]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: move the jbd2 wrapper functions out of super.c
[thirdparty/kernel/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
4a092d73 48#include "ext4_extents.h" /* Needed for trace points definition */
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
TT		 72static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 74static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 75static int ext4_freeze(struct super_block *sb);
152a0836
AV		 76static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
2035e776
TT		 78static inline int ext2_feature_set_ok(struct super_block *sb);
79static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 80static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 81static void ext4_destroy_lazyinit_thread(void);
82static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 83static void ext4_clear_request_list(void);
ac27a0ec 84
2035e776
TT		 85#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
86static struct file_system_type ext2_fs_type = {
87 .owner = THIS_MODULE,
88 .name = "ext2",
89 .mount = ext4_mount,
90 .kill_sb = kill_block_super,
91 .fs_flags = FS_REQUIRES_DEV,
92};
93#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
94#else
95#define IS_EXT2_SB(sb) (0)
96#endif
97
98
ba69f9ab
JK		 99#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
100static struct file_system_type ext3_fs_type = {
101 .owner = THIS_MODULE,
102 .name = "ext3",
152a0836 103 .mount = ext4_mount,
ba69f9ab
JK		 104 .kill_sb = kill_block_super,
105 .fs_flags = FS_REQUIRES_DEV,
106};
107#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
108#else
109#define IS_EXT3_SB(sb) (0)
110#endif
bd81d8ee 111
d25425f8
DW		 112static int ext4_verify_csum_type(struct super_block *sb,
113 struct ext4_super_block *es)
114{
115 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
116 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
117 return 1;
118
119 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
120}
121
a9c47317
DW		 122static __le32 ext4_superblock_csum(struct super_block *sb,
123 struct ext4_super_block *es)
124{
125 struct ext4_sb_info *sbi = EXT4_SB(sb);
126 int offset = offsetof(struct ext4_super_block, s_checksum);
127 __u32 csum;
128
129 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
130
131 return cpu_to_le32(csum);
132}
133
134int ext4_superblock_csum_verify(struct super_block *sb,
135 struct ext4_super_block *es)
136{
137 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
138 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
139 return 1;
140
141 return es->s_checksum == ext4_superblock_csum(sb, es);
142}
143
06db49e6 144void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 145{
06db49e6
TT		 146 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
147
a9c47317
DW		 148 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
149 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
150 return;
151
152 es->s_checksum = ext4_superblock_csum(sb, es);
153}
154
9933fc0a
TT		 155void *ext4_kvmalloc(size_t size, gfp_t flags)
156{
157 void *ret;
158
159 ret = kmalloc(size, flags);
160 if (!ret)
161 ret = __vmalloc(size, flags, PAGE_KERNEL);
162 return ret;
163}
164
165void *ext4_kvzalloc(size_t size, gfp_t flags)
166{
167 void *ret;
168
db9481c0 169 ret = kzalloc(size, flags);
9933fc0a
TT		 170 if (!ret)
171 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
172 return ret;
173}
174
175void ext4_kvfree(void *ptr)
176{
177 if (is_vmalloc_addr(ptr))
178 vfree(ptr);
179 else
180 kfree(ptr);
181
182}
183
8fadc143
AR		 184ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
185 struct ext4_group_desc *bg)
bd81d8ee 186{
3a14589c 187 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 188 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 189 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 190}
191
8fadc143
AR		 192ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
193 struct ext4_group_desc *bg)
bd81d8ee 194{
5272f837 195 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 196 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 197 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 198}
199
8fadc143
AR		 200ext4_fsblk_t ext4_inode_table(struct super_block *sb,
201 struct ext4_group_desc *bg)
bd81d8ee 202{
5272f837 203 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 204 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 205 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 206}
207
021b65bb
TT		 208__u32 ext4_free_group_clusters(struct super_block *sb,
209 struct ext4_group_desc *bg)
560671a0
AK		 210{
211 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
212 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 213 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 214}
215
216__u32 ext4_free_inodes_count(struct super_block *sb,
217 struct ext4_group_desc *bg)
218{
219 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
220 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 221 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 222}
223
224__u32 ext4_used_dirs_count(struct super_block *sb,
225 struct ext4_group_desc *bg)
226{
227 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
228 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 229 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 230}
231
232__u32 ext4_itable_unused_count(struct super_block *sb,
233 struct ext4_group_desc *bg)
234{
235 return le16_to_cpu(bg->bg_itable_unused_lo) |
236 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 237 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 238}
239
8fadc143
AR		 240void ext4_block_bitmap_set(struct super_block *sb,
241 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 242{
3a14589c 243 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 244 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
245 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 246}
247
8fadc143
AR		 248void ext4_inode_bitmap_set(struct super_block *sb,
249 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 250{
5272f837 251 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 252 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
253 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 254}
255
8fadc143
AR		 256void ext4_inode_table_set(struct super_block *sb,
257 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 258{
5272f837 259 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 260 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
261 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 262}
263
021b65bb
TT		 264void ext4_free_group_clusters_set(struct super_block *sb,
265 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 266{
267 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
268 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
269 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
270}
271
272void ext4_free_inodes_set(struct super_block *sb,
273 struct ext4_group_desc *bg, __u32 count)
274{
275 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
276 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
277 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
278}
279
280void ext4_used_dirs_set(struct super_block *sb,
281 struct ext4_group_desc *bg, __u32 count)
282{
283 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
284 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
285 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
286}
287
288void ext4_itable_unused_set(struct super_block *sb,
289 struct ext4_group_desc *bg, __u32 count)
290{
291 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
292 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
293 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
294}
295
d3d1faf6 296
1c13d5c0
TT		 297static void __save_error_info(struct super_block *sb, const char *func,
298 unsigned int line)
299{
300 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
301
302 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
303 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
304 es->s_last_error_time = cpu_to_le32(get_seconds());
305 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
306 es->s_last_error_line = cpu_to_le32(line);
307 if (!es->s_first_error_time) {
308 es->s_first_error_time = es->s_last_error_time;
309 strncpy(es->s_first_error_func, func,
310 sizeof(es->s_first_error_func));
311 es->s_first_error_line = cpu_to_le32(line);
312 es->s_first_error_ino = es->s_last_error_ino;
313 es->s_first_error_block = es->s_last_error_block;
314 }
66e61a9e
TT		 315 /*
316 * Start the daily error reporting function if it hasn't been
317 * started already
318 */
319 if (!es->s_error_count)
320 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 321 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 322}
323
324static void save_error_info(struct super_block *sb, const char *func,
325 unsigned int line)
326{
327 __save_error_info(sb, func, line);
328 ext4_commit_super(sb, 1);
329}
330
7c2e7087
TT		 331/*
332 * The del_gendisk() function uninitializes the disk-specific data
333 * structures, including the bdi structure, without telling anyone
334 * else. Once this happens, any attempt to call mark_buffer_dirty()
335 * (for example, by ext4_commit_super), will cause a kernel OOPS.
336 * This is a kludge to prevent these oops until we can put in a proper
337 * hook in del_gendisk() to inform the VFS and file system layers.
338 */
339static int block_device_ejected(struct super_block *sb)
340{
341 struct inode *bd_inode = sb->s_bdev->bd_inode;
342 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
343
344 return bdi->dev == NULL;
345}
346
18aadd47
BJ		 347static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
348{
349 struct super_block *sb = journal->j_private;
350 struct ext4_sb_info *sbi = EXT4_SB(sb);
351 int error = is_journal_aborted(journal);
352 struct ext4_journal_cb_entry *jce, *tmp;
353
354 spin_lock(&sbi->s_md_lock);
355 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
356 list_del_init(&jce->jce_list);
357 spin_unlock(&sbi->s_md_lock);
358 jce->jce_func(sb, jce, error);
359 spin_lock(&sbi->s_md_lock);
360 }
361 spin_unlock(&sbi->s_md_lock);
362}
1c13d5c0 363
ac27a0ec
DK		 364/* Deal with the reporting of failure conditions on a filesystem such as
365 * inconsistencies detected or read IO failures.
366 *
367 * On ext2, we can store the error state of the filesystem in the
617ba13b 368 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 369 * write ordering constraints on the superblock which prevent us from
370 * writing it out straight away; and given that the journal is about to
371 * be aborted, we can't rely on the current, or future, transactions to
372 * write out the superblock safely.
373 *
dab291af 374 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 375 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 376 * that error until we've noted it down and cleared it.
377 */
378
617ba13b 379static void ext4_handle_error(struct super_block *sb)
ac27a0ec 380{
ac27a0ec
DK		 381 if (sb->s_flags & MS_RDONLY)
382 return;
383
2b2d6d01 384 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 385 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 386
4ab2f15b 387 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 388 if (journal)
dab291af 389 jbd2_journal_abort(journal, -EIO);
ac27a0ec 390 }
2b2d6d01 391 if (test_opt(sb, ERRORS_RO)) {
b31e1552 392 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 393 sb->s_flags |= MS_RDONLY;
394 }
ac27a0ec 395 if (test_opt(sb, ERRORS_PANIC))
617ba13b 396 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 397 sb->s_id);
398}
399
12062ddd 400void __ext4_error(struct super_block *sb, const char *function,
c398eda0 401 unsigned int line, const char *fmt, ...)
ac27a0ec 402{
0ff2ea7d 403 struct va_format vaf;
ac27a0ec
DK		 404 va_list args;
405
406 va_start(args, fmt);
0ff2ea7d
JP		 407 vaf.fmt = fmt;
408 vaf.va = &args;
409 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
410 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec 411 va_end(args);
f3fc0210 412 save_error_info(sb, function, line);
ac27a0ec 413
617ba13b 414 ext4_handle_error(sb);
ac27a0ec
DK		 415}
416
c398eda0
TT		 417void ext4_error_inode(struct inode *inode, const char *function,
418 unsigned int line, ext4_fsblk_t block,
273df556
FM		 419 const char *fmt, ...)
420{
421 va_list args;
f7c21177 422 struct va_format vaf;
1c13d5c0 423 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 424
1c13d5c0
TT		 425 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
426 es->s_last_error_block = cpu_to_le64(block);
427 save_error_info(inode->i_sb, function, line);
273df556 428 va_start(args, fmt);
f7c21177
TT		 429 vaf.fmt = fmt;
430 vaf.va = &args;
c398eda0 431 if (block)
d9ee81da
JP		 432 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
433 "inode #%lu: block %llu: comm %s: %pV\n",
434 inode->i_sb->s_id, function, line, inode->i_ino,
435 block, current->comm, &vaf);
436 else
437 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
438 "inode #%lu: comm %s: %pV\n",
439 inode->i_sb->s_id, function, line, inode->i_ino,
440 current->comm, &vaf);
273df556
FM		 441 va_end(args);
442
443 ext4_handle_error(inode->i_sb);
444}
445
c398eda0 446void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 447 unsigned int line, ext4_fsblk_t block,
448 const char *fmt, ...)
273df556
FM		 449{
450 va_list args;
f7c21177 451 struct va_format vaf;
1c13d5c0 452 struct ext4_super_block *es;
273df556
FM		 453 struct inode *inode = file->f_dentry->d_inode;
454 char pathname[80], *path;
455
1c13d5c0
TT		 456 es = EXT4_SB(inode->i_sb)->s_es;
457 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
458 save_error_info(inode->i_sb, function, line);
273df556 459 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 460 if (IS_ERR(path))
273df556 461 path = "(unknown)";
f7c21177
TT		 462 va_start(args, fmt);
463 vaf.fmt = fmt;
464 vaf.va = &args;
d9ee81da
JP		 465 if (block)
466 printk(KERN_CRIT
467 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
468 "block %llu: comm %s: path %s: %pV\n",
469 inode->i_sb->s_id, function, line, inode->i_ino,
470 block, current->comm, path, &vaf);
471 else
472 printk(KERN_CRIT
473 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
474 "comm %s: path %s: %pV\n",
475 inode->i_sb->s_id, function, line, inode->i_ino,
476 current->comm, path, &vaf);
273df556
FM		 477 va_end(args);
478
479 ext4_handle_error(inode->i_sb);
480}
481
722887dd
TT		 482const char *ext4_decode_error(struct super_block *sb, int errno,
483 char nbuf[16])
ac27a0ec
DK		 484{
485 char *errstr = NULL;
486
487 switch (errno) {
488 case -EIO:
489 errstr = "IO failure";
490 break;
491 case -ENOMEM:
492 errstr = "Out of memory";
493 break;
494 case -EROFS:
78f1ddbb
TT		 495 if (!sb || (EXT4_SB(sb)->s_journal &&
496 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 497 errstr = "Journal has aborted";
498 else
499 errstr = "Readonly filesystem";
500 break;
501 default:
502 /* If the caller passed in an extra buffer for unknown
503 * errors, textualise them now. Else we just return
504 * NULL. */
505 if (nbuf) {
506 /* Check for truncated error codes... */
507 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
508 errstr = nbuf;
509 }
510 break;
511 }
512
513 return errstr;
514}
515
617ba13b 516/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 517 * automatically and invokes the appropriate error response. */
518
c398eda0
TT		 519void __ext4_std_error(struct super_block *sb, const char *function,
520 unsigned int line, int errno)
ac27a0ec
DK		 521{
522 char nbuf[16];
523 const char *errstr;
524
525 /* Special case: if the error is EROFS, and we're not already
526 * inside a transaction, then there's really no point in logging
527 * an error. */
528 if (errno == -EROFS && journal_current_handle() == NULL &&
529 (sb->s_flags & MS_RDONLY))
530 return;
531
617ba13b 532 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 533 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
534 sb->s_id, function, line, errstr);
1c13d5c0 535 save_error_info(sb, function, line);
ac27a0ec 536
617ba13b 537 ext4_handle_error(sb);
ac27a0ec
DK		 538}
539
540/*
617ba13b 541 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 542 * abort function may be used to deal with unrecoverable failures such
543 * as journal IO errors or ENOMEM at a critical moment in log management.
544 *
545 * We unconditionally force the filesystem into an ABORT|READONLY state,
546 * unless the error response on the fs has been set to panic in which
547 * case we take the easy way out and panic immediately.
548 */
549
c67d859e 550void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 551 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 552{
553 va_list args;
554
1c13d5c0 555 save_error_info(sb, function, line);
ac27a0ec 556 va_start(args, fmt);
c398eda0
TT		 557 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
558 function, line);
ac27a0ec
DK		 559 vprintk(fmt, args);
560 printk("\n");
561 va_end(args);
562
1c13d5c0
TT		 563 if ((sb->s_flags & MS_RDONLY) == 0) {
564 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
565 sb->s_flags |= MS_RDONLY;
566 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
567 if (EXT4_SB(sb)->s_journal)
568 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
569 save_error_info(sb, function, line);
570 }
ac27a0ec 571 if (test_opt(sb, ERRORS_PANIC))
617ba13b 572 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 573}
574
0ff2ea7d 575void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 576{
0ff2ea7d 577 struct va_format vaf;
b31e1552
ES		 578 va_list args;
579
580 va_start(args, fmt);
0ff2ea7d
JP		 581 vaf.fmt = fmt;
582 vaf.va = &args;
583 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 584 va_end(args);
585}
586
12062ddd 587void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 588 unsigned int line, const char *fmt, ...)
ac27a0ec 589{
0ff2ea7d 590 struct va_format vaf;
ac27a0ec
DK		 591 va_list args;
592
593 va_start(args, fmt);
0ff2ea7d
JP		 594 vaf.fmt = fmt;
595 vaf.va = &args;
596 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
597 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 598 va_end(args);
599}
600
e29136f8
TT		 601void __ext4_grp_locked_error(const char *function, unsigned int line,
602 struct super_block *sb, ext4_group_t grp,
603 unsigned long ino, ext4_fsblk_t block,
604 const char *fmt, ...)
5d1b1b3f
AK		 605__releases(bitlock)
606__acquires(bitlock)
607{
0ff2ea7d 608 struct va_format vaf;
5d1b1b3f
AK		 609 va_list args;
610 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
611
1c13d5c0
TT		 612 es->s_last_error_ino = cpu_to_le32(ino);
613 es->s_last_error_block = cpu_to_le64(block);
614 __save_error_info(sb, function, line);
0ff2ea7d 615
5d1b1b3f 616 va_start(args, fmt);
0ff2ea7d
JP		 617
618 vaf.fmt = fmt;
619 vaf.va = &args;
21149d61 620 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 621 sb->s_id, function, line, grp);
622 if (ino)
0ff2ea7d 623 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 624 if (block)
0ff2ea7d
JP		 625 printk(KERN_CONT "block %llu:", (unsigned long long) block);
626 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 627 va_end(args);
628
629 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 630 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 631 return;
632 }
1c13d5c0 633
5d1b1b3f
AK		 634 ext4_unlock_group(sb, grp);
635 ext4_handle_error(sb);
636 /*
637 * We only get here in the ERRORS_RO case; relocking the group
638 * may be dangerous, but nothing bad will happen since the
639 * filesystem will have already been marked read/only and the
640 * journal has been aborted. We return 1 as a hint to callers
641 * who might what to use the return value from
25985edc 642 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 643 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
644 * aggressively from the ext4 function in question, with a
645 * more appropriate error code.
646 */
647 ext4_lock_group(sb, grp);
648 return;
649}
650
617ba13b 651void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 652{
617ba13b 653 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 654
617ba13b 655 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 656 return;
657
12062ddd 658 ext4_warning(sb,
ac27a0ec
DK		 659 "updating to rev %d because of new feature flag, "
660 "running e2fsck is recommended",
617ba13b 661 EXT4_DYNAMIC_REV);
ac27a0ec 662
617ba13b
MC		 663 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
664 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
665 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 666 /* leave es->s_feature_*compat flags alone */
667 /* es->s_uuid will be set by e2fsck if empty */
668
669 /*
670 * The rest of the superblock fields should be zero, and if not it
671 * means they are likely already in use, so leave them alone. We
672 * can leave it up to e2fsck to clean up any inconsistencies there.
673 */
674}
675
676/*
677 * Open the external journal device
678 */
b31e1552 679static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 680{
681 struct block_device *bdev;
682 char b[BDEVNAME_SIZE];
683
d4d77629 684 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 685 if (IS_ERR(bdev))
686 goto fail;
687 return bdev;
688
689fail:
b31e1552 690 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 691 __bdevname(dev, b), PTR_ERR(bdev));
692 return NULL;
693}
694
695/*
696 * Release the journal device
697 */
617ba13b 698static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 699{
e525fd89 700 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 701}
702
617ba13b 703static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 704{
705 struct block_device *bdev;
706 int ret = -ENODEV;
707
708 bdev = sbi->journal_bdev;
709 if (bdev) {
617ba13b 710 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 711 sbi->journal_bdev = NULL;
712 }
713 return ret;
714}
715
716static inline struct inode *orphan_list_entry(struct list_head *l)
717{
617ba13b 718 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 719}
720
617ba13b 721static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 722{
723 struct list_head *l;
724
b31e1552
ES		 725 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
726 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 727
728 printk(KERN_ERR "sb_info orphan list:\n");
729 list_for_each(l, &sbi->s_orphan) {
730 struct inode *inode = orphan_list_entry(l);
731 printk(KERN_ERR " "
732 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
733 inode->i_sb->s_id, inode->i_ino, inode,
734 inode->i_mode, inode->i_nlink,
735 NEXT_ORPHAN(inode));
736 }
737}
738
2b2d6d01 739static void ext4_put_super(struct super_block *sb)
ac27a0ec 740{
617ba13b
MC		 741 struct ext4_sb_info *sbi = EXT4_SB(sb);
742 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 743 int i, err;
ac27a0ec 744
857ac889 745 ext4_unregister_li_request(sb);
e0ccfd95
CH		 746 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
747
4c0425ff
MC		 748 flush_workqueue(sbi->dio_unwritten_wq);
749 destroy_workqueue(sbi->dio_unwritten_wq);
750
0390131b
FM		 751 if (sbi->s_journal) {
752 err = jbd2_journal_destroy(sbi->s_journal);
753 sbi->s_journal = NULL;
754 if (err < 0)
c67d859e 755 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 756 }
d4edac31 757
a1c6c569 758 del_timer(&sbi->s_err_report);
d4edac31
JB		 759 ext4_release_system_zone(sb);
760 ext4_mb_release(sb);
761 ext4_ext_release(sb);
762 ext4_xattr_put_super(sb);
763
ac27a0ec 764 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 765 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 766 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 767 }
58c5873a 768 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 769 ext4_commit_super(sb, 1);
770
240799cd 771 if (sbi->s_proc) {
66acdcf4 772 remove_proc_entry("options", sbi->s_proc);
9f6200bb 773 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 774 }
3197ebdb 775 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 776
777 for (i = 0; i < sbi->s_gdb_count; i++)
778 brelse(sbi->s_group_desc[i]);
f18a5f21 779 ext4_kvfree(sbi->s_group_desc);
9933fc0a 780 ext4_kvfree(sbi->s_flex_groups);
57042651 781 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 782 percpu_counter_destroy(&sbi->s_freeinodes_counter);
783 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 784 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 785 brelse(sbi->s_sbh);
786#ifdef CONFIG_QUOTA
787 for (i = 0; i < MAXQUOTAS; i++)
788 kfree(sbi->s_qf_names[i]);
789#endif
790
791 /* Debugging code just in case the in-memory inode orphan list
792 * isn't empty. The on-disk one can be non-empty if we've
793 * detected an error and taken the fs readonly, but the
794 * in-memory list had better be clean by this point. */
795 if (!list_empty(&sbi->s_orphan))
796 dump_orphan_list(sb, sbi);
797 J_ASSERT(list_empty(&sbi->s_orphan));
798
f98393a6 799 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 800 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
801 /*
802 * Invalidate the journal device's buffers. We don't want them
803 * floating about in memory - the physical journal device may
804 * hotswapped, and it breaks the `ro-after' testing code.
805 */
806 sync_blockdev(sbi->journal_bdev);
f98393a6 807 invalidate_bdev(sbi->journal_bdev);
617ba13b 808 ext4_blkdev_remove(sbi);
ac27a0ec 809 }
c5e06d10
JL		 810 if (sbi->s_mmp_tsk)
811 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 812 sb->s_fs_info = NULL;
3197ebdb
TT		 813 /*
814 * Now that we are completely done shutting down the
815 * superblock, we need to actually destroy the kobject.
816 */
3197ebdb
TT		 817 kobject_put(&sbi->s_kobj);
818 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 819 if (sbi->s_chksum_driver)
820 crypto_free_shash(sbi->s_chksum_driver);
705895b6 821 kfree(sbi->s_blockgroup_lock);
ac27a0ec 822 kfree(sbi);
ac27a0ec
DK		 823}
824
e18b890b 825static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 826
827/*
828 * Called inside transaction, so use GFP_NOFS
829 */
617ba13b 830static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 831{
617ba13b 832 struct ext4_inode_info *ei;
ac27a0ec 833
e6b4f8da 834 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 835 if (!ei)
836 return NULL;
0b8e58a1 837
ac27a0ec 838 ei->vfs_inode.i_version = 1;
a86c6181 839 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 840 INIT_LIST_HEAD(&ei->i_prealloc_list);
841 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 842 ext4_es_init_tree(&ei->i_es_tree);
843 rwlock_init(&ei->i_es_lock);
d2a17637
MC		 844 ei->i_reserved_data_blocks = 0;
845 ei->i_reserved_meta_blocks = 0;
846 ei->i_allocated_meta_blocks = 0;
9d0be502 847 ei->i_da_metadata_calc_len = 0;
7e731bc9 848 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 849 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 850#ifdef CONFIG_QUOTA
851 ei->i_reserved_quota = 0;
852#endif
8aefcd55 853 ei->jinode = NULL;
c7064ef1 854 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 855 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 856 ei->i_sync_tid = 0;
857 ei->i_datasync_tid = 0;
f7ad6d2e 858 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 859 atomic_set(&ei->i_unwritten, 0);
84c17543 860 INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
0b8e58a1 861
ac27a0ec
DK		 862 return &ei->vfs_inode;
863}
864
7ff9c073
TT		 865static int ext4_drop_inode(struct inode *inode)
866{
867 int drop = generic_drop_inode(inode);
868
869 trace_ext4_drop_inode(inode, drop);
870 return drop;
871}
872
fa0d7e3d
NP		 873static void ext4_i_callback(struct rcu_head *head)
874{
875 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 876 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
877}
878
617ba13b 879static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 880{
9f7dd93d 881 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 882 ext4_msg(inode->i_sb, KERN_ERR,
883 "Inode %lu (%p): orphan list check failed!",
884 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 885 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
886 EXT4_I(inode), sizeof(struct ext4_inode_info),
887 true);
888 dump_stack();
889 }
fa0d7e3d 890 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 891}
892
51cc5068 893static void init_once(void *foo)
ac27a0ec 894{
617ba13b 895 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 896
a35afb83 897 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 898 init_rwsem(&ei->xattr_sem);
0e855ac8 899 init_rwsem(&ei->i_data_sem);
a35afb83 900 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 901}
902
903static int init_inodecache(void)
904{
617ba13b
MC		 905 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
906 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 907 0, (SLAB_RECLAIM_ACCOUNT|
908 SLAB_MEM_SPREAD),
20c2df83 909 init_once);
617ba13b 910 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 911 return -ENOMEM;
912 return 0;
913}
914
915static void destroy_inodecache(void)
916{
8c0a8537
KS		 917 /*
918 * Make sure all delayed rcu free inodes are flushed before we
919 * destroy cache.
920 */
921 rcu_barrier();
617ba13b 922 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 923}
924
0930fcc1 925void ext4_clear_inode(struct inode *inode)
ac27a0ec 926{
0930fcc1 927 invalidate_inode_buffers(inode);
dbd5768f 928 clear_inode(inode);
9f754758 929 dquot_drop(inode);
c2ea3fde 930 ext4_discard_preallocations(inode);
51865fda 931 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 932 if (EXT4_I(inode)->jinode) {
933 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
934 EXT4_I(inode)->jinode);
935 jbd2_free_inode(EXT4_I(inode)->jinode);
936 EXT4_I(inode)->jinode = NULL;
937 }
ac27a0ec
DK		 938}
939
1b961ac0 940static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 941 u64 ino, u32 generation)
ac27a0ec 942{
ac27a0ec 943 struct inode *inode;
ac27a0ec 944
617ba13b 945 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 946 return ERR_PTR(-ESTALE);
617ba13b 947 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 948 return ERR_PTR(-ESTALE);
949
950 /* iget isn't really right if the inode is currently unallocated!!
951 *
617ba13b 952 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 953 * deleted, so we should be safe.
954 *
955 * Currently we don't know the generation for parent directory, so
956 * a generation of 0 means "accept any"
957 */
1d1fe1ee
DH		 958 inode = ext4_iget(sb, ino);
959 if (IS_ERR(inode))
960 return ERR_CAST(inode);
961 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 962 iput(inode);
963 return ERR_PTR(-ESTALE);
964 }
1b961ac0
CH		 965
966 return inode;
967}
968
969static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 970 int fh_len, int fh_type)
1b961ac0
CH		 971{
972 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
973 ext4_nfs_get_inode);
974}
975
976static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 977 int fh_len, int fh_type)
1b961ac0
CH		 978{
979 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
980 ext4_nfs_get_inode);
ac27a0ec
DK		 981}
982
c39a7f84
TO		 983/*
984 * Try to release metadata pages (indirect blocks, directories) which are
985 * mapped via the block device. Since these pages could have journal heads
986 * which would prevent try_to_free_buffers() from freeing them, we must use
987 * jbd2 layer's try_to_free_buffers() function to release them.
988 */
0b8e58a1
AD		 989static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
990 gfp_t wait)
c39a7f84
TO		 991{
992 journal_t *journal = EXT4_SB(sb)->s_journal;
993
994 WARN_ON(PageChecked(page));
995 if (!page_has_buffers(page))
996 return 0;
997 if (journal)
998 return jbd2_journal_try_to_free_buffers(journal, page,
999 wait & ~__GFP_WAIT);
1000 return try_to_free_buffers(page);
1001}
1002
ac27a0ec 1003#ifdef CONFIG_QUOTA
af5bc92d 1004#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1005#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1006
617ba13b
MC		 1007static int ext4_write_dquot(struct dquot *dquot);
1008static int ext4_acquire_dquot(struct dquot *dquot);
1009static int ext4_release_dquot(struct dquot *dquot);
1010static int ext4_mark_dquot_dirty(struct dquot *dquot);
1011static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1012static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1013 struct path *path);
7c319d32
AK		 1014static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1015 int format_id);
ca0e05e4 1016static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1017static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1018static int ext4_quota_on_mount(struct super_block *sb, int type);
1019static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1020 size_t len, loff_t off);
617ba13b 1021static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1022 const char *data, size_t len, loff_t off);
7c319d32
AK		 1023static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1024 unsigned int flags);
1025static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1026
61e225dc 1027static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1028 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1029 .write_dquot = ext4_write_dquot,
1030 .acquire_dquot = ext4_acquire_dquot,
1031 .release_dquot = ext4_release_dquot,
1032 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1033 .write_info = ext4_write_info,
1034 .alloc_dquot = dquot_alloc,
1035 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1036};
1037
0d54b217 1038static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1039 .quota_on = ext4_quota_on,
ca0e05e4 1040 .quota_off = ext4_quota_off,
287a8095
CH		 1041 .quota_sync = dquot_quota_sync,
1042 .get_info = dquot_get_dqinfo,
1043 .set_info = dquot_set_dqinfo,
1044 .get_dqblk = dquot_get_dqblk,
1045 .set_dqblk = dquot_set_dqblk
ac27a0ec 1046};
7c319d32
AK		 1047
1048static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1049 .quota_on_meta = ext4_quota_on_sysfile,
1050 .quota_off = ext4_quota_off_sysfile,
1051 .quota_sync = dquot_quota_sync,
1052 .get_info = dquot_get_dqinfo,
1053 .set_info = dquot_set_dqinfo,
1054 .get_dqblk = dquot_get_dqblk,
1055 .set_dqblk = dquot_set_dqblk
1056};
ac27a0ec
DK		 1057#endif
1058
ee9b6d61 1059static const struct super_operations ext4_sops = {
617ba13b
MC		 1060 .alloc_inode = ext4_alloc_inode,
1061 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1062 .write_inode = ext4_write_inode,
1063 .dirty_inode = ext4_dirty_inode,
7ff9c073 1064 .drop_inode = ext4_drop_inode,
0930fcc1 1065 .evict_inode = ext4_evict_inode,
617ba13b 1066 .put_super = ext4_put_super,
617ba13b 1067 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1068 .freeze_fs = ext4_freeze,
1069 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1070 .statfs = ext4_statfs,
1071 .remount_fs = ext4_remount,
617ba13b 1072 .show_options = ext4_show_options,
ac27a0ec 1073#ifdef CONFIG_QUOTA
617ba13b
MC		 1074 .quota_read = ext4_quota_read,
1075 .quota_write = ext4_quota_write,
ac27a0ec 1076#endif
c39a7f84 1077 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1078};
1079
9ca92389
TT		 1080static const struct super_operations ext4_nojournal_sops = {
1081 .alloc_inode = ext4_alloc_inode,
1082 .destroy_inode = ext4_destroy_inode,
1083 .write_inode = ext4_write_inode,
1084 .dirty_inode = ext4_dirty_inode,
7ff9c073 1085 .drop_inode = ext4_drop_inode,
0930fcc1 1086 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1087 .put_super = ext4_put_super,
1088 .statfs = ext4_statfs,
1089 .remount_fs = ext4_remount,
9ca92389
TT		 1090 .show_options = ext4_show_options,
1091#ifdef CONFIG_QUOTA
1092 .quota_read = ext4_quota_read,
1093 .quota_write = ext4_quota_write,
1094#endif
1095 .bdev_try_to_free_page = bdev_try_to_free_page,
1096};
1097
39655164 1098static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1099 .fh_to_dentry = ext4_fh_to_dentry,
1100 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1101 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1102};
1103
1104enum {
1105 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1106 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1107 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1108 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1109 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1110 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1111 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1112 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1113 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1114 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1115 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1116 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1117 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1118 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1119 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1120 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1121 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1122 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1123 Opt_max_dir_size_kb,
ac27a0ec
DK		 1124};
1125
a447c093 1126static const match_table_t tokens = {
ac27a0ec
DK		 1127 {Opt_bsd_df, "bsddf"},
1128 {Opt_minix_df, "minixdf"},
1129 {Opt_grpid, "grpid"},
1130 {Opt_grpid, "bsdgroups"},
1131 {Opt_nogrpid, "nogrpid"},
1132 {Opt_nogrpid, "sysvgroups"},
1133 {Opt_resgid, "resgid=%u"},
1134 {Opt_resuid, "resuid=%u"},
1135 {Opt_sb, "sb=%u"},
1136 {Opt_err_cont, "errors=continue"},
1137 {Opt_err_panic, "errors=panic"},
1138 {Opt_err_ro, "errors=remount-ro"},
1139 {Opt_nouid32, "nouid32"},
ac27a0ec 1140 {Opt_debug, "debug"},
72578c33
TT		 1141 {Opt_removed, "oldalloc"},
1142 {Opt_removed, "orlov"},
ac27a0ec
DK		 1143 {Opt_user_xattr, "user_xattr"},
1144 {Opt_nouser_xattr, "nouser_xattr"},
1145 {Opt_acl, "acl"},
1146 {Opt_noacl, "noacl"},
e3bb52ae 1147 {Opt_noload, "norecovery"},
5a916be1 1148 {Opt_noload, "noload"},
72578c33
TT		 1149 {Opt_removed, "nobh"},
1150 {Opt_removed, "bh"},
ac27a0ec 1151 {Opt_commit, "commit=%u"},
30773840
TT		 1152 {Opt_min_batch_time, "min_batch_time=%u"},
1153 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1154 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1155 {Opt_journal_checksum, "journal_checksum"},
1156 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1157 {Opt_abort, "abort"},
1158 {Opt_data_journal, "data=journal"},
1159 {Opt_data_ordered, "data=ordered"},
1160 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1161 {Opt_data_err_abort, "data_err=abort"},
1162 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1163 {Opt_offusrjquota, "usrjquota="},
1164 {Opt_usrjquota, "usrjquota=%s"},
1165 {Opt_offgrpjquota, "grpjquota="},
1166 {Opt_grpjquota, "grpjquota=%s"},
1167 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1168 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1169 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1170 {Opt_grpquota, "grpquota"},
1171 {Opt_noquota, "noquota"},
1172 {Opt_quota, "quota"},
1173 {Opt_usrquota, "usrquota"},
1174 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1175 {Opt_barrier, "barrier"},
1176 {Opt_nobarrier, "nobarrier"},
25ec56b5 1177 {Opt_i_version, "i_version"},
c9de560d 1178 {Opt_stripe, "stripe=%u"},
64769240 1179 {Opt_delalloc, "delalloc"},
dd919b98 1180 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1181 {Opt_removed, "mblk_io_submit"},
1182 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1183 {Opt_block_validity, "block_validity"},
1184 {Opt_noblock_validity, "noblock_validity"},
240799cd 1185 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1186 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1187 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1188 {Opt_auto_da_alloc, "auto_da_alloc"},
1189 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1190 {Opt_dioread_nolock, "dioread_nolock"},
1191 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1192 {Opt_discard, "discard"},
1193 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1194 {Opt_init_itable, "init_itable=%u"},
1195 {Opt_init_itable, "init_itable"},
1196 {Opt_noinit_itable, "noinit_itable"},
df981d03 1197 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1198 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1199 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1200 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1201 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1202 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1203 {Opt_err, NULL},
ac27a0ec
DK		 1204};
1205
617ba13b 1206static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1207{
617ba13b 1208 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1209 char *options = (char *) *data;
1210
1211 if (!options || strncmp(options, "sb=", 3) != 0)
1212 return 1; /* Default location */
0b8e58a1 1213
ac27a0ec 1214 options += 3;
0b8e58a1 1215 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1216 sb_block = simple_strtoul(options, &options, 0);
1217 if (*options && *options != ',') {
4776004f 1218 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1219 (char *) *data);
1220 return 1;
1221 }
1222 if (*options == ',')
1223 options++;
1224 *data = (void *) options;
0b8e58a1 1225
ac27a0ec
DK		 1226 return sb_block;
1227}
1228
b3881f74 1229#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1230static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1231 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1232
56c50f11
DM		 1233#ifdef CONFIG_QUOTA
1234static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1235{
1236 struct ext4_sb_info *sbi = EXT4_SB(sb);
1237 char *qname;
03dafb5f 1238 int ret = -1;
56c50f11
DM		 1239
1240 if (sb_any_quota_loaded(sb) &&
1241 !sbi->s_qf_names[qtype]) {
1242 ext4_msg(sb, KERN_ERR,
1243 "Cannot change journaled "
1244 "quota options when quota turned on");
57f73c2c 1245 return -1;
56c50f11
DM		 1246 }
1247 qname = match_strdup(args);
1248 if (!qname) {
1249 ext4_msg(sb, KERN_ERR,
1250 "Not enough memory for storing quotafile name");
57f73c2c 1251 return -1;
56c50f11 1252 }
03dafb5f
CG		 1253 if (sbi->s_qf_names[qtype]) {
1254 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1255 ret = 1;
1256 else
1257 ext4_msg(sb, KERN_ERR,
1258 "%s quota file already specified",
1259 QTYPE2NAME(qtype));
1260 goto errout;
56c50f11 1261 }
03dafb5f 1262 if (strchr(qname, '/')) {
56c50f11
DM		 1263 ext4_msg(sb, KERN_ERR,
1264 "quotafile must be on filesystem root");
03dafb5f 1265 goto errout;
56c50f11 1266 }
03dafb5f 1267 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1268 set_opt(sb, QUOTA);
56c50f11 1269 return 1;
03dafb5f
CG		 1270errout:
1271 kfree(qname);
1272 return ret;
56c50f11
DM		 1273}
1274
1275static int clear_qf_name(struct super_block *sb, int qtype)
1276{
1277
1278 struct ext4_sb_info *sbi = EXT4_SB(sb);
1279
1280 if (sb_any_quota_loaded(sb) &&
1281 sbi->s_qf_names[qtype]) {
1282 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1283 " when quota turned on");
57f73c2c 1284 return -1;
56c50f11 1285 }
03dafb5f 1286 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1287 sbi->s_qf_names[qtype] = NULL;
1288 return 1;
1289}
1290#endif
1291
26092bf5
TT		 1292#define MOPT_SET 0x0001
1293#define MOPT_CLEAR 0x0002
1294#define MOPT_NOSUPPORT 0x0004
1295#define MOPT_EXPLICIT 0x0008
1296#define MOPT_CLEAR_ERR 0x0010
1297#define MOPT_GTE0 0x0020
ac27a0ec 1298#ifdef CONFIG_QUOTA
26092bf5
TT		 1299#define MOPT_Q 0
1300#define MOPT_QFMT 0x0040
1301#else
1302#define MOPT_Q MOPT_NOSUPPORT
1303#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1304#endif
26092bf5 1305#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1306#define MOPT_NO_EXT2 0x0100
1307#define MOPT_NO_EXT3 0x0200
1308#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
26092bf5
TT		 1309
1310static const struct mount_opts {
1311 int token;
1312 int mount_opt;
1313 int flags;
1314} ext4_mount_opts[] = {
1315 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1316 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1317 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1318 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1319 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1320 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1321 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1322 MOPT_EXT4_ONLY | MOPT_SET},
1323 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1324 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1325 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1326 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1327 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1328 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1329 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1330 MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1331 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1332 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1333 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1334 EXT4_MOUNT_JOURNAL_CHECKSUM),
1335 MOPT_EXT4_ONLY | MOPT_SET},
1336 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1337 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1338 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1339 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1340 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1341 MOPT_NO_EXT2 | MOPT_SET},
1342 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1343 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1344 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1345 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1346 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1347 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1348 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1349 {Opt_commit, 0, MOPT_GTE0},
1350 {Opt_max_batch_time, 0, MOPT_GTE0},
1351 {Opt_min_batch_time, 0, MOPT_GTE0},
1352 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1353 {Opt_init_itable, 0, MOPT_GTE0},
1354 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1355 {Opt_resuid, 0, MOPT_GTE0},
1356 {Opt_resgid, 0, MOPT_GTE0},
1357 {Opt_journal_dev, 0, MOPT_GTE0},
1358 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1359 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1360 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1361 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1362 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1363 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1364 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1365#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1366 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1367 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1368#else
26092bf5
TT		 1369 {Opt_acl, 0, MOPT_NOSUPPORT},
1370 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1371#endif
26092bf5
TT		 1372 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1373 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1374 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1375 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1376 MOPT_SET | MOPT_Q},
1377 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1378 MOPT_SET | MOPT_Q},
1379 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1380 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1381 {Opt_usrjquota, 0, MOPT_Q},
1382 {Opt_grpjquota, 0, MOPT_Q},
1383 {Opt_offusrjquota, 0, MOPT_Q},
1384 {Opt_offgrpjquota, 0, MOPT_Q},
1385 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1386 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1387 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1388 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1389 {Opt_err, 0, 0}
1390};
1391
1392static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1393 substring_t *args, unsigned long *journal_devnum,
1394 unsigned int *journal_ioprio, int is_remount)
1395{
1396 struct ext4_sb_info *sbi = EXT4_SB(sb);
1397 const struct mount_opts *m;
08cefc7a
EB		 1398 kuid_t uid;
1399 kgid_t gid;
26092bf5
TT		 1400 int arg = 0;
1401
57f73c2c
TT		 1402#ifdef CONFIG_QUOTA
1403 if (token == Opt_usrjquota)
1404 return set_qf_name(sb, USRQUOTA, &args[0]);
1405 else if (token == Opt_grpjquota)
1406 return set_qf_name(sb, GRPQUOTA, &args[0]);
1407 else if (token == Opt_offusrjquota)
1408 return clear_qf_name(sb, USRQUOTA);
1409 else if (token == Opt_offgrpjquota)
1410 return clear_qf_name(sb, GRPQUOTA);
1411#endif
26092bf5 1412 switch (token) {
f7048605
TT		 1413 case Opt_noacl:
1414 case Opt_nouser_xattr:
1415 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1416 break;
26092bf5
TT		 1417 case Opt_sb:
1418 return 1; /* handled by get_sb_block() */
1419 case Opt_removed:
5f3633e3 1420 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1421 return 1;
26092bf5
TT		 1422 case Opt_abort:
1423 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1424 return 1;
1425 case Opt_i_version:
1426 sb->s_flags |= MS_I_VERSION;
1427 return 1;
26092bf5
TT		 1428 }
1429
5f3633e3
JK		 1430 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1431 if (token == m->token)
1432 break;
1433
1434 if (m->token == Opt_err) {
1435 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1436 "or missing value", opt);
1437 return -1;
1438 }
1439
8dc0aa8c
TT		 1440 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1441 ext4_msg(sb, KERN_ERR,
1442 "Mount option \"%s\" incompatible with ext2", opt);
1443 return -1;
1444 }
1445 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1446 ext4_msg(sb, KERN_ERR,
1447 "Mount option \"%s\" incompatible with ext3", opt);
1448 return -1;
1449 }
1450
5f3633e3
JK		 1451 if (args->from && match_int(args, &arg))
1452 return -1;
1453 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1454 return -1;
1455 if (m->flags & MOPT_EXPLICIT)
1456 set_opt2(sb, EXPLICIT_DELALLOC);
1457 if (m->flags & MOPT_CLEAR_ERR)
1458 clear_opt(sb, ERRORS_MASK);
1459 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1460 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1461 "options when quota turned on");
1462 return -1;
1463 }
1464
1465 if (m->flags & MOPT_NOSUPPORT) {
1466 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1467 } else if (token == Opt_commit) {
1468 if (arg == 0)
1469 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1470 sbi->s_commit_interval = HZ * arg;
1471 } else if (token == Opt_max_batch_time) {
1472 if (arg == 0)
1473 arg = EXT4_DEF_MAX_BATCH_TIME;
1474 sbi->s_max_batch_time = arg;
1475 } else if (token == Opt_min_batch_time) {
1476 sbi->s_min_batch_time = arg;
1477 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1478 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1479 ext4_msg(sb, KERN_ERR,
1480 "EXT4-fs: inode_readahead_blks must be "
1481 "0 or a power of 2 smaller than 2^31");
26092bf5 1482 return -1;
5f3633e3
JK		 1483 }
1484 sbi->s_inode_readahead_blks = arg;
1485 } else if (token == Opt_init_itable) {
1486 set_opt(sb, INIT_INODE_TABLE);
1487 if (!args->from)
1488 arg = EXT4_DEF_LI_WAIT_MULT;
1489 sbi->s_li_wait_mult = arg;
1490 } else if (token == Opt_max_dir_size_kb) {
1491 sbi->s_max_dir_size_kb = arg;
1492 } else if (token == Opt_stripe) {
1493 sbi->s_stripe = arg;
1494 } else if (token == Opt_resuid) {
1495 uid = make_kuid(current_user_ns(), arg);
1496 if (!uid_valid(uid)) {
1497 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1498 return -1;
1499 }
5f3633e3
JK		 1500 sbi->s_resuid = uid;
1501 } else if (token == Opt_resgid) {
1502 gid = make_kgid(current_user_ns(), arg);
1503 if (!gid_valid(gid)) {
1504 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1505 return -1;
1506 }
1507 sbi->s_resgid = gid;
1508 } else if (token == Opt_journal_dev) {
1509 if (is_remount) {
1510 ext4_msg(sb, KERN_ERR,
1511 "Cannot specify journal on remount");
1512 return -1;
1513 }
1514 *journal_devnum = arg;
1515 } else if (token == Opt_journal_ioprio) {
1516 if (arg > 7) {
1517 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1518 " (must be 0-7)");
1519 return -1;
1520 }
1521 *journal_ioprio =
1522 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1523 } else if (m->flags & MOPT_DATAJ) {
1524 if (is_remount) {
1525 if (!sbi->s_journal)
1526 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1527 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1528 ext4_msg(sb, KERN_ERR,
26092bf5 1529 "Cannot change data mode on remount");
26092bf5 1530 return -1;
ac27a0ec 1531 }
26092bf5 1532 } else {
5f3633e3
JK		 1533 clear_opt(sb, DATA_FLAGS);
1534 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1535 }
5f3633e3
JK		 1536#ifdef CONFIG_QUOTA
1537 } else if (m->flags & MOPT_QFMT) {
1538 if (sb_any_quota_loaded(sb) &&
1539 sbi->s_jquota_fmt != m->mount_opt) {
1540 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1541 "quota options when quota turned on");
1542 return -1;
1543 }
1544 sbi->s_jquota_fmt = m->mount_opt;
1545#endif
1546 } else {
1547 if (!args->from)
1548 arg = 1;
1549 if (m->flags & MOPT_CLEAR)
1550 arg = !arg;
1551 else if (unlikely(!(m->flags & MOPT_SET))) {
1552 ext4_msg(sb, KERN_WARNING,
1553 "buggy handling of option %s", opt);
1554 WARN_ON(1);
1555 return -1;
1556 }
1557 if (arg != 0)
1558 sbi->s_mount_opt |= m->mount_opt;
1559 else
1560 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1561 }
5f3633e3 1562 return 1;
26092bf5
TT		 1563}
1564
1565static int parse_options(char *options, struct super_block *sb,
1566 unsigned long *journal_devnum,
1567 unsigned int *journal_ioprio,
1568 int is_remount)
1569{
1570 struct ext4_sb_info *sbi = EXT4_SB(sb);
1571 char *p;
1572 substring_t args[MAX_OPT_ARGS];
1573 int token;
1574
1575 if (!options)
1576 return 1;
1577
1578 while ((p = strsep(&options, ",")) != NULL) {
1579 if (!*p)
1580 continue;
1581 /*
1582 * Initialize args struct so we know whether arg was
1583 * found; some options take optional arguments.
1584 */
caecd0af 1585 args[0].to = args[0].from = NULL;
26092bf5
TT		 1586 token = match_token(p, tokens, args);
1587 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1588 journal_ioprio, is_remount) < 0)
1589 return 0;
ac27a0ec
DK		 1590 }
1591#ifdef CONFIG_QUOTA
1592 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1593 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1594 clear_opt(sb, USRQUOTA);
ac27a0ec 1595
482a7425 1596 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1597 clear_opt(sb, GRPQUOTA);
ac27a0ec 1598
56c50f11 1599 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1600 ext4_msg(sb, KERN_ERR, "old and new quota "
1601 "format mixing");
ac27a0ec
DK		 1602 return 0;
1603 }
1604
1605 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1606 ext4_msg(sb, KERN_ERR, "journaled quota format "
1607 "not specified");
ac27a0ec
DK		 1608 return 0;
1609 }
1610 } else {
1611 if (sbi->s_jquota_fmt) {
b31e1552 1612 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1613 "specified with no journaling "
b31e1552 1614 "enabled");
ac27a0ec
DK		 1615 return 0;
1616 }
1617 }
1618#endif
261cb20c
JK		 1619 if (test_opt(sb, DIOREAD_NOLOCK)) {
1620 int blocksize =
1621 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1622
1623 if (blocksize < PAGE_CACHE_SIZE) {
1624 ext4_msg(sb, KERN_ERR, "can't mount with "
1625 "dioread_nolock if block size != PAGE_SIZE");
1626 return 0;
1627 }
1628 }
ac27a0ec
DK		 1629 return 1;
1630}
1631
2adf6da8
TT		 1632static inline void ext4_show_quota_options(struct seq_file *seq,
1633 struct super_block *sb)
1634{
1635#if defined(CONFIG_QUOTA)
1636 struct ext4_sb_info *sbi = EXT4_SB(sb);
1637
1638 if (sbi->s_jquota_fmt) {
1639 char *fmtname = "";
1640
1641 switch (sbi->s_jquota_fmt) {
1642 case QFMT_VFS_OLD:
1643 fmtname = "vfsold";
1644 break;
1645 case QFMT_VFS_V0:
1646 fmtname = "vfsv0";
1647 break;
1648 case QFMT_VFS_V1:
1649 fmtname = "vfsv1";
1650 break;
1651 }
1652 seq_printf(seq, ",jqfmt=%s", fmtname);
1653 }
1654
1655 if (sbi->s_qf_names[USRQUOTA])
1656 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1657
1658 if (sbi->s_qf_names[GRPQUOTA])
1659 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1660
1661 if (test_opt(sb, USRQUOTA))
1662 seq_puts(seq, ",usrquota");
1663
1664 if (test_opt(sb, GRPQUOTA))
1665 seq_puts(seq, ",grpquota");
1666#endif
1667}
1668
5a916be1
TT		 1669static const char *token2str(int token)
1670{
50df9fd5 1671 const struct match_token *t;
5a916be1
TT		 1672
1673 for (t = tokens; t->token != Opt_err; t++)
1674 if (t->token == token && !strchr(t->pattern, '='))
1675 break;
1676 return t->pattern;
1677}
1678
2adf6da8
TT		 1679/*
1680 * Show an option if
1681 * - it's set to a non-default value OR
1682 * - if the per-sb default is different from the global default
1683 */
66acdcf4
TT		 1684static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1685 int nodefs)
2adf6da8 1686{
2adf6da8
TT		 1687 struct ext4_sb_info *sbi = EXT4_SB(sb);
1688 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1689 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1690 const struct mount_opts *m;
66acdcf4 1691 char sep = nodefs ? '\n' : ',';
2adf6da8 1692
66acdcf4
TT		 1693#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1694#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1695
1696 if (sbi->s_sb_block != 1)
5a916be1
TT		 1697 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1698
1699 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1700 int want_set = m->flags & MOPT_SET;
1701 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1702 (m->flags & MOPT_CLEAR_ERR))
1703 continue;
66acdcf4 1704 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1705 continue; /* skip if same as the default */
1706 if ((want_set &&
1707 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1708 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1709 continue; /* select Opt_noFoo vs Opt_Foo */
1710 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1711 }
5a916be1 1712
08cefc7a 1713 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1714 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1715 SEQ_OPTS_PRINT("resuid=%u",
1716 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1717 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1718 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1719 SEQ_OPTS_PRINT("resgid=%u",
1720 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1721 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1722 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1723 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1724 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1725 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1726 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1727 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1728 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1729 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1730 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1731 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1732 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1733 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1734 if (sb->s_flags & MS_I_VERSION)
5a916be1 1735 SEQ_OPTS_PUTS("i_version");
66acdcf4 1736 if (nodefs || sbi->s_stripe)
5a916be1 1737 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1738 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1739 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1740 SEQ_OPTS_PUTS("data=journal");
1741 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1742 SEQ_OPTS_PUTS("data=ordered");
1743 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1744 SEQ_OPTS_PUTS("data=writeback");
1745 }
66acdcf4
TT		 1746 if (nodefs ||
1747 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1748 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1749 sbi->s_inode_readahead_blks);
2adf6da8 1750
66acdcf4
TT		 1751 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1752 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1753 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1754 if (nodefs || sbi->s_max_dir_size_kb)
1755 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1756
1757 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1758 return 0;
1759}
1760
66acdcf4
TT		 1761static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1762{
1763 return _ext4_show_options(seq, root->d_sb, 0);
1764}
1765
1766static int options_seq_show(struct seq_file *seq, void *offset)
1767{
1768 struct super_block *sb = seq->private;
1769 int rc;
1770
1771 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1772 rc = _ext4_show_options(seq, sb, 1);
1773 seq_puts(seq, "\n");
1774 return rc;
1775}
1776
1777static int options_open_fs(struct inode *inode, struct file *file)
1778{
1779 return single_open(file, options_seq_show, PDE(inode)->data);
1780}
1781
1782static const struct file_operations ext4_seq_options_fops = {
1783 .owner = THIS_MODULE,
1784 .open = options_open_fs,
1785 .read = seq_read,
1786 .llseek = seq_lseek,
1787 .release = single_release,
1788};
1789
617ba13b 1790static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1791 int read_only)
1792{
617ba13b 1793 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1794 int res = 0;
1795
617ba13b 1796 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1797 ext4_msg(sb, KERN_ERR, "revision level too high, "
1798 "forcing read-only mode");
ac27a0ec
DK		 1799 res = MS_RDONLY;
1800 }
1801 if (read_only)
281b5995 1802 goto done;
617ba13b 1803 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1804 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1805 "running e2fsck is recommended");
617ba13b 1806 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1807 ext4_msg(sb, KERN_WARNING,
1808 "warning: mounting fs with errors, "
1809 "running e2fsck is recommended");
ed3ce80a 1810 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1811 le16_to_cpu(es->s_mnt_count) >=
1812 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1813 ext4_msg(sb, KERN_WARNING,
1814 "warning: maximal mount count reached, "
1815 "running e2fsck is recommended");
ac27a0ec
DK		 1816 else if (le32_to_cpu(es->s_checkinterval) &&
1817 (le32_to_cpu(es->s_lastcheck) +
1818 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1819 ext4_msg(sb, KERN_WARNING,
1820 "warning: checktime reached, "
1821 "running e2fsck is recommended");
0b8e58a1 1822 if (!sbi->s_journal)
0390131b 1823 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1824 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1825 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1826 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1827 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1828 ext4_update_dynamic_rev(sb);
0390131b
FM		 1829 if (sbi->s_journal)
1830 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1831
e2d67052 1832 ext4_commit_super(sb, 1);
281b5995 1833done:
ac27a0ec 1834 if (test_opt(sb, DEBUG))
a9df9a49 1835 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1836 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1837 sb->s_blocksize,
1838 sbi->s_groups_count,
617ba13b
MC		 1839 EXT4_BLOCKS_PER_GROUP(sb),
1840 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1841 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1842
7abc52c2 1843 cleancache_init_fs(sb);
ac27a0ec
DK		 1844 return res;
1845}
1846
117fff10
TT		 1847int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1848{
1849 struct ext4_sb_info *sbi = EXT4_SB(sb);
1850 struct flex_groups *new_groups;
1851 int size;
1852
1853 if (!sbi->s_log_groups_per_flex)
1854 return 0;
1855
1856 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1857 if (size <= sbi->s_flex_groups_allocated)
1858 return 0;
1859
1860 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1861 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1862 if (!new_groups) {
1863 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1864 size / (int) sizeof(struct flex_groups));
1865 return -ENOMEM;
1866 }
1867
1868 if (sbi->s_flex_groups) {
1869 memcpy(new_groups, sbi->s_flex_groups,
1870 (sbi->s_flex_groups_allocated *
1871 sizeof(struct flex_groups)));
1872 ext4_kvfree(sbi->s_flex_groups);
1873 }
1874 sbi->s_flex_groups = new_groups;
1875 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1876 return 0;
1877}
1878
772cb7c8
JS		 1879static int ext4_fill_flex_info(struct super_block *sb)
1880{
1881 struct ext4_sb_info *sbi = EXT4_SB(sb);
1882 struct ext4_group_desc *gdp = NULL;
772cb7c8 1883 ext4_group_t flex_group;
d50f2ab6 1884 unsigned int groups_per_flex = 0;
117fff10 1885 int i, err;
772cb7c8 1886
503358ae 1887 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1888 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1889 sbi->s_log_groups_per_flex = 0;
1890 return 1;
1891 }
76495ec1 1892 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
772cb7c8 1893
117fff10
TT		 1894 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1895 if (err)
9933fc0a 1896 goto failed;
772cb7c8 1897
772cb7c8 1898 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1899 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1900
1901 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1902 atomic_add(ext4_free_inodes_count(sb, gdp),
1903 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1904 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1905 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1906 atomic_add(ext4_used_dirs_count(sb, gdp),
1907 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1908 }
1909
1910 return 1;
1911failed:
1912 return 0;
1913}
1914
feb0ab32
DW		 1915static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1916 struct ext4_group_desc *gdp)
717d50e4 1917{
feb0ab32 1918 int offset;
717d50e4 1919 __u16 crc = 0;
feb0ab32 1920 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1921
feb0ab32
DW		 1922 if ((sbi->s_es->s_feature_ro_compat &
1923 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1924 /* Use new metadata_csum algorithm */
1925 __u16 old_csum;
1926 __u32 csum32;
1927
1928 old_csum = gdp->bg_checksum;
1929 gdp->bg_checksum = 0;
1930 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1931 sizeof(le_group));
1932 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1933 sbi->s_desc_size);
1934 gdp->bg_checksum = old_csum;
1935
1936 crc = csum32 & 0xFFFF;
1937 goto out;
717d50e4
AD		 1938 }
1939
feb0ab32
DW		 1940 /* old crc16 code */
1941 offset = offsetof(struct ext4_group_desc, bg_checksum);
1942
1943 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1944 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1945 crc = crc16(crc, (__u8 *)gdp, offset);
1946 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1947 /* for checksum of struct ext4_group_desc do the rest...*/
1948 if ((sbi->s_es->s_feature_incompat &
1949 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1950 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1951 crc = crc16(crc, (__u8 *)gdp + offset,
1952 le16_to_cpu(sbi->s_es->s_desc_size) -
1953 offset);
1954
1955out:
717d50e4
AD		 1956 return cpu_to_le16(crc);
1957}
1958
feb0ab32 1959int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 1960 struct ext4_group_desc *gdp)
1961{
feb0ab32
DW		 1962 if (ext4_has_group_desc_csum(sb) &&
1963 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1964 block_group, gdp)))
717d50e4
AD		 1965 return 0;
1966
1967 return 1;
1968}
1969
feb0ab32
DW		 1970void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1971 struct ext4_group_desc *gdp)
1972{
1973 if (!ext4_has_group_desc_csum(sb))
1974 return;
1975 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
1976}
1977
ac27a0ec 1978/* Called at mount-time, super-block is locked */
bfff6873
LC		 1979static int ext4_check_descriptors(struct super_block *sb,
1980 ext4_group_t *first_not_zeroed)
ac27a0ec 1981{
617ba13b
MC		 1982 struct ext4_sb_info *sbi = EXT4_SB(sb);
1983 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1984 ext4_fsblk_t last_block;
bd81d8ee
LV		 1985 ext4_fsblk_t block_bitmap;
1986 ext4_fsblk_t inode_bitmap;
1987 ext4_fsblk_t inode_table;
ce421581 1988 int flexbg_flag = 0;
bfff6873 1989 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 1990
ce421581
JS		 1991 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1992 flexbg_flag = 1;
1993
af5bc92d 1994 ext4_debug("Checking group descriptors");
ac27a0ec 1995
197cd65a
AM		 1996 for (i = 0; i < sbi->s_groups_count; i++) {
1997 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1998
ce421581 1999 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2000 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2001 else
2002 last_block = first_block +
617ba13b 2003 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2004
bfff6873
LC		 2005 if ((grp == sbi->s_groups_count) &&
2006 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2007 grp = i;
2008
8fadc143 2009 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2010 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2011 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2012 "Block bitmap for group %u not in group "
b31e1552 2013 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2014 return 0;
2015 }
8fadc143 2016 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2017 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2018 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2019 "Inode bitmap for group %u not in group "
b31e1552 2020 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2021 return 0;
2022 }
8fadc143 2023 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2024 if (inode_table < first_block ||
2b2d6d01 2025 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2026 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2027 "Inode table for group %u not in group "
b31e1552 2028 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2029 return 0;
2030 }
955ce5f5 2031 ext4_lock_group(sb, i);
feb0ab32 2032 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2033 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2034 "Checksum for group %u failed (%u!=%u)",
2035 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2036 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2037 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2038 ext4_unlock_group(sb, i);
8a266467 2039 return 0;
7ee1ec4c 2040 }
717d50e4 2041 }
955ce5f5 2042 ext4_unlock_group(sb, i);
ce421581
JS		 2043 if (!flexbg_flag)
2044 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2045 }
bfff6873
LC		 2046 if (NULL != first_not_zeroed)
2047 *first_not_zeroed = grp;
ac27a0ec 2048
5dee5437
TT		 2049 ext4_free_blocks_count_set(sbi->s_es,
2050 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2051 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2052 return 1;
2053}
2054
617ba13b 2055/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2056 * the superblock) which were deleted from all directories, but held open by
2057 * a process at the time of a crash. We walk the list and try to delete these
2058 * inodes at recovery time (only with a read-write filesystem).
2059 *
2060 * In order to keep the orphan inode chain consistent during traversal (in
2061 * case of crash during recovery), we link each inode into the superblock
2062 * orphan list_head and handle it the same way as an inode deletion during
2063 * normal operation (which journals the operations for us).
2064 *
2065 * We only do an iget() and an iput() on each inode, which is very safe if we
2066 * accidentally point at an in-use or already deleted inode. The worst that
2067 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2068 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2069 * e2fsck was run on this filesystem, and it must have already done the orphan
2070 * inode cleanup for us, so we can safely abort without any further action.
2071 */
2b2d6d01
TT		 2072static void ext4_orphan_cleanup(struct super_block *sb,
2073 struct ext4_super_block *es)
ac27a0ec
DK		 2074{
2075 unsigned int s_flags = sb->s_flags;
2076 int nr_orphans = 0, nr_truncates = 0;
2077#ifdef CONFIG_QUOTA
2078 int i;
2079#endif
2080 if (!es->s_last_orphan) {
2081 jbd_debug(4, "no orphan inodes to clean up\n");
2082 return;
2083 }
2084
a8f48a95 2085 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2086 ext4_msg(sb, KERN_ERR, "write access "
2087 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2088 return;
2089 }
2090
d39195c3
AG		 2091 /* Check if feature set would not allow a r/w mount */
2092 if (!ext4_feature_set_ok(sb, 0)) {
2093 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2094 "unknown ROCOMPAT features");
2095 return;
2096 }
2097
617ba13b 2098 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2099 /* don't clear list on RO mount w/ errors */
2100 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2101 jbd_debug(1, "Errors on filesystem, "
2102 "clearing orphan list.\n");
c25f9bc6
ES		 2103 es->s_last_orphan = 0;
2104 }
ac27a0ec
DK		 2105 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2106 return;
2107 }
2108
2109 if (s_flags & MS_RDONLY) {
b31e1552 2110 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2111 sb->s_flags &= ~MS_RDONLY;
2112 }
2113#ifdef CONFIG_QUOTA
2114 /* Needed for iput() to work correctly and not trash data */
2115 sb->s_flags |= MS_ACTIVE;
2116 /* Turn on quotas so that they are updated correctly */
2117 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2118 if (EXT4_SB(sb)->s_qf_names[i]) {
2119 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2120 if (ret < 0)
b31e1552
ES		 2121 ext4_msg(sb, KERN_ERR,
2122 "Cannot turn on journaled "
2123 "quota: error %d", ret);
ac27a0ec
DK		 2124 }
2125 }
2126#endif
2127
2128 while (es->s_last_orphan) {
2129 struct inode *inode;
2130
97bd42b9
JB		 2131 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2132 if (IS_ERR(inode)) {
ac27a0ec
DK		 2133 es->s_last_orphan = 0;
2134 break;
2135 }
2136
617ba13b 2137 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2138 dquot_initialize(inode);
ac27a0ec 2139 if (inode->i_nlink) {
b31e1552
ES		 2140 ext4_msg(sb, KERN_DEBUG,
2141 "%s: truncating inode %lu to %lld bytes",
46e665e9 2142 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2143 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2144 inode->i_ino, inode->i_size);
721e3eba 2145 mutex_lock(&inode->i_mutex);
617ba13b 2146 ext4_truncate(inode);
721e3eba 2147 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2148 nr_truncates++;
2149 } else {
b31e1552
ES		 2150 ext4_msg(sb, KERN_DEBUG,
2151 "%s: deleting unreferenced inode %lu",
46e665e9 2152 __func__, inode->i_ino);
ac27a0ec
DK		 2153 jbd_debug(2, "deleting unreferenced inode %lu\n",
2154 inode->i_ino);
2155 nr_orphans++;
2156 }
2157 iput(inode); /* The delete magic happens here! */
2158 }
2159
2b2d6d01 2160#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2161
2162 if (nr_orphans)
b31e1552
ES		 2163 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2164 PLURAL(nr_orphans));
ac27a0ec 2165 if (nr_truncates)
b31e1552
ES		 2166 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2167 PLURAL(nr_truncates));
ac27a0ec
DK		 2168#ifdef CONFIG_QUOTA
2169 /* Turn quotas off */
2170 for (i = 0; i < MAXQUOTAS; i++) {
2171 if (sb_dqopt(sb)->files[i])
287a8095 2172 dquot_quota_off(sb, i);
ac27a0ec
DK		 2173 }
2174#endif
2175 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2176}
0b8e58a1 2177
cd2291a4
ES		 2178/*
2179 * Maximal extent format file size.
2180 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2181 * extent format containers, within a sector_t, and within i_blocks
2182 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2183 * so that won't be a limiting factor.
2184 *
f17722f9
LC		 2185 * However there is other limiting factor. We do store extents in the form
2186 * of starting block and length, hence the resulting length of the extent
2187 * covering maximum file size must fit into on-disk format containers as
2188 * well. Given that length is always by 1 unit bigger than max unit (because
2189 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2190 *
cd2291a4
ES		 2191 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2192 */
f287a1a5 2193static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2194{
2195 loff_t res;
2196 loff_t upper_limit = MAX_LFS_FILESIZE;
2197
2198 /* small i_blocks in vfs inode? */
f287a1a5 2199 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2200 /*
90c699a9 2201 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2202 * i_block represent total blocks in 512 bytes
2203 * 32 == size of vfs inode i_blocks * 8
2204 */
2205 upper_limit = (1LL << 32) - 1;
2206
2207 /* total blocks in file system block size */
2208 upper_limit >>= (blkbits - 9);
2209 upper_limit <<= blkbits;
2210 }
2211
f17722f9
LC		 2212 /*
2213 * 32-bit extent-start container, ee_block. We lower the maxbytes
2214 * by one fs block, so ee_len can cover the extent of maximum file
2215 * size
2216 */
2217 res = (1LL << 32) - 1;
cd2291a4 2218 res <<= blkbits;
cd2291a4
ES		 2219
2220 /* Sanity check against vm- & vfs- imposed limits */
2221 if (res > upper_limit)
2222 res = upper_limit;
2223
2224 return res;
2225}
ac27a0ec 2226
ac27a0ec 2227/*
cd2291a4 2228 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2229 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2230 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2231 */
f287a1a5 2232static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2233{
617ba13b 2234 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2235 int meta_blocks;
2236 loff_t upper_limit;
0b8e58a1
AD		 2237 /* This is calculated to be the largest file size for a dense, block
2238 * mapped file such that the file's total number of 512-byte sectors,
2239 * including data and all indirect blocks, does not exceed (2^48 - 1).
2240 *
2241 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2242 * number of 512-byte sectors of the file.
0fc1b451
AK		 2243 */
2244
f287a1a5 2245 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2246 /*
90c699a9 2247 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2248 * the inode i_block field represents total file blocks in
2249 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2250 */
2251 upper_limit = (1LL << 32) - 1;
2252
2253 /* total blocks in file system block size */
2254 upper_limit >>= (bits - 9);
2255
2256 } else {
8180a562
AK		 2257 /*
2258 * We use 48 bit ext4_inode i_blocks
2259 * With EXT4_HUGE_FILE_FL set the i_blocks
2260 * represent total number of blocks in
2261 * file system block size
2262 */
0fc1b451
AK		 2263 upper_limit = (1LL << 48) - 1;
2264
0fc1b451
AK		 2265 }
2266
2267 /* indirect blocks */
2268 meta_blocks = 1;
2269 /* double indirect blocks */
2270 meta_blocks += 1 + (1LL << (bits-2));
2271 /* tripple indirect blocks */
2272 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2273
2274 upper_limit -= meta_blocks;
2275 upper_limit <<= bits;
ac27a0ec
DK		 2276
2277 res += 1LL << (bits-2);
2278 res += 1LL << (2*(bits-2));
2279 res += 1LL << (3*(bits-2));
2280 res <<= bits;
2281 if (res > upper_limit)
2282 res = upper_limit;
0fc1b451
AK		 2283
2284 if (res > MAX_LFS_FILESIZE)
2285 res = MAX_LFS_FILESIZE;
2286
ac27a0ec
DK		 2287 return res;
2288}
2289
617ba13b 2290static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2291 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2292{
617ba13b 2293 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2294 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2295 int has_super = 0;
2296
2297 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2298
617ba13b 2299 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2300 nr < first_meta_bg)
70bbb3e0 2301 return logical_sb_block + nr + 1;
ac27a0ec 2302 bg = sbi->s_desc_per_block * nr;
617ba13b 2303 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2304 has_super = 1;
0b8e58a1 2305
617ba13b 2306 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2307}
2308
c9de560d
AT		 2309/**
2310 * ext4_get_stripe_size: Get the stripe size.
2311 * @sbi: In memory super block info
2312 *
2313 * If we have specified it via mount option, then
2314 * use the mount option value. If the value specified at mount time is
2315 * greater than the blocks per group use the super block value.
2316 * If the super block value is greater than blocks per group return 0.
2317 * Allocator needs it be less than blocks per group.
2318 *
2319 */
2320static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2321{
2322 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2323 unsigned long stripe_width =
2324 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2325 int ret;
c9de560d
AT		 2326
2327 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2328 ret = sbi->s_stripe;
2329 else if (stripe_width <= sbi->s_blocks_per_group)
2330 ret = stripe_width;
2331 else if (stride <= sbi->s_blocks_per_group)
2332 ret = stride;
2333 else
2334 ret = 0;
c9de560d 2335
3eb08658
DE		 2336 /*
2337 * If the stripe width is 1, this makes no sense and
2338 * we set it to 0 to turn off stripe handling code.
2339 */
2340 if (ret <= 1)
2341 ret = 0;
c9de560d 2342
3eb08658 2343 return ret;
c9de560d 2344}
ac27a0ec 2345
3197ebdb
TT		 2346/* sysfs supprt */
2347
2348struct ext4_attr {
2349 struct attribute attr;
2350 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2351 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2352 const char *, size_t);
2353 int offset;
2354};
2355
2356static int parse_strtoul(const char *buf,
2357 unsigned long max, unsigned long *value)
2358{
2359 char *endp;
2360
e7d2860b
AGR		 2361 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2362 endp = skip_spaces(endp);
3197ebdb
TT		 2363 if (*endp || *value > max)
2364 return -EINVAL;
2365
2366 return 0;
2367}
2368
2369static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2370 struct ext4_sb_info *sbi,
2371 char *buf)
2372{
2373 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2374 (s64) EXT4_C2B(sbi,
2375 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2376}
2377
2378static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2379 struct ext4_sb_info *sbi, char *buf)
2380{
2381 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2382
f613dfcb
TT		 2383 if (!sb->s_bdev->bd_part)
2384 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2385 return snprintf(buf, PAGE_SIZE, "%lu\n",
2386 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2387 sbi->s_sectors_written_start) >> 1);
2388}
2389
2390static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2391 struct ext4_sb_info *sbi, char *buf)
2392{
2393 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2394
f613dfcb
TT		 2395 if (!sb->s_bdev->bd_part)
2396 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2397 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2398 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2399 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2400 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2401}
2402
2403static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2404 struct ext4_sb_info *sbi,
2405 const char *buf, size_t count)
2406{
2407 unsigned long t;
2408
2409 if (parse_strtoul(buf, 0x40000000, &t))
2410 return -EINVAL;
2411
5dbd571d 2412 if (t && !is_power_of_2(t))
3197ebdb
TT		 2413 return -EINVAL;
2414
2415 sbi->s_inode_readahead_blks = t;
2416 return count;
2417}
2418
2419static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2420 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2421{
2422 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2423
2424 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2425}
2426
2427static ssize_t sbi_ui_store(struct ext4_attr *a,
2428 struct ext4_sb_info *sbi,
2429 const char *buf, size_t count)
2430{
2431 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2432 unsigned long t;
2433
2434 if (parse_strtoul(buf, 0xffffffff, &t))
2435 return -EINVAL;
2436 *ui = t;
2437 return count;
2438}
2439
2c0544b2
TT		 2440static ssize_t trigger_test_error(struct ext4_attr *a,
2441 struct ext4_sb_info *sbi,
2442 const char *buf, size_t count)
2443{
2444 int len = count;
2445
2446 if (!capable(CAP_SYS_ADMIN))
2447 return -EPERM;
2448
2449 if (len && buf[len-1] == '\n')
2450 len--;
2451
2452 if (len)
2453 ext4_error(sbi->s_sb, "%.*s", len, buf);
2454 return count;
2455}
2456
3197ebdb
TT		 2457#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2458static struct ext4_attr ext4_attr_##_name = { \
2459 .attr = {.name = __stringify(_name), .mode = _mode }, \
2460 .show = _show, \
2461 .store = _store, \
2462 .offset = offsetof(struct ext4_sb_info, _elname), \
2463}
2464#define EXT4_ATTR(name, mode, show, store) \
2465static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2466
857ac889 2467#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2468#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2469#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2470#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2471 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2472#define ATTR_LIST(name) &ext4_attr_##name.attr
2473
2474EXT4_RO_ATTR(delayed_allocation_blocks);
2475EXT4_RO_ATTR(session_write_kbytes);
2476EXT4_RO_ATTR(lifetime_write_kbytes);
2477EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2478 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2479EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2480EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2481EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2482EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2483EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2484EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2485EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2486EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
67a5da56 2487EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2488EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2489
2490static struct attribute *ext4_attrs[] = {
2491 ATTR_LIST(delayed_allocation_blocks),
2492 ATTR_LIST(session_write_kbytes),
2493 ATTR_LIST(lifetime_write_kbytes),
2494 ATTR_LIST(inode_readahead_blks),
11013911 2495 ATTR_LIST(inode_goal),
3197ebdb
TT		 2496 ATTR_LIST(mb_stats),
2497 ATTR_LIST(mb_max_to_scan),
2498 ATTR_LIST(mb_min_to_scan),
2499 ATTR_LIST(mb_order2_req),
2500 ATTR_LIST(mb_stream_req),
2501 ATTR_LIST(mb_group_prealloc),
55138e0b 2502 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2503 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2504 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2505 NULL,
2506};
2507
857ac889
LC		 2508/* Features this copy of ext4 supports */
2509EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2510EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2511EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2512
2513static struct attribute *ext4_feat_attrs[] = {
2514 ATTR_LIST(lazy_itable_init),
27ee40df 2515 ATTR_LIST(batched_discard),
5e7bbef1 2516 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2517 NULL,
2518};
2519
3197ebdb
TT		 2520static ssize_t ext4_attr_show(struct kobject *kobj,
2521 struct attribute *attr, char *buf)
2522{
2523 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2524 s_kobj);
2525 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2526
2527 return a->show ? a->show(a, sbi, buf) : 0;
2528}
2529
2530static ssize_t ext4_attr_store(struct kobject *kobj,
2531 struct attribute *attr,
2532 const char *buf, size_t len)
2533{
2534 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2535 s_kobj);
2536 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2537
2538 return a->store ? a->store(a, sbi, buf, len) : 0;
2539}
2540
2541static void ext4_sb_release(struct kobject *kobj)
2542{
2543 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2544 s_kobj);
2545 complete(&sbi->s_kobj_unregister);
2546}
2547
52cf25d0 2548static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2549 .show = ext4_attr_show,
2550 .store = ext4_attr_store,
2551};
2552
2553static struct kobj_type ext4_ktype = {
2554 .default_attrs = ext4_attrs,
2555 .sysfs_ops = &ext4_attr_ops,
2556 .release = ext4_sb_release,
2557};
2558
857ac889
LC		 2559static void ext4_feat_release(struct kobject *kobj)
2560{
2561 complete(&ext4_feat->f_kobj_unregister);
2562}
2563
2564static struct kobj_type ext4_feat_ktype = {
2565 .default_attrs = ext4_feat_attrs,
2566 .sysfs_ops = &ext4_attr_ops,
2567 .release = ext4_feat_release,
2568};
2569
a13fb1a4
ES		 2570/*
2571 * Check whether this filesystem can be mounted based on
2572 * the features present and the RDONLY/RDWR mount requested.
2573 * Returns 1 if this filesystem can be mounted as requested,
2574 * 0 if it cannot be.
2575 */
2576static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2577{
2578 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2579 ext4_msg(sb, KERN_ERR,
2580 "Couldn't mount because of "
2581 "unsupported optional features (%x)",
2582 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2583 ~EXT4_FEATURE_INCOMPAT_SUPP));
2584 return 0;
2585 }
2586
2587 if (readonly)
2588 return 1;
2589
2590 /* Check that feature set is OK for a read-write mount */
2591 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2592 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2593 "unsupported optional features (%x)",
2594 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2595 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2596 return 0;
2597 }
2598 /*
2599 * Large file size enabled file system can only be mounted
2600 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2601 */
2602 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2603 if (sizeof(blkcnt_t) < sizeof(u64)) {
2604 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2605 "cannot be mounted RDWR without "
2606 "CONFIG_LBDAF");
2607 return 0;
2608 }
2609 }
bab08ab9
TT		 2610 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2611 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2612 ext4_msg(sb, KERN_ERR,
2613 "Can't support bigalloc feature without "
2614 "extents feature\n");
2615 return 0;
2616 }
7c319d32
AK		 2617
2618#ifndef CONFIG_QUOTA
2619 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2620 !readonly) {
2621 ext4_msg(sb, KERN_ERR,
2622 "Filesystem with quota feature cannot be mounted RDWR "
2623 "without CONFIG_QUOTA");
2624 return 0;
2625 }
2626#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2627 return 1;
2628}
2629
66e61a9e
TT		 2630/*
2631 * This function is called once a day if we have errors logged
2632 * on the file system
2633 */
2634static void print_daily_error_info(unsigned long arg)
2635{
2636 struct super_block *sb = (struct super_block *) arg;
2637 struct ext4_sb_info *sbi;
2638 struct ext4_super_block *es;
2639
2640 sbi = EXT4_SB(sb);
2641 es = sbi->s_es;
2642
2643 if (es->s_error_count)
2644 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2645 le32_to_cpu(es->s_error_count));
2646 if (es->s_first_error_time) {
2647 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2648 sb->s_id, le32_to_cpu(es->s_first_error_time),
2649 (int) sizeof(es->s_first_error_func),
2650 es->s_first_error_func,
2651 le32_to_cpu(es->s_first_error_line));
2652 if (es->s_first_error_ino)
2653 printk(": inode %u",
2654 le32_to_cpu(es->s_first_error_ino));
2655 if (es->s_first_error_block)
2656 printk(": block %llu", (unsigned long long)
2657 le64_to_cpu(es->s_first_error_block));
2658 printk("\n");
2659 }
2660 if (es->s_last_error_time) {
2661 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2662 sb->s_id, le32_to_cpu(es->s_last_error_time),
2663 (int) sizeof(es->s_last_error_func),
2664 es->s_last_error_func,
2665 le32_to_cpu(es->s_last_error_line));
2666 if (es->s_last_error_ino)
2667 printk(": inode %u",
2668 le32_to_cpu(es->s_last_error_ino));
2669 if (es->s_last_error_block)
2670 printk(": block %llu", (unsigned long long)
2671 le64_to_cpu(es->s_last_error_block));
2672 printk("\n");
2673 }
2674 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2675}
2676
bfff6873
LC		 2677/* Find next suitable group and run ext4_init_inode_table */
2678static int ext4_run_li_request(struct ext4_li_request *elr)
2679{
2680 struct ext4_group_desc *gdp = NULL;
2681 ext4_group_t group, ngroups;
2682 struct super_block *sb;
2683 unsigned long timeout = 0;
2684 int ret = 0;
2685
2686 sb = elr->lr_super;
2687 ngroups = EXT4_SB(sb)->s_groups_count;
2688
8e8ad8a5 2689 sb_start_write(sb);
bfff6873
LC		 2690 for (group = elr->lr_next_group; group < ngroups; group++) {
2691 gdp = ext4_get_group_desc(sb, group, NULL);
2692 if (!gdp) {
2693 ret = 1;
2694 break;
2695 }
2696
2697 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2698 break;
2699 }
2700
7f511862 2701 if (group >= ngroups)
bfff6873
LC		 2702 ret = 1;
2703
2704 if (!ret) {
2705 timeout = jiffies;
2706 ret = ext4_init_inode_table(sb, group,
2707 elr->lr_timeout ? 0 : 1);
2708 if (elr->lr_timeout == 0) {
51ce6511
LC		 2709 timeout = (jiffies - timeout) *
2710 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2711 elr->lr_timeout = timeout;
2712 }
2713 elr->lr_next_sched = jiffies + elr->lr_timeout;
2714 elr->lr_next_group = group + 1;
2715 }
8e8ad8a5 2716 sb_end_write(sb);
bfff6873
LC		 2717
2718 return ret;
2719}
2720
2721/*
2722 * Remove lr_request from the list_request and free the
4ed5c033 2723 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2724 */
2725static void ext4_remove_li_request(struct ext4_li_request *elr)
2726{
2727 struct ext4_sb_info *sbi;
2728
2729 if (!elr)
2730 return;
2731
2732 sbi = elr->lr_sbi;
2733
2734 list_del(&elr->lr_request);
2735 sbi->s_li_request = NULL;
2736 kfree(elr);
2737}
2738
2739static void ext4_unregister_li_request(struct super_block *sb)
2740{
1bb933fb
LC		 2741 mutex_lock(&ext4_li_mtx);
2742 if (!ext4_li_info) {
2743 mutex_unlock(&ext4_li_mtx);
bfff6873 2744 return;
1bb933fb 2745 }
bfff6873
LC		 2746
2747 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2748 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2749 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2750 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2751}
2752
8f1f7453
ES		 2753static struct task_struct *ext4_lazyinit_task;
2754
bfff6873
LC		 2755/*
2756 * This is the function where ext4lazyinit thread lives. It walks
2757 * through the request list searching for next scheduled filesystem.
2758 * When such a fs is found, run the lazy initialization request
2759 * (ext4_rn_li_request) and keep track of the time spend in this
2760 * function. Based on that time we compute next schedule time of
2761 * the request. When walking through the list is complete, compute
2762 * next waking time and put itself into sleep.
2763 */
2764static int ext4_lazyinit_thread(void *arg)
2765{
2766 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2767 struct list_head *pos, *n;
2768 struct ext4_li_request *elr;
4ed5c033 2769 unsigned long next_wakeup, cur;
bfff6873
LC		 2770
2771 BUG_ON(NULL == eli);
2772
bfff6873
LC		 2773cont_thread:
2774 while (true) {
2775 next_wakeup = MAX_JIFFY_OFFSET;
2776
2777 mutex_lock(&eli->li_list_mtx);
2778 if (list_empty(&eli->li_request_list)) {
2779 mutex_unlock(&eli->li_list_mtx);
2780 goto exit_thread;
2781 }
2782
2783 list_for_each_safe(pos, n, &eli->li_request_list) {
2784 elr = list_entry(pos, struct ext4_li_request,
2785 lr_request);
2786
b2c78cd0
TT		 2787 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2788 if (ext4_run_li_request(elr) != 0) {
2789 /* error, remove the lazy_init job */
2790 ext4_remove_li_request(elr);
2791 continue;
2792 }
bfff6873
LC		 2793 }
2794
2795 if (time_before(elr->lr_next_sched, next_wakeup))
2796 next_wakeup = elr->lr_next_sched;
2797 }
2798 mutex_unlock(&eli->li_list_mtx);
2799
a0acae0e 2800 try_to_freeze();
bfff6873 2801
4ed5c033
LC		 2802 cur = jiffies;
2803 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2804 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2805 cond_resched();
2806 continue;
2807 }
2808
4ed5c033
LC		 2809 schedule_timeout_interruptible(next_wakeup - cur);
2810
8f1f7453
ES		 2811 if (kthread_should_stop()) {
2812 ext4_clear_request_list();
2813 goto exit_thread;
2814 }
bfff6873
LC		 2815 }
2816
2817exit_thread:
2818 /*
2819 * It looks like the request list is empty, but we need
2820 * to check it under the li_list_mtx lock, to prevent any
2821 * additions into it, and of course we should lock ext4_li_mtx
2822 * to atomically free the list and ext4_li_info, because at
2823 * this point another ext4 filesystem could be registering
2824 * new one.
2825 */
2826 mutex_lock(&ext4_li_mtx);
2827 mutex_lock(&eli->li_list_mtx);
2828 if (!list_empty(&eli->li_request_list)) {
2829 mutex_unlock(&eli->li_list_mtx);
2830 mutex_unlock(&ext4_li_mtx);
2831 goto cont_thread;
2832 }
2833 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2834 kfree(ext4_li_info);
2835 ext4_li_info = NULL;
2836 mutex_unlock(&ext4_li_mtx);
2837
2838 return 0;
2839}
2840
2841static void ext4_clear_request_list(void)
2842{
2843 struct list_head *pos, *n;
2844 struct ext4_li_request *elr;
2845
2846 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2847 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2848 elr = list_entry(pos, struct ext4_li_request,
2849 lr_request);
2850 ext4_remove_li_request(elr);
2851 }
2852 mutex_unlock(&ext4_li_info->li_list_mtx);
2853}
2854
2855static int ext4_run_lazyinit_thread(void)
2856{
8f1f7453
ES		 2857 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2858 ext4_li_info, "ext4lazyinit");
2859 if (IS_ERR(ext4_lazyinit_task)) {
2860 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2861 ext4_clear_request_list();
bfff6873
LC		 2862 kfree(ext4_li_info);
2863 ext4_li_info = NULL;
92b97816 2864 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2865 "initialization thread\n",
2866 err);
2867 return err;
2868 }
2869 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2870 return 0;
2871}
2872
2873/*
2874 * Check whether it make sense to run itable init. thread or not.
2875 * If there is at least one uninitialized inode table, return
2876 * corresponding group number, else the loop goes through all
2877 * groups and return total number of groups.
2878 */
2879static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2880{
2881 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2882 struct ext4_group_desc *gdp = NULL;
2883
2884 for (group = 0; group < ngroups; group++) {
2885 gdp = ext4_get_group_desc(sb, group, NULL);
2886 if (!gdp)
2887 continue;
2888
2889 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2890 break;
2891 }
2892
2893 return group;
2894}
2895
2896static int ext4_li_info_new(void)
2897{
2898 struct ext4_lazy_init *eli = NULL;
2899
2900 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2901 if (!eli)
2902 return -ENOMEM;
2903
bfff6873
LC		 2904 INIT_LIST_HEAD(&eli->li_request_list);
2905 mutex_init(&eli->li_list_mtx);
2906
bfff6873
LC		 2907 eli->li_state |= EXT4_LAZYINIT_QUIT;
2908
2909 ext4_li_info = eli;
2910
2911 return 0;
2912}
2913
2914static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2915 ext4_group_t start)
2916{
2917 struct ext4_sb_info *sbi = EXT4_SB(sb);
2918 struct ext4_li_request *elr;
2919 unsigned long rnd;
2920
2921 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2922 if (!elr)
2923 return NULL;
2924
2925 elr->lr_super = sb;
2926 elr->lr_sbi = sbi;
2927 elr->lr_next_group = start;
2928
2929 /*
2930 * Randomize first schedule time of the request to
2931 * spread the inode table initialization requests
2932 * better.
2933 */
2934 get_random_bytes(&rnd, sizeof(rnd));
2935 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2936 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2937
2938 return elr;
2939}
2940
7f511862
TT		 2941int ext4_register_li_request(struct super_block *sb,
2942 ext4_group_t first_not_zeroed)
bfff6873
LC		 2943{
2944 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 2945 struct ext4_li_request *elr = NULL;
bfff6873 2946 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2947 int ret = 0;
bfff6873 2948
7f511862 2949 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 2950 if (sbi->s_li_request != NULL) {
2951 /*
2952 * Reset timeout so it can be computed again, because
2953 * s_li_wait_mult might have changed.
2954 */
2955 sbi->s_li_request->lr_timeout = 0;
7f511862 2956 goto out;
51ce6511 2957 }
bfff6873
LC		 2958
2959 if (first_not_zeroed == ngroups ||
2960 (sb->s_flags & MS_RDONLY) ||
55ff3840 2961 !test_opt(sb, INIT_INODE_TABLE))
7f511862 2962 goto out;
bfff6873
LC		 2963
2964 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 2965 if (!elr) {
2966 ret = -ENOMEM;
2967 goto out;
2968 }
bfff6873
LC		 2969
2970 if (NULL == ext4_li_info) {
2971 ret = ext4_li_info_new();
2972 if (ret)
2973 goto out;
2974 }
2975
2976 mutex_lock(&ext4_li_info->li_list_mtx);
2977 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2978 mutex_unlock(&ext4_li_info->li_list_mtx);
2979
2980 sbi->s_li_request = elr;
46e4690b
TM		 2981 /*
2982 * set elr to NULL here since it has been inserted to
2983 * the request_list and the removal and free of it is
2984 * handled by ext4_clear_request_list from now on.
2985 */
2986 elr = NULL;
bfff6873
LC		 2987
2988 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2989 ret = ext4_run_lazyinit_thread();
2990 if (ret)
2991 goto out;
2992 }
bfff6873 2993out:
beed5ecb
NK		 2994 mutex_unlock(&ext4_li_mtx);
2995 if (ret)
bfff6873 2996 kfree(elr);
bfff6873
LC		 2997 return ret;
2998}
2999
3000/*
3001 * We do not need to lock anything since this is called on
3002 * module unload.
3003 */
3004static void ext4_destroy_lazyinit_thread(void)
3005{
3006 /*
3007 * If thread exited earlier
3008 * there's nothing to be done.
3009 */
8f1f7453 3010 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3011 return;
3012
8f1f7453 3013 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3014}
3015
25ed6e8a
DW		 3016static int set_journal_csum_feature_set(struct super_block *sb)
3017{
3018 int ret = 1;
3019 int compat, incompat;
3020 struct ext4_sb_info *sbi = EXT4_SB(sb);
3021
3022 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3023 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3024 /* journal checksum v2 */
3025 compat = 0;
3026 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3027 } else {
3028 /* journal checksum v1 */
3029 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3030 incompat = 0;
3031 }
3032
3033 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3034 ret = jbd2_journal_set_features(sbi->s_journal,
3035 compat, 0,
3036 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3037 incompat);
3038 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3039 ret = jbd2_journal_set_features(sbi->s_journal,
3040 compat, 0,
3041 incompat);
3042 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3043 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3044 } else {
3045 jbd2_journal_clear_features(sbi->s_journal,
3046 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3047 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3048 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3049 }
3050
3051 return ret;
3052}
3053
952fc18e
TT		 3054/*
3055 * Note: calculating the overhead so we can be compatible with
3056 * historical BSD practice is quite difficult in the face of
3057 * clusters/bigalloc. This is because multiple metadata blocks from
3058 * different block group can end up in the same allocation cluster.
3059 * Calculating the exact overhead in the face of clustered allocation
3060 * requires either O(all block bitmaps) in memory or O(number of block
3061 * groups**2) in time. We will still calculate the superblock for
3062 * older file systems --- and if we come across with a bigalloc file
3063 * system with zero in s_overhead_clusters the estimate will be close to
3064 * correct especially for very large cluster sizes --- but for newer
3065 * file systems, it's better to calculate this figure once at mkfs
3066 * time, and store it in the superblock. If the superblock value is
3067 * present (even for non-bigalloc file systems), we will use it.
3068 */
3069static int count_overhead(struct super_block *sb, ext4_group_t grp,
3070 char *buf)
3071{
3072 struct ext4_sb_info *sbi = EXT4_SB(sb);
3073 struct ext4_group_desc *gdp;
3074 ext4_fsblk_t first_block, last_block, b;
3075 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3076 int s, j, count = 0;
3077
0548bbb8
TT		 3078 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3079 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3080 sbi->s_itb_per_group + 2);
3081
952fc18e
TT		 3082 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3083 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3084 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3085 for (i = 0; i < ngroups; i++) {
3086 gdp = ext4_get_group_desc(sb, i, NULL);
3087 b = ext4_block_bitmap(sb, gdp);
3088 if (b >= first_block && b <= last_block) {
3089 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3090 count++;
3091 }
3092 b = ext4_inode_bitmap(sb, gdp);
3093 if (b >= first_block && b <= last_block) {
3094 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3095 count++;
3096 }
3097 b = ext4_inode_table(sb, gdp);
3098 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3099 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3100 int c = EXT4_B2C(sbi, b - first_block);
3101 ext4_set_bit(c, buf);
3102 count++;
3103 }
3104 if (i != grp)
3105 continue;
3106 s = 0;
3107 if (ext4_bg_has_super(sb, grp)) {
3108 ext4_set_bit(s++, buf);
3109 count++;
3110 }
3111 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3112 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3113 count++;
3114 }
3115 }
3116 if (!count)
3117 return 0;
3118 return EXT4_CLUSTERS_PER_GROUP(sb) -
3119 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3120}
3121
3122/*
3123 * Compute the overhead and stash it in sbi->s_overhead
3124 */
3125int ext4_calculate_overhead(struct super_block *sb)
3126{
3127 struct ext4_sb_info *sbi = EXT4_SB(sb);
3128 struct ext4_super_block *es = sbi->s_es;
3129 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3130 ext4_fsblk_t overhead = 0;
3131 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3132
952fc18e
TT		 3133 if (!buf)
3134 return -ENOMEM;
3135
3136 /*
3137 * Compute the overhead (FS structures). This is constant
3138 * for a given filesystem unless the number of block groups
3139 * changes so we cache the previous value until it does.
3140 */
3141
3142 /*
3143 * All of the blocks before first_data_block are overhead
3144 */
3145 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3146
3147 /*
3148 * Add the overhead found in each block group
3149 */
3150 for (i = 0; i < ngroups; i++) {
3151 int blks;
3152
3153 blks = count_overhead(sb, i, buf);
3154 overhead += blks;
3155 if (blks)
3156 memset(buf, 0, PAGE_SIZE);
3157 cond_resched();
3158 }
0875a2b4
ES		 3159 /* Add the journal blocks as well */
3160 if (sbi->s_journal)
3161 overhead += EXT4_B2C(sbi, sbi->s_journal->j_maxlen);
3162
952fc18e
TT		 3163 sbi->s_overhead = overhead;
3164 smp_wmb();
3165 free_page((unsigned long) buf);
3166 return 0;
3167}
3168
2b2d6d01 3169static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3170{
d4c402d9 3171 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3172 struct buffer_head *bh;
617ba13b
MC		 3173 struct ext4_super_block *es = NULL;
3174 struct ext4_sb_info *sbi;
3175 ext4_fsblk_t block;
3176 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3177 ext4_fsblk_t logical_sb_block;
ac27a0ec 3178 unsigned long offset = 0;
ac27a0ec
DK		 3179 unsigned long journal_devnum = 0;
3180 unsigned long def_mount_opts;
3181 struct inode *root;
9f6200bb 3182 char *cp;
0390131b 3183 const char *descr;
dcc7dae3 3184 int ret = -ENOMEM;
281b5995 3185 int blocksize, clustersize;
4ec11028
TT		 3186 unsigned int db_count;
3187 unsigned int i;
281b5995 3188 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3189 __u64 blocks_count;
07aa2ea1 3190 int err = 0;
b3881f74 3191 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3192 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3193
3194 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3195 if (!sbi)
dcc7dae3 3196 goto out_free_orig;
705895b6
PE		 3197
3198 sbi->s_blockgroup_lock =
3199 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3200 if (!sbi->s_blockgroup_lock) {
3201 kfree(sbi);
dcc7dae3 3202 goto out_free_orig;
705895b6 3203 }
ac27a0ec 3204 sb->s_fs_info = sbi;
2c0544b2 3205 sbi->s_sb = sb;
240799cd 3206 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3207 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3208 if (sb->s_bdev->bd_part)
3209 sbi->s_sectors_written_start =
3210 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3211
9f6200bb
TT		 3212 /* Cleanup superblock name */
3213 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3214 *cp = '!';
3215
07aa2ea1 3216 /* -EINVAL is default */
dcc7dae3 3217 ret = -EINVAL;
617ba13b 3218 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3219 if (!blocksize) {
b31e1552 3220 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3221 goto out_fail;
3222 }
3223
3224 /*
617ba13b 3225 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3226 * block sizes. We need to calculate the offset from buffer start.
3227 */
617ba13b 3228 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3229 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3230 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3231 } else {
70bbb3e0 3232 logical_sb_block = sb_block;
ac27a0ec
DK		 3233 }
3234
70bbb3e0 3235 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3236 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3237 goto out_fail;
3238 }
3239 /*
3240 * Note: s_es must be initialized as soon as possible because
617ba13b 3241 * some ext4 macro-instructions depend on its value
ac27a0ec 3242 */
2716b802 3243 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3244 sbi->s_es = es;
3245 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3246 if (sb->s_magic != EXT4_SUPER_MAGIC)
3247 goto cantfind_ext4;
afc32f7e 3248 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3249
feb0ab32
DW		 3250 /* Warn if metadata_csum and gdt_csum are both set. */
3251 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3252 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3253 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3254 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3255 "redundant flags; please run fsck.");
3256
d25425f8
DW		 3257 /* Check for a known checksum algorithm */
3258 if (!ext4_verify_csum_type(sb, es)) {
3259 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3260 "unknown checksum algorithm.");
3261 silent = 1;
3262 goto cantfind_ext4;
3263 }
3264
0441984a
DW		 3265 /* Load the checksum driver */
3266 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3267 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3268 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3269 if (IS_ERR(sbi->s_chksum_driver)) {
3270 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3271 ret = PTR_ERR(sbi->s_chksum_driver);
3272 sbi->s_chksum_driver = NULL;
3273 goto failed_mount;
3274 }
3275 }
3276
a9c47317
DW		 3277 /* Check superblock checksum */
3278 if (!ext4_superblock_csum_verify(sb, es)) {
3279 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3280 "invalid superblock checksum. Run e2fsck?");
3281 silent = 1;
3282 goto cantfind_ext4;
3283 }
3284
3285 /* Precompute checksum seed for all metadata */
3286 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3287 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3288 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3289 sizeof(es->s_uuid));
3290
ac27a0ec
DK		 3291 /* Set defaults before we parse the mount options */
3292 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3293 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3294 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3295 set_opt(sb, DEBUG);
87f26807 3296 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3297 set_opt(sb, GRPID);
617ba13b 3298 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3299 set_opt(sb, NO_UID32);
ea663336 3300 /* xattr user namespace & acls are now defaulted on */
ea663336 3301 set_opt(sb, XATTR_USER);
03010a33 3302#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3303 set_opt(sb, POSIX_ACL);
2e7842b8 3304#endif
617ba13b 3305 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3306 set_opt(sb, JOURNAL_DATA);
617ba13b 3307 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3308 set_opt(sb, ORDERED_DATA);
617ba13b 3309 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3310 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3311
3312 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3313 set_opt(sb, ERRORS_PANIC);
bb4f397a 3314 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3315 set_opt(sb, ERRORS_CONT);
bb4f397a 3316 else
fd8c37ec 3317 set_opt(sb, ERRORS_RO);
8b67f04a 3318 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3319 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3320 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3321 set_opt(sb, DISCARD);
ac27a0ec 3322
08cefc7a
EB		 3323 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3324 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3325 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3326 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3327 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3328
8b67f04a 3329 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3330 set_opt(sb, BARRIER);
ac27a0ec 3331
dd919b98
AK		 3332 /*
3333 * enable delayed allocation by default
3334 * Use -o nodelalloc to turn it off
3335 */
bc0b75f7 3336 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3337 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3338 set_opt(sb, DELALLOC);
dd919b98 3339
51ce6511
LC		 3340 /*
3341 * set default s_li_wait_mult for lazyinit, for the case there is
3342 * no mount option specified.
3343 */
3344 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3345
8b67f04a 3346 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3347 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3348 ext4_msg(sb, KERN_WARNING,
3349 "failed to parse options in superblock: %s",
3350 sbi->s_es->s_mount_opts);
3351 }
5a916be1 3352 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3353 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3354 &journal_ioprio, 0))
ac27a0ec
DK		 3355 goto failed_mount;
3356
56889787
TT		 3357 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3358 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3359 "with data=journal disables delayed "
3360 "allocation and O_DIRECT support!\n");
3361 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3362 ext4_msg(sb, KERN_ERR, "can't mount with "
3363 "both data=journal and delalloc");
3364 goto failed_mount;
3365 }
3366 if (test_opt(sb, DIOREAD_NOLOCK)) {
3367 ext4_msg(sb, KERN_ERR, "can't mount with "
3368 "both data=journal and delalloc");
3369 goto failed_mount;
3370 }
3371 if (test_opt(sb, DELALLOC))
3372 clear_opt(sb, DELALLOC);
3373 }
3374
ac27a0ec 3375 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3376 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3377
617ba13b
MC		 3378 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3379 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3380 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3381 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3382 ext4_msg(sb, KERN_WARNING,
3383 "feature flags set on rev 0 fs, "
3384 "running e2fsck is recommended");
469108ff 3385
2035e776
TT		 3386 if (IS_EXT2_SB(sb)) {
3387 if (ext2_feature_set_ok(sb))
3388 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3389 "using the ext4 subsystem");
3390 else {
3391 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3392 "to feature incompatibilities");
3393 goto failed_mount;
3394 }
3395 }
3396
3397 if (IS_EXT3_SB(sb)) {
3398 if (ext3_feature_set_ok(sb))
3399 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3400 "using the ext4 subsystem");
3401 else {
3402 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3403 "to feature incompatibilities");
3404 goto failed_mount;
3405 }
3406 }
3407
ac27a0ec
DK		 3408 /*
3409 * Check feature flags regardless of the revision level, since we
3410 * previously didn't change the revision level when setting the flags,
3411 * so there is a chance incompat flags are set on a rev 0 filesystem.
3412 */
a13fb1a4 3413 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3414 goto failed_mount;
a13fb1a4 3415
261cb20c 3416 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3417 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3418 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3419 ext4_msg(sb, KERN_ERR,
3420 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3421 goto failed_mount;
3422 }
3423
ac27a0ec 3424 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3425 /* Validate the filesystem blocksize */
3426 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3427 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3428 blocksize);
ac27a0ec
DK		 3429 goto failed_mount;
3430 }
3431
2b2d6d01 3432 brelse(bh);
70bbb3e0
AM		 3433 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3434 offset = do_div(logical_sb_block, blocksize);
3435 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3436 if (!bh) {
b31e1552
ES		 3437 ext4_msg(sb, KERN_ERR,
3438 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3439 goto failed_mount;
3440 }
2716b802 3441 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3442 sbi->s_es = es;
617ba13b 3443 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3444 ext4_msg(sb, KERN_ERR,
3445 "Magic mismatch, very weird!");
ac27a0ec
DK		 3446 goto failed_mount;
3447 }
3448 }
3449
a13fb1a4
ES		 3450 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3451 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3452 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3453 has_huge_files);
3454 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3455
617ba13b
MC		 3456 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3457 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3458 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3459 } else {
3460 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3461 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3462 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3463 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3464 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3465 ext4_msg(sb, KERN_ERR,
3466 "unsupported inode size: %d",
2b2d6d01 3467 sbi->s_inode_size);
ac27a0ec
DK		 3468 goto failed_mount;
3469 }
ef7f3835
KS		 3470 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3471 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3472 }
0b8e58a1 3473
0d1ee42f
AR		 3474 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3475 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3476 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3477 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3478 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3479 ext4_msg(sb, KERN_ERR,
3480 "unsupported descriptor size %lu",
0d1ee42f
AR		 3481 sbi->s_desc_size);
3482 goto failed_mount;
3483 }
3484 } else
3485 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3486
ac27a0ec 3487 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3488 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3489 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3490 goto cantfind_ext4;
0b8e58a1 3491
617ba13b 3492 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3493 if (sbi->s_inodes_per_block == 0)
617ba13b 3494 goto cantfind_ext4;
ac27a0ec
DK		 3495 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3496 sbi->s_inodes_per_block;
0d1ee42f 3497 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3498 sbi->s_sbh = bh;
3499 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3500 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3501 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3502
2b2d6d01 3503 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3504 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3505 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3506 i = le32_to_cpu(es->s_flags);
3507 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3508 sbi->s_hash_unsigned = 3;
3509 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3510#ifdef __CHAR_UNSIGNED__
3511 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3512 sbi->s_hash_unsigned = 3;
3513#else
3514 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3515#endif
f99b2589 3516 }
ac27a0ec 3517
281b5995
TT		 3518 /* Handle clustersize */
3519 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3520 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3521 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3522 if (has_bigalloc) {
3523 if (clustersize < blocksize) {
3524 ext4_msg(sb, KERN_ERR,
3525 "cluster size (%d) smaller than "
3526 "block size (%d)", clustersize, blocksize);
3527 goto failed_mount;
3528 }
3529 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3530 le32_to_cpu(es->s_log_block_size);
3531 sbi->s_clusters_per_group =
3532 le32_to_cpu(es->s_clusters_per_group);
3533 if (sbi->s_clusters_per_group > blocksize * 8) {
3534 ext4_msg(sb, KERN_ERR,
3535 "#clusters per group too big: %lu",
3536 sbi->s_clusters_per_group);
3537 goto failed_mount;
3538 }
3539 if (sbi->s_blocks_per_group !=
3540 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3541 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3542 "clusters per group (%lu) inconsistent",
3543 sbi->s_blocks_per_group,
3544 sbi->s_clusters_per_group);
3545 goto failed_mount;
3546 }
3547 } else {
3548 if (clustersize != blocksize) {
3549 ext4_warning(sb, "fragment/cluster size (%d) != "
3550 "block size (%d)", clustersize,
3551 blocksize);
3552 clustersize = blocksize;
3553 }
3554 if (sbi->s_blocks_per_group > blocksize * 8) {
3555 ext4_msg(sb, KERN_ERR,
3556 "#blocks per group too big: %lu",
3557 sbi->s_blocks_per_group);
3558 goto failed_mount;
3559 }
3560 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3561 sbi->s_cluster_bits = 0;
ac27a0ec 3562 }
281b5995
TT		 3563 sbi->s_cluster_ratio = clustersize / blocksize;
3564
ac27a0ec 3565 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3566 ext4_msg(sb, KERN_ERR,
3567 "#inodes per group too big: %lu",
2b2d6d01 3568 sbi->s_inodes_per_group);
ac27a0ec
DK		 3569 goto failed_mount;
3570 }
3571
bf43d84b
ES		 3572 /*
3573 * Test whether we have more sectors than will fit in sector_t,
3574 * and whether the max offset is addressable by the page cache.
3575 */
5a9ae68a 3576 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3577 ext4_blocks_count(es));
5a9ae68a 3578 if (err) {
b31e1552 3579 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3580 " too large to mount safely on this system");
ac27a0ec 3581 if (sizeof(sector_t) < 8)
90c699a9 3582 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3583 goto failed_mount;
3584 }
3585
617ba13b
MC		 3586 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3587 goto cantfind_ext4;
e7c95593 3588
0f2ddca6
FTN		 3589 /* check blocks count against device size */
3590 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3591 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3592 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3593 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3594 ext4_blocks_count(es), blocks_count);
3595 goto failed_mount;
3596 }
3597
0b8e58a1
AD		 3598 /*
3599 * It makes no sense for the first data block to be beyond the end
3600 * of the filesystem.
3601 */
3602 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3603 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3604 "block %u is beyond end of filesystem (%llu)",
3605 le32_to_cpu(es->s_first_data_block),
3606 ext4_blocks_count(es));
e7c95593
ES		 3607 goto failed_mount;
3608 }
bd81d8ee
LV		 3609 blocks_count = (ext4_blocks_count(es) -
3610 le32_to_cpu(es->s_first_data_block) +
3611 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3612 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3613 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3614 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3615 "(block count %llu, first data block %u, "
b31e1552 3616 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3617 ext4_blocks_count(es),
3618 le32_to_cpu(es->s_first_data_block),
3619 EXT4_BLOCKS_PER_GROUP(sb));
3620 goto failed_mount;
3621 }
bd81d8ee 3622 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3623 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3624 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3625 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3626 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3627 sbi->s_group_desc = ext4_kvmalloc(db_count *
3628 sizeof(struct buffer_head *),
3629 GFP_KERNEL);
ac27a0ec 3630 if (sbi->s_group_desc == NULL) {
b31e1552 3631 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3632 ret = -ENOMEM;
ac27a0ec
DK		 3633 goto failed_mount;
3634 }
3635
9f6200bb
TT		 3636 if (ext4_proc_root)
3637 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3638
66acdcf4
TT		 3639 if (sbi->s_proc)
3640 proc_create_data("options", S_IRUGO, sbi->s_proc,
3641 &ext4_seq_options_fops, sb);
3642
705895b6 3643 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3644
3645 for (i = 0; i < db_count; i++) {
70bbb3e0 3646 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3647 sbi->s_group_desc[i] = sb_bread(sb, block);
3648 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3649 ext4_msg(sb, KERN_ERR,
3650 "can't read group descriptor %d", i);
ac27a0ec
DK		 3651 db_count = i;
3652 goto failed_mount2;
3653 }
3654 }
bfff6873 3655 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3656 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3657 goto failed_mount2;
3658 }
772cb7c8
JS		 3659 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3660 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3661 ext4_msg(sb, KERN_ERR,
3662 "unable to initialize "
3663 "flex_bg meta info!");
772cb7c8
JS		 3664 goto failed_mount2;
3665 }
3666
ac27a0ec
DK		 3667 sbi->s_gdb_count = db_count;
3668 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3669 spin_lock_init(&sbi->s_next_gen_lock);
3670
04496411
TM		 3671 init_timer(&sbi->s_err_report);
3672 sbi->s_err_report.function = print_daily_error_info;
3673 sbi->s_err_report.data = (unsigned long) sb;
3674
57042651 3675 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3676 ext4_count_free_clusters(sb));
ce7e010a
TT		 3677 if (!err) {
3678 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3679 ext4_count_free_inodes(sb));
3680 }
3681 if (!err) {
3682 err = percpu_counter_init(&sbi->s_dirs_counter,
3683 ext4_count_dirs(sb));
3684 }
3685 if (!err) {
57042651 3686 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3687 }
3688 if (err) {
3689 ext4_msg(sb, KERN_ERR, "insufficient memory");
3690 goto failed_mount3;
3691 }
3692
c9de560d 3693 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3694 sbi->s_max_writeback_mb_bump = 128;
67a5da56 3695 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3696
ac27a0ec
DK		 3697 /*
3698 * set up enough so that it can read an inode
3699 */
9ca92389
TT		 3700 if (!test_opt(sb, NOLOAD) &&
3701 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3702 sb->s_op = &ext4_sops;
3703 else
3704 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3705 sb->s_export_op = &ext4_export_ops;
3706 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3707#ifdef CONFIG_QUOTA
617ba13b
MC		 3708 sb->s_qcop = &ext4_qctl_operations;
3709 sb->dq_op = &ext4_quota_operations;
7c319d32
AK		 3710
3711 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3712 /* Use qctl operations for hidden quota files. */
3713 sb->s_qcop = &ext4_qctl_sysfile_operations;
3714 }
ac27a0ec 3715#endif
f2fa2ffc
AK		 3716 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3717
ac27a0ec 3718 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3719 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3720
3721 sb->s_root = NULL;
3722
3723 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3724 EXT4_HAS_INCOMPAT_FEATURE(sb,
3725 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3726
c5e06d10
JL		 3727 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3728 !(sb->s_flags & MS_RDONLY))
3729 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3730 goto failed_mount3;
3731
ac27a0ec
DK		 3732 /*
3733 * The first inode we look at is the journal inode. Don't try
3734 * root first: it may be modified in the journal!
3735 */
3736 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3737 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3738 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3739 goto failed_mount3;
0390131b
FM		 3740 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3741 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3742 ext4_msg(sb, KERN_ERR, "required journal recovery "
3743 "suppressed and not mounted read-only");
744692dc 3744 goto failed_mount_wq;
ac27a0ec 3745 } else {
fd8c37ec 3746 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3747 sbi->s_journal = NULL;
3748 needs_recovery = 0;
3749 goto no_journal;
ac27a0ec
DK		 3750 }
3751
f32aaf2d 3752 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3753 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3754 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3755 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3756 goto failed_mount_wq;
eb40a09c
JS		 3757 }
3758
25ed6e8a
DW		 3759 if (!set_journal_csum_feature_set(sb)) {
3760 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3761 "feature set");
3762 goto failed_mount_wq;
d4da6c9c 3763 }
818d276c 3764
ac27a0ec
DK		 3765 /* We have now updated the journal if required, so we can
3766 * validate the data journaling mode. */
3767 switch (test_opt(sb, DATA_FLAGS)) {
3768 case 0:
3769 /* No mode set, assume a default based on the journal
63f57933
AM		 3770 * capabilities: ORDERED_DATA if the journal can
3771 * cope, else JOURNAL_DATA
3772 */
dab291af
MC		 3773 if (jbd2_journal_check_available_features
3774 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3775 set_opt(sb, ORDERED_DATA);
ac27a0ec 3776 else
fd8c37ec 3777 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3778 break;
3779
617ba13b
MC		 3780 case EXT4_MOUNT_ORDERED_DATA:
3781 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3782 if (!jbd2_journal_check_available_features
3783 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3784 ext4_msg(sb, KERN_ERR, "Journal does not support "
3785 "requested data journaling mode");
744692dc 3786 goto failed_mount_wq;
ac27a0ec
DK		 3787 }
3788 default:
3789 break;
3790 }
b3881f74 3791 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3792
18aadd47
BJ		 3793 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3794
ce7e010a
TT		 3795 /*
3796 * The journal may have updated the bg summary counts, so we
3797 * need to update the global counters.
3798 */
57042651 3799 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3800 ext4_count_free_clusters(sb));
ce7e010a
TT		 3801 percpu_counter_set(&sbi->s_freeinodes_counter,
3802 ext4_count_free_inodes(sb));
3803 percpu_counter_set(&sbi->s_dirs_counter,
3804 ext4_count_dirs(sb));
57042651 3805 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3806
ce7e010a 3807no_journal:
952fc18e
TT		 3808 /*
3809 * Get the # of file system overhead blocks from the
3810 * superblock if present.
3811 */
3812 if (es->s_overhead_clusters)
3813 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3814 else {
07aa2ea1
LC		 3815 err = ext4_calculate_overhead(sb);
3816 if (err)
952fc18e
TT		 3817 goto failed_mount_wq;
3818 }
3819
fd89d5f2
TH		 3820 /*
3821 * The maximum number of concurrent works can be high and
3822 * concurrency isn't really necessary. Limit it to 1.
3823 */
3824 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3825 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3826 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3827 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
07aa2ea1 3828 ret = -ENOMEM;
4c0425ff
MC		 3829 goto failed_mount_wq;
3830 }
3831
ac27a0ec 3832 /*
dab291af 3833 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3834 * so we can safely mount the rest of the filesystem now.
3835 */
3836
1d1fe1ee
DH		 3837 root = ext4_iget(sb, EXT4_ROOT_INO);
3838 if (IS_ERR(root)) {
b31e1552 3839 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3840 ret = PTR_ERR(root);
32a9bb57 3841 root = NULL;
ac27a0ec
DK		 3842 goto failed_mount4;
3843 }
3844 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3845 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3846 iput(root);
ac27a0ec
DK		 3847 goto failed_mount4;
3848 }
48fde701 3849 sb->s_root = d_make_root(root);
1d1fe1ee 3850 if (!sb->s_root) {
b31e1552 3851 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3852 ret = -ENOMEM;
3853 goto failed_mount4;
3854 }
ac27a0ec 3855
7e84b621
ES		 3856 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3857 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3858
3859 /* determine the minimum size of new large inodes, if present */
3860 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3861 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3862 EXT4_GOOD_OLD_INODE_SIZE;
3863 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3864 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3865 if (sbi->s_want_extra_isize <
3866 le16_to_cpu(es->s_want_extra_isize))
3867 sbi->s_want_extra_isize =
3868 le16_to_cpu(es->s_want_extra_isize);
3869 if (sbi->s_want_extra_isize <
3870 le16_to_cpu(es->s_min_extra_isize))
3871 sbi->s_want_extra_isize =
3872 le16_to_cpu(es->s_min_extra_isize);
3873 }
3874 }
3875 /* Check if enough inode space is available */
3876 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3877 sbi->s_inode_size) {
3878 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3879 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3880 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3881 "available");
ef7f3835
KS		 3882 }
3883
6fd058f7
TT		 3884 err = ext4_setup_system_zone(sb);
3885 if (err) {
b31e1552 3886 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3887 "zone (%d)", err);
94bf608a 3888 goto failed_mount4a;
6fd058f7
TT		 3889 }
3890
c2774d84 3891 ext4_ext_init(sb);
9d99012f 3892 err = ext4_mb_init(sb);
c2774d84 3893 if (err) {
421f91d2 3894 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3895 err);
dcf2d804 3896 goto failed_mount5;
c2774d84
AK		 3897 }
3898
bfff6873
LC		 3899 err = ext4_register_li_request(sb, first_not_zeroed);
3900 if (err)
dcf2d804 3901 goto failed_mount6;
bfff6873 3902
3197ebdb
TT		 3903 sbi->s_kobj.kset = ext4_kset;
3904 init_completion(&sbi->s_kobj_unregister);
3905 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3906 "%s", sb->s_id);
dcf2d804
TM		 3907 if (err)
3908 goto failed_mount7;
3197ebdb 3909
617ba13b
MC		 3910 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3911 ext4_orphan_cleanup(sb, es);
3912 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3913 if (needs_recovery) {
b31e1552 3914 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3915 ext4_mark_recovery_complete(sb, es);
3916 }
3917 if (EXT4_SB(sb)->s_journal) {
3918 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3919 descr = " journalled data mode";
3920 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3921 descr = " ordered data mode";
3922 else
3923 descr = " writeback data mode";
3924 } else
3925 descr = "out journal";
3926
7c319d32
AK		 3927#ifdef CONFIG_QUOTA
3928 /* Enable quota usage during mount. */
3929 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3930 !(sb->s_flags & MS_RDONLY)) {
07aa2ea1
LC		 3931 err = ext4_enable_quotas(sb);
3932 if (err)
72ba7450 3933 goto failed_mount8;
7c319d32
AK		 3934 }
3935#endif /* CONFIG_QUOTA */
3936
79add3a3
LC		 3937 if (test_opt(sb, DISCARD)) {
3938 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3939 if (!blk_queue_discard(q))
3940 ext4_msg(sb, KERN_WARNING,
3941 "mounting with \"discard\" option, but "
3942 "the device does not support discard");
3943 }
3944
d4c402d9 3945 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3946 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3947 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3948
66e61a9e
TT		 3949 if (es->s_error_count)
3950 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3951
d4c402d9 3952 kfree(orig_data);
ac27a0ec
DK		 3953 return 0;
3954
617ba13b 3955cantfind_ext4:
ac27a0ec 3956 if (!silent)
b31e1552 3957 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3958 goto failed_mount;
3959
72ba7450
TT		 3960#ifdef CONFIG_QUOTA
3961failed_mount8:
3962 kobject_del(&sbi->s_kobj);
3963#endif
dcf2d804
TM		 3964failed_mount7:
3965 ext4_unregister_li_request(sb);
3966failed_mount6:
dcf2d804 3967 ext4_mb_release(sb);
94bf608a
AV		 3968failed_mount5:
3969 ext4_ext_release(sb);
dcf2d804 3970 ext4_release_system_zone(sb);
94bf608a
AV		 3971failed_mount4a:
3972 dput(sb->s_root);
32a9bb57 3973 sb->s_root = NULL;
94bf608a 3974failed_mount4:
b31e1552 3975 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3976 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3977failed_mount_wq:
0390131b
FM		 3978 if (sbi->s_journal) {
3979 jbd2_journal_destroy(sbi->s_journal);
3980 sbi->s_journal = NULL;
3981 }
ac27a0ec 3982failed_mount3:
04496411 3983 del_timer(&sbi->s_err_report);
9933fc0a
TT		 3984 if (sbi->s_flex_groups)
3985 ext4_kvfree(sbi->s_flex_groups);
57042651 3986 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 3987 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3988 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 3989 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 3990 if (sbi->s_mmp_tsk)
3991 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3992failed_mount2:
3993 for (i = 0; i < db_count; i++)
3994 brelse(sbi->s_group_desc[i]);
f18a5f21 3995 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 3996failed_mount:
0441984a
DW		 3997 if (sbi->s_chksum_driver)
3998 crypto_free_shash(sbi->s_chksum_driver);
240799cd 3999 if (sbi->s_proc) {
66acdcf4 4000 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4001 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4002 }
ac27a0ec
DK		 4003#ifdef CONFIG_QUOTA
4004 for (i = 0; i < MAXQUOTAS; i++)
4005 kfree(sbi->s_qf_names[i]);
4006#endif
617ba13b 4007 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4008 brelse(bh);
4009out_fail:
4010 sb->s_fs_info = NULL;
f6830165 4011 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4012 kfree(sbi);
dcc7dae3 4013out_free_orig:
d4c402d9 4014 kfree(orig_data);
07aa2ea1 4015 return err ? err : ret;
ac27a0ec
DK		 4016}
4017
4018/*
4019 * Setup any per-fs journal parameters now. We'll do this both on
4020 * initial mount, once the journal has been initialised but before we've
4021 * done any recovery; and again on any subsequent remount.
4022 */
617ba13b 4023static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4024{
617ba13b 4025 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4026
30773840
TT		 4027 journal->j_commit_interval = sbi->s_commit_interval;
4028 journal->j_min_batch_time = sbi->s_min_batch_time;
4029 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4030
a931da6a 4031 write_lock(&journal->j_state_lock);
ac27a0ec 4032 if (test_opt(sb, BARRIER))
dab291af 4033 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4034 else
dab291af 4035 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4036 if (test_opt(sb, DATA_ERR_ABORT))
4037 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4038 else
4039 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4040 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4041}
4042
617ba13b 4043static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4044 unsigned int journal_inum)
4045{
4046 struct inode *journal_inode;
4047 journal_t *journal;
4048
0390131b
FM		 4049 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4050
ac27a0ec
DK		 4051 /* First, test for the existence of a valid inode on disk. Bad
4052 * things happen if we iget() an unused inode, as the subsequent
4053 * iput() will try to delete it. */
4054
1d1fe1ee
DH		 4055 journal_inode = ext4_iget(sb, journal_inum);
4056 if (IS_ERR(journal_inode)) {
b31e1552 4057 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4058 return NULL;
4059 }
4060 if (!journal_inode->i_nlink) {
4061 make_bad_inode(journal_inode);
4062 iput(journal_inode);
b31e1552 4063 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4064 return NULL;
4065 }
4066
e5f8eab8 4067 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4068 journal_inode, journal_inode->i_size);
1d1fe1ee 4069 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4070 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4071 iput(journal_inode);
4072 return NULL;
4073 }
4074
dab291af 4075 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4076 if (!journal) {
b31e1552 4077 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4078 iput(journal_inode);
4079 return NULL;
4080 }
4081 journal->j_private = sb;
617ba13b 4082 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4083 return journal;
4084}
4085
617ba13b 4086static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4087 dev_t j_dev)
4088{
2b2d6d01 4089 struct buffer_head *bh;
ac27a0ec 4090 journal_t *journal;
617ba13b
MC		 4091 ext4_fsblk_t start;
4092 ext4_fsblk_t len;
ac27a0ec 4093 int hblock, blocksize;
617ba13b 4094 ext4_fsblk_t sb_block;
ac27a0ec 4095 unsigned long offset;
2b2d6d01 4096 struct ext4_super_block *es;
ac27a0ec
DK		 4097 struct block_device *bdev;
4098
0390131b
FM		 4099 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4100
b31e1552 4101 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4102 if (bdev == NULL)
4103 return NULL;
4104
ac27a0ec 4105 blocksize = sb->s_blocksize;
e1defc4f 4106 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4107 if (blocksize < hblock) {
b31e1552
ES		 4108 ext4_msg(sb, KERN_ERR,
4109 "blocksize too small for journal device");
ac27a0ec
DK		 4110 goto out_bdev;
4111 }
4112
617ba13b
MC		 4113 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4114 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4115 set_blocksize(bdev, blocksize);
4116 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4117 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4118 "external journal");
ac27a0ec
DK		 4119 goto out_bdev;
4120 }
4121
2716b802 4122 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4123 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4124 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4125 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4126 ext4_msg(sb, KERN_ERR, "external journal has "
4127 "bad superblock");
ac27a0ec
DK		 4128 brelse(bh);
4129 goto out_bdev;
4130 }
4131
617ba13b 4132 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4133 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4134 brelse(bh);
4135 goto out_bdev;
4136 }
4137
bd81d8ee 4138 len = ext4_blocks_count(es);
ac27a0ec
DK		 4139 start = sb_block + 1;
4140 brelse(bh); /* we're done with the superblock */
4141
dab291af 4142 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4143 start, len, blocksize);
4144 if (!journal) {
b31e1552 4145 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4146 goto out_bdev;
4147 }
4148 journal->j_private = sb;
4149 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4150 wait_on_buffer(journal->j_sb_buffer);
4151 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4152 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4153 goto out_journal;
4154 }
4155 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4156 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4157 "user (unsupported) - %d",
ac27a0ec
DK		 4158 be32_to_cpu(journal->j_superblock->s_nr_users));
4159 goto out_journal;
4160 }
617ba13b
MC		 4161 EXT4_SB(sb)->journal_bdev = bdev;
4162 ext4_init_journal_params(sb, journal);
ac27a0ec 4163 return journal;
0b8e58a1 4164
ac27a0ec 4165out_journal:
dab291af 4166 jbd2_journal_destroy(journal);
ac27a0ec 4167out_bdev:
617ba13b 4168 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4169 return NULL;
4170}
4171
617ba13b
MC		 4172static int ext4_load_journal(struct super_block *sb,
4173 struct ext4_super_block *es,
ac27a0ec
DK		 4174 unsigned long journal_devnum)
4175{
4176 journal_t *journal;
4177 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4178 dev_t journal_dev;
4179 int err = 0;
4180 int really_read_only;
4181
0390131b
FM		 4182 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4183
ac27a0ec
DK		 4184 if (journal_devnum &&
4185 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4186 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4187 "numbers have changed");
ac27a0ec
DK		 4188 journal_dev = new_decode_dev(journal_devnum);
4189 } else
4190 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4191
4192 really_read_only = bdev_read_only(sb->s_bdev);
4193
4194 /*
4195 * Are we loading a blank journal or performing recovery after a
4196 * crash? For recovery, we need to check in advance whether we
4197 * can get read-write access to the device.
4198 */
617ba13b 4199 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4200 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4201 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4202 "required on readonly filesystem");
ac27a0ec 4203 if (really_read_only) {
b31e1552
ES		 4204 ext4_msg(sb, KERN_ERR, "write access "
4205 "unavailable, cannot proceed");
ac27a0ec
DK		 4206 return -EROFS;
4207 }
b31e1552
ES		 4208 ext4_msg(sb, KERN_INFO, "write access will "
4209 "be enabled during recovery");
ac27a0ec
DK		 4210 }
4211 }
4212
4213 if (journal_inum && journal_dev) {
b31e1552
ES		 4214 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4215 "and inode journals!");
ac27a0ec
DK		 4216 return -EINVAL;
4217 }
4218
4219 if (journal_inum) {
617ba13b 4220 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4221 return -EINVAL;
4222 } else {
617ba13b 4223 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4224 return -EINVAL;
4225 }
4226
90576c0b 4227 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4228 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4229
617ba13b 4230 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4231 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4232 if (!err) {
4233 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4234 if (save)
4235 memcpy(save, ((char *) es) +
4236 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4237 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4238 if (save)
4239 memcpy(((char *) es) + EXT4_S_ERR_START,
4240 save, EXT4_S_ERR_LEN);
4241 kfree(save);
4242 }
ac27a0ec
DK		 4243
4244 if (err) {
b31e1552 4245 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4246 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4247 return err;
4248 }
4249
617ba13b
MC		 4250 EXT4_SB(sb)->s_journal = journal;
4251 ext4_clear_journal_err(sb, es);
ac27a0ec 4252
c41303ce 4253 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4254 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4255 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4256
4257 /* Make sure we flush the recovery flag to disk. */
e2d67052 4258 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4259 }
4260
4261 return 0;
4262}
4263
e2d67052 4264static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4265{
e2d67052 4266 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4267 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4268 int error = 0;
ac27a0ec 4269
7c2e7087 4270 if (!sbh || block_device_ejected(sb))
c4be0c1d 4271 return error;
914258bf
TT		 4272 if (buffer_write_io_error(sbh)) {
4273 /*
4274 * Oh, dear. A previous attempt to write the
4275 * superblock failed. This could happen because the
4276 * USB device was yanked out. Or it could happen to
4277 * be a transient write error and maybe the block will
4278 * be remapped. Nothing we can do but to retry the
4279 * write and hope for the best.
4280 */
b31e1552
ES		 4281 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4282 "superblock detected");
914258bf
TT		 4283 clear_buffer_write_io_error(sbh);
4284 set_buffer_uptodate(sbh);
4285 }
71290b36
TT		 4286 /*
4287 * If the file system is mounted read-only, don't update the
4288 * superblock write time. This avoids updating the superblock
4289 * write time when we are mounting the root file system
4290 * read/only but we need to replay the journal; at that point,
4291 * for people who are east of GMT and who make their clock
4292 * tick in localtime for Windows bug-for-bug compatibility,
4293 * the clock is set in the future, and this will cause e2fsck
4294 * to complain and force a full file system check.
4295 */
4296 if (!(sb->s_flags & MS_RDONLY))
4297 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4298 if (sb->s_bdev->bd_part)
4299 es->s_kbytes_written =
4300 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4301 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4302 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4303 else
4304 es->s_kbytes_written =
4305 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4306 ext4_free_blocks_count_set(es,
4307 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4308 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4309 es->s_free_inodes_count =
4310 cpu_to_le32(percpu_counter_sum_positive(
4311 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4312 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4313 ext4_superblock_csum_set(sb);
ac27a0ec 4314 mark_buffer_dirty(sbh);
914258bf 4315 if (sync) {
c4be0c1d
TS		 4316 error = sync_dirty_buffer(sbh);
4317 if (error)
4318 return error;
4319
4320 error = buffer_write_io_error(sbh);
4321 if (error) {
b31e1552
ES		 4322 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4323 "superblock");
914258bf
TT		 4324 clear_buffer_write_io_error(sbh);
4325 set_buffer_uptodate(sbh);
4326 }
4327 }
c4be0c1d 4328 return error;
ac27a0ec
DK		 4329}
4330
ac27a0ec
DK		 4331/*
4332 * Have we just finished recovery? If so, and if we are mounting (or
4333 * remounting) the filesystem readonly, then we will end up with a
4334 * consistent fs on disk. Record that fact.
4335 */
2b2d6d01
TT		 4336static void ext4_mark_recovery_complete(struct super_block *sb,
4337 struct ext4_super_block *es)
ac27a0ec 4338{
617ba13b 4339 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4340
0390131b
FM		 4341 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4342 BUG_ON(journal != NULL);
4343 return;
4344 }
dab291af 4345 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4346 if (jbd2_journal_flush(journal) < 0)
4347 goto out;
4348
617ba13b 4349 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4350 sb->s_flags & MS_RDONLY) {
617ba13b 4351 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4352 ext4_commit_super(sb, 1);
ac27a0ec 4353 }
7ffe1ea8
HK		 4354
4355out:
dab291af 4356 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4357}
4358
4359/*
4360 * If we are mounting (or read-write remounting) a filesystem whose journal
4361 * has recorded an error from a previous lifetime, move that error to the
4362 * main filesystem now.
4363 */
2b2d6d01
TT		 4364static void ext4_clear_journal_err(struct super_block *sb,
4365 struct ext4_super_block *es)
ac27a0ec
DK		 4366{
4367 journal_t *journal;
4368 int j_errno;
4369 const char *errstr;
4370
0390131b
FM		 4371 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4372
617ba13b 4373 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4374
4375 /*
4376 * Now check for any error status which may have been recorded in the
617ba13b 4377 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4378 */
4379
dab291af 4380 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4381 if (j_errno) {
4382 char nbuf[16];
4383
617ba13b 4384 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4385 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4386 "from previous mount: %s", errstr);
12062ddd 4387 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4388
617ba13b
MC		 4389 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4390 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4391 ext4_commit_super(sb, 1);
ac27a0ec 4392
dab291af 4393 jbd2_journal_clear_err(journal);
d796c52e 4394 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4395 }
4396}
4397
4398/*
4399 * Force the running and committing transactions to commit,
4400 * and wait on the commit.
4401 */
617ba13b 4402int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4403{
4404 journal_t *journal;
ac27a0ec
DK		 4405
4406 if (sb->s_flags & MS_RDONLY)
4407 return 0;
4408
617ba13b 4409 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4410 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4411}
4412
617ba13b 4413static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4414{
14ce0cb4 4415 int ret = 0;
9eddacf9 4416 tid_t target;
8d5d02e6 4417 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4418
9bffad1e 4419 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4420 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4421 /*
4422 * Writeback quota in non-journalled quota case - journalled quota has
4423 * no dirty dquots
4424 */
4425 dquot_writeback_dquots(sb, -1);
8d5d02e6 4426 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4427 if (wait)
8d5d02e6 4428 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4429 }
14ce0cb4 4430 return ret;
ac27a0ec
DK		 4431}
4432
4433/*
4434 * LVM calls this function before a (read-only) snapshot is created. This
4435 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4436 *
4437 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4438 * state independently. It relies on upper layer to stop all data & metadata
4439 * modifications.
ac27a0ec 4440 */
c4be0c1d 4441static int ext4_freeze(struct super_block *sb)
ac27a0ec 4442{
c4be0c1d
TS		 4443 int error = 0;
4444 journal_t *journal;
ac27a0ec 4445
9ca92389
TT		 4446 if (sb->s_flags & MS_RDONLY)
4447 return 0;
ac27a0ec 4448
9ca92389 4449 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4450
9ca92389
TT		 4451 /* Now we set up the journal barrier. */
4452 jbd2_journal_lock_updates(journal);
ac27a0ec 4453
9ca92389
TT		 4454 /*
4455 * Don't clear the needs_recovery flag if we failed to flush
4456 * the journal.
4457 */
4458 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4459 if (error < 0)
4460 goto out;
9ca92389
TT		 4461
4462 /* Journal blocked and flushed, clear needs_recovery flag. */
4463 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4464 error = ext4_commit_super(sb, 1);
6b0310fb 4465out:
8e8ad8a5 4466 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4467 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4468 return error;
ac27a0ec
DK		 4469}
4470
4471/*
4472 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4473 * flag here, even though the filesystem is not technically dirty yet.
4474 */
c4be0c1d 4475static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4476{
9ca92389
TT		 4477 if (sb->s_flags & MS_RDONLY)
4478 return 0;
4479
9ca92389
TT		 4480 /* Reset the needs_recovery flag before the fs is unlocked. */
4481 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4482 ext4_commit_super(sb, 1);
c4be0c1d 4483 return 0;
ac27a0ec
DK		 4484}
4485
673c6100
TT		 4486/*
4487 * Structure to save mount options for ext4_remount's benefit
4488 */
4489struct ext4_mount_options {
4490 unsigned long s_mount_opt;
a2595b8a 4491 unsigned long s_mount_opt2;
08cefc7a
EB		 4492 kuid_t s_resuid;
4493 kgid_t s_resgid;
673c6100
TT		 4494 unsigned long s_commit_interval;
4495 u32 s_min_batch_time, s_max_batch_time;
4496#ifdef CONFIG_QUOTA
4497 int s_jquota_fmt;
4498 char *s_qf_names[MAXQUOTAS];
4499#endif
4500};
4501
2b2d6d01 4502static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4503{
2b2d6d01 4504 struct ext4_super_block *es;
617ba13b 4505 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4506 unsigned long old_sb_flags;
617ba13b 4507 struct ext4_mount_options old_opts;
c79d967d 4508 int enable_quota = 0;
8a266467 4509 ext4_group_t g;
b3881f74 4510 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4511 int err = 0;
ac27a0ec 4512#ifdef CONFIG_QUOTA
03dafb5f 4513 int i, j;
ac27a0ec 4514#endif
d4c402d9 4515 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4516
4517 /* Store the original options */
4518 old_sb_flags = sb->s_flags;
4519 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4520 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4521 old_opts.s_resuid = sbi->s_resuid;
4522 old_opts.s_resgid = sbi->s_resgid;
4523 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4524 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4525 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4526#ifdef CONFIG_QUOTA
4527 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4528 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f
CG		 4529 if (sbi->s_qf_names[i]) {
4530 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4531 GFP_KERNEL);
4532 if (!old_opts.s_qf_names[i]) {
4533 for (j = 0; j < i; j++)
4534 kfree(old_opts.s_qf_names[j]);
4535 return -ENOMEM;
4536 }
4537 } else
4538 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4539#endif
b3881f74
TT		 4540 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4541 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4542
4543 /*
4544 * Allow the "check" option to be passed as a remount option.
4545 */
661aa520 4546 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4547 err = -EINVAL;
4548 goto restore_opts;
4549 }
4550
4ab2f15b 4551 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4552 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4553
4554 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4555 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4556
4557 es = sbi->s_es;
4558
b3881f74 4559 if (sbi->s_journal) {
0390131b 4560 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4561 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4562 }
ac27a0ec 4563
661aa520 4564 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4565 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4566 err = -EROFS;
4567 goto restore_opts;
4568 }
4569
4570 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4571 err = dquot_suspend(sb, -1);
4572 if (err < 0)
c79d967d 4573 goto restore_opts;
c79d967d 4574
ac27a0ec
DK		 4575 /*
4576 * First of all, the unconditional stuff we have to do
4577 * to disable replay of the journal when we next remount
4578 */
4579 sb->s_flags |= MS_RDONLY;
4580
4581 /*
4582 * OK, test if we are remounting a valid rw partition
4583 * readonly, and if so set the rdonly flag and then
4584 * mark the partition as valid again.
4585 */
617ba13b
MC		 4586 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4587 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4588 es->s_state = cpu_to_le16(sbi->s_mount_state);
4589
a63c9eb2 4590 if (sbi->s_journal)
0390131b 4591 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4592 } else {
a13fb1a4
ES		 4593 /* Make sure we can mount this feature set readwrite */
4594 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4595 err = -EROFS;
4596 goto restore_opts;
4597 }
8a266467
TT		 4598 /*
4599 * Make sure the group descriptor checksums
0b8e58a1 4600 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4601 */
4602 for (g = 0; g < sbi->s_groups_count; g++) {
4603 struct ext4_group_desc *gdp =
4604 ext4_get_group_desc(sb, g, NULL);
4605
feb0ab32 4606 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4607 ext4_msg(sb, KERN_ERR,
4608 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4609 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4610 le16_to_cpu(gdp->bg_checksum));
4611 err = -EINVAL;
4612 goto restore_opts;
4613 }
4614 }
4615
ead6596b
ES		 4616 /*
4617 * If we have an unprocessed orphan list hanging
4618 * around from a previously readonly bdev mount,
4619 * require a full umount/remount for now.
4620 */
4621 if (es->s_last_orphan) {
b31e1552 4622 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4623 "remount RDWR because of unprocessed "
4624 "orphan inode list. Please "
b31e1552 4625 "umount/remount instead");
ead6596b
ES		 4626 err = -EINVAL;
4627 goto restore_opts;
4628 }
4629
ac27a0ec
DK		 4630 /*
4631 * Mounting a RDONLY partition read-write, so reread
4632 * and store the current valid flag. (It may have
4633 * been changed by e2fsck since we originally mounted
4634 * the partition.)
4635 */
0390131b
FM		 4636 if (sbi->s_journal)
4637 ext4_clear_journal_err(sb, es);
ac27a0ec 4638 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4639 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4640 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4641 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4642 EXT4_FEATURE_INCOMPAT_MMP))
4643 if (ext4_multi_mount_protect(sb,
4644 le64_to_cpu(es->s_mmp_block))) {
4645 err = -EROFS;
4646 goto restore_opts;
4647 }
c79d967d 4648 enable_quota = 1;
ac27a0ec
DK		 4649 }
4650 }
bfff6873
LC		 4651
4652 /*
4653 * Reinitialize lazy itable initialization thread based on
4654 * current settings
4655 */
4656 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4657 ext4_unregister_li_request(sb);
4658 else {
4659 ext4_group_t first_not_zeroed;
4660 first_not_zeroed = ext4_has_uninit_itable(sb);
4661 ext4_register_li_request(sb, first_not_zeroed);
4662 }
4663
6fd058f7 4664 ext4_setup_system_zone(sb);
d096ad0f 4665 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4666 ext4_commit_super(sb, 1);
0390131b 4667
ac27a0ec
DK		 4668#ifdef CONFIG_QUOTA
4669 /* Release old quota file names */
4670 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f 4671 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4672 if (enable_quota) {
4673 if (sb_any_quota_suspended(sb))
4674 dquot_resume(sb, -1);
4675 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4676 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4677 err = ext4_enable_quotas(sb);
07724f98 4678 if (err)
7c319d32 4679 goto restore_opts;
7c319d32
AK		 4680 }
4681 }
ac27a0ec 4682#endif
d4c402d9
CW		 4683
4684 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4685 kfree(orig_data);
ac27a0ec 4686 return 0;
0b8e58a1 4687
ac27a0ec
DK		 4688restore_opts:
4689 sb->s_flags = old_sb_flags;
4690 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4691 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4692 sbi->s_resuid = old_opts.s_resuid;
4693 sbi->s_resgid = old_opts.s_resgid;
4694 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4695 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4696 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4697#ifdef CONFIG_QUOTA
4698 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4699 for (i = 0; i < MAXQUOTAS; i++) {
03dafb5f 4700 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4701 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4702 }
4703#endif
d4c402d9 4704 kfree(orig_data);
ac27a0ec
DK		 4705 return err;
4706}
4707
2b2d6d01 4708static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4709{
4710 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4711 struct ext4_sb_info *sbi = EXT4_SB(sb);
4712 struct ext4_super_block *es = sbi->s_es;
952fc18e 4713 ext4_fsblk_t overhead = 0;
960cc398 4714 u64 fsid;
d02a9391 4715 s64 bfree;
ac27a0ec 4716
952fc18e
TT		 4717 if (!test_opt(sb, MINIX_DF))
4718 overhead = sbi->s_overhead;
ac27a0ec 4719
617ba13b 4720 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4721 buf->f_bsize = sb->s_blocksize;
b72f78cb 4722 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4723 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4724 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4725 /* prevent underflow in case that few free space is available */
57042651 4726 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4727 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4728 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4729 buf->f_bavail = 0;
4730 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4731 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4732 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4733 fsid = le64_to_cpup((void *)es->s_uuid) ^
4734 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4735 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4736 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4737
ac27a0ec
DK		 4738 return 0;
4739}
4740
0b8e58a1
AD		 4741/* Helper function for writing quotas on sync - we need to start transaction
4742 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4743 * Process 1 Process 2
617ba13b 4744 * ext4_create() quota_sync()
a269eb18 4745 * jbd2_journal_start() write_dquot()
871a2931 4746 * dquot_initialize() down(dqio_mutex)
dab291af 4747 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4748 *
4749 */
4750
4751#ifdef CONFIG_QUOTA
4752
4753static inline struct inode *dquot_to_inode(struct dquot *dquot)
4754{
4c376dca 4755 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4756}
4757
617ba13b 4758static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4759{
4760 int ret, err;
4761 handle_t *handle;
4762 struct inode *inode;
4763
4764 inode = dquot_to_inode(dquot);
617ba13b 4765 handle = ext4_journal_start(inode,
0b8e58a1 4766 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4767 if (IS_ERR(handle))
4768 return PTR_ERR(handle);
4769 ret = dquot_commit(dquot);
617ba13b 4770 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4771 if (!ret)
4772 ret = err;
4773 return ret;
4774}
4775
617ba13b 4776static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4777{
4778 int ret, err;
4779 handle_t *handle;
4780
617ba13b 4781 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4782 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4783 if (IS_ERR(handle))
4784 return PTR_ERR(handle);
4785 ret = dquot_acquire(dquot);
617ba13b 4786 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4787 if (!ret)
4788 ret = err;
4789 return ret;
4790}
4791
617ba13b 4792static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4793{
4794 int ret, err;
4795 handle_t *handle;
4796
617ba13b 4797 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4798 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4799 if (IS_ERR(handle)) {
4800 /* Release dquot anyway to avoid endless cycle in dqput() */
4801 dquot_release(dquot);
ac27a0ec 4802 return PTR_ERR(handle);
9c3013e9 4803 }
ac27a0ec 4804 ret = dquot_release(dquot);
617ba13b 4805 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4806 if (!ret)
4807 ret = err;
4808 return ret;
4809}
4810
617ba13b 4811static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4812{
2c8be6b2 4813 /* Are we journaling quotas? */
617ba13b
MC		 4814 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4815 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4816 dquot_mark_dquot_dirty(dquot);
617ba13b 4817 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4818 } else {
4819 return dquot_mark_dquot_dirty(dquot);
4820 }
4821}
4822
617ba13b 4823static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4824{
4825 int ret, err;
4826 handle_t *handle;
4827
4828 /* Data block + inode block */
617ba13b 4829 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4830 if (IS_ERR(handle))
4831 return PTR_ERR(handle);
4832 ret = dquot_commit_info(sb, type);
617ba13b 4833 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4834 if (!ret)
4835 ret = err;
4836 return ret;
4837}
4838
4839/*
4840 * Turn on quotas during mount time - we need to find
4841 * the quota file and such...
4842 */
617ba13b 4843static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4844{
287a8095
CH		 4845 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4846 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4847}
4848
4849/*
4850 * Standard function to be called on quota_on
4851 */
617ba13b 4852static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4853 struct path *path)
ac27a0ec
DK		 4854{
4855 int err;
ac27a0ec
DK		 4856
4857 if (!test_opt(sb, QUOTA))
4858 return -EINVAL;
0623543b 4859
ac27a0ec 4860 /* Quotafile not on the same filesystem? */
d8c9584e 4861 if (path->dentry->d_sb != sb)
ac27a0ec 4862 return -EXDEV;
0623543b
JK		 4863 /* Journaling quota? */
4864 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4865 /* Quotafile not in fs root? */
f00c9e44 4866 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4867 ext4_msg(sb, KERN_WARNING,
4868 "Quota file not on filesystem root. "
4869 "Journaled quota will not work");
2b2d6d01 4870 }
0623543b
JK		 4871
4872 /*
4873 * When we journal data on quota file, we have to flush journal to see
4874 * all updates to the file when we bypass pagecache...
4875 */
0390131b 4876 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4877 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4878 /*
4879 * We don't need to lock updates but journal_flush() could
4880 * otherwise be livelocked...
4881 */
4882 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4883 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4884 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4885 if (err)
7ffe1ea8 4886 return err;
0623543b
JK		 4887 }
4888
f00c9e44 4889 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4890}
4891
7c319d32
AK		 4892static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4893 unsigned int flags)
4894{
4895 int err;
4896 struct inode *qf_inode;
4897 unsigned long qf_inums[MAXQUOTAS] = {
4898 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4899 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4900 };
4901
4902 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4903
4904 if (!qf_inums[type])
4905 return -EPERM;
4906
4907 qf_inode = ext4_iget(sb, qf_inums[type]);
4908 if (IS_ERR(qf_inode)) {
4909 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4910 return PTR_ERR(qf_inode);
4911 }
4912
4913 err = dquot_enable(qf_inode, type, format_id, flags);
4914 iput(qf_inode);
4915
4916 return err;
4917}
4918
4919/* Enable usage tracking for all quota types. */
4920static int ext4_enable_quotas(struct super_block *sb)
4921{
4922 int type, err = 0;
4923 unsigned long qf_inums[MAXQUOTAS] = {
4924 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4925 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4926 };
4927
4928 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4929 for (type = 0; type < MAXQUOTAS; type++) {
4930 if (qf_inums[type]) {
4931 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4932 DQUOT_USAGE_ENABLED);
4933 if (err) {
4934 ext4_warning(sb,
72ba7450
TT		 4935 "Failed to enable quota tracking "
4936 "(type=%d, err=%d). Please run "
4937 "e2fsck to fix.", type, err);
7c319d32
AK		 4938 return err;
4939 }
4940 }
4941 }
4942 return 0;
4943}
4944
4945/*
4946 * quota_on function that is used when QUOTA feature is set.
4947 */
4948static int ext4_quota_on_sysfile(struct super_block *sb, int type,
4949 int format_id)
4950{
4951 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4952 return -EINVAL;
4953
4954 /*
4955 * USAGE was enabled at mount time. Only need to enable LIMITS now.
4956 */
4957 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
4958}
4959
ca0e05e4
DM		 4960static int ext4_quota_off(struct super_block *sb, int type)
4961{
21f97697
JK		 4962 struct inode *inode = sb_dqopt(sb)->files[type];
4963 handle_t *handle;
4964
87009d86
DM		 4965 /* Force all delayed allocation blocks to be allocated.
4966 * Caller already holds s_umount sem */
4967 if (test_opt(sb, DELALLOC))
ca0e05e4 4968 sync_filesystem(sb);
ca0e05e4 4969
0b268590
AG		 4970 if (!inode)
4971 goto out;
4972
21f97697
JK		 4973 /* Update modification times of quota files when userspace can
4974 * start looking at them */
4975 handle = ext4_journal_start(inode, 1);
4976 if (IS_ERR(handle))
4977 goto out;
4978 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4979 ext4_mark_inode_dirty(handle, inode);
4980 ext4_journal_stop(handle);
4981
4982out:
ca0e05e4
DM		 4983 return dquot_quota_off(sb, type);
4984}
4985
7c319d32
AK		 4986/*
4987 * quota_off function that is used when QUOTA feature is set.
4988 */
4989static int ext4_quota_off_sysfile(struct super_block *sb, int type)
4990{
4991 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4992 return -EINVAL;
4993
4994 /* Disable only the limits. */
4995 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
4996}
4997
ac27a0ec
DK		 4998/* Read data from quotafile - avoid pagecache and such because we cannot afford
4999 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5000 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5001 * we don't have to be afraid of races */
617ba13b 5002static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5003 size_t len, loff_t off)
5004{
5005 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5006 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5007 int err = 0;
5008 int offset = off & (sb->s_blocksize - 1);
5009 int tocopy;
5010 size_t toread;
5011 struct buffer_head *bh;
5012 loff_t i_size = i_size_read(inode);
5013
5014 if (off > i_size)
5015 return 0;
5016 if (off+len > i_size)
5017 len = i_size-off;
5018 toread = len;
5019 while (toread > 0) {
5020 tocopy = sb->s_blocksize - offset < toread ?
5021 sb->s_blocksize - offset : toread;
617ba13b 5022 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5023 if (err)
5024 return err;
5025 if (!bh) /* A hole? */
5026 memset(data, 0, tocopy);
5027 else
5028 memcpy(data, bh->b_data+offset, tocopy);
5029 brelse(bh);
5030 offset = 0;
5031 toread -= tocopy;
5032 data += tocopy;
5033 blk++;
5034 }
5035 return len;
5036}
5037
5038/* Write to quotafile (we know the transaction is already started and has
5039 * enough credits) */
617ba13b 5040static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5041 const char *data, size_t len, loff_t off)
5042{
5043 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5044 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5045 int err = 0;
5046 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5047 struct buffer_head *bh;
5048 handle_t *handle = journal_current_handle();
5049
0390131b 5050 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5051 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5052 " cancelled because transaction is not started",
9c3013e9
JK		 5053 (unsigned long long)off, (unsigned long long)len);
5054 return -EIO;
5055 }
67eeb568
DM		 5056 /*
5057 * Since we account only one data block in transaction credits,
5058 * then it is impossible to cross a block boundary.
5059 */
5060 if (sb->s_blocksize - offset < len) {
5061 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5062 " cancelled because not block aligned",
5063 (unsigned long long)off, (unsigned long long)len);
5064 return -EIO;
5065 }
5066
67eeb568
DM		 5067 bh = ext4_bread(handle, inode, blk, 1, &err);
5068 if (!bh)
5069 goto out;
62d2b5f2
JK		 5070 err = ext4_journal_get_write_access(handle, bh);
5071 if (err) {
5072 brelse(bh);
5073 goto out;
ac27a0ec 5074 }
67eeb568
DM		 5075 lock_buffer(bh);
5076 memcpy(bh->b_data+offset, data, len);
5077 flush_dcache_page(bh->b_page);
5078 unlock_buffer(bh);
62d2b5f2 5079 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5080 brelse(bh);
ac27a0ec 5081out:
0b7f7cef 5082 if (err)
ac27a0ec 5083 return err;
67eeb568
DM		 5084 if (inode->i_size < off + len) {
5085 i_size_write(inode, off + len);
617ba13b 5086 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5087 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5088 }
67eeb568 5089 return len;
ac27a0ec
DK		 5090}
5091
5092#endif
5093
152a0836
AV		 5094static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5095 const char *dev_name, void *data)
ac27a0ec 5096{
152a0836 5097 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5098}
5099
37f328eb 5100#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5101static inline void register_as_ext2(void)
5102{
5103 int err = register_filesystem(&ext2_fs_type);
5104 if (err)
5105 printk(KERN_WARNING
5106 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5107}
5108
5109static inline void unregister_as_ext2(void)
5110{
5111 unregister_filesystem(&ext2_fs_type);
5112}
2035e776
TT		 5113
5114static inline int ext2_feature_set_ok(struct super_block *sb)
5115{
5116 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5117 return 0;
5118 if (sb->s_flags & MS_RDONLY)
5119 return 1;
5120 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5121 return 0;
5122 return 1;
5123}
51b7e3c9 5124MODULE_ALIAS("ext2");
24b58424
TT		 5125#else
5126static inline void register_as_ext2(void) { }
5127static inline void unregister_as_ext2(void) { }
2035e776 5128static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5129#endif
5130
37f328eb 5131#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5132static inline void register_as_ext3(void)
5133{
5134 int err = register_filesystem(&ext3_fs_type);
5135 if (err)
5136 printk(KERN_WARNING
5137 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5138}
5139
5140static inline void unregister_as_ext3(void)
5141{
5142 unregister_filesystem(&ext3_fs_type);
5143}
2035e776
TT		 5144
5145static inline int ext3_feature_set_ok(struct super_block *sb)
5146{
5147 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5148 return 0;
5149 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5150 return 0;
5151 if (sb->s_flags & MS_RDONLY)
5152 return 1;
5153 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5154 return 0;
5155 return 1;
5156}
51b7e3c9 5157MODULE_ALIAS("ext3");
24b58424
TT		 5158#else
5159static inline void register_as_ext3(void) { }
5160static inline void unregister_as_ext3(void) { }
2035e776 5161static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5162#endif
5163
03010a33
TT		 5164static struct file_system_type ext4_fs_type = {
5165 .owner = THIS_MODULE,
5166 .name = "ext4",
152a0836 5167 .mount = ext4_mount,
03010a33
TT		 5168 .kill_sb = kill_block_super,
5169 .fs_flags = FS_REQUIRES_DEV,
5170};
5171
8f021222 5172static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5173{
5174 struct ext4_features *ef;
5175 int ret = -ENOMEM;
5176
5177 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5178 if (!ef)
5179 goto out;
5180
5181 ef->f_kobj.kset = ext4_kset;
5182 init_completion(&ef->f_kobj_unregister);
5183 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5184 "features");
5185 if (ret) {
5186 kfree(ef);
5187 goto out;
5188 }
5189
5190 ext4_feat = ef;
5191 ret = 0;
5192out:
5193 return ret;
5194}
5195
8f021222
LC		 5196static void ext4_exit_feat_adverts(void)
5197{
5198 kobject_put(&ext4_feat->f_kobj);
5199 wait_for_completion(&ext4_feat->f_kobj_unregister);
5200 kfree(ext4_feat);
5201}
5202
e9e3bcec
ES		 5203/* Shared across all ext4 file systems */
5204wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5205struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5206
5dabfc78 5207static int __init ext4_init_fs(void)
ac27a0ec 5208{
e9e3bcec 5209 int i, err;
c9de560d 5210
07c0c5d8
AV		 5211 ext4_li_info = NULL;
5212 mutex_init(&ext4_li_mtx);
5213
9a4c8019 5214 /* Build-time check for flags consistency */
12e9b892 5215 ext4_check_flag_values();
e9e3bcec
ES		 5216
5217 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5218 mutex_init(&ext4__aio_mutex[i]);
5219 init_waitqueue_head(&ext4__ioend_wq[i]);
5220 }
5221
51865fda 5222 err = ext4_init_es();
6fd058f7
TT		 5223 if (err)
5224 return err;
51865fda
ZL		 5225
5226 err = ext4_init_pageio();
5227 if (err)
5228 goto out7;
5229
5dabfc78 5230 err = ext4_init_system_zone();
bd2d0210 5231 if (err)
d44651d0 5232 goto out6;
3197ebdb 5233 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5234 if (!ext4_kset) {
5235 err = -ENOMEM;
dd68314c 5236 goto out5;
0e376b1e 5237 }
d44651d0 5238 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5239
5240 err = ext4_init_feat_adverts();
dd68314c
TT		 5241 if (err)
5242 goto out4;
857ac889 5243
5dabfc78 5244 err = ext4_init_mballoc();
ac27a0ec 5245 if (err)
6fd058f7 5246 goto out3;
c9de560d 5247
5dabfc78 5248 err = ext4_init_xattr();
c9de560d
AT		 5249 if (err)
5250 goto out2;
ac27a0ec
DK		 5251 err = init_inodecache();
5252 if (err)
5253 goto out1;
24b58424 5254 register_as_ext3();
2035e776 5255 register_as_ext2();
03010a33 5256 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5257 if (err)
5258 goto out;
bfff6873 5259
ac27a0ec
DK		 5260 return 0;
5261out:
24b58424
TT		 5262 unregister_as_ext2();
5263 unregister_as_ext3();
ac27a0ec
DK		 5264 destroy_inodecache();
5265out1:
5dabfc78 5266 ext4_exit_xattr();
c9de560d 5267out2:
5dabfc78 5268 ext4_exit_mballoc();
6fd058f7 5269out3:
8f021222 5270 ext4_exit_feat_adverts();
dd68314c 5271out4:
d44651d0
FJ		 5272 if (ext4_proc_root)
5273 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5274 kset_unregister(ext4_kset);
d44651d0 5275out5:
5dabfc78 5276 ext4_exit_system_zone();
d44651d0 5277out6:
5dabfc78 5278 ext4_exit_pageio();
51865fda
ZL		 5279out7:
5280 ext4_exit_es();
5281
ac27a0ec
DK		 5282 return err;
5283}
5284
5dabfc78 5285static void __exit ext4_exit_fs(void)
ac27a0ec 5286{
bfff6873 5287 ext4_destroy_lazyinit_thread();
24b58424
TT		 5288 unregister_as_ext2();
5289 unregister_as_ext3();
03010a33 5290 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5291 destroy_inodecache();
5dabfc78
TT		 5292 ext4_exit_xattr();
5293 ext4_exit_mballoc();
8f021222 5294 ext4_exit_feat_adverts();
9f6200bb 5295 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5296 kset_unregister(ext4_kset);
5dabfc78
TT		 5297 ext4_exit_system_zone();
5298 ext4_exit_pageio();
ac27a0ec
DK		 5299}
5300
5301MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5302MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5303MODULE_LICENSE("GPL");
5dabfc78
TT		 5304module_init(ext4_init_fs)
5305module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository