]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Fix endianness bug when reading the MMP block
[mirror_ubuntu-artful-kernel.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
6f91bc5f 48#include "ext4_extents.h"
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
e2d67052 65static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 66static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
617ba13b 70static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 71static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 72 char nbuf[16]);
2b2d6d01
TT		 73static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 75static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 76static void ext4_write_super(struct super_block *sb);
c4be0c1d 77static int ext4_freeze(struct super_block *sb);
152a0836
AV		 78static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
2035e776
TT		 80static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 82static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 83static void ext4_destroy_lazyinit_thread(void);
84static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 85static void ext4_clear_request_list(void);
ac27a0ec 86
2035e776
TT		 87#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94};
95#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96#else
97#define IS_EXT2_SB(sb) (0)
98#endif
99
100
ba69f9ab
JK		 101#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
104 .name = "ext3",
152a0836 105 .mount = ext4_mount,
ba69f9ab
JK		 106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
108};
109#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110#else
111#define IS_EXT3_SB(sb) (0)
112#endif
bd81d8ee 113
9933fc0a
TT		 114void *ext4_kvmalloc(size_t size, gfp_t flags)
115{
116 void *ret;
117
118 ret = kmalloc(size, flags);
119 if (!ret)
120 ret = __vmalloc(size, flags, PAGE_KERNEL);
121 return ret;
122}
123
124void *ext4_kvzalloc(size_t size, gfp_t flags)
125{
126 void *ret;
127
db9481c0 128 ret = kzalloc(size, flags);
9933fc0a
TT		 129 if (!ret)
130 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
131 return ret;
132}
133
134void ext4_kvfree(void *ptr)
135{
136 if (is_vmalloc_addr(ptr))
137 vfree(ptr);
138 else
139 kfree(ptr);
140
141}
142
8fadc143
AR		 143ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
144 struct ext4_group_desc *bg)
bd81d8ee 145{
3a14589c 146 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 148 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 149}
150
8fadc143
AR		 151ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
152 struct ext4_group_desc *bg)
bd81d8ee 153{
5272f837 154 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 155 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 156 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 157}
158
8fadc143
AR		 159ext4_fsblk_t ext4_inode_table(struct super_block *sb,
160 struct ext4_group_desc *bg)
bd81d8ee 161{
5272f837 162 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 163 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 164 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 165}
166
021b65bb
TT		 167__u32 ext4_free_group_clusters(struct super_block *sb,
168 struct ext4_group_desc *bg)
560671a0
AK		 169{
170 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
171 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 172 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 173}
174
175__u32 ext4_free_inodes_count(struct super_block *sb,
176 struct ext4_group_desc *bg)
177{
178 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
179 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 180 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 181}
182
183__u32 ext4_used_dirs_count(struct super_block *sb,
184 struct ext4_group_desc *bg)
185{
186 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
187 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 188 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 189}
190
191__u32 ext4_itable_unused_count(struct super_block *sb,
192 struct ext4_group_desc *bg)
193{
194 return le16_to_cpu(bg->bg_itable_unused_lo) |
195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 196 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 197}
198
8fadc143
AR		 199void ext4_block_bitmap_set(struct super_block *sb,
200 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 201{
3a14589c 202 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 205}
206
8fadc143
AR		 207void ext4_inode_bitmap_set(struct super_block *sb,
208 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 209{
5272f837 210 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 211 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
212 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 213}
214
8fadc143
AR		 215void ext4_inode_table_set(struct super_block *sb,
216 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 217{
5272f837 218 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 219 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
220 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 221}
222
021b65bb
TT		 223void ext4_free_group_clusters_set(struct super_block *sb,
224 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 225{
226 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
227 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
228 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
229}
230
231void ext4_free_inodes_set(struct super_block *sb,
232 struct ext4_group_desc *bg, __u32 count)
233{
234 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
235 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
236 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
237}
238
239void ext4_used_dirs_set(struct super_block *sb,
240 struct ext4_group_desc *bg, __u32 count)
241{
242 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
243 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
244 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
245}
246
247void ext4_itable_unused_set(struct super_block *sb,
248 struct ext4_group_desc *bg, __u32 count)
249{
250 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
253}
254
d3d1faf6
CW		 255
256/* Just increment the non-pointer handle value */
257static handle_t *ext4_get_nojournal(void)
258{
259 handle_t *handle = current->journal_info;
260 unsigned long ref_cnt = (unsigned long)handle;
261
262 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
263
264 ref_cnt++;
265 handle = (handle_t *)ref_cnt;
266
267 current->journal_info = handle;
268 return handle;
269}
270
271
272/* Decrement the non-pointer handle value */
273static void ext4_put_nojournal(handle_t *handle)
274{
275 unsigned long ref_cnt = (unsigned long)handle;
276
277 BUG_ON(ref_cnt == 0);
278
279 ref_cnt--;
280 handle = (handle_t *)ref_cnt;
281
282 current->journal_info = handle;
283}
284
ac27a0ec 285/*
dab291af 286 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 287 *
288 * The only special thing we need to do here is to make sure that all
289 * journal_end calls result in the superblock being marked dirty, so
290 * that sync() will call the filesystem's write_super callback if
291 * appropriate.
be4f27d3
YY		 292 *
293 * To avoid j_barrier hold in userspace when a user calls freeze(),
294 * ext4 prevents a new handle from being started by s_frozen, which
295 * is in an upper layer.
ac27a0ec 296 */
617ba13b 297handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 298{
299 journal_t *journal;
be4f27d3 300 handle_t *handle;
ac27a0ec 301
12706394 302 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 303 if (sb->s_flags & MS_RDONLY)
304 return ERR_PTR(-EROFS);
305
617ba13b 306 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 307 handle = ext4_journal_current_handle();
308
309 /*
310 * If a handle has been started, it should be allowed to
311 * finish, otherwise deadlock could happen between freeze
312 * and others(e.g. truncate) due to the restart of the
313 * journal handle if the filesystem is forzen and active
314 * handles are not stopped.
315 */
316 if (!handle)
317 vfs_check_frozen(sb, SB_FREEZE_TRANS);
318
319 if (!journal)
320 return ext4_get_nojournal();
321 /*
322 * Special case here: if the journal has aborted behind our
323 * backs (eg. EIO in the commit thread), then we still need to
324 * take the FS itself readonly cleanly.
325 */
326 if (is_journal_aborted(journal)) {
327 ext4_abort(sb, "Detected aborted journal");
328 return ERR_PTR(-EROFS);
ac27a0ec 329 }
be4f27d3 330 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 331}
332
333/*
334 * The only special thing we need to do here is to make sure that all
dab291af 335 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 336 * that sync() will call the filesystem's write_super callback if
337 * appropriate.
338 */
c398eda0 339int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 340{
341 struct super_block *sb;
342 int err;
343 int rc;
344
0390131b 345 if (!ext4_handle_valid(handle)) {
d3d1faf6 346 ext4_put_nojournal(handle);
0390131b
FM		 347 return 0;
348 }
ac27a0ec
DK		 349 sb = handle->h_transaction->t_journal->j_private;
350 err = handle->h_err;
dab291af 351 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 352
353 if (!err)
354 err = rc;
355 if (err)
c398eda0 356 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 357 return err;
358}
359
90c7201b
TT		 360void ext4_journal_abort_handle(const char *caller, unsigned int line,
361 const char *err_fn, struct buffer_head *bh,
362 handle_t *handle, int err)
ac27a0ec
DK		 363{
364 char nbuf[16];
617ba13b 365 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 366
0390131b
FM		 367 BUG_ON(!ext4_handle_valid(handle));
368
ac27a0ec
DK		 369 if (bh)
370 BUFFER_TRACE(bh, "abort");
371
372 if (!handle->h_err)
373 handle->h_err = err;
374
375 if (is_handle_aborted(handle))
376 return;
377
90c7201b
TT		 378 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
379 caller, line, errstr, err_fn);
ac27a0ec 380
dab291af 381 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 382}
383
1c13d5c0
TT		 384static void __save_error_info(struct super_block *sb, const char *func,
385 unsigned int line)
386{
387 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
388
389 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
390 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
391 es->s_last_error_time = cpu_to_le32(get_seconds());
392 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
393 es->s_last_error_line = cpu_to_le32(line);
394 if (!es->s_first_error_time) {
395 es->s_first_error_time = es->s_last_error_time;
396 strncpy(es->s_first_error_func, func,
397 sizeof(es->s_first_error_func));
398 es->s_first_error_line = cpu_to_le32(line);
399 es->s_first_error_ino = es->s_last_error_ino;
400 es->s_first_error_block = es->s_last_error_block;
401 }
66e61a9e
TT		 402 /*
403 * Start the daily error reporting function if it hasn't been
404 * started already
405 */
406 if (!es->s_error_count)
407 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 408 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
409}
410
411static void save_error_info(struct super_block *sb, const char *func,
412 unsigned int line)
413{
414 __save_error_info(sb, func, line);
415 ext4_commit_super(sb, 1);
416}
417
7c2e7087
TT		 418/*
419 * The del_gendisk() function uninitializes the disk-specific data
420 * structures, including the bdi structure, without telling anyone
421 * else. Once this happens, any attempt to call mark_buffer_dirty()
422 * (for example, by ext4_commit_super), will cause a kernel OOPS.
423 * This is a kludge to prevent these oops until we can put in a proper
424 * hook in del_gendisk() to inform the VFS and file system layers.
425 */
426static int block_device_ejected(struct super_block *sb)
427{
428 struct inode *bd_inode = sb->s_bdev->bd_inode;
429 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
430
431 return bdi->dev == NULL;
432}
433
18aadd47
BJ		 434static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
435{
436 struct super_block *sb = journal->j_private;
437 struct ext4_sb_info *sbi = EXT4_SB(sb);
438 int error = is_journal_aborted(journal);
439 struct ext4_journal_cb_entry *jce, *tmp;
440
441 spin_lock(&sbi->s_md_lock);
442 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
443 list_del_init(&jce->jce_list);
444 spin_unlock(&sbi->s_md_lock);
445 jce->jce_func(sb, jce, error);
446 spin_lock(&sbi->s_md_lock);
447 }
448 spin_unlock(&sbi->s_md_lock);
449}
1c13d5c0 450
ac27a0ec
DK		 451/* Deal with the reporting of failure conditions on a filesystem such as
452 * inconsistencies detected or read IO failures.
453 *
454 * On ext2, we can store the error state of the filesystem in the
617ba13b 455 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 456 * write ordering constraints on the superblock which prevent us from
457 * writing it out straight away; and given that the journal is about to
458 * be aborted, we can't rely on the current, or future, transactions to
459 * write out the superblock safely.
460 *
dab291af 461 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 462 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 463 * that error until we've noted it down and cleared it.
464 */
465
617ba13b 466static void ext4_handle_error(struct super_block *sb)
ac27a0ec 467{
ac27a0ec
DK		 468 if (sb->s_flags & MS_RDONLY)
469 return;
470
2b2d6d01 471 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 472 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 473
4ab2f15b 474 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 475 if (journal)
dab291af 476 jbd2_journal_abort(journal, -EIO);
ac27a0ec 477 }
2b2d6d01 478 if (test_opt(sb, ERRORS_RO)) {
b31e1552 479 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 480 sb->s_flags |= MS_RDONLY;
481 }
ac27a0ec 482 if (test_opt(sb, ERRORS_PANIC))
617ba13b 483 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 484 sb->s_id);
485}
486
12062ddd 487void __ext4_error(struct super_block *sb, const char *function,
c398eda0 488 unsigned int line, const char *fmt, ...)
ac27a0ec 489{
0ff2ea7d 490 struct va_format vaf;
ac27a0ec
DK		 491 va_list args;
492
493 va_start(args, fmt);
0ff2ea7d
JP		 494 vaf.fmt = fmt;
495 vaf.va = &args;
496 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
497 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec
DK		 498 va_end(args);
499
617ba13b 500 ext4_handle_error(sb);
ac27a0ec
DK		 501}
502
c398eda0
TT		 503void ext4_error_inode(struct inode *inode, const char *function,
504 unsigned int line, ext4_fsblk_t block,
273df556
FM		 505 const char *fmt, ...)
506{
507 va_list args;
f7c21177 508 struct va_format vaf;
1c13d5c0 509 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 510
1c13d5c0
TT		 511 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
512 es->s_last_error_block = cpu_to_le64(block);
513 save_error_info(inode->i_sb, function, line);
273df556 514 va_start(args, fmt);
f7c21177
TT		 515 vaf.fmt = fmt;
516 vaf.va = &args;
c398eda0
TT		 517 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
518 inode->i_sb->s_id, function, line, inode->i_ino);
519 if (block)
f7c21177
TT		 520 printk(KERN_CONT "block %llu: ", block);
521 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
273df556
FM		 522 va_end(args);
523
524 ext4_handle_error(inode->i_sb);
525}
526
c398eda0 527void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 528 unsigned int line, ext4_fsblk_t block,
529 const char *fmt, ...)
273df556
FM		 530{
531 va_list args;
f7c21177 532 struct va_format vaf;
1c13d5c0 533 struct ext4_super_block *es;
273df556
FM		 534 struct inode *inode = file->f_dentry->d_inode;
535 char pathname[80], *path;
536
1c13d5c0
TT		 537 es = EXT4_SB(inode->i_sb)->s_es;
538 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
539 save_error_info(inode->i_sb, function, line);
273df556 540 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 541 if (IS_ERR(path))
273df556
FM		 542 path = "(unknown)";
543 printk(KERN_CRIT
f7c21177
TT		 544 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
545 inode->i_sb->s_id, function, line, inode->i_ino);
546 if (block)
547 printk(KERN_CONT "block %llu: ", block);
548 va_start(args, fmt);
549 vaf.fmt = fmt;
550 vaf.va = &args;
551 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
273df556
FM		 552 va_end(args);
553
554 ext4_handle_error(inode->i_sb);
555}
556
2b2d6d01 557static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 558 char nbuf[16])
559{
560 char *errstr = NULL;
561
562 switch (errno) {
563 case -EIO:
564 errstr = "IO failure";
565 break;
566 case -ENOMEM:
567 errstr = "Out of memory";
568 break;
569 case -EROFS:
78f1ddbb
TT		 570 if (!sb || (EXT4_SB(sb)->s_journal &&
571 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 572 errstr = "Journal has aborted";
573 else
574 errstr = "Readonly filesystem";
575 break;
576 default:
577 /* If the caller passed in an extra buffer for unknown
578 * errors, textualise them now. Else we just return
579 * NULL. */
580 if (nbuf) {
581 /* Check for truncated error codes... */
582 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
583 errstr = nbuf;
584 }
585 break;
586 }
587
588 return errstr;
589}
590
617ba13b 591/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 592 * automatically and invokes the appropriate error response. */
593
c398eda0
TT		 594void __ext4_std_error(struct super_block *sb, const char *function,
595 unsigned int line, int errno)
ac27a0ec
DK		 596{
597 char nbuf[16];
598 const char *errstr;
599
600 /* Special case: if the error is EROFS, and we're not already
601 * inside a transaction, then there's really no point in logging
602 * an error. */
603 if (errno == -EROFS && journal_current_handle() == NULL &&
604 (sb->s_flags & MS_RDONLY))
605 return;
606
617ba13b 607 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 608 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
609 sb->s_id, function, line, errstr);
1c13d5c0 610 save_error_info(sb, function, line);
ac27a0ec 611
617ba13b 612 ext4_handle_error(sb);
ac27a0ec
DK		 613}
614
615/*
617ba13b 616 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 617 * abort function may be used to deal with unrecoverable failures such
618 * as journal IO errors or ENOMEM at a critical moment in log management.
619 *
620 * We unconditionally force the filesystem into an ABORT|READONLY state,
621 * unless the error response on the fs has been set to panic in which
622 * case we take the easy way out and panic immediately.
623 */
624
c67d859e 625void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 626 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 627{
628 va_list args;
629
1c13d5c0 630 save_error_info(sb, function, line);
ac27a0ec 631 va_start(args, fmt);
c398eda0
TT		 632 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
633 function, line);
ac27a0ec
DK		 634 vprintk(fmt, args);
635 printk("\n");
636 va_end(args);
637
1c13d5c0
TT		 638 if ((sb->s_flags & MS_RDONLY) == 0) {
639 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
640 sb->s_flags |= MS_RDONLY;
641 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
642 if (EXT4_SB(sb)->s_journal)
643 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
644 save_error_info(sb, function, line);
645 }
ac27a0ec 646 if (test_opt(sb, ERRORS_PANIC))
617ba13b 647 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 648}
649
0ff2ea7d 650void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 651{
0ff2ea7d 652 struct va_format vaf;
b31e1552
ES		 653 va_list args;
654
655 va_start(args, fmt);
0ff2ea7d
JP		 656 vaf.fmt = fmt;
657 vaf.va = &args;
658 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 659 va_end(args);
660}
661
12062ddd 662void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 663 unsigned int line, const char *fmt, ...)
ac27a0ec 664{
0ff2ea7d 665 struct va_format vaf;
ac27a0ec
DK		 666 va_list args;
667
668 va_start(args, fmt);
0ff2ea7d
JP		 669 vaf.fmt = fmt;
670 vaf.va = &args;
671 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
672 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 673 va_end(args);
674}
675
e29136f8
TT		 676void __ext4_grp_locked_error(const char *function, unsigned int line,
677 struct super_block *sb, ext4_group_t grp,
678 unsigned long ino, ext4_fsblk_t block,
679 const char *fmt, ...)
5d1b1b3f
AK		 680__releases(bitlock)
681__acquires(bitlock)
682{
0ff2ea7d 683 struct va_format vaf;
5d1b1b3f
AK		 684 va_list args;
685 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
686
1c13d5c0
TT		 687 es->s_last_error_ino = cpu_to_le32(ino);
688 es->s_last_error_block = cpu_to_le64(block);
689 __save_error_info(sb, function, line);
0ff2ea7d 690
5d1b1b3f 691 va_start(args, fmt);
0ff2ea7d
JP		 692
693 vaf.fmt = fmt;
694 vaf.va = &args;
21149d61 695 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 696 sb->s_id, function, line, grp);
697 if (ino)
0ff2ea7d 698 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 699 if (block)
0ff2ea7d
JP		 700 printk(KERN_CONT "block %llu:", (unsigned long long) block);
701 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 702 va_end(args);
703
704 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 705 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 706 return;
707 }
1c13d5c0 708
5d1b1b3f
AK		 709 ext4_unlock_group(sb, grp);
710 ext4_handle_error(sb);
711 /*
712 * We only get here in the ERRORS_RO case; relocking the group
713 * may be dangerous, but nothing bad will happen since the
714 * filesystem will have already been marked read/only and the
715 * journal has been aborted. We return 1 as a hint to callers
716 * who might what to use the return value from
25985edc 717 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 718 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
719 * aggressively from the ext4 function in question, with a
720 * more appropriate error code.
721 */
722 ext4_lock_group(sb, grp);
723 return;
724}
725
617ba13b 726void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 727{
617ba13b 728 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 729
617ba13b 730 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 731 return;
732
12062ddd 733 ext4_warning(sb,
ac27a0ec
DK		 734 "updating to rev %d because of new feature flag, "
735 "running e2fsck is recommended",
617ba13b 736 EXT4_DYNAMIC_REV);
ac27a0ec 737
617ba13b
MC		 738 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
739 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
740 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 741 /* leave es->s_feature_*compat flags alone */
742 /* es->s_uuid will be set by e2fsck if empty */
743
744 /*
745 * The rest of the superblock fields should be zero, and if not it
746 * means they are likely already in use, so leave them alone. We
747 * can leave it up to e2fsck to clean up any inconsistencies there.
748 */
749}
750
751/*
752 * Open the external journal device
753 */
b31e1552 754static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 755{
756 struct block_device *bdev;
757 char b[BDEVNAME_SIZE];
758
d4d77629 759 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 760 if (IS_ERR(bdev))
761 goto fail;
762 return bdev;
763
764fail:
b31e1552 765 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 766 __bdevname(dev, b), PTR_ERR(bdev));
767 return NULL;
768}
769
770/*
771 * Release the journal device
772 */
617ba13b 773static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 774{
e525fd89 775 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 776}
777
617ba13b 778static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 779{
780 struct block_device *bdev;
781 int ret = -ENODEV;
782
783 bdev = sbi->journal_bdev;
784 if (bdev) {
617ba13b 785 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 786 sbi->journal_bdev = NULL;
787 }
788 return ret;
789}
790
791static inline struct inode *orphan_list_entry(struct list_head *l)
792{
617ba13b 793 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 794}
795
617ba13b 796static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 797{
798 struct list_head *l;
799
b31e1552
ES		 800 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
801 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 802
803 printk(KERN_ERR "sb_info orphan list:\n");
804 list_for_each(l, &sbi->s_orphan) {
805 struct inode *inode = orphan_list_entry(l);
806 printk(KERN_ERR " "
807 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
808 inode->i_sb->s_id, inode->i_ino, inode,
809 inode->i_mode, inode->i_nlink,
810 NEXT_ORPHAN(inode));
811 }
812}
813
2b2d6d01 814static void ext4_put_super(struct super_block *sb)
ac27a0ec 815{
617ba13b
MC		 816 struct ext4_sb_info *sbi = EXT4_SB(sb);
817 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 818 int i, err;
ac27a0ec 819
857ac889 820 ext4_unregister_li_request(sb);
e0ccfd95
CH		 821 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
822
4c0425ff
MC		 823 flush_workqueue(sbi->dio_unwritten_wq);
824 destroy_workqueue(sbi->dio_unwritten_wq);
825
a9e220f8 826 lock_super(sb);
8c85e125 827 if (sb->s_dirt)
ebc1ac16 828 ext4_commit_super(sb, 1);
8c85e125 829
0390131b
FM		 830 if (sbi->s_journal) {
831 err = jbd2_journal_destroy(sbi->s_journal);
832 sbi->s_journal = NULL;
833 if (err < 0)
c67d859e 834 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 835 }
d4edac31 836
a1c6c569 837 del_timer(&sbi->s_err_report);
d4edac31
JB		 838 ext4_release_system_zone(sb);
839 ext4_mb_release(sb);
840 ext4_ext_release(sb);
841 ext4_xattr_put_super(sb);
842
ac27a0ec 843 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 844 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 845 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 846 ext4_commit_super(sb, 1);
ac27a0ec 847 }
240799cd 848 if (sbi->s_proc) {
9f6200bb 849 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 850 }
3197ebdb 851 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 852
853 for (i = 0; i < sbi->s_gdb_count; i++)
854 brelse(sbi->s_group_desc[i]);
f18a5f21 855 ext4_kvfree(sbi->s_group_desc);
9933fc0a 856 ext4_kvfree(sbi->s_flex_groups);
57042651 857 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 858 percpu_counter_destroy(&sbi->s_freeinodes_counter);
859 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 860 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 861 brelse(sbi->s_sbh);
862#ifdef CONFIG_QUOTA
863 for (i = 0; i < MAXQUOTAS; i++)
864 kfree(sbi->s_qf_names[i]);
865#endif
866
867 /* Debugging code just in case the in-memory inode orphan list
868 * isn't empty. The on-disk one can be non-empty if we've
869 * detected an error and taken the fs readonly, but the
870 * in-memory list had better be clean by this point. */
871 if (!list_empty(&sbi->s_orphan))
872 dump_orphan_list(sb, sbi);
873 J_ASSERT(list_empty(&sbi->s_orphan));
874
f98393a6 875 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 876 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
877 /*
878 * Invalidate the journal device's buffers. We don't want them
879 * floating about in memory - the physical journal device may
880 * hotswapped, and it breaks the `ro-after' testing code.
881 */
882 sync_blockdev(sbi->journal_bdev);
f98393a6 883 invalidate_bdev(sbi->journal_bdev);
617ba13b 884 ext4_blkdev_remove(sbi);
ac27a0ec 885 }
c5e06d10
JL		 886 if (sbi->s_mmp_tsk)
887 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 888 sb->s_fs_info = NULL;
3197ebdb
TT		 889 /*
890 * Now that we are completely done shutting down the
891 * superblock, we need to actually destroy the kobject.
892 */
3197ebdb
TT		 893 unlock_super(sb);
894 kobject_put(&sbi->s_kobj);
895 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 896 kfree(sbi->s_blockgroup_lock);
ac27a0ec 897 kfree(sbi);
ac27a0ec
DK		 898}
899
e18b890b 900static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 901
902/*
903 * Called inside transaction, so use GFP_NOFS
904 */
617ba13b 905static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 906{
617ba13b 907 struct ext4_inode_info *ei;
ac27a0ec 908
e6b4f8da 909 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 910 if (!ei)
911 return NULL;
0b8e58a1 912
ac27a0ec 913 ei->vfs_inode.i_version = 1;
91246c00 914 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 915 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 916 INIT_LIST_HEAD(&ei->i_prealloc_list);
917 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 918 ei->i_reserved_data_blocks = 0;
919 ei->i_reserved_meta_blocks = 0;
920 ei->i_allocated_meta_blocks = 0;
9d0be502 921 ei->i_da_metadata_calc_len = 0;
d2a17637 922 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 923#ifdef CONFIG_QUOTA
924 ei->i_reserved_quota = 0;
925#endif
8aefcd55 926 ei->jinode = NULL;
c7064ef1 927 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 928 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 929 ei->cur_aio_dio = NULL;
b436b9be
JK		 930 ei->i_sync_tid = 0;
931 ei->i_datasync_tid = 0;
f7ad6d2e 932 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 933 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 934
ac27a0ec
DK		 935 return &ei->vfs_inode;
936}
937
7ff9c073
TT		 938static int ext4_drop_inode(struct inode *inode)
939{
940 int drop = generic_drop_inode(inode);
941
942 trace_ext4_drop_inode(inode, drop);
943 return drop;
944}
945
fa0d7e3d
NP		 946static void ext4_i_callback(struct rcu_head *head)
947{
948 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 949 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
950}
951
617ba13b 952static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 953{
9f7dd93d 954 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 955 ext4_msg(inode->i_sb, KERN_ERR,
956 "Inode %lu (%p): orphan list check failed!",
957 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 958 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
959 EXT4_I(inode), sizeof(struct ext4_inode_info),
960 true);
961 dump_stack();
962 }
fa0d7e3d 963 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 964}
965
51cc5068 966static void init_once(void *foo)
ac27a0ec 967{
617ba13b 968 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 969
a35afb83 970 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 971#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 972 init_rwsem(&ei->xattr_sem);
ac27a0ec 973#endif
0e855ac8 974 init_rwsem(&ei->i_data_sem);
a35afb83 975 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 976}
977
978static int init_inodecache(void)
979{
617ba13b
MC		 980 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
981 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 982 0, (SLAB_RECLAIM_ACCOUNT|
983 SLAB_MEM_SPREAD),
20c2df83 984 init_once);
617ba13b 985 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 986 return -ENOMEM;
987 return 0;
988}
989
990static void destroy_inodecache(void)
991{
617ba13b 992 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 993}
994
0930fcc1 995void ext4_clear_inode(struct inode *inode)
ac27a0ec 996{
0930fcc1
AV		 997 invalidate_inode_buffers(inode);
998 end_writeback(inode);
9f754758 999 dquot_drop(inode);
c2ea3fde 1000 ext4_discard_preallocations(inode);
8aefcd55
TT		 1001 if (EXT4_I(inode)->jinode) {
1002 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1003 EXT4_I(inode)->jinode);
1004 jbd2_free_inode(EXT4_I(inode)->jinode);
1005 EXT4_I(inode)->jinode = NULL;
1006 }
ac27a0ec
DK		 1007}
1008
2b2d6d01
TT		 1009static inline void ext4_show_quota_options(struct seq_file *seq,
1010 struct super_block *sb)
ac27a0ec
DK		 1011{
1012#if defined(CONFIG_QUOTA)
617ba13b 1013 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 1014
5a20bdfc
JK		 1015 if (sbi->s_jquota_fmt) {
1016 char *fmtname = "";
1017
1018 switch (sbi->s_jquota_fmt) {
1019 case QFMT_VFS_OLD:
1020 fmtname = "vfsold";
1021 break;
1022 case QFMT_VFS_V0:
1023 fmtname = "vfsv0";
1024 break;
1025 case QFMT_VFS_V1:
1026 fmtname = "vfsv1";
1027 break;
1028 }
1029 seq_printf(seq, ",jqfmt=%s", fmtname);
1030 }
ac27a0ec
DK		 1031
1032 if (sbi->s_qf_names[USRQUOTA])
1033 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1034
1035 if (sbi->s_qf_names[GRPQUOTA])
1036 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1037
482a7425 1038 if (test_opt(sb, USRQUOTA))
ac27a0ec
DK		 1039 seq_puts(seq, ",usrquota");
1040
482a7425 1041 if (test_opt(sb, GRPQUOTA))
ac27a0ec
DK		 1042 seq_puts(seq, ",grpquota");
1043#endif
1044}
1045
d9c9bef1
MS		 1046/*
1047 * Show an option if
1048 * - it's set to a non-default value OR
1049 * - if the per-sb default is different from the global default
1050 */
34c80b1d 1051static int ext4_show_options(struct seq_file *seq, struct dentry *root)
ac27a0ec 1052{
aa22df2c
AK		 1053 int def_errors;
1054 unsigned long def_mount_opts;
34c80b1d 1055 struct super_block *sb = root->d_sb;
d9c9bef1
MS		 1056 struct ext4_sb_info *sbi = EXT4_SB(sb);
1057 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 1058
1059 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 1060 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 1061
1062 if (sbi->s_sb_block != 1)
1063 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1064 if (test_opt(sb, MINIX_DF))
1065 seq_puts(seq, ",minixdf");
aa22df2c 1066 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 1067 seq_puts(seq, ",grpid");
1068 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1069 seq_puts(seq, ",nogrpid");
1070 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1071 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1072 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1073 }
1074 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1075 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1076 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1077 }
bb4f397a 1078 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 1079 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 1080 def_errors == EXT4_ERRORS_CONTINUE) {
1081 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 1082 }
1083 }
aa22df2c 1084 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 1085 seq_puts(seq, ",errors=continue");
aa22df2c 1086 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 1087 seq_puts(seq, ",errors=panic");
aa22df2c 1088 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 1089 seq_puts(seq, ",nouid32");
aa22df2c 1090 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1 1091 seq_puts(seq, ",debug");
03010a33 1092#ifdef CONFIG_EXT4_FS_XATTR
ea663336 1093 if (test_opt(sb, XATTR_USER))
d9c9bef1 1094 seq_puts(seq, ",user_xattr");
ea663336 1095 if (!test_opt(sb, XATTR_USER))
d9c9bef1 1096 seq_puts(seq, ",nouser_xattr");
d9c9bef1 1097#endif
03010a33 1098#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 1099 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 1100 seq_puts(seq, ",acl");
1101 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1102 seq_puts(seq, ",noacl");
1103#endif
30773840 1104 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 1105 seq_printf(seq, ",commit=%u",
1106 (unsigned) (sbi->s_commit_interval / HZ));
1107 }
30773840
TT		 1108 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1109 seq_printf(seq, ",min_batch_time=%u",
1110 (unsigned) sbi->s_min_batch_time);
1111 }
1112 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1113 seq_printf(seq, ",max_batch_time=%u",
1d526fc9 1114 (unsigned) sbi->s_max_batch_time);
30773840
TT		 1115 }
1116
571640ca
ES		 1117 /*
1118 * We're changing the default of barrier mount option, so
1119 * let's always display its mount state so it's clear what its
1120 * status is.
1121 */
1122 seq_puts(seq, ",barrier=");
1123 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 1124 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1125 seq_puts(seq, ",journal_async_commit");
39a4bade
JK		 1126 else if (test_opt(sb, JOURNAL_CHECKSUM))
1127 seq_puts(seq, ",journal_checksum");
25ec56b5
JNC		 1128 if (test_opt(sb, I_VERSION))
1129 seq_puts(seq, ",i_version");
8b67f04a
TT		 1130 if (!test_opt(sb, DELALLOC) &&
1131 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
dd919b98
AK		 1132 seq_puts(seq, ",nodelalloc");
1133
6fd7a467
TT		 1134 if (!test_opt(sb, MBLK_IO_SUBMIT))
1135 seq_puts(seq, ",nomblk_io_submit");
cb45bbe4
MS		 1136 if (sbi->s_stripe)
1137 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 1138 /*
1139 * journal mode get enabled in different ways
1140 * So just print the value even if we didn't specify it
1141 */
617ba13b 1142 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 1143 seq_puts(seq, ",data=journal");
617ba13b 1144 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 1145 seq_puts(seq, ",data=ordered");
617ba13b 1146 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 1147 seq_puts(seq, ",data=writeback");
1148
240799cd
TT		 1149 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1150 seq_printf(seq, ",inode_readahead_blks=%u",
1151 sbi->s_inode_readahead_blks);
1152
5bf5683a
HK		 1153 if (test_opt(sb, DATA_ERR_ABORT))
1154 seq_puts(seq, ",data_err=abort");
1155
afd4672d 1156 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 1157 seq_puts(seq, ",noauto_da_alloc");
afd4672d 1158
8b67f04a 1159 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
5328e635
ES		 1160 seq_puts(seq, ",discard");
1161
e3bb52ae
ES		 1162 if (test_opt(sb, NOLOAD))
1163 seq_puts(seq, ",norecovery");
1164
744692dc
JZ		 1165 if (test_opt(sb, DIOREAD_NOLOCK))
1166 seq_puts(seq, ",dioread_nolock");
1167
8b67f04a
TT		 1168 if (test_opt(sb, BLOCK_VALIDITY) &&
1169 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1170 seq_puts(seq, ",block_validity");
1171
bfff6873 1172 if (!test_opt(sb, INIT_INODE_TABLE))
fc6cb1cd 1173 seq_puts(seq, ",noinit_itable");
51ce6511 1174 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
fc6cb1cd 1175 seq_printf(seq, ",init_itable=%u",
bfff6873
LC		 1176 (unsigned) sbi->s_li_wait_mult);
1177
617ba13b 1178 ext4_show_quota_options(seq, sb);
0b8e58a1 1179
ac27a0ec
DK		 1180 return 0;
1181}
1182
1b961ac0 1183static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1184 u64 ino, u32 generation)
ac27a0ec 1185{
ac27a0ec 1186 struct inode *inode;
ac27a0ec 1187
617ba13b 1188 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1189 return ERR_PTR(-ESTALE);
617ba13b 1190 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1191 return ERR_PTR(-ESTALE);
1192
1193 /* iget isn't really right if the inode is currently unallocated!!
1194 *
617ba13b 1195 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1196 * deleted, so we should be safe.
1197 *
1198 * Currently we don't know the generation for parent directory, so
1199 * a generation of 0 means "accept any"
1200 */
1d1fe1ee
DH		 1201 inode = ext4_iget(sb, ino);
1202 if (IS_ERR(inode))
1203 return ERR_CAST(inode);
1204 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1205 iput(inode);
1206 return ERR_PTR(-ESTALE);
1207 }
1b961ac0
CH		 1208
1209 return inode;
1210}
1211
1212static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1213 int fh_len, int fh_type)
1b961ac0
CH		 1214{
1215 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1216 ext4_nfs_get_inode);
1217}
1218
1219static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1220 int fh_len, int fh_type)
1b961ac0
CH		 1221{
1222 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1223 ext4_nfs_get_inode);
ac27a0ec
DK		 1224}
1225
c39a7f84
TO		 1226/*
1227 * Try to release metadata pages (indirect blocks, directories) which are
1228 * mapped via the block device. Since these pages could have journal heads
1229 * which would prevent try_to_free_buffers() from freeing them, we must use
1230 * jbd2 layer's try_to_free_buffers() function to release them.
1231 */
0b8e58a1
AD		 1232static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1233 gfp_t wait)
c39a7f84
TO		 1234{
1235 journal_t *journal = EXT4_SB(sb)->s_journal;
1236
1237 WARN_ON(PageChecked(page));
1238 if (!page_has_buffers(page))
1239 return 0;
1240 if (journal)
1241 return jbd2_journal_try_to_free_buffers(journal, page,
1242 wait & ~__GFP_WAIT);
1243 return try_to_free_buffers(page);
1244}
1245
ac27a0ec 1246#ifdef CONFIG_QUOTA
af5bc92d 1247#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1248#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1249
617ba13b
MC		 1250static int ext4_write_dquot(struct dquot *dquot);
1251static int ext4_acquire_dquot(struct dquot *dquot);
1252static int ext4_release_dquot(struct dquot *dquot);
1253static int ext4_mark_dquot_dirty(struct dquot *dquot);
1254static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1255static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1256 struct path *path);
ca0e05e4 1257static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1258static int ext4_quota_on_mount(struct super_block *sb, int type);
1259static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1260 size_t len, loff_t off);
617ba13b 1261static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1262 const char *data, size_t len, loff_t off);
1263
61e225dc 1264static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1265 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1266 .write_dquot = ext4_write_dquot,
1267 .acquire_dquot = ext4_acquire_dquot,
1268 .release_dquot = ext4_release_dquot,
1269 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1270 .write_info = ext4_write_info,
1271 .alloc_dquot = dquot_alloc,
1272 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1273};
1274
0d54b217 1275static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1276 .quota_on = ext4_quota_on,
ca0e05e4 1277 .quota_off = ext4_quota_off,
287a8095
CH		 1278 .quota_sync = dquot_quota_sync,
1279 .get_info = dquot_get_dqinfo,
1280 .set_info = dquot_set_dqinfo,
1281 .get_dqblk = dquot_get_dqblk,
1282 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1283};
1284#endif
1285
ee9b6d61 1286static const struct super_operations ext4_sops = {
617ba13b
MC		 1287 .alloc_inode = ext4_alloc_inode,
1288 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1289 .write_inode = ext4_write_inode,
1290 .dirty_inode = ext4_dirty_inode,
7ff9c073 1291 .drop_inode = ext4_drop_inode,
0930fcc1 1292 .evict_inode = ext4_evict_inode,
617ba13b 1293 .put_super = ext4_put_super,
617ba13b 1294 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1295 .freeze_fs = ext4_freeze,
1296 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1297 .statfs = ext4_statfs,
1298 .remount_fs = ext4_remount,
617ba13b 1299 .show_options = ext4_show_options,
ac27a0ec 1300#ifdef CONFIG_QUOTA
617ba13b
MC		 1301 .quota_read = ext4_quota_read,
1302 .quota_write = ext4_quota_write,
ac27a0ec 1303#endif
c39a7f84 1304 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1305};
1306
9ca92389
TT		 1307static const struct super_operations ext4_nojournal_sops = {
1308 .alloc_inode = ext4_alloc_inode,
1309 .destroy_inode = ext4_destroy_inode,
1310 .write_inode = ext4_write_inode,
1311 .dirty_inode = ext4_dirty_inode,
7ff9c073 1312 .drop_inode = ext4_drop_inode,
0930fcc1 1313 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1314 .write_super = ext4_write_super,
1315 .put_super = ext4_put_super,
1316 .statfs = ext4_statfs,
1317 .remount_fs = ext4_remount,
9ca92389
TT		 1318 .show_options = ext4_show_options,
1319#ifdef CONFIG_QUOTA
1320 .quota_read = ext4_quota_read,
1321 .quota_write = ext4_quota_write,
1322#endif
1323 .bdev_try_to_free_page = bdev_try_to_free_page,
1324};
1325
39655164 1326static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1327 .fh_to_dentry = ext4_fh_to_dentry,
1328 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1329 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1330};
1331
1332enum {
1333 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1334 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1335 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1336 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1337 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1338 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1339 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1340 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1341 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1342 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1343 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1344 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1345 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1346 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1347 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1348 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1349 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1350 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
ac27a0ec
DK		 1351};
1352
a447c093 1353static const match_table_t tokens = {
ac27a0ec
DK		 1354 {Opt_bsd_df, "bsddf"},
1355 {Opt_minix_df, "minixdf"},
1356 {Opt_grpid, "grpid"},
1357 {Opt_grpid, "bsdgroups"},
1358 {Opt_nogrpid, "nogrpid"},
1359 {Opt_nogrpid, "sysvgroups"},
1360 {Opt_resgid, "resgid=%u"},
1361 {Opt_resuid, "resuid=%u"},
1362 {Opt_sb, "sb=%u"},
1363 {Opt_err_cont, "errors=continue"},
1364 {Opt_err_panic, "errors=panic"},
1365 {Opt_err_ro, "errors=remount-ro"},
1366 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1367 {Opt_debug, "debug"},
1368 {Opt_oldalloc, "oldalloc"},
1369 {Opt_orlov, "orlov"},
1370 {Opt_user_xattr, "user_xattr"},
1371 {Opt_nouser_xattr, "nouser_xattr"},
1372 {Opt_acl, "acl"},
1373 {Opt_noacl, "noacl"},
ac27a0ec 1374 {Opt_noload, "noload"},
e3bb52ae 1375 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1376 {Opt_nobh, "nobh"},
1377 {Opt_bh, "bh"},
1378 {Opt_commit, "commit=%u"},
30773840
TT		 1379 {Opt_min_batch_time, "min_batch_time=%u"},
1380 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1381 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1382 {Opt_journal_checksum, "journal_checksum"},
1383 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1384 {Opt_abort, "abort"},
1385 {Opt_data_journal, "data=journal"},
1386 {Opt_data_ordered, "data=ordered"},
1387 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1388 {Opt_data_err_abort, "data_err=abort"},
1389 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1390 {Opt_offusrjquota, "usrjquota="},
1391 {Opt_usrjquota, "usrjquota=%s"},
1392 {Opt_offgrpjquota, "grpjquota="},
1393 {Opt_grpjquota, "grpjquota=%s"},
1394 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1395 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1396 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1397 {Opt_grpquota, "grpquota"},
1398 {Opt_noquota, "noquota"},
1399 {Opt_quota, "quota"},
1400 {Opt_usrquota, "usrquota"},
1401 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1402 {Opt_barrier, "barrier"},
1403 {Opt_nobarrier, "nobarrier"},
25ec56b5 1404 {Opt_i_version, "i_version"},
c9de560d 1405 {Opt_stripe, "stripe=%u"},
64769240 1406 {Opt_delalloc, "delalloc"},
dd919b98 1407 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1408 {Opt_mblk_io_submit, "mblk_io_submit"},
1409 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1410 {Opt_block_validity, "block_validity"},
1411 {Opt_noblock_validity, "noblock_validity"},
240799cd 1412 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1413 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1414 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1415 {Opt_auto_da_alloc, "auto_da_alloc"},
1416 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1417 {Opt_dioread_nolock, "dioread_nolock"},
1418 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1419 {Opt_discard, "discard"},
1420 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1421 {Opt_init_itable, "init_itable=%u"},
1422 {Opt_init_itable, "init_itable"},
1423 {Opt_noinit_itable, "noinit_itable"},
f3f12faa 1424 {Opt_err, NULL},
ac27a0ec
DK		 1425};
1426
617ba13b 1427static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1428{
617ba13b 1429 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1430 char *options = (char *) *data;
1431
1432 if (!options || strncmp(options, "sb=", 3) != 0)
1433 return 1; /* Default location */
0b8e58a1 1434
ac27a0ec 1435 options += 3;
0b8e58a1 1436 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1437 sb_block = simple_strtoul(options, &options, 0);
1438 if (*options && *options != ',') {
4776004f 1439 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1440 (char *) *data);
1441 return 1;
1442 }
1443 if (*options == ',')
1444 options++;
1445 *data = (void *) options;
0b8e58a1 1446
ac27a0ec
DK		 1447 return sb_block;
1448}
1449
b3881f74 1450#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1451static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1452 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1453
56c50f11
DM		 1454#ifdef CONFIG_QUOTA
1455static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1456{
1457 struct ext4_sb_info *sbi = EXT4_SB(sb);
1458 char *qname;
1459
1460 if (sb_any_quota_loaded(sb) &&
1461 !sbi->s_qf_names[qtype]) {
1462 ext4_msg(sb, KERN_ERR,
1463 "Cannot change journaled "
1464 "quota options when quota turned on");
1465 return 0;
1466 }
1467 qname = match_strdup(args);
1468 if (!qname) {
1469 ext4_msg(sb, KERN_ERR,
1470 "Not enough memory for storing quotafile name");
1471 return 0;
1472 }
1473 if (sbi->s_qf_names[qtype] &&
1474 strcmp(sbi->s_qf_names[qtype], qname)) {
1475 ext4_msg(sb, KERN_ERR,
1476 "%s quota file already specified", QTYPE2NAME(qtype));
1477 kfree(qname);
1478 return 0;
1479 }
1480 sbi->s_qf_names[qtype] = qname;
1481 if (strchr(sbi->s_qf_names[qtype], '/')) {
1482 ext4_msg(sb, KERN_ERR,
1483 "quotafile must be on filesystem root");
1484 kfree(sbi->s_qf_names[qtype]);
1485 sbi->s_qf_names[qtype] = NULL;
1486 return 0;
1487 }
fd8c37ec 1488 set_opt(sb, QUOTA);
56c50f11
DM		 1489 return 1;
1490}
1491
1492static int clear_qf_name(struct super_block *sb, int qtype)
1493{
1494
1495 struct ext4_sb_info *sbi = EXT4_SB(sb);
1496
1497 if (sb_any_quota_loaded(sb) &&
1498 sbi->s_qf_names[qtype]) {
1499 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1500 " when quota turned on");
1501 return 0;
1502 }
1503 /*
1504 * The space will be released later when all options are confirmed
1505 * to be correct
1506 */
1507 sbi->s_qf_names[qtype] = NULL;
1508 return 1;
1509}
1510#endif
1511
2b2d6d01 1512static int parse_options(char *options, struct super_block *sb,
c3191067 1513 unsigned long *journal_devnum,
b3881f74 1514 unsigned int *journal_ioprio,
661aa520 1515 int is_remount)
ac27a0ec 1516{
617ba13b 1517 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1518 char *p;
ac27a0ec
DK		 1519 substring_t args[MAX_OPT_ARGS];
1520 int data_opt = 0;
1521 int option;
1522#ifdef CONFIG_QUOTA
56c50f11 1523 int qfmt;
ac27a0ec
DK		 1524#endif
1525
1526 if (!options)
1527 return 1;
1528
2b2d6d01 1529 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1530 int token;
1531 if (!*p)
1532 continue;
1533
15121c18
ES		 1534 /*
1535 * Initialize args struct so we know whether arg was
1536 * found; some options take optional arguments.
1537 */
7dc57615 1538 args[0].to = args[0].from = NULL;
ac27a0ec
DK		 1539 token = match_token(p, tokens, args);
1540 switch (token) {
1541 case Opt_bsd_df:
437ca0fd 1542 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1543 clear_opt(sb, MINIX_DF);
ac27a0ec
DK		 1544 break;
1545 case Opt_minix_df:
437ca0fd 1546 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1547 set_opt(sb, MINIX_DF);
437ca0fd 1548
ac27a0ec
DK		 1549 break;
1550 case Opt_grpid:
437ca0fd 1551 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1552 set_opt(sb, GRPID);
437ca0fd 1553
ac27a0ec
DK		 1554 break;
1555 case Opt_nogrpid:
437ca0fd 1556 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1557 clear_opt(sb, GRPID);
437ca0fd 1558
ac27a0ec
DK		 1559 break;
1560 case Opt_resuid:
1561 if (match_int(&args[0], &option))
1562 return 0;
1563 sbi->s_resuid = option;
1564 break;
1565 case Opt_resgid:
1566 if (match_int(&args[0], &option))
1567 return 0;
1568 sbi->s_resgid = option;
1569 break;
1570 case Opt_sb:
1571 /* handled by get_sb_block() instead of here */
1572 /* *sb_block = match_int(&args[0]); */
1573 break;
1574 case Opt_err_panic:
fd8c37ec
TT		 1575 clear_opt(sb, ERRORS_CONT);
1576 clear_opt(sb, ERRORS_RO);
1577 set_opt(sb, ERRORS_PANIC);
ac27a0ec
DK		 1578 break;
1579 case Opt_err_ro:
fd8c37ec
TT		 1580 clear_opt(sb, ERRORS_CONT);
1581 clear_opt(sb, ERRORS_PANIC);
1582 set_opt(sb, ERRORS_RO);
ac27a0ec
DK		 1583 break;
1584 case Opt_err_cont:
fd8c37ec
TT		 1585 clear_opt(sb, ERRORS_RO);
1586 clear_opt(sb, ERRORS_PANIC);
1587 set_opt(sb, ERRORS_CONT);
ac27a0ec
DK		 1588 break;
1589 case Opt_nouid32:
fd8c37ec 1590 set_opt(sb, NO_UID32);
ac27a0ec 1591 break;
ac27a0ec 1592 case Opt_debug:
fd8c37ec 1593 set_opt(sb, DEBUG);
ac27a0ec
DK		 1594 break;
1595 case Opt_oldalloc:
4113c4ca
LC		 1596 ext4_msg(sb, KERN_WARNING,
1597 "Ignoring deprecated oldalloc option");
ac27a0ec
DK		 1598 break;
1599 case Opt_orlov:
4113c4ca
LC		 1600 ext4_msg(sb, KERN_WARNING,
1601 "Ignoring deprecated orlov option");
ac27a0ec 1602 break;
03010a33 1603#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1604 case Opt_user_xattr:
fd8c37ec 1605 set_opt(sb, XATTR_USER);
ac27a0ec
DK		 1606 break;
1607 case Opt_nouser_xattr:
fd8c37ec 1608 clear_opt(sb, XATTR_USER);
ac27a0ec
DK		 1609 break;
1610#else
1611 case Opt_user_xattr:
1612 case Opt_nouser_xattr:
b31e1552 1613 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1614 break;
1615#endif
03010a33 1616#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec 1617 case Opt_acl:
fd8c37ec 1618 set_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1619 break;
1620 case Opt_noacl:
fd8c37ec 1621 clear_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1622 break;
1623#else
1624 case Opt_acl:
1625 case Opt_noacl:
b31e1552 1626 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1627 break;
1628#endif
ac27a0ec
DK		 1629 case Opt_journal_dev:
1630 if (is_remount) {
b31e1552
ES		 1631 ext4_msg(sb, KERN_ERR,
1632 "Cannot specify journal on remount");
ac27a0ec
DK		 1633 return 0;
1634 }
1635 if (match_int(&args[0], &option))
1636 return 0;
1637 *journal_devnum = option;
1638 break;
818d276c 1639 case Opt_journal_checksum:
fd8c37ec 1640 set_opt(sb, JOURNAL_CHECKSUM);
d4da6c9c 1641 break;
818d276c 1642 case Opt_journal_async_commit:
fd8c37ec
TT		 1643 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1644 set_opt(sb, JOURNAL_CHECKSUM);
818d276c 1645 break;
ac27a0ec 1646 case Opt_noload:
fd8c37ec 1647 set_opt(sb, NOLOAD);
ac27a0ec
DK		 1648 break;
1649 case Opt_commit:
1650 if (match_int(&args[0], &option))
1651 return 0;
1652 if (option < 0)
1653 return 0;
1654 if (option == 0)
cd02ff0b 1655 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1656 sbi->s_commit_interval = HZ * option;
1657 break;
30773840
TT		 1658 case Opt_max_batch_time:
1659 if (match_int(&args[0], &option))
1660 return 0;
1661 if (option < 0)
1662 return 0;
1663 if (option == 0)
1664 option = EXT4_DEF_MAX_BATCH_TIME;
1665 sbi->s_max_batch_time = option;
1666 break;
1667 case Opt_min_batch_time:
1668 if (match_int(&args[0], &option))
1669 return 0;
1670 if (option < 0)
1671 return 0;
1672 sbi->s_min_batch_time = option;
1673 break;
ac27a0ec 1674 case Opt_data_journal:
617ba13b 1675 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1676 goto datacheck;
1677 case Opt_data_ordered:
617ba13b 1678 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1679 goto datacheck;
1680 case Opt_data_writeback:
617ba13b 1681 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1682 datacheck:
1683 if (is_remount) {
eb513689
TT		 1684 if (!sbi->s_journal)
1685 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1686 else if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1687 ext4_msg(sb, KERN_ERR,
1688 "Cannot change data mode on remount");
ac27a0ec
DK		 1689 return 0;
1690 }
1691 } else {
fd8c37ec 1692 clear_opt(sb, DATA_FLAGS);
ac27a0ec
DK		 1693 sbi->s_mount_opt |= data_opt;
1694 }
1695 break;
5bf5683a 1696 case Opt_data_err_abort:
fd8c37ec 1697 set_opt(sb, DATA_ERR_ABORT);
5bf5683a
HK		 1698 break;
1699 case Opt_data_err_ignore:
fd8c37ec 1700 clear_opt(sb, DATA_ERR_ABORT);
5bf5683a 1701 break;
ac27a0ec
DK		 1702#ifdef CONFIG_QUOTA
1703 case Opt_usrjquota:
56c50f11 1704 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1705 return 0;
56c50f11
DM		 1706 break;
1707 case Opt_grpjquota:
1708 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1709 return 0;
ac27a0ec
DK		 1710 break;
1711 case Opt_offusrjquota:
56c50f11
DM		 1712 if (!clear_qf_name(sb, USRQUOTA))
1713 return 0;
1714 break;
ac27a0ec 1715 case Opt_offgrpjquota:
56c50f11 1716 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1717 return 0;
ac27a0ec 1718 break;
56c50f11 1719
ac27a0ec 1720 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1721 qfmt = QFMT_VFS_OLD;
1722 goto set_qf_format;
ac27a0ec 1723 case Opt_jqfmt_vfsv0:
dfc5d03f 1724 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1725 goto set_qf_format;
1726 case Opt_jqfmt_vfsv1:
1727 qfmt = QFMT_VFS_V1;
dfc5d03f 1728set_qf_format:
17bd13b3 1729 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1730 sbi->s_jquota_fmt != qfmt) {
b31e1552 1731 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1732 "journaled quota options when "
b31e1552 1733 "quota turned on");
dfc5d03f
JK		 1734 return 0;
1735 }
1736 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1737 break;
1738 case Opt_quota:
1739 case Opt_usrquota:
fd8c37ec
TT		 1740 set_opt(sb, QUOTA);
1741 set_opt(sb, USRQUOTA);
ac27a0ec
DK		 1742 break;
1743 case Opt_grpquota:
fd8c37ec
TT		 1744 set_opt(sb, QUOTA);
1745 set_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1746 break;
1747 case Opt_noquota:
17bd13b3 1748 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1749 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1750 "options when quota turned on");
ac27a0ec
DK		 1751 return 0;
1752 }
fd8c37ec
TT		 1753 clear_opt(sb, QUOTA);
1754 clear_opt(sb, USRQUOTA);
1755 clear_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1756 break;
1757#else
1758 case Opt_quota:
1759 case Opt_usrquota:
1760 case Opt_grpquota:
b31e1552
ES		 1761 ext4_msg(sb, KERN_ERR,
1762 "quota options not supported");
cd59e7b9 1763 break;
ac27a0ec
DK		 1764 case Opt_usrjquota:
1765 case Opt_grpjquota:
1766 case Opt_offusrjquota:
1767 case Opt_offgrpjquota:
1768 case Opt_jqfmt_vfsold:
1769 case Opt_jqfmt_vfsv0:
5a20bdfc 1770 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1771 ext4_msg(sb, KERN_ERR,
1772 "journaled quota options not supported");
ac27a0ec
DK		 1773 break;
1774 case Opt_noquota:
1775 break;
1776#endif
1777 case Opt_abort:
4ab2f15b 1778 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1779 break;
06705bff 1780 case Opt_nobarrier:
fd8c37ec 1781 clear_opt(sb, BARRIER);
06705bff 1782 break;
ac27a0ec 1783 case Opt_barrier:
15121c18
ES		 1784 if (args[0].from) {
1785 if (match_int(&args[0], &option))
1786 return 0;
1787 } else
1788 option = 1; /* No argument, default to 1 */
ac27a0ec 1789 if (option)
fd8c37ec 1790 set_opt(sb, BARRIER);
ac27a0ec 1791 else
fd8c37ec 1792 clear_opt(sb, BARRIER);
ac27a0ec
DK		 1793 break;
1794 case Opt_ignore:
1795 break;
ac27a0ec 1796 case Opt_nobh:
206f7ab4
CH		 1797 ext4_msg(sb, KERN_WARNING,
1798 "Ignoring deprecated nobh option");
ac27a0ec
DK		 1799 break;
1800 case Opt_bh:
206f7ab4
CH		 1801 ext4_msg(sb, KERN_WARNING,
1802 "Ignoring deprecated bh option");
ac27a0ec 1803 break;
25ec56b5 1804 case Opt_i_version:
fd8c37ec 1805 set_opt(sb, I_VERSION);
25ec56b5
JNC		 1806 sb->s_flags |= MS_I_VERSION;
1807 break;
dd919b98 1808 case Opt_nodelalloc:
fd8c37ec 1809 clear_opt(sb, DELALLOC);
56889787 1810 clear_opt2(sb, EXPLICIT_DELALLOC);
dd919b98 1811 break;
1449032b 1812 case Opt_mblk_io_submit:
fd8c37ec 1813 set_opt(sb, MBLK_IO_SUBMIT);
1449032b
TT		 1814 break;
1815 case Opt_nomblk_io_submit:
fd8c37ec 1816 clear_opt(sb, MBLK_IO_SUBMIT);
1449032b 1817 break;
c9de560d
AT		 1818 case Opt_stripe:
1819 if (match_int(&args[0], &option))
1820 return 0;
1821 if (option < 0)
1822 return 0;
1823 sbi->s_stripe = option;
1824 break;
64769240 1825 case Opt_delalloc:
fd8c37ec 1826 set_opt(sb, DELALLOC);
56889787 1827 set_opt2(sb, EXPLICIT_DELALLOC);
64769240 1828 break;
6fd058f7 1829 case Opt_block_validity:
fd8c37ec 1830 set_opt(sb, BLOCK_VALIDITY);
6fd058f7
TT		 1831 break;
1832 case Opt_noblock_validity:
fd8c37ec 1833 clear_opt(sb, BLOCK_VALIDITY);
6fd058f7 1834 break;
240799cd
TT		 1835 case Opt_inode_readahead_blks:
1836 if (match_int(&args[0], &option))
1837 return 0;
1838 if (option < 0 || option > (1 << 30))
1839 return 0;
5dbd571d 1840 if (option && !is_power_of_2(option)) {
b31e1552
ES		 1841 ext4_msg(sb, KERN_ERR,
1842 "EXT4-fs: inode_readahead_blks"
1843 " must be a power of 2");
3197ebdb
TT		 1844 return 0;
1845 }
240799cd
TT		 1846 sbi->s_inode_readahead_blks = option;
1847 break;
b3881f74
TT		 1848 case Opt_journal_ioprio:
1849 if (match_int(&args[0], &option))
1850 return 0;
1851 if (option < 0 || option > 7)
1852 break;
1853 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1854 option);
1855 break;
06705bff 1856 case Opt_noauto_da_alloc:
fd8c37ec 1857 set_opt(sb, NO_AUTO_DA_ALLOC);
06705bff 1858 break;
afd4672d 1859 case Opt_auto_da_alloc:
15121c18
ES		 1860 if (args[0].from) {
1861 if (match_int(&args[0], &option))
1862 return 0;
1863 } else
1864 option = 1; /* No argument, default to 1 */
afd4672d 1865 if (option)
fd8c37ec 1866 clear_opt(sb, NO_AUTO_DA_ALLOC);
afd4672d 1867 else
fd8c37ec 1868 set_opt(sb,NO_AUTO_DA_ALLOC);
afd4672d 1869 break;
5328e635 1870 case Opt_discard:
fd8c37ec 1871 set_opt(sb, DISCARD);
5328e635
ES		 1872 break;
1873 case Opt_nodiscard:
fd8c37ec 1874 clear_opt(sb, DISCARD);
5328e635 1875 break;
744692dc 1876 case Opt_dioread_nolock:
fd8c37ec 1877 set_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 1878 break;
1879 case Opt_dioread_lock:
fd8c37ec 1880 clear_opt(sb, DIOREAD_NOLOCK);
744692dc 1881 break;
fc6cb1cd 1882 case Opt_init_itable:
fd8c37ec 1883 set_opt(sb, INIT_INODE_TABLE);
bfff6873
LC		 1884 if (args[0].from) {
1885 if (match_int(&args[0], &option))
1886 return 0;
1887 } else
1888 option = EXT4_DEF_LI_WAIT_MULT;
1889 if (option < 0)
1890 return 0;
1891 sbi->s_li_wait_mult = option;
1892 break;
fc6cb1cd 1893 case Opt_noinit_itable:
fd8c37ec 1894 clear_opt(sb, INIT_INODE_TABLE);
bfff6873 1895 break;
ac27a0ec 1896 default:
b31e1552
ES		 1897 ext4_msg(sb, KERN_ERR,
1898 "Unrecognized mount option \"%s\" "
1899 "or missing value", p);
ac27a0ec
DK		 1900 return 0;
1901 }
1902 }
1903#ifdef CONFIG_QUOTA
1904 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1905 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1906 clear_opt(sb, USRQUOTA);
ac27a0ec 1907
482a7425 1908 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1909 clear_opt(sb, GRPQUOTA);
ac27a0ec 1910
56c50f11 1911 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1912 ext4_msg(sb, KERN_ERR, "old and new quota "
1913 "format mixing");
ac27a0ec
DK		 1914 return 0;
1915 }
1916
1917 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1918 ext4_msg(sb, KERN_ERR, "journaled quota format "
1919 "not specified");
ac27a0ec
DK		 1920 return 0;
1921 }
1922 } else {
1923 if (sbi->s_jquota_fmt) {
b31e1552 1924 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1925 "specified with no journaling "
b31e1552 1926 "enabled");
ac27a0ec
DK		 1927 return 0;
1928 }
1929 }
1930#endif
1931 return 1;
1932}
1933
617ba13b 1934static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1935 int read_only)
1936{
617ba13b 1937 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1938 int res = 0;
1939
617ba13b 1940 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1941 ext4_msg(sb, KERN_ERR, "revision level too high, "
1942 "forcing read-only mode");
ac27a0ec
DK		 1943 res = MS_RDONLY;
1944 }
1945 if (read_only)
281b5995 1946 goto done;
617ba13b 1947 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1948 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1949 "running e2fsck is recommended");
617ba13b 1950 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1951 ext4_msg(sb, KERN_WARNING,
1952 "warning: mounting fs with errors, "
1953 "running e2fsck is recommended");
ed3ce80a 1954 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1955 le16_to_cpu(es->s_mnt_count) >=
1956 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1957 ext4_msg(sb, KERN_WARNING,
1958 "warning: maximal mount count reached, "
1959 "running e2fsck is recommended");
ac27a0ec
DK		 1960 else if (le32_to_cpu(es->s_checkinterval) &&
1961 (le32_to_cpu(es->s_lastcheck) +
1962 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1963 ext4_msg(sb, KERN_WARNING,
1964 "warning: checktime reached, "
1965 "running e2fsck is recommended");
0b8e58a1 1966 if (!sbi->s_journal)
0390131b 1967 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1968 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1969 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1970 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1971 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1972 ext4_update_dynamic_rev(sb);
0390131b
FM		 1973 if (sbi->s_journal)
1974 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1975
e2d67052 1976 ext4_commit_super(sb, 1);
281b5995 1977done:
ac27a0ec 1978 if (test_opt(sb, DEBUG))
a9df9a49 1979 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1980 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1981 sb->s_blocksize,
1982 sbi->s_groups_count,
617ba13b
MC		 1983 EXT4_BLOCKS_PER_GROUP(sb),
1984 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1985 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1986
7abc52c2 1987 cleancache_init_fs(sb);
ac27a0ec
DK		 1988 return res;
1989}
1990
772cb7c8
JS		 1991static int ext4_fill_flex_info(struct super_block *sb)
1992{
1993 struct ext4_sb_info *sbi = EXT4_SB(sb);
1994 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1995 ext4_group_t flex_group_count;
1996 ext4_group_t flex_group;
d50f2ab6 1997 unsigned int groups_per_flex = 0;
c5ca7c76 1998 size_t size;
772cb7c8
JS		 1999 int i;
2000
503358ae 2001 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2002 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2003 sbi->s_log_groups_per_flex = 0;
2004 return 1;
2005 }
d50f2ab6 2006 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
772cb7c8 2007
c62a11fd
FB		 2008 /* We allocate both existing and potentially added groups */
2009 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 2010 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2011 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 2012 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 2013 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 2014 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 2015 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2016 flex_group_count);
2017 goto failed;
772cb7c8 2018 }
772cb7c8 2019
772cb7c8 2020 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2021 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2022
2023 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2024 atomic_add(ext4_free_inodes_count(sb, gdp),
2025 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 2026 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 2027 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2028 atomic_add(ext4_used_dirs_count(sb, gdp),
2029 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2030 }
2031
2032 return 1;
2033failed:
2034 return 0;
2035}
2036
717d50e4
AD		 2037__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2038 struct ext4_group_desc *gdp)
2039{
2040 __u16 crc = 0;
2041
2042 if (sbi->s_es->s_feature_ro_compat &
2043 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2044 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2045 __le32 le_group = cpu_to_le32(block_group);
2046
2047 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2048 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2049 crc = crc16(crc, (__u8 *)gdp, offset);
2050 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2051 /* for checksum of struct ext4_group_desc do the rest...*/
2052 if ((sbi->s_es->s_feature_incompat &
2053 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2054 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2055 crc = crc16(crc, (__u8 *)gdp + offset,
2056 le16_to_cpu(sbi->s_es->s_desc_size) -
2057 offset);
2058 }
2059
2060 return cpu_to_le16(crc);
2061}
2062
2063int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2064 struct ext4_group_desc *gdp)
2065{
2066 if ((sbi->s_es->s_feature_ro_compat &
2067 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2068 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2069 return 0;
2070
2071 return 1;
2072}
2073
ac27a0ec 2074/* Called at mount-time, super-block is locked */
bfff6873
LC		 2075static int ext4_check_descriptors(struct super_block *sb,
2076 ext4_group_t *first_not_zeroed)
ac27a0ec 2077{
617ba13b
MC		 2078 struct ext4_sb_info *sbi = EXT4_SB(sb);
2079 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2080 ext4_fsblk_t last_block;
bd81d8ee
LV		 2081 ext4_fsblk_t block_bitmap;
2082 ext4_fsblk_t inode_bitmap;
2083 ext4_fsblk_t inode_table;
ce421581 2084 int flexbg_flag = 0;
bfff6873 2085 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2086
ce421581
JS		 2087 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2088 flexbg_flag = 1;
2089
af5bc92d 2090 ext4_debug("Checking group descriptors");
ac27a0ec 2091
197cd65a
AM		 2092 for (i = 0; i < sbi->s_groups_count; i++) {
2093 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2094
ce421581 2095 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2096 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2097 else
2098 last_block = first_block +
617ba13b 2099 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2100
bfff6873
LC		 2101 if ((grp == sbi->s_groups_count) &&
2102 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2103 grp = i;
2104
8fadc143 2105 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2106 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2107 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2108 "Block bitmap for group %u not in group "
b31e1552 2109 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2110 return 0;
2111 }
8fadc143 2112 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2113 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2114 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2115 "Inode bitmap for group %u not in group "
b31e1552 2116 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2117 return 0;
2118 }
8fadc143 2119 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2120 if (inode_table < first_block ||
2b2d6d01 2121 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2122 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2123 "Inode table for group %u not in group "
b31e1552 2124 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2125 return 0;
2126 }
955ce5f5 2127 ext4_lock_group(sb, i);
717d50e4 2128 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2129 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2130 "Checksum for group %u failed (%u!=%u)",
2131 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2132 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2133 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2134 ext4_unlock_group(sb, i);
8a266467 2135 return 0;
7ee1ec4c 2136 }
717d50e4 2137 }
955ce5f5 2138 ext4_unlock_group(sb, i);
ce421581
JS		 2139 if (!flexbg_flag)
2140 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2141 }
bfff6873
LC		 2142 if (NULL != first_not_zeroed)
2143 *first_not_zeroed = grp;
ac27a0ec 2144
5dee5437
TT		 2145 ext4_free_blocks_count_set(sbi->s_es,
2146 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2147 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2148 return 1;
2149}
2150
617ba13b 2151/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2152 * the superblock) which were deleted from all directories, but held open by
2153 * a process at the time of a crash. We walk the list and try to delete these
2154 * inodes at recovery time (only with a read-write filesystem).
2155 *
2156 * In order to keep the orphan inode chain consistent during traversal (in
2157 * case of crash during recovery), we link each inode into the superblock
2158 * orphan list_head and handle it the same way as an inode deletion during
2159 * normal operation (which journals the operations for us).
2160 *
2161 * We only do an iget() and an iput() on each inode, which is very safe if we
2162 * accidentally point at an in-use or already deleted inode. The worst that
2163 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2164 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2165 * e2fsck was run on this filesystem, and it must have already done the orphan
2166 * inode cleanup for us, so we can safely abort without any further action.
2167 */
2b2d6d01
TT		 2168static void ext4_orphan_cleanup(struct super_block *sb,
2169 struct ext4_super_block *es)
ac27a0ec
DK		 2170{
2171 unsigned int s_flags = sb->s_flags;
2172 int nr_orphans = 0, nr_truncates = 0;
2173#ifdef CONFIG_QUOTA
2174 int i;
2175#endif
2176 if (!es->s_last_orphan) {
2177 jbd_debug(4, "no orphan inodes to clean up\n");
2178 return;
2179 }
2180
a8f48a95 2181 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2182 ext4_msg(sb, KERN_ERR, "write access "
2183 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2184 return;
2185 }
2186
d39195c3
AG		 2187 /* Check if feature set would not allow a r/w mount */
2188 if (!ext4_feature_set_ok(sb, 0)) {
2189 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2190 "unknown ROCOMPAT features");
2191 return;
2192 }
2193
617ba13b 2194 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2195 if (es->s_last_orphan)
2196 jbd_debug(1, "Errors on filesystem, "
2197 "clearing orphan list.\n");
2198 es->s_last_orphan = 0;
2199 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2200 return;
2201 }
2202
2203 if (s_flags & MS_RDONLY) {
b31e1552 2204 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2205 sb->s_flags &= ~MS_RDONLY;
2206 }
2207#ifdef CONFIG_QUOTA
2208 /* Needed for iput() to work correctly and not trash data */
2209 sb->s_flags |= MS_ACTIVE;
2210 /* Turn on quotas so that they are updated correctly */
2211 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2212 if (EXT4_SB(sb)->s_qf_names[i]) {
2213 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2214 if (ret < 0)
b31e1552
ES		 2215 ext4_msg(sb, KERN_ERR,
2216 "Cannot turn on journaled "
2217 "quota: error %d", ret);
ac27a0ec
DK		 2218 }
2219 }
2220#endif
2221
2222 while (es->s_last_orphan) {
2223 struct inode *inode;
2224
97bd42b9
JB		 2225 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2226 if (IS_ERR(inode)) {
ac27a0ec
DK		 2227 es->s_last_orphan = 0;
2228 break;
2229 }
2230
617ba13b 2231 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2232 dquot_initialize(inode);
ac27a0ec 2233 if (inode->i_nlink) {
b31e1552
ES		 2234 ext4_msg(sb, KERN_DEBUG,
2235 "%s: truncating inode %lu to %lld bytes",
46e665e9 2236 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2237 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2238 inode->i_ino, inode->i_size);
617ba13b 2239 ext4_truncate(inode);
ac27a0ec
DK		 2240 nr_truncates++;
2241 } else {
b31e1552
ES		 2242 ext4_msg(sb, KERN_DEBUG,
2243 "%s: deleting unreferenced inode %lu",
46e665e9 2244 __func__, inode->i_ino);
ac27a0ec
DK		 2245 jbd_debug(2, "deleting unreferenced inode %lu\n",
2246 inode->i_ino);
2247 nr_orphans++;
2248 }
2249 iput(inode); /* The delete magic happens here! */
2250 }
2251
2b2d6d01 2252#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2253
2254 if (nr_orphans)
b31e1552
ES		 2255 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2256 PLURAL(nr_orphans));
ac27a0ec 2257 if (nr_truncates)
b31e1552
ES		 2258 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2259 PLURAL(nr_truncates));
ac27a0ec
DK		 2260#ifdef CONFIG_QUOTA
2261 /* Turn quotas off */
2262 for (i = 0; i < MAXQUOTAS; i++) {
2263 if (sb_dqopt(sb)->files[i])
287a8095 2264 dquot_quota_off(sb, i);
ac27a0ec
DK		 2265 }
2266#endif
2267 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2268}
0b8e58a1 2269
cd2291a4
ES		 2270/*
2271 * Maximal extent format file size.
2272 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2273 * extent format containers, within a sector_t, and within i_blocks
2274 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2275 * so that won't be a limiting factor.
2276 *
f17722f9
LC		 2277 * However there is other limiting factor. We do store extents in the form
2278 * of starting block and length, hence the resulting length of the extent
2279 * covering maximum file size must fit into on-disk format containers as
2280 * well. Given that length is always by 1 unit bigger than max unit (because
2281 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2282 *
cd2291a4
ES		 2283 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2284 */
f287a1a5 2285static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2286{
2287 loff_t res;
2288 loff_t upper_limit = MAX_LFS_FILESIZE;
2289
2290 /* small i_blocks in vfs inode? */
f287a1a5 2291 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2292 /*
90c699a9 2293 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2294 * i_block represent total blocks in 512 bytes
2295 * 32 == size of vfs inode i_blocks * 8
2296 */
2297 upper_limit = (1LL << 32) - 1;
2298
2299 /* total blocks in file system block size */
2300 upper_limit >>= (blkbits - 9);
2301 upper_limit <<= blkbits;
2302 }
2303
f17722f9
LC		 2304 /*
2305 * 32-bit extent-start container, ee_block. We lower the maxbytes
2306 * by one fs block, so ee_len can cover the extent of maximum file
2307 * size
2308 */
2309 res = (1LL << 32) - 1;
cd2291a4 2310 res <<= blkbits;
cd2291a4
ES		 2311
2312 /* Sanity check against vm- & vfs- imposed limits */
2313 if (res > upper_limit)
2314 res = upper_limit;
2315
2316 return res;
2317}
ac27a0ec 2318
ac27a0ec 2319/*
cd2291a4 2320 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2321 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2322 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2323 */
f287a1a5 2324static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2325{
617ba13b 2326 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2327 int meta_blocks;
2328 loff_t upper_limit;
0b8e58a1
AD		 2329 /* This is calculated to be the largest file size for a dense, block
2330 * mapped file such that the file's total number of 512-byte sectors,
2331 * including data and all indirect blocks, does not exceed (2^48 - 1).
2332 *
2333 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2334 * number of 512-byte sectors of the file.
0fc1b451
AK		 2335 */
2336
f287a1a5 2337 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2338 /*
90c699a9 2339 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2340 * the inode i_block field represents total file blocks in
2341 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2342 */
2343 upper_limit = (1LL << 32) - 1;
2344
2345 /* total blocks in file system block size */
2346 upper_limit >>= (bits - 9);
2347
2348 } else {
8180a562
AK		 2349 /*
2350 * We use 48 bit ext4_inode i_blocks
2351 * With EXT4_HUGE_FILE_FL set the i_blocks
2352 * represent total number of blocks in
2353 * file system block size
2354 */
0fc1b451
AK		 2355 upper_limit = (1LL << 48) - 1;
2356
0fc1b451
AK		 2357 }
2358
2359 /* indirect blocks */
2360 meta_blocks = 1;
2361 /* double indirect blocks */
2362 meta_blocks += 1 + (1LL << (bits-2));
2363 /* tripple indirect blocks */
2364 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2365
2366 upper_limit -= meta_blocks;
2367 upper_limit <<= bits;
ac27a0ec
DK		 2368
2369 res += 1LL << (bits-2);
2370 res += 1LL << (2*(bits-2));
2371 res += 1LL << (3*(bits-2));
2372 res <<= bits;
2373 if (res > upper_limit)
2374 res = upper_limit;
0fc1b451
AK		 2375
2376 if (res > MAX_LFS_FILESIZE)
2377 res = MAX_LFS_FILESIZE;
2378
ac27a0ec
DK		 2379 return res;
2380}
2381
617ba13b 2382static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2383 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2384{
617ba13b 2385 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2386 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2387 int has_super = 0;
2388
2389 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2390
617ba13b 2391 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2392 nr < first_meta_bg)
70bbb3e0 2393 return logical_sb_block + nr + 1;
ac27a0ec 2394 bg = sbi->s_desc_per_block * nr;
617ba13b 2395 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2396 has_super = 1;
0b8e58a1 2397
617ba13b 2398 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2399}
2400
c9de560d
AT		 2401/**
2402 * ext4_get_stripe_size: Get the stripe size.
2403 * @sbi: In memory super block info
2404 *
2405 * If we have specified it via mount option, then
2406 * use the mount option value. If the value specified at mount time is
2407 * greater than the blocks per group use the super block value.
2408 * If the super block value is greater than blocks per group return 0.
2409 * Allocator needs it be less than blocks per group.
2410 *
2411 */
2412static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2413{
2414 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2415 unsigned long stripe_width =
2416 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2417 int ret;
c9de560d
AT		 2418
2419 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2420 ret = sbi->s_stripe;
2421 else if (stripe_width <= sbi->s_blocks_per_group)
2422 ret = stripe_width;
2423 else if (stride <= sbi->s_blocks_per_group)
2424 ret = stride;
2425 else
2426 ret = 0;
c9de560d 2427
3eb08658
DE		 2428 /*
2429 * If the stripe width is 1, this makes no sense and
2430 * we set it to 0 to turn off stripe handling code.
2431 */
2432 if (ret <= 1)
2433 ret = 0;
c9de560d 2434
3eb08658 2435 return ret;
c9de560d 2436}
ac27a0ec 2437
3197ebdb
TT		 2438/* sysfs supprt */
2439
2440struct ext4_attr {
2441 struct attribute attr;
2442 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2443 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2444 const char *, size_t);
2445 int offset;
2446};
2447
2448static int parse_strtoul(const char *buf,
2449 unsigned long max, unsigned long *value)
2450{
2451 char *endp;
2452
e7d2860b
AGR		 2453 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2454 endp = skip_spaces(endp);
3197ebdb
TT		 2455 if (*endp || *value > max)
2456 return -EINVAL;
2457
2458 return 0;
2459}
2460
2461static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2462 struct ext4_sb_info *sbi,
2463 char *buf)
2464{
2465 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2466 (s64) EXT4_C2B(sbi,
2467 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2468}
2469
2470static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2471 struct ext4_sb_info *sbi, char *buf)
2472{
2473 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2474
f613dfcb
TT		 2475 if (!sb->s_bdev->bd_part)
2476 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2477 return snprintf(buf, PAGE_SIZE, "%lu\n",
2478 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2479 sbi->s_sectors_written_start) >> 1);
2480}
2481
2482static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2483 struct ext4_sb_info *sbi, char *buf)
2484{
2485 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2486
f613dfcb
TT		 2487 if (!sb->s_bdev->bd_part)
2488 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2489 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2490 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2491 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2492 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2493}
2494
77f4135f
VH		 2495static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2496 struct ext4_sb_info *sbi, char *buf)
2497{
2498 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2499}
2500
2501static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2502 struct ext4_sb_info *sbi, char *buf)
2503{
2504 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2505}
2506
3197ebdb
TT		 2507static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2508 struct ext4_sb_info *sbi,
2509 const char *buf, size_t count)
2510{
2511 unsigned long t;
2512
2513 if (parse_strtoul(buf, 0x40000000, &t))
2514 return -EINVAL;
2515
5dbd571d 2516 if (t && !is_power_of_2(t))
3197ebdb
TT		 2517 return -EINVAL;
2518
2519 sbi->s_inode_readahead_blks = t;
2520 return count;
2521}
2522
2523static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2524 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2525{
2526 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2527
2528 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2529}
2530
2531static ssize_t sbi_ui_store(struct ext4_attr *a,
2532 struct ext4_sb_info *sbi,
2533 const char *buf, size_t count)
2534{
2535 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2536 unsigned long t;
2537
2538 if (parse_strtoul(buf, 0xffffffff, &t))
2539 return -EINVAL;
2540 *ui = t;
2541 return count;
2542}
2543
2544#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2545static struct ext4_attr ext4_attr_##_name = { \
2546 .attr = {.name = __stringify(_name), .mode = _mode }, \
2547 .show = _show, \
2548 .store = _store, \
2549 .offset = offsetof(struct ext4_sb_info, _elname), \
2550}
2551#define EXT4_ATTR(name, mode, show, store) \
2552static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2553
857ac889 2554#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2555#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2556#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2557#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2558 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2559#define ATTR_LIST(name) &ext4_attr_##name.attr
2560
2561EXT4_RO_ATTR(delayed_allocation_blocks);
2562EXT4_RO_ATTR(session_write_kbytes);
2563EXT4_RO_ATTR(lifetime_write_kbytes);
77f4135f
VH		 2564EXT4_RO_ATTR(extent_cache_hits);
2565EXT4_RO_ATTR(extent_cache_misses);
3197ebdb
TT		 2566EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2567 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2568EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2569EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2570EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2571EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2572EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2573EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2574EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2575EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2576
2577static struct attribute *ext4_attrs[] = {
2578 ATTR_LIST(delayed_allocation_blocks),
2579 ATTR_LIST(session_write_kbytes),
2580 ATTR_LIST(lifetime_write_kbytes),
77f4135f
VH		 2581 ATTR_LIST(extent_cache_hits),
2582 ATTR_LIST(extent_cache_misses),
3197ebdb 2583 ATTR_LIST(inode_readahead_blks),
11013911 2584 ATTR_LIST(inode_goal),
3197ebdb
TT		 2585 ATTR_LIST(mb_stats),
2586 ATTR_LIST(mb_max_to_scan),
2587 ATTR_LIST(mb_min_to_scan),
2588 ATTR_LIST(mb_order2_req),
2589 ATTR_LIST(mb_stream_req),
2590 ATTR_LIST(mb_group_prealloc),
55138e0b 2591 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2592 NULL,
2593};
2594
857ac889
LC		 2595/* Features this copy of ext4 supports */
2596EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2597EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2598
2599static struct attribute *ext4_feat_attrs[] = {
2600 ATTR_LIST(lazy_itable_init),
27ee40df 2601 ATTR_LIST(batched_discard),
857ac889
LC		 2602 NULL,
2603};
2604
3197ebdb
TT		 2605static ssize_t ext4_attr_show(struct kobject *kobj,
2606 struct attribute *attr, char *buf)
2607{
2608 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2609 s_kobj);
2610 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2611
2612 return a->show ? a->show(a, sbi, buf) : 0;
2613}
2614
2615static ssize_t ext4_attr_store(struct kobject *kobj,
2616 struct attribute *attr,
2617 const char *buf, size_t len)
2618{
2619 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2620 s_kobj);
2621 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2622
2623 return a->store ? a->store(a, sbi, buf, len) : 0;
2624}
2625
2626static void ext4_sb_release(struct kobject *kobj)
2627{
2628 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2629 s_kobj);
2630 complete(&sbi->s_kobj_unregister);
2631}
2632
52cf25d0 2633static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2634 .show = ext4_attr_show,
2635 .store = ext4_attr_store,
2636};
2637
2638static struct kobj_type ext4_ktype = {
2639 .default_attrs = ext4_attrs,
2640 .sysfs_ops = &ext4_attr_ops,
2641 .release = ext4_sb_release,
2642};
2643
857ac889
LC		 2644static void ext4_feat_release(struct kobject *kobj)
2645{
2646 complete(&ext4_feat->f_kobj_unregister);
2647}
2648
2649static struct kobj_type ext4_feat_ktype = {
2650 .default_attrs = ext4_feat_attrs,
2651 .sysfs_ops = &ext4_attr_ops,
2652 .release = ext4_feat_release,
2653};
2654
a13fb1a4
ES		 2655/*
2656 * Check whether this filesystem can be mounted based on
2657 * the features present and the RDONLY/RDWR mount requested.
2658 * Returns 1 if this filesystem can be mounted as requested,
2659 * 0 if it cannot be.
2660 */
2661static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2662{
2663 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2664 ext4_msg(sb, KERN_ERR,
2665 "Couldn't mount because of "
2666 "unsupported optional features (%x)",
2667 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2668 ~EXT4_FEATURE_INCOMPAT_SUPP));
2669 return 0;
2670 }
2671
2672 if (readonly)
2673 return 1;
2674
2675 /* Check that feature set is OK for a read-write mount */
2676 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2677 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2678 "unsupported optional features (%x)",
2679 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2680 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2681 return 0;
2682 }
2683 /*
2684 * Large file size enabled file system can only be mounted
2685 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2686 */
2687 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2688 if (sizeof(blkcnt_t) < sizeof(u64)) {
2689 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2690 "cannot be mounted RDWR without "
2691 "CONFIG_LBDAF");
2692 return 0;
2693 }
2694 }
bab08ab9
TT		 2695 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2696 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2697 ext4_msg(sb, KERN_ERR,
2698 "Can't support bigalloc feature without "
2699 "extents feature\n");
2700 return 0;
2701 }
a13fb1a4
ES		 2702 return 1;
2703}
2704
66e61a9e
TT		 2705/*
2706 * This function is called once a day if we have errors logged
2707 * on the file system
2708 */
2709static void print_daily_error_info(unsigned long arg)
2710{
2711 struct super_block *sb = (struct super_block *) arg;
2712 struct ext4_sb_info *sbi;
2713 struct ext4_super_block *es;
2714
2715 sbi = EXT4_SB(sb);
2716 es = sbi->s_es;
2717
2718 if (es->s_error_count)
2719 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2720 le32_to_cpu(es->s_error_count));
2721 if (es->s_first_error_time) {
2722 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2723 sb->s_id, le32_to_cpu(es->s_first_error_time),
2724 (int) sizeof(es->s_first_error_func),
2725 es->s_first_error_func,
2726 le32_to_cpu(es->s_first_error_line));
2727 if (es->s_first_error_ino)
2728 printk(": inode %u",
2729 le32_to_cpu(es->s_first_error_ino));
2730 if (es->s_first_error_block)
2731 printk(": block %llu", (unsigned long long)
2732 le64_to_cpu(es->s_first_error_block));
2733 printk("\n");
2734 }
2735 if (es->s_last_error_time) {
2736 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2737 sb->s_id, le32_to_cpu(es->s_last_error_time),
2738 (int) sizeof(es->s_last_error_func),
2739 es->s_last_error_func,
2740 le32_to_cpu(es->s_last_error_line));
2741 if (es->s_last_error_ino)
2742 printk(": inode %u",
2743 le32_to_cpu(es->s_last_error_ino));
2744 if (es->s_last_error_block)
2745 printk(": block %llu", (unsigned long long)
2746 le64_to_cpu(es->s_last_error_block));
2747 printk("\n");
2748 }
2749 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2750}
2751
bfff6873
LC		 2752/* Find next suitable group and run ext4_init_inode_table */
2753static int ext4_run_li_request(struct ext4_li_request *elr)
2754{
2755 struct ext4_group_desc *gdp = NULL;
2756 ext4_group_t group, ngroups;
2757 struct super_block *sb;
2758 unsigned long timeout = 0;
2759 int ret = 0;
2760
2761 sb = elr->lr_super;
2762 ngroups = EXT4_SB(sb)->s_groups_count;
2763
2764 for (group = elr->lr_next_group; group < ngroups; group++) {
2765 gdp = ext4_get_group_desc(sb, group, NULL);
2766 if (!gdp) {
2767 ret = 1;
2768 break;
2769 }
2770
2771 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2772 break;
2773 }
2774
2775 if (group == ngroups)
2776 ret = 1;
2777
2778 if (!ret) {
2779 timeout = jiffies;
2780 ret = ext4_init_inode_table(sb, group,
2781 elr->lr_timeout ? 0 : 1);
2782 if (elr->lr_timeout == 0) {
51ce6511
LC		 2783 timeout = (jiffies - timeout) *
2784 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2785 elr->lr_timeout = timeout;
2786 }
2787 elr->lr_next_sched = jiffies + elr->lr_timeout;
2788 elr->lr_next_group = group + 1;
2789 }
2790
2791 return ret;
2792}
2793
2794/*
2795 * Remove lr_request from the list_request and free the
4ed5c033 2796 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2797 */
2798static void ext4_remove_li_request(struct ext4_li_request *elr)
2799{
2800 struct ext4_sb_info *sbi;
2801
2802 if (!elr)
2803 return;
2804
2805 sbi = elr->lr_sbi;
2806
2807 list_del(&elr->lr_request);
2808 sbi->s_li_request = NULL;
2809 kfree(elr);
2810}
2811
2812static void ext4_unregister_li_request(struct super_block *sb)
2813{
1bb933fb
LC		 2814 mutex_lock(&ext4_li_mtx);
2815 if (!ext4_li_info) {
2816 mutex_unlock(&ext4_li_mtx);
bfff6873 2817 return;
1bb933fb 2818 }
bfff6873
LC		 2819
2820 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2821 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2822 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2823 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2824}
2825
8f1f7453
ES		 2826static struct task_struct *ext4_lazyinit_task;
2827
bfff6873
LC		 2828/*
2829 * This is the function where ext4lazyinit thread lives. It walks
2830 * through the request list searching for next scheduled filesystem.
2831 * When such a fs is found, run the lazy initialization request
2832 * (ext4_rn_li_request) and keep track of the time spend in this
2833 * function. Based on that time we compute next schedule time of
2834 * the request. When walking through the list is complete, compute
2835 * next waking time and put itself into sleep.
2836 */
2837static int ext4_lazyinit_thread(void *arg)
2838{
2839 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2840 struct list_head *pos, *n;
2841 struct ext4_li_request *elr;
4ed5c033 2842 unsigned long next_wakeup, cur;
bfff6873
LC		 2843
2844 BUG_ON(NULL == eli);
2845
bfff6873
LC		 2846cont_thread:
2847 while (true) {
2848 next_wakeup = MAX_JIFFY_OFFSET;
2849
2850 mutex_lock(&eli->li_list_mtx);
2851 if (list_empty(&eli->li_request_list)) {
2852 mutex_unlock(&eli->li_list_mtx);
2853 goto exit_thread;
2854 }
2855
2856 list_for_each_safe(pos, n, &eli->li_request_list) {
2857 elr = list_entry(pos, struct ext4_li_request,
2858 lr_request);
2859
b2c78cd0
TT		 2860 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2861 if (ext4_run_li_request(elr) != 0) {
2862 /* error, remove the lazy_init job */
2863 ext4_remove_li_request(elr);
2864 continue;
2865 }
bfff6873
LC		 2866 }
2867
2868 if (time_before(elr->lr_next_sched, next_wakeup))
2869 next_wakeup = elr->lr_next_sched;
2870 }
2871 mutex_unlock(&eli->li_list_mtx);
2872
a0acae0e 2873 try_to_freeze();
bfff6873 2874
4ed5c033
LC		 2875 cur = jiffies;
2876 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2877 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2878 cond_resched();
2879 continue;
2880 }
2881
4ed5c033
LC		 2882 schedule_timeout_interruptible(next_wakeup - cur);
2883
8f1f7453
ES		 2884 if (kthread_should_stop()) {
2885 ext4_clear_request_list();
2886 goto exit_thread;
2887 }
bfff6873
LC		 2888 }
2889
2890exit_thread:
2891 /*
2892 * It looks like the request list is empty, but we need
2893 * to check it under the li_list_mtx lock, to prevent any
2894 * additions into it, and of course we should lock ext4_li_mtx
2895 * to atomically free the list and ext4_li_info, because at
2896 * this point another ext4 filesystem could be registering
2897 * new one.
2898 */
2899 mutex_lock(&ext4_li_mtx);
2900 mutex_lock(&eli->li_list_mtx);
2901 if (!list_empty(&eli->li_request_list)) {
2902 mutex_unlock(&eli->li_list_mtx);
2903 mutex_unlock(&ext4_li_mtx);
2904 goto cont_thread;
2905 }
2906 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2907 kfree(ext4_li_info);
2908 ext4_li_info = NULL;
2909 mutex_unlock(&ext4_li_mtx);
2910
2911 return 0;
2912}
2913
2914static void ext4_clear_request_list(void)
2915{
2916 struct list_head *pos, *n;
2917 struct ext4_li_request *elr;
2918
2919 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2920 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2921 elr = list_entry(pos, struct ext4_li_request,
2922 lr_request);
2923 ext4_remove_li_request(elr);
2924 }
2925 mutex_unlock(&ext4_li_info->li_list_mtx);
2926}
2927
2928static int ext4_run_lazyinit_thread(void)
2929{
8f1f7453
ES		 2930 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2931 ext4_li_info, "ext4lazyinit");
2932 if (IS_ERR(ext4_lazyinit_task)) {
2933 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2934 ext4_clear_request_list();
bfff6873
LC		 2935 kfree(ext4_li_info);
2936 ext4_li_info = NULL;
2937 printk(KERN_CRIT "EXT4: error %d creating inode table "
2938 "initialization thread\n",
2939 err);
2940 return err;
2941 }
2942 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2943 return 0;
2944}
2945
2946/*
2947 * Check whether it make sense to run itable init. thread or not.
2948 * If there is at least one uninitialized inode table, return
2949 * corresponding group number, else the loop goes through all
2950 * groups and return total number of groups.
2951 */
2952static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2953{
2954 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2955 struct ext4_group_desc *gdp = NULL;
2956
2957 for (group = 0; group < ngroups; group++) {
2958 gdp = ext4_get_group_desc(sb, group, NULL);
2959 if (!gdp)
2960 continue;
2961
2962 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2963 break;
2964 }
2965
2966 return group;
2967}
2968
2969static int ext4_li_info_new(void)
2970{
2971 struct ext4_lazy_init *eli = NULL;
2972
2973 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2974 if (!eli)
2975 return -ENOMEM;
2976
bfff6873
LC		 2977 INIT_LIST_HEAD(&eli->li_request_list);
2978 mutex_init(&eli->li_list_mtx);
2979
bfff6873
LC		 2980 eli->li_state |= EXT4_LAZYINIT_QUIT;
2981
2982 ext4_li_info = eli;
2983
2984 return 0;
2985}
2986
2987static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2988 ext4_group_t start)
2989{
2990 struct ext4_sb_info *sbi = EXT4_SB(sb);
2991 struct ext4_li_request *elr;
2992 unsigned long rnd;
2993
2994 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2995 if (!elr)
2996 return NULL;
2997
2998 elr->lr_super = sb;
2999 elr->lr_sbi = sbi;
3000 elr->lr_next_group = start;
3001
3002 /*
3003 * Randomize first schedule time of the request to
3004 * spread the inode table initialization requests
3005 * better.
3006 */
3007 get_random_bytes(&rnd, sizeof(rnd));
3008 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3009 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3010
3011 return elr;
3012}
3013
3014static int ext4_register_li_request(struct super_block *sb,
3015 ext4_group_t first_not_zeroed)
3016{
3017 struct ext4_sb_info *sbi = EXT4_SB(sb);
3018 struct ext4_li_request *elr;
3019 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3020 int ret = 0;
bfff6873 3021
51ce6511
LC		 3022 if (sbi->s_li_request != NULL) {
3023 /*
3024 * Reset timeout so it can be computed again, because
3025 * s_li_wait_mult might have changed.
3026 */
3027 sbi->s_li_request->lr_timeout = 0;
beed5ecb 3028 return 0;
51ce6511 3029 }
bfff6873
LC		 3030
3031 if (first_not_zeroed == ngroups ||
3032 (sb->s_flags & MS_RDONLY) ||
55ff3840 3033 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 3034 return 0;
bfff6873
LC		 3035
3036 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 3037 if (!elr)
3038 return -ENOMEM;
bfff6873
LC		 3039
3040 mutex_lock(&ext4_li_mtx);
3041
3042 if (NULL == ext4_li_info) {
3043 ret = ext4_li_info_new();
3044 if (ret)
3045 goto out;
3046 }
3047
3048 mutex_lock(&ext4_li_info->li_list_mtx);
3049 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3050 mutex_unlock(&ext4_li_info->li_list_mtx);
3051
3052 sbi->s_li_request = elr;
46e4690b
TM		 3053 /*
3054 * set elr to NULL here since it has been inserted to
3055 * the request_list and the removal and free of it is
3056 * handled by ext4_clear_request_list from now on.
3057 */
3058 elr = NULL;
bfff6873
LC		 3059
3060 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3061 ret = ext4_run_lazyinit_thread();
3062 if (ret)
3063 goto out;
3064 }
bfff6873 3065out:
beed5ecb
NK		 3066 mutex_unlock(&ext4_li_mtx);
3067 if (ret)
bfff6873 3068 kfree(elr);
bfff6873
LC		 3069 return ret;
3070}
3071
3072/*
3073 * We do not need to lock anything since this is called on
3074 * module unload.
3075 */
3076static void ext4_destroy_lazyinit_thread(void)
3077{
3078 /*
3079 * If thread exited earlier
3080 * there's nothing to be done.
3081 */
8f1f7453 3082 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3083 return;
3084
8f1f7453 3085 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3086}
3087
2b2d6d01 3088static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3089{
d4c402d9 3090 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3091 struct buffer_head *bh;
617ba13b
MC		 3092 struct ext4_super_block *es = NULL;
3093 struct ext4_sb_info *sbi;
3094 ext4_fsblk_t block;
3095 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3096 ext4_fsblk_t logical_sb_block;
ac27a0ec 3097 unsigned long offset = 0;
ac27a0ec
DK		 3098 unsigned long journal_devnum = 0;
3099 unsigned long def_mount_opts;
3100 struct inode *root;
9f6200bb 3101 char *cp;
0390131b 3102 const char *descr;
dcc7dae3 3103 int ret = -ENOMEM;
281b5995 3104 int blocksize, clustersize;
4ec11028
TT		 3105 unsigned int db_count;
3106 unsigned int i;
281b5995 3107 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3108 __u64 blocks_count;
833f4077 3109 int err;
b3881f74 3110 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3111 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3112
3113 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3114 if (!sbi)
dcc7dae3 3115 goto out_free_orig;
705895b6
PE		 3116
3117 sbi->s_blockgroup_lock =
3118 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3119 if (!sbi->s_blockgroup_lock) {
3120 kfree(sbi);
dcc7dae3 3121 goto out_free_orig;
705895b6 3122 }
ac27a0ec
DK		 3123 sb->s_fs_info = sbi;
3124 sbi->s_mount_opt = 0;
617ba13b
MC		 3125 sbi->s_resuid = EXT4_DEF_RESUID;
3126 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 3127 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3128 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3129 if (sb->s_bdev->bd_part)
3130 sbi->s_sectors_written_start =
3131 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3132
9f6200bb
TT		 3133 /* Cleanup superblock name */
3134 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3135 *cp = '!';
3136
dcc7dae3 3137 ret = -EINVAL;
617ba13b 3138 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3139 if (!blocksize) {
b31e1552 3140 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3141 goto out_fail;
3142 }
3143
3144 /*
617ba13b 3145 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3146 * block sizes. We need to calculate the offset from buffer start.
3147 */
617ba13b 3148 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3149 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3150 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3151 } else {
70bbb3e0 3152 logical_sb_block = sb_block;
ac27a0ec
DK		 3153 }
3154
70bbb3e0 3155 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3156 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3157 goto out_fail;
3158 }
3159 /*
3160 * Note: s_es must be initialized as soon as possible because
617ba13b 3161 * some ext4 macro-instructions depend on its value
ac27a0ec 3162 */
617ba13b 3163 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3164 sbi->s_es = es;
3165 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3166 if (sb->s_magic != EXT4_SUPER_MAGIC)
3167 goto cantfind_ext4;
afc32f7e 3168 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 3169
3170 /* Set defaults before we parse the mount options */
3171 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3172 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3173 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3174 set_opt(sb, DEBUG);
437ca0fd
DM		 3175 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3176 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3177 "2.6.38");
fd8c37ec 3178 set_opt(sb, GRPID);
437ca0fd 3179 }
617ba13b 3180 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3181 set_opt(sb, NO_UID32);
ea663336 3182 /* xattr user namespace & acls are now defaulted on */
03010a33 3183#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3184 set_opt(sb, XATTR_USER);
2e7842b8 3185#endif
03010a33 3186#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3187 set_opt(sb, POSIX_ACL);
2e7842b8 3188#endif
6fd7a467 3189 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3190 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3191 set_opt(sb, JOURNAL_DATA);
617ba13b 3192 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3193 set_opt(sb, ORDERED_DATA);
617ba13b 3194 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3195 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3196
3197 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3198 set_opt(sb, ERRORS_PANIC);
bb4f397a 3199 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3200 set_opt(sb, ERRORS_CONT);
bb4f397a 3201 else
fd8c37ec 3202 set_opt(sb, ERRORS_RO);
8b67f04a 3203 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3204 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3205 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3206 set_opt(sb, DISCARD);
ac27a0ec
DK		 3207
3208 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3209 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3210 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3211 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3212 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3213
8b67f04a 3214 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3215 set_opt(sb, BARRIER);
ac27a0ec 3216
dd919b98
AK		 3217 /*
3218 * enable delayed allocation by default
3219 * Use -o nodelalloc to turn it off
3220 */
8b67f04a
TT		 3221 if (!IS_EXT3_SB(sb) &&
3222 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3223 set_opt(sb, DELALLOC);
dd919b98 3224
51ce6511
LC		 3225 /*
3226 * set default s_li_wait_mult for lazyinit, for the case there is
3227 * no mount option specified.
3228 */
3229 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3230
8b67f04a 3231 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3232 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3233 ext4_msg(sb, KERN_WARNING,
3234 "failed to parse options in superblock: %s",
3235 sbi->s_es->s_mount_opts);
3236 }
b3881f74 3237 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3238 &journal_ioprio, 0))
ac27a0ec
DK		 3239 goto failed_mount;
3240
56889787
TT		 3241 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3242 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3243 "with data=journal disables delayed "
3244 "allocation and O_DIRECT support!\n");
3245 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3246 ext4_msg(sb, KERN_ERR, "can't mount with "
3247 "both data=journal and delalloc");
3248 goto failed_mount;
3249 }
3250 if (test_opt(sb, DIOREAD_NOLOCK)) {
3251 ext4_msg(sb, KERN_ERR, "can't mount with "
3252 "both data=journal and delalloc");
3253 goto failed_mount;
3254 }
3255 if (test_opt(sb, DELALLOC))
3256 clear_opt(sb, DELALLOC);
3257 }
3258
3259 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3260 if (test_opt(sb, DIOREAD_NOLOCK)) {
3261 if (blocksize < PAGE_SIZE) {
3262 ext4_msg(sb, KERN_ERR, "can't mount with "
3263 "dioread_nolock if block size != PAGE_SIZE");
3264 goto failed_mount;
3265 }
3266 }
3267
ac27a0ec 3268 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3269 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3270
617ba13b
MC		 3271 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3272 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3273 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3274 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3275 ext4_msg(sb, KERN_WARNING,
3276 "feature flags set on rev 0 fs, "
3277 "running e2fsck is recommended");
469108ff 3278
2035e776
TT		 3279 if (IS_EXT2_SB(sb)) {
3280 if (ext2_feature_set_ok(sb))
3281 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3282 "using the ext4 subsystem");
3283 else {
3284 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3285 "to feature incompatibilities");
3286 goto failed_mount;
3287 }
3288 }
3289
3290 if (IS_EXT3_SB(sb)) {
3291 if (ext3_feature_set_ok(sb))
3292 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3293 "using the ext4 subsystem");
3294 else {
3295 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3296 "to feature incompatibilities");
3297 goto failed_mount;
3298 }
3299 }
3300
ac27a0ec
DK		 3301 /*
3302 * Check feature flags regardless of the revision level, since we
3303 * previously didn't change the revision level when setting the flags,
3304 * so there is a chance incompat flags are set on a rev 0 filesystem.
3305 */
a13fb1a4 3306 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3307 goto failed_mount;
a13fb1a4 3308
617ba13b
MC		 3309 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3310 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3311 ext4_msg(sb, KERN_ERR,
3312 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3313 goto failed_mount;
3314 }
3315
ac27a0ec 3316 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3317 /* Validate the filesystem blocksize */
3318 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3319 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3320 blocksize);
ac27a0ec
DK		 3321 goto failed_mount;
3322 }
3323
2b2d6d01 3324 brelse(bh);
70bbb3e0
AM		 3325 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3326 offset = do_div(logical_sb_block, blocksize);
3327 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3328 if (!bh) {
b31e1552
ES		 3329 ext4_msg(sb, KERN_ERR,
3330 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3331 goto failed_mount;
3332 }
617ba13b 3333 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3334 sbi->s_es = es;
617ba13b 3335 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3336 ext4_msg(sb, KERN_ERR,
3337 "Magic mismatch, very weird!");
ac27a0ec
DK		 3338 goto failed_mount;
3339 }
3340 }
3341
a13fb1a4
ES		 3342 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3343 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3344 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3345 has_huge_files);
3346 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3347
617ba13b
MC		 3348 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3349 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3350 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3351 } else {
3352 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3353 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3354 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3355 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3356 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3357 ext4_msg(sb, KERN_ERR,
3358 "unsupported inode size: %d",
2b2d6d01 3359 sbi->s_inode_size);
ac27a0ec
DK		 3360 goto failed_mount;
3361 }
ef7f3835
KS		 3362 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3363 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3364 }
0b8e58a1 3365
0d1ee42f
AR		 3366 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3367 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3368 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3369 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3370 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3371 ext4_msg(sb, KERN_ERR,
3372 "unsupported descriptor size %lu",
0d1ee42f
AR		 3373 sbi->s_desc_size);
3374 goto failed_mount;
3375 }
3376 } else
3377 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3378
ac27a0ec 3379 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3380 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3381 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3382 goto cantfind_ext4;
0b8e58a1 3383
617ba13b 3384 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3385 if (sbi->s_inodes_per_block == 0)
617ba13b 3386 goto cantfind_ext4;
ac27a0ec
DK		 3387 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3388 sbi->s_inodes_per_block;
0d1ee42f 3389 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3390 sbi->s_sbh = bh;
3391 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3392 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3393 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3394
2b2d6d01 3395 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3396 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3397 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3398 i = le32_to_cpu(es->s_flags);
3399 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3400 sbi->s_hash_unsigned = 3;
3401 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3402#ifdef __CHAR_UNSIGNED__
3403 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3404 sbi->s_hash_unsigned = 3;
3405#else
3406 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3407#endif
3408 sb->s_dirt = 1;
3409 }
ac27a0ec 3410
281b5995
TT		 3411 /* Handle clustersize */
3412 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3413 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3414 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3415 if (has_bigalloc) {
3416 if (clustersize < blocksize) {
3417 ext4_msg(sb, KERN_ERR,
3418 "cluster size (%d) smaller than "
3419 "block size (%d)", clustersize, blocksize);
3420 goto failed_mount;
3421 }
3422 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3423 le32_to_cpu(es->s_log_block_size);
3424 sbi->s_clusters_per_group =
3425 le32_to_cpu(es->s_clusters_per_group);
3426 if (sbi->s_clusters_per_group > blocksize * 8) {
3427 ext4_msg(sb, KERN_ERR,
3428 "#clusters per group too big: %lu",
3429 sbi->s_clusters_per_group);
3430 goto failed_mount;
3431 }
3432 if (sbi->s_blocks_per_group !=
3433 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3434 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3435 "clusters per group (%lu) inconsistent",
3436 sbi->s_blocks_per_group,
3437 sbi->s_clusters_per_group);
3438 goto failed_mount;
3439 }
3440 } else {
3441 if (clustersize != blocksize) {
3442 ext4_warning(sb, "fragment/cluster size (%d) != "
3443 "block size (%d)", clustersize,
3444 blocksize);
3445 clustersize = blocksize;
3446 }
3447 if (sbi->s_blocks_per_group > blocksize * 8) {
3448 ext4_msg(sb, KERN_ERR,
3449 "#blocks per group too big: %lu",
3450 sbi->s_blocks_per_group);
3451 goto failed_mount;
3452 }
3453 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3454 sbi->s_cluster_bits = 0;
ac27a0ec 3455 }
281b5995
TT		 3456 sbi->s_cluster_ratio = clustersize / blocksize;
3457
ac27a0ec 3458 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3459 ext4_msg(sb, KERN_ERR,
3460 "#inodes per group too big: %lu",
2b2d6d01 3461 sbi->s_inodes_per_group);
ac27a0ec
DK		 3462 goto failed_mount;
3463 }
3464
bf43d84b
ES		 3465 /*
3466 * Test whether we have more sectors than will fit in sector_t,
3467 * and whether the max offset is addressable by the page cache.
3468 */
5a9ae68a 3469 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3470 ext4_blocks_count(es));
5a9ae68a 3471 if (err) {
b31e1552 3472 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3473 " too large to mount safely on this system");
ac27a0ec 3474 if (sizeof(sector_t) < 8)
90c699a9 3475 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3476 ret = err;
ac27a0ec
DK		 3477 goto failed_mount;
3478 }
3479
617ba13b
MC		 3480 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3481 goto cantfind_ext4;
e7c95593 3482
0f2ddca6
FTN		 3483 /* check blocks count against device size */
3484 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3485 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3486 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3487 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3488 ext4_blocks_count(es), blocks_count);
3489 goto failed_mount;
3490 }
3491
0b8e58a1
AD		 3492 /*
3493 * It makes no sense for the first data block to be beyond the end
3494 * of the filesystem.
3495 */
3496 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3497 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3498 "block %u is beyond end of filesystem (%llu)",
3499 le32_to_cpu(es->s_first_data_block),
3500 ext4_blocks_count(es));
e7c95593
ES		 3501 goto failed_mount;
3502 }
bd81d8ee
LV		 3503 blocks_count = (ext4_blocks_count(es) -
3504 le32_to_cpu(es->s_first_data_block) +
3505 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3506 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3507 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3508 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3509 "(block count %llu, first data block %u, "
b31e1552 3510 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3511 ext4_blocks_count(es),
3512 le32_to_cpu(es->s_first_data_block),
3513 EXT4_BLOCKS_PER_GROUP(sb));
3514 goto failed_mount;
3515 }
bd81d8ee 3516 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3517 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3518 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3519 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3520 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3521 sbi->s_group_desc = ext4_kvmalloc(db_count *
3522 sizeof(struct buffer_head *),
3523 GFP_KERNEL);
ac27a0ec 3524 if (sbi->s_group_desc == NULL) {
b31e1552 3525 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3526 goto failed_mount;
3527 }
3528
9f6200bb
TT		 3529 if (ext4_proc_root)
3530 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3531
705895b6 3532 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3533
3534 for (i = 0; i < db_count; i++) {
70bbb3e0 3535 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3536 sbi->s_group_desc[i] = sb_bread(sb, block);
3537 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3538 ext4_msg(sb, KERN_ERR,
3539 "can't read group descriptor %d", i);
ac27a0ec
DK		 3540 db_count = i;
3541 goto failed_mount2;
3542 }
3543 }
bfff6873 3544 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3545 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3546 goto failed_mount2;
3547 }
772cb7c8
JS		 3548 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3549 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3550 ext4_msg(sb, KERN_ERR,
3551 "unable to initialize "
3552 "flex_bg meta info!");
772cb7c8
JS		 3553 goto failed_mount2;
3554 }
3555
ac27a0ec
DK		 3556 sbi->s_gdb_count = db_count;
3557 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3558 spin_lock_init(&sbi->s_next_gen_lock);
3559
04496411
TM		 3560 init_timer(&sbi->s_err_report);
3561 sbi->s_err_report.function = print_daily_error_info;
3562 sbi->s_err_report.data = (unsigned long) sb;
3563
57042651 3564 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3565 ext4_count_free_clusters(sb));
ce7e010a
TT		 3566 if (!err) {
3567 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3568 ext4_count_free_inodes(sb));
3569 }
3570 if (!err) {
3571 err = percpu_counter_init(&sbi->s_dirs_counter,
3572 ext4_count_dirs(sb));
3573 }
3574 if (!err) {
57042651 3575 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3576 }
3577 if (err) {
3578 ext4_msg(sb, KERN_ERR, "insufficient memory");
3579 goto failed_mount3;
3580 }
3581
c9de560d 3582 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3583 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3584
ac27a0ec
DK		 3585 /*
3586 * set up enough so that it can read an inode
3587 */
9ca92389
TT		 3588 if (!test_opt(sb, NOLOAD) &&
3589 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3590 sb->s_op = &ext4_sops;
3591 else
3592 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3593 sb->s_export_op = &ext4_export_ops;
3594 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3595#ifdef CONFIG_QUOTA
617ba13b
MC		 3596 sb->s_qcop = &ext4_qctl_operations;
3597 sb->dq_op = &ext4_quota_operations;
ac27a0ec 3598#endif
f2fa2ffc
AK		 3599 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3600
ac27a0ec 3601 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3602 mutex_init(&sbi->s_orphan_lock);
8f82f840 3603 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3604
3605 sb->s_root = NULL;
3606
3607 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3608 EXT4_HAS_INCOMPAT_FEATURE(sb,
3609 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3610
c5e06d10
JL		 3611 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3612 !(sb->s_flags & MS_RDONLY))
3613 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3614 goto failed_mount3;
3615
ac27a0ec
DK		 3616 /*
3617 * The first inode we look at is the journal inode. Don't try
3618 * root first: it may be modified in the journal!
3619 */
3620 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3621 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3622 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3623 goto failed_mount3;
0390131b
FM		 3624 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3625 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3626 ext4_msg(sb, KERN_ERR, "required journal recovery "
3627 "suppressed and not mounted read-only");
744692dc 3628 goto failed_mount_wq;
ac27a0ec 3629 } else {
fd8c37ec 3630 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3631 sbi->s_journal = NULL;
3632 needs_recovery = 0;
3633 goto no_journal;
ac27a0ec
DK		 3634 }
3635
eb40a09c
JS		 3636 if (ext4_blocks_count(es) > 0xffffffffULL &&
3637 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3638 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3639 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3640 goto failed_mount_wq;
eb40a09c
JS		 3641 }
3642
d4da6c9c
LT		 3643 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3644 jbd2_journal_set_features(sbi->s_journal,
3645 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3646 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3647 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3648 jbd2_journal_set_features(sbi->s_journal,
3649 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3650 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3651 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3652 } else {
3653 jbd2_journal_clear_features(sbi->s_journal,
3654 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3655 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3656 }
818d276c 3657
ac27a0ec
DK		 3658 /* We have now updated the journal if required, so we can
3659 * validate the data journaling mode. */
3660 switch (test_opt(sb, DATA_FLAGS)) {
3661 case 0:
3662 /* No mode set, assume a default based on the journal
63f57933
AM		 3663 * capabilities: ORDERED_DATA if the journal can
3664 * cope, else JOURNAL_DATA
3665 */
dab291af
MC		 3666 if (jbd2_journal_check_available_features
3667 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3668 set_opt(sb, ORDERED_DATA);
ac27a0ec 3669 else
fd8c37ec 3670 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3671 break;
3672
617ba13b
MC		 3673 case EXT4_MOUNT_ORDERED_DATA:
3674 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3675 if (!jbd2_journal_check_available_features
3676 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3677 ext4_msg(sb, KERN_ERR, "Journal does not support "
3678 "requested data journaling mode");
744692dc 3679 goto failed_mount_wq;
ac27a0ec
DK		 3680 }
3681 default:
3682 break;
3683 }
b3881f74 3684 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3685
18aadd47
BJ		 3686 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3687
ce7e010a
TT		 3688 /*
3689 * The journal may have updated the bg summary counts, so we
3690 * need to update the global counters.
3691 */
57042651 3692 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3693 ext4_count_free_clusters(sb));
ce7e010a
TT		 3694 percpu_counter_set(&sbi->s_freeinodes_counter,
3695 ext4_count_free_inodes(sb));
3696 percpu_counter_set(&sbi->s_dirs_counter,
3697 ext4_count_dirs(sb));
57042651 3698 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3699
ce7e010a 3700no_journal:
fd89d5f2
TH		 3701 /*
3702 * The maximum number of concurrent works can be high and
3703 * concurrency isn't really necessary. Limit it to 1.
3704 */
3705 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3706 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3707 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3708 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3709 goto failed_mount_wq;
3710 }
3711
ac27a0ec 3712 /*
dab291af 3713 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3714 * so we can safely mount the rest of the filesystem now.
3715 */
3716
1d1fe1ee
DH		 3717 root = ext4_iget(sb, EXT4_ROOT_INO);
3718 if (IS_ERR(root)) {
b31e1552 3719 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3720 ret = PTR_ERR(root);
32a9bb57 3721 root = NULL;
ac27a0ec
DK		 3722 goto failed_mount4;
3723 }
3724 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3725 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3726 iput(root);
ac27a0ec
DK		 3727 goto failed_mount4;
3728 }
1d1fe1ee
DH		 3729 sb->s_root = d_alloc_root(root);
3730 if (!sb->s_root) {
94bf608a 3731 iput(root);
b31e1552 3732 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3733 ret = -ENOMEM;
3734 goto failed_mount4;
3735 }
ac27a0ec 3736
2b2d6d01 3737 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3738
3739 /* determine the minimum size of new large inodes, if present */
3740 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3741 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3742 EXT4_GOOD_OLD_INODE_SIZE;
3743 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3744 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3745 if (sbi->s_want_extra_isize <
3746 le16_to_cpu(es->s_want_extra_isize))
3747 sbi->s_want_extra_isize =
3748 le16_to_cpu(es->s_want_extra_isize);
3749 if (sbi->s_want_extra_isize <
3750 le16_to_cpu(es->s_min_extra_isize))
3751 sbi->s_want_extra_isize =
3752 le16_to_cpu(es->s_min_extra_isize);
3753 }
3754 }
3755 /* Check if enough inode space is available */
3756 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3757 sbi->s_inode_size) {
3758 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3759 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3760 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3761 "available");
ef7f3835
KS		 3762 }
3763
6fd058f7
TT		 3764 err = ext4_setup_system_zone(sb);
3765 if (err) {
b31e1552 3766 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3767 "zone (%d)", err);
94bf608a 3768 goto failed_mount4a;
6fd058f7
TT		 3769 }
3770
c2774d84
AK		 3771 ext4_ext_init(sb);
3772 err = ext4_mb_init(sb, needs_recovery);
3773 if (err) {
421f91d2 3774 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3775 err);
dcf2d804 3776 goto failed_mount5;
c2774d84
AK		 3777 }
3778
bfff6873
LC		 3779 err = ext4_register_li_request(sb, first_not_zeroed);
3780 if (err)
dcf2d804 3781 goto failed_mount6;
bfff6873 3782
3197ebdb
TT		 3783 sbi->s_kobj.kset = ext4_kset;
3784 init_completion(&sbi->s_kobj_unregister);
3785 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3786 "%s", sb->s_id);
dcf2d804
TM		 3787 if (err)
3788 goto failed_mount7;
3197ebdb 3789
617ba13b
MC		 3790 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3791 ext4_orphan_cleanup(sb, es);
3792 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3793 if (needs_recovery) {
b31e1552 3794 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3795 ext4_mark_recovery_complete(sb, es);
3796 }
3797 if (EXT4_SB(sb)->s_journal) {
3798 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3799 descr = " journalled data mode";
3800 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3801 descr = " ordered data mode";
3802 else
3803 descr = " writeback data mode";
3804 } else
3805 descr = "out journal";
3806
d4c402d9 3807 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3808 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3809 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3810
66e61a9e
TT		 3811 if (es->s_error_count)
3812 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3813
d4c402d9 3814 kfree(orig_data);
ac27a0ec
DK		 3815 return 0;
3816
617ba13b 3817cantfind_ext4:
ac27a0ec 3818 if (!silent)
b31e1552 3819 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3820 goto failed_mount;
3821
dcf2d804
TM		 3822failed_mount7:
3823 ext4_unregister_li_request(sb);
3824failed_mount6:
dcf2d804 3825 ext4_mb_release(sb);
94bf608a
AV		 3826failed_mount5:
3827 ext4_ext_release(sb);
dcf2d804 3828 ext4_release_system_zone(sb);
94bf608a
AV		 3829failed_mount4a:
3830 dput(sb->s_root);
32a9bb57 3831 sb->s_root = NULL;
94bf608a 3832failed_mount4:
b31e1552 3833 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3834 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3835failed_mount_wq:
0390131b
FM		 3836 if (sbi->s_journal) {
3837 jbd2_journal_destroy(sbi->s_journal);
3838 sbi->s_journal = NULL;
3839 }
ac27a0ec 3840failed_mount3:
04496411 3841 del_timer(&sbi->s_err_report);
9933fc0a
TT		 3842 if (sbi->s_flex_groups)
3843 ext4_kvfree(sbi->s_flex_groups);
57042651 3844 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 3845 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3846 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 3847 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 3848 if (sbi->s_mmp_tsk)
3849 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3850failed_mount2:
3851 for (i = 0; i < db_count; i++)
3852 brelse(sbi->s_group_desc[i]);
f18a5f21 3853 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 3854failed_mount:
240799cd 3855 if (sbi->s_proc) {
9f6200bb 3856 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3857 }
ac27a0ec
DK		 3858#ifdef CONFIG_QUOTA
3859 for (i = 0; i < MAXQUOTAS; i++)
3860 kfree(sbi->s_qf_names[i]);
3861#endif
617ba13b 3862 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3863 brelse(bh);
3864out_fail:
3865 sb->s_fs_info = NULL;
f6830165 3866 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3867 kfree(sbi);
dcc7dae3 3868out_free_orig:
d4c402d9 3869 kfree(orig_data);
1d1fe1ee 3870 return ret;
ac27a0ec
DK		 3871}
3872
3873/*
3874 * Setup any per-fs journal parameters now. We'll do this both on
3875 * initial mount, once the journal has been initialised but before we've
3876 * done any recovery; and again on any subsequent remount.
3877 */
617ba13b 3878static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3879{
617ba13b 3880 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3881
30773840
TT		 3882 journal->j_commit_interval = sbi->s_commit_interval;
3883 journal->j_min_batch_time = sbi->s_min_batch_time;
3884 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3885
a931da6a 3886 write_lock(&journal->j_state_lock);
ac27a0ec 3887 if (test_opt(sb, BARRIER))
dab291af 3888 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3889 else
dab291af 3890 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3891 if (test_opt(sb, DATA_ERR_ABORT))
3892 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3893 else
3894 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3895 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3896}
3897
617ba13b 3898static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3899 unsigned int journal_inum)
3900{
3901 struct inode *journal_inode;
3902 journal_t *journal;
3903
0390131b
FM		 3904 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3905
ac27a0ec
DK		 3906 /* First, test for the existence of a valid inode on disk. Bad
3907 * things happen if we iget() an unused inode, as the subsequent
3908 * iput() will try to delete it. */
3909
1d1fe1ee
DH		 3910 journal_inode = ext4_iget(sb, journal_inum);
3911 if (IS_ERR(journal_inode)) {
b31e1552 3912 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3913 return NULL;
3914 }
3915 if (!journal_inode->i_nlink) {
3916 make_bad_inode(journal_inode);
3917 iput(journal_inode);
b31e1552 3918 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3919 return NULL;
3920 }
3921
e5f8eab8 3922 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3923 journal_inode, journal_inode->i_size);
1d1fe1ee 3924 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3925 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3926 iput(journal_inode);
3927 return NULL;
3928 }
3929
dab291af 3930 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3931 if (!journal) {
b31e1552 3932 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3933 iput(journal_inode);
3934 return NULL;
3935 }
3936 journal->j_private = sb;
617ba13b 3937 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3938 return journal;
3939}
3940
617ba13b 3941static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3942 dev_t j_dev)
3943{
2b2d6d01 3944 struct buffer_head *bh;
ac27a0ec 3945 journal_t *journal;
617ba13b
MC		 3946 ext4_fsblk_t start;
3947 ext4_fsblk_t len;
ac27a0ec 3948 int hblock, blocksize;
617ba13b 3949 ext4_fsblk_t sb_block;
ac27a0ec 3950 unsigned long offset;
2b2d6d01 3951 struct ext4_super_block *es;
ac27a0ec
DK		 3952 struct block_device *bdev;
3953
0390131b
FM		 3954 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3955
b31e1552 3956 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3957 if (bdev == NULL)
3958 return NULL;
3959
ac27a0ec 3960 blocksize = sb->s_blocksize;
e1defc4f 3961 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3962 if (blocksize < hblock) {
b31e1552
ES		 3963 ext4_msg(sb, KERN_ERR,
3964 "blocksize too small for journal device");
ac27a0ec
DK		 3965 goto out_bdev;
3966 }
3967
617ba13b
MC		 3968 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3969 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3970 set_blocksize(bdev, blocksize);
3971 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3972 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3973 "external journal");
ac27a0ec
DK		 3974 goto out_bdev;
3975 }
3976
617ba13b
MC		 3977 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3978 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3979 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3980 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3981 ext4_msg(sb, KERN_ERR, "external journal has "
3982 "bad superblock");
ac27a0ec
DK		 3983 brelse(bh);
3984 goto out_bdev;
3985 }
3986
617ba13b 3987 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3988 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3989 brelse(bh);
3990 goto out_bdev;
3991 }
3992
bd81d8ee 3993 len = ext4_blocks_count(es);
ac27a0ec
DK		 3994 start = sb_block + 1;
3995 brelse(bh); /* we're done with the superblock */
3996
dab291af 3997 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3998 start, len, blocksize);
3999 if (!journal) {
b31e1552 4000 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4001 goto out_bdev;
4002 }
4003 journal->j_private = sb;
4004 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4005 wait_on_buffer(journal->j_sb_buffer);
4006 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4007 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4008 goto out_journal;
4009 }
4010 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4011 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4012 "user (unsupported) - %d",
ac27a0ec
DK		 4013 be32_to_cpu(journal->j_superblock->s_nr_users));
4014 goto out_journal;
4015 }
617ba13b
MC		 4016 EXT4_SB(sb)->journal_bdev = bdev;
4017 ext4_init_journal_params(sb, journal);
ac27a0ec 4018 return journal;
0b8e58a1 4019
ac27a0ec 4020out_journal:
dab291af 4021 jbd2_journal_destroy(journal);
ac27a0ec 4022out_bdev:
617ba13b 4023 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4024 return NULL;
4025}
4026
617ba13b
MC		 4027static int ext4_load_journal(struct super_block *sb,
4028 struct ext4_super_block *es,
ac27a0ec
DK		 4029 unsigned long journal_devnum)
4030{
4031 journal_t *journal;
4032 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4033 dev_t journal_dev;
4034 int err = 0;
4035 int really_read_only;
4036
0390131b
FM		 4037 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4038
ac27a0ec
DK		 4039 if (journal_devnum &&
4040 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4041 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4042 "numbers have changed");
ac27a0ec
DK		 4043 journal_dev = new_decode_dev(journal_devnum);
4044 } else
4045 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4046
4047 really_read_only = bdev_read_only(sb->s_bdev);
4048
4049 /*
4050 * Are we loading a blank journal or performing recovery after a
4051 * crash? For recovery, we need to check in advance whether we
4052 * can get read-write access to the device.
4053 */
617ba13b 4054 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4055 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4056 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4057 "required on readonly filesystem");
ac27a0ec 4058 if (really_read_only) {
b31e1552
ES		 4059 ext4_msg(sb, KERN_ERR, "write access "
4060 "unavailable, cannot proceed");
ac27a0ec
DK		 4061 return -EROFS;
4062 }
b31e1552
ES		 4063 ext4_msg(sb, KERN_INFO, "write access will "
4064 "be enabled during recovery");
ac27a0ec
DK		 4065 }
4066 }
4067
4068 if (journal_inum && journal_dev) {
b31e1552
ES		 4069 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4070 "and inode journals!");
ac27a0ec
DK		 4071 return -EINVAL;
4072 }
4073
4074 if (journal_inum) {
617ba13b 4075 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4076 return -EINVAL;
4077 } else {
617ba13b 4078 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4079 return -EINVAL;
4080 }
4081
90576c0b 4082 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4083 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4084
617ba13b 4085 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4086 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4087 if (!err) {
4088 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4089 if (save)
4090 memcpy(save, ((char *) es) +
4091 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4092 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4093 if (save)
4094 memcpy(((char *) es) + EXT4_S_ERR_START,
4095 save, EXT4_S_ERR_LEN);
4096 kfree(save);
4097 }
ac27a0ec
DK		 4098
4099 if (err) {
b31e1552 4100 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4101 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4102 return err;
4103 }
4104
617ba13b
MC		 4105 EXT4_SB(sb)->s_journal = journal;
4106 ext4_clear_journal_err(sb, es);
ac27a0ec 4107
c41303ce 4108 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4109 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4110 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4111
4112 /* Make sure we flush the recovery flag to disk. */
e2d67052 4113 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4114 }
4115
4116 return 0;
4117}
4118
e2d67052 4119static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4120{
e2d67052 4121 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4122 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4123 int error = 0;
ac27a0ec 4124
7c2e7087 4125 if (!sbh || block_device_ejected(sb))
c4be0c1d 4126 return error;
914258bf
TT		 4127 if (buffer_write_io_error(sbh)) {
4128 /*
4129 * Oh, dear. A previous attempt to write the
4130 * superblock failed. This could happen because the
4131 * USB device was yanked out. Or it could happen to
4132 * be a transient write error and maybe the block will
4133 * be remapped. Nothing we can do but to retry the
4134 * write and hope for the best.
4135 */
b31e1552
ES		 4136 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4137 "superblock detected");
914258bf
TT		 4138 clear_buffer_write_io_error(sbh);
4139 set_buffer_uptodate(sbh);
4140 }
71290b36
TT		 4141 /*
4142 * If the file system is mounted read-only, don't update the
4143 * superblock write time. This avoids updating the superblock
4144 * write time when we are mounting the root file system
4145 * read/only but we need to replay the journal; at that point,
4146 * for people who are east of GMT and who make their clock
4147 * tick in localtime for Windows bug-for-bug compatibility,
4148 * the clock is set in the future, and this will cause e2fsck
4149 * to complain and force a full file system check.
4150 */
4151 if (!(sb->s_flags & MS_RDONLY))
4152 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4153 if (sb->s_bdev->bd_part)
4154 es->s_kbytes_written =
4155 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4156 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4157 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4158 else
4159 es->s_kbytes_written =
4160 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4161 ext4_free_blocks_count_set(es,
4162 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4163 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4164 es->s_free_inodes_count =
4165 cpu_to_le32(percpu_counter_sum_positive(
4166 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 4167 sb->s_dirt = 0;
ac27a0ec
DK		 4168 BUFFER_TRACE(sbh, "marking dirty");
4169 mark_buffer_dirty(sbh);
914258bf 4170 if (sync) {
c4be0c1d
TS		 4171 error = sync_dirty_buffer(sbh);
4172 if (error)
4173 return error;
4174
4175 error = buffer_write_io_error(sbh);
4176 if (error) {
b31e1552
ES		 4177 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4178 "superblock");
914258bf
TT		 4179 clear_buffer_write_io_error(sbh);
4180 set_buffer_uptodate(sbh);
4181 }
4182 }
c4be0c1d 4183 return error;
ac27a0ec
DK		 4184}
4185
ac27a0ec
DK		 4186/*
4187 * Have we just finished recovery? If so, and if we are mounting (or
4188 * remounting) the filesystem readonly, then we will end up with a
4189 * consistent fs on disk. Record that fact.
4190 */
2b2d6d01
TT		 4191static void ext4_mark_recovery_complete(struct super_block *sb,
4192 struct ext4_super_block *es)
ac27a0ec 4193{
617ba13b 4194 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4195
0390131b
FM		 4196 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4197 BUG_ON(journal != NULL);
4198 return;
4199 }
dab291af 4200 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4201 if (jbd2_journal_flush(journal) < 0)
4202 goto out;
4203
617ba13b 4204 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4205 sb->s_flags & MS_RDONLY) {
617ba13b 4206 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4207 ext4_commit_super(sb, 1);
ac27a0ec 4208 }
7ffe1ea8
HK		 4209
4210out:
dab291af 4211 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4212}
4213
4214/*
4215 * If we are mounting (or read-write remounting) a filesystem whose journal
4216 * has recorded an error from a previous lifetime, move that error to the
4217 * main filesystem now.
4218 */
2b2d6d01
TT		 4219static void ext4_clear_journal_err(struct super_block *sb,
4220 struct ext4_super_block *es)
ac27a0ec
DK		 4221{
4222 journal_t *journal;
4223 int j_errno;
4224 const char *errstr;
4225
0390131b
FM		 4226 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4227
617ba13b 4228 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4229
4230 /*
4231 * Now check for any error status which may have been recorded in the
617ba13b 4232 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4233 */
4234
dab291af 4235 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4236 if (j_errno) {
4237 char nbuf[16];
4238
617ba13b 4239 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4240 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4241 "from previous mount: %s", errstr);
12062ddd 4242 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4243
617ba13b
MC		 4244 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4245 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4246 ext4_commit_super(sb, 1);
ac27a0ec 4247
dab291af 4248 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4249 }
4250}
4251
4252/*
4253 * Force the running and committing transactions to commit,
4254 * and wait on the commit.
4255 */
617ba13b 4256int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4257{
4258 journal_t *journal;
0390131b 4259 int ret = 0;
ac27a0ec
DK		 4260
4261 if (sb->s_flags & MS_RDONLY)
4262 return 0;
4263
617ba13b 4264 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4265 if (journal) {
437f88cc 4266 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4267 ret = ext4_journal_force_commit(journal);
6b0310fb 4268 }
0390131b 4269
ac27a0ec
DK		 4270 return ret;
4271}
4272
2b2d6d01 4273static void ext4_write_super(struct super_block *sb)
ac27a0ec 4274{
ebc1ac16 4275 lock_super(sb);
9ca92389 4276 ext4_commit_super(sb, 1);
ebc1ac16 4277 unlock_super(sb);
ac27a0ec
DK		 4278}
4279
617ba13b 4280static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4281{
14ce0cb4 4282 int ret = 0;
9eddacf9 4283 tid_t target;
8d5d02e6 4284 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4285
9bffad1e 4286 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4287 flush_workqueue(sbi->dio_unwritten_wq);
4288 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4289 if (wait)
8d5d02e6 4290 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4291 }
14ce0cb4 4292 return ret;
ac27a0ec
DK		 4293}
4294
4295/*
4296 * LVM calls this function before a (read-only) snapshot is created. This
4297 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4298 *
4299 * Note that only this function cannot bring a filesystem to be in a clean
4300 * state independently, because ext4 prevents a new handle from being started
4301 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4302 * the upper layer.
ac27a0ec 4303 */
c4be0c1d 4304static int ext4_freeze(struct super_block *sb)
ac27a0ec 4305{
c4be0c1d
TS		 4306 int error = 0;
4307 journal_t *journal;
ac27a0ec 4308
9ca92389
TT		 4309 if (sb->s_flags & MS_RDONLY)
4310 return 0;
ac27a0ec 4311
9ca92389 4312 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4313
9ca92389
TT		 4314 /* Now we set up the journal barrier. */
4315 jbd2_journal_lock_updates(journal);
ac27a0ec 4316
9ca92389
TT		 4317 /*
4318 * Don't clear the needs_recovery flag if we failed to flush
4319 * the journal.
4320 */
4321 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4322 if (error < 0)
4323 goto out;
9ca92389
TT		 4324
4325 /* Journal blocked and flushed, clear needs_recovery flag. */
4326 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4327 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4328out:
4329 /* we rely on s_frozen to stop further updates */
4330 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4331 return error;
ac27a0ec
DK		 4332}
4333
4334/*
4335 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4336 * flag here, even though the filesystem is not technically dirty yet.
4337 */
c4be0c1d 4338static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4339{
9ca92389
TT		 4340 if (sb->s_flags & MS_RDONLY)
4341 return 0;
4342
4343 lock_super(sb);
4344 /* Reset the needs_recovery flag before the fs is unlocked. */
4345 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4346 ext4_commit_super(sb, 1);
4347 unlock_super(sb);
c4be0c1d 4348 return 0;
ac27a0ec
DK		 4349}
4350
673c6100
TT		 4351/*
4352 * Structure to save mount options for ext4_remount's benefit
4353 */
4354struct ext4_mount_options {
4355 unsigned long s_mount_opt;
a2595b8a 4356 unsigned long s_mount_opt2;
673c6100
TT		 4357 uid_t s_resuid;
4358 gid_t s_resgid;
4359 unsigned long s_commit_interval;
4360 u32 s_min_batch_time, s_max_batch_time;
4361#ifdef CONFIG_QUOTA
4362 int s_jquota_fmt;
4363 char *s_qf_names[MAXQUOTAS];
4364#endif
4365};
4366
2b2d6d01 4367static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4368{
2b2d6d01 4369 struct ext4_super_block *es;
617ba13b 4370 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4371 unsigned long old_sb_flags;
617ba13b 4372 struct ext4_mount_options old_opts;
c79d967d 4373 int enable_quota = 0;
8a266467 4374 ext4_group_t g;
b3881f74 4375 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4376 int err = 0;
ac27a0ec
DK		 4377#ifdef CONFIG_QUOTA
4378 int i;
4379#endif
d4c402d9 4380 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4381
4382 /* Store the original options */
bbd6851a 4383 lock_super(sb);
ac27a0ec
DK		 4384 old_sb_flags = sb->s_flags;
4385 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4386 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4387 old_opts.s_resuid = sbi->s_resuid;
4388 old_opts.s_resgid = sbi->s_resgid;
4389 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4390 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4391 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4392#ifdef CONFIG_QUOTA
4393 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4394 for (i = 0; i < MAXQUOTAS; i++)
4395 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4396#endif
b3881f74
TT		 4397 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4398 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4399
4400 /*
4401 * Allow the "check" option to be passed as a remount option.
4402 */
661aa520 4403 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4404 err = -EINVAL;
4405 goto restore_opts;
4406 }
4407
4ab2f15b 4408 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4409 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4410
4411 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4412 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4413
4414 es = sbi->s_es;
4415
b3881f74 4416 if (sbi->s_journal) {
0390131b 4417 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4418 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4419 }
ac27a0ec 4420
661aa520 4421 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4422 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4423 err = -EROFS;
4424 goto restore_opts;
4425 }
4426
4427 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4428 err = dquot_suspend(sb, -1);
4429 if (err < 0)
c79d967d 4430 goto restore_opts;
c79d967d 4431
ac27a0ec
DK		 4432 /*
4433 * First of all, the unconditional stuff we have to do
4434 * to disable replay of the journal when we next remount
4435 */
4436 sb->s_flags |= MS_RDONLY;
4437
4438 /*
4439 * OK, test if we are remounting a valid rw partition
4440 * readonly, and if so set the rdonly flag and then
4441 * mark the partition as valid again.
4442 */
617ba13b
MC		 4443 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4444 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4445 es->s_state = cpu_to_le16(sbi->s_mount_state);
4446
a63c9eb2 4447 if (sbi->s_journal)
0390131b 4448 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4449 } else {
a13fb1a4
ES		 4450 /* Make sure we can mount this feature set readwrite */
4451 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4452 err = -EROFS;
4453 goto restore_opts;
4454 }
8a266467
TT		 4455 /*
4456 * Make sure the group descriptor checksums
0b8e58a1 4457 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4458 */
4459 for (g = 0; g < sbi->s_groups_count; g++) {
4460 struct ext4_group_desc *gdp =
4461 ext4_get_group_desc(sb, g, NULL);
4462
4463 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4464 ext4_msg(sb, KERN_ERR,
4465 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4466 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4467 le16_to_cpu(gdp->bg_checksum));
4468 err = -EINVAL;
4469 goto restore_opts;
4470 }
4471 }
4472
ead6596b
ES		 4473 /*
4474 * If we have an unprocessed orphan list hanging
4475 * around from a previously readonly bdev mount,
4476 * require a full umount/remount for now.
4477 */
4478 if (es->s_last_orphan) {
b31e1552 4479 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4480 "remount RDWR because of unprocessed "
4481 "orphan inode list. Please "
b31e1552 4482 "umount/remount instead");
ead6596b
ES		 4483 err = -EINVAL;
4484 goto restore_opts;
4485 }
4486
ac27a0ec
DK		 4487 /*
4488 * Mounting a RDONLY partition read-write, so reread
4489 * and store the current valid flag. (It may have
4490 * been changed by e2fsck since we originally mounted
4491 * the partition.)
4492 */
0390131b
FM		 4493 if (sbi->s_journal)
4494 ext4_clear_journal_err(sb, es);
ac27a0ec 4495 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4496 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4497 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4498 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4499 EXT4_FEATURE_INCOMPAT_MMP))
4500 if (ext4_multi_mount_protect(sb,
4501 le64_to_cpu(es->s_mmp_block))) {
4502 err = -EROFS;
4503 goto restore_opts;
4504 }
c79d967d 4505 enable_quota = 1;
ac27a0ec
DK		 4506 }
4507 }
bfff6873
LC		 4508
4509 /*
4510 * Reinitialize lazy itable initialization thread based on
4511 * current settings
4512 */
4513 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4514 ext4_unregister_li_request(sb);
4515 else {
4516 ext4_group_t first_not_zeroed;
4517 first_not_zeroed = ext4_has_uninit_itable(sb);
4518 ext4_register_li_request(sb, first_not_zeroed);
4519 }
4520
6fd058f7 4521 ext4_setup_system_zone(sb);
0390131b 4522 if (sbi->s_journal == NULL)
e2d67052 4523 ext4_commit_super(sb, 1);
0390131b 4524
ac27a0ec
DK		 4525#ifdef CONFIG_QUOTA
4526 /* Release old quota file names */
4527 for (i = 0; i < MAXQUOTAS; i++)
4528 if (old_opts.s_qf_names[i] &&
4529 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4530 kfree(old_opts.s_qf_names[i]);
4531#endif
bbd6851a 4532 unlock_super(sb);
c79d967d 4533 if (enable_quota)
0f0dd62f 4534 dquot_resume(sb, -1);
d4c402d9
CW		 4535
4536 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4537 kfree(orig_data);
ac27a0ec 4538 return 0;
0b8e58a1 4539
ac27a0ec
DK		 4540restore_opts:
4541 sb->s_flags = old_sb_flags;
4542 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4543 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4544 sbi->s_resuid = old_opts.s_resuid;
4545 sbi->s_resgid = old_opts.s_resgid;
4546 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4547 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4548 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4549#ifdef CONFIG_QUOTA
4550 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4551 for (i = 0; i < MAXQUOTAS; i++) {
4552 if (sbi->s_qf_names[i] &&
4553 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4554 kfree(sbi->s_qf_names[i]);
4555 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4556 }
4557#endif
bbd6851a 4558 unlock_super(sb);
d4c402d9 4559 kfree(orig_data);
ac27a0ec
DK		 4560 return err;
4561}
4562
f975d6bc
TT		 4563/*
4564 * Note: calculating the overhead so we can be compatible with
4565 * historical BSD practice is quite difficult in the face of
4566 * clusters/bigalloc. This is because multiple metadata blocks from
4567 * different block group can end up in the same allocation cluster.
4568 * Calculating the exact overhead in the face of clustered allocation
4569 * requires either O(all block bitmaps) in memory or O(number of block
4570 * groups**2) in time. We will still calculate the superblock for
4571 * older file systems --- and if we come across with a bigalloc file
4572 * system with zero in s_overhead_clusters the estimate will be close to
4573 * correct especially for very large cluster sizes --- but for newer
4574 * file systems, it's better to calculate this figure once at mkfs
4575 * time, and store it in the superblock. If the superblock value is
4576 * present (even for non-bigalloc file systems), we will use it.
4577 */
2b2d6d01 4578static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4579{
4580 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4581 struct ext4_sb_info *sbi = EXT4_SB(sb);
4582 struct ext4_super_block *es = sbi->s_es;
f975d6bc 4583 struct ext4_group_desc *gdp;
960cc398 4584 u64 fsid;
d02a9391 4585 s64 bfree;
ac27a0ec 4586
5e70030d
BP		 4587 if (test_opt(sb, MINIX_DF)) {
4588 sbi->s_overhead_last = 0;
f975d6bc
TT		 4589 } else if (es->s_overhead_clusters) {
4590 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
6bc9feff 4591 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4592 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4593 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4594
4595 /*
5e70030d
BP		 4596 * Compute the overhead (FS structures). This is constant
4597 * for a given filesystem unless the number of block groups
4598 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4599 */
4600
4601 /*
4602 * All of the blocks before first_data_block are
4603 * overhead
4604 */
f975d6bc 4605 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
ac27a0ec
DK		 4606
4607 /*
f975d6bc 4608 * Add the overhead found in each block group
ac27a0ec
DK		 4609 */
4610 for (i = 0; i < ngroups; i++) {
f975d6bc
TT		 4611 gdp = ext4_get_group_desc(sb, i, NULL);
4612 overhead += ext4_num_overhead_clusters(sb, i, gdp);
ac27a0ec
DK		 4613 cond_resched();
4614 }
5e70030d
BP		 4615 sbi->s_overhead_last = overhead;
4616 smp_wmb();
6bc9feff 4617 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4618 }
4619
617ba13b 4620 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4621 buf->f_bsize = sb->s_blocksize;
f975d6bc
TT		 4622 buf->f_blocks = (ext4_blocks_count(es) -
4623 EXT4_C2B(sbi, sbi->s_overhead_last));
57042651
TT		 4624 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4625 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4626 /* prevent underflow in case that few free space is available */
57042651 4627 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4628 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4629 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4630 buf->f_bavail = 0;
4631 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4632 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4633 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4634 fsid = le64_to_cpup((void *)es->s_uuid) ^
4635 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4636 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4637 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4638
ac27a0ec
DK		 4639 return 0;
4640}
4641
0b8e58a1
AD		 4642/* Helper function for writing quotas on sync - we need to start transaction
4643 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4644 * Process 1 Process 2
617ba13b 4645 * ext4_create() quota_sync()
a269eb18 4646 * jbd2_journal_start() write_dquot()
871a2931 4647 * dquot_initialize() down(dqio_mutex)
dab291af 4648 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4649 *
4650 */
4651
4652#ifdef CONFIG_QUOTA
4653
4654static inline struct inode *dquot_to_inode(struct dquot *dquot)
4655{
4656 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4657}
4658
617ba13b 4659static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4660{
4661 int ret, err;
4662 handle_t *handle;
4663 struct inode *inode;
4664
4665 inode = dquot_to_inode(dquot);
617ba13b 4666 handle = ext4_journal_start(inode,
0b8e58a1 4667 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4668 if (IS_ERR(handle))
4669 return PTR_ERR(handle);
4670 ret = dquot_commit(dquot);
617ba13b 4671 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4672 if (!ret)
4673 ret = err;
4674 return ret;
4675}
4676
617ba13b 4677static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4678{
4679 int ret, err;
4680 handle_t *handle;
4681
617ba13b 4682 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4683 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4684 if (IS_ERR(handle))
4685 return PTR_ERR(handle);
4686 ret = dquot_acquire(dquot);
617ba13b 4687 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4688 if (!ret)
4689 ret = err;
4690 return ret;
4691}
4692
617ba13b 4693static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4694{
4695 int ret, err;
4696 handle_t *handle;
4697
617ba13b 4698 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4699 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4700 if (IS_ERR(handle)) {
4701 /* Release dquot anyway to avoid endless cycle in dqput() */
4702 dquot_release(dquot);
ac27a0ec 4703 return PTR_ERR(handle);
9c3013e9 4704 }
ac27a0ec 4705 ret = dquot_release(dquot);
617ba13b 4706 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4707 if (!ret)
4708 ret = err;
4709 return ret;
4710}
4711
617ba13b 4712static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4713{
2c8be6b2 4714 /* Are we journaling quotas? */
617ba13b
MC		 4715 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4716 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4717 dquot_mark_dquot_dirty(dquot);
617ba13b 4718 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4719 } else {
4720 return dquot_mark_dquot_dirty(dquot);
4721 }
4722}
4723
617ba13b 4724static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4725{
4726 int ret, err;
4727 handle_t *handle;
4728
4729 /* Data block + inode block */
617ba13b 4730 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4731 if (IS_ERR(handle))
4732 return PTR_ERR(handle);
4733 ret = dquot_commit_info(sb, type);
617ba13b 4734 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4735 if (!ret)
4736 ret = err;
4737 return ret;
4738}
4739
4740/*
4741 * Turn on quotas during mount time - we need to find
4742 * the quota file and such...
4743 */
617ba13b 4744static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4745{
287a8095
CH		 4746 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4747 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4748}
4749
4750/*
4751 * Standard function to be called on quota_on
4752 */
617ba13b 4753static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4754 struct path *path)
ac27a0ec
DK		 4755{
4756 int err;
ac27a0ec
DK		 4757
4758 if (!test_opt(sb, QUOTA))
4759 return -EINVAL;
0623543b 4760
ac27a0ec 4761 /* Quotafile not on the same filesystem? */
d8c9584e 4762 if (path->dentry->d_sb != sb)
ac27a0ec 4763 return -EXDEV;
0623543b
JK		 4764 /* Journaling quota? */
4765 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4766 /* Quotafile not in fs root? */
f00c9e44 4767 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4768 ext4_msg(sb, KERN_WARNING,
4769 "Quota file not on filesystem root. "
4770 "Journaled quota will not work");
2b2d6d01 4771 }
0623543b
JK		 4772
4773 /*
4774 * When we journal data on quota file, we have to flush journal to see
4775 * all updates to the file when we bypass pagecache...
4776 */
0390131b 4777 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4778 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4779 /*
4780 * We don't need to lock updates but journal_flush() could
4781 * otherwise be livelocked...
4782 */
4783 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4784 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4785 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4786 if (err)
7ffe1ea8 4787 return err;
0623543b
JK		 4788 }
4789
f00c9e44 4790 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4791}
4792
ca0e05e4
DM		 4793static int ext4_quota_off(struct super_block *sb, int type)
4794{
21f97697
JK		 4795 struct inode *inode = sb_dqopt(sb)->files[type];
4796 handle_t *handle;
4797
87009d86
DM		 4798 /* Force all delayed allocation blocks to be allocated.
4799 * Caller already holds s_umount sem */
4800 if (test_opt(sb, DELALLOC))
ca0e05e4 4801 sync_filesystem(sb);
ca0e05e4 4802
0b268590
AG		 4803 if (!inode)
4804 goto out;
4805
21f97697
JK		 4806 /* Update modification times of quota files when userspace can
4807 * start looking at them */
4808 handle = ext4_journal_start(inode, 1);
4809 if (IS_ERR(handle))
4810 goto out;
4811 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4812 ext4_mark_inode_dirty(handle, inode);
4813 ext4_journal_stop(handle);
4814
4815out:
ca0e05e4
DM		 4816 return dquot_quota_off(sb, type);
4817}
4818
ac27a0ec
DK		 4819/* Read data from quotafile - avoid pagecache and such because we cannot afford
4820 * acquiring the locks... As quota files are never truncated and quota code
25985edc 4821 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 4822 * we don't have to be afraid of races */
617ba13b 4823static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4824 size_t len, loff_t off)
4825{
4826 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4827 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4828 int err = 0;
4829 int offset = off & (sb->s_blocksize - 1);
4830 int tocopy;
4831 size_t toread;
4832 struct buffer_head *bh;
4833 loff_t i_size = i_size_read(inode);
4834
4835 if (off > i_size)
4836 return 0;
4837 if (off+len > i_size)
4838 len = i_size-off;
4839 toread = len;
4840 while (toread > 0) {
4841 tocopy = sb->s_blocksize - offset < toread ?
4842 sb->s_blocksize - offset : toread;
617ba13b 4843 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4844 if (err)
4845 return err;
4846 if (!bh) /* A hole? */
4847 memset(data, 0, tocopy);
4848 else
4849 memcpy(data, bh->b_data+offset, tocopy);
4850 brelse(bh);
4851 offset = 0;
4852 toread -= tocopy;
4853 data += tocopy;
4854 blk++;
4855 }
4856 return len;
4857}
4858
4859/* Write to quotafile (we know the transaction is already started and has
4860 * enough credits) */
617ba13b 4861static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4862 const char *data, size_t len, loff_t off)
4863{
4864 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4865 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4866 int err = 0;
4867 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4868 struct buffer_head *bh;
4869 handle_t *handle = journal_current_handle();
4870
0390131b 4871 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4872 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4873 " cancelled because transaction is not started",
9c3013e9
JK		 4874 (unsigned long long)off, (unsigned long long)len);
4875 return -EIO;
4876 }
67eeb568
DM		 4877 /*
4878 * Since we account only one data block in transaction credits,
4879 * then it is impossible to cross a block boundary.
4880 */
4881 if (sb->s_blocksize - offset < len) {
4882 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4883 " cancelled because not block aligned",
4884 (unsigned long long)off, (unsigned long long)len);
4885 return -EIO;
4886 }
4887
ac27a0ec 4888 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4889 bh = ext4_bread(handle, inode, blk, 1, &err);
4890 if (!bh)
4891 goto out;
62d2b5f2
JK		 4892 err = ext4_journal_get_write_access(handle, bh);
4893 if (err) {
4894 brelse(bh);
4895 goto out;
ac27a0ec 4896 }
67eeb568
DM		 4897 lock_buffer(bh);
4898 memcpy(bh->b_data+offset, data, len);
4899 flush_dcache_page(bh->b_page);
4900 unlock_buffer(bh);
62d2b5f2 4901 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4902 brelse(bh);
ac27a0ec 4903out:
67eeb568 4904 if (err) {
4d04e4fb 4905 mutex_unlock(&inode->i_mutex);
ac27a0ec 4906 return err;
4d04e4fb 4907 }
67eeb568
DM		 4908 if (inode->i_size < off + len) {
4909 i_size_write(inode, off + len);
617ba13b 4910 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 4911 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4912 }
ac27a0ec 4913 mutex_unlock(&inode->i_mutex);
67eeb568 4914 return len;
ac27a0ec
DK		 4915}
4916
4917#endif
4918
152a0836
AV		 4919static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4920 const char *dev_name, void *data)
ac27a0ec 4921{
152a0836 4922 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4923}
4924
37f328eb 4925#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4926static inline void register_as_ext2(void)
4927{
4928 int err = register_filesystem(&ext2_fs_type);
4929 if (err)
4930 printk(KERN_WARNING
4931 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4932}
4933
4934static inline void unregister_as_ext2(void)
4935{
4936 unregister_filesystem(&ext2_fs_type);
4937}
2035e776
TT		 4938
4939static inline int ext2_feature_set_ok(struct super_block *sb)
4940{
4941 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4942 return 0;
4943 if (sb->s_flags & MS_RDONLY)
4944 return 1;
4945 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4946 return 0;
4947 return 1;
4948}
51b7e3c9 4949MODULE_ALIAS("ext2");
24b58424
TT		 4950#else
4951static inline void register_as_ext2(void) { }
4952static inline void unregister_as_ext2(void) { }
2035e776 4953static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4954#endif
4955
37f328eb 4956#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4957static inline void register_as_ext3(void)
4958{
4959 int err = register_filesystem(&ext3_fs_type);
4960 if (err)
4961 printk(KERN_WARNING
4962 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4963}
4964
4965static inline void unregister_as_ext3(void)
4966{
4967 unregister_filesystem(&ext3_fs_type);
4968}
2035e776
TT		 4969
4970static inline int ext3_feature_set_ok(struct super_block *sb)
4971{
4972 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4973 return 0;
4974 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4975 return 0;
4976 if (sb->s_flags & MS_RDONLY)
4977 return 1;
4978 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4979 return 0;
4980 return 1;
4981}
51b7e3c9 4982MODULE_ALIAS("ext3");
24b58424
TT		 4983#else
4984static inline void register_as_ext3(void) { }
4985static inline void unregister_as_ext3(void) { }
2035e776 4986static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4987#endif
4988
03010a33
TT		 4989static struct file_system_type ext4_fs_type = {
4990 .owner = THIS_MODULE,
4991 .name = "ext4",
152a0836 4992 .mount = ext4_mount,
03010a33
TT		 4993 .kill_sb = kill_block_super,
4994 .fs_flags = FS_REQUIRES_DEV,
4995};
4996
8f021222 4997static int __init ext4_init_feat_adverts(void)
857ac889
LC		 4998{
4999 struct ext4_features *ef;
5000 int ret = -ENOMEM;
5001
5002 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5003 if (!ef)
5004 goto out;
5005
5006 ef->f_kobj.kset = ext4_kset;
5007 init_completion(&ef->f_kobj_unregister);
5008 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5009 "features");
5010 if (ret) {
5011 kfree(ef);
5012 goto out;
5013 }
5014
5015 ext4_feat = ef;
5016 ret = 0;
5017out:
5018 return ret;
5019}
5020
8f021222
LC		 5021static void ext4_exit_feat_adverts(void)
5022{
5023 kobject_put(&ext4_feat->f_kobj);
5024 wait_for_completion(&ext4_feat->f_kobj_unregister);
5025 kfree(ext4_feat);
5026}
5027
e9e3bcec
ES		 5028/* Shared across all ext4 file systems */
5029wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5030struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5031
5dabfc78 5032static int __init ext4_init_fs(void)
ac27a0ec 5033{
e9e3bcec 5034 int i, err;
c9de560d 5035
12e9b892 5036 ext4_check_flag_values();
e9e3bcec
ES		 5037
5038 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5039 mutex_init(&ext4__aio_mutex[i]);
5040 init_waitqueue_head(&ext4__ioend_wq[i]);
5041 }
5042
5dabfc78 5043 err = ext4_init_pageio();
6fd058f7
TT		 5044 if (err)
5045 return err;
5dabfc78 5046 err = ext4_init_system_zone();
bd2d0210 5047 if (err)
d44651d0 5048 goto out6;
3197ebdb
TT		 5049 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5050 if (!ext4_kset)
dd68314c 5051 goto out5;
d44651d0 5052 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5053
5054 err = ext4_init_feat_adverts();
dd68314c
TT		 5055 if (err)
5056 goto out4;
857ac889 5057
5dabfc78 5058 err = ext4_init_mballoc();
ac27a0ec 5059 if (err)
6fd058f7 5060 goto out3;
c9de560d 5061
5dabfc78 5062 err = ext4_init_xattr();
c9de560d
AT		 5063 if (err)
5064 goto out2;
ac27a0ec
DK		 5065 err = init_inodecache();
5066 if (err)
5067 goto out1;
24b58424 5068 register_as_ext3();
2035e776 5069 register_as_ext2();
03010a33 5070 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5071 if (err)
5072 goto out;
bfff6873
LC		 5073
5074 ext4_li_info = NULL;
5075 mutex_init(&ext4_li_mtx);
ac27a0ec
DK		 5076 return 0;
5077out:
24b58424
TT		 5078 unregister_as_ext2();
5079 unregister_as_ext3();
ac27a0ec
DK		 5080 destroy_inodecache();
5081out1:
5dabfc78 5082 ext4_exit_xattr();
c9de560d 5083out2:
5dabfc78 5084 ext4_exit_mballoc();
6fd058f7 5085out3:
8f021222 5086 ext4_exit_feat_adverts();
dd68314c 5087out4:
d44651d0
FJ		 5088 if (ext4_proc_root)
5089 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5090 kset_unregister(ext4_kset);
d44651d0 5091out5:
5dabfc78 5092 ext4_exit_system_zone();
d44651d0 5093out6:
5dabfc78 5094 ext4_exit_pageio();
ac27a0ec
DK		 5095 return err;
5096}
5097
5dabfc78 5098static void __exit ext4_exit_fs(void)
ac27a0ec 5099{
bfff6873 5100 ext4_destroy_lazyinit_thread();
24b58424
TT		 5101 unregister_as_ext2();
5102 unregister_as_ext3();
03010a33 5103 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5104 destroy_inodecache();
5dabfc78
TT		 5105 ext4_exit_xattr();
5106 ext4_exit_mballoc();
8f021222 5107 ext4_exit_feat_adverts();
9f6200bb 5108 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5109 kset_unregister(ext4_kset);
5dabfc78
TT		 5110 ext4_exit_system_zone();
5111 ext4_exit_pageio();
ac27a0ec
DK		 5112}
5113
5114MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5115MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5116MODULE_LICENSE("GPL");
5dabfc78
TT		 5117module_init(ext4_init_fs)
5118module_exit(ext4_exit_fs)
Ubuntu Artful kernel mirror