]> 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
fs: Protect write paths by sb_start_write - sb_end_write
[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);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 72static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 73 char nbuf[16]);
2b2d6d01
TT		 74static int ext4_remount(struct super_block *sb, int *flags, char *data);
75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 76static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 77static void ext4_write_super(struct super_block *sb);
c4be0c1d 78static int ext4_freeze(struct super_block *sb);
152a0836
AV		 79static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 const char *dev_name, void *data);
2035e776
TT		 81static inline int ext2_feature_set_ok(struct super_block *sb);
82static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 83static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 84static void ext4_destroy_lazyinit_thread(void);
85static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 86static void ext4_clear_request_list(void);
ac27a0ec 87
2035e776
TT		 88#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89static struct file_system_type ext2_fs_type = {
90 .owner = THIS_MODULE,
91 .name = "ext2",
92 .mount = ext4_mount,
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
95};
96#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97#else
98#define IS_EXT2_SB(sb) (0)
99#endif
100
101
ba69f9ab
JK		 102#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
152a0836 106 .mount = ext4_mount,
ba69f9ab
JK		 107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109};
110#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111#else
112#define IS_EXT3_SB(sb) (0)
113#endif
bd81d8ee 114
d25425f8
DW		 115static int ext4_verify_csum_type(struct super_block *sb,
116 struct ext4_super_block *es)
117{
118 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
119 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
120 return 1;
121
122 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
123}
124
a9c47317
DW		 125static __le32 ext4_superblock_csum(struct super_block *sb,
126 struct ext4_super_block *es)
127{
128 struct ext4_sb_info *sbi = EXT4_SB(sb);
129 int offset = offsetof(struct ext4_super_block, s_checksum);
130 __u32 csum;
131
132 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
133
134 return cpu_to_le32(csum);
135}
136
137int ext4_superblock_csum_verify(struct super_block *sb,
138 struct ext4_super_block *es)
139{
140 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
141 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
142 return 1;
143
144 return es->s_checksum == ext4_superblock_csum(sb, es);
145}
146
147void ext4_superblock_csum_set(struct super_block *sb,
148 struct ext4_super_block *es)
149{
150 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
151 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
152 return;
153
154 es->s_checksum = ext4_superblock_csum(sb, es);
155}
156
9933fc0a
TT		 157void *ext4_kvmalloc(size_t size, gfp_t flags)
158{
159 void *ret;
160
161 ret = kmalloc(size, flags);
162 if (!ret)
163 ret = __vmalloc(size, flags, PAGE_KERNEL);
164 return ret;
165}
166
167void *ext4_kvzalloc(size_t size, gfp_t flags)
168{
169 void *ret;
170
db9481c0 171 ret = kzalloc(size, flags);
9933fc0a
TT		 172 if (!ret)
173 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
174 return ret;
175}
176
177void ext4_kvfree(void *ptr)
178{
179 if (is_vmalloc_addr(ptr))
180 vfree(ptr);
181 else
182 kfree(ptr);
183
184}
185
8fadc143
AR		 186ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
bd81d8ee 188{
3a14589c 189 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 191 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 192}
193
8fadc143
AR		 194ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
195 struct ext4_group_desc *bg)
bd81d8ee 196{
5272f837 197 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 200}
201
8fadc143
AR		 202ext4_fsblk_t ext4_inode_table(struct super_block *sb,
203 struct ext4_group_desc *bg)
bd81d8ee 204{
5272f837 205 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 207 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 208}
209
021b65bb
TT		 210__u32 ext4_free_group_clusters(struct super_block *sb,
211 struct ext4_group_desc *bg)
560671a0
AK		 212{
213 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 215 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 216}
217
218__u32 ext4_free_inodes_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220{
221 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 223 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 224}
225
226__u32 ext4_used_dirs_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
228{
229 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 231 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 232}
233
234__u32 ext4_itable_unused_count(struct super_block *sb,
235 struct ext4_group_desc *bg)
236{
237 return le16_to_cpu(bg->bg_itable_unused_lo) |
238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 239 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 240}
241
8fadc143
AR		 242void ext4_block_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 244{
3a14589c 245 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 248}
249
8fadc143
AR		 250void ext4_inode_bitmap_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 252{
5272f837 253 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 256}
257
8fadc143
AR		 258void ext4_inode_table_set(struct super_block *sb,
259 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 260{
5272f837 261 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 264}
265
021b65bb
TT		 266void ext4_free_group_clusters_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 268{
269 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
272}
273
274void ext4_free_inodes_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276{
277 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
280}
281
282void ext4_used_dirs_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284{
285 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
288}
289
290void ext4_itable_unused_set(struct super_block *sb,
291 struct ext4_group_desc *bg, __u32 count)
292{
293 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
296}
297
d3d1faf6
CW		 298
299/* Just increment the non-pointer handle value */
300static handle_t *ext4_get_nojournal(void)
301{
302 handle_t *handle = current->journal_info;
303 unsigned long ref_cnt = (unsigned long)handle;
304
305 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
306
307 ref_cnt++;
308 handle = (handle_t *)ref_cnt;
309
310 current->journal_info = handle;
311 return handle;
312}
313
314
315/* Decrement the non-pointer handle value */
316static void ext4_put_nojournal(handle_t *handle)
317{
318 unsigned long ref_cnt = (unsigned long)handle;
319
320 BUG_ON(ref_cnt == 0);
321
322 ref_cnt--;
323 handle = (handle_t *)ref_cnt;
324
325 current->journal_info = handle;
326}
327
ac27a0ec 328/*
dab291af 329 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 330 *
331 * The only special thing we need to do here is to make sure that all
332 * journal_end calls result in the superblock being marked dirty, so
333 * that sync() will call the filesystem's write_super callback if
334 * appropriate.
be4f27d3
YY		 335 *
336 * To avoid j_barrier hold in userspace when a user calls freeze(),
337 * ext4 prevents a new handle from being started by s_frozen, which
338 * is in an upper layer.
ac27a0ec 339 */
617ba13b 340handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 341{
342 journal_t *journal;
be4f27d3 343 handle_t *handle;
ac27a0ec 344
12706394 345 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 346 if (sb->s_flags & MS_RDONLY)
347 return ERR_PTR(-EROFS);
348
617ba13b 349 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 350 handle = ext4_journal_current_handle();
351
352 /*
353 * If a handle has been started, it should be allowed to
354 * finish, otherwise deadlock could happen between freeze
355 * and others(e.g. truncate) due to the restart of the
356 * journal handle if the filesystem is forzen and active
357 * handles are not stopped.
358 */
359 if (!handle)
360 vfs_check_frozen(sb, SB_FREEZE_TRANS);
361
362 if (!journal)
363 return ext4_get_nojournal();
364 /*
365 * Special case here: if the journal has aborted behind our
366 * backs (eg. EIO in the commit thread), then we still need to
367 * take the FS itself readonly cleanly.
368 */
369 if (is_journal_aborted(journal)) {
370 ext4_abort(sb, "Detected aborted journal");
371 return ERR_PTR(-EROFS);
ac27a0ec 372 }
be4f27d3 373 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 374}
375
376/*
377 * The only special thing we need to do here is to make sure that all
dab291af 378 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 379 * that sync() will call the filesystem's write_super callback if
380 * appropriate.
381 */
c398eda0 382int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 383{
384 struct super_block *sb;
385 int err;
386 int rc;
387
0390131b 388 if (!ext4_handle_valid(handle)) {
d3d1faf6 389 ext4_put_nojournal(handle);
0390131b
FM		 390 return 0;
391 }
ac27a0ec
DK		 392 sb = handle->h_transaction->t_journal->j_private;
393 err = handle->h_err;
dab291af 394 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 395
396 if (!err)
397 err = rc;
398 if (err)
c398eda0 399 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 400 return err;
401}
402
90c7201b
TT		 403void ext4_journal_abort_handle(const char *caller, unsigned int line,
404 const char *err_fn, struct buffer_head *bh,
405 handle_t *handle, int err)
ac27a0ec
DK		 406{
407 char nbuf[16];
617ba13b 408 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 409
0390131b
FM		 410 BUG_ON(!ext4_handle_valid(handle));
411
ac27a0ec
DK		 412 if (bh)
413 BUFFER_TRACE(bh, "abort");
414
415 if (!handle->h_err)
416 handle->h_err = err;
417
418 if (is_handle_aborted(handle))
419 return;
420
92b97816 421 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
90c7201b 422 caller, line, errstr, err_fn);
ac27a0ec 423
dab291af 424 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 425}
426
1c13d5c0
TT		 427static void __save_error_info(struct super_block *sb, const char *func,
428 unsigned int line)
429{
430 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
431
432 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
433 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
434 es->s_last_error_time = cpu_to_le32(get_seconds());
435 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
436 es->s_last_error_line = cpu_to_le32(line);
437 if (!es->s_first_error_time) {
438 es->s_first_error_time = es->s_last_error_time;
439 strncpy(es->s_first_error_func, func,
440 sizeof(es->s_first_error_func));
441 es->s_first_error_line = cpu_to_le32(line);
442 es->s_first_error_ino = es->s_last_error_ino;
443 es->s_first_error_block = es->s_last_error_block;
444 }
66e61a9e
TT		 445 /*
446 * Start the daily error reporting function if it hasn't been
447 * started already
448 */
449 if (!es->s_error_count)
450 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 451 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
452}
453
454static void save_error_info(struct super_block *sb, const char *func,
455 unsigned int line)
456{
457 __save_error_info(sb, func, line);
458 ext4_commit_super(sb, 1);
459}
460
7c2e7087
TT		 461/*
462 * The del_gendisk() function uninitializes the disk-specific data
463 * structures, including the bdi structure, without telling anyone
464 * else. Once this happens, any attempt to call mark_buffer_dirty()
465 * (for example, by ext4_commit_super), will cause a kernel OOPS.
466 * This is a kludge to prevent these oops until we can put in a proper
467 * hook in del_gendisk() to inform the VFS and file system layers.
468 */
469static int block_device_ejected(struct super_block *sb)
470{
471 struct inode *bd_inode = sb->s_bdev->bd_inode;
472 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
473
474 return bdi->dev == NULL;
475}
476
18aadd47
BJ		 477static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
478{
479 struct super_block *sb = journal->j_private;
480 struct ext4_sb_info *sbi = EXT4_SB(sb);
481 int error = is_journal_aborted(journal);
482 struct ext4_journal_cb_entry *jce, *tmp;
483
484 spin_lock(&sbi->s_md_lock);
485 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
486 list_del_init(&jce->jce_list);
487 spin_unlock(&sbi->s_md_lock);
488 jce->jce_func(sb, jce, error);
489 spin_lock(&sbi->s_md_lock);
490 }
491 spin_unlock(&sbi->s_md_lock);
492}
1c13d5c0 493
ac27a0ec
DK		 494/* Deal with the reporting of failure conditions on a filesystem such as
495 * inconsistencies detected or read IO failures.
496 *
497 * On ext2, we can store the error state of the filesystem in the
617ba13b 498 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 499 * write ordering constraints on the superblock which prevent us from
500 * writing it out straight away; and given that the journal is about to
501 * be aborted, we can't rely on the current, or future, transactions to
502 * write out the superblock safely.
503 *
dab291af 504 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 505 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 506 * that error until we've noted it down and cleared it.
507 */
508
617ba13b 509static void ext4_handle_error(struct super_block *sb)
ac27a0ec 510{
ac27a0ec
DK		 511 if (sb->s_flags & MS_RDONLY)
512 return;
513
2b2d6d01 514 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 515 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 516
4ab2f15b 517 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 518 if (journal)
dab291af 519 jbd2_journal_abort(journal, -EIO);
ac27a0ec 520 }
2b2d6d01 521 if (test_opt(sb, ERRORS_RO)) {
b31e1552 522 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 523 sb->s_flags |= MS_RDONLY;
524 }
ac27a0ec 525 if (test_opt(sb, ERRORS_PANIC))
617ba13b 526 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 527 sb->s_id);
528}
529
12062ddd 530void __ext4_error(struct super_block *sb, const char *function,
c398eda0 531 unsigned int line, const char *fmt, ...)
ac27a0ec 532{
0ff2ea7d 533 struct va_format vaf;
ac27a0ec
DK		 534 va_list args;
535
536 va_start(args, fmt);
0ff2ea7d
JP		 537 vaf.fmt = fmt;
538 vaf.va = &args;
539 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
540 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec 541 va_end(args);
f3fc0210 542 save_error_info(sb, function, line);
ac27a0ec 543
617ba13b 544 ext4_handle_error(sb);
ac27a0ec
DK		 545}
546
c398eda0
TT		 547void ext4_error_inode(struct inode *inode, const char *function,
548 unsigned int line, ext4_fsblk_t block,
273df556
FM		 549 const char *fmt, ...)
550{
551 va_list args;
f7c21177 552 struct va_format vaf;
1c13d5c0 553 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 554
1c13d5c0
TT		 555 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
556 es->s_last_error_block = cpu_to_le64(block);
557 save_error_info(inode->i_sb, function, line);
273df556 558 va_start(args, fmt);
f7c21177
TT		 559 vaf.fmt = fmt;
560 vaf.va = &args;
c398eda0 561 if (block)
d9ee81da
JP		 562 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
563 "inode #%lu: block %llu: comm %s: %pV\n",
564 inode->i_sb->s_id, function, line, inode->i_ino,
565 block, current->comm, &vaf);
566 else
567 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
568 "inode #%lu: comm %s: %pV\n",
569 inode->i_sb->s_id, function, line, inode->i_ino,
570 current->comm, &vaf);
273df556
FM		 571 va_end(args);
572
573 ext4_handle_error(inode->i_sb);
574}
575
c398eda0 576void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 577 unsigned int line, ext4_fsblk_t block,
578 const char *fmt, ...)
273df556
FM		 579{
580 va_list args;
f7c21177 581 struct va_format vaf;
1c13d5c0 582 struct ext4_super_block *es;
273df556
FM		 583 struct inode *inode = file->f_dentry->d_inode;
584 char pathname[80], *path;
585
1c13d5c0
TT		 586 es = EXT4_SB(inode->i_sb)->s_es;
587 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
588 save_error_info(inode->i_sb, function, line);
273df556 589 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 590 if (IS_ERR(path))
273df556 591 path = "(unknown)";
f7c21177
TT		 592 va_start(args, fmt);
593 vaf.fmt = fmt;
594 vaf.va = &args;
d9ee81da
JP		 595 if (block)
596 printk(KERN_CRIT
597 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
598 "block %llu: comm %s: path %s: %pV\n",
599 inode->i_sb->s_id, function, line, inode->i_ino,
600 block, current->comm, path, &vaf);
601 else
602 printk(KERN_CRIT
603 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
604 "comm %s: path %s: %pV\n",
605 inode->i_sb->s_id, function, line, inode->i_ino,
606 current->comm, path, &vaf);
273df556
FM		 607 va_end(args);
608
609 ext4_handle_error(inode->i_sb);
610}
611
2b2d6d01 612static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 613 char nbuf[16])
614{
615 char *errstr = NULL;
616
617 switch (errno) {
618 case -EIO:
619 errstr = "IO failure";
620 break;
621 case -ENOMEM:
622 errstr = "Out of memory";
623 break;
624 case -EROFS:
78f1ddbb
TT		 625 if (!sb || (EXT4_SB(sb)->s_journal &&
626 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 627 errstr = "Journal has aborted";
628 else
629 errstr = "Readonly filesystem";
630 break;
631 default:
632 /* If the caller passed in an extra buffer for unknown
633 * errors, textualise them now. Else we just return
634 * NULL. */
635 if (nbuf) {
636 /* Check for truncated error codes... */
637 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
638 errstr = nbuf;
639 }
640 break;
641 }
642
643 return errstr;
644}
645
617ba13b 646/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 647 * automatically and invokes the appropriate error response. */
648
c398eda0
TT		 649void __ext4_std_error(struct super_block *sb, const char *function,
650 unsigned int line, int errno)
ac27a0ec
DK		 651{
652 char nbuf[16];
653 const char *errstr;
654
655 /* Special case: if the error is EROFS, and we're not already
656 * inside a transaction, then there's really no point in logging
657 * an error. */
658 if (errno == -EROFS && journal_current_handle() == NULL &&
659 (sb->s_flags & MS_RDONLY))
660 return;
661
617ba13b 662 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 663 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
664 sb->s_id, function, line, errstr);
1c13d5c0 665 save_error_info(sb, function, line);
ac27a0ec 666
617ba13b 667 ext4_handle_error(sb);
ac27a0ec
DK		 668}
669
670/*
617ba13b 671 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 672 * abort function may be used to deal with unrecoverable failures such
673 * as journal IO errors or ENOMEM at a critical moment in log management.
674 *
675 * We unconditionally force the filesystem into an ABORT|READONLY state,
676 * unless the error response on the fs has been set to panic in which
677 * case we take the easy way out and panic immediately.
678 */
679
c67d859e 680void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 681 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 682{
683 va_list args;
684
1c13d5c0 685 save_error_info(sb, function, line);
ac27a0ec 686 va_start(args, fmt);
c398eda0
TT		 687 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
688 function, line);
ac27a0ec
DK		 689 vprintk(fmt, args);
690 printk("\n");
691 va_end(args);
692
1c13d5c0
TT		 693 if ((sb->s_flags & MS_RDONLY) == 0) {
694 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
695 sb->s_flags |= MS_RDONLY;
696 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
697 if (EXT4_SB(sb)->s_journal)
698 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
699 save_error_info(sb, function, line);
700 }
ac27a0ec 701 if (test_opt(sb, ERRORS_PANIC))
617ba13b 702 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 703}
704
0ff2ea7d 705void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 706{
0ff2ea7d 707 struct va_format vaf;
b31e1552
ES		 708 va_list args;
709
710 va_start(args, fmt);
0ff2ea7d
JP		 711 vaf.fmt = fmt;
712 vaf.va = &args;
713 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 714 va_end(args);
715}
716
12062ddd 717void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 718 unsigned int line, const char *fmt, ...)
ac27a0ec 719{
0ff2ea7d 720 struct va_format vaf;
ac27a0ec
DK		 721 va_list args;
722
723 va_start(args, fmt);
0ff2ea7d
JP		 724 vaf.fmt = fmt;
725 vaf.va = &args;
726 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
727 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 728 va_end(args);
729}
730
e29136f8
TT		 731void __ext4_grp_locked_error(const char *function, unsigned int line,
732 struct super_block *sb, ext4_group_t grp,
733 unsigned long ino, ext4_fsblk_t block,
734 const char *fmt, ...)
5d1b1b3f
AK		 735__releases(bitlock)
736__acquires(bitlock)
737{
0ff2ea7d 738 struct va_format vaf;
5d1b1b3f
AK		 739 va_list args;
740 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
741
1c13d5c0
TT		 742 es->s_last_error_ino = cpu_to_le32(ino);
743 es->s_last_error_block = cpu_to_le64(block);
744 __save_error_info(sb, function, line);
0ff2ea7d 745
5d1b1b3f 746 va_start(args, fmt);
0ff2ea7d
JP		 747
748 vaf.fmt = fmt;
749 vaf.va = &args;
21149d61 750 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 751 sb->s_id, function, line, grp);
752 if (ino)
0ff2ea7d 753 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 754 if (block)
0ff2ea7d
JP		 755 printk(KERN_CONT "block %llu:", (unsigned long long) block);
756 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 757 va_end(args);
758
759 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 760 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 761 return;
762 }
1c13d5c0 763
5d1b1b3f
AK		 764 ext4_unlock_group(sb, grp);
765 ext4_handle_error(sb);
766 /*
767 * We only get here in the ERRORS_RO case; relocking the group
768 * may be dangerous, but nothing bad will happen since the
769 * filesystem will have already been marked read/only and the
770 * journal has been aborted. We return 1 as a hint to callers
771 * who might what to use the return value from
25985edc 772 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 773 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
774 * aggressively from the ext4 function in question, with a
775 * more appropriate error code.
776 */
777 ext4_lock_group(sb, grp);
778 return;
779}
780
617ba13b 781void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 782{
617ba13b 783 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 784
617ba13b 785 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 786 return;
787
12062ddd 788 ext4_warning(sb,
ac27a0ec
DK		 789 "updating to rev %d because of new feature flag, "
790 "running e2fsck is recommended",
617ba13b 791 EXT4_DYNAMIC_REV);
ac27a0ec 792
617ba13b
MC		 793 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
794 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
795 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 796 /* leave es->s_feature_*compat flags alone */
797 /* es->s_uuid will be set by e2fsck if empty */
798
799 /*
800 * The rest of the superblock fields should be zero, and if not it
801 * means they are likely already in use, so leave them alone. We
802 * can leave it up to e2fsck to clean up any inconsistencies there.
803 */
804}
805
806/*
807 * Open the external journal device
808 */
b31e1552 809static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 810{
811 struct block_device *bdev;
812 char b[BDEVNAME_SIZE];
813
d4d77629 814 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 815 if (IS_ERR(bdev))
816 goto fail;
817 return bdev;
818
819fail:
b31e1552 820 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 821 __bdevname(dev, b), PTR_ERR(bdev));
822 return NULL;
823}
824
825/*
826 * Release the journal device
827 */
617ba13b 828static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 829{
e525fd89 830 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 831}
832
617ba13b 833static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 834{
835 struct block_device *bdev;
836 int ret = -ENODEV;
837
838 bdev = sbi->journal_bdev;
839 if (bdev) {
617ba13b 840 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 841 sbi->journal_bdev = NULL;
842 }
843 return ret;
844}
845
846static inline struct inode *orphan_list_entry(struct list_head *l)
847{
617ba13b 848 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 849}
850
617ba13b 851static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 852{
853 struct list_head *l;
854
b31e1552
ES		 855 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
856 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 857
858 printk(KERN_ERR "sb_info orphan list:\n");
859 list_for_each(l, &sbi->s_orphan) {
860 struct inode *inode = orphan_list_entry(l);
861 printk(KERN_ERR " "
862 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
863 inode->i_sb->s_id, inode->i_ino, inode,
864 inode->i_mode, inode->i_nlink,
865 NEXT_ORPHAN(inode));
866 }
867}
868
2b2d6d01 869static void ext4_put_super(struct super_block *sb)
ac27a0ec 870{
617ba13b
MC		 871 struct ext4_sb_info *sbi = EXT4_SB(sb);
872 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 873 int i, err;
ac27a0ec 874
857ac889 875 ext4_unregister_li_request(sb);
e0ccfd95
CH		 876 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
877
4c0425ff
MC		 878 flush_workqueue(sbi->dio_unwritten_wq);
879 destroy_workqueue(sbi->dio_unwritten_wq);
880
a9e220f8 881 lock_super(sb);
0390131b
FM		 882 if (sbi->s_journal) {
883 err = jbd2_journal_destroy(sbi->s_journal);
884 sbi->s_journal = NULL;
885 if (err < 0)
c67d859e 886 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 887 }
d4edac31 888
a1c6c569 889 del_timer(&sbi->s_err_report);
d4edac31
JB		 890 ext4_release_system_zone(sb);
891 ext4_mb_release(sb);
892 ext4_ext_release(sb);
893 ext4_xattr_put_super(sb);
894
ac27a0ec 895 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 896 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 897 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 898 }
a8e25a83
AB		 899 if (sb->s_dirt || !(sb->s_flags & MS_RDONLY))
900 ext4_commit_super(sb, 1);
901
240799cd 902 if (sbi->s_proc) {
66acdcf4 903 remove_proc_entry("options", sbi->s_proc);
9f6200bb 904 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 905 }
3197ebdb 906 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 907
908 for (i = 0; i < sbi->s_gdb_count; i++)
909 brelse(sbi->s_group_desc[i]);
f18a5f21 910 ext4_kvfree(sbi->s_group_desc);
9933fc0a 911 ext4_kvfree(sbi->s_flex_groups);
57042651 912 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 913 percpu_counter_destroy(&sbi->s_freeinodes_counter);
914 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 915 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 916 brelse(sbi->s_sbh);
917#ifdef CONFIG_QUOTA
918 for (i = 0; i < MAXQUOTAS; i++)
919 kfree(sbi->s_qf_names[i]);
920#endif
921
922 /* Debugging code just in case the in-memory inode orphan list
923 * isn't empty. The on-disk one can be non-empty if we've
924 * detected an error and taken the fs readonly, but the
925 * in-memory list had better be clean by this point. */
926 if (!list_empty(&sbi->s_orphan))
927 dump_orphan_list(sb, sbi);
928 J_ASSERT(list_empty(&sbi->s_orphan));
929
f98393a6 930 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 931 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
932 /*
933 * Invalidate the journal device's buffers. We don't want them
934 * floating about in memory - the physical journal device may
935 * hotswapped, and it breaks the `ro-after' testing code.
936 */
937 sync_blockdev(sbi->journal_bdev);
f98393a6 938 invalidate_bdev(sbi->journal_bdev);
617ba13b 939 ext4_blkdev_remove(sbi);
ac27a0ec 940 }
c5e06d10
JL		 941 if (sbi->s_mmp_tsk)
942 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 943 sb->s_fs_info = NULL;
3197ebdb
TT		 944 /*
945 * Now that we are completely done shutting down the
946 * superblock, we need to actually destroy the kobject.
947 */
3197ebdb
TT		 948 unlock_super(sb);
949 kobject_put(&sbi->s_kobj);
950 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 951 if (sbi->s_chksum_driver)
952 crypto_free_shash(sbi->s_chksum_driver);
705895b6 953 kfree(sbi->s_blockgroup_lock);
ac27a0ec 954 kfree(sbi);
ac27a0ec
DK		 955}
956
e18b890b 957static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 958
959/*
960 * Called inside transaction, so use GFP_NOFS
961 */
617ba13b 962static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 963{
617ba13b 964 struct ext4_inode_info *ei;
ac27a0ec 965
e6b4f8da 966 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 967 if (!ei)
968 return NULL;
0b8e58a1 969
ac27a0ec 970 ei->vfs_inode.i_version = 1;
91246c00 971 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 972 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 973 INIT_LIST_HEAD(&ei->i_prealloc_list);
974 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 975 ei->i_reserved_data_blocks = 0;
976 ei->i_reserved_meta_blocks = 0;
977 ei->i_allocated_meta_blocks = 0;
9d0be502 978 ei->i_da_metadata_calc_len = 0;
d2a17637 979 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 980#ifdef CONFIG_QUOTA
981 ei->i_reserved_quota = 0;
982#endif
8aefcd55 983 ei->jinode = NULL;
c7064ef1 984 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 985 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 986 ei->cur_aio_dio = NULL;
b436b9be
JK		 987 ei->i_sync_tid = 0;
988 ei->i_datasync_tid = 0;
f7ad6d2e 989 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 990 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 991
ac27a0ec
DK		 992 return &ei->vfs_inode;
993}
994
7ff9c073
TT		 995static int ext4_drop_inode(struct inode *inode)
996{
997 int drop = generic_drop_inode(inode);
998
999 trace_ext4_drop_inode(inode, drop);
1000 return drop;
1001}
1002
fa0d7e3d
NP		 1003static void ext4_i_callback(struct rcu_head *head)
1004{
1005 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 1006 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1007}
1008
617ba13b 1009static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1010{
9f7dd93d 1011 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1012 ext4_msg(inode->i_sb, KERN_ERR,
1013 "Inode %lu (%p): orphan list check failed!",
1014 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1015 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1016 EXT4_I(inode), sizeof(struct ext4_inode_info),
1017 true);
1018 dump_stack();
1019 }
fa0d7e3d 1020 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1021}
1022
51cc5068 1023static void init_once(void *foo)
ac27a0ec 1024{
617ba13b 1025 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1026
a35afb83 1027 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 1028#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 1029 init_rwsem(&ei->xattr_sem);
ac27a0ec 1030#endif
0e855ac8 1031 init_rwsem(&ei->i_data_sem);
a35afb83 1032 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1033}
1034
1035static int init_inodecache(void)
1036{
617ba13b
MC		 1037 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1038 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 1039 0, (SLAB_RECLAIM_ACCOUNT|
1040 SLAB_MEM_SPREAD),
20c2df83 1041 init_once);
617ba13b 1042 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1043 return -ENOMEM;
1044 return 0;
1045}
1046
1047static void destroy_inodecache(void)
1048{
617ba13b 1049 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1050}
1051
0930fcc1 1052void ext4_clear_inode(struct inode *inode)
ac27a0ec 1053{
0930fcc1 1054 invalidate_inode_buffers(inode);
dbd5768f 1055 clear_inode(inode);
9f754758 1056 dquot_drop(inode);
c2ea3fde 1057 ext4_discard_preallocations(inode);
8aefcd55
TT		 1058 if (EXT4_I(inode)->jinode) {
1059 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1060 EXT4_I(inode)->jinode);
1061 jbd2_free_inode(EXT4_I(inode)->jinode);
1062 EXT4_I(inode)->jinode = NULL;
1063 }
ac27a0ec
DK		 1064}
1065
1b961ac0 1066static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1067 u64 ino, u32 generation)
ac27a0ec 1068{
ac27a0ec 1069 struct inode *inode;
ac27a0ec 1070
617ba13b 1071 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1072 return ERR_PTR(-ESTALE);
617ba13b 1073 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1074 return ERR_PTR(-ESTALE);
1075
1076 /* iget isn't really right if the inode is currently unallocated!!
1077 *
617ba13b 1078 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1079 * deleted, so we should be safe.
1080 *
1081 * Currently we don't know the generation for parent directory, so
1082 * a generation of 0 means "accept any"
1083 */
1d1fe1ee
DH		 1084 inode = ext4_iget(sb, ino);
1085 if (IS_ERR(inode))
1086 return ERR_CAST(inode);
1087 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1088 iput(inode);
1089 return ERR_PTR(-ESTALE);
1090 }
1b961ac0
CH		 1091
1092 return inode;
1093}
1094
1095static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1096 int fh_len, int fh_type)
1b961ac0
CH		 1097{
1098 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1099 ext4_nfs_get_inode);
1100}
1101
1102static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1103 int fh_len, int fh_type)
1b961ac0
CH		 1104{
1105 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1106 ext4_nfs_get_inode);
ac27a0ec
DK		 1107}
1108
c39a7f84
TO		 1109/*
1110 * Try to release metadata pages (indirect blocks, directories) which are
1111 * mapped via the block device. Since these pages could have journal heads
1112 * which would prevent try_to_free_buffers() from freeing them, we must use
1113 * jbd2 layer's try_to_free_buffers() function to release them.
1114 */
0b8e58a1
AD		 1115static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1116 gfp_t wait)
c39a7f84
TO		 1117{
1118 journal_t *journal = EXT4_SB(sb)->s_journal;
1119
1120 WARN_ON(PageChecked(page));
1121 if (!page_has_buffers(page))
1122 return 0;
1123 if (journal)
1124 return jbd2_journal_try_to_free_buffers(journal, page,
1125 wait & ~__GFP_WAIT);
1126 return try_to_free_buffers(page);
1127}
1128
ac27a0ec 1129#ifdef CONFIG_QUOTA
af5bc92d 1130#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1131#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1132
617ba13b
MC		 1133static int ext4_write_dquot(struct dquot *dquot);
1134static int ext4_acquire_dquot(struct dquot *dquot);
1135static int ext4_release_dquot(struct dquot *dquot);
1136static int ext4_mark_dquot_dirty(struct dquot *dquot);
1137static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1138static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1139 struct path *path);
ca0e05e4 1140static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1141static int ext4_quota_on_mount(struct super_block *sb, int type);
1142static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1143 size_t len, loff_t off);
617ba13b 1144static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1145 const char *data, size_t len, loff_t off);
1146
61e225dc 1147static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1148 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1149 .write_dquot = ext4_write_dquot,
1150 .acquire_dquot = ext4_acquire_dquot,
1151 .release_dquot = ext4_release_dquot,
1152 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1153 .write_info = ext4_write_info,
1154 .alloc_dquot = dquot_alloc,
1155 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1156};
1157
0d54b217 1158static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1159 .quota_on = ext4_quota_on,
ca0e05e4 1160 .quota_off = ext4_quota_off,
287a8095
CH		 1161 .quota_sync = dquot_quota_sync,
1162 .get_info = dquot_get_dqinfo,
1163 .set_info = dquot_set_dqinfo,
1164 .get_dqblk = dquot_get_dqblk,
1165 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1166};
1167#endif
1168
ee9b6d61 1169static const struct super_operations ext4_sops = {
617ba13b
MC		 1170 .alloc_inode = ext4_alloc_inode,
1171 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1172 .write_inode = ext4_write_inode,
1173 .dirty_inode = ext4_dirty_inode,
7ff9c073 1174 .drop_inode = ext4_drop_inode,
0930fcc1 1175 .evict_inode = ext4_evict_inode,
617ba13b 1176 .put_super = ext4_put_super,
617ba13b 1177 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1178 .freeze_fs = ext4_freeze,
1179 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1180 .statfs = ext4_statfs,
1181 .remount_fs = ext4_remount,
617ba13b 1182 .show_options = ext4_show_options,
ac27a0ec 1183#ifdef CONFIG_QUOTA
617ba13b
MC		 1184 .quota_read = ext4_quota_read,
1185 .quota_write = ext4_quota_write,
ac27a0ec 1186#endif
c39a7f84 1187 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1188};
1189
9ca92389
TT		 1190static const struct super_operations ext4_nojournal_sops = {
1191 .alloc_inode = ext4_alloc_inode,
1192 .destroy_inode = ext4_destroy_inode,
1193 .write_inode = ext4_write_inode,
1194 .dirty_inode = ext4_dirty_inode,
7ff9c073 1195 .drop_inode = ext4_drop_inode,
0930fcc1 1196 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1197 .write_super = ext4_write_super,
1198 .put_super = ext4_put_super,
1199 .statfs = ext4_statfs,
1200 .remount_fs = ext4_remount,
9ca92389
TT		 1201 .show_options = ext4_show_options,
1202#ifdef CONFIG_QUOTA
1203 .quota_read = ext4_quota_read,
1204 .quota_write = ext4_quota_write,
1205#endif
1206 .bdev_try_to_free_page = bdev_try_to_free_page,
1207};
1208
39655164 1209static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1210 .fh_to_dentry = ext4_fh_to_dentry,
1211 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1212 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1213};
1214
1215enum {
1216 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1217 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1218 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1219 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1220 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1221 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1222 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1223 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1224 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1225 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1226 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1227 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1228 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1229 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1230 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1231 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1232 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1233 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
ac27a0ec
DK		 1234};
1235
a447c093 1236static const match_table_t tokens = {
ac27a0ec
DK		 1237 {Opt_bsd_df, "bsddf"},
1238 {Opt_minix_df, "minixdf"},
1239 {Opt_grpid, "grpid"},
1240 {Opt_grpid, "bsdgroups"},
1241 {Opt_nogrpid, "nogrpid"},
1242 {Opt_nogrpid, "sysvgroups"},
1243 {Opt_resgid, "resgid=%u"},
1244 {Opt_resuid, "resuid=%u"},
1245 {Opt_sb, "sb=%u"},
1246 {Opt_err_cont, "errors=continue"},
1247 {Opt_err_panic, "errors=panic"},
1248 {Opt_err_ro, "errors=remount-ro"},
1249 {Opt_nouid32, "nouid32"},
ac27a0ec 1250 {Opt_debug, "debug"},
72578c33
TT		 1251 {Opt_removed, "oldalloc"},
1252 {Opt_removed, "orlov"},
ac27a0ec
DK		 1253 {Opt_user_xattr, "user_xattr"},
1254 {Opt_nouser_xattr, "nouser_xattr"},
1255 {Opt_acl, "acl"},
1256 {Opt_noacl, "noacl"},
e3bb52ae 1257 {Opt_noload, "norecovery"},
5a916be1 1258 {Opt_noload, "noload"},
72578c33
TT		 1259 {Opt_removed, "nobh"},
1260 {Opt_removed, "bh"},
ac27a0ec 1261 {Opt_commit, "commit=%u"},
30773840
TT		 1262 {Opt_min_batch_time, "min_batch_time=%u"},
1263 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1264 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1265 {Opt_journal_checksum, "journal_checksum"},
1266 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1267 {Opt_abort, "abort"},
1268 {Opt_data_journal, "data=journal"},
1269 {Opt_data_ordered, "data=ordered"},
1270 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1271 {Opt_data_err_abort, "data_err=abort"},
1272 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1273 {Opt_offusrjquota, "usrjquota="},
1274 {Opt_usrjquota, "usrjquota=%s"},
1275 {Opt_offgrpjquota, "grpjquota="},
1276 {Opt_grpjquota, "grpjquota=%s"},
1277 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1278 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1279 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1280 {Opt_grpquota, "grpquota"},
1281 {Opt_noquota, "noquota"},
1282 {Opt_quota, "quota"},
1283 {Opt_usrquota, "usrquota"},
1284 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1285 {Opt_barrier, "barrier"},
1286 {Opt_nobarrier, "nobarrier"},
25ec56b5 1287 {Opt_i_version, "i_version"},
c9de560d 1288 {Opt_stripe, "stripe=%u"},
64769240 1289 {Opt_delalloc, "delalloc"},
dd919b98 1290 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1291 {Opt_mblk_io_submit, "mblk_io_submit"},
1292 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1293 {Opt_block_validity, "block_validity"},
1294 {Opt_noblock_validity, "noblock_validity"},
240799cd 1295 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1296 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1297 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1298 {Opt_auto_da_alloc, "auto_da_alloc"},
1299 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1300 {Opt_dioread_nolock, "dioread_nolock"},
1301 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1302 {Opt_discard, "discard"},
1303 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1304 {Opt_init_itable, "init_itable=%u"},
1305 {Opt_init_itable, "init_itable"},
1306 {Opt_noinit_itable, "noinit_itable"},
c7198b9c
TT		 1307 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1308 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1309 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1310 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1311 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1312 {Opt_err, NULL},
ac27a0ec
DK		 1313};
1314
617ba13b 1315static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1316{
617ba13b 1317 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1318 char *options = (char *) *data;
1319
1320 if (!options || strncmp(options, "sb=", 3) != 0)
1321 return 1; /* Default location */
0b8e58a1 1322
ac27a0ec 1323 options += 3;
0b8e58a1 1324 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1325 sb_block = simple_strtoul(options, &options, 0);
1326 if (*options && *options != ',') {
4776004f 1327 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1328 (char *) *data);
1329 return 1;
1330 }
1331 if (*options == ',')
1332 options++;
1333 *data = (void *) options;
0b8e58a1 1334
ac27a0ec
DK		 1335 return sb_block;
1336}
1337
b3881f74 1338#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1339static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1340 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1341
56c50f11
DM		 1342#ifdef CONFIG_QUOTA
1343static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1344{
1345 struct ext4_sb_info *sbi = EXT4_SB(sb);
1346 char *qname;
1347
1348 if (sb_any_quota_loaded(sb) &&
1349 !sbi->s_qf_names[qtype]) {
1350 ext4_msg(sb, KERN_ERR,
1351 "Cannot change journaled "
1352 "quota options when quota turned on");
57f73c2c 1353 return -1;
56c50f11
DM		 1354 }
1355 qname = match_strdup(args);
1356 if (!qname) {
1357 ext4_msg(sb, KERN_ERR,
1358 "Not enough memory for storing quotafile name");
57f73c2c 1359 return -1;
56c50f11
DM		 1360 }
1361 if (sbi->s_qf_names[qtype] &&
1362 strcmp(sbi->s_qf_names[qtype], qname)) {
1363 ext4_msg(sb, KERN_ERR,
1364 "%s quota file already specified", QTYPE2NAME(qtype));
1365 kfree(qname);
57f73c2c 1366 return -1;
56c50f11
DM		 1367 }
1368 sbi->s_qf_names[qtype] = qname;
1369 if (strchr(sbi->s_qf_names[qtype], '/')) {
1370 ext4_msg(sb, KERN_ERR,
1371 "quotafile must be on filesystem root");
1372 kfree(sbi->s_qf_names[qtype]);
1373 sbi->s_qf_names[qtype] = NULL;
57f73c2c 1374 return -1;
56c50f11 1375 }
fd8c37ec 1376 set_opt(sb, QUOTA);
56c50f11
DM		 1377 return 1;
1378}
1379
1380static int clear_qf_name(struct super_block *sb, int qtype)
1381{
1382
1383 struct ext4_sb_info *sbi = EXT4_SB(sb);
1384
1385 if (sb_any_quota_loaded(sb) &&
1386 sbi->s_qf_names[qtype]) {
1387 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1388 " when quota turned on");
57f73c2c 1389 return -1;
56c50f11
DM		 1390 }
1391 /*
1392 * The space will be released later when all options are confirmed
1393 * to be correct
1394 */
1395 sbi->s_qf_names[qtype] = NULL;
1396 return 1;
1397}
1398#endif
1399
26092bf5
TT		 1400#define MOPT_SET 0x0001
1401#define MOPT_CLEAR 0x0002
1402#define MOPT_NOSUPPORT 0x0004
1403#define MOPT_EXPLICIT 0x0008
1404#define MOPT_CLEAR_ERR 0x0010
1405#define MOPT_GTE0 0x0020
ac27a0ec 1406#ifdef CONFIG_QUOTA
26092bf5
TT		 1407#define MOPT_Q 0
1408#define MOPT_QFMT 0x0040
1409#else
1410#define MOPT_Q MOPT_NOSUPPORT
1411#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1412#endif
26092bf5
TT		 1413#define MOPT_DATAJ 0x0080
1414
1415static const struct mount_opts {
1416 int token;
1417 int mount_opt;
1418 int flags;
1419} ext4_mount_opts[] = {
1420 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1421 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1422 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1423 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1424 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1425 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1426 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1427 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1428 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1429 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1430 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1431 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1432 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1433 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1434 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1435 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1436 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1437 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1438 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1439 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1440 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1441 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1442 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1443 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1444 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1445 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1446 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1447 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1448 {Opt_commit, 0, MOPT_GTE0},
1449 {Opt_max_batch_time, 0, MOPT_GTE0},
1450 {Opt_min_batch_time, 0, MOPT_GTE0},
1451 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1452 {Opt_init_itable, 0, MOPT_GTE0},
1453 {Opt_stripe, 0, MOPT_GTE0},
1454 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1455 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1456 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
03010a33 1457#ifdef CONFIG_EXT4_FS_XATTR
26092bf5
TT		 1458 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1459 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
ac27a0ec 1460#else
26092bf5
TT		 1461 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1462 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
ac27a0ec 1463#endif
03010a33 1464#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1465 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1466 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1467#else
26092bf5
TT		 1468 {Opt_acl, 0, MOPT_NOSUPPORT},
1469 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1470#endif
26092bf5
TT		 1471 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1472 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1473 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1474 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1475 MOPT_SET | MOPT_Q},
1476 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1477 MOPT_SET | MOPT_Q},
1478 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1479 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1480 {Opt_usrjquota, 0, MOPT_Q},
1481 {Opt_grpjquota, 0, MOPT_Q},
1482 {Opt_offusrjquota, 0, MOPT_Q},
1483 {Opt_offgrpjquota, 0, MOPT_Q},
1484 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1485 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1486 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1487 {Opt_err, 0, 0}
1488};
1489
1490static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1491 substring_t *args, unsigned long *journal_devnum,
1492 unsigned int *journal_ioprio, int is_remount)
1493{
1494 struct ext4_sb_info *sbi = EXT4_SB(sb);
1495 const struct mount_opts *m;
08cefc7a
EB		 1496 kuid_t uid;
1497 kgid_t gid;
26092bf5
TT		 1498 int arg = 0;
1499
57f73c2c
TT		 1500#ifdef CONFIG_QUOTA
1501 if (token == Opt_usrjquota)
1502 return set_qf_name(sb, USRQUOTA, &args[0]);
1503 else if (token == Opt_grpjquota)
1504 return set_qf_name(sb, GRPQUOTA, &args[0]);
1505 else if (token == Opt_offusrjquota)
1506 return clear_qf_name(sb, USRQUOTA);
1507 else if (token == Opt_offgrpjquota)
1508 return clear_qf_name(sb, GRPQUOTA);
1509#endif
26092bf5
TT		 1510 if (args->from && match_int(args, &arg))
1511 return -1;
1512 switch (token) {
f7048605
TT		 1513 case Opt_noacl:
1514 case Opt_nouser_xattr:
1515 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1516 break;
26092bf5
TT		 1517 case Opt_sb:
1518 return 1; /* handled by get_sb_block() */
1519 case Opt_removed:
1520 ext4_msg(sb, KERN_WARNING,
1521 "Ignoring removed %s option", opt);
1522 return 1;
1523 case Opt_resuid:
08cefc7a
EB		 1524 uid = make_kuid(current_user_ns(), arg);
1525 if (!uid_valid(uid)) {
1526 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1527 return -1;
1528 }
1529 sbi->s_resuid = uid;
26092bf5
TT		 1530 return 1;
1531 case Opt_resgid:
08cefc7a
EB		 1532 gid = make_kgid(current_user_ns(), arg);
1533 if (!gid_valid(gid)) {
1534 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1535 return -1;
1536 }
1537 sbi->s_resgid = gid;
26092bf5
TT		 1538 return 1;
1539 case Opt_abort:
1540 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1541 return 1;
1542 case Opt_i_version:
1543 sb->s_flags |= MS_I_VERSION;
1544 return 1;
1545 case Opt_journal_dev:
1546 if (is_remount) {
1547 ext4_msg(sb, KERN_ERR,
1548 "Cannot specify journal on remount");
1549 return -1;
1550 }
1551 *journal_devnum = arg;
1552 return 1;
1553 case Opt_journal_ioprio:
1554 if (arg < 0 || arg > 7)
1555 return -1;
1556 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1557 return 1;
1558 }
1559
1560 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1561 if (token != m->token)
1562 continue;
1563 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1564 return -1;
1565 if (m->flags & MOPT_EXPLICIT)
1566 set_opt2(sb, EXPLICIT_DELALLOC);
1567 if (m->flags & MOPT_CLEAR_ERR)
1568 clear_opt(sb, ERRORS_MASK);
1569 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1570 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1571 "options when quota turned on");
1572 return -1;
1573 }
1574
1575 if (m->flags & MOPT_NOSUPPORT) {
1576 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1577 } else if (token == Opt_commit) {
1578 if (arg == 0)
1579 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1580 sbi->s_commit_interval = HZ * arg;
1581 } else if (token == Opt_max_batch_time) {
1582 if (arg == 0)
1583 arg = EXT4_DEF_MAX_BATCH_TIME;
1584 sbi->s_max_batch_time = arg;
1585 } else if (token == Opt_min_batch_time) {
1586 sbi->s_min_batch_time = arg;
1587 } else if (token == Opt_inode_readahead_blks) {
1588 if (arg > (1 << 30))
1589 return -1;
1590 if (arg && !is_power_of_2(arg)) {
b31e1552 1591 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1592 "EXT4-fs: inode_readahead_blks"
1593 " must be a power of 2");
1594 return -1;
ac27a0ec 1595 }
26092bf5
TT		 1596 sbi->s_inode_readahead_blks = arg;
1597 } else if (token == Opt_init_itable) {
1598 set_opt(sb, INIT_INODE_TABLE);
1599 if (!args->from)
1600 arg = EXT4_DEF_LI_WAIT_MULT;
1601 sbi->s_li_wait_mult = arg;
1602 } else if (token == Opt_stripe) {
1603 sbi->s_stripe = arg;
1604 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1605 if (is_remount) {
eb513689
TT		 1606 if (!sbi->s_journal)
1607 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1608 else if (test_opt(sb, DATA_FLAGS) !=
1609 m->mount_opt) {
b31e1552 1610 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1611 "Cannot change data mode on remount");
1612 return -1;
ac27a0ec
DK		 1613 }
1614 } else {
fd8c37ec 1615 clear_opt(sb, DATA_FLAGS);
26092bf5 1616 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1617 }
ac27a0ec 1618#ifdef CONFIG_QUOTA
26092bf5 1619 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1620 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1621 sbi->s_jquota_fmt != m->mount_opt) {
1622 ext4_msg(sb, KERN_ERR, "Cannot "
1623 "change journaled quota options "
1624 "when quota turned on");
1625 return -1;
ac27a0ec 1626 }
26092bf5 1627 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1628#endif
26092bf5
TT		 1629 } else {
1630 if (!args->from)
1631 arg = 1;
1632 if (m->flags & MOPT_CLEAR)
1633 arg = !arg;
1634 else if (unlikely(!(m->flags & MOPT_SET))) {
1635 ext4_msg(sb, KERN_WARNING,
1636 "buggy handling of option %s", opt);
1637 WARN_ON(1);
1638 return -1;
3197ebdb 1639 }
26092bf5
TT		 1640 if (arg != 0)
1641 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1642 else
26092bf5 1643 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1644 }
26092bf5
TT		 1645 return 1;
1646 }
1647 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1648 "or missing value", opt);
1649 return -1;
1650}
1651
1652static int parse_options(char *options, struct super_block *sb,
1653 unsigned long *journal_devnum,
1654 unsigned int *journal_ioprio,
1655 int is_remount)
1656{
db7e5c66 1657#ifdef CONFIG_QUOTA
26092bf5 1658 struct ext4_sb_info *sbi = EXT4_SB(sb);
db7e5c66 1659#endif
26092bf5
TT		 1660 char *p;
1661 substring_t args[MAX_OPT_ARGS];
1662 int token;
1663
1664 if (!options)
1665 return 1;
1666
1667 while ((p = strsep(&options, ",")) != NULL) {
1668 if (!*p)
1669 continue;
1670 /*
1671 * Initialize args struct so we know whether arg was
1672 * found; some options take optional arguments.
1673 */
1674 args[0].to = args[0].from = 0;
1675 token = match_token(p, tokens, args);
1676 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1677 journal_ioprio, is_remount) < 0)
1678 return 0;
ac27a0ec
DK		 1679 }
1680#ifdef CONFIG_QUOTA
1681 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1682 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1683 clear_opt(sb, USRQUOTA);
ac27a0ec 1684
482a7425 1685 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1686 clear_opt(sb, GRPQUOTA);
ac27a0ec 1687
56c50f11 1688 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1689 ext4_msg(sb, KERN_ERR, "old and new quota "
1690 "format mixing");
ac27a0ec
DK		 1691 return 0;
1692 }
1693
1694 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1695 ext4_msg(sb, KERN_ERR, "journaled quota format "
1696 "not specified");
ac27a0ec
DK		 1697 return 0;
1698 }
1699 } else {
1700 if (sbi->s_jquota_fmt) {
b31e1552 1701 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1702 "specified with no journaling "
b31e1552 1703 "enabled");
ac27a0ec
DK		 1704 return 0;
1705 }
1706 }
1707#endif
1708 return 1;
1709}
1710
2adf6da8
TT		 1711static inline void ext4_show_quota_options(struct seq_file *seq,
1712 struct super_block *sb)
1713{
1714#if defined(CONFIG_QUOTA)
1715 struct ext4_sb_info *sbi = EXT4_SB(sb);
1716
1717 if (sbi->s_jquota_fmt) {
1718 char *fmtname = "";
1719
1720 switch (sbi->s_jquota_fmt) {
1721 case QFMT_VFS_OLD:
1722 fmtname = "vfsold";
1723 break;
1724 case QFMT_VFS_V0:
1725 fmtname = "vfsv0";
1726 break;
1727 case QFMT_VFS_V1:
1728 fmtname = "vfsv1";
1729 break;
1730 }
1731 seq_printf(seq, ",jqfmt=%s", fmtname);
1732 }
1733
1734 if (sbi->s_qf_names[USRQUOTA])
1735 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1736
1737 if (sbi->s_qf_names[GRPQUOTA])
1738 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1739
1740 if (test_opt(sb, USRQUOTA))
1741 seq_puts(seq, ",usrquota");
1742
1743 if (test_opt(sb, GRPQUOTA))
1744 seq_puts(seq, ",grpquota");
1745#endif
1746}
1747
5a916be1
TT		 1748static const char *token2str(int token)
1749{
1750 static const struct match_token *t;
1751
1752 for (t = tokens; t->token != Opt_err; t++)
1753 if (t->token == token && !strchr(t->pattern, '='))
1754 break;
1755 return t->pattern;
1756}
1757
2adf6da8
TT		 1758/*
1759 * Show an option if
1760 * - it's set to a non-default value OR
1761 * - if the per-sb default is different from the global default
1762 */
66acdcf4
TT		 1763static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1764 int nodefs)
2adf6da8 1765{
2adf6da8
TT		 1766 struct ext4_sb_info *sbi = EXT4_SB(sb);
1767 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1768 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1769 const struct mount_opts *m;
66acdcf4 1770 char sep = nodefs ? '\n' : ',';
2adf6da8 1771
66acdcf4
TT		 1772#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1773#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1774
1775 if (sbi->s_sb_block != 1)
5a916be1
TT		 1776 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1777
1778 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1779 int want_set = m->flags & MOPT_SET;
1780 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1781 (m->flags & MOPT_CLEAR_ERR))
1782 continue;
66acdcf4 1783 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1784 continue; /* skip if same as the default */
1785 if ((want_set &&
1786 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1787 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1788 continue; /* select Opt_noFoo vs Opt_Foo */
1789 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1790 }
5a916be1 1791
08cefc7a 1792 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1793 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1794 SEQ_OPTS_PRINT("resuid=%u",
1795 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1796 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1797 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1798 SEQ_OPTS_PRINT("resgid=%u",
1799 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1800 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1801 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1802 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1803 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1804 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1805 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1806 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1807 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1808 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1809 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1810 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1811 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1812 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1813 if (sb->s_flags & MS_I_VERSION)
5a916be1 1814 SEQ_OPTS_PUTS("i_version");
66acdcf4 1815 if (nodefs || sbi->s_stripe)
5a916be1 1816 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1817 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1818 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1819 SEQ_OPTS_PUTS("data=journal");
1820 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1821 SEQ_OPTS_PUTS("data=ordered");
1822 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1823 SEQ_OPTS_PUTS("data=writeback");
1824 }
66acdcf4
TT		 1825 if (nodefs ||
1826 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1827 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1828 sbi->s_inode_readahead_blks);
2adf6da8 1829
66acdcf4
TT		 1830 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1831 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1832 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2adf6da8
TT		 1833
1834 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1835 return 0;
1836}
1837
66acdcf4
TT		 1838static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1839{
1840 return _ext4_show_options(seq, root->d_sb, 0);
1841}
1842
1843static int options_seq_show(struct seq_file *seq, void *offset)
1844{
1845 struct super_block *sb = seq->private;
1846 int rc;
1847
1848 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1849 rc = _ext4_show_options(seq, sb, 1);
1850 seq_puts(seq, "\n");
1851 return rc;
1852}
1853
1854static int options_open_fs(struct inode *inode, struct file *file)
1855{
1856 return single_open(file, options_seq_show, PDE(inode)->data);
1857}
1858
1859static const struct file_operations ext4_seq_options_fops = {
1860 .owner = THIS_MODULE,
1861 .open = options_open_fs,
1862 .read = seq_read,
1863 .llseek = seq_lseek,
1864 .release = single_release,
1865};
1866
617ba13b 1867static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1868 int read_only)
1869{
617ba13b 1870 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1871 int res = 0;
1872
617ba13b 1873 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1874 ext4_msg(sb, KERN_ERR, "revision level too high, "
1875 "forcing read-only mode");
ac27a0ec
DK		 1876 res = MS_RDONLY;
1877 }
1878 if (read_only)
281b5995 1879 goto done;
617ba13b 1880 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1881 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1882 "running e2fsck is recommended");
617ba13b 1883 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1884 ext4_msg(sb, KERN_WARNING,
1885 "warning: mounting fs with errors, "
1886 "running e2fsck is recommended");
ed3ce80a 1887 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1888 le16_to_cpu(es->s_mnt_count) >=
1889 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1890 ext4_msg(sb, KERN_WARNING,
1891 "warning: maximal mount count reached, "
1892 "running e2fsck is recommended");
ac27a0ec
DK		 1893 else if (le32_to_cpu(es->s_checkinterval) &&
1894 (le32_to_cpu(es->s_lastcheck) +
1895 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1896 ext4_msg(sb, KERN_WARNING,
1897 "warning: checktime reached, "
1898 "running e2fsck is recommended");
0b8e58a1 1899 if (!sbi->s_journal)
0390131b 1900 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1901 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1902 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1903 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1904 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1905 ext4_update_dynamic_rev(sb);
0390131b
FM		 1906 if (sbi->s_journal)
1907 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1908
e2d67052 1909 ext4_commit_super(sb, 1);
281b5995 1910done:
ac27a0ec 1911 if (test_opt(sb, DEBUG))
a9df9a49 1912 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1913 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1914 sb->s_blocksize,
1915 sbi->s_groups_count,
617ba13b
MC		 1916 EXT4_BLOCKS_PER_GROUP(sb),
1917 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1918 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1919
7abc52c2 1920 cleancache_init_fs(sb);
ac27a0ec
DK		 1921 return res;
1922}
1923
772cb7c8
JS		 1924static int ext4_fill_flex_info(struct super_block *sb)
1925{
1926 struct ext4_sb_info *sbi = EXT4_SB(sb);
1927 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1928 ext4_group_t flex_group_count;
1929 ext4_group_t flex_group;
d50f2ab6 1930 unsigned int groups_per_flex = 0;
c5ca7c76 1931 size_t size;
772cb7c8
JS		 1932 int i;
1933
503358ae 1934 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1935 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1936 sbi->s_log_groups_per_flex = 0;
1937 return 1;
1938 }
d50f2ab6 1939 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
772cb7c8 1940
c62a11fd
FB		 1941 /* We allocate both existing and potentially added groups */
1942 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1943 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1944 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 1945 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 1946 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 1947 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 1948 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1949 flex_group_count);
1950 goto failed;
772cb7c8 1951 }
772cb7c8 1952
772cb7c8 1953 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1954 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1955
1956 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1957 atomic_add(ext4_free_inodes_count(sb, gdp),
1958 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1959 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1960 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1961 atomic_add(ext4_used_dirs_count(sb, gdp),
1962 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1963 }
1964
1965 return 1;
1966failed:
1967 return 0;
1968}
1969
feb0ab32
DW		 1970static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1971 struct ext4_group_desc *gdp)
717d50e4 1972{
feb0ab32 1973 int offset;
717d50e4 1974 __u16 crc = 0;
feb0ab32 1975 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1976
feb0ab32
DW		 1977 if ((sbi->s_es->s_feature_ro_compat &
1978 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1979 /* Use new metadata_csum algorithm */
1980 __u16 old_csum;
1981 __u32 csum32;
1982
1983 old_csum = gdp->bg_checksum;
1984 gdp->bg_checksum = 0;
1985 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1986 sizeof(le_group));
1987 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1988 sbi->s_desc_size);
1989 gdp->bg_checksum = old_csum;
1990
1991 crc = csum32 & 0xFFFF;
1992 goto out;
717d50e4
AD		 1993 }
1994
feb0ab32
DW		 1995 /* old crc16 code */
1996 offset = offsetof(struct ext4_group_desc, bg_checksum);
1997
1998 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1999 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2000 crc = crc16(crc, (__u8 *)gdp, offset);
2001 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2002 /* for checksum of struct ext4_group_desc do the rest...*/
2003 if ((sbi->s_es->s_feature_incompat &
2004 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2005 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2006 crc = crc16(crc, (__u8 *)gdp + offset,
2007 le16_to_cpu(sbi->s_es->s_desc_size) -
2008 offset);
2009
2010out:
717d50e4
AD		 2011 return cpu_to_le16(crc);
2012}
2013
feb0ab32 2014int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2015 struct ext4_group_desc *gdp)
2016{
feb0ab32
DW		 2017 if (ext4_has_group_desc_csum(sb) &&
2018 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2019 block_group, gdp)))
717d50e4
AD		 2020 return 0;
2021
2022 return 1;
2023}
2024
feb0ab32
DW		 2025void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2026 struct ext4_group_desc *gdp)
2027{
2028 if (!ext4_has_group_desc_csum(sb))
2029 return;
2030 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2031}
2032
ac27a0ec 2033/* Called at mount-time, super-block is locked */
bfff6873
LC		 2034static int ext4_check_descriptors(struct super_block *sb,
2035 ext4_group_t *first_not_zeroed)
ac27a0ec 2036{
617ba13b
MC		 2037 struct ext4_sb_info *sbi = EXT4_SB(sb);
2038 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2039 ext4_fsblk_t last_block;
bd81d8ee
LV		 2040 ext4_fsblk_t block_bitmap;
2041 ext4_fsblk_t inode_bitmap;
2042 ext4_fsblk_t inode_table;
ce421581 2043 int flexbg_flag = 0;
bfff6873 2044 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2045
ce421581
JS		 2046 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2047 flexbg_flag = 1;
2048
af5bc92d 2049 ext4_debug("Checking group descriptors");
ac27a0ec 2050
197cd65a
AM		 2051 for (i = 0; i < sbi->s_groups_count; i++) {
2052 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2053
ce421581 2054 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2055 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2056 else
2057 last_block = first_block +
617ba13b 2058 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2059
bfff6873
LC		 2060 if ((grp == sbi->s_groups_count) &&
2061 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2062 grp = i;
2063
8fadc143 2064 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2065 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2066 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2067 "Block bitmap for group %u not in group "
b31e1552 2068 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2069 return 0;
2070 }
8fadc143 2071 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2072 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2073 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2074 "Inode bitmap for group %u not in group "
b31e1552 2075 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2076 return 0;
2077 }
8fadc143 2078 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2079 if (inode_table < first_block ||
2b2d6d01 2080 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2081 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2082 "Inode table for group %u not in group "
b31e1552 2083 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2084 return 0;
2085 }
955ce5f5 2086 ext4_lock_group(sb, i);
feb0ab32 2087 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2088 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2089 "Checksum for group %u failed (%u!=%u)",
2090 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2091 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2092 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2093 ext4_unlock_group(sb, i);
8a266467 2094 return 0;
7ee1ec4c 2095 }
717d50e4 2096 }
955ce5f5 2097 ext4_unlock_group(sb, i);
ce421581
JS		 2098 if (!flexbg_flag)
2099 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2100 }
bfff6873
LC		 2101 if (NULL != first_not_zeroed)
2102 *first_not_zeroed = grp;
ac27a0ec 2103
5dee5437
TT		 2104 ext4_free_blocks_count_set(sbi->s_es,
2105 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2106 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2107 return 1;
2108}
2109
617ba13b 2110/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2111 * the superblock) which were deleted from all directories, but held open by
2112 * a process at the time of a crash. We walk the list and try to delete these
2113 * inodes at recovery time (only with a read-write filesystem).
2114 *
2115 * In order to keep the orphan inode chain consistent during traversal (in
2116 * case of crash during recovery), we link each inode into the superblock
2117 * orphan list_head and handle it the same way as an inode deletion during
2118 * normal operation (which journals the operations for us).
2119 *
2120 * We only do an iget() and an iput() on each inode, which is very safe if we
2121 * accidentally point at an in-use or already deleted inode. The worst that
2122 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2123 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2124 * e2fsck was run on this filesystem, and it must have already done the orphan
2125 * inode cleanup for us, so we can safely abort without any further action.
2126 */
2b2d6d01
TT		 2127static void ext4_orphan_cleanup(struct super_block *sb,
2128 struct ext4_super_block *es)
ac27a0ec
DK		 2129{
2130 unsigned int s_flags = sb->s_flags;
2131 int nr_orphans = 0, nr_truncates = 0;
2132#ifdef CONFIG_QUOTA
2133 int i;
2134#endif
2135 if (!es->s_last_orphan) {
2136 jbd_debug(4, "no orphan inodes to clean up\n");
2137 return;
2138 }
2139
a8f48a95 2140 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2141 ext4_msg(sb, KERN_ERR, "write access "
2142 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2143 return;
2144 }
2145
d39195c3
AG		 2146 /* Check if feature set would not allow a r/w mount */
2147 if (!ext4_feature_set_ok(sb, 0)) {
2148 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2149 "unknown ROCOMPAT features");
2150 return;
2151 }
2152
617ba13b 2153 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2154 if (es->s_last_orphan)
2155 jbd_debug(1, "Errors on filesystem, "
2156 "clearing orphan list.\n");
2157 es->s_last_orphan = 0;
2158 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2159 return;
2160 }
2161
2162 if (s_flags & MS_RDONLY) {
b31e1552 2163 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2164 sb->s_flags &= ~MS_RDONLY;
2165 }
2166#ifdef CONFIG_QUOTA
2167 /* Needed for iput() to work correctly and not trash data */
2168 sb->s_flags |= MS_ACTIVE;
2169 /* Turn on quotas so that they are updated correctly */
2170 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2171 if (EXT4_SB(sb)->s_qf_names[i]) {
2172 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2173 if (ret < 0)
b31e1552
ES		 2174 ext4_msg(sb, KERN_ERR,
2175 "Cannot turn on journaled "
2176 "quota: error %d", ret);
ac27a0ec
DK		 2177 }
2178 }
2179#endif
2180
2181 while (es->s_last_orphan) {
2182 struct inode *inode;
2183
97bd42b9
JB		 2184 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2185 if (IS_ERR(inode)) {
ac27a0ec
DK		 2186 es->s_last_orphan = 0;
2187 break;
2188 }
2189
617ba13b 2190 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2191 dquot_initialize(inode);
ac27a0ec 2192 if (inode->i_nlink) {
b31e1552
ES		 2193 ext4_msg(sb, KERN_DEBUG,
2194 "%s: truncating inode %lu to %lld bytes",
46e665e9 2195 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2196 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2197 inode->i_ino, inode->i_size);
617ba13b 2198 ext4_truncate(inode);
ac27a0ec
DK		 2199 nr_truncates++;
2200 } else {
b31e1552
ES		 2201 ext4_msg(sb, KERN_DEBUG,
2202 "%s: deleting unreferenced inode %lu",
46e665e9 2203 __func__, inode->i_ino);
ac27a0ec
DK		 2204 jbd_debug(2, "deleting unreferenced inode %lu\n",
2205 inode->i_ino);
2206 nr_orphans++;
2207 }
2208 iput(inode); /* The delete magic happens here! */
2209 }
2210
2b2d6d01 2211#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2212
2213 if (nr_orphans)
b31e1552
ES		 2214 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2215 PLURAL(nr_orphans));
ac27a0ec 2216 if (nr_truncates)
b31e1552
ES		 2217 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2218 PLURAL(nr_truncates));
ac27a0ec
DK		 2219#ifdef CONFIG_QUOTA
2220 /* Turn quotas off */
2221 for (i = 0; i < MAXQUOTAS; i++) {
2222 if (sb_dqopt(sb)->files[i])
287a8095 2223 dquot_quota_off(sb, i);
ac27a0ec
DK		 2224 }
2225#endif
2226 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2227}
0b8e58a1 2228
cd2291a4
ES		 2229/*
2230 * Maximal extent format file size.
2231 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2232 * extent format containers, within a sector_t, and within i_blocks
2233 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2234 * so that won't be a limiting factor.
2235 *
f17722f9
LC		 2236 * However there is other limiting factor. We do store extents in the form
2237 * of starting block and length, hence the resulting length of the extent
2238 * covering maximum file size must fit into on-disk format containers as
2239 * well. Given that length is always by 1 unit bigger than max unit (because
2240 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2241 *
cd2291a4
ES		 2242 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2243 */
f287a1a5 2244static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2245{
2246 loff_t res;
2247 loff_t upper_limit = MAX_LFS_FILESIZE;
2248
2249 /* small i_blocks in vfs inode? */
f287a1a5 2250 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2251 /*
90c699a9 2252 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2253 * i_block represent total blocks in 512 bytes
2254 * 32 == size of vfs inode i_blocks * 8
2255 */
2256 upper_limit = (1LL << 32) - 1;
2257
2258 /* total blocks in file system block size */
2259 upper_limit >>= (blkbits - 9);
2260 upper_limit <<= blkbits;
2261 }
2262
f17722f9
LC		 2263 /*
2264 * 32-bit extent-start container, ee_block. We lower the maxbytes
2265 * by one fs block, so ee_len can cover the extent of maximum file
2266 * size
2267 */
2268 res = (1LL << 32) - 1;
cd2291a4 2269 res <<= blkbits;
cd2291a4
ES		 2270
2271 /* Sanity check against vm- & vfs- imposed limits */
2272 if (res > upper_limit)
2273 res = upper_limit;
2274
2275 return res;
2276}
ac27a0ec 2277
ac27a0ec 2278/*
cd2291a4 2279 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2280 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2281 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2282 */
f287a1a5 2283static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2284{
617ba13b 2285 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2286 int meta_blocks;
2287 loff_t upper_limit;
0b8e58a1
AD		 2288 /* This is calculated to be the largest file size for a dense, block
2289 * mapped file such that the file's total number of 512-byte sectors,
2290 * including data and all indirect blocks, does not exceed (2^48 - 1).
2291 *
2292 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2293 * number of 512-byte sectors of the file.
0fc1b451
AK		 2294 */
2295
f287a1a5 2296 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2297 /*
90c699a9 2298 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2299 * the inode i_block field represents total file blocks in
2300 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2301 */
2302 upper_limit = (1LL << 32) - 1;
2303
2304 /* total blocks in file system block size */
2305 upper_limit >>= (bits - 9);
2306
2307 } else {
8180a562
AK		 2308 /*
2309 * We use 48 bit ext4_inode i_blocks
2310 * With EXT4_HUGE_FILE_FL set the i_blocks
2311 * represent total number of blocks in
2312 * file system block size
2313 */
0fc1b451
AK		 2314 upper_limit = (1LL << 48) - 1;
2315
0fc1b451
AK		 2316 }
2317
2318 /* indirect blocks */
2319 meta_blocks = 1;
2320 /* double indirect blocks */
2321 meta_blocks += 1 + (1LL << (bits-2));
2322 /* tripple indirect blocks */
2323 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2324
2325 upper_limit -= meta_blocks;
2326 upper_limit <<= bits;
ac27a0ec
DK		 2327
2328 res += 1LL << (bits-2);
2329 res += 1LL << (2*(bits-2));
2330 res += 1LL << (3*(bits-2));
2331 res <<= bits;
2332 if (res > upper_limit)
2333 res = upper_limit;
0fc1b451
AK		 2334
2335 if (res > MAX_LFS_FILESIZE)
2336 res = MAX_LFS_FILESIZE;
2337
ac27a0ec
DK		 2338 return res;
2339}
2340
617ba13b 2341static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2342 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2343{
617ba13b 2344 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2345 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2346 int has_super = 0;
2347
2348 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2349
617ba13b 2350 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2351 nr < first_meta_bg)
70bbb3e0 2352 return logical_sb_block + nr + 1;
ac27a0ec 2353 bg = sbi->s_desc_per_block * nr;
617ba13b 2354 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2355 has_super = 1;
0b8e58a1 2356
617ba13b 2357 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2358}
2359
c9de560d
AT		 2360/**
2361 * ext4_get_stripe_size: Get the stripe size.
2362 * @sbi: In memory super block info
2363 *
2364 * If we have specified it via mount option, then
2365 * use the mount option value. If the value specified at mount time is
2366 * greater than the blocks per group use the super block value.
2367 * If the super block value is greater than blocks per group return 0.
2368 * Allocator needs it be less than blocks per group.
2369 *
2370 */
2371static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2372{
2373 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2374 unsigned long stripe_width =
2375 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2376 int ret;
c9de560d
AT		 2377
2378 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2379 ret = sbi->s_stripe;
2380 else if (stripe_width <= sbi->s_blocks_per_group)
2381 ret = stripe_width;
2382 else if (stride <= sbi->s_blocks_per_group)
2383 ret = stride;
2384 else
2385 ret = 0;
c9de560d 2386
3eb08658
DE		 2387 /*
2388 * If the stripe width is 1, this makes no sense and
2389 * we set it to 0 to turn off stripe handling code.
2390 */
2391 if (ret <= 1)
2392 ret = 0;
c9de560d 2393
3eb08658 2394 return ret;
c9de560d 2395}
ac27a0ec 2396
3197ebdb
TT		 2397/* sysfs supprt */
2398
2399struct ext4_attr {
2400 struct attribute attr;
2401 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2402 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2403 const char *, size_t);
2404 int offset;
2405};
2406
2407static int parse_strtoul(const char *buf,
2408 unsigned long max, unsigned long *value)
2409{
2410 char *endp;
2411
e7d2860b
AGR		 2412 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2413 endp = skip_spaces(endp);
3197ebdb
TT		 2414 if (*endp || *value > max)
2415 return -EINVAL;
2416
2417 return 0;
2418}
2419
2420static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2421 struct ext4_sb_info *sbi,
2422 char *buf)
2423{
2424 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2425 (s64) EXT4_C2B(sbi,
2426 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2427}
2428
2429static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2430 struct ext4_sb_info *sbi, char *buf)
2431{
2432 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2433
f613dfcb
TT		 2434 if (!sb->s_bdev->bd_part)
2435 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2436 return snprintf(buf, PAGE_SIZE, "%lu\n",
2437 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2438 sbi->s_sectors_written_start) >> 1);
2439}
2440
2441static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2442 struct ext4_sb_info *sbi, char *buf)
2443{
2444 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2445
f613dfcb
TT		 2446 if (!sb->s_bdev->bd_part)
2447 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2448 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2449 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2450 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2451 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2452}
2453
2454static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2455 struct ext4_sb_info *sbi,
2456 const char *buf, size_t count)
2457{
2458 unsigned long t;
2459
2460 if (parse_strtoul(buf, 0x40000000, &t))
2461 return -EINVAL;
2462
5dbd571d 2463 if (t && !is_power_of_2(t))
3197ebdb
TT		 2464 return -EINVAL;
2465
2466 sbi->s_inode_readahead_blks = t;
2467 return count;
2468}
2469
2470static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2471 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2472{
2473 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2474
2475 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2476}
2477
2478static ssize_t sbi_ui_store(struct ext4_attr *a,
2479 struct ext4_sb_info *sbi,
2480 const char *buf, size_t count)
2481{
2482 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2483 unsigned long t;
2484
2485 if (parse_strtoul(buf, 0xffffffff, &t))
2486 return -EINVAL;
2487 *ui = t;
2488 return count;
2489}
2490
2c0544b2
TT		 2491static ssize_t trigger_test_error(struct ext4_attr *a,
2492 struct ext4_sb_info *sbi,
2493 const char *buf, size_t count)
2494{
2495 int len = count;
2496
2497 if (!capable(CAP_SYS_ADMIN))
2498 return -EPERM;
2499
2500 if (len && buf[len-1] == '\n')
2501 len--;
2502
2503 if (len)
2504 ext4_error(sbi->s_sb, "%.*s", len, buf);
2505 return count;
2506}
2507
3197ebdb
TT		 2508#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2509static struct ext4_attr ext4_attr_##_name = { \
2510 .attr = {.name = __stringify(_name), .mode = _mode }, \
2511 .show = _show, \
2512 .store = _store, \
2513 .offset = offsetof(struct ext4_sb_info, _elname), \
2514}
2515#define EXT4_ATTR(name, mode, show, store) \
2516static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2517
857ac889 2518#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2519#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2520#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2521#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2522 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2523#define ATTR_LIST(name) &ext4_attr_##name.attr
2524
2525EXT4_RO_ATTR(delayed_allocation_blocks);
2526EXT4_RO_ATTR(session_write_kbytes);
2527EXT4_RO_ATTR(lifetime_write_kbytes);
2528EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2529 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2530EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2531EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2532EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2533EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2534EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2535EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2536EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2537EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2c0544b2 2538EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2539
2540static struct attribute *ext4_attrs[] = {
2541 ATTR_LIST(delayed_allocation_blocks),
2542 ATTR_LIST(session_write_kbytes),
2543 ATTR_LIST(lifetime_write_kbytes),
2544 ATTR_LIST(inode_readahead_blks),
11013911 2545 ATTR_LIST(inode_goal),
3197ebdb
TT		 2546 ATTR_LIST(mb_stats),
2547 ATTR_LIST(mb_max_to_scan),
2548 ATTR_LIST(mb_min_to_scan),
2549 ATTR_LIST(mb_order2_req),
2550 ATTR_LIST(mb_stream_req),
2551 ATTR_LIST(mb_group_prealloc),
55138e0b 2552 ATTR_LIST(max_writeback_mb_bump),
2c0544b2 2553 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2554 NULL,
2555};
2556
857ac889
LC		 2557/* Features this copy of ext4 supports */
2558EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2559EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2560
2561static struct attribute *ext4_feat_attrs[] = {
2562 ATTR_LIST(lazy_itable_init),
27ee40df 2563 ATTR_LIST(batched_discard),
857ac889
LC		 2564 NULL,
2565};
2566
3197ebdb
TT		 2567static ssize_t ext4_attr_show(struct kobject *kobj,
2568 struct attribute *attr, char *buf)
2569{
2570 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2571 s_kobj);
2572 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2573
2574 return a->show ? a->show(a, sbi, buf) : 0;
2575}
2576
2577static ssize_t ext4_attr_store(struct kobject *kobj,
2578 struct attribute *attr,
2579 const char *buf, size_t len)
2580{
2581 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2582 s_kobj);
2583 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2584
2585 return a->store ? a->store(a, sbi, buf, len) : 0;
2586}
2587
2588static void ext4_sb_release(struct kobject *kobj)
2589{
2590 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2591 s_kobj);
2592 complete(&sbi->s_kobj_unregister);
2593}
2594
52cf25d0 2595static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2596 .show = ext4_attr_show,
2597 .store = ext4_attr_store,
2598};
2599
2600static struct kobj_type ext4_ktype = {
2601 .default_attrs = ext4_attrs,
2602 .sysfs_ops = &ext4_attr_ops,
2603 .release = ext4_sb_release,
2604};
2605
857ac889
LC		 2606static void ext4_feat_release(struct kobject *kobj)
2607{
2608 complete(&ext4_feat->f_kobj_unregister);
2609}
2610
2611static struct kobj_type ext4_feat_ktype = {
2612 .default_attrs = ext4_feat_attrs,
2613 .sysfs_ops = &ext4_attr_ops,
2614 .release = ext4_feat_release,
2615};
2616
a13fb1a4
ES		 2617/*
2618 * Check whether this filesystem can be mounted based on
2619 * the features present and the RDONLY/RDWR mount requested.
2620 * Returns 1 if this filesystem can be mounted as requested,
2621 * 0 if it cannot be.
2622 */
2623static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2624{
2625 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2626 ext4_msg(sb, KERN_ERR,
2627 "Couldn't mount because of "
2628 "unsupported optional features (%x)",
2629 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2630 ~EXT4_FEATURE_INCOMPAT_SUPP));
2631 return 0;
2632 }
2633
2634 if (readonly)
2635 return 1;
2636
2637 /* Check that feature set is OK for a read-write mount */
2638 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2639 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2640 "unsupported optional features (%x)",
2641 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2642 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2643 return 0;
2644 }
2645 /*
2646 * Large file size enabled file system can only be mounted
2647 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2648 */
2649 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2650 if (sizeof(blkcnt_t) < sizeof(u64)) {
2651 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2652 "cannot be mounted RDWR without "
2653 "CONFIG_LBDAF");
2654 return 0;
2655 }
2656 }
bab08ab9
TT		 2657 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2658 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2659 ext4_msg(sb, KERN_ERR,
2660 "Can't support bigalloc feature without "
2661 "extents feature\n");
2662 return 0;
2663 }
a13fb1a4
ES		 2664 return 1;
2665}
2666
66e61a9e
TT		 2667/*
2668 * This function is called once a day if we have errors logged
2669 * on the file system
2670 */
2671static void print_daily_error_info(unsigned long arg)
2672{
2673 struct super_block *sb = (struct super_block *) arg;
2674 struct ext4_sb_info *sbi;
2675 struct ext4_super_block *es;
2676
2677 sbi = EXT4_SB(sb);
2678 es = sbi->s_es;
2679
2680 if (es->s_error_count)
2681 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2682 le32_to_cpu(es->s_error_count));
2683 if (es->s_first_error_time) {
2684 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2685 sb->s_id, le32_to_cpu(es->s_first_error_time),
2686 (int) sizeof(es->s_first_error_func),
2687 es->s_first_error_func,
2688 le32_to_cpu(es->s_first_error_line));
2689 if (es->s_first_error_ino)
2690 printk(": inode %u",
2691 le32_to_cpu(es->s_first_error_ino));
2692 if (es->s_first_error_block)
2693 printk(": block %llu", (unsigned long long)
2694 le64_to_cpu(es->s_first_error_block));
2695 printk("\n");
2696 }
2697 if (es->s_last_error_time) {
2698 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2699 sb->s_id, le32_to_cpu(es->s_last_error_time),
2700 (int) sizeof(es->s_last_error_func),
2701 es->s_last_error_func,
2702 le32_to_cpu(es->s_last_error_line));
2703 if (es->s_last_error_ino)
2704 printk(": inode %u",
2705 le32_to_cpu(es->s_last_error_ino));
2706 if (es->s_last_error_block)
2707 printk(": block %llu", (unsigned long long)
2708 le64_to_cpu(es->s_last_error_block));
2709 printk("\n");
2710 }
2711 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2712}
2713
bfff6873
LC		 2714/* Find next suitable group and run ext4_init_inode_table */
2715static int ext4_run_li_request(struct ext4_li_request *elr)
2716{
2717 struct ext4_group_desc *gdp = NULL;
2718 ext4_group_t group, ngroups;
2719 struct super_block *sb;
2720 unsigned long timeout = 0;
2721 int ret = 0;
2722
2723 sb = elr->lr_super;
2724 ngroups = EXT4_SB(sb)->s_groups_count;
2725
2726 for (group = elr->lr_next_group; group < ngroups; group++) {
2727 gdp = ext4_get_group_desc(sb, group, NULL);
2728 if (!gdp) {
2729 ret = 1;
2730 break;
2731 }
2732
2733 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2734 break;
2735 }
2736
2737 if (group == ngroups)
2738 ret = 1;
2739
2740 if (!ret) {
2741 timeout = jiffies;
2742 ret = ext4_init_inode_table(sb, group,
2743 elr->lr_timeout ? 0 : 1);
2744 if (elr->lr_timeout == 0) {
51ce6511
LC		 2745 timeout = (jiffies - timeout) *
2746 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2747 elr->lr_timeout = timeout;
2748 }
2749 elr->lr_next_sched = jiffies + elr->lr_timeout;
2750 elr->lr_next_group = group + 1;
2751 }
2752
2753 return ret;
2754}
2755
2756/*
2757 * Remove lr_request from the list_request and free the
4ed5c033 2758 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2759 */
2760static void ext4_remove_li_request(struct ext4_li_request *elr)
2761{
2762 struct ext4_sb_info *sbi;
2763
2764 if (!elr)
2765 return;
2766
2767 sbi = elr->lr_sbi;
2768
2769 list_del(&elr->lr_request);
2770 sbi->s_li_request = NULL;
2771 kfree(elr);
2772}
2773
2774static void ext4_unregister_li_request(struct super_block *sb)
2775{
1bb933fb
LC		 2776 mutex_lock(&ext4_li_mtx);
2777 if (!ext4_li_info) {
2778 mutex_unlock(&ext4_li_mtx);
bfff6873 2779 return;
1bb933fb 2780 }
bfff6873
LC		 2781
2782 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2783 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2784 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2785 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2786}
2787
8f1f7453
ES		 2788static struct task_struct *ext4_lazyinit_task;
2789
bfff6873
LC		 2790/*
2791 * This is the function where ext4lazyinit thread lives. It walks
2792 * through the request list searching for next scheduled filesystem.
2793 * When such a fs is found, run the lazy initialization request
2794 * (ext4_rn_li_request) and keep track of the time spend in this
2795 * function. Based on that time we compute next schedule time of
2796 * the request. When walking through the list is complete, compute
2797 * next waking time and put itself into sleep.
2798 */
2799static int ext4_lazyinit_thread(void *arg)
2800{
2801 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2802 struct list_head *pos, *n;
2803 struct ext4_li_request *elr;
4ed5c033 2804 unsigned long next_wakeup, cur;
bfff6873
LC		 2805
2806 BUG_ON(NULL == eli);
2807
bfff6873
LC		 2808cont_thread:
2809 while (true) {
2810 next_wakeup = MAX_JIFFY_OFFSET;
2811
2812 mutex_lock(&eli->li_list_mtx);
2813 if (list_empty(&eli->li_request_list)) {
2814 mutex_unlock(&eli->li_list_mtx);
2815 goto exit_thread;
2816 }
2817
2818 list_for_each_safe(pos, n, &eli->li_request_list) {
2819 elr = list_entry(pos, struct ext4_li_request,
2820 lr_request);
2821
b2c78cd0
TT		 2822 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2823 if (ext4_run_li_request(elr) != 0) {
2824 /* error, remove the lazy_init job */
2825 ext4_remove_li_request(elr);
2826 continue;
2827 }
bfff6873
LC		 2828 }
2829
2830 if (time_before(elr->lr_next_sched, next_wakeup))
2831 next_wakeup = elr->lr_next_sched;
2832 }
2833 mutex_unlock(&eli->li_list_mtx);
2834
a0acae0e 2835 try_to_freeze();
bfff6873 2836
4ed5c033
LC		 2837 cur = jiffies;
2838 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2839 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2840 cond_resched();
2841 continue;
2842 }
2843
4ed5c033
LC		 2844 schedule_timeout_interruptible(next_wakeup - cur);
2845
8f1f7453
ES		 2846 if (kthread_should_stop()) {
2847 ext4_clear_request_list();
2848 goto exit_thread;
2849 }
bfff6873
LC		 2850 }
2851
2852exit_thread:
2853 /*
2854 * It looks like the request list is empty, but we need
2855 * to check it under the li_list_mtx lock, to prevent any
2856 * additions into it, and of course we should lock ext4_li_mtx
2857 * to atomically free the list and ext4_li_info, because at
2858 * this point another ext4 filesystem could be registering
2859 * new one.
2860 */
2861 mutex_lock(&ext4_li_mtx);
2862 mutex_lock(&eli->li_list_mtx);
2863 if (!list_empty(&eli->li_request_list)) {
2864 mutex_unlock(&eli->li_list_mtx);
2865 mutex_unlock(&ext4_li_mtx);
2866 goto cont_thread;
2867 }
2868 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2869 kfree(ext4_li_info);
2870 ext4_li_info = NULL;
2871 mutex_unlock(&ext4_li_mtx);
2872
2873 return 0;
2874}
2875
2876static void ext4_clear_request_list(void)
2877{
2878 struct list_head *pos, *n;
2879 struct ext4_li_request *elr;
2880
2881 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2882 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2883 elr = list_entry(pos, struct ext4_li_request,
2884 lr_request);
2885 ext4_remove_li_request(elr);
2886 }
2887 mutex_unlock(&ext4_li_info->li_list_mtx);
2888}
2889
2890static int ext4_run_lazyinit_thread(void)
2891{
8f1f7453
ES		 2892 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2893 ext4_li_info, "ext4lazyinit");
2894 if (IS_ERR(ext4_lazyinit_task)) {
2895 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2896 ext4_clear_request_list();
bfff6873
LC		 2897 kfree(ext4_li_info);
2898 ext4_li_info = NULL;
92b97816 2899 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2900 "initialization thread\n",
2901 err);
2902 return err;
2903 }
2904 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2905 return 0;
2906}
2907
2908/*
2909 * Check whether it make sense to run itable init. thread or not.
2910 * If there is at least one uninitialized inode table, return
2911 * corresponding group number, else the loop goes through all
2912 * groups and return total number of groups.
2913 */
2914static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2915{
2916 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2917 struct ext4_group_desc *gdp = NULL;
2918
2919 for (group = 0; group < ngroups; group++) {
2920 gdp = ext4_get_group_desc(sb, group, NULL);
2921 if (!gdp)
2922 continue;
2923
2924 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2925 break;
2926 }
2927
2928 return group;
2929}
2930
2931static int ext4_li_info_new(void)
2932{
2933 struct ext4_lazy_init *eli = NULL;
2934
2935 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2936 if (!eli)
2937 return -ENOMEM;
2938
bfff6873
LC		 2939 INIT_LIST_HEAD(&eli->li_request_list);
2940 mutex_init(&eli->li_list_mtx);
2941
bfff6873
LC		 2942 eli->li_state |= EXT4_LAZYINIT_QUIT;
2943
2944 ext4_li_info = eli;
2945
2946 return 0;
2947}
2948
2949static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2950 ext4_group_t start)
2951{
2952 struct ext4_sb_info *sbi = EXT4_SB(sb);
2953 struct ext4_li_request *elr;
2954 unsigned long rnd;
2955
2956 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2957 if (!elr)
2958 return NULL;
2959
2960 elr->lr_super = sb;
2961 elr->lr_sbi = sbi;
2962 elr->lr_next_group = start;
2963
2964 /*
2965 * Randomize first schedule time of the request to
2966 * spread the inode table initialization requests
2967 * better.
2968 */
2969 get_random_bytes(&rnd, sizeof(rnd));
2970 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2971 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2972
2973 return elr;
2974}
2975
2976static int ext4_register_li_request(struct super_block *sb,
2977 ext4_group_t first_not_zeroed)
2978{
2979 struct ext4_sb_info *sbi = EXT4_SB(sb);
2980 struct ext4_li_request *elr;
2981 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2982 int ret = 0;
bfff6873 2983
51ce6511
LC		 2984 if (sbi->s_li_request != NULL) {
2985 /*
2986 * Reset timeout so it can be computed again, because
2987 * s_li_wait_mult might have changed.
2988 */
2989 sbi->s_li_request->lr_timeout = 0;
beed5ecb 2990 return 0;
51ce6511 2991 }
bfff6873
LC		 2992
2993 if (first_not_zeroed == ngroups ||
2994 (sb->s_flags & MS_RDONLY) ||
55ff3840 2995 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 2996 return 0;
bfff6873
LC		 2997
2998 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 2999 if (!elr)
3000 return -ENOMEM;
bfff6873
LC		 3001
3002 mutex_lock(&ext4_li_mtx);
3003
3004 if (NULL == ext4_li_info) {
3005 ret = ext4_li_info_new();
3006 if (ret)
3007 goto out;
3008 }
3009
3010 mutex_lock(&ext4_li_info->li_list_mtx);
3011 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3012 mutex_unlock(&ext4_li_info->li_list_mtx);
3013
3014 sbi->s_li_request = elr;
46e4690b
TM		 3015 /*
3016 * set elr to NULL here since it has been inserted to
3017 * the request_list and the removal and free of it is
3018 * handled by ext4_clear_request_list from now on.
3019 */
3020 elr = NULL;
bfff6873
LC		 3021
3022 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3023 ret = ext4_run_lazyinit_thread();
3024 if (ret)
3025 goto out;
3026 }
bfff6873 3027out:
beed5ecb
NK		 3028 mutex_unlock(&ext4_li_mtx);
3029 if (ret)
bfff6873 3030 kfree(elr);
bfff6873
LC		 3031 return ret;
3032}
3033
3034/*
3035 * We do not need to lock anything since this is called on
3036 * module unload.
3037 */
3038static void ext4_destroy_lazyinit_thread(void)
3039{
3040 /*
3041 * If thread exited earlier
3042 * there's nothing to be done.
3043 */
8f1f7453 3044 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3045 return;
3046
8f1f7453 3047 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3048}
3049
25ed6e8a
DW		 3050static int set_journal_csum_feature_set(struct super_block *sb)
3051{
3052 int ret = 1;
3053 int compat, incompat;
3054 struct ext4_sb_info *sbi = EXT4_SB(sb);
3055
3056 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3057 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3058 /* journal checksum v2 */
3059 compat = 0;
3060 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3061 } else {
3062 /* journal checksum v1 */
3063 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3064 incompat = 0;
3065 }
3066
3067 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3068 ret = jbd2_journal_set_features(sbi->s_journal,
3069 compat, 0,
3070 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3071 incompat);
3072 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3073 ret = jbd2_journal_set_features(sbi->s_journal,
3074 compat, 0,
3075 incompat);
3076 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3077 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3078 } else {
3079 jbd2_journal_clear_features(sbi->s_journal,
3080 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3081 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3082 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3083 }
3084
3085 return ret;
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 3123 sb->s_fs_info = sbi;
2c0544b2 3124 sbi->s_sb = sb;
ac27a0ec 3125 sbi->s_mount_opt = 0;
08cefc7a
EB		 3126 sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3127 sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
240799cd 3128 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3129 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3130 if (sb->s_bdev->bd_part)
3131 sbi->s_sectors_written_start =
3132 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3133
9f6200bb
TT		 3134 /* Cleanup superblock name */
3135 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3136 *cp = '!';
3137
dcc7dae3 3138 ret = -EINVAL;
617ba13b 3139 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3140 if (!blocksize) {
b31e1552 3141 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3142 goto out_fail;
3143 }
3144
3145 /*
617ba13b 3146 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3147 * block sizes. We need to calculate the offset from buffer start.
3148 */
617ba13b 3149 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3150 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3151 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3152 } else {
70bbb3e0 3153 logical_sb_block = sb_block;
ac27a0ec
DK		 3154 }
3155
70bbb3e0 3156 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3157 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3158 goto out_fail;
3159 }
3160 /*
3161 * Note: s_es must be initialized as soon as possible because
617ba13b 3162 * some ext4 macro-instructions depend on its value
ac27a0ec 3163 */
2716b802 3164 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3165 sbi->s_es = es;
3166 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3167 if (sb->s_magic != EXT4_SUPER_MAGIC)
3168 goto cantfind_ext4;
afc32f7e 3169 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3170
feb0ab32
DW		 3171 /* Warn if metadata_csum and gdt_csum are both set. */
3172 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3173 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3174 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3175 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3176 "redundant flags; please run fsck.");
3177
d25425f8
DW		 3178 /* Check for a known checksum algorithm */
3179 if (!ext4_verify_csum_type(sb, es)) {
3180 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3181 "unknown checksum algorithm.");
3182 silent = 1;
3183 goto cantfind_ext4;
3184 }
3185
0441984a
DW		 3186 /* Load the checksum driver */
3187 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3188 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3189 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3190 if (IS_ERR(sbi->s_chksum_driver)) {
3191 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3192 ret = PTR_ERR(sbi->s_chksum_driver);
3193 sbi->s_chksum_driver = NULL;
3194 goto failed_mount;
3195 }
3196 }
3197
a9c47317
DW		 3198 /* Check superblock checksum */
3199 if (!ext4_superblock_csum_verify(sb, es)) {
3200 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3201 "invalid superblock checksum. Run e2fsck?");
3202 silent = 1;
3203 goto cantfind_ext4;
3204 }
3205
3206 /* Precompute checksum seed for all metadata */
3207 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3208 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3209 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3210 sizeof(es->s_uuid));
3211
ac27a0ec
DK		 3212 /* Set defaults before we parse the mount options */
3213 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3214 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3215 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3216 set_opt(sb, DEBUG);
87f26807 3217 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3218 set_opt(sb, GRPID);
617ba13b 3219 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3220 set_opt(sb, NO_UID32);
ea663336 3221 /* xattr user namespace & acls are now defaulted on */
03010a33 3222#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3223 set_opt(sb, XATTR_USER);
2e7842b8 3224#endif
03010a33 3225#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3226 set_opt(sb, POSIX_ACL);
2e7842b8 3227#endif
6fd7a467 3228 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3229 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3230 set_opt(sb, JOURNAL_DATA);
617ba13b 3231 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3232 set_opt(sb, ORDERED_DATA);
617ba13b 3233 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3234 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3235
3236 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3237 set_opt(sb, ERRORS_PANIC);
bb4f397a 3238 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3239 set_opt(sb, ERRORS_CONT);
bb4f397a 3240 else
fd8c37ec 3241 set_opt(sb, ERRORS_RO);
8b67f04a 3242 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3243 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3244 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3245 set_opt(sb, DISCARD);
ac27a0ec 3246
08cefc7a
EB		 3247 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3248 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3249 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3250 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3251 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3252
8b67f04a 3253 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3254 set_opt(sb, BARRIER);
ac27a0ec 3255
dd919b98
AK		 3256 /*
3257 * enable delayed allocation by default
3258 * Use -o nodelalloc to turn it off
3259 */
8b67f04a
TT		 3260 if (!IS_EXT3_SB(sb) &&
3261 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3262 set_opt(sb, DELALLOC);
dd919b98 3263
51ce6511
LC		 3264 /*
3265 * set default s_li_wait_mult for lazyinit, for the case there is
3266 * no mount option specified.
3267 */
3268 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3269
8b67f04a 3270 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3271 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3272 ext4_msg(sb, KERN_WARNING,
3273 "failed to parse options in superblock: %s",
3274 sbi->s_es->s_mount_opts);
3275 }
5a916be1 3276 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3277 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3278 &journal_ioprio, 0))
ac27a0ec
DK		 3279 goto failed_mount;
3280
56889787
TT		 3281 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3282 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3283 "with data=journal disables delayed "
3284 "allocation and O_DIRECT support!\n");
3285 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3286 ext4_msg(sb, KERN_ERR, "can't mount with "
3287 "both data=journal and delalloc");
3288 goto failed_mount;
3289 }
3290 if (test_opt(sb, DIOREAD_NOLOCK)) {
3291 ext4_msg(sb, KERN_ERR, "can't mount with "
3292 "both data=journal and delalloc");
3293 goto failed_mount;
3294 }
3295 if (test_opt(sb, DELALLOC))
3296 clear_opt(sb, DELALLOC);
3297 }
3298
3299 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3300 if (test_opt(sb, DIOREAD_NOLOCK)) {
3301 if (blocksize < PAGE_SIZE) {
3302 ext4_msg(sb, KERN_ERR, "can't mount with "
3303 "dioread_nolock if block size != PAGE_SIZE");
3304 goto failed_mount;
3305 }
3306 }
3307
ac27a0ec 3308 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3309 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3310
617ba13b
MC		 3311 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3312 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3313 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3314 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3315 ext4_msg(sb, KERN_WARNING,
3316 "feature flags set on rev 0 fs, "
3317 "running e2fsck is recommended");
469108ff 3318
2035e776
TT		 3319 if (IS_EXT2_SB(sb)) {
3320 if (ext2_feature_set_ok(sb))
3321 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3322 "using the ext4 subsystem");
3323 else {
3324 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3325 "to feature incompatibilities");
3326 goto failed_mount;
3327 }
3328 }
3329
3330 if (IS_EXT3_SB(sb)) {
3331 if (ext3_feature_set_ok(sb))
3332 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3333 "using the ext4 subsystem");
3334 else {
3335 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3336 "to feature incompatibilities");
3337 goto failed_mount;
3338 }
3339 }
3340
ac27a0ec
DK		 3341 /*
3342 * Check feature flags regardless of the revision level, since we
3343 * previously didn't change the revision level when setting the flags,
3344 * so there is a chance incompat flags are set on a rev 0 filesystem.
3345 */
a13fb1a4 3346 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3347 goto failed_mount;
a13fb1a4 3348
617ba13b
MC		 3349 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3350 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3351 ext4_msg(sb, KERN_ERR,
3352 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3353 goto failed_mount;
3354 }
3355
ac27a0ec 3356 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3357 /* Validate the filesystem blocksize */
3358 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3359 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3360 blocksize);
ac27a0ec
DK		 3361 goto failed_mount;
3362 }
3363
2b2d6d01 3364 brelse(bh);
70bbb3e0
AM		 3365 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3366 offset = do_div(logical_sb_block, blocksize);
3367 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3368 if (!bh) {
b31e1552
ES		 3369 ext4_msg(sb, KERN_ERR,
3370 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3371 goto failed_mount;
3372 }
2716b802 3373 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3374 sbi->s_es = es;
617ba13b 3375 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3376 ext4_msg(sb, KERN_ERR,
3377 "Magic mismatch, very weird!");
ac27a0ec
DK		 3378 goto failed_mount;
3379 }
3380 }
3381
a13fb1a4
ES		 3382 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3383 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3384 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3385 has_huge_files);
3386 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3387
617ba13b
MC		 3388 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3389 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3390 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3391 } else {
3392 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3393 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3394 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3395 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3396 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3397 ext4_msg(sb, KERN_ERR,
3398 "unsupported inode size: %d",
2b2d6d01 3399 sbi->s_inode_size);
ac27a0ec
DK		 3400 goto failed_mount;
3401 }
ef7f3835
KS		 3402 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3403 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3404 }
0b8e58a1 3405
0d1ee42f
AR		 3406 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3407 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3408 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3409 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3410 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3411 ext4_msg(sb, KERN_ERR,
3412 "unsupported descriptor size %lu",
0d1ee42f
AR		 3413 sbi->s_desc_size);
3414 goto failed_mount;
3415 }
3416 } else
3417 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3418
ac27a0ec 3419 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3420 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3421 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3422 goto cantfind_ext4;
0b8e58a1 3423
617ba13b 3424 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3425 if (sbi->s_inodes_per_block == 0)
617ba13b 3426 goto cantfind_ext4;
ac27a0ec
DK		 3427 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3428 sbi->s_inodes_per_block;
0d1ee42f 3429 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3430 sbi->s_sbh = bh;
3431 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3432 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3433 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3434
2b2d6d01 3435 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3436 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3437 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3438 i = le32_to_cpu(es->s_flags);
3439 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3440 sbi->s_hash_unsigned = 3;
3441 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3442#ifdef __CHAR_UNSIGNED__
3443 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3444 sbi->s_hash_unsigned = 3;
3445#else
3446 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3447#endif
f99b2589 3448 }
ac27a0ec 3449
281b5995
TT		 3450 /* Handle clustersize */
3451 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3452 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3453 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3454 if (has_bigalloc) {
3455 if (clustersize < blocksize) {
3456 ext4_msg(sb, KERN_ERR,
3457 "cluster size (%d) smaller than "
3458 "block size (%d)", clustersize, blocksize);
3459 goto failed_mount;
3460 }
3461 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3462 le32_to_cpu(es->s_log_block_size);
3463 sbi->s_clusters_per_group =
3464 le32_to_cpu(es->s_clusters_per_group);
3465 if (sbi->s_clusters_per_group > blocksize * 8) {
3466 ext4_msg(sb, KERN_ERR,
3467 "#clusters per group too big: %lu",
3468 sbi->s_clusters_per_group);
3469 goto failed_mount;
3470 }
3471 if (sbi->s_blocks_per_group !=
3472 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3473 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3474 "clusters per group (%lu) inconsistent",
3475 sbi->s_blocks_per_group,
3476 sbi->s_clusters_per_group);
3477 goto failed_mount;
3478 }
3479 } else {
3480 if (clustersize != blocksize) {
3481 ext4_warning(sb, "fragment/cluster size (%d) != "
3482 "block size (%d)", clustersize,
3483 blocksize);
3484 clustersize = blocksize;
3485 }
3486 if (sbi->s_blocks_per_group > blocksize * 8) {
3487 ext4_msg(sb, KERN_ERR,
3488 "#blocks per group too big: %lu",
3489 sbi->s_blocks_per_group);
3490 goto failed_mount;
3491 }
3492 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3493 sbi->s_cluster_bits = 0;
ac27a0ec 3494 }
281b5995
TT		 3495 sbi->s_cluster_ratio = clustersize / blocksize;
3496
ac27a0ec 3497 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3498 ext4_msg(sb, KERN_ERR,
3499 "#inodes per group too big: %lu",
2b2d6d01 3500 sbi->s_inodes_per_group);
ac27a0ec
DK		 3501 goto failed_mount;
3502 }
3503
bf43d84b
ES		 3504 /*
3505 * Test whether we have more sectors than will fit in sector_t,
3506 * and whether the max offset is addressable by the page cache.
3507 */
5a9ae68a 3508 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3509 ext4_blocks_count(es));
5a9ae68a 3510 if (err) {
b31e1552 3511 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3512 " too large to mount safely on this system");
ac27a0ec 3513 if (sizeof(sector_t) < 8)
90c699a9 3514 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3515 ret = err;
ac27a0ec
DK		 3516 goto failed_mount;
3517 }
3518
617ba13b
MC		 3519 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3520 goto cantfind_ext4;
e7c95593 3521
0f2ddca6
FTN		 3522 /* check blocks count against device size */
3523 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3524 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3525 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3526 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3527 ext4_blocks_count(es), blocks_count);
3528 goto failed_mount;
3529 }
3530
0b8e58a1
AD		 3531 /*
3532 * It makes no sense for the first data block to be beyond the end
3533 * of the filesystem.
3534 */
3535 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3536 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3537 "block %u is beyond end of filesystem (%llu)",
3538 le32_to_cpu(es->s_first_data_block),
3539 ext4_blocks_count(es));
e7c95593
ES		 3540 goto failed_mount;
3541 }
bd81d8ee
LV		 3542 blocks_count = (ext4_blocks_count(es) -
3543 le32_to_cpu(es->s_first_data_block) +
3544 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3545 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3546 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3547 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3548 "(block count %llu, first data block %u, "
b31e1552 3549 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3550 ext4_blocks_count(es),
3551 le32_to_cpu(es->s_first_data_block),
3552 EXT4_BLOCKS_PER_GROUP(sb));
3553 goto failed_mount;
3554 }
bd81d8ee 3555 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3556 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3557 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3558 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3559 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3560 sbi->s_group_desc = ext4_kvmalloc(db_count *
3561 sizeof(struct buffer_head *),
3562 GFP_KERNEL);
ac27a0ec 3563 if (sbi->s_group_desc == NULL) {
b31e1552 3564 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3565 ret = -ENOMEM;
ac27a0ec
DK		 3566 goto failed_mount;
3567 }
3568
9f6200bb
TT		 3569 if (ext4_proc_root)
3570 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3571
66acdcf4
TT		 3572 if (sbi->s_proc)
3573 proc_create_data("options", S_IRUGO, sbi->s_proc,
3574 &ext4_seq_options_fops, sb);
3575
705895b6 3576 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3577
3578 for (i = 0; i < db_count; i++) {
70bbb3e0 3579 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3580 sbi->s_group_desc[i] = sb_bread(sb, block);
3581 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3582 ext4_msg(sb, KERN_ERR,
3583 "can't read group descriptor %d", i);
ac27a0ec
DK		 3584 db_count = i;
3585 goto failed_mount2;
3586 }
3587 }
bfff6873 3588 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3589 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3590 goto failed_mount2;
3591 }
772cb7c8
JS		 3592 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3593 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3594 ext4_msg(sb, KERN_ERR,
3595 "unable to initialize "
3596 "flex_bg meta info!");
772cb7c8
JS		 3597 goto failed_mount2;
3598 }
3599
ac27a0ec
DK		 3600 sbi->s_gdb_count = db_count;
3601 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3602 spin_lock_init(&sbi->s_next_gen_lock);
3603
04496411
TM		 3604 init_timer(&sbi->s_err_report);
3605 sbi->s_err_report.function = print_daily_error_info;
3606 sbi->s_err_report.data = (unsigned long) sb;
3607
57042651 3608 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3609 ext4_count_free_clusters(sb));
ce7e010a
TT		 3610 if (!err) {
3611 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3612 ext4_count_free_inodes(sb));
3613 }
3614 if (!err) {
3615 err = percpu_counter_init(&sbi->s_dirs_counter,
3616 ext4_count_dirs(sb));
3617 }
3618 if (!err) {
57042651 3619 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3620 }
3621 if (err) {
3622 ext4_msg(sb, KERN_ERR, "insufficient memory");
2cde417d 3623 ret = err;
ce7e010a
TT		 3624 goto failed_mount3;
3625 }
3626
c9de560d 3627 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3628 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3629
ac27a0ec
DK		 3630 /*
3631 * set up enough so that it can read an inode
3632 */
9ca92389
TT		 3633 if (!test_opt(sb, NOLOAD) &&
3634 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3635 sb->s_op = &ext4_sops;
3636 else
3637 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3638 sb->s_export_op = &ext4_export_ops;
3639 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3640#ifdef CONFIG_QUOTA
617ba13b
MC		 3641 sb->s_qcop = &ext4_qctl_operations;
3642 sb->dq_op = &ext4_quota_operations;
ac27a0ec 3643#endif
f2fa2ffc
AK		 3644 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3645
ac27a0ec 3646 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3647 mutex_init(&sbi->s_orphan_lock);
8f82f840 3648 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3649
3650 sb->s_root = NULL;
3651
3652 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3653 EXT4_HAS_INCOMPAT_FEATURE(sb,
3654 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3655
c5e06d10
JL		 3656 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3657 !(sb->s_flags & MS_RDONLY))
3658 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3659 goto failed_mount3;
3660
ac27a0ec
DK		 3661 /*
3662 * The first inode we look at is the journal inode. Don't try
3663 * root first: it may be modified in the journal!
3664 */
3665 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3666 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3667 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3668 goto failed_mount3;
0390131b
FM		 3669 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3670 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3671 ext4_msg(sb, KERN_ERR, "required journal recovery "
3672 "suppressed and not mounted read-only");
744692dc 3673 goto failed_mount_wq;
ac27a0ec 3674 } else {
fd8c37ec 3675 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3676 sbi->s_journal = NULL;
3677 needs_recovery = 0;
3678 goto no_journal;
ac27a0ec
DK		 3679 }
3680
f32aaf2d 3681 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3682 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3683 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3684 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3685 goto failed_mount_wq;
eb40a09c
JS		 3686 }
3687
25ed6e8a
DW		 3688 if (!set_journal_csum_feature_set(sb)) {
3689 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3690 "feature set");
3691 goto failed_mount_wq;
d4da6c9c 3692 }
818d276c 3693
ac27a0ec
DK		 3694 /* We have now updated the journal if required, so we can
3695 * validate the data journaling mode. */
3696 switch (test_opt(sb, DATA_FLAGS)) {
3697 case 0:
3698 /* No mode set, assume a default based on the journal
63f57933
AM		 3699 * capabilities: ORDERED_DATA if the journal can
3700 * cope, else JOURNAL_DATA
3701 */
dab291af
MC		 3702 if (jbd2_journal_check_available_features
3703 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3704 set_opt(sb, ORDERED_DATA);
ac27a0ec 3705 else
fd8c37ec 3706 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3707 break;
3708
617ba13b
MC		 3709 case EXT4_MOUNT_ORDERED_DATA:
3710 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3711 if (!jbd2_journal_check_available_features
3712 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3713 ext4_msg(sb, KERN_ERR, "Journal does not support "
3714 "requested data journaling mode");
744692dc 3715 goto failed_mount_wq;
ac27a0ec
DK		 3716 }
3717 default:
3718 break;
3719 }
b3881f74 3720 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3721
18aadd47
BJ		 3722 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3723
ce7e010a
TT		 3724 /*
3725 * The journal may have updated the bg summary counts, so we
3726 * need to update the global counters.
3727 */
57042651 3728 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3729 ext4_count_free_clusters(sb));
ce7e010a
TT		 3730 percpu_counter_set(&sbi->s_freeinodes_counter,
3731 ext4_count_free_inodes(sb));
3732 percpu_counter_set(&sbi->s_dirs_counter,
3733 ext4_count_dirs(sb));
57042651 3734 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3735
ce7e010a 3736no_journal:
fd89d5f2
TH		 3737 /*
3738 * The maximum number of concurrent works can be high and
3739 * concurrency isn't really necessary. Limit it to 1.
3740 */
3741 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3742 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3743 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3744 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3745 goto failed_mount_wq;
3746 }
3747
ac27a0ec 3748 /*
dab291af 3749 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3750 * so we can safely mount the rest of the filesystem now.
3751 */
3752
1d1fe1ee
DH		 3753 root = ext4_iget(sb, EXT4_ROOT_INO);
3754 if (IS_ERR(root)) {
b31e1552 3755 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3756 ret = PTR_ERR(root);
32a9bb57 3757 root = NULL;
ac27a0ec
DK		 3758 goto failed_mount4;
3759 }
3760 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3761 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3762 iput(root);
ac27a0ec
DK		 3763 goto failed_mount4;
3764 }
48fde701 3765 sb->s_root = d_make_root(root);
1d1fe1ee 3766 if (!sb->s_root) {
b31e1552 3767 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3768 ret = -ENOMEM;
3769 goto failed_mount4;
3770 }
ac27a0ec 3771
7e84b621
ES		 3772 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3773 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3774
3775 /* determine the minimum size of new large inodes, if present */
3776 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3777 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3778 EXT4_GOOD_OLD_INODE_SIZE;
3779 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3780 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3781 if (sbi->s_want_extra_isize <
3782 le16_to_cpu(es->s_want_extra_isize))
3783 sbi->s_want_extra_isize =
3784 le16_to_cpu(es->s_want_extra_isize);
3785 if (sbi->s_want_extra_isize <
3786 le16_to_cpu(es->s_min_extra_isize))
3787 sbi->s_want_extra_isize =
3788 le16_to_cpu(es->s_min_extra_isize);
3789 }
3790 }
3791 /* Check if enough inode space is available */
3792 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3793 sbi->s_inode_size) {
3794 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3795 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3796 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3797 "available");
ef7f3835
KS		 3798 }
3799
6fd058f7
TT		 3800 err = ext4_setup_system_zone(sb);
3801 if (err) {
b31e1552 3802 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3803 "zone (%d)", err);
94bf608a 3804 goto failed_mount4a;
6fd058f7
TT		 3805 }
3806
c2774d84 3807 ext4_ext_init(sb);
9d99012f 3808 err = ext4_mb_init(sb);
c2774d84 3809 if (err) {
421f91d2 3810 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3811 err);
dcf2d804 3812 goto failed_mount5;
c2774d84
AK		 3813 }
3814
bfff6873
LC		 3815 err = ext4_register_li_request(sb, first_not_zeroed);
3816 if (err)
dcf2d804 3817 goto failed_mount6;
bfff6873 3818
3197ebdb
TT		 3819 sbi->s_kobj.kset = ext4_kset;
3820 init_completion(&sbi->s_kobj_unregister);
3821 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3822 "%s", sb->s_id);
dcf2d804
TM		 3823 if (err)
3824 goto failed_mount7;
3197ebdb 3825
617ba13b
MC		 3826 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3827 ext4_orphan_cleanup(sb, es);
3828 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3829 if (needs_recovery) {
b31e1552 3830 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3831 ext4_mark_recovery_complete(sb, es);
3832 }
3833 if (EXT4_SB(sb)->s_journal) {
3834 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3835 descr = " journalled data mode";
3836 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3837 descr = " ordered data mode";
3838 else
3839 descr = " writeback data mode";
3840 } else
3841 descr = "out journal";
3842
d4c402d9 3843 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3844 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3845 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3846
66e61a9e
TT		 3847 if (es->s_error_count)
3848 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3849
d4c402d9 3850 kfree(orig_data);
ac27a0ec
DK		 3851 return 0;
3852
617ba13b 3853cantfind_ext4:
ac27a0ec 3854 if (!silent)
b31e1552 3855 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3856 goto failed_mount;
3857
dcf2d804
TM		 3858failed_mount7:
3859 ext4_unregister_li_request(sb);
3860failed_mount6:
dcf2d804 3861 ext4_mb_release(sb);
94bf608a
AV		 3862failed_mount5:
3863 ext4_ext_release(sb);
dcf2d804 3864 ext4_release_system_zone(sb);
94bf608a
AV		 3865failed_mount4a:
3866 dput(sb->s_root);
32a9bb57 3867 sb->s_root = NULL;
94bf608a 3868failed_mount4:
b31e1552 3869 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3870 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3871failed_mount_wq:
0390131b
FM		 3872 if (sbi->s_journal) {
3873 jbd2_journal_destroy(sbi->s_journal);
3874 sbi->s_journal = NULL;
3875 }
ac27a0ec 3876failed_mount3:
04496411 3877 del_timer(&sbi->s_err_report);
9933fc0a
TT		 3878 if (sbi->s_flex_groups)
3879 ext4_kvfree(sbi->s_flex_groups);
57042651 3880 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 3881 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3882 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 3883 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 3884 if (sbi->s_mmp_tsk)
3885 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3886failed_mount2:
3887 for (i = 0; i < db_count; i++)
3888 brelse(sbi->s_group_desc[i]);
f18a5f21 3889 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 3890failed_mount:
0441984a
DW		 3891 if (sbi->s_chksum_driver)
3892 crypto_free_shash(sbi->s_chksum_driver);
240799cd 3893 if (sbi->s_proc) {
66acdcf4 3894 remove_proc_entry("options", sbi->s_proc);
9f6200bb 3895 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3896 }
ac27a0ec
DK		 3897#ifdef CONFIG_QUOTA
3898 for (i = 0; i < MAXQUOTAS; i++)
3899 kfree(sbi->s_qf_names[i]);
3900#endif
617ba13b 3901 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3902 brelse(bh);
3903out_fail:
3904 sb->s_fs_info = NULL;
f6830165 3905 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3906 kfree(sbi);
dcc7dae3 3907out_free_orig:
d4c402d9 3908 kfree(orig_data);
1d1fe1ee 3909 return ret;
ac27a0ec
DK		 3910}
3911
3912/*
3913 * Setup any per-fs journal parameters now. We'll do this both on
3914 * initial mount, once the journal has been initialised but before we've
3915 * done any recovery; and again on any subsequent remount.
3916 */
617ba13b 3917static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3918{
617ba13b 3919 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3920
30773840
TT		 3921 journal->j_commit_interval = sbi->s_commit_interval;
3922 journal->j_min_batch_time = sbi->s_min_batch_time;
3923 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3924
a931da6a 3925 write_lock(&journal->j_state_lock);
ac27a0ec 3926 if (test_opt(sb, BARRIER))
dab291af 3927 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3928 else
dab291af 3929 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3930 if (test_opt(sb, DATA_ERR_ABORT))
3931 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3932 else
3933 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3934 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3935}
3936
617ba13b 3937static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3938 unsigned int journal_inum)
3939{
3940 struct inode *journal_inode;
3941 journal_t *journal;
3942
0390131b
FM		 3943 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3944
ac27a0ec
DK		 3945 /* First, test for the existence of a valid inode on disk. Bad
3946 * things happen if we iget() an unused inode, as the subsequent
3947 * iput() will try to delete it. */
3948
1d1fe1ee
DH		 3949 journal_inode = ext4_iget(sb, journal_inum);
3950 if (IS_ERR(journal_inode)) {
b31e1552 3951 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3952 return NULL;
3953 }
3954 if (!journal_inode->i_nlink) {
3955 make_bad_inode(journal_inode);
3956 iput(journal_inode);
b31e1552 3957 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3958 return NULL;
3959 }
3960
e5f8eab8 3961 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3962 journal_inode, journal_inode->i_size);
1d1fe1ee 3963 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3964 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3965 iput(journal_inode);
3966 return NULL;
3967 }
3968
dab291af 3969 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3970 if (!journal) {
b31e1552 3971 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3972 iput(journal_inode);
3973 return NULL;
3974 }
3975 journal->j_private = sb;
617ba13b 3976 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3977 return journal;
3978}
3979
617ba13b 3980static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3981 dev_t j_dev)
3982{
2b2d6d01 3983 struct buffer_head *bh;
ac27a0ec 3984 journal_t *journal;
617ba13b
MC		 3985 ext4_fsblk_t start;
3986 ext4_fsblk_t len;
ac27a0ec 3987 int hblock, blocksize;
617ba13b 3988 ext4_fsblk_t sb_block;
ac27a0ec 3989 unsigned long offset;
2b2d6d01 3990 struct ext4_super_block *es;
ac27a0ec
DK		 3991 struct block_device *bdev;
3992
0390131b
FM		 3993 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3994
b31e1552 3995 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3996 if (bdev == NULL)
3997 return NULL;
3998
ac27a0ec 3999 blocksize = sb->s_blocksize;
e1defc4f 4000 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4001 if (blocksize < hblock) {
b31e1552
ES		 4002 ext4_msg(sb, KERN_ERR,
4003 "blocksize too small for journal device");
ac27a0ec
DK		 4004 goto out_bdev;
4005 }
4006
617ba13b
MC		 4007 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4008 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4009 set_blocksize(bdev, blocksize);
4010 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4011 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4012 "external journal");
ac27a0ec
DK		 4013 goto out_bdev;
4014 }
4015
2716b802 4016 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4017 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4018 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4019 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4020 ext4_msg(sb, KERN_ERR, "external journal has "
4021 "bad superblock");
ac27a0ec
DK		 4022 brelse(bh);
4023 goto out_bdev;
4024 }
4025
617ba13b 4026 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4027 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4028 brelse(bh);
4029 goto out_bdev;
4030 }
4031
bd81d8ee 4032 len = ext4_blocks_count(es);
ac27a0ec
DK		 4033 start = sb_block + 1;
4034 brelse(bh); /* we're done with the superblock */
4035
dab291af 4036 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4037 start, len, blocksize);
4038 if (!journal) {
b31e1552 4039 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4040 goto out_bdev;
4041 }
4042 journal->j_private = sb;
4043 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4044 wait_on_buffer(journal->j_sb_buffer);
4045 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4046 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4047 goto out_journal;
4048 }
4049 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4050 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4051 "user (unsupported) - %d",
ac27a0ec
DK		 4052 be32_to_cpu(journal->j_superblock->s_nr_users));
4053 goto out_journal;
4054 }
617ba13b
MC		 4055 EXT4_SB(sb)->journal_bdev = bdev;
4056 ext4_init_journal_params(sb, journal);
ac27a0ec 4057 return journal;
0b8e58a1 4058
ac27a0ec 4059out_journal:
dab291af 4060 jbd2_journal_destroy(journal);
ac27a0ec 4061out_bdev:
617ba13b 4062 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4063 return NULL;
4064}
4065
617ba13b
MC		 4066static int ext4_load_journal(struct super_block *sb,
4067 struct ext4_super_block *es,
ac27a0ec
DK		 4068 unsigned long journal_devnum)
4069{
4070 journal_t *journal;
4071 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4072 dev_t journal_dev;
4073 int err = 0;
4074 int really_read_only;
4075
0390131b
FM		 4076 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4077
ac27a0ec
DK		 4078 if (journal_devnum &&
4079 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4080 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4081 "numbers have changed");
ac27a0ec
DK		 4082 journal_dev = new_decode_dev(journal_devnum);
4083 } else
4084 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4085
4086 really_read_only = bdev_read_only(sb->s_bdev);
4087
4088 /*
4089 * Are we loading a blank journal or performing recovery after a
4090 * crash? For recovery, we need to check in advance whether we
4091 * can get read-write access to the device.
4092 */
617ba13b 4093 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4094 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4095 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4096 "required on readonly filesystem");
ac27a0ec 4097 if (really_read_only) {
b31e1552
ES		 4098 ext4_msg(sb, KERN_ERR, "write access "
4099 "unavailable, cannot proceed");
ac27a0ec
DK		 4100 return -EROFS;
4101 }
b31e1552
ES		 4102 ext4_msg(sb, KERN_INFO, "write access will "
4103 "be enabled during recovery");
ac27a0ec
DK		 4104 }
4105 }
4106
4107 if (journal_inum && journal_dev) {
b31e1552
ES		 4108 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4109 "and inode journals!");
ac27a0ec
DK		 4110 return -EINVAL;
4111 }
4112
4113 if (journal_inum) {
617ba13b 4114 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4115 return -EINVAL;
4116 } else {
617ba13b 4117 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4118 return -EINVAL;
4119 }
4120
90576c0b 4121 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4122 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4123
617ba13b 4124 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4125 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4126 if (!err) {
4127 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4128 if (save)
4129 memcpy(save, ((char *) es) +
4130 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4131 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4132 if (save)
4133 memcpy(((char *) es) + EXT4_S_ERR_START,
4134 save, EXT4_S_ERR_LEN);
4135 kfree(save);
4136 }
ac27a0ec
DK		 4137
4138 if (err) {
b31e1552 4139 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4140 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4141 return err;
4142 }
4143
617ba13b
MC		 4144 EXT4_SB(sb)->s_journal = journal;
4145 ext4_clear_journal_err(sb, es);
ac27a0ec 4146
c41303ce 4147 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4148 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4149 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4150
4151 /* Make sure we flush the recovery flag to disk. */
e2d67052 4152 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4153 }
4154
4155 return 0;
4156}
4157
e2d67052 4158static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4159{
e2d67052 4160 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4161 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4162 int error = 0;
ac27a0ec 4163
7c2e7087 4164 if (!sbh || block_device_ejected(sb))
c4be0c1d 4165 return error;
914258bf
TT		 4166 if (buffer_write_io_error(sbh)) {
4167 /*
4168 * Oh, dear. A previous attempt to write the
4169 * superblock failed. This could happen because the
4170 * USB device was yanked out. Or it could happen to
4171 * be a transient write error and maybe the block will
4172 * be remapped. Nothing we can do but to retry the
4173 * write and hope for the best.
4174 */
b31e1552
ES		 4175 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4176 "superblock detected");
914258bf
TT		 4177 clear_buffer_write_io_error(sbh);
4178 set_buffer_uptodate(sbh);
4179 }
71290b36
TT		 4180 /*
4181 * If the file system is mounted read-only, don't update the
4182 * superblock write time. This avoids updating the superblock
4183 * write time when we are mounting the root file system
4184 * read/only but we need to replay the journal; at that point,
4185 * for people who are east of GMT and who make their clock
4186 * tick in localtime for Windows bug-for-bug compatibility,
4187 * the clock is set in the future, and this will cause e2fsck
4188 * to complain and force a full file system check.
4189 */
4190 if (!(sb->s_flags & MS_RDONLY))
4191 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4192 if (sb->s_bdev->bd_part)
4193 es->s_kbytes_written =
4194 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4195 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4196 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4197 else
4198 es->s_kbytes_written =
4199 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4200 ext4_free_blocks_count_set(es,
4201 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4202 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4203 es->s_free_inodes_count =
4204 cpu_to_le32(percpu_counter_sum_positive(
4205 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 4206 sb->s_dirt = 0;
ac27a0ec 4207 BUFFER_TRACE(sbh, "marking dirty");
a9c47317 4208 ext4_superblock_csum_set(sb, es);
ac27a0ec 4209 mark_buffer_dirty(sbh);
914258bf 4210 if (sync) {
c4be0c1d
TS		 4211 error = sync_dirty_buffer(sbh);
4212 if (error)
4213 return error;
4214
4215 error = buffer_write_io_error(sbh);
4216 if (error) {
b31e1552
ES		 4217 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4218 "superblock");
914258bf
TT		 4219 clear_buffer_write_io_error(sbh);
4220 set_buffer_uptodate(sbh);
4221 }
4222 }
c4be0c1d 4223 return error;
ac27a0ec
DK		 4224}
4225
ac27a0ec
DK		 4226/*
4227 * Have we just finished recovery? If so, and if we are mounting (or
4228 * remounting) the filesystem readonly, then we will end up with a
4229 * consistent fs on disk. Record that fact.
4230 */
2b2d6d01
TT		 4231static void ext4_mark_recovery_complete(struct super_block *sb,
4232 struct ext4_super_block *es)
ac27a0ec 4233{
617ba13b 4234 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4235
0390131b
FM		 4236 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4237 BUG_ON(journal != NULL);
4238 return;
4239 }
dab291af 4240 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4241 if (jbd2_journal_flush(journal) < 0)
4242 goto out;
4243
617ba13b 4244 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4245 sb->s_flags & MS_RDONLY) {
617ba13b 4246 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4247 ext4_commit_super(sb, 1);
ac27a0ec 4248 }
7ffe1ea8
HK		 4249
4250out:
dab291af 4251 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4252}
4253
4254/*
4255 * If we are mounting (or read-write remounting) a filesystem whose journal
4256 * has recorded an error from a previous lifetime, move that error to the
4257 * main filesystem now.
4258 */
2b2d6d01
TT		 4259static void ext4_clear_journal_err(struct super_block *sb,
4260 struct ext4_super_block *es)
ac27a0ec
DK		 4261{
4262 journal_t *journal;
4263 int j_errno;
4264 const char *errstr;
4265
0390131b
FM		 4266 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4267
617ba13b 4268 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4269
4270 /*
4271 * Now check for any error status which may have been recorded in the
617ba13b 4272 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4273 */
4274
dab291af 4275 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4276 if (j_errno) {
4277 char nbuf[16];
4278
617ba13b 4279 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4280 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4281 "from previous mount: %s", errstr);
12062ddd 4282 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4283
617ba13b
MC		 4284 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4285 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4286 ext4_commit_super(sb, 1);
ac27a0ec 4287
dab291af 4288 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4289 }
4290}
4291
4292/*
4293 * Force the running and committing transactions to commit,
4294 * and wait on the commit.
4295 */
617ba13b 4296int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4297{
4298 journal_t *journal;
0390131b 4299 int ret = 0;
ac27a0ec
DK		 4300
4301 if (sb->s_flags & MS_RDONLY)
4302 return 0;
4303
617ba13b 4304 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4305 if (journal) {
437f88cc 4306 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4307 ret = ext4_journal_force_commit(journal);
6b0310fb 4308 }
0390131b 4309
ac27a0ec
DK		 4310 return ret;
4311}
4312
2b2d6d01 4313static void ext4_write_super(struct super_block *sb)
ac27a0ec 4314{
ebc1ac16 4315 lock_super(sb);
9ca92389 4316 ext4_commit_super(sb, 1);
ebc1ac16 4317 unlock_super(sb);
ac27a0ec
DK		 4318}
4319
617ba13b 4320static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4321{
14ce0cb4 4322 int ret = 0;
9eddacf9 4323 tid_t target;
8d5d02e6 4324 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4325
9bffad1e 4326 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4327 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4328 /*
4329 * Writeback quota in non-journalled quota case - journalled quota has
4330 * no dirty dquots
4331 */
4332 dquot_writeback_dquots(sb, -1);
8d5d02e6 4333 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4334 if (wait)
8d5d02e6 4335 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4336 }
14ce0cb4 4337 return ret;
ac27a0ec
DK		 4338}
4339
4340/*
4341 * LVM calls this function before a (read-only) snapshot is created. This
4342 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4343 *
4344 * Note that only this function cannot bring a filesystem to be in a clean
4345 * state independently, because ext4 prevents a new handle from being started
4346 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4347 * the upper layer.
ac27a0ec 4348 */
c4be0c1d 4349static int ext4_freeze(struct super_block *sb)
ac27a0ec 4350{
c4be0c1d
TS		 4351 int error = 0;
4352 journal_t *journal;
ac27a0ec 4353
9ca92389
TT		 4354 if (sb->s_flags & MS_RDONLY)
4355 return 0;
ac27a0ec 4356
9ca92389 4357 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4358
9ca92389
TT		 4359 /* Now we set up the journal barrier. */
4360 jbd2_journal_lock_updates(journal);
ac27a0ec 4361
9ca92389
TT		 4362 /*
4363 * Don't clear the needs_recovery flag if we failed to flush
4364 * the journal.
4365 */
4366 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4367 if (error < 0)
4368 goto out;
9ca92389
TT		 4369
4370 /* Journal blocked and flushed, clear needs_recovery flag. */
4371 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4372 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4373out:
4374 /* we rely on s_frozen to stop further updates */
4375 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4376 return error;
ac27a0ec
DK		 4377}
4378
4379/*
4380 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4381 * flag here, even though the filesystem is not technically dirty yet.
4382 */
c4be0c1d 4383static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4384{
9ca92389
TT		 4385 if (sb->s_flags & MS_RDONLY)
4386 return 0;
4387
4388 lock_super(sb);
4389 /* Reset the needs_recovery flag before the fs is unlocked. */
4390 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4391 ext4_commit_super(sb, 1);
4392 unlock_super(sb);
c4be0c1d 4393 return 0;
ac27a0ec
DK		 4394}
4395
673c6100
TT		 4396/*
4397 * Structure to save mount options for ext4_remount's benefit
4398 */
4399struct ext4_mount_options {
4400 unsigned long s_mount_opt;
a2595b8a 4401 unsigned long s_mount_opt2;
08cefc7a
EB		 4402 kuid_t s_resuid;
4403 kgid_t s_resgid;
673c6100
TT		 4404 unsigned long s_commit_interval;
4405 u32 s_min_batch_time, s_max_batch_time;
4406#ifdef CONFIG_QUOTA
4407 int s_jquota_fmt;
4408 char *s_qf_names[MAXQUOTAS];
4409#endif
4410};
4411
2b2d6d01 4412static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4413{
2b2d6d01 4414 struct ext4_super_block *es;
617ba13b 4415 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4416 unsigned long old_sb_flags;
617ba13b 4417 struct ext4_mount_options old_opts;
c79d967d 4418 int enable_quota = 0;
8a266467 4419 ext4_group_t g;
b3881f74 4420 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4421 int err = 0;
ac27a0ec
DK		 4422#ifdef CONFIG_QUOTA
4423 int i;
4424#endif
d4c402d9 4425 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4426
4427 /* Store the original options */
bbd6851a 4428 lock_super(sb);
ac27a0ec
DK		 4429 old_sb_flags = sb->s_flags;
4430 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4431 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4432 old_opts.s_resuid = sbi->s_resuid;
4433 old_opts.s_resgid = sbi->s_resgid;
4434 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4435 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4436 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4437#ifdef CONFIG_QUOTA
4438 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4439 for (i = 0; i < MAXQUOTAS; i++)
4440 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4441#endif
b3881f74
TT		 4442 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4443 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4444
4445 /*
4446 * Allow the "check" option to be passed as a remount option.
4447 */
661aa520 4448 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4449 err = -EINVAL;
4450 goto restore_opts;
4451 }
4452
4ab2f15b 4453 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4454 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4455
4456 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4457 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4458
4459 es = sbi->s_es;
4460
b3881f74 4461 if (sbi->s_journal) {
0390131b 4462 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4463 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4464 }
ac27a0ec 4465
661aa520 4466 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4467 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4468 err = -EROFS;
4469 goto restore_opts;
4470 }
4471
4472 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4473 err = dquot_suspend(sb, -1);
4474 if (err < 0)
c79d967d 4475 goto restore_opts;
c79d967d 4476
ac27a0ec
DK		 4477 /*
4478 * First of all, the unconditional stuff we have to do
4479 * to disable replay of the journal when we next remount
4480 */
4481 sb->s_flags |= MS_RDONLY;
4482
4483 /*
4484 * OK, test if we are remounting a valid rw partition
4485 * readonly, and if so set the rdonly flag and then
4486 * mark the partition as valid again.
4487 */
617ba13b
MC		 4488 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4489 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4490 es->s_state = cpu_to_le16(sbi->s_mount_state);
4491
a63c9eb2 4492 if (sbi->s_journal)
0390131b 4493 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4494 } else {
a13fb1a4
ES		 4495 /* Make sure we can mount this feature set readwrite */
4496 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4497 err = -EROFS;
4498 goto restore_opts;
4499 }
8a266467
TT		 4500 /*
4501 * Make sure the group descriptor checksums
0b8e58a1 4502 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4503 */
4504 for (g = 0; g < sbi->s_groups_count; g++) {
4505 struct ext4_group_desc *gdp =
4506 ext4_get_group_desc(sb, g, NULL);
4507
feb0ab32 4508 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4509 ext4_msg(sb, KERN_ERR,
4510 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4511 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4512 le16_to_cpu(gdp->bg_checksum));
4513 err = -EINVAL;
4514 goto restore_opts;
4515 }
4516 }
4517
ead6596b
ES		 4518 /*
4519 * If we have an unprocessed orphan list hanging
4520 * around from a previously readonly bdev mount,
4521 * require a full umount/remount for now.
4522 */
4523 if (es->s_last_orphan) {
b31e1552 4524 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4525 "remount RDWR because of unprocessed "
4526 "orphan inode list. Please "
b31e1552 4527 "umount/remount instead");
ead6596b
ES		 4528 err = -EINVAL;
4529 goto restore_opts;
4530 }
4531
ac27a0ec
DK		 4532 /*
4533 * Mounting a RDONLY partition read-write, so reread
4534 * and store the current valid flag. (It may have
4535 * been changed by e2fsck since we originally mounted
4536 * the partition.)
4537 */
0390131b
FM		 4538 if (sbi->s_journal)
4539 ext4_clear_journal_err(sb, es);
ac27a0ec 4540 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4541 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4542 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4543 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4544 EXT4_FEATURE_INCOMPAT_MMP))
4545 if (ext4_multi_mount_protect(sb,
4546 le64_to_cpu(es->s_mmp_block))) {
4547 err = -EROFS;
4548 goto restore_opts;
4549 }
c79d967d 4550 enable_quota = 1;
ac27a0ec
DK		 4551 }
4552 }
bfff6873
LC		 4553
4554 /*
4555 * Reinitialize lazy itable initialization thread based on
4556 * current settings
4557 */
4558 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4559 ext4_unregister_li_request(sb);
4560 else {
4561 ext4_group_t first_not_zeroed;
4562 first_not_zeroed = ext4_has_uninit_itable(sb);
4563 ext4_register_li_request(sb, first_not_zeroed);
4564 }
4565
6fd058f7 4566 ext4_setup_system_zone(sb);
0390131b 4567 if (sbi->s_journal == NULL)
e2d67052 4568 ext4_commit_super(sb, 1);
0390131b 4569
ac27a0ec
DK		 4570#ifdef CONFIG_QUOTA
4571 /* Release old quota file names */
4572 for (i = 0; i < MAXQUOTAS; i++)
4573 if (old_opts.s_qf_names[i] &&
4574 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4575 kfree(old_opts.s_qf_names[i]);
4576#endif
bbd6851a 4577 unlock_super(sb);
c79d967d 4578 if (enable_quota)
0f0dd62f 4579 dquot_resume(sb, -1);
d4c402d9
CW		 4580
4581 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4582 kfree(orig_data);
ac27a0ec 4583 return 0;
0b8e58a1 4584
ac27a0ec
DK		 4585restore_opts:
4586 sb->s_flags = old_sb_flags;
4587 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4588 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4589 sbi->s_resuid = old_opts.s_resuid;
4590 sbi->s_resgid = old_opts.s_resgid;
4591 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4592 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4593 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4594#ifdef CONFIG_QUOTA
4595 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4596 for (i = 0; i < MAXQUOTAS; i++) {
4597 if (sbi->s_qf_names[i] &&
4598 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4599 kfree(sbi->s_qf_names[i]);
4600 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4601 }
4602#endif
bbd6851a 4603 unlock_super(sb);
d4c402d9 4604 kfree(orig_data);
ac27a0ec
DK		 4605 return err;
4606}
4607
f975d6bc
TT		 4608/*
4609 * Note: calculating the overhead so we can be compatible with
4610 * historical BSD practice is quite difficult in the face of
4611 * clusters/bigalloc. This is because multiple metadata blocks from
4612 * different block group can end up in the same allocation cluster.
4613 * Calculating the exact overhead in the face of clustered allocation
4614 * requires either O(all block bitmaps) in memory or O(number of block
4615 * groups**2) in time. We will still calculate the superblock for
4616 * older file systems --- and if we come across with a bigalloc file
4617 * system with zero in s_overhead_clusters the estimate will be close to
4618 * correct especially for very large cluster sizes --- but for newer
4619 * file systems, it's better to calculate this figure once at mkfs
4620 * time, and store it in the superblock. If the superblock value is
4621 * present (even for non-bigalloc file systems), we will use it.
4622 */
2b2d6d01 4623static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4624{
4625 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4626 struct ext4_sb_info *sbi = EXT4_SB(sb);
4627 struct ext4_super_block *es = sbi->s_es;
f975d6bc 4628 struct ext4_group_desc *gdp;
960cc398 4629 u64 fsid;
d02a9391 4630 s64 bfree;
ac27a0ec 4631
5e70030d
BP		 4632 if (test_opt(sb, MINIX_DF)) {
4633 sbi->s_overhead_last = 0;
f975d6bc
TT		 4634 } else if (es->s_overhead_clusters) {
4635 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
6bc9feff 4636 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4637 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4638 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4639
4640 /*
5e70030d
BP		 4641 * Compute the overhead (FS structures). This is constant
4642 * for a given filesystem unless the number of block groups
4643 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4644 */
4645
4646 /*
4647 * All of the blocks before first_data_block are
4648 * overhead
4649 */
f975d6bc 4650 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
ac27a0ec
DK		 4651
4652 /*
f975d6bc 4653 * Add the overhead found in each block group
ac27a0ec
DK		 4654 */
4655 for (i = 0; i < ngroups; i++) {
f975d6bc
TT		 4656 gdp = ext4_get_group_desc(sb, i, NULL);
4657 overhead += ext4_num_overhead_clusters(sb, i, gdp);
ac27a0ec
DK		 4658 cond_resched();
4659 }
5e70030d
BP		 4660 sbi->s_overhead_last = overhead;
4661 smp_wmb();
6bc9feff 4662 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4663 }
4664
617ba13b 4665 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4666 buf->f_bsize = sb->s_blocksize;
f975d6bc
TT		 4667 buf->f_blocks = (ext4_blocks_count(es) -
4668 EXT4_C2B(sbi, sbi->s_overhead_last));
57042651
TT		 4669 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4670 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4671 /* prevent underflow in case that few free space is available */
57042651 4672 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4673 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4674 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4675 buf->f_bavail = 0;
4676 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4677 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4678 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4679 fsid = le64_to_cpup((void *)es->s_uuid) ^
4680 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4681 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4682 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4683
ac27a0ec
DK		 4684 return 0;
4685}
4686
0b8e58a1
AD		 4687/* Helper function for writing quotas on sync - we need to start transaction
4688 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4689 * Process 1 Process 2
617ba13b 4690 * ext4_create() quota_sync()
a269eb18 4691 * jbd2_journal_start() write_dquot()
871a2931 4692 * dquot_initialize() down(dqio_mutex)
dab291af 4693 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4694 *
4695 */
4696
4697#ifdef CONFIG_QUOTA
4698
4699static inline struct inode *dquot_to_inode(struct dquot *dquot)
4700{
4701 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4702}
4703
617ba13b 4704static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4705{
4706 int ret, err;
4707 handle_t *handle;
4708 struct inode *inode;
4709
4710 inode = dquot_to_inode(dquot);
617ba13b 4711 handle = ext4_journal_start(inode,
0b8e58a1 4712 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4713 if (IS_ERR(handle))
4714 return PTR_ERR(handle);
4715 ret = dquot_commit(dquot);
617ba13b 4716 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4717 if (!ret)
4718 ret = err;
4719 return ret;
4720}
4721
617ba13b 4722static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4723{
4724 int ret, err;
4725 handle_t *handle;
4726
617ba13b 4727 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4728 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4729 if (IS_ERR(handle))
4730 return PTR_ERR(handle);
4731 ret = dquot_acquire(dquot);
617ba13b 4732 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4733 if (!ret)
4734 ret = err;
4735 return ret;
4736}
4737
617ba13b 4738static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4739{
4740 int ret, err;
4741 handle_t *handle;
4742
617ba13b 4743 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4744 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4745 if (IS_ERR(handle)) {
4746 /* Release dquot anyway to avoid endless cycle in dqput() */
4747 dquot_release(dquot);
ac27a0ec 4748 return PTR_ERR(handle);
9c3013e9 4749 }
ac27a0ec 4750 ret = dquot_release(dquot);
617ba13b 4751 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4752 if (!ret)
4753 ret = err;
4754 return ret;
4755}
4756
617ba13b 4757static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4758{
2c8be6b2 4759 /* Are we journaling quotas? */
617ba13b
MC		 4760 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4761 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4762 dquot_mark_dquot_dirty(dquot);
617ba13b 4763 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4764 } else {
4765 return dquot_mark_dquot_dirty(dquot);
4766 }
4767}
4768
617ba13b 4769static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4770{
4771 int ret, err;
4772 handle_t *handle;
4773
4774 /* Data block + inode block */
617ba13b 4775 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4776 if (IS_ERR(handle))
4777 return PTR_ERR(handle);
4778 ret = dquot_commit_info(sb, type);
617ba13b 4779 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4780 if (!ret)
4781 ret = err;
4782 return ret;
4783}
4784
4785/*
4786 * Turn on quotas during mount time - we need to find
4787 * the quota file and such...
4788 */
617ba13b 4789static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4790{
287a8095
CH		 4791 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4792 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4793}
4794
4795/*
4796 * Standard function to be called on quota_on
4797 */
617ba13b 4798static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4799 struct path *path)
ac27a0ec
DK		 4800{
4801 int err;
ac27a0ec
DK		 4802
4803 if (!test_opt(sb, QUOTA))
4804 return -EINVAL;
0623543b 4805
ac27a0ec 4806 /* Quotafile not on the same filesystem? */
d8c9584e 4807 if (path->dentry->d_sb != sb)
ac27a0ec 4808 return -EXDEV;
0623543b
JK		 4809 /* Journaling quota? */
4810 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4811 /* Quotafile not in fs root? */
f00c9e44 4812 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4813 ext4_msg(sb, KERN_WARNING,
4814 "Quota file not on filesystem root. "
4815 "Journaled quota will not work");
2b2d6d01 4816 }
0623543b
JK		 4817
4818 /*
4819 * When we journal data on quota file, we have to flush journal to see
4820 * all updates to the file when we bypass pagecache...
4821 */
0390131b 4822 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4823 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4824 /*
4825 * We don't need to lock updates but journal_flush() could
4826 * otherwise be livelocked...
4827 */
4828 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4829 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4830 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4831 if (err)
7ffe1ea8 4832 return err;
0623543b
JK		 4833 }
4834
f00c9e44 4835 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4836}
4837
ca0e05e4
DM		 4838static int ext4_quota_off(struct super_block *sb, int type)
4839{
21f97697
JK		 4840 struct inode *inode = sb_dqopt(sb)->files[type];
4841 handle_t *handle;
4842
87009d86
DM		 4843 /* Force all delayed allocation blocks to be allocated.
4844 * Caller already holds s_umount sem */
4845 if (test_opt(sb, DELALLOC))
ca0e05e4 4846 sync_filesystem(sb);
ca0e05e4 4847
0b268590
AG		 4848 if (!inode)
4849 goto out;
4850
21f97697
JK		 4851 /* Update modification times of quota files when userspace can
4852 * start looking at them */
4853 handle = ext4_journal_start(inode, 1);
4854 if (IS_ERR(handle))
4855 goto out;
4856 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4857 ext4_mark_inode_dirty(handle, inode);
4858 ext4_journal_stop(handle);
4859
4860out:
ca0e05e4
DM		 4861 return dquot_quota_off(sb, type);
4862}
4863
ac27a0ec
DK		 4864/* Read data from quotafile - avoid pagecache and such because we cannot afford
4865 * acquiring the locks... As quota files are never truncated and quota code
25985edc 4866 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 4867 * we don't have to be afraid of races */
617ba13b 4868static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4869 size_t len, loff_t off)
4870{
4871 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4872 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4873 int err = 0;
4874 int offset = off & (sb->s_blocksize - 1);
4875 int tocopy;
4876 size_t toread;
4877 struct buffer_head *bh;
4878 loff_t i_size = i_size_read(inode);
4879
4880 if (off > i_size)
4881 return 0;
4882 if (off+len > i_size)
4883 len = i_size-off;
4884 toread = len;
4885 while (toread > 0) {
4886 tocopy = sb->s_blocksize - offset < toread ?
4887 sb->s_blocksize - offset : toread;
617ba13b 4888 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4889 if (err)
4890 return err;
4891 if (!bh) /* A hole? */
4892 memset(data, 0, tocopy);
4893 else
4894 memcpy(data, bh->b_data+offset, tocopy);
4895 brelse(bh);
4896 offset = 0;
4897 toread -= tocopy;
4898 data += tocopy;
4899 blk++;
4900 }
4901 return len;
4902}
4903
4904/* Write to quotafile (we know the transaction is already started and has
4905 * enough credits) */
617ba13b 4906static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4907 const char *data, size_t len, loff_t off)
4908{
4909 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4910 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4911 int err = 0;
4912 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4913 struct buffer_head *bh;
4914 handle_t *handle = journal_current_handle();
4915
0390131b 4916 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4917 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4918 " cancelled because transaction is not started",
9c3013e9
JK		 4919 (unsigned long long)off, (unsigned long long)len);
4920 return -EIO;
4921 }
67eeb568
DM		 4922 /*
4923 * Since we account only one data block in transaction credits,
4924 * then it is impossible to cross a block boundary.
4925 */
4926 if (sb->s_blocksize - offset < len) {
4927 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4928 " cancelled because not block aligned",
4929 (unsigned long long)off, (unsigned long long)len);
4930 return -EIO;
4931 }
4932
67eeb568
DM		 4933 bh = ext4_bread(handle, inode, blk, 1, &err);
4934 if (!bh)
4935 goto out;
62d2b5f2
JK		 4936 err = ext4_journal_get_write_access(handle, bh);
4937 if (err) {
4938 brelse(bh);
4939 goto out;
ac27a0ec 4940 }
67eeb568
DM		 4941 lock_buffer(bh);
4942 memcpy(bh->b_data+offset, data, len);
4943 flush_dcache_page(bh->b_page);
4944 unlock_buffer(bh);
62d2b5f2 4945 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4946 brelse(bh);
ac27a0ec 4947out:
0b7f7cef 4948 if (err)
ac27a0ec 4949 return err;
67eeb568
DM		 4950 if (inode->i_size < off + len) {
4951 i_size_write(inode, off + len);
617ba13b 4952 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 4953 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4954 }
67eeb568 4955 return len;
ac27a0ec
DK		 4956}
4957
4958#endif
4959
152a0836
AV		 4960static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4961 const char *dev_name, void *data)
ac27a0ec 4962{
152a0836 4963 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4964}
4965
37f328eb 4966#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4967static inline void register_as_ext2(void)
4968{
4969 int err = register_filesystem(&ext2_fs_type);
4970 if (err)
4971 printk(KERN_WARNING
4972 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4973}
4974
4975static inline void unregister_as_ext2(void)
4976{
4977 unregister_filesystem(&ext2_fs_type);
4978}
2035e776
TT		 4979
4980static inline int ext2_feature_set_ok(struct super_block *sb)
4981{
4982 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4983 return 0;
4984 if (sb->s_flags & MS_RDONLY)
4985 return 1;
4986 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4987 return 0;
4988 return 1;
4989}
51b7e3c9 4990MODULE_ALIAS("ext2");
24b58424
TT		 4991#else
4992static inline void register_as_ext2(void) { }
4993static inline void unregister_as_ext2(void) { }
2035e776 4994static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4995#endif
4996
37f328eb 4997#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4998static inline void register_as_ext3(void)
4999{
5000 int err = register_filesystem(&ext3_fs_type);
5001 if (err)
5002 printk(KERN_WARNING
5003 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5004}
5005
5006static inline void unregister_as_ext3(void)
5007{
5008 unregister_filesystem(&ext3_fs_type);
5009}
2035e776
TT		 5010
5011static inline int ext3_feature_set_ok(struct super_block *sb)
5012{
5013 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5014 return 0;
5015 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5016 return 0;
5017 if (sb->s_flags & MS_RDONLY)
5018 return 1;
5019 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5020 return 0;
5021 return 1;
5022}
51b7e3c9 5023MODULE_ALIAS("ext3");
24b58424
TT		 5024#else
5025static inline void register_as_ext3(void) { }
5026static inline void unregister_as_ext3(void) { }
2035e776 5027static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5028#endif
5029
03010a33
TT		 5030static struct file_system_type ext4_fs_type = {
5031 .owner = THIS_MODULE,
5032 .name = "ext4",
152a0836 5033 .mount = ext4_mount,
03010a33
TT		 5034 .kill_sb = kill_block_super,
5035 .fs_flags = FS_REQUIRES_DEV,
5036};
5037
8f021222 5038static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5039{
5040 struct ext4_features *ef;
5041 int ret = -ENOMEM;
5042
5043 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5044 if (!ef)
5045 goto out;
5046
5047 ef->f_kobj.kset = ext4_kset;
5048 init_completion(&ef->f_kobj_unregister);
5049 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5050 "features");
5051 if (ret) {
5052 kfree(ef);
5053 goto out;
5054 }
5055
5056 ext4_feat = ef;
5057 ret = 0;
5058out:
5059 return ret;
5060}
5061
8f021222
LC		 5062static void ext4_exit_feat_adverts(void)
5063{
5064 kobject_put(&ext4_feat->f_kobj);
5065 wait_for_completion(&ext4_feat->f_kobj_unregister);
5066 kfree(ext4_feat);
5067}
5068
e9e3bcec
ES		 5069/* Shared across all ext4 file systems */
5070wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5071struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5072
5dabfc78 5073static int __init ext4_init_fs(void)
ac27a0ec 5074{
e9e3bcec 5075 int i, err;
c9de560d 5076
07c0c5d8
AV		 5077 ext4_li_info = NULL;
5078 mutex_init(&ext4_li_mtx);
5079
12e9b892 5080 ext4_check_flag_values();
e9e3bcec
ES		 5081
5082 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5083 mutex_init(&ext4__aio_mutex[i]);
5084 init_waitqueue_head(&ext4__ioend_wq[i]);
5085 }
5086
5dabfc78 5087 err = ext4_init_pageio();
6fd058f7
TT		 5088 if (err)
5089 return err;
5dabfc78 5090 err = ext4_init_system_zone();
bd2d0210 5091 if (err)
d44651d0 5092 goto out6;
3197ebdb
TT		 5093 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5094 if (!ext4_kset)
dd68314c 5095 goto out5;
d44651d0 5096 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5097
5098 err = ext4_init_feat_adverts();
dd68314c
TT		 5099 if (err)
5100 goto out4;
857ac889 5101
5dabfc78 5102 err = ext4_init_mballoc();
ac27a0ec 5103 if (err)
6fd058f7 5104 goto out3;
c9de560d 5105
5dabfc78 5106 err = ext4_init_xattr();
c9de560d
AT		 5107 if (err)
5108 goto out2;
ac27a0ec
DK		 5109 err = init_inodecache();
5110 if (err)
5111 goto out1;
24b58424 5112 register_as_ext3();
2035e776 5113 register_as_ext2();
03010a33 5114 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5115 if (err)
5116 goto out;
bfff6873 5117
ac27a0ec
DK		 5118 return 0;
5119out:
24b58424
TT		 5120 unregister_as_ext2();
5121 unregister_as_ext3();
ac27a0ec
DK		 5122 destroy_inodecache();
5123out1:
5dabfc78 5124 ext4_exit_xattr();
c9de560d 5125out2:
5dabfc78 5126 ext4_exit_mballoc();
6fd058f7 5127out3:
8f021222 5128 ext4_exit_feat_adverts();
dd68314c 5129out4:
d44651d0
FJ		 5130 if (ext4_proc_root)
5131 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5132 kset_unregister(ext4_kset);
d44651d0 5133out5:
5dabfc78 5134 ext4_exit_system_zone();
d44651d0 5135out6:
5dabfc78 5136 ext4_exit_pageio();
ac27a0ec
DK		 5137 return err;
5138}
5139
5dabfc78 5140static void __exit ext4_exit_fs(void)
ac27a0ec 5141{
bfff6873 5142 ext4_destroy_lazyinit_thread();
24b58424
TT		 5143 unregister_as_ext2();
5144 unregister_as_ext3();
03010a33 5145 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5146 destroy_inodecache();
5dabfc78
TT		 5147 ext4_exit_xattr();
5148 ext4_exit_mballoc();
8f021222 5149 ext4_exit_feat_adverts();
9f6200bb 5150 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5151 kset_unregister(ext4_kset);
5dabfc78
TT		 5152 ext4_exit_system_zone();
5153 ext4_exit_pageio();
ac27a0ec
DK		 5154}
5155
5156MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5157MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5158MODULE_LICENSE("GPL");
5dabfc78
TT		 5159module_init(ext4_init_fs)
5160module_exit(ext4_exit_fs)
Ubuntu Artful kernel mirror