lu\n",
2238 inode->i_ino);
2239 nr_orphans++;
2240 }
2241 iput(inode); /* The delete magic happens here! */
2242 }
2243
2b2d6d01 2244#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2245
2246 if (nr_orphans)
b31e1552
ES		 2247 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2248 PLURAL(nr_orphans));
ac27a0ec 2249 if (nr_truncates)
b31e1552
ES		 2250 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2251 PLURAL(nr_truncates));
ac27a0ec
DK		 2252#ifdef CONFIG_QUOTA
2253 /* Turn quotas off */
2254 for (i = 0; i < MAXQUOTAS; i++) {
2255 if (sb_dqopt(sb)->files[i])
287a8095 2256 dquot_quota_off(sb, i);
ac27a0ec
DK		 2257 }
2258#endif
2259 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2260}
0b8e58a1 2261
cd2291a4
ES		 2262/*
2263 * Maximal extent format file size.
2264 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2265 * extent format containers, within a sector_t, and within i_blocks
2266 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2267 * so that won't be a limiting factor.
2268 *
f17722f9
LC		 2269 * However there is other limiting factor. We do store extents in the form
2270 * of starting block and length, hence the resulting length of the extent
2271 * covering maximum file size must fit into on-disk format containers as
2272 * well. Given that length is always by 1 unit bigger than max unit (because
2273 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2274 *
cd2291a4
ES		 2275 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2276 */
f287a1a5 2277static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2278{
2279 loff_t res;
2280 loff_t upper_limit = MAX_LFS_FILESIZE;
2281
2282 /* small i_blocks in vfs inode? */
f287a1a5 2283 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2284 /*
90c699a9 2285 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2286 * i_block represent total blocks in 512 bytes
2287 * 32 == size of vfs inode i_blocks * 8
2288 */
2289 upper_limit = (1LL << 32) - 1;
2290
2291 /* total blocks in file system block size */
2292 upper_limit >>= (blkbits - 9);
2293 upper_limit <<= blkbits;
2294 }
2295
f17722f9
LC		 2296 /*
2297 * 32-bit extent-start container, ee_block. We lower the maxbytes
2298 * by one fs block, so ee_len can cover the extent of maximum file
2299 * size
2300 */
2301 res = (1LL << 32) - 1;
cd2291a4 2302 res <<= blkbits;
cd2291a4
ES		 2303
2304 /* Sanity check against vm- & vfs- imposed limits */
2305 if (res > upper_limit)
2306 res = upper_limit;
2307
2308 return res;
2309}
ac27a0ec 2310
ac27a0ec 2311/*
cd2291a4 2312 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2313 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2314 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2315 */
f287a1a5 2316static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2317{
617ba13b 2318 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2319 int meta_blocks;
2320 loff_t upper_limit;
0b8e58a1
AD		 2321 /* This is calculated to be the largest file size for a dense, block
2322 * mapped file such that the file's total number of 512-byte sectors,
2323 * including data and all indirect blocks, does not exceed (2^48 - 1).
2324 *
2325 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2326 * number of 512-byte sectors of the file.
0fc1b451
AK		 2327 */
2328
f287a1a5 2329 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2330 /*
90c699a9 2331 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2332 * the inode i_block field represents total file blocks in
2333 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2334 */
2335 upper_limit = (1LL << 32) - 1;
2336
2337 /* total blocks in file system block size */
2338 upper_limit >>= (bits - 9);
2339
2340 } else {
8180a562
AK		 2341 /*
2342 * We use 48 bit ext4_inode i_blocks
2343 * With EXT4_HUGE_FILE_FL set the i_blocks
2344 * represent total number of blocks in
2345 * file system block size
2346 */
0fc1b451
AK		 2347 upper_limit = (1LL << 48) - 1;
2348
0fc1b451
AK		 2349 }
2350
2351 /* indirect blocks */
2352 meta_blocks = 1;
2353 /* double indirect blocks */
2354 meta_blocks += 1 + (1LL << (bits-2));
2355 /* tripple indirect blocks */
2356 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2357
2358 upper_limit -= meta_blocks;
2359 upper_limit <<= bits;
ac27a0ec
DK		 2360
2361 res += 1LL << (bits-2);
2362 res += 1LL << (2*(bits-2));
2363 res += 1LL << (3*(bits-2));
2364 res <<= bits;
2365 if (res > upper_limit)
2366 res = upper_limit;
0fc1b451
AK		 2367
2368 if (res > MAX_LFS_FILESIZE)
2369 res = MAX_LFS_FILESIZE;
2370
ac27a0ec
DK		 2371 return res;
2372}
2373
617ba13b 2374static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2375 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2376{
617ba13b 2377 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2378 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2379 int has_super = 0;
2380
2381 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2382
617ba13b 2383 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2384 nr < first_meta_bg)
70bbb3e0 2385 return logical_sb_block + nr + 1;
ac27a0ec 2386 bg = sbi->s_desc_per_block * nr;
617ba13b 2387 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2388 has_super = 1;
0b8e58a1 2389
617ba13b 2390 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2391}
2392
c9de560d
AT		 2393/**
2394 * ext4_get_stripe_size: Get the stripe size.
2395 * @sbi: In memory super block info
2396 *
2397 * If we have specified it via mount option, then
2398 * use the mount option value. If the value specified at mount time is
2399 * greater than the blocks per group use the super block value.
2400 * If the super block value is greater than blocks per group return 0.
2401 * Allocator needs it be less than blocks per group.
2402 *
2403 */
2404static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2405{
2406 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2407 unsigned long stripe_width =
2408 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2409 int ret;
c9de560d
AT		 2410
2411 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2412 ret = sbi->s_stripe;
2413 else if (stripe_width <= sbi->s_blocks_per_group)
2414 ret = stripe_width;
2415 else if (stride <= sbi->s_blocks_per_group)
2416 ret = stride;
2417 else
2418 ret = 0;
c9de560d 2419
3eb08658
DE		 2420 /*
2421 * If the stripe width is 1, this makes no sense and
2422 * we set it to 0 to turn off stripe handling code.
2423 */
2424 if (ret <= 1)
2425 ret = 0;
c9de560d 2426
3eb08658 2427 return ret;
c9de560d 2428}
ac27a0ec 2429
3197ebdb
TT		 2430/* sysfs supprt */
2431
2432struct ext4_attr {
2433 struct attribute attr;
2434 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2435 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2436 const char *, size_t);
2437 int offset;
2438};
2439
2440static int parse_strtoul(const char *buf,
2441 unsigned long max, unsigned long *value)
2442{
2443 char *endp;
2444
e7d2860b
AGR		 2445 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2446 endp = skip_spaces(endp);
3197ebdb
TT		 2447 if (*endp || *value > max)
2448 return -EINVAL;
2449
2450 return 0;
2451}
2452
2453static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2454 struct ext4_sb_info *sbi,
2455 char *buf)
2456{
2457 return snprintf(buf, PAGE_SIZE, "%llu\n",
2458 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2459}
2460
2461static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2462 struct ext4_sb_info *sbi, char *buf)
2463{
2464 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2465
f613dfcb
TT		 2466 if (!sb->s_bdev->bd_part)
2467 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2468 return snprintf(buf, PAGE_SIZE, "%lu\n",
2469 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2470 sbi->s_sectors_written_start) >> 1);
2471}
2472
2473static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2474 struct ext4_sb_info *sbi, char *buf)
2475{
2476 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2477
f613dfcb
TT		 2478 if (!sb->s_bdev->bd_part)
2479 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2480 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2481 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2482 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2483 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2484}
2485
77f4135f
VH		 2486static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2487 struct ext4_sb_info *sbi, char *buf)
2488{
2489 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2490}
2491
2492static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2493 struct ext4_sb_info *sbi, char *buf)
2494{
2495 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2496}
2497
3197ebdb
TT		 2498static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2499 struct ext4_sb_info *sbi,
2500 const char *buf, size_t count)
2501{
2502 unsigned long t;
2503
2504 if (parse_strtoul(buf, 0x40000000, &t))
2505 return -EINVAL;
2506
5dbd571d 2507 if (t && !is_power_of_2(t))
3197ebdb
TT		 2508 return -EINVAL;
2509
2510 sbi->s_inode_readahead_blks = t;
2511 return count;
2512}
2513
2514static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2515 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2516{
2517 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2518
2519 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2520}
2521
2522static ssize_t sbi_ui_store(struct ext4_attr *a,
2523 struct ext4_sb_info *sbi,
2524 const char *buf, size_t count)
2525{
2526 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2527 unsigned long t;
2528
2529 if (parse_strtoul(buf, 0xffffffff, &t))
2530 return -EINVAL;
2531 *ui = t;
2532 return count;
2533}
2534
2535#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2536static struct ext4_attr ext4_attr_##_name = { \
2537 .attr = {.name = __stringify(_name), .mode = _mode }, \
2538 .show = _show, \
2539 .store = _store, \
2540 .offset = offsetof(struct ext4_sb_info, _elname), \
2541}
2542#define EXT4_ATTR(name, mode, show, store) \
2543static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2544
857ac889 2545#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2546#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2547#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2548#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2549 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2550#define ATTR_LIST(name) &ext4_attr_##name.attr
2551
2552EXT4_RO_ATTR(delayed_allocation_blocks);
2553EXT4_RO_ATTR(session_write_kbytes);
2554EXT4_RO_ATTR(lifetime_write_kbytes);
77f4135f
VH		 2555EXT4_RO_ATTR(extent_cache_hits);
2556EXT4_RO_ATTR(extent_cache_misses);
3197ebdb
TT		 2557EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2558 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2559EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2560EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2561EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2562EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2563EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2564EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2565EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2566EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2567
2568static struct attribute *ext4_attrs[] = {
2569 ATTR_LIST(delayed_allocation_blocks),
2570 ATTR_LIST(session_write_kbytes),
2571 ATTR_LIST(lifetime_write_kbytes),
77f4135f
VH		 2572 ATTR_LIST(extent_cache_hits),
2573 ATTR_LIST(extent_cache_misses),
3197ebdb 2574 ATTR_LIST(inode_readahead_blks),
11013911 2575 ATTR_LIST(inode_goal),
3197ebdb
TT		 2576 ATTR_LIST(mb_stats),
2577 ATTR_LIST(mb_max_to_scan),
2578 ATTR_LIST(mb_min_to_scan),
2579 ATTR_LIST(mb_order2_req),
2580 ATTR_LIST(mb_stream_req),
2581 ATTR_LIST(mb_group_prealloc),
55138e0b 2582 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2583 NULL,
2584};
2585
857ac889
LC		 2586/* Features this copy of ext4 supports */
2587EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2588EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2589
2590static struct attribute *ext4_feat_attrs[] = {
2591 ATTR_LIST(lazy_itable_init),
27ee40df 2592 ATTR_LIST(batched_discard),
857ac889
LC		 2593 NULL,
2594};
2595
3197ebdb
TT		 2596static ssize_t ext4_attr_show(struct kobject *kobj,
2597 struct attribute *attr, char *buf)
2598{
2599 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2600 s_kobj);
2601 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2602
2603 return a->show ? a->show(a, sbi, buf) : 0;
2604}
2605
2606static ssize_t ext4_attr_store(struct kobject *kobj,
2607 struct attribute *attr,
2608 const char *buf, size_t len)
2609{
2610 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2611 s_kobj);
2612 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2613
2614 return a->store ? a->store(a, sbi, buf, len) : 0;
2615}
2616
2617static void ext4_sb_release(struct kobject *kobj)
2618{
2619 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2620 s_kobj);
2621 complete(&sbi->s_kobj_unregister);
2622}
2623
52cf25d0 2624static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2625 .show = ext4_attr_show,
2626 .store = ext4_attr_store,
2627};
2628
2629static struct kobj_type ext4_ktype = {
2630 .default_attrs = ext4_attrs,
2631 .sysfs_ops = &ext4_attr_ops,
2632 .release = ext4_sb_release,
2633};
2634
857ac889
LC		 2635static void ext4_feat_release(struct kobject *kobj)
2636{
2637 complete(&ext4_feat->f_kobj_unregister);
2638}
2639
2640static struct kobj_type ext4_feat_ktype = {
2641 .default_attrs = ext4_feat_attrs,
2642 .sysfs_ops = &ext4_attr_ops,
2643 .release = ext4_feat_release,
2644};
2645
a13fb1a4
ES		 2646/*
2647 * Check whether this filesystem can be mounted based on
2648 * the features present and the RDONLY/RDWR mount requested.
2649 * Returns 1 if this filesystem can be mounted as requested,
2650 * 0 if it cannot be.
2651 */
2652static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2653{
2654 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2655 ext4_msg(sb, KERN_ERR,
2656 "Couldn't mount because of "
2657 "unsupported optional features (%x)",
2658 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2659 ~EXT4_FEATURE_INCOMPAT_SUPP));
2660 return 0;
2661 }
2662
2663 if (readonly)
2664 return 1;
2665
2666 /* Check that feature set is OK for a read-write mount */
2667 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2668 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2669 "unsupported optional features (%x)",
2670 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2671 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2672 return 0;
2673 }
2674 /*
2675 * Large file size enabled file system can only be mounted
2676 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2677 */
2678 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2679 if (sizeof(blkcnt_t) < sizeof(u64)) {
2680 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2681 "cannot be mounted RDWR without "
2682 "CONFIG_LBDAF");
2683 return 0;
2684 }
2685 }
2686 return 1;
2687}
2688
66e61a9e
TT		 2689/*
2690 * This function is called once a day if we have errors logged
2691 * on the file system
2692 */
2693static void print_daily_error_info(unsigned long arg)
2694{
2695 struct super_block *sb = (struct super_block *) arg;
2696 struct ext4_sb_info *sbi;
2697 struct ext4_super_block *es;
2698
2699 sbi = EXT4_SB(sb);
2700 es = sbi->s_es;
2701
2702 if (es->s_error_count)
2703 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2704 le32_to_cpu(es->s_error_count));
2705 if (es->s_first_error_time) {
2706 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2707 sb->s_id, le32_to_cpu(es->s_first_error_time),
2708 (int) sizeof(es->s_first_error_func),
2709 es->s_first_error_func,
2710 le32_to_cpu(es->s_first_error_line));
2711 if (es->s_first_error_ino)
2712 printk(": inode %u",
2713 le32_to_cpu(es->s_first_error_ino));
2714 if (es->s_first_error_block)
2715 printk(": block %llu", (unsigned long long)
2716 le64_to_cpu(es->s_first_error_block));
2717 printk("\n");
2718 }
2719 if (es->s_last_error_time) {
2720 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2721 sb->s_id, le32_to_cpu(es->s_last_error_time),
2722 (int) sizeof(es->s_last_error_func),
2723 es->s_last_error_func,
2724 le32_to_cpu(es->s_last_error_line));
2725 if (es->s_last_error_ino)
2726 printk(": inode %u",
2727 le32_to_cpu(es->s_last_error_ino));
2728 if (es->s_last_error_block)
2729 printk(": block %llu", (unsigned long long)
2730 le64_to_cpu(es->s_last_error_block));
2731 printk("\n");
2732 }
2733 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2734}
2735
bfff6873
LC		 2736/* Find next suitable group and run ext4_init_inode_table */
2737static int ext4_run_li_request(struct ext4_li_request *elr)
2738{
2739 struct ext4_group_desc *gdp = NULL;
2740 ext4_group_t group, ngroups;
2741 struct super_block *sb;
2742 unsigned long timeout = 0;
2743 int ret = 0;
2744
2745 sb = elr->lr_super;
2746 ngroups = EXT4_SB(sb)->s_groups_count;
2747
2748 for (group = elr->lr_next_group; group < ngroups; group++) {
2749 gdp = ext4_get_group_desc(sb, group, NULL);
2750 if (!gdp) {
2751 ret = 1;
2752 break;
2753 }
2754
2755 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2756 break;
2757 }
2758
2759 if (group == ngroups)
2760 ret = 1;
2761
2762 if (!ret) {
2763 timeout = jiffies;
2764 ret = ext4_init_inode_table(sb, group,
2765 elr->lr_timeout ? 0 : 1);
2766 if (elr->lr_timeout == 0) {
51ce6511
LC		 2767 timeout = (jiffies - timeout) *
2768 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2769 elr->lr_timeout = timeout;
2770 }
2771 elr->lr_next_sched = jiffies + elr->lr_timeout;
2772 elr->lr_next_group = group + 1;
2773 }
2774
2775 return ret;
2776}
2777
2778/*
2779 * Remove lr_request from the list_request and free the
4ed5c033 2780 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2781 */
2782static void ext4_remove_li_request(struct ext4_li_request *elr)
2783{
2784 struct ext4_sb_info *sbi;
2785
2786 if (!elr)
2787 return;
2788
2789 sbi = elr->lr_sbi;
2790
2791 list_del(&elr->lr_request);
2792 sbi->s_li_request = NULL;
2793 kfree(elr);
2794}
2795
2796static void ext4_unregister_li_request(struct super_block *sb)
2797{
1bb933fb
LC		 2798 mutex_lock(&ext4_li_mtx);
2799 if (!ext4_li_info) {
2800 mutex_unlock(&ext4_li_mtx);
bfff6873 2801 return;
1bb933fb 2802 }
bfff6873
LC		 2803
2804 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2805 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2806 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2807 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2808}
2809
8f1f7453
ES		 2810static struct task_struct *ext4_lazyinit_task;
2811
bfff6873
LC		 2812/*
2813 * This is the function where ext4lazyinit thread lives. It walks
2814 * through the request list searching for next scheduled filesystem.
2815 * When such a fs is found, run the lazy initialization request
2816 * (ext4_rn_li_request) and keep track of the time spend in this
2817 * function. Based on that time we compute next schedule time of
2818 * the request. When walking through the list is complete, compute
2819 * next waking time and put itself into sleep.
2820 */
2821static int ext4_lazyinit_thread(void *arg)
2822{
2823 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2824 struct list_head *pos, *n;
2825 struct ext4_li_request *elr;
4ed5c033 2826 unsigned long next_wakeup, cur;
bfff6873
LC		 2827
2828 BUG_ON(NULL == eli);
2829
bfff6873
LC		 2830cont_thread:
2831 while (true) {
2832 next_wakeup = MAX_JIFFY_OFFSET;
2833
2834 mutex_lock(&eli->li_list_mtx);
2835 if (list_empty(&eli->li_request_list)) {
2836 mutex_unlock(&eli->li_list_mtx);
2837 goto exit_thread;
2838 }
2839
2840 list_for_each_safe(pos, n, &eli->li_request_list) {
2841 elr = list_entry(pos, struct ext4_li_request,
2842 lr_request);
2843
b2c78cd0
TT		 2844 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2845 if (ext4_run_li_request(elr) != 0) {
2846 /* error, remove the lazy_init job */
2847 ext4_remove_li_request(elr);
2848 continue;
2849 }
bfff6873
LC		 2850 }
2851
2852 if (time_before(elr->lr_next_sched, next_wakeup))
2853 next_wakeup = elr->lr_next_sched;
2854 }
2855 mutex_unlock(&eli->li_list_mtx);
2856
2857 if (freezing(current))
2858 refrigerator();
2859
4ed5c033
LC		 2860 cur = jiffies;
2861 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2862 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2863 cond_resched();
2864 continue;
2865 }
2866
4ed5c033
LC		 2867 schedule_timeout_interruptible(next_wakeup - cur);
2868
8f1f7453
ES		 2869 if (kthread_should_stop()) {
2870 ext4_clear_request_list();
2871 goto exit_thread;
2872 }
bfff6873
LC		 2873 }
2874
2875exit_thread:
2876 /*
2877 * It looks like the request list is empty, but we need
2878 * to check it under the li_list_mtx lock, to prevent any
2879 * additions into it, and of course we should lock ext4_li_mtx
2880 * to atomically free the list and ext4_li_info, because at
2881 * this point another ext4 filesystem could be registering
2882 * new one.
2883 */
2884 mutex_lock(&ext4_li_mtx);
2885 mutex_lock(&eli->li_list_mtx);
2886 if (!list_empty(&eli->li_request_list)) {
2887 mutex_unlock(&eli->li_list_mtx);
2888 mutex_unlock(&ext4_li_mtx);
2889 goto cont_thread;
2890 }
2891 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2892 kfree(ext4_li_info);
2893 ext4_li_info = NULL;
2894 mutex_unlock(&ext4_li_mtx);
2895
2896 return 0;
2897}
2898
2899static void ext4_clear_request_list(void)
2900{
2901 struct list_head *pos, *n;
2902 struct ext4_li_request *elr;
2903
2904 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2905 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2906 elr = list_entry(pos, struct ext4_li_request,
2907 lr_request);
2908 ext4_remove_li_request(elr);
2909 }
2910 mutex_unlock(&ext4_li_info->li_list_mtx);
2911}
2912
2913static int ext4_run_lazyinit_thread(void)
2914{
8f1f7453
ES		 2915 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2916 ext4_li_info, "ext4lazyinit");
2917 if (IS_ERR(ext4_lazyinit_task)) {
2918 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2919 ext4_clear_request_list();
bfff6873
LC		 2920 kfree(ext4_li_info);
2921 ext4_li_info = NULL;
2922 printk(KERN_CRIT "EXT4: error %d creating inode table "
2923 "initialization thread\n",
2924 err);
2925 return err;
2926 }
2927 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2928 return 0;
2929}
2930
2931/*
2932 * Check whether it make sense to run itable init. thread or not.
2933 * If there is at least one uninitialized inode table, return
2934 * corresponding group number, else the loop goes through all
2935 * groups and return total number of groups.
2936 */
2937static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2938{
2939 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2940 struct ext4_group_desc *gdp = NULL;
2941
2942 for (group = 0; group < ngroups; group++) {
2943 gdp = ext4_get_group_desc(sb, group, NULL);
2944 if (!gdp)
2945 continue;
2946
2947 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2948 break;
2949 }
2950
2951 return group;
2952}
2953
2954static int ext4_li_info_new(void)
2955{
2956 struct ext4_lazy_init *eli = NULL;
2957
2958 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2959 if (!eli)
2960 return -ENOMEM;
2961
bfff6873
LC		 2962 INIT_LIST_HEAD(&eli->li_request_list);
2963 mutex_init(&eli->li_list_mtx);
2964
bfff6873
LC		 2965 eli->li_state |= EXT4_LAZYINIT_QUIT;
2966
2967 ext4_li_info = eli;
2968
2969 return 0;
2970}
2971
2972static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2973 ext4_group_t start)
2974{
2975 struct ext4_sb_info *sbi = EXT4_SB(sb);
2976 struct ext4_li_request *elr;
2977 unsigned long rnd;
2978
2979 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2980 if (!elr)
2981 return NULL;
2982
2983 elr->lr_super = sb;
2984 elr->lr_sbi = sbi;
2985 elr->lr_next_group = start;
2986
2987 /*
2988 * Randomize first schedule time of the request to
2989 * spread the inode table initialization requests
2990 * better.
2991 */
2992 get_random_bytes(&rnd, sizeof(rnd));
2993 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2994 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2995
2996 return elr;
2997}
2998
2999static int ext4_register_li_request(struct super_block *sb,
3000 ext4_group_t first_not_zeroed)
3001{
3002 struct ext4_sb_info *sbi = EXT4_SB(sb);
3003 struct ext4_li_request *elr;
3004 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3005 int ret = 0;
bfff6873 3006
51ce6511
LC		 3007 if (sbi->s_li_request != NULL) {
3008 /*
3009 * Reset timeout so it can be computed again, because
3010 * s_li_wait_mult might have changed.
3011 */
3012 sbi->s_li_request->lr_timeout = 0;
beed5ecb 3013 return 0;
51ce6511 3014 }
bfff6873
LC		 3015
3016 if (first_not_zeroed == ngroups ||
3017 (sb->s_flags & MS_RDONLY) ||
55ff3840 3018 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 3019 return 0;
bfff6873
LC		 3020
3021 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 3022 if (!elr)
3023 return -ENOMEM;
bfff6873
LC		 3024
3025 mutex_lock(&ext4_li_mtx);
3026
3027 if (NULL == ext4_li_info) {
3028 ret = ext4_li_info_new();
3029 if (ret)
3030 goto out;
3031 }
3032
3033 mutex_lock(&ext4_li_info->li_list_mtx);
3034 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3035 mutex_unlock(&ext4_li_info->li_list_mtx);
3036
3037 sbi->s_li_request = elr;
46e4690b
TM		 3038 /*
3039 * set elr to NULL here since it has been inserted to
3040 * the request_list and the removal and free of it is
3041 * handled by ext4_clear_request_list from now on.
3042 */
3043 elr = NULL;
bfff6873
LC		 3044
3045 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3046 ret = ext4_run_lazyinit_thread();
3047 if (ret)
3048 goto out;
3049 }
bfff6873 3050out:
beed5ecb
NK		 3051 mutex_unlock(&ext4_li_mtx);
3052 if (ret)
bfff6873 3053 kfree(elr);
bfff6873
LC		 3054 return ret;
3055}
3056
3057/*
3058 * We do not need to lock anything since this is called on
3059 * module unload.
3060 */
3061static void ext4_destroy_lazyinit_thread(void)
3062{
3063 /*
3064 * If thread exited earlier
3065 * there's nothing to be done.
3066 */
8f1f7453 3067 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3068 return;
3069
8f1f7453 3070 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3071}
3072
2b2d6d01 3073static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 3074 __releases(kernel_lock)
3075 __acquires(kernel_lock)
ac27a0ec 3076{
d4c402d9 3077 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3078 struct buffer_head *bh;
617ba13b
MC		 3079 struct ext4_super_block *es = NULL;
3080 struct ext4_sb_info *sbi;
3081 ext4_fsblk_t block;
3082 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3083 ext4_fsblk_t logical_sb_block;
ac27a0ec 3084 unsigned long offset = 0;
ac27a0ec
DK		 3085 unsigned long journal_devnum = 0;
3086 unsigned long def_mount_opts;
3087 struct inode *root;
9f6200bb 3088 char *cp;
0390131b 3089 const char *descr;
dcc7dae3 3090 int ret = -ENOMEM;
ac27a0ec 3091 int blocksize;
4ec11028
TT		 3092 unsigned int db_count;
3093 unsigned int i;
f287a1a5 3094 int needs_recovery, has_huge_files;
bd81d8ee 3095 __u64 blocks_count;
833f4077 3096 int err;
b3881f74 3097 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3098 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3099
3100 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3101 if (!sbi)
dcc7dae3 3102 goto out_free_orig;
705895b6
PE		 3103
3104 sbi->s_blockgroup_lock =
3105 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3106 if (!sbi->s_blockgroup_lock) {
3107 kfree(sbi);
dcc7dae3 3108 goto out_free_orig;
705895b6 3109 }
ac27a0ec
DK		 3110 sb->s_fs_info = sbi;
3111 sbi->s_mount_opt = 0;
617ba13b
MC		 3112 sbi->s_resuid = EXT4_DEF_RESUID;
3113 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 3114 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3115 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3116 if (sb->s_bdev->bd_part)
3117 sbi->s_sectors_written_start =
3118 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3119
9f6200bb
TT		 3120 /* Cleanup superblock name */
3121 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3122 *cp = '!';
3123
dcc7dae3 3124 ret = -EINVAL;
617ba13b 3125 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3126 if (!blocksize) {
b31e1552 3127 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3128 goto out_fail;
3129 }
3130
3131 /*
617ba13b 3132 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3133 * block sizes. We need to calculate the offset from buffer start.
3134 */
617ba13b 3135 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3136 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3137 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3138 } else {
70bbb3e0 3139 logical_sb_block = sb_block;
ac27a0ec
DK		 3140 }
3141
70bbb3e0 3142 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3143 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3144 goto out_fail;
3145 }
3146 /*
3147 * Note: s_es must be initialized as soon as possible because
617ba13b 3148 * some ext4 macro-instructions depend on its value
ac27a0ec 3149 */
617ba13b 3150 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3151 sbi->s_es = es;
3152 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3153 if (sb->s_magic != EXT4_SUPER_MAGIC)
3154 goto cantfind_ext4;
afc32f7e 3155 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 3156
3157 /* Set defaults before we parse the mount options */
3158 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3159 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3160 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3161 set_opt(sb, DEBUG);
437ca0fd
DM		 3162 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3163 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3164 "2.6.38");
fd8c37ec 3165 set_opt(sb, GRPID);
437ca0fd 3166 }
617ba13b 3167 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3168 set_opt(sb, NO_UID32);
ea663336 3169 /* xattr user namespace & acls are now defaulted on */
03010a33 3170#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3171 set_opt(sb, XATTR_USER);
2e7842b8 3172#endif
03010a33 3173#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3174 set_opt(sb, POSIX_ACL);
2e7842b8 3175#endif
6fd7a467 3176 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3177 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3178 set_opt(sb, JOURNAL_DATA);
617ba13b 3179 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3180 set_opt(sb, ORDERED_DATA);
617ba13b 3181 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3182 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3183
3184 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3185 set_opt(sb, ERRORS_PANIC);
bb4f397a 3186 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3187 set_opt(sb, ERRORS_CONT);
bb4f397a 3188 else
fd8c37ec 3189 set_opt(sb, ERRORS_RO);
8b67f04a 3190 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3191 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3192 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3193 set_opt(sb, DISCARD);
ac27a0ec
DK		 3194
3195 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3196 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3197 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3198 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3199 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3200
8b67f04a 3201 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3202 set_opt(sb, BARRIER);
ac27a0ec 3203
dd919b98
AK		 3204 /*
3205 * enable delayed allocation by default
3206 * Use -o nodelalloc to turn it off
3207 */
8b67f04a
TT		 3208 if (!IS_EXT3_SB(sb) &&
3209 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3210 set_opt(sb, DELALLOC);
dd919b98 3211
51ce6511
LC		 3212 /*
3213 * set default s_li_wait_mult for lazyinit, for the case there is
3214 * no mount option specified.
3215 */
3216 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3217
8b67f04a
TT		 3218 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3219 &journal_devnum, &journal_ioprio, NULL, 0)) {
3220 ext4_msg(sb, KERN_WARNING,
3221 "failed to parse options in superblock: %s",
3222 sbi->s_es->s_mount_opts);
3223 }
b3881f74
TT		 3224 if (!parse_options((char *) data, sb, &journal_devnum,
3225 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 3226 goto failed_mount;
3227
3228 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3229 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3230
617ba13b
MC		 3231 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3232 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3233 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3234 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3235 ext4_msg(sb, KERN_WARNING,
3236 "feature flags set on rev 0 fs, "
3237 "running e2fsck is recommended");
469108ff 3238
2035e776
TT		 3239 if (IS_EXT2_SB(sb)) {
3240 if (ext2_feature_set_ok(sb))
3241 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3242 "using the ext4 subsystem");
3243 else {
3244 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3245 "to feature incompatibilities");
3246 goto failed_mount;
3247 }
3248 }
3249
3250 if (IS_EXT3_SB(sb)) {
3251 if (ext3_feature_set_ok(sb))
3252 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3253 "using the ext4 subsystem");
3254 else {
3255 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3256 "to feature incompatibilities");
3257 goto failed_mount;
3258 }
3259 }
3260
ac27a0ec
DK		 3261 /*
3262 * Check feature flags regardless of the revision level, since we
3263 * previously didn't change the revision level when setting the flags,
3264 * so there is a chance incompat flags are set on a rev 0 filesystem.
3265 */
a13fb1a4 3266 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3267 goto failed_mount;
a13fb1a4 3268
ac27a0ec
DK		 3269 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3270
617ba13b
MC		 3271 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3272 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3273 ext4_msg(sb, KERN_ERR,
3274 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3275 goto failed_mount;
3276 }
3277
ac27a0ec 3278 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3279 /* Validate the filesystem blocksize */
3280 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3281 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3282 blocksize);
ac27a0ec
DK		 3283 goto failed_mount;
3284 }
3285
2b2d6d01 3286 brelse(bh);
70bbb3e0
AM		 3287 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3288 offset = do_div(logical_sb_block, blocksize);
3289 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3290 if (!bh) {
b31e1552
ES		 3291 ext4_msg(sb, KERN_ERR,
3292 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3293 goto failed_mount;
3294 }
617ba13b 3295 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3296 sbi->s_es = es;
617ba13b 3297 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3298 ext4_msg(sb, KERN_ERR,
3299 "Magic mismatch, very weird!");
ac27a0ec
DK		 3300 goto failed_mount;
3301 }
3302 }
3303
a13fb1a4
ES		 3304 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3305 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3306 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3307 has_huge_files);
3308 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3309
617ba13b
MC		 3310 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3311 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3312 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3313 } else {
3314 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3315 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3316 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3317 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3318 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3319 ext4_msg(sb, KERN_ERR,
3320 "unsupported inode size: %d",
2b2d6d01 3321 sbi->s_inode_size);
ac27a0ec
DK		 3322 goto failed_mount;
3323 }
ef7f3835
KS		 3324 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3325 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3326 }
0b8e58a1 3327
0d1ee42f
AR		 3328 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3329 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3330 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3331 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3332 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3333 ext4_msg(sb, KERN_ERR,
3334 "unsupported descriptor size %lu",
0d1ee42f
AR		 3335 sbi->s_desc_size);
3336 goto failed_mount;
3337 }
3338 } else
3339 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3340
ac27a0ec 3341 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3342 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3343 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3344 goto cantfind_ext4;
0b8e58a1 3345
617ba13b 3346 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3347 if (sbi->s_inodes_per_block == 0)
617ba13b 3348 goto cantfind_ext4;
ac27a0ec
DK		 3349 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3350 sbi->s_inodes_per_block;
0d1ee42f 3351 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3352 sbi->s_sbh = bh;
3353 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3354 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3355 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3356
2b2d6d01 3357 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3358 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3359 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3360 i = le32_to_cpu(es->s_flags);
3361 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3362 sbi->s_hash_unsigned = 3;
3363 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3364#ifdef __CHAR_UNSIGNED__
3365 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3366 sbi->s_hash_unsigned = 3;
3367#else
3368 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3369#endif
3370 sb->s_dirt = 1;
3371 }
ac27a0ec
DK		 3372
3373 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 3374 ext4_msg(sb, KERN_ERR,
3375 "#blocks per group too big: %lu",
2b2d6d01 3376 sbi->s_blocks_per_group);
ac27a0ec
DK		 3377 goto failed_mount;
3378 }
ac27a0ec 3379 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3380 ext4_msg(sb, KERN_ERR,
3381 "#inodes per group too big: %lu",
2b2d6d01 3382 sbi->s_inodes_per_group);
ac27a0ec
DK		 3383 goto failed_mount;
3384 }
3385
bf43d84b
ES		 3386 /*
3387 * Test whether we have more sectors than will fit in sector_t,
3388 * and whether the max offset is addressable by the page cache.
3389 */
5a9ae68a 3390 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3391 ext4_blocks_count(es));
5a9ae68a 3392 if (err) {
b31e1552 3393 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3394 " too large to mount safely on this system");
ac27a0ec 3395 if (sizeof(sector_t) < 8)
90c699a9 3396 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3397 ret = err;
ac27a0ec
DK		 3398 goto failed_mount;
3399 }
3400
617ba13b
MC		 3401 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3402 goto cantfind_ext4;
e7c95593 3403
0f2ddca6
FTN		 3404 /* check blocks count against device size */
3405 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3406 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3407 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3408 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3409 ext4_blocks_count(es), blocks_count);
3410 goto failed_mount;
3411 }
3412
0b8e58a1
AD		 3413 /*
3414 * It makes no sense for the first data block to be beyond the end
3415 * of the filesystem.
3416 */
3417 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 3418 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3419 "block %u is beyond end of filesystem (%llu)",
3420 le32_to_cpu(es->s_first_data_block),
3421 ext4_blocks_count(es));
e7c95593
ES		 3422 goto failed_mount;
3423 }
bd81d8ee
LV		 3424 blocks_count = (ext4_blocks_count(es) -
3425 le32_to_cpu(es->s_first_data_block) +
3426 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3427 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3428 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3429 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3430 "(block count %llu, first data block %u, "
b31e1552 3431 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3432 ext4_blocks_count(es),
3433 le32_to_cpu(es->s_first_data_block),
3434 EXT4_BLOCKS_PER_GROUP(sb));
3435 goto failed_mount;
3436 }
bd81d8ee 3437 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3438 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3439 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3440 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3441 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 3442 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 3443 GFP_KERNEL);
3444 if (sbi->s_group_desc == NULL) {
b31e1552 3445 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3446 goto failed_mount;
3447 }
3448
3244fcb1 3449#ifdef CONFIG_PROC_FS
9f6200bb
TT		 3450 if (ext4_proc_root)
3451 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 3452#endif
240799cd 3453
705895b6 3454 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3455
3456 for (i = 0; i < db_count; i++) {
70bbb3e0 3457 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3458 sbi->s_group_desc[i] = sb_bread(sb, block);
3459 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3460 ext4_msg(sb, KERN_ERR,
3461 "can't read group descriptor %d", i);
ac27a0ec
DK		 3462 db_count = i;
3463 goto failed_mount2;
3464 }
3465 }
bfff6873 3466 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3467 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3468 goto failed_mount2;
3469 }
772cb7c8
JS		 3470 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3471 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3472 ext4_msg(sb, KERN_ERR,
3473 "unable to initialize "
3474 "flex_bg meta info!");
772cb7c8
JS		 3475 goto failed_mount2;
3476 }
3477
ac27a0ec
DK		 3478 sbi->s_gdb_count = db_count;
3479 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3480 spin_lock_init(&sbi->s_next_gen_lock);
3481
04496411
TM		 3482 init_timer(&sbi->s_err_report);
3483 sbi->s_err_report.function = print_daily_error_info;
3484 sbi->s_err_report.data = (unsigned long) sb;
3485
ce7e010a
TT		 3486 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3487 ext4_count_free_blocks(sb));
3488 if (!err) {
3489 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3490 ext4_count_free_inodes(sb));
3491 }
3492 if (!err) {
3493 err = percpu_counter_init(&sbi->s_dirs_counter,
3494 ext4_count_dirs(sb));
3495 }
3496 if (!err) {
3497 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3498 }
3499 if (err) {
3500 ext4_msg(sb, KERN_ERR, "insufficient memory");
3501 goto failed_mount3;
3502 }
3503
c9de560d 3504 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3505 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3506
ac27a0ec
DK		 3507 /*
3508 * set up enough so that it can read an inode
3509 */
9ca92389
TT		 3510 if (!test_opt(sb, NOLOAD) &&
3511 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3512 sb->s_op = &ext4_sops;
3513 else
3514 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3515 sb->s_export_op = &ext4_export_ops;
3516 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3517#ifdef CONFIG_QUOTA
617ba13b
MC		 3518 sb->s_qcop = &ext4_qctl_operations;
3519 sb->dq_op = &ext4_quota_operations;
ac27a0ec 3520#endif
f2fa2ffc
AK		 3521 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3522
ac27a0ec 3523 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3524 mutex_init(&sbi->s_orphan_lock);
8f82f840 3525 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3526
3527 sb->s_root = NULL;
3528
3529 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3530 EXT4_HAS_INCOMPAT_FEATURE(sb,
3531 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3532
c5e06d10
JL		 3533 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3534 !(sb->s_flags & MS_RDONLY))
3535 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3536 goto failed_mount3;
3537
ac27a0ec
DK		 3538 /*
3539 * The first inode we look at is the journal inode. Don't try
3540 * root first: it may be modified in the journal!
3541 */
3542 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3543 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3544 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3545 goto failed_mount3;
0390131b
FM		 3546 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3547 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3548 ext4_msg(sb, KERN_ERR, "required journal recovery "
3549 "suppressed and not mounted read-only");
744692dc 3550 goto failed_mount_wq;
ac27a0ec 3551 } else {
fd8c37ec 3552 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3553 sbi->s_journal = NULL;
3554 needs_recovery = 0;
3555 goto no_journal;
ac27a0ec
DK		 3556 }
3557
eb40a09c
JS		 3558 if (ext4_blocks_count(es) > 0xffffffffULL &&
3559 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3560 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3561 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3562 goto failed_mount_wq;
eb40a09c
JS		 3563 }
3564
d4da6c9c
LT		 3565 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3566 jbd2_journal_set_features(sbi->s_journal,
3567 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3568 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3569 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3570 jbd2_journal_set_features(sbi->s_journal,
3571 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3572 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3573 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3574 } else {
3575 jbd2_journal_clear_features(sbi->s_journal,
3576 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3577 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3578 }
818d276c 3579
ac27a0ec
DK		 3580 /* We have now updated the journal if required, so we can
3581 * validate the data journaling mode. */
3582 switch (test_opt(sb, DATA_FLAGS)) {
3583 case 0:
3584 /* No mode set, assume a default based on the journal
63f57933
AM		 3585 * capabilities: ORDERED_DATA if the journal can
3586 * cope, else JOURNAL_DATA
3587 */
dab291af
MC		 3588 if (jbd2_journal_check_available_features
3589 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3590 set_opt(sb, ORDERED_DATA);
ac27a0ec 3591 else
fd8c37ec 3592 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3593 break;
3594
617ba13b
MC		 3595 case EXT4_MOUNT_ORDERED_DATA:
3596 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3597 if (!jbd2_journal_check_available_features
3598 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3599 ext4_msg(sb, KERN_ERR, "Journal does not support "
3600 "requested data journaling mode");
744692dc 3601 goto failed_mount_wq;
ac27a0ec
DK		 3602 }
3603 default:
3604 break;
3605 }
b3881f74 3606 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3607
ce7e010a
TT		 3608 /*
3609 * The journal may have updated the bg summary counts, so we
3610 * need to update the global counters.
3611 */
3612 percpu_counter_set(&sbi->s_freeblocks_counter,
3613 ext4_count_free_blocks(sb));
3614 percpu_counter_set(&sbi->s_freeinodes_counter,
3615 ext4_count_free_inodes(sb));
3616 percpu_counter_set(&sbi->s_dirs_counter,
3617 ext4_count_dirs(sb));
3618 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
206f7ab4 3619
ce7e010a 3620no_journal:
fd89d5f2
TH		 3621 /*
3622 * The maximum number of concurrent works can be high and
3623 * concurrency isn't really necessary. Limit it to 1.
3624 */
3625 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3626 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3627 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3628 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3629 goto failed_mount_wq;
3630 }
3631
ac27a0ec 3632 /*
dab291af 3633 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3634 * so we can safely mount the rest of the filesystem now.
3635 */
3636
1d1fe1ee
DH		 3637 root = ext4_iget(sb, EXT4_ROOT_INO);
3638 if (IS_ERR(root)) {
b31e1552 3639 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3640 ret = PTR_ERR(root);
32a9bb57 3641 root = NULL;
ac27a0ec
DK		 3642 goto failed_mount4;
3643 }
3644 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3645 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 3646 goto failed_mount4;
3647 }
1d1fe1ee
DH		 3648 sb->s_root = d_alloc_root(root);
3649 if (!sb->s_root) {
b31e1552 3650 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3651 ret = -ENOMEM;
3652 goto failed_mount4;
3653 }
ac27a0ec 3654
2b2d6d01 3655 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3656
3657 /* determine the minimum size of new large inodes, if present */
3658 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3659 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3660 EXT4_GOOD_OLD_INODE_SIZE;
3661 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3662 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3663 if (sbi->s_want_extra_isize <
3664 le16_to_cpu(es->s_want_extra_isize))
3665 sbi->s_want_extra_isize =
3666 le16_to_cpu(es->s_want_extra_isize);
3667 if (sbi->s_want_extra_isize <
3668 le16_to_cpu(es->s_min_extra_isize))
3669 sbi->s_want_extra_isize =
3670 le16_to_cpu(es->s_min_extra_isize);
3671 }
3672 }
3673 /* Check if enough inode space is available */
3674 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3675 sbi->s_inode_size) {
3676 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3677 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3678 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3679 "available");
ef7f3835
KS		 3680 }
3681
90576c0b
TT		 3682 if (test_opt(sb, DELALLOC) &&
3683 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
b31e1552
ES		 3684 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3685 "requested data journaling mode");
fd8c37ec 3686 clear_opt(sb, DELALLOC);
90576c0b 3687 }
744692dc
JZ		 3688 if (test_opt(sb, DIOREAD_NOLOCK)) {
3689 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3690 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3691 "option - requested data journaling mode");
fd8c37ec 3692 clear_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 3693 }
3694 if (sb->s_blocksize < PAGE_SIZE) {
3695 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3696 "option - block size is too small");
fd8c37ec 3697 clear_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 3698 }
3699 }
c2774d84 3700
6fd058f7
TT		 3701 err = ext4_setup_system_zone(sb);
3702 if (err) {
b31e1552 3703 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3704 "zone (%d)", err);
6fd058f7
TT		 3705 goto failed_mount4;
3706 }
3707
c2774d84
AK		 3708 ext4_ext_init(sb);
3709 err = ext4_mb_init(sb, needs_recovery);
3710 if (err) {
421f91d2 3711 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3712 err);
c2774d84
AK		 3713 goto failed_mount4;
3714 }
3715
bfff6873
LC		 3716 err = ext4_register_li_request(sb, first_not_zeroed);
3717 if (err)
3718 goto failed_mount4;
3719
3197ebdb
TT		 3720 sbi->s_kobj.kset = ext4_kset;
3721 init_completion(&sbi->s_kobj_unregister);
3722 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3723 "%s", sb->s_id);
3724 if (err) {
3725 ext4_mb_release(sb);
3726 ext4_ext_release(sb);
3727 goto failed_mount4;
3728 };
3729
617ba13b
MC		 3730 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3731 ext4_orphan_cleanup(sb, es);
3732 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3733 if (needs_recovery) {
b31e1552 3734 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3735 ext4_mark_recovery_complete(sb, es);
3736 }
3737 if (EXT4_SB(sb)->s_journal) {
3738 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3739 descr = " journalled data mode";
3740 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3741 descr = " ordered data mode";
3742 else
3743 descr = " writeback data mode";
3744 } else
3745 descr = "out journal";
3746
d4c402d9 3747 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3748 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3749 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3750
66e61a9e
TT		 3751 if (es->s_error_count)
3752 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3753
d4c402d9 3754 kfree(orig_data);
ac27a0ec
DK		 3755 return 0;
3756
617ba13b 3757cantfind_ext4:
ac27a0ec 3758 if (!silent)
b31e1552 3759 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3760 goto failed_mount;
3761
3762failed_mount4:
32a9bb57
MK		 3763 iput(root);
3764 sb->s_root = NULL;
b31e1552 3765 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3766 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3767failed_mount_wq:
6fd058f7 3768 ext4_release_system_zone(sb);
0390131b
FM		 3769 if (sbi->s_journal) {
3770 jbd2_journal_destroy(sbi->s_journal);
3771 sbi->s_journal = NULL;
3772 }
ac27a0ec 3773failed_mount3:
04496411 3774 del_timer(&sbi->s_err_report);
9933fc0a
TT		 3775 if (sbi->s_flex_groups)
3776 ext4_kvfree(sbi->s_flex_groups);
ce7e010a
TT		 3777 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3778 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3779 percpu_counter_destroy(&sbi->s_dirs_counter);
3780 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
c5e06d10
JL		 3781 if (sbi->s_mmp_tsk)
3782 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3783failed_mount2:
3784 for (i = 0; i < db_count; i++)
3785 brelse(sbi->s_group_desc[i]);
3786 kfree(sbi->s_group_desc);
3787failed_mount:
240799cd 3788 if (sbi->s_proc) {
9f6200bb 3789 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3790 }
ac27a0ec
DK		 3791#ifdef CONFIG_QUOTA
3792 for (i = 0; i < MAXQUOTAS; i++)
3793 kfree(sbi->s_qf_names[i]);
3794#endif
617ba13b 3795 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3796 brelse(bh);
3797out_fail:
3798 sb->s_fs_info = NULL;
f6830165 3799 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3800 kfree(sbi);
dcc7dae3 3801out_free_orig:
d4c402d9 3802 kfree(orig_data);
1d1fe1ee 3803 return ret;
ac27a0ec
DK		 3804}
3805
3806/*
3807 * Setup any per-fs journal parameters now. We'll do this both on
3808 * initial mount, once the journal has been initialised but before we've
3809 * done any recovery; and again on any subsequent remount.
3810 */
617ba13b 3811static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3812{
617ba13b 3813 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3814
30773840
TT		 3815 journal->j_commit_interval = sbi->s_commit_interval;
3816 journal->j_min_batch_time = sbi->s_min_batch_time;
3817 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3818
a931da6a 3819 write_lock(&journal->j_state_lock);
ac27a0ec 3820 if (test_opt(sb, BARRIER))
dab291af 3821 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3822 else
dab291af 3823 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3824 if (test_opt(sb, DATA_ERR_ABORT))
3825 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3826 else
3827 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3828 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3829}
3830
617ba13b 3831static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3832 unsigned int journal_inum)
3833{
3834 struct inode *journal_inode;
3835 journal_t *journal;
3836
0390131b
FM		 3837 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3838
ac27a0ec
DK		 3839 /* First, test for the existence of a valid inode on disk. Bad
3840 * things happen if we iget() an unused inode, as the subsequent
3841 * iput() will try to delete it. */
3842
1d1fe1ee
DH		 3843 journal_inode = ext4_iget(sb, journal_inum);
3844 if (IS_ERR(journal_inode)) {
b31e1552 3845 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3846 return NULL;
3847 }
3848 if (!journal_inode->i_nlink) {
3849 make_bad_inode(journal_inode);
3850 iput(journal_inode);
b31e1552 3851 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3852 return NULL;
3853 }
3854
e5f8eab8 3855 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3856 journal_inode, journal_inode->i_size);
1d1fe1ee 3857 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3858 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3859 iput(journal_inode);
3860 return NULL;
3861 }
3862
dab291af 3863 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3864 if (!journal) {
b31e1552 3865 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3866 iput(journal_inode);
3867 return NULL;
3868 }
3869 journal->j_private = sb;
617ba13b 3870 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3871 return journal;
3872}
3873
617ba13b 3874static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3875 dev_t j_dev)
3876{
2b2d6d01 3877 struct buffer_head *bh;
ac27a0ec 3878 journal_t *journal;
617ba13b
MC		 3879 ext4_fsblk_t start;
3880 ext4_fsblk_t len;
ac27a0ec 3881 int hblock, blocksize;
617ba13b 3882 ext4_fsblk_t sb_block;
ac27a0ec 3883 unsigned long offset;
2b2d6d01 3884 struct ext4_super_block *es;
ac27a0ec
DK		 3885 struct block_device *bdev;
3886
0390131b
FM		 3887 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3888
b31e1552 3889 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3890 if (bdev == NULL)
3891 return NULL;
3892
ac27a0ec 3893 blocksize = sb->s_blocksize;
e1defc4f 3894 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3895 if (blocksize < hblock) {
b31e1552
ES		 3896 ext4_msg(sb, KERN_ERR,
3897 "blocksize too small for journal device");
ac27a0ec
DK		 3898 goto out_bdev;
3899 }
3900
617ba13b
MC		 3901 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3902 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3903 set_blocksize(bdev, blocksize);
3904 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3905 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3906 "external journal");
ac27a0ec
DK		 3907 goto out_bdev;
3908 }
3909
617ba13b
MC		 3910 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3911 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3912 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3913 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3914 ext4_msg(sb, KERN_ERR, "external journal has "
3915 "bad superblock");
ac27a0ec
DK		 3916 brelse(bh);
3917 goto out_bdev;
3918 }
3919
617ba13b 3920 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3921 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3922 brelse(bh);
3923 goto out_bdev;
3924 }
3925
bd81d8ee 3926 len = ext4_blocks_count(es);
ac27a0ec
DK		 3927 start = sb_block + 1;
3928 brelse(bh); /* we're done with the superblock */
3929
dab291af 3930 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3931 start, len, blocksize);
3932 if (!journal) {
b31e1552 3933 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3934 goto out_bdev;
3935 }
3936 journal->j_private = sb;
3937 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3938 wait_on_buffer(journal->j_sb_buffer);
3939 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3940 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3941 goto out_journal;
3942 }
3943 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3944 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3945 "user (unsupported) - %d",
ac27a0ec
DK		 3946 be32_to_cpu(journal->j_superblock->s_nr_users));
3947 goto out_journal;
3948 }
617ba13b
MC		 3949 EXT4_SB(sb)->journal_bdev = bdev;
3950 ext4_init_journal_params(sb, journal);
ac27a0ec 3951 return journal;
0b8e58a1 3952
ac27a0ec 3953out_journal:
dab291af 3954 jbd2_journal_destroy(journal);
ac27a0ec 3955out_bdev:
617ba13b 3956 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3957 return NULL;
3958}
3959
617ba13b
MC		 3960static int ext4_load_journal(struct super_block *sb,
3961 struct ext4_super_block *es,
ac27a0ec
DK		 3962 unsigned long journal_devnum)
3963{
3964 journal_t *journal;
3965 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3966 dev_t journal_dev;
3967 int err = 0;
3968 int really_read_only;
3969
0390131b
FM		 3970 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3971
ac27a0ec
DK		 3972 if (journal_devnum &&
3973 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3974 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3975 "numbers have changed");
ac27a0ec
DK		 3976 journal_dev = new_decode_dev(journal_devnum);
3977 } else
3978 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3979
3980 really_read_only = bdev_read_only(sb->s_bdev);
3981
3982 /*
3983 * Are we loading a blank journal or performing recovery after a
3984 * crash? For recovery, we need to check in advance whether we
3985 * can get read-write access to the device.
3986 */
617ba13b 3987 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3988 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3989 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3990 "required on readonly filesystem");
ac27a0ec 3991 if (really_read_only) {
b31e1552
ES		 3992 ext4_msg(sb, KERN_ERR, "write access "
3993 "unavailable, cannot proceed");
ac27a0ec
DK		 3994 return -EROFS;
3995 }
b31e1552
ES		 3996 ext4_msg(sb, KERN_INFO, "write access will "
3997 "be enabled during recovery");
ac27a0ec
DK		 3998 }
3999 }
4000
4001 if (journal_inum && journal_dev) {
b31e1552
ES		 4002 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4003 "and inode journals!");
ac27a0ec
DK		 4004 return -EINVAL;
4005 }
4006
4007 if (journal_inum) {
617ba13b 4008 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4009 return -EINVAL;
4010 } else {
617ba13b 4011 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4012 return -EINVAL;
4013 }
4014
90576c0b 4015 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4016 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4017
ac27a0ec 4018 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 4019 err = jbd2_journal_update_format(journal);
ac27a0ec 4020 if (err) {
b31e1552 4021 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 4022 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4023 return err;
4024 }
4025 }
4026
617ba13b 4027 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4028 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4029 if (!err) {
4030 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4031 if (save)
4032 memcpy(save, ((char *) es) +
4033 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4034 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4035 if (save)
4036 memcpy(((char *) es) + EXT4_S_ERR_START,
4037 save, EXT4_S_ERR_LEN);
4038 kfree(save);
4039 }
ac27a0ec
DK		 4040
4041 if (err) {
b31e1552 4042 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4043 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4044 return err;
4045 }
4046
617ba13b
MC		 4047 EXT4_SB(sb)->s_journal = journal;
4048 ext4_clear_journal_err(sb, es);
ac27a0ec 4049
c41303ce 4050 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4051 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4052 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4053
4054 /* Make sure we flush the recovery flag to disk. */
e2d67052 4055 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4056 }
4057
4058 return 0;
4059}
4060
e2d67052 4061static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4062{
e2d67052 4063 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4064 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4065 int error = 0;
ac27a0ec
DK		 4066
4067 if (!sbh)
c4be0c1d 4068 return error;
914258bf
TT		 4069 if (buffer_write_io_error(sbh)) {
4070 /*
4071 * Oh, dear. A previous attempt to write the
4072 * superblock failed. This could happen because the
4073 * USB device was yanked out. Or it could happen to
4074 * be a transient write error and maybe the block will
4075 * be remapped. Nothing we can do but to retry the
4076 * write and hope for the best.
4077 */
b31e1552
ES		 4078 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4079 "superblock detected");
914258bf
TT		 4080 clear_buffer_write_io_error(sbh);
4081 set_buffer_uptodate(sbh);
4082 }
71290b36
TT		 4083 /*
4084 * If the file system is mounted read-only, don't update the
4085 * superblock write time. This avoids updating the superblock
4086 * write time when we are mounting the root file system
4087 * read/only but we need to replay the journal; at that point,
4088 * for people who are east of GMT and who make their clock
4089 * tick in localtime for Windows bug-for-bug compatibility,
4090 * the clock is set in the future, and this will cause e2fsck
4091 * to complain and force a full file system check.
4092 */
4093 if (!(sb->s_flags & MS_RDONLY))
4094 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4095 if (sb->s_bdev->bd_part)
4096 es->s_kbytes_written =
4097 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4098 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4099 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4100 else
4101 es->s_kbytes_written =
4102 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
ce7e010a
TT		 4103 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
4104 &EXT4_SB(sb)->s_freeblocks_counter));
4105 es->s_free_inodes_count =
4106 cpu_to_le32(percpu_counter_sum_positive(
4107 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 4108 sb->s_dirt = 0;
ac27a0ec
DK		 4109 BUFFER_TRACE(sbh, "marking dirty");
4110 mark_buffer_dirty(sbh);
914258bf 4111 if (sync) {
c4be0c1d
TS		 4112 error = sync_dirty_buffer(sbh);
4113 if (error)
4114 return error;
4115
4116 error = buffer_write_io_error(sbh);
4117 if (error) {
b31e1552
ES		 4118 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4119 "superblock");
914258bf
TT		 4120 clear_buffer_write_io_error(sbh);
4121 set_buffer_uptodate(sbh);
4122 }
4123 }
c4be0c1d 4124 return error;
ac27a0ec
DK		 4125}
4126
ac27a0ec
DK		 4127/*
4128 * Have we just finished recovery? If so, and if we are mounting (or
4129 * remounting) the filesystem readonly, then we will end up with a
4130 * consistent fs on disk. Record that fact.
4131 */
2b2d6d01
TT		 4132static void ext4_mark_recovery_complete(struct super_block *sb,
4133 struct ext4_super_block *es)
ac27a0ec 4134{
617ba13b 4135 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4136
0390131b
FM		 4137 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4138 BUG_ON(journal != NULL);
4139 return;
4140 }
dab291af 4141 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4142 if (jbd2_journal_flush(journal) < 0)
4143 goto out;
4144
617ba13b 4145 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4146 sb->s_flags & MS_RDONLY) {
617ba13b 4147 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4148 ext4_commit_super(sb, 1);
ac27a0ec 4149 }
7ffe1ea8
HK		 4150
4151out:
dab291af 4152 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4153}
4154
4155/*
4156 * If we are mounting (or read-write remounting) a filesystem whose journal
4157 * has recorded an error from a previous lifetime, move that error to the
4158 * main filesystem now.
4159 */
2b2d6d01
TT		 4160static void ext4_clear_journal_err(struct super_block *sb,
4161 struct ext4_super_block *es)
ac27a0ec
DK		 4162{
4163 journal_t *journal;
4164 int j_errno;
4165 const char *errstr;
4166
0390131b
FM		 4167 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4168
617ba13b 4169 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4170
4171 /*
4172 * Now check for any error status which may have been recorded in the
617ba13b 4173 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4174 */
4175
dab291af 4176 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4177 if (j_errno) {
4178 char nbuf[16];
4179
617ba13b 4180 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4181 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4182 "from previous mount: %s", errstr);
12062ddd 4183 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4184
617ba13b
MC		 4185 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4186 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4187 ext4_commit_super(sb, 1);
ac27a0ec 4188
dab291af 4189 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4190 }
4191}
4192
4193/*
4194 * Force the running and committing transactions to commit,
4195 * and wait on the commit.
4196 */
617ba13b 4197int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4198{
4199 journal_t *journal;
0390131b 4200 int ret = 0;
ac27a0ec
DK		 4201
4202 if (sb->s_flags & MS_RDONLY)
4203 return 0;
4204
617ba13b 4205 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4206 if (journal) {
437f88cc 4207 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4208 ret = ext4_journal_force_commit(journal);
6b0310fb 4209 }
0390131b 4210
ac27a0ec
DK		 4211 return ret;
4212}
4213
2b2d6d01 4214static void ext4_write_super(struct super_block *sb)
ac27a0ec 4215{
ebc1ac16 4216 lock_super(sb);
9ca92389 4217 ext4_commit_super(sb, 1);
ebc1ac16 4218 unlock_super(sb);
ac27a0ec
DK		 4219}
4220
617ba13b 4221static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4222{
14ce0cb4 4223 int ret = 0;
9eddacf9 4224 tid_t target;
8d5d02e6 4225 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4226
9bffad1e 4227 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4228 flush_workqueue(sbi->dio_unwritten_wq);
4229 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4230 if (wait)
8d5d02e6 4231 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4232 }
14ce0cb4 4233 return ret;
ac27a0ec
DK		 4234}
4235
4236/*
4237 * LVM calls this function before a (read-only) snapshot is created. This
4238 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4239 *
4240 * Note that only this function cannot bring a filesystem to be in a clean
4241 * state independently, because ext4 prevents a new handle from being started
4242 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4243 * the upper layer.
ac27a0ec 4244 */
c4be0c1d 4245static int ext4_freeze(struct super_block *sb)
ac27a0ec 4246{
c4be0c1d
TS		 4247 int error = 0;
4248 journal_t *journal;
ac27a0ec 4249
9ca92389
TT		 4250 if (sb->s_flags & MS_RDONLY)
4251 return 0;
ac27a0ec 4252
9ca92389 4253 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4254
9ca92389
TT		 4255 /* Now we set up the journal barrier. */
4256 jbd2_journal_lock_updates(journal);
ac27a0ec 4257
9ca92389
TT		 4258 /*
4259 * Don't clear the needs_recovery flag if we failed to flush
4260 * the journal.
4261 */
4262 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4263 if (error < 0)
4264 goto out;
9ca92389
TT		 4265
4266 /* Journal blocked and flushed, clear needs_recovery flag. */
4267 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4268 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4269out:
4270 /* we rely on s_frozen to stop further updates */
4271 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4272 return error;
ac27a0ec
DK		 4273}
4274
4275/*
4276 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4277 * flag here, even though the filesystem is not technically dirty yet.
4278 */
c4be0c1d 4279static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4280{
9ca92389
TT		 4281 if (sb->s_flags & MS_RDONLY)
4282 return 0;
4283
4284 lock_super(sb);
4285 /* Reset the needs_recovery flag before the fs is unlocked. */
4286 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4287 ext4_commit_super(sb, 1);
4288 unlock_super(sb);
c4be0c1d 4289 return 0;
ac27a0ec
DK		 4290}
4291
673c6100
TT		 4292/*
4293 * Structure to save mount options for ext4_remount's benefit
4294 */
4295struct ext4_mount_options {
4296 unsigned long s_mount_opt;
a2595b8a 4297 unsigned long s_mount_opt2;
673c6100
TT		 4298 uid_t s_resuid;
4299 gid_t s_resgid;
4300 unsigned long s_commit_interval;
4301 u32 s_min_batch_time, s_max_batch_time;
4302#ifdef CONFIG_QUOTA
4303 int s_jquota_fmt;
4304 char *s_qf_names[MAXQUOTAS];
4305#endif
4306};
4307
2b2d6d01 4308static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4309{
2b2d6d01 4310 struct ext4_super_block *es;
617ba13b
MC		 4311 struct ext4_sb_info *sbi = EXT4_SB(sb);
4312 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 4313 unsigned long old_sb_flags;
617ba13b 4314 struct ext4_mount_options old_opts;
c79d967d 4315 int enable_quota = 0;
8a266467 4316 ext4_group_t g;
b3881f74 4317 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4318 int err = 0;
ac27a0ec
DK		 4319#ifdef CONFIG_QUOTA
4320 int i;
4321#endif
d4c402d9 4322 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4323
4324 /* Store the original options */
bbd6851a 4325 lock_super(sb);
ac27a0ec
DK		 4326 old_sb_flags = sb->s_flags;
4327 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4328 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4329 old_opts.s_resuid = sbi->s_resuid;
4330 old_opts.s_resgid = sbi->s_resgid;
4331 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4332 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4333 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4334#ifdef CONFIG_QUOTA
4335 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4336 for (i = 0; i < MAXQUOTAS; i++)
4337 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4338#endif
b3881f74
TT		 4339 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4340 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4341
4342 /*
4343 * Allow the "check" option to be passed as a remount option.
4344 */
b3881f74
TT		 4345 if (!parse_options(data, sb, NULL, &journal_ioprio,
4346 &n_blocks_count, 1)) {
ac27a0ec
DK		 4347 err = -EINVAL;
4348 goto restore_opts;
4349 }
4350
4ab2f15b 4351 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4352 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4353
4354 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4355 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4356
4357 es = sbi->s_es;
4358
b3881f74 4359 if (sbi->s_journal) {
0390131b 4360 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4361 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4362 }
ac27a0ec
DK		 4363
4364 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 4365 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 4366 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4367 err = -EROFS;
4368 goto restore_opts;
4369 }
4370
4371 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4372 err = dquot_suspend(sb, -1);
4373 if (err < 0)
c79d967d 4374 goto restore_opts;
c79d967d 4375
ac27a0ec
DK		 4376 /*
4377 * First of all, the unconditional stuff we have to do
4378 * to disable replay of the journal when we next remount
4379 */
4380 sb->s_flags |= MS_RDONLY;
4381
4382 /*
4383 * OK, test if we are remounting a valid rw partition
4384 * readonly, and if so set the rdonly flag and then
4385 * mark the partition as valid again.
4386 */
617ba13b
MC		 4387 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4388 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4389 es->s_state = cpu_to_le16(sbi->s_mount_state);
4390
a63c9eb2 4391 if (sbi->s_journal)
0390131b 4392 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4393 } else {
a13fb1a4
ES		 4394 /* Make sure we can mount this feature set readwrite */
4395 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4396 err = -EROFS;
4397 goto restore_opts;
4398 }
8a266467
TT		 4399 /*
4400 * Make sure the group descriptor checksums
0b8e58a1 4401 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4402 */
4403 for (g = 0; g < sbi->s_groups_count; g++) {
4404 struct ext4_group_desc *gdp =
4405 ext4_get_group_desc(sb, g, NULL);
4406
4407 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4408 ext4_msg(sb, KERN_ERR,
4409 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4410 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4411 le16_to_cpu(gdp->bg_checksum));
4412 err = -EINVAL;
4413 goto restore_opts;
4414 }
4415 }
4416
ead6596b
ES		 4417 /*
4418 * If we have an unprocessed orphan list hanging
4419 * around from a previously readonly bdev mount,
4420 * require a full umount/remount for now.
4421 */
4422 if (es->s_last_orphan) {
b31e1552 4423 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4424 "remount RDWR because of unprocessed "
4425 "orphan inode list. Please "
b31e1552 4426 "umount/remount instead");
ead6596b
ES		 4427 err = -EINVAL;
4428 goto restore_opts;
4429 }
4430
ac27a0ec
DK		 4431 /*
4432 * Mounting a RDONLY partition read-write, so reread
4433 * and store the current valid flag. (It may have
4434 * been changed by e2fsck since we originally mounted
4435 * the partition.)
4436 */
0390131b
FM		 4437 if (sbi->s_journal)
4438 ext4_clear_journal_err(sb, es);
ac27a0ec 4439 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 4440 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 4441 goto restore_opts;
2b2d6d01 4442 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4443 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4444 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4445 EXT4_FEATURE_INCOMPAT_MMP))
4446 if (ext4_multi_mount_protect(sb,
4447 le64_to_cpu(es->s_mmp_block))) {
4448 err = -EROFS;
4449 goto restore_opts;
4450 }
c79d967d 4451 enable_quota = 1;
ac27a0ec
DK		 4452 }
4453 }
bfff6873
LC		 4454
4455 /*
4456 * Reinitialize lazy itable initialization thread based on
4457 * current settings
4458 */
4459 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4460 ext4_unregister_li_request(sb);
4461 else {
4462 ext4_group_t first_not_zeroed;
4463 first_not_zeroed = ext4_has_uninit_itable(sb);
4464 ext4_register_li_request(sb, first_not_zeroed);
4465 }
4466
6fd058f7 4467 ext4_setup_system_zone(sb);
0390131b 4468 if (sbi->s_journal == NULL)
e2d67052 4469 ext4_commit_super(sb, 1);
0390131b 4470
ac27a0ec
DK		 4471#ifdef CONFIG_QUOTA
4472 /* Release old quota file names */
4473 for (i = 0; i < MAXQUOTAS; i++)
4474 if (old_opts.s_qf_names[i] &&
4475 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4476 kfree(old_opts.s_qf_names[i]);
4477#endif
bbd6851a 4478 unlock_super(sb);
c79d967d 4479 if (enable_quota)
0f0dd62f 4480 dquot_resume(sb, -1);
d4c402d9
CW		 4481
4482 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4483 kfree(orig_data);
ac27a0ec 4484 return 0;
0b8e58a1 4485
ac27a0ec
DK		 4486restore_opts:
4487 sb->s_flags = old_sb_flags;
4488 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4489 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4490 sbi->s_resuid = old_opts.s_resuid;
4491 sbi->s_resgid = old_opts.s_resgid;
4492 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4493 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4494 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4495#ifdef CONFIG_QUOTA
4496 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4497 for (i = 0; i < MAXQUOTAS; i++) {
4498 if (sbi->s_qf_names[i] &&
4499 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4500 kfree(sbi->s_qf_names[i]);
4501 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4502 }
4503#endif
bbd6851a 4504 unlock_super(sb);
d4c402d9 4505 kfree(orig_data);
ac27a0ec
DK		 4506 return err;
4507}
4508
2b2d6d01 4509static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4510{
4511 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4512 struct ext4_sb_info *sbi = EXT4_SB(sb);
4513 struct ext4_super_block *es = sbi->s_es;
960cc398 4514 u64 fsid;
d02a9391 4515 s64 bfree;
ac27a0ec 4516
5e70030d
BP		 4517 if (test_opt(sb, MINIX_DF)) {
4518 sbi->s_overhead_last = 0;
6bc9feff 4519 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4520 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4521 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4522
4523 /*
5e70030d
BP		 4524 * Compute the overhead (FS structures). This is constant
4525 * for a given filesystem unless the number of block groups
4526 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4527 */
4528
4529 /*
4530 * All of the blocks before first_data_block are
4531 * overhead
4532 */
4533 overhead = le32_to_cpu(es->s_first_data_block);
4534
4535 /*
4536 * Add the overhead attributed to the superblock and
4537 * block group descriptors. If the sparse superblocks
4538 * feature is turned on, then not all groups have this.
4539 */
4540 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 4541 overhead += ext4_bg_has_super(sb, i) +
4542 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 4543 cond_resched();
4544 }
4545
4546 /*
4547 * Every block group has an inode bitmap, a block
4548 * bitmap, and an inode table.
4549 */
5e70030d
BP		 4550 overhead += ngroups * (2 + sbi->s_itb_per_group);
4551 sbi->s_overhead_last = overhead;
4552 smp_wmb();
6bc9feff 4553 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4554 }
4555
617ba13b 4556 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4557 buf->f_bsize = sb->s_blocksize;
5e70030d 4558 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
d02a9391 4559 bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
6bc6e63f 4560 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
d02a9391
KM		 4561 /* prevent underflow in case that few free space is available */
4562 buf->f_bfree = max_t(s64, bfree, 0);
bd81d8ee
LV		 4563 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4564 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4565 buf->f_bavail = 0;
4566 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4567 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4568 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4569 fsid = le64_to_cpup((void *)es->s_uuid) ^
4570 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4571 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4572 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4573
ac27a0ec
DK		 4574 return 0;
4575}
4576
0b8e58a1
AD		 4577/* Helper function for writing quotas on sync - we need to start transaction
4578 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4579 * Process 1 Process 2
617ba13b 4580 * ext4_create() quota_sync()
a269eb18 4581 * jbd2_journal_start() write_dquot()
871a2931 4582 * dquot_initialize() down(dqio_mutex)
dab291af 4583 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4584 *
4585 */
4586
4587#ifdef CONFIG_QUOTA
4588
4589static inline struct inode *dquot_to_inode(struct dquot *dquot)
4590{
4591 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4592}
4593
617ba13b 4594static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4595{
4596 int ret, err;
4597 handle_t *handle;
4598 struct inode *inode;
4599
4600 inode = dquot_to_inode(dquot);
617ba13b 4601 handle = ext4_journal_start(inode,
0b8e58a1 4602 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4603 if (IS_ERR(handle))
4604 return PTR_ERR(handle);
4605 ret = dquot_commit(dquot);
617ba13b 4606 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4607 if (!ret)
4608 ret = err;
4609 return ret;
4610}
4611
617ba13b 4612static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4613{
4614 int ret, err;
4615 handle_t *handle;
4616
617ba13b 4617 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4618 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4619 if (IS_ERR(handle))
4620 return PTR_ERR(handle);
4621 ret = dquot_acquire(dquot);
617ba13b 4622 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4623 if (!ret)
4624 ret = err;
4625 return ret;
4626}
4627
617ba13b 4628static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4629{
4630 int ret, err;
4631 handle_t *handle;
4632
617ba13b 4633 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4634 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4635 if (IS_ERR(handle)) {
4636 /* Release dquot anyway to avoid endless cycle in dqput() */
4637 dquot_release(dquot);
ac27a0ec 4638 return PTR_ERR(handle);
9c3013e9 4639 }
ac27a0ec 4640 ret = dquot_release(dquot);
617ba13b 4641 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4642 if (!ret)
4643 ret = err;
4644 return ret;
4645}
4646
617ba13b 4647static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4648{
2c8be6b2 4649 /* Are we journaling quotas? */
617ba13b
MC		 4650 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4651 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4652 dquot_mark_dquot_dirty(dquot);
617ba13b 4653 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4654 } else {
4655 return dquot_mark_dquot_dirty(dquot);
4656 }
4657}
4658
617ba13b 4659static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4660{
4661 int ret, err;
4662 handle_t *handle;
4663
4664 /* Data block + inode block */
617ba13b 4665 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4666 if (IS_ERR(handle))
4667 return PTR_ERR(handle);
4668 ret = dquot_commit_info(sb, type);
617ba13b 4669 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4670 if (!ret)
4671 ret = err;
4672 return ret;
4673}
4674
4675/*
4676 * Turn on quotas during mount time - we need to find
4677 * the quota file and such...
4678 */
617ba13b 4679static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4680{
287a8095
CH		 4681 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4682 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4683}
4684
4685/*
4686 * Standard function to be called on quota_on
4687 */
617ba13b 4688static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4689 struct path *path)
ac27a0ec
DK		 4690{
4691 int err;
ac27a0ec
DK		 4692
4693 if (!test_opt(sb, QUOTA))
4694 return -EINVAL;
0623543b 4695
ac27a0ec 4696 /* Quotafile not on the same filesystem? */
f00c9e44 4697 if (path->mnt->mnt_sb != sb)
ac27a0ec 4698 return -EXDEV;
0623543b
JK		 4699 /* Journaling quota? */
4700 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4701 /* Quotafile not in fs root? */
f00c9e44 4702 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4703 ext4_msg(sb, KERN_WARNING,
4704 "Quota file not on filesystem root. "
4705 "Journaled quota will not work");
2b2d6d01 4706 }
0623543b
JK		 4707
4708 /*
4709 * When we journal data on quota file, we have to flush journal to see
4710 * all updates to the file when we bypass pagecache...
4711 */
0390131b 4712 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4713 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4714 /*
4715 * We don't need to lock updates but journal_flush() could
4716 * otherwise be livelocked...
4717 */
4718 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4719 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4720 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4721 if (err)
7ffe1ea8 4722 return err;
0623543b
JK		 4723 }
4724
f00c9e44 4725 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4726}
4727
ca0e05e4
DM		 4728static int ext4_quota_off(struct super_block *sb, int type)
4729{
21f97697
JK		 4730 struct inode *inode = sb_dqopt(sb)->files[type];
4731 handle_t *handle;
4732
87009d86
DM		 4733 /* Force all delayed allocation blocks to be allocated.
4734 * Caller already holds s_umount sem */
4735 if (test_opt(sb, DELALLOC))
ca0e05e4 4736 sync_filesystem(sb);
ca0e05e4 4737
0b268590
AG		 4738 if (!inode)
4739 goto out;
4740
21f97697
JK		 4741 /* Update modification times of quota files when userspace can
4742 * start looking at them */
4743 handle = ext4_journal_start(inode, 1);
4744 if (IS_ERR(handle))
4745 goto out;
4746 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4747 ext4_mark_inode_dirty(handle, inode);
4748 ext4_journal_stop(handle);
4749
4750out:
ca0e05e4
DM		 4751 return dquot_quota_off(sb, type);
4752}
4753
ac27a0ec
DK		 4754/* Read data from quotafile - avoid pagecache and such because we cannot afford
4755 * acquiring the locks... As quota files are never truncated and quota code
25985edc 4756 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 4757 * we don't have to be afraid of races */
617ba13b 4758static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4759 size_t len, loff_t off)
4760{
4761 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4762 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4763 int err = 0;
4764 int offset = off & (sb->s_blocksize - 1);
4765 int tocopy;
4766 size_t toread;
4767 struct buffer_head *bh;
4768 loff_t i_size = i_size_read(inode);
4769
4770 if (off > i_size)
4771 return 0;
4772 if (off+len > i_size)
4773 len = i_size-off;
4774 toread = len;
4775 while (toread > 0) {
4776 tocopy = sb->s_blocksize - offset < toread ?
4777 sb->s_blocksize - offset : toread;
617ba13b 4778 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4779 if (err)
4780 return err;
4781 if (!bh) /* A hole? */
4782 memset(data, 0, tocopy);
4783 else
4784 memcpy(data, bh->b_data+offset, tocopy);
4785 brelse(bh);
4786 offset = 0;
4787 toread -= tocopy;
4788 data += tocopy;
4789 blk++;
4790 }
4791 return len;
4792}
4793
4794/* Write to quotafile (we know the transaction is already started and has
4795 * enough credits) */
617ba13b 4796static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4797 const char *data, size_t len, loff_t off)
4798{
4799 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4800 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4801 int err = 0;
4802 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4803 struct buffer_head *bh;
4804 handle_t *handle = journal_current_handle();
4805
0390131b 4806 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4807 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4808 " cancelled because transaction is not started",
9c3013e9
JK		 4809 (unsigned long long)off, (unsigned long long)len);
4810 return -EIO;
4811 }
67eeb568
DM		 4812 /*
4813 * Since we account only one data block in transaction credits,
4814 * then it is impossible to cross a block boundary.
4815 */
4816 if (sb->s_blocksize - offset < len) {
4817 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4818 " cancelled because not block aligned",
4819 (unsigned long long)off, (unsigned long long)len);
4820 return -EIO;
4821 }
4822
ac27a0ec 4823 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4824 bh = ext4_bread(handle, inode, blk, 1, &err);
4825 if (!bh)
4826 goto out;
62d2b5f2
JK		 4827 err = ext4_journal_get_write_access(handle, bh);
4828 if (err) {
4829 brelse(bh);
4830 goto out;
ac27a0ec 4831 }
67eeb568
DM		 4832 lock_buffer(bh);
4833 memcpy(bh->b_data+offset, data, len);
4834 flush_dcache_page(bh->b_page);
4835 unlock_buffer(bh);
62d2b5f2 4836 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4837 brelse(bh);
ac27a0ec 4838out:
67eeb568 4839 if (err) {
4d04e4fb 4840 mutex_unlock(&inode->i_mutex);
ac27a0ec 4841 return err;
4d04e4fb 4842 }
67eeb568
DM		 4843 if (inode->i_size < off + len) {
4844 i_size_write(inode, off + len);
617ba13b 4845 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 4846 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4847 }
ac27a0ec 4848 mutex_unlock(&inode->i_mutex);
67eeb568 4849 return len;
ac27a0ec
DK		 4850}
4851
4852#endif
4853
152a0836
AV		 4854static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4855 const char *dev_name, void *data)
ac27a0ec 4856{
152a0836 4857 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4858}
4859
37f328eb 4860#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4861static inline void register_as_ext2(void)
4862{
4863 int err = register_filesystem(&ext2_fs_type);
4864 if (err)
4865 printk(KERN_WARNING
4866 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4867}
4868
4869static inline void unregister_as_ext2(void)
4870{
4871 unregister_filesystem(&ext2_fs_type);
4872}
2035e776
TT		 4873
4874static inline int ext2_feature_set_ok(struct super_block *sb)
4875{
4876 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4877 return 0;
4878 if (sb->s_flags & MS_RDONLY)
4879 return 1;
4880 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4881 return 0;
4882 return 1;
4883}
51b7e3c9 4884MODULE_ALIAS("ext2");
24b58424
TT		 4885#else
4886static inline void register_as_ext2(void) { }
4887static inline void unregister_as_ext2(void) { }
2035e776 4888static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4889#endif
4890
37f328eb 4891#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4892static inline void register_as_ext3(void)
4893{
4894 int err = register_filesystem(&ext3_fs_type);
4895 if (err)
4896 printk(KERN_WARNING
4897 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4898}
4899
4900static inline void unregister_as_ext3(void)
4901{
4902 unregister_filesystem(&ext3_fs_type);
4903}
2035e776
TT		 4904
4905static inline int ext3_feature_set_ok(struct super_block *sb)
4906{
4907 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4908 return 0;
4909 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4910 return 0;
4911 if (sb->s_flags & MS_RDONLY)
4912 return 1;
4913 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4914 return 0;
4915 return 1;
4916}
51b7e3c9 4917MODULE_ALIAS("ext3");
24b58424
TT		 4918#else
4919static inline void register_as_ext3(void) { }
4920static inline void unregister_as_ext3(void) { }
2035e776 4921static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4922#endif
4923
03010a33
TT		 4924static struct file_system_type ext4_fs_type = {
4925 .owner = THIS_MODULE,
4926 .name = "ext4",
152a0836 4927 .mount = ext4_mount,
03010a33
TT		 4928 .kill_sb = kill_block_super,
4929 .fs_flags = FS_REQUIRES_DEV,
4930};
4931
8f021222 4932static int __init ext4_init_feat_adverts(void)
857ac889
LC		 4933{
4934 struct ext4_features *ef;
4935 int ret = -ENOMEM;
4936
4937 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4938 if (!ef)
4939 goto out;
4940
4941 ef->f_kobj.kset = ext4_kset;
4942 init_completion(&ef->f_kobj_unregister);
4943 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4944 "features");
4945 if (ret) {
4946 kfree(ef);
4947 goto out;
4948 }
4949
4950 ext4_feat = ef;
4951 ret = 0;
4952out:
4953 return ret;
4954}
4955
8f021222
LC		 4956static void ext4_exit_feat_adverts(void)
4957{
4958 kobject_put(&ext4_feat->f_kobj);
4959 wait_for_completion(&ext4_feat->f_kobj_unregister);
4960 kfree(ext4_feat);
4961}
4962
e9e3bcec
ES		 4963/* Shared across all ext4 file systems */
4964wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4965struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4966
5dabfc78 4967static int __init ext4_init_fs(void)
ac27a0ec 4968{
e9e3bcec 4969 int i, err;
c9de560d 4970
12e9b892 4971 ext4_check_flag_values();
e9e3bcec
ES		 4972
4973 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4974 mutex_init(&ext4__aio_mutex[i]);
4975 init_waitqueue_head(&ext4__ioend_wq[i]);
4976 }
4977
5dabfc78 4978 err = ext4_init_pageio();
6fd058f7
TT		 4979 if (err)
4980 return err;
5dabfc78 4981 err = ext4_init_system_zone();
bd2d0210 4982 if (err)
dd68314c 4983 goto out7;
3197ebdb
TT		 4984 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4985 if (!ext4_kset)
dd68314c 4986 goto out6;
9f6200bb 4987 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
dd68314c
TT		 4988 if (!ext4_proc_root)
4989 goto out5;
857ac889
LC		 4990
4991 err = ext4_init_feat_adverts();
dd68314c
TT		 4992 if (err)
4993 goto out4;
857ac889 4994
5dabfc78 4995 err = ext4_init_mballoc();
ac27a0ec 4996 if (err)
6fd058f7 4997 goto out3;
c9de560d 4998
5dabfc78 4999 err = ext4_init_xattr();
c9de560d
AT		 5000 if (err)
5001 goto out2;
ac27a0ec
DK		 5002 err = init_inodecache();
5003 if (err)
5004 goto out1;
24b58424 5005 register_as_ext3();
2035e776 5006 register_as_ext2();
03010a33 5007 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5008 if (err)
5009 goto out;
bfff6873
LC		 5010
5011 ext4_li_info = NULL;
5012 mutex_init(&ext4_li_mtx);
ac27a0ec
DK		 5013 return 0;
5014out:
24b58424
TT		 5015 unregister_as_ext2();
5016 unregister_as_ext3();
ac27a0ec
DK		 5017 destroy_inodecache();
5018out1:
5dabfc78 5019 ext4_exit_xattr();
c9de560d 5020out2:
5dabfc78 5021 ext4_exit_mballoc();
6fd058f7 5022out3:
8f021222 5023 ext4_exit_feat_adverts();
dd68314c 5024out4:
6fd058f7 5025 remove_proc_entry("fs/ext4", NULL);
dd68314c 5026out5:
6fd058f7 5027 kset_unregister(ext4_kset);
dd68314c 5028out6:
5dabfc78 5029 ext4_exit_system_zone();
dd68314c 5030out7:
5dabfc78 5031 ext4_exit_pageio();
ac27a0ec
DK		 5032 return err;
5033}
5034
5dabfc78 5035static void __exit ext4_exit_fs(void)
ac27a0ec 5036{
bfff6873 5037 ext4_destroy_lazyinit_thread();
24b58424
TT		 5038 unregister_as_ext2();
5039 unregister_as_ext3();
03010a33 5040 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5041 destroy_inodecache();
5dabfc78
TT		 5042 ext4_exit_xattr();
5043 ext4_exit_mballoc();
8f021222 5044 ext4_exit_feat_adverts();
9f6200bb 5045 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5046 kset_unregister(ext4_kset);
5dabfc78
TT		 5047 ext4_exit_system_zone();
5048 ext4_exit_pageio();
ac27a0ec
DK		 5049}
5050
5051MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5052MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5053MODULE_LICENSE("GPL");
5dabfc78
TT		 5054module_init(ext4_init_fs)
5055module_exit(ext4_exit_fs)
kernel source for telekom puls (alcatel/mediatek)