projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
direct-io: Implement generic deferred AIO completions
[linux-2.6-block.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);
06a407f1 72static int ext4_sync_fs_nojournal(struct super_block *sb, int wait);
2b2d6d01
TT		 73static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 75static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 76static int ext4_freeze(struct super_block *sb);
152a0836
AV		 77static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
2035e776
TT		 79static inline int ext2_feature_set_ok(struct super_block *sb);
80static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 81static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 82static void ext4_destroy_lazyinit_thread(void);
83static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 84static void ext4_clear_request_list(void);
27dd4385 85static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
ac27a0ec 86
2035e776
TT		 87#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94};
7f78e035 95MODULE_ALIAS_FS("ext2");
fa7614dd 96MODULE_ALIAS("ext2");
2035e776
TT		 97#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98#else
99#define IS_EXT2_SB(sb) (0)
100#endif
101
102
ba69f9ab
JK		 103#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
106 .name = "ext3",
152a0836 107 .mount = ext4_mount,
ba69f9ab
JK		 108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
110};
7f78e035 111MODULE_ALIAS_FS("ext3");
fa7614dd 112MODULE_ALIAS("ext3");
ba69f9ab
JK		 113#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114#else
115#define IS_EXT3_SB(sb) (0)
116#endif
bd81d8ee 117
d25425f8
DW		 118static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
120{
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
123 return 1;
124
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
126}
127
a9c47317
DW		 128static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
130{
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
133 __u32 csum;
134
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136
137 return cpu_to_le32(csum);
138}
139
140int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
142{
143 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
144 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
145 return 1;
146
147 return es->s_checksum == ext4_superblock_csum(sb, es);
148}
149
06db49e6 150void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 151{
06db49e6
TT		 152 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
153
a9c47317
DW		 154 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
155 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
156 return;
157
158 es->s_checksum = ext4_superblock_csum(sb, es);
159}
160
9933fc0a
TT		 161void *ext4_kvmalloc(size_t size, gfp_t flags)
162{
163 void *ret;
164
165 ret = kmalloc(size, flags);
166 if (!ret)
167 ret = __vmalloc(size, flags, PAGE_KERNEL);
168 return ret;
169}
170
171void *ext4_kvzalloc(size_t size, gfp_t flags)
172{
173 void *ret;
174
db9481c0 175 ret = kzalloc(size, flags);
9933fc0a
TT		 176 if (!ret)
177 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
178 return ret;
179}
180
181void ext4_kvfree(void *ptr)
182{
183 if (is_vmalloc_addr(ptr))
184 vfree(ptr);
185 else
186 kfree(ptr);
187
188}
189
8fadc143
AR		 190ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
191 struct ext4_group_desc *bg)
bd81d8ee 192{
3a14589c 193 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 194 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 195 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 196}
197
8fadc143
AR		 198ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
199 struct ext4_group_desc *bg)
bd81d8ee 200{
5272f837 201 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 202 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 203 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 204}
205
8fadc143
AR		 206ext4_fsblk_t ext4_inode_table(struct super_block *sb,
207 struct ext4_group_desc *bg)
bd81d8ee 208{
5272f837 209 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 210 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 211 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 212}
213
021b65bb
TT		 214__u32 ext4_free_group_clusters(struct super_block *sb,
215 struct ext4_group_desc *bg)
560671a0
AK		 216{
217 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
218 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 219 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 220}
221
222__u32 ext4_free_inodes_count(struct super_block *sb,
223 struct ext4_group_desc *bg)
224{
225 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
226 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 227 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 228}
229
230__u32 ext4_used_dirs_count(struct super_block *sb,
231 struct ext4_group_desc *bg)
232{
233 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
234 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 235 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 236}
237
238__u32 ext4_itable_unused_count(struct super_block *sb,
239 struct ext4_group_desc *bg)
240{
241 return le16_to_cpu(bg->bg_itable_unused_lo) |
242 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 243 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 244}
245
8fadc143
AR		 246void ext4_block_bitmap_set(struct super_block *sb,
247 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 248{
3a14589c 249 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 250 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
251 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 252}
253
8fadc143
AR		 254void ext4_inode_bitmap_set(struct super_block *sb,
255 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 256{
5272f837 257 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 258 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
259 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 260}
261
8fadc143
AR		 262void ext4_inode_table_set(struct super_block *sb,
263 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 264{
5272f837 265 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 266 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
267 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 268}
269
021b65bb
TT		 270void ext4_free_group_clusters_set(struct super_block *sb,
271 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 272{
273 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
274 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
275 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
276}
277
278void ext4_free_inodes_set(struct super_block *sb,
279 struct ext4_group_desc *bg, __u32 count)
280{
281 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
282 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
283 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
284}
285
286void ext4_used_dirs_set(struct super_block *sb,
287 struct ext4_group_desc *bg, __u32 count)
288{
289 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
290 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
291 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
292}
293
294void ext4_itable_unused_set(struct super_block *sb,
295 struct ext4_group_desc *bg, __u32 count)
296{
297 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
298 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
299 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
300}
301
d3d1faf6 302
1c13d5c0
TT		 303static void __save_error_info(struct super_block *sb, const char *func,
304 unsigned int line)
305{
306 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
307
308 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
309 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
310 es->s_last_error_time = cpu_to_le32(get_seconds());
311 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
312 es->s_last_error_line = cpu_to_le32(line);
313 if (!es->s_first_error_time) {
314 es->s_first_error_time = es->s_last_error_time;
315 strncpy(es->s_first_error_func, func,
316 sizeof(es->s_first_error_func));
317 es->s_first_error_line = cpu_to_le32(line);
318 es->s_first_error_ino = es->s_last_error_ino;
319 es->s_first_error_block = es->s_last_error_block;
320 }
66e61a9e
TT		 321 /*
322 * Start the daily error reporting function if it hasn't been
323 * started already
324 */
325 if (!es->s_error_count)
326 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 327 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 328}
329
330static void save_error_info(struct super_block *sb, const char *func,
331 unsigned int line)
332{
333 __save_error_info(sb, func, line);
334 ext4_commit_super(sb, 1);
335}
336
7c2e7087
TT		 337/*
338 * The del_gendisk() function uninitializes the disk-specific data
339 * structures, including the bdi structure, without telling anyone
340 * else. Once this happens, any attempt to call mark_buffer_dirty()
341 * (for example, by ext4_commit_super), will cause a kernel OOPS.
342 * This is a kludge to prevent these oops until we can put in a proper
343 * hook in del_gendisk() to inform the VFS and file system layers.
344 */
345static int block_device_ejected(struct super_block *sb)
346{
347 struct inode *bd_inode = sb->s_bdev->bd_inode;
348 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
349
350 return bdi->dev == NULL;
351}
352
18aadd47
BJ		 353static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
354{
355 struct super_block *sb = journal->j_private;
356 struct ext4_sb_info *sbi = EXT4_SB(sb);
357 int error = is_journal_aborted(journal);
5d3ee208 358 struct ext4_journal_cb_entry *jce;
18aadd47 359
5d3ee208 360 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 361 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 362 while (!list_empty(&txn->t_private_list)) {
363 jce = list_entry(txn->t_private_list.next,
364 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 365 list_del_init(&jce->jce_list);
366 spin_unlock(&sbi->s_md_lock);
367 jce->jce_func(sb, jce, error);
368 spin_lock(&sbi->s_md_lock);
369 }
370 spin_unlock(&sbi->s_md_lock);
371}
1c13d5c0 372
ac27a0ec
DK		 373/* Deal with the reporting of failure conditions on a filesystem such as
374 * inconsistencies detected or read IO failures.
375 *
376 * On ext2, we can store the error state of the filesystem in the
617ba13b 377 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 378 * write ordering constraints on the superblock which prevent us from
379 * writing it out straight away; and given that the journal is about to
380 * be aborted, we can't rely on the current, or future, transactions to
381 * write out the superblock safely.
382 *
dab291af 383 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 384 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 385 * that error until we've noted it down and cleared it.
386 */
387
617ba13b 388static void ext4_handle_error(struct super_block *sb)
ac27a0ec 389{
ac27a0ec
DK		 390 if (sb->s_flags & MS_RDONLY)
391 return;
392
2b2d6d01 393 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 394 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 395
4ab2f15b 396 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 397 if (journal)
dab291af 398 jbd2_journal_abort(journal, -EIO);
ac27a0ec 399 }
2b2d6d01 400 if (test_opt(sb, ERRORS_RO)) {
b31e1552 401 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 402 /*
403 * Make sure updated value of ->s_mount_flags will be visible
404 * before ->s_flags update
405 */
406 smp_wmb();
ac27a0ec
DK		 407 sb->s_flags |= MS_RDONLY;
408 }
ac27a0ec 409 if (test_opt(sb, ERRORS_PANIC))
617ba13b 410 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 411 sb->s_id);
412}
413
12062ddd 414void __ext4_error(struct super_block *sb, const char *function,
c398eda0 415 unsigned int line, const char *fmt, ...)
ac27a0ec 416{
0ff2ea7d 417 struct va_format vaf;
ac27a0ec
DK		 418 va_list args;
419
420 va_start(args, fmt);
0ff2ea7d
JP		 421 vaf.fmt = fmt;
422 vaf.va = &args;
423 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
424 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec 425 va_end(args);
f3fc0210 426 save_error_info(sb, function, line);
ac27a0ec 427
617ba13b 428 ext4_handle_error(sb);
ac27a0ec
DK		 429}
430
e7c96e8e
JP		 431void __ext4_error_inode(struct inode *inode, const char *function,
432 unsigned int line, ext4_fsblk_t block,
433 const char *fmt, ...)
273df556
FM		 434{
435 va_list args;
f7c21177 436 struct va_format vaf;
1c13d5c0 437 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 438
1c13d5c0
TT		 439 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
440 es->s_last_error_block = cpu_to_le64(block);
441 save_error_info(inode->i_sb, function, line);
273df556 442 va_start(args, fmt);
f7c21177
TT		 443 vaf.fmt = fmt;
444 vaf.va = &args;
c398eda0 445 if (block)
d9ee81da
JP		 446 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
447 "inode #%lu: block %llu: comm %s: %pV\n",
448 inode->i_sb->s_id, function, line, inode->i_ino,
449 block, current->comm, &vaf);
450 else
451 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
452 "inode #%lu: comm %s: %pV\n",
453 inode->i_sb->s_id, function, line, inode->i_ino,
454 current->comm, &vaf);
273df556
FM		 455 va_end(args);
456
457 ext4_handle_error(inode->i_sb);
458}
459
e7c96e8e
JP		 460void __ext4_error_file(struct file *file, const char *function,
461 unsigned int line, ext4_fsblk_t block,
462 const char *fmt, ...)
273df556
FM		 463{
464 va_list args;
f7c21177 465 struct va_format vaf;
1c13d5c0 466 struct ext4_super_block *es;
496ad9aa 467 struct inode *inode = file_inode(file);
273df556
FM		 468 char pathname[80], *path;
469
1c13d5c0
TT		 470 es = EXT4_SB(inode->i_sb)->s_es;
471 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
472 save_error_info(inode->i_sb, function, line);
273df556 473 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 474 if (IS_ERR(path))
273df556 475 path = "(unknown)";
f7c21177
TT		 476 va_start(args, fmt);
477 vaf.fmt = fmt;
478 vaf.va = &args;
d9ee81da
JP		 479 if (block)
480 printk(KERN_CRIT
481 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
482 "block %llu: comm %s: path %s: %pV\n",
483 inode->i_sb->s_id, function, line, inode->i_ino,
484 block, current->comm, path, &vaf);
485 else
486 printk(KERN_CRIT
487 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
488 "comm %s: path %s: %pV\n",
489 inode->i_sb->s_id, function, line, inode->i_ino,
490 current->comm, path, &vaf);
273df556
FM		 491 va_end(args);
492
493 ext4_handle_error(inode->i_sb);
494}
495
722887dd
TT		 496const char *ext4_decode_error(struct super_block *sb, int errno,
497 char nbuf[16])
ac27a0ec
DK		 498{
499 char *errstr = NULL;
500
501 switch (errno) {
502 case -EIO:
503 errstr = "IO failure";
504 break;
505 case -ENOMEM:
506 errstr = "Out of memory";
507 break;
508 case -EROFS:
78f1ddbb
TT		 509 if (!sb || (EXT4_SB(sb)->s_journal &&
510 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 511 errstr = "Journal has aborted";
512 else
513 errstr = "Readonly filesystem";
514 break;
515 default:
516 /* If the caller passed in an extra buffer for unknown
517 * errors, textualise them now. Else we just return
518 * NULL. */
519 if (nbuf) {
520 /* Check for truncated error codes... */
521 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
522 errstr = nbuf;
523 }
524 break;
525 }
526
527 return errstr;
528}
529
617ba13b 530/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 531 * automatically and invokes the appropriate error response. */
532
c398eda0
TT		 533void __ext4_std_error(struct super_block *sb, const char *function,
534 unsigned int line, int errno)
ac27a0ec
DK		 535{
536 char nbuf[16];
537 const char *errstr;
538
539 /* Special case: if the error is EROFS, and we're not already
540 * inside a transaction, then there's really no point in logging
541 * an error. */
542 if (errno == -EROFS && journal_current_handle() == NULL &&
543 (sb->s_flags & MS_RDONLY))
544 return;
545
617ba13b 546 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 547 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
548 sb->s_id, function, line, errstr);
1c13d5c0 549 save_error_info(sb, function, line);
ac27a0ec 550
617ba13b 551 ext4_handle_error(sb);
ac27a0ec
DK		 552}
553
554/*
617ba13b 555 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 556 * abort function may be used to deal with unrecoverable failures such
557 * as journal IO errors or ENOMEM at a critical moment in log management.
558 *
559 * We unconditionally force the filesystem into an ABORT|READONLY state,
560 * unless the error response on the fs has been set to panic in which
561 * case we take the easy way out and panic immediately.
562 */
563
c67d859e 564void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 565 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 566{
567 va_list args;
568
1c13d5c0 569 save_error_info(sb, function, line);
ac27a0ec 570 va_start(args, fmt);
c398eda0
TT		 571 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
572 function, line);
ac27a0ec
DK		 573 vprintk(fmt, args);
574 printk("\n");
575 va_end(args);
576
1c13d5c0
TT		 577 if ((sb->s_flags & MS_RDONLY) == 0) {
578 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 579 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 580 /*
581 * Make sure updated value of ->s_mount_flags will be visible
582 * before ->s_flags update
583 */
584 smp_wmb();
585 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 586 if (EXT4_SB(sb)->s_journal)
587 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
588 save_error_info(sb, function, line);
589 }
ac27a0ec 590 if (test_opt(sb, ERRORS_PANIC))
617ba13b 591 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 592}
593
e7c96e8e
JP		 594void __ext4_msg(struct super_block *sb,
595 const char *prefix, const char *fmt, ...)
b31e1552 596{
0ff2ea7d 597 struct va_format vaf;
b31e1552
ES		 598 va_list args;
599
600 va_start(args, fmt);
0ff2ea7d
JP		 601 vaf.fmt = fmt;
602 vaf.va = &args;
603 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 604 va_end(args);
605}
606
12062ddd 607void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 608 unsigned int line, const char *fmt, ...)
ac27a0ec 609{
0ff2ea7d 610 struct va_format vaf;
ac27a0ec
DK		 611 va_list args;
612
613 va_start(args, fmt);
0ff2ea7d
JP		 614 vaf.fmt = fmt;
615 vaf.va = &args;
616 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
617 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 618 va_end(args);
619}
620
e29136f8
TT		 621void __ext4_grp_locked_error(const char *function, unsigned int line,
622 struct super_block *sb, ext4_group_t grp,
623 unsigned long ino, ext4_fsblk_t block,
624 const char *fmt, ...)
5d1b1b3f
AK		 625__releases(bitlock)
626__acquires(bitlock)
627{
0ff2ea7d 628 struct va_format vaf;
5d1b1b3f
AK		 629 va_list args;
630 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
631
1c13d5c0
TT		 632 es->s_last_error_ino = cpu_to_le32(ino);
633 es->s_last_error_block = cpu_to_le64(block);
634 __save_error_info(sb, function, line);
0ff2ea7d 635
5d1b1b3f 636 va_start(args, fmt);
0ff2ea7d
JP		 637
638 vaf.fmt = fmt;
639 vaf.va = &args;
21149d61 640 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 641 sb->s_id, function, line, grp);
642 if (ino)
0ff2ea7d 643 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 644 if (block)
0ff2ea7d
JP		 645 printk(KERN_CONT "block %llu:", (unsigned long long) block);
646 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 647 va_end(args);
648
649 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 650 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 651 return;
652 }
1c13d5c0 653
5d1b1b3f
AK		 654 ext4_unlock_group(sb, grp);
655 ext4_handle_error(sb);
656 /*
657 * We only get here in the ERRORS_RO case; relocking the group
658 * may be dangerous, but nothing bad will happen since the
659 * filesystem will have already been marked read/only and the
660 * journal has been aborted. We return 1 as a hint to callers
661 * who might what to use the return value from
25985edc 662 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 663 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
664 * aggressively from the ext4 function in question, with a
665 * more appropriate error code.
666 */
667 ext4_lock_group(sb, grp);
668 return;
669}
670
617ba13b 671void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 672{
617ba13b 673 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 674
617ba13b 675 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 676 return;
677
12062ddd 678 ext4_warning(sb,
ac27a0ec
DK		 679 "updating to rev %d because of new feature flag, "
680 "running e2fsck is recommended",
617ba13b 681 EXT4_DYNAMIC_REV);
ac27a0ec 682
617ba13b
MC		 683 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
684 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
685 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 686 /* leave es->s_feature_*compat flags alone */
687 /* es->s_uuid will be set by e2fsck if empty */
688
689 /*
690 * The rest of the superblock fields should be zero, and if not it
691 * means they are likely already in use, so leave them alone. We
692 * can leave it up to e2fsck to clean up any inconsistencies there.
693 */
694}
695
696/*
697 * Open the external journal device
698 */
b31e1552 699static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 700{
701 struct block_device *bdev;
702 char b[BDEVNAME_SIZE];
703
d4d77629 704 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 705 if (IS_ERR(bdev))
706 goto fail;
707 return bdev;
708
709fail:
b31e1552 710 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 711 __bdevname(dev, b), PTR_ERR(bdev));
712 return NULL;
713}
714
715/*
716 * Release the journal device
717 */
4385bab1 718static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 719{
4385bab1 720 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 721}
722
4385bab1 723static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 724{
725 struct block_device *bdev;
ac27a0ec
DK		 726 bdev = sbi->journal_bdev;
727 if (bdev) {
4385bab1 728 ext4_blkdev_put(bdev);
ac27a0ec
DK		 729 sbi->journal_bdev = NULL;
730 }
ac27a0ec
DK		 731}
732
733static inline struct inode *orphan_list_entry(struct list_head *l)
734{
617ba13b 735 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 736}
737
617ba13b 738static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 739{
740 struct list_head *l;
741
b31e1552
ES		 742 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
743 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 744
745 printk(KERN_ERR "sb_info orphan list:\n");
746 list_for_each(l, &sbi->s_orphan) {
747 struct inode *inode = orphan_list_entry(l);
748 printk(KERN_ERR " "
749 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
750 inode->i_sb->s_id, inode->i_ino, inode,
751 inode->i_mode, inode->i_nlink,
752 NEXT_ORPHAN(inode));
753 }
754}
755
2b2d6d01 756static void ext4_put_super(struct super_block *sb)
ac27a0ec 757{
617ba13b
MC		 758 struct ext4_sb_info *sbi = EXT4_SB(sb);
759 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 760 int i, err;
ac27a0ec 761
857ac889 762 ext4_unregister_li_request(sb);
e0ccfd95
CH		 763 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
764
2e8fa54e 765 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 766 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 767
0390131b
FM		 768 if (sbi->s_journal) {
769 err = jbd2_journal_destroy(sbi->s_journal);
770 sbi->s_journal = NULL;
771 if (err < 0)
c67d859e 772 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 773 }
d4edac31 774
d3922a77 775 ext4_es_unregister_shrinker(sbi);
a1c6c569 776 del_timer(&sbi->s_err_report);
d4edac31
JB		 777 ext4_release_system_zone(sb);
778 ext4_mb_release(sb);
779 ext4_ext_release(sb);
780 ext4_xattr_put_super(sb);
781
ac27a0ec 782 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 783 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 784 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 785 }
58c5873a 786 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 787 ext4_commit_super(sb, 1);
788
240799cd 789 if (sbi->s_proc) {
66acdcf4 790 remove_proc_entry("options", sbi->s_proc);
9f6200bb 791 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 792 }
3197ebdb 793 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 794
795 for (i = 0; i < sbi->s_gdb_count; i++)
796 brelse(sbi->s_group_desc[i]);
f18a5f21 797 ext4_kvfree(sbi->s_group_desc);
9933fc0a 798 ext4_kvfree(sbi->s_flex_groups);
57042651 799 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 800 percpu_counter_destroy(&sbi->s_freeinodes_counter);
801 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 802 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 803 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
ac27a0ec
DK		 804 brelse(sbi->s_sbh);
805#ifdef CONFIG_QUOTA
806 for (i = 0; i < MAXQUOTAS; i++)
807 kfree(sbi->s_qf_names[i]);
808#endif
809
810 /* Debugging code just in case the in-memory inode orphan list
811 * isn't empty. The on-disk one can be non-empty if we've
812 * detected an error and taken the fs readonly, but the
813 * in-memory list had better be clean by this point. */
814 if (!list_empty(&sbi->s_orphan))
815 dump_orphan_list(sb, sbi);
816 J_ASSERT(list_empty(&sbi->s_orphan));
817
f98393a6 818 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 819 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
820 /*
821 * Invalidate the journal device's buffers. We don't want them
822 * floating about in memory - the physical journal device may
823 * hotswapped, and it breaks the `ro-after' testing code.
824 */
825 sync_blockdev(sbi->journal_bdev);
f98393a6 826 invalidate_bdev(sbi->journal_bdev);
617ba13b 827 ext4_blkdev_remove(sbi);
ac27a0ec 828 }
c5e06d10
JL		 829 if (sbi->s_mmp_tsk)
830 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 831 sb->s_fs_info = NULL;
3197ebdb
TT		 832 /*
833 * Now that we are completely done shutting down the
834 * superblock, we need to actually destroy the kobject.
835 */
3197ebdb
TT		 836 kobject_put(&sbi->s_kobj);
837 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 838 if (sbi->s_chksum_driver)
839 crypto_free_shash(sbi->s_chksum_driver);
705895b6 840 kfree(sbi->s_blockgroup_lock);
ac27a0ec 841 kfree(sbi);
ac27a0ec
DK		 842}
843
e18b890b 844static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 845
846/*
847 * Called inside transaction, so use GFP_NOFS
848 */
617ba13b 849static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 850{
617ba13b 851 struct ext4_inode_info *ei;
ac27a0ec 852
e6b4f8da 853 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 854 if (!ei)
855 return NULL;
0b8e58a1 856
ac27a0ec 857 ei->vfs_inode.i_version = 1;
c9de560d
AT		 858 INIT_LIST_HEAD(&ei->i_prealloc_list);
859 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 860 ext4_es_init_tree(&ei->i_es_tree);
861 rwlock_init(&ei->i_es_lock);
74cd15cd
ZL		 862 INIT_LIST_HEAD(&ei->i_es_lru);
863 ei->i_es_lru_nr = 0;
d3922a77 864 ei->i_touch_when = 0;
d2a17637
MC		 865 ei->i_reserved_data_blocks = 0;
866 ei->i_reserved_meta_blocks = 0;
867 ei->i_allocated_meta_blocks = 0;
9d0be502 868 ei->i_da_metadata_calc_len = 0;
7e731bc9 869 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 870 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 871#ifdef CONFIG_QUOTA
872 ei->i_reserved_quota = 0;
873#endif
8aefcd55 874 ei->jinode = NULL;
2e8fa54e 875 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 876 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 877 ei->i_sync_tid = 0;
878 ei->i_datasync_tid = 0;
f7ad6d2e 879 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 880 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 881 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
0b8e58a1 882
ac27a0ec
DK		 883 return &ei->vfs_inode;
884}
885
7ff9c073
TT		 886static int ext4_drop_inode(struct inode *inode)
887{
888 int drop = generic_drop_inode(inode);
889
890 trace_ext4_drop_inode(inode, drop);
891 return drop;
892}
893
fa0d7e3d
NP		 894static void ext4_i_callback(struct rcu_head *head)
895{
896 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 897 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
898}
899
617ba13b 900static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 901{
9f7dd93d 902 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 903 ext4_msg(inode->i_sb, KERN_ERR,
904 "Inode %lu (%p): orphan list check failed!",
905 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 906 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
907 EXT4_I(inode), sizeof(struct ext4_inode_info),
908 true);
909 dump_stack();
910 }
fa0d7e3d 911 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 912}
913
51cc5068 914static void init_once(void *foo)
ac27a0ec 915{
617ba13b 916 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 917
a35afb83 918 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 919 init_rwsem(&ei->xattr_sem);
0e855ac8 920 init_rwsem(&ei->i_data_sem);
a35afb83 921 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 922}
923
924static int init_inodecache(void)
925{
617ba13b
MC		 926 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
927 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 928 0, (SLAB_RECLAIM_ACCOUNT|
929 SLAB_MEM_SPREAD),
20c2df83 930 init_once);
617ba13b 931 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 932 return -ENOMEM;
933 return 0;
934}
935
936static void destroy_inodecache(void)
937{
8c0a8537
KS		 938 /*
939 * Make sure all delayed rcu free inodes are flushed before we
940 * destroy cache.
941 */
942 rcu_barrier();
617ba13b 943 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 944}
945
0930fcc1 946void ext4_clear_inode(struct inode *inode)
ac27a0ec 947{
0930fcc1 948 invalidate_inode_buffers(inode);
dbd5768f 949 clear_inode(inode);
9f754758 950 dquot_drop(inode);
c2ea3fde 951 ext4_discard_preallocations(inode);
51865fda 952 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
74cd15cd 953 ext4_es_lru_del(inode);
8aefcd55
TT		 954 if (EXT4_I(inode)->jinode) {
955 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
956 EXT4_I(inode)->jinode);
957 jbd2_free_inode(EXT4_I(inode)->jinode);
958 EXT4_I(inode)->jinode = NULL;
959 }
ac27a0ec
DK		 960}
961
1b961ac0 962static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 963 u64 ino, u32 generation)
ac27a0ec 964{
ac27a0ec 965 struct inode *inode;
ac27a0ec 966
617ba13b 967 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 968 return ERR_PTR(-ESTALE);
617ba13b 969 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 970 return ERR_PTR(-ESTALE);
971
972 /* iget isn't really right if the inode is currently unallocated!!
973 *
617ba13b 974 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 975 * deleted, so we should be safe.
976 *
977 * Currently we don't know the generation for parent directory, so
978 * a generation of 0 means "accept any"
979 */
1d1fe1ee
DH		 980 inode = ext4_iget(sb, ino);
981 if (IS_ERR(inode))
982 return ERR_CAST(inode);
983 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 984 iput(inode);
985 return ERR_PTR(-ESTALE);
986 }
1b961ac0
CH		 987
988 return inode;
989}
990
991static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 992 int fh_len, int fh_type)
1b961ac0
CH		 993{
994 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
995 ext4_nfs_get_inode);
996}
997
998static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 999 int fh_len, int fh_type)
1b961ac0
CH		 1000{
1001 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1002 ext4_nfs_get_inode);
ac27a0ec
DK		 1003}
1004
c39a7f84
TO		 1005/*
1006 * Try to release metadata pages (indirect blocks, directories) which are
1007 * mapped via the block device. Since these pages could have journal heads
1008 * which would prevent try_to_free_buffers() from freeing them, we must use
1009 * jbd2 layer's try_to_free_buffers() function to release them.
1010 */
0b8e58a1
AD		 1011static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1012 gfp_t wait)
c39a7f84
TO		 1013{
1014 journal_t *journal = EXT4_SB(sb)->s_journal;
1015
1016 WARN_ON(PageChecked(page));
1017 if (!page_has_buffers(page))
1018 return 0;
1019 if (journal)
1020 return jbd2_journal_try_to_free_buffers(journal, page,
1021 wait & ~__GFP_WAIT);
1022 return try_to_free_buffers(page);
1023}
1024
ac27a0ec 1025#ifdef CONFIG_QUOTA
af5bc92d 1026#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1027#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1028
617ba13b
MC		 1029static int ext4_write_dquot(struct dquot *dquot);
1030static int ext4_acquire_dquot(struct dquot *dquot);
1031static int ext4_release_dquot(struct dquot *dquot);
1032static int ext4_mark_dquot_dirty(struct dquot *dquot);
1033static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1034static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1035 struct path *path);
7c319d32
AK		 1036static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1037 int format_id);
ca0e05e4 1038static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1039static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1040static int ext4_quota_on_mount(struct super_block *sb, int type);
1041static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1042 size_t len, loff_t off);
617ba13b 1043static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1044 const char *data, size_t len, loff_t off);
7c319d32
AK		 1045static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1046 unsigned int flags);
1047static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1048
61e225dc 1049static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1050 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1051 .write_dquot = ext4_write_dquot,
1052 .acquire_dquot = ext4_acquire_dquot,
1053 .release_dquot = ext4_release_dquot,
1054 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1055 .write_info = ext4_write_info,
1056 .alloc_dquot = dquot_alloc,
1057 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1058};
1059
0d54b217 1060static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1061 .quota_on = ext4_quota_on,
ca0e05e4 1062 .quota_off = ext4_quota_off,
287a8095
CH		 1063 .quota_sync = dquot_quota_sync,
1064 .get_info = dquot_get_dqinfo,
1065 .set_info = dquot_set_dqinfo,
1066 .get_dqblk = dquot_get_dqblk,
1067 .set_dqblk = dquot_set_dqblk
ac27a0ec 1068};
7c319d32
AK		 1069
1070static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1071 .quota_on_meta = ext4_quota_on_sysfile,
1072 .quota_off = ext4_quota_off_sysfile,
1073 .quota_sync = dquot_quota_sync,
1074 .get_info = dquot_get_dqinfo,
1075 .set_info = dquot_set_dqinfo,
1076 .get_dqblk = dquot_get_dqblk,
1077 .set_dqblk = dquot_set_dqblk
1078};
ac27a0ec
DK		 1079#endif
1080
ee9b6d61 1081static const struct super_operations ext4_sops = {
617ba13b
MC		 1082 .alloc_inode = ext4_alloc_inode,
1083 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1084 .write_inode = ext4_write_inode,
1085 .dirty_inode = ext4_dirty_inode,
7ff9c073 1086 .drop_inode = ext4_drop_inode,
0930fcc1 1087 .evict_inode = ext4_evict_inode,
617ba13b 1088 .put_super = ext4_put_super,
617ba13b 1089 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1090 .freeze_fs = ext4_freeze,
1091 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1092 .statfs = ext4_statfs,
1093 .remount_fs = ext4_remount,
617ba13b 1094 .show_options = ext4_show_options,
ac27a0ec 1095#ifdef CONFIG_QUOTA
617ba13b
MC		 1096 .quota_read = ext4_quota_read,
1097 .quota_write = ext4_quota_write,
ac27a0ec 1098#endif
c39a7f84 1099 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1100};
1101
9ca92389
TT		 1102static const struct super_operations ext4_nojournal_sops = {
1103 .alloc_inode = ext4_alloc_inode,
1104 .destroy_inode = ext4_destroy_inode,
1105 .write_inode = ext4_write_inode,
1106 .dirty_inode = ext4_dirty_inode,
7ff9c073 1107 .drop_inode = ext4_drop_inode,
0930fcc1 1108 .evict_inode = ext4_evict_inode,
06a407f1 1109 .sync_fs = ext4_sync_fs_nojournal,
9ca92389
TT		 1110 .put_super = ext4_put_super,
1111 .statfs = ext4_statfs,
1112 .remount_fs = ext4_remount,
9ca92389
TT		 1113 .show_options = ext4_show_options,
1114#ifdef CONFIG_QUOTA
1115 .quota_read = ext4_quota_read,
1116 .quota_write = ext4_quota_write,
1117#endif
1118 .bdev_try_to_free_page = bdev_try_to_free_page,
1119};
1120
39655164 1121static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1122 .fh_to_dentry = ext4_fh_to_dentry,
1123 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1124 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1125};
1126
1127enum {
1128 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1129 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1130 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1131 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1132 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1133 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1134 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1135 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1136 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1137 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1138 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1139 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1140 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1141 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1142 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1143 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1144 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1145 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1146 Opt_max_dir_size_kb,
ac27a0ec
DK		 1147};
1148
a447c093 1149static const match_table_t tokens = {
ac27a0ec
DK		 1150 {Opt_bsd_df, "bsddf"},
1151 {Opt_minix_df, "minixdf"},
1152 {Opt_grpid, "grpid"},
1153 {Opt_grpid, "bsdgroups"},
1154 {Opt_nogrpid, "nogrpid"},
1155 {Opt_nogrpid, "sysvgroups"},
1156 {Opt_resgid, "resgid=%u"},
1157 {Opt_resuid, "resuid=%u"},
1158 {Opt_sb, "sb=%u"},
1159 {Opt_err_cont, "errors=continue"},
1160 {Opt_err_panic, "errors=panic"},
1161 {Opt_err_ro, "errors=remount-ro"},
1162 {Opt_nouid32, "nouid32"},
ac27a0ec 1163 {Opt_debug, "debug"},
72578c33
TT		 1164 {Opt_removed, "oldalloc"},
1165 {Opt_removed, "orlov"},
ac27a0ec
DK		 1166 {Opt_user_xattr, "user_xattr"},
1167 {Opt_nouser_xattr, "nouser_xattr"},
1168 {Opt_acl, "acl"},
1169 {Opt_noacl, "noacl"},
e3bb52ae 1170 {Opt_noload, "norecovery"},
5a916be1 1171 {Opt_noload, "noload"},
72578c33
TT		 1172 {Opt_removed, "nobh"},
1173 {Opt_removed, "bh"},
ac27a0ec 1174 {Opt_commit, "commit=%u"},
30773840
TT		 1175 {Opt_min_batch_time, "min_batch_time=%u"},
1176 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1177 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1178 {Opt_journal_checksum, "journal_checksum"},
1179 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1180 {Opt_abort, "abort"},
1181 {Opt_data_journal, "data=journal"},
1182 {Opt_data_ordered, "data=ordered"},
1183 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1184 {Opt_data_err_abort, "data_err=abort"},
1185 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1186 {Opt_offusrjquota, "usrjquota="},
1187 {Opt_usrjquota, "usrjquota=%s"},
1188 {Opt_offgrpjquota, "grpjquota="},
1189 {Opt_grpjquota, "grpjquota=%s"},
1190 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1191 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1192 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1193 {Opt_grpquota, "grpquota"},
1194 {Opt_noquota, "noquota"},
1195 {Opt_quota, "quota"},
1196 {Opt_usrquota, "usrquota"},
1197 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1198 {Opt_barrier, "barrier"},
1199 {Opt_nobarrier, "nobarrier"},
25ec56b5 1200 {Opt_i_version, "i_version"},
c9de560d 1201 {Opt_stripe, "stripe=%u"},
64769240 1202 {Opt_delalloc, "delalloc"},
dd919b98 1203 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1204 {Opt_removed, "mblk_io_submit"},
1205 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1206 {Opt_block_validity, "block_validity"},
1207 {Opt_noblock_validity, "noblock_validity"},
240799cd 1208 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1209 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1210 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1211 {Opt_auto_da_alloc, "auto_da_alloc"},
1212 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1213 {Opt_dioread_nolock, "dioread_nolock"},
1214 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1215 {Opt_discard, "discard"},
1216 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1217 {Opt_init_itable, "init_itable=%u"},
1218 {Opt_init_itable, "init_itable"},
1219 {Opt_noinit_itable, "noinit_itable"},
df981d03 1220 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1221 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1222 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1223 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1224 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1225 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1226 {Opt_err, NULL},
ac27a0ec
DK		 1227};
1228
617ba13b 1229static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1230{
617ba13b 1231 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1232 char *options = (char *) *data;
1233
1234 if (!options || strncmp(options, "sb=", 3) != 0)
1235 return 1; /* Default location */
0b8e58a1 1236
ac27a0ec 1237 options += 3;
0b8e58a1 1238 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1239 sb_block = simple_strtoul(options, &options, 0);
1240 if (*options && *options != ',') {
4776004f 1241 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1242 (char *) *data);
1243 return 1;
1244 }
1245 if (*options == ',')
1246 options++;
1247 *data = (void *) options;
0b8e58a1 1248
ac27a0ec
DK		 1249 return sb_block;
1250}
1251
b3881f74 1252#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1253static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1254 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1255
56c50f11
DM		 1256#ifdef CONFIG_QUOTA
1257static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1258{
1259 struct ext4_sb_info *sbi = EXT4_SB(sb);
1260 char *qname;
03dafb5f 1261 int ret = -1;
56c50f11
DM		 1262
1263 if (sb_any_quota_loaded(sb) &&
1264 !sbi->s_qf_names[qtype]) {
1265 ext4_msg(sb, KERN_ERR,
1266 "Cannot change journaled "
1267 "quota options when quota turned on");
57f73c2c 1268 return -1;
56c50f11 1269 }
262b4662
JK		 1270 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1271 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1272 "when QUOTA feature is enabled");
1273 return -1;
1274 }
56c50f11
DM		 1275 qname = match_strdup(args);
1276 if (!qname) {
1277 ext4_msg(sb, KERN_ERR,
1278 "Not enough memory for storing quotafile name");
57f73c2c 1279 return -1;
56c50f11 1280 }
03dafb5f
CG		 1281 if (sbi->s_qf_names[qtype]) {
1282 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1283 ret = 1;
1284 else
1285 ext4_msg(sb, KERN_ERR,
1286 "%s quota file already specified",
1287 QTYPE2NAME(qtype));
1288 goto errout;
56c50f11 1289 }
03dafb5f 1290 if (strchr(qname, '/')) {
56c50f11
DM		 1291 ext4_msg(sb, KERN_ERR,
1292 "quotafile must be on filesystem root");
03dafb5f 1293 goto errout;
56c50f11 1294 }
03dafb5f 1295 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1296 set_opt(sb, QUOTA);
56c50f11 1297 return 1;
03dafb5f
CG		 1298errout:
1299 kfree(qname);
1300 return ret;
56c50f11
DM		 1301}
1302
1303static int clear_qf_name(struct super_block *sb, int qtype)
1304{
1305
1306 struct ext4_sb_info *sbi = EXT4_SB(sb);
1307
1308 if (sb_any_quota_loaded(sb) &&
1309 sbi->s_qf_names[qtype]) {
1310 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1311 " when quota turned on");
57f73c2c 1312 return -1;
56c50f11 1313 }
03dafb5f 1314 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1315 sbi->s_qf_names[qtype] = NULL;
1316 return 1;
1317}
1318#endif
1319
26092bf5
TT		 1320#define MOPT_SET 0x0001
1321#define MOPT_CLEAR 0x0002
1322#define MOPT_NOSUPPORT 0x0004
1323#define MOPT_EXPLICIT 0x0008
1324#define MOPT_CLEAR_ERR 0x0010
1325#define MOPT_GTE0 0x0020
ac27a0ec 1326#ifdef CONFIG_QUOTA
26092bf5
TT		 1327#define MOPT_Q 0
1328#define MOPT_QFMT 0x0040
1329#else
1330#define MOPT_Q MOPT_NOSUPPORT
1331#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1332#endif
26092bf5 1333#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1334#define MOPT_NO_EXT2 0x0100
1335#define MOPT_NO_EXT3 0x0200
1336#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
26092bf5
TT		 1337
1338static const struct mount_opts {
1339 int token;
1340 int mount_opt;
1341 int flags;
1342} ext4_mount_opts[] = {
1343 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1344 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1345 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1346 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1347 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1348 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1349 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1350 MOPT_EXT4_ONLY | MOPT_SET},
1351 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1352 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1353 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1354 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1355 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1356 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1357 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1358 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c
TT		 1359 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1360 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1361 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1362 EXT4_MOUNT_JOURNAL_CHECKSUM),
1363 MOPT_EXT4_ONLY | MOPT_SET},
1364 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1365 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1366 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1367 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1368 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1369 MOPT_NO_EXT2 | MOPT_SET},
1370 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1371 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1372 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1373 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1374 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1375 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1376 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1377 {Opt_commit, 0, MOPT_GTE0},
1378 {Opt_max_batch_time, 0, MOPT_GTE0},
1379 {Opt_min_batch_time, 0, MOPT_GTE0},
1380 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1381 {Opt_init_itable, 0, MOPT_GTE0},
1382 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1383 {Opt_resuid, 0, MOPT_GTE0},
1384 {Opt_resgid, 0, MOPT_GTE0},
1385 {Opt_journal_dev, 0, MOPT_GTE0},
1386 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1387 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1388 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1389 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1390 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1391 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1392 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1393#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1394 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1395 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1396#else
26092bf5
TT		 1397 {Opt_acl, 0, MOPT_NOSUPPORT},
1398 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1399#endif
26092bf5
TT		 1400 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1401 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1402 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1403 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1404 MOPT_SET | MOPT_Q},
1405 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1406 MOPT_SET | MOPT_Q},
1407 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1408 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1409 {Opt_usrjquota, 0, MOPT_Q},
1410 {Opt_grpjquota, 0, MOPT_Q},
1411 {Opt_offusrjquota, 0, MOPT_Q},
1412 {Opt_offgrpjquota, 0, MOPT_Q},
1413 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1414 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1415 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1416 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1417 {Opt_err, 0, 0}
1418};
1419
1420static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1421 substring_t *args, unsigned long *journal_devnum,
1422 unsigned int *journal_ioprio, int is_remount)
1423{
1424 struct ext4_sb_info *sbi = EXT4_SB(sb);
1425 const struct mount_opts *m;
08cefc7a
EB		 1426 kuid_t uid;
1427 kgid_t gid;
26092bf5
TT		 1428 int arg = 0;
1429
57f73c2c
TT		 1430#ifdef CONFIG_QUOTA
1431 if (token == Opt_usrjquota)
1432 return set_qf_name(sb, USRQUOTA, &args[0]);
1433 else if (token == Opt_grpjquota)
1434 return set_qf_name(sb, GRPQUOTA, &args[0]);
1435 else if (token == Opt_offusrjquota)
1436 return clear_qf_name(sb, USRQUOTA);
1437 else if (token == Opt_offgrpjquota)
1438 return clear_qf_name(sb, GRPQUOTA);
1439#endif
26092bf5 1440 switch (token) {
f7048605
TT		 1441 case Opt_noacl:
1442 case Opt_nouser_xattr:
1443 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1444 break;
26092bf5
TT		 1445 case Opt_sb:
1446 return 1; /* handled by get_sb_block() */
1447 case Opt_removed:
5f3633e3 1448 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1449 return 1;
26092bf5
TT		 1450 case Opt_abort:
1451 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1452 return 1;
1453 case Opt_i_version:
1454 sb->s_flags |= MS_I_VERSION;
1455 return 1;
26092bf5
TT		 1456 }
1457
5f3633e3
JK		 1458 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1459 if (token == m->token)
1460 break;
1461
1462 if (m->token == Opt_err) {
1463 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1464 "or missing value", opt);
1465 return -1;
1466 }
1467
8dc0aa8c
TT		 1468 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1469 ext4_msg(sb, KERN_ERR,
1470 "Mount option \"%s\" incompatible with ext2", opt);
1471 return -1;
1472 }
1473 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1474 ext4_msg(sb, KERN_ERR,
1475 "Mount option \"%s\" incompatible with ext3", opt);
1476 return -1;
1477 }
1478
5f3633e3
JK		 1479 if (args->from && match_int(args, &arg))
1480 return -1;
1481 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1482 return -1;
1483 if (m->flags & MOPT_EXPLICIT)
1484 set_opt2(sb, EXPLICIT_DELALLOC);
1485 if (m->flags & MOPT_CLEAR_ERR)
1486 clear_opt(sb, ERRORS_MASK);
1487 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1488 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1489 "options when quota turned on");
1490 return -1;
1491 }
1492
1493 if (m->flags & MOPT_NOSUPPORT) {
1494 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1495 } else if (token == Opt_commit) {
1496 if (arg == 0)
1497 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1498 sbi->s_commit_interval = HZ * arg;
1499 } else if (token == Opt_max_batch_time) {
1500 if (arg == 0)
1501 arg = EXT4_DEF_MAX_BATCH_TIME;
1502 sbi->s_max_batch_time = arg;
1503 } else if (token == Opt_min_batch_time) {
1504 sbi->s_min_batch_time = arg;
1505 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1506 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1507 ext4_msg(sb, KERN_ERR,
1508 "EXT4-fs: inode_readahead_blks must be "
1509 "0 or a power of 2 smaller than 2^31");
26092bf5 1510 return -1;
5f3633e3
JK		 1511 }
1512 sbi->s_inode_readahead_blks = arg;
1513 } else if (token == Opt_init_itable) {
1514 set_opt(sb, INIT_INODE_TABLE);
1515 if (!args->from)
1516 arg = EXT4_DEF_LI_WAIT_MULT;
1517 sbi->s_li_wait_mult = arg;
1518 } else if (token == Opt_max_dir_size_kb) {
1519 sbi->s_max_dir_size_kb = arg;
1520 } else if (token == Opt_stripe) {
1521 sbi->s_stripe = arg;
1522 } else if (token == Opt_resuid) {
1523 uid = make_kuid(current_user_ns(), arg);
1524 if (!uid_valid(uid)) {
1525 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1526 return -1;
1527 }
5f3633e3
JK		 1528 sbi->s_resuid = uid;
1529 } else if (token == Opt_resgid) {
1530 gid = make_kgid(current_user_ns(), arg);
1531 if (!gid_valid(gid)) {
1532 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1533 return -1;
1534 }
1535 sbi->s_resgid = gid;
1536 } else if (token == Opt_journal_dev) {
1537 if (is_remount) {
1538 ext4_msg(sb, KERN_ERR,
1539 "Cannot specify journal on remount");
1540 return -1;
1541 }
1542 *journal_devnum = arg;
1543 } else if (token == Opt_journal_ioprio) {
1544 if (arg > 7) {
1545 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1546 " (must be 0-7)");
1547 return -1;
1548 }
1549 *journal_ioprio =
1550 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1551 } else if (m->flags & MOPT_DATAJ) {
1552 if (is_remount) {
1553 if (!sbi->s_journal)
1554 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1555 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1556 ext4_msg(sb, KERN_ERR,
26092bf5 1557 "Cannot change data mode on remount");
26092bf5 1558 return -1;
ac27a0ec 1559 }
26092bf5 1560 } else {
5f3633e3
JK		 1561 clear_opt(sb, DATA_FLAGS);
1562 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1563 }
5f3633e3
JK		 1564#ifdef CONFIG_QUOTA
1565 } else if (m->flags & MOPT_QFMT) {
1566 if (sb_any_quota_loaded(sb) &&
1567 sbi->s_jquota_fmt != m->mount_opt) {
1568 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1569 "quota options when quota turned on");
1570 return -1;
1571 }
262b4662
JK		 1572 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1573 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1574 ext4_msg(sb, KERN_ERR,
1575 "Cannot set journaled quota options "
1576 "when QUOTA feature is enabled");
1577 return -1;
1578 }
5f3633e3
JK		 1579 sbi->s_jquota_fmt = m->mount_opt;
1580#endif
1581 } else {
1582 if (!args->from)
1583 arg = 1;
1584 if (m->flags & MOPT_CLEAR)
1585 arg = !arg;
1586 else if (unlikely(!(m->flags & MOPT_SET))) {
1587 ext4_msg(sb, KERN_WARNING,
1588 "buggy handling of option %s", opt);
1589 WARN_ON(1);
1590 return -1;
1591 }
1592 if (arg != 0)
1593 sbi->s_mount_opt |= m->mount_opt;
1594 else
1595 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1596 }
5f3633e3 1597 return 1;
26092bf5
TT		 1598}
1599
1600static int parse_options(char *options, struct super_block *sb,
1601 unsigned long *journal_devnum,
1602 unsigned int *journal_ioprio,
1603 int is_remount)
1604{
1605 struct ext4_sb_info *sbi = EXT4_SB(sb);
1606 char *p;
1607 substring_t args[MAX_OPT_ARGS];
1608 int token;
1609
1610 if (!options)
1611 return 1;
1612
1613 while ((p = strsep(&options, ",")) != NULL) {
1614 if (!*p)
1615 continue;
1616 /*
1617 * Initialize args struct so we know whether arg was
1618 * found; some options take optional arguments.
1619 */
caecd0af 1620 args[0].to = args[0].from = NULL;
26092bf5
TT		 1621 token = match_token(p, tokens, args);
1622 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1623 journal_ioprio, is_remount) < 0)
1624 return 0;
ac27a0ec
DK		 1625 }
1626#ifdef CONFIG_QUOTA
262b4662
JK		 1627 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1628 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1629 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1630 "feature is enabled");
1631 return 0;
1632 }
ac27a0ec 1633 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1634 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1635 clear_opt(sb, USRQUOTA);
ac27a0ec 1636
482a7425 1637 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1638 clear_opt(sb, GRPQUOTA);
ac27a0ec 1639
56c50f11 1640 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1641 ext4_msg(sb, KERN_ERR, "old and new quota "
1642 "format mixing");
ac27a0ec
DK		 1643 return 0;
1644 }
1645
1646 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1647 ext4_msg(sb, KERN_ERR, "journaled quota format "
1648 "not specified");
ac27a0ec
DK		 1649 return 0;
1650 }
1651 } else {
1652 if (sbi->s_jquota_fmt) {
b31e1552 1653 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1654 "specified with no journaling "
b31e1552 1655 "enabled");
ac27a0ec
DK		 1656 return 0;
1657 }
1658 }
1659#endif
261cb20c
JK		 1660 if (test_opt(sb, DIOREAD_NOLOCK)) {
1661 int blocksize =
1662 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1663
1664 if (blocksize < PAGE_CACHE_SIZE) {
1665 ext4_msg(sb, KERN_ERR, "can't mount with "
1666 "dioread_nolock if block size != PAGE_SIZE");
1667 return 0;
1668 }
1669 }
ac27a0ec
DK		 1670 return 1;
1671}
1672
2adf6da8
TT		 1673static inline void ext4_show_quota_options(struct seq_file *seq,
1674 struct super_block *sb)
1675{
1676#if defined(CONFIG_QUOTA)
1677 struct ext4_sb_info *sbi = EXT4_SB(sb);
1678
1679 if (sbi->s_jquota_fmt) {
1680 char *fmtname = "";
1681
1682 switch (sbi->s_jquota_fmt) {
1683 case QFMT_VFS_OLD:
1684 fmtname = "vfsold";
1685 break;
1686 case QFMT_VFS_V0:
1687 fmtname = "vfsv0";
1688 break;
1689 case QFMT_VFS_V1:
1690 fmtname = "vfsv1";
1691 break;
1692 }
1693 seq_printf(seq, ",jqfmt=%s", fmtname);
1694 }
1695
1696 if (sbi->s_qf_names[USRQUOTA])
1697 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1698
1699 if (sbi->s_qf_names[GRPQUOTA])
1700 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1701#endif
1702}
1703
5a916be1
TT		 1704static const char *token2str(int token)
1705{
50df9fd5 1706 const struct match_token *t;
5a916be1
TT		 1707
1708 for (t = tokens; t->token != Opt_err; t++)
1709 if (t->token == token && !strchr(t->pattern, '='))
1710 break;
1711 return t->pattern;
1712}
1713
2adf6da8
TT		 1714/*
1715 * Show an option if
1716 * - it's set to a non-default value OR
1717 * - if the per-sb default is different from the global default
1718 */
66acdcf4
TT		 1719static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1720 int nodefs)
2adf6da8 1721{
2adf6da8
TT		 1722 struct ext4_sb_info *sbi = EXT4_SB(sb);
1723 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1724 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1725 const struct mount_opts *m;
66acdcf4 1726 char sep = nodefs ? '\n' : ',';
2adf6da8 1727
66acdcf4
TT		 1728#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1729#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1730
1731 if (sbi->s_sb_block != 1)
5a916be1
TT		 1732 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1733
1734 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1735 int want_set = m->flags & MOPT_SET;
1736 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1737 (m->flags & MOPT_CLEAR_ERR))
1738 continue;
66acdcf4 1739 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1740 continue; /* skip if same as the default */
1741 if ((want_set &&
1742 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1743 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1744 continue; /* select Opt_noFoo vs Opt_Foo */
1745 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1746 }
5a916be1 1747
08cefc7a 1748 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1749 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1750 SEQ_OPTS_PRINT("resuid=%u",
1751 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1752 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1753 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1754 SEQ_OPTS_PRINT("resgid=%u",
1755 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1756 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1757 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1758 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1759 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1760 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1761 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1762 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1763 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1764 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1765 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1766 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1767 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1768 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1769 if (sb->s_flags & MS_I_VERSION)
5a916be1 1770 SEQ_OPTS_PUTS("i_version");
66acdcf4 1771 if (nodefs || sbi->s_stripe)
5a916be1 1772 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1773 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1774 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1775 SEQ_OPTS_PUTS("data=journal");
1776 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1777 SEQ_OPTS_PUTS("data=ordered");
1778 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1779 SEQ_OPTS_PUTS("data=writeback");
1780 }
66acdcf4
TT		 1781 if (nodefs ||
1782 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1783 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1784 sbi->s_inode_readahead_blks);
2adf6da8 1785
66acdcf4
TT		 1786 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1787 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1788 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1789 if (nodefs || sbi->s_max_dir_size_kb)
1790 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1791
1792 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1793 return 0;
1794}
1795
66acdcf4
TT		 1796static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1797{
1798 return _ext4_show_options(seq, root->d_sb, 0);
1799}
1800
1801static int options_seq_show(struct seq_file *seq, void *offset)
1802{
1803 struct super_block *sb = seq->private;
1804 int rc;
1805
1806 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1807 rc = _ext4_show_options(seq, sb, 1);
1808 seq_puts(seq, "\n");
1809 return rc;
1810}
1811
1812static int options_open_fs(struct inode *inode, struct file *file)
1813{
d9dda78b 1814 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1815}
1816
1817static const struct file_operations ext4_seq_options_fops = {
1818 .owner = THIS_MODULE,
1819 .open = options_open_fs,
1820 .read = seq_read,
1821 .llseek = seq_lseek,
1822 .release = single_release,
1823};
1824
617ba13b 1825static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1826 int read_only)
1827{
617ba13b 1828 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1829 int res = 0;
1830
617ba13b 1831 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1832 ext4_msg(sb, KERN_ERR, "revision level too high, "
1833 "forcing read-only mode");
ac27a0ec
DK		 1834 res = MS_RDONLY;
1835 }
1836 if (read_only)
281b5995 1837 goto done;
617ba13b 1838 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1839 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1840 "running e2fsck is recommended");
617ba13b 1841 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1842 ext4_msg(sb, KERN_WARNING,
1843 "warning: mounting fs with errors, "
1844 "running e2fsck is recommended");
ed3ce80a 1845 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1846 le16_to_cpu(es->s_mnt_count) >=
1847 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1848 ext4_msg(sb, KERN_WARNING,
1849 "warning: maximal mount count reached, "
1850 "running e2fsck is recommended");
ac27a0ec
DK		 1851 else if (le32_to_cpu(es->s_checkinterval) &&
1852 (le32_to_cpu(es->s_lastcheck) +
1853 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1854 ext4_msg(sb, KERN_WARNING,
1855 "warning: checktime reached, "
1856 "running e2fsck is recommended");
0b8e58a1 1857 if (!sbi->s_journal)
0390131b 1858 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1859 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1860 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1861 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1862 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1863 ext4_update_dynamic_rev(sb);
0390131b
FM		 1864 if (sbi->s_journal)
1865 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1866
e2d67052 1867 ext4_commit_super(sb, 1);
281b5995 1868done:
ac27a0ec 1869 if (test_opt(sb, DEBUG))
a9df9a49 1870 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1871 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1872 sb->s_blocksize,
1873 sbi->s_groups_count,
617ba13b
MC		 1874 EXT4_BLOCKS_PER_GROUP(sb),
1875 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1876 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1877
7abc52c2 1878 cleancache_init_fs(sb);
ac27a0ec
DK		 1879 return res;
1880}
1881
117fff10
TT		 1882int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1883{
1884 struct ext4_sb_info *sbi = EXT4_SB(sb);
1885 struct flex_groups *new_groups;
1886 int size;
1887
1888 if (!sbi->s_log_groups_per_flex)
1889 return 0;
1890
1891 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1892 if (size <= sbi->s_flex_groups_allocated)
1893 return 0;
1894
1895 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1896 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1897 if (!new_groups) {
1898 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1899 size / (int) sizeof(struct flex_groups));
1900 return -ENOMEM;
1901 }
1902
1903 if (sbi->s_flex_groups) {
1904 memcpy(new_groups, sbi->s_flex_groups,
1905 (sbi->s_flex_groups_allocated *
1906 sizeof(struct flex_groups)));
1907 ext4_kvfree(sbi->s_flex_groups);
1908 }
1909 sbi->s_flex_groups = new_groups;
1910 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1911 return 0;
1912}
1913
772cb7c8
JS		 1914static int ext4_fill_flex_info(struct super_block *sb)
1915{
1916 struct ext4_sb_info *sbi = EXT4_SB(sb);
1917 struct ext4_group_desc *gdp = NULL;
772cb7c8 1918 ext4_group_t flex_group;
117fff10 1919 int i, err;
772cb7c8 1920
503358ae 1921 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1922 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1923 sbi->s_log_groups_per_flex = 0;
1924 return 1;
1925 }
1926
117fff10
TT		 1927 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1928 if (err)
9933fc0a 1929 goto failed;
772cb7c8 1930
772cb7c8 1931 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1932 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1933
1934 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1935 atomic_add(ext4_free_inodes_count(sb, gdp),
1936 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 1937 atomic64_add(ext4_free_group_clusters(sb, gdp),
1938 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1939 atomic_add(ext4_used_dirs_count(sb, gdp),
1940 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1941 }
1942
1943 return 1;
1944failed:
1945 return 0;
1946}
1947
feb0ab32
DW		 1948static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1949 struct ext4_group_desc *gdp)
717d50e4 1950{
feb0ab32 1951 int offset;
717d50e4 1952 __u16 crc = 0;
feb0ab32 1953 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1954
feb0ab32
DW		 1955 if ((sbi->s_es->s_feature_ro_compat &
1956 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1957 /* Use new metadata_csum algorithm */
d6a77105 1958 __le16 save_csum;
feb0ab32
DW		 1959 __u32 csum32;
1960
d6a77105 1961 save_csum = gdp->bg_checksum;
feb0ab32
DW		 1962 gdp->bg_checksum = 0;
1963 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1964 sizeof(le_group));
1965 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1966 sbi->s_desc_size);
d6a77105 1967 gdp->bg_checksum = save_csum;
feb0ab32
DW		 1968
1969 crc = csum32 & 0xFFFF;
1970 goto out;
717d50e4
AD		 1971 }
1972
feb0ab32
DW		 1973 /* old crc16 code */
1974 offset = offsetof(struct ext4_group_desc, bg_checksum);
1975
1976 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1977 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1978 crc = crc16(crc, (__u8 *)gdp, offset);
1979 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1980 /* for checksum of struct ext4_group_desc do the rest...*/
1981 if ((sbi->s_es->s_feature_incompat &
1982 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1983 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1984 crc = crc16(crc, (__u8 *)gdp + offset,
1985 le16_to_cpu(sbi->s_es->s_desc_size) -
1986 offset);
1987
1988out:
717d50e4
AD		 1989 return cpu_to_le16(crc);
1990}
1991
feb0ab32 1992int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 1993 struct ext4_group_desc *gdp)
1994{
feb0ab32
DW		 1995 if (ext4_has_group_desc_csum(sb) &&
1996 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1997 block_group, gdp)))
717d50e4
AD		 1998 return 0;
1999
2000 return 1;
2001}
2002
feb0ab32
DW		 2003void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2004 struct ext4_group_desc *gdp)
2005{
2006 if (!ext4_has_group_desc_csum(sb))
2007 return;
2008 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2009}
2010
ac27a0ec 2011/* Called at mount-time, super-block is locked */
bfff6873
LC		 2012static int ext4_check_descriptors(struct super_block *sb,
2013 ext4_group_t *first_not_zeroed)
ac27a0ec 2014{
617ba13b
MC		 2015 struct ext4_sb_info *sbi = EXT4_SB(sb);
2016 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2017 ext4_fsblk_t last_block;
bd81d8ee
LV		 2018 ext4_fsblk_t block_bitmap;
2019 ext4_fsblk_t inode_bitmap;
2020 ext4_fsblk_t inode_table;
ce421581 2021 int flexbg_flag = 0;
bfff6873 2022 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2023
ce421581
JS		 2024 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2025 flexbg_flag = 1;
2026
af5bc92d 2027 ext4_debug("Checking group descriptors");
ac27a0ec 2028
197cd65a
AM		 2029 for (i = 0; i < sbi->s_groups_count; i++) {
2030 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2031
ce421581 2032 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2033 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2034 else
2035 last_block = first_block +
617ba13b 2036 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2037
bfff6873
LC		 2038 if ((grp == sbi->s_groups_count) &&
2039 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2040 grp = i;
2041
8fadc143 2042 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2043 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2044 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2045 "Block bitmap for group %u not in group "
b31e1552 2046 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2047 return 0;
2048 }
8fadc143 2049 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2050 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2051 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2052 "Inode bitmap for group %u not in group "
b31e1552 2053 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2054 return 0;
2055 }
8fadc143 2056 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2057 if (inode_table < first_block ||
2b2d6d01 2058 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2059 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2060 "Inode table for group %u not in group "
b31e1552 2061 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2062 return 0;
2063 }
955ce5f5 2064 ext4_lock_group(sb, i);
feb0ab32 2065 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2066 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2067 "Checksum for group %u failed (%u!=%u)",
2068 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2069 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2070 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2071 ext4_unlock_group(sb, i);
8a266467 2072 return 0;
7ee1ec4c 2073 }
717d50e4 2074 }
955ce5f5 2075 ext4_unlock_group(sb, i);
ce421581
JS		 2076 if (!flexbg_flag)
2077 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2078 }
bfff6873
LC		 2079 if (NULL != first_not_zeroed)
2080 *first_not_zeroed = grp;
ac27a0ec 2081
5dee5437
TT		 2082 ext4_free_blocks_count_set(sbi->s_es,
2083 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2084 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2085 return 1;
2086}
2087
617ba13b 2088/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2089 * the superblock) which were deleted from all directories, but held open by
2090 * a process at the time of a crash. We walk the list and try to delete these
2091 * inodes at recovery time (only with a read-write filesystem).
2092 *
2093 * In order to keep the orphan inode chain consistent during traversal (in
2094 * case of crash during recovery), we link each inode into the superblock
2095 * orphan list_head and handle it the same way as an inode deletion during
2096 * normal operation (which journals the operations for us).
2097 *
2098 * We only do an iget() and an iput() on each inode, which is very safe if we
2099 * accidentally point at an in-use or already deleted inode. The worst that
2100 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2101 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2102 * e2fsck was run on this filesystem, and it must have already done the orphan
2103 * inode cleanup for us, so we can safely abort without any further action.
2104 */
2b2d6d01
TT		 2105static void ext4_orphan_cleanup(struct super_block *sb,
2106 struct ext4_super_block *es)
ac27a0ec
DK		 2107{
2108 unsigned int s_flags = sb->s_flags;
2109 int nr_orphans = 0, nr_truncates = 0;
2110#ifdef CONFIG_QUOTA
2111 int i;
2112#endif
2113 if (!es->s_last_orphan) {
2114 jbd_debug(4, "no orphan inodes to clean up\n");
2115 return;
2116 }
2117
a8f48a95 2118 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2119 ext4_msg(sb, KERN_ERR, "write access "
2120 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2121 return;
2122 }
2123
d39195c3
AG		 2124 /* Check if feature set would not allow a r/w mount */
2125 if (!ext4_feature_set_ok(sb, 0)) {
2126 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2127 "unknown ROCOMPAT features");
2128 return;
2129 }
2130
617ba13b 2131 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2132 /* don't clear list on RO mount w/ errors */
2133 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2134 jbd_debug(1, "Errors on filesystem, "
2135 "clearing orphan list.\n");
c25f9bc6
ES		 2136 es->s_last_orphan = 0;
2137 }
ac27a0ec
DK		 2138 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2139 return;
2140 }
2141
2142 if (s_flags & MS_RDONLY) {
b31e1552 2143 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2144 sb->s_flags &= ~MS_RDONLY;
2145 }
2146#ifdef CONFIG_QUOTA
2147 /* Needed for iput() to work correctly and not trash data */
2148 sb->s_flags |= MS_ACTIVE;
2149 /* Turn on quotas so that they are updated correctly */
2150 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2151 if (EXT4_SB(sb)->s_qf_names[i]) {
2152 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2153 if (ret < 0)
b31e1552
ES		 2154 ext4_msg(sb, KERN_ERR,
2155 "Cannot turn on journaled "
2156 "quota: error %d", ret);
ac27a0ec
DK		 2157 }
2158 }
2159#endif
2160
2161 while (es->s_last_orphan) {
2162 struct inode *inode;
2163
97bd42b9
JB		 2164 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2165 if (IS_ERR(inode)) {
ac27a0ec
DK		 2166 es->s_last_orphan = 0;
2167 break;
2168 }
2169
617ba13b 2170 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2171 dquot_initialize(inode);
ac27a0ec 2172 if (inode->i_nlink) {
566370a2
PT		 2173 if (test_opt(sb, DEBUG))
2174 ext4_msg(sb, KERN_DEBUG,
2175 "%s: truncating inode %lu to %lld bytes",
2176 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2177 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2178 inode->i_ino, inode->i_size);
721e3eba 2179 mutex_lock(&inode->i_mutex);
55f252c9 2180 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2181 ext4_truncate(inode);
721e3eba 2182 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2183 nr_truncates++;
2184 } else {
566370a2
PT		 2185 if (test_opt(sb, DEBUG))
2186 ext4_msg(sb, KERN_DEBUG,
2187 "%s: deleting unreferenced inode %lu",
2188 __func__, inode->i_ino);
ac27a0ec
DK		 2189 jbd_debug(2, "deleting unreferenced inode %lu\n",
2190 inode->i_ino);
2191 nr_orphans++;
2192 }
2193 iput(inode); /* The delete magic happens here! */
2194 }
2195
2b2d6d01 2196#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2197
2198 if (nr_orphans)
b31e1552
ES		 2199 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2200 PLURAL(nr_orphans));
ac27a0ec 2201 if (nr_truncates)
b31e1552
ES		 2202 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2203 PLURAL(nr_truncates));
ac27a0ec
DK		 2204#ifdef CONFIG_QUOTA
2205 /* Turn quotas off */
2206 for (i = 0; i < MAXQUOTAS; i++) {
2207 if (sb_dqopt(sb)->files[i])
287a8095 2208 dquot_quota_off(sb, i);
ac27a0ec
DK		 2209 }
2210#endif
2211 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2212}
0b8e58a1 2213
cd2291a4
ES		 2214/*
2215 * Maximal extent format file size.
2216 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2217 * extent format containers, within a sector_t, and within i_blocks
2218 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2219 * so that won't be a limiting factor.
2220 *
f17722f9
LC		 2221 * However there is other limiting factor. We do store extents in the form
2222 * of starting block and length, hence the resulting length of the extent
2223 * covering maximum file size must fit into on-disk format containers as
2224 * well. Given that length is always by 1 unit bigger than max unit (because
2225 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2226 *
cd2291a4
ES		 2227 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2228 */
f287a1a5 2229static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2230{
2231 loff_t res;
2232 loff_t upper_limit = MAX_LFS_FILESIZE;
2233
2234 /* small i_blocks in vfs inode? */
f287a1a5 2235 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2236 /*
90c699a9 2237 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2238 * i_block represent total blocks in 512 bytes
2239 * 32 == size of vfs inode i_blocks * 8
2240 */
2241 upper_limit = (1LL << 32) - 1;
2242
2243 /* total blocks in file system block size */
2244 upper_limit >>= (blkbits - 9);
2245 upper_limit <<= blkbits;
2246 }
2247
f17722f9
LC		 2248 /*
2249 * 32-bit extent-start container, ee_block. We lower the maxbytes
2250 * by one fs block, so ee_len can cover the extent of maximum file
2251 * size
2252 */
2253 res = (1LL << 32) - 1;
cd2291a4 2254 res <<= blkbits;
cd2291a4
ES		 2255
2256 /* Sanity check against vm- & vfs- imposed limits */
2257 if (res > upper_limit)
2258 res = upper_limit;
2259
2260 return res;
2261}
ac27a0ec 2262
ac27a0ec 2263/*
cd2291a4 2264 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2265 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2266 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2267 */
f287a1a5 2268static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2269{
617ba13b 2270 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2271 int meta_blocks;
2272 loff_t upper_limit;
0b8e58a1
AD		 2273 /* This is calculated to be the largest file size for a dense, block
2274 * mapped file such that the file's total number of 512-byte sectors,
2275 * including data and all indirect blocks, does not exceed (2^48 - 1).
2276 *
2277 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2278 * number of 512-byte sectors of the file.
0fc1b451
AK		 2279 */
2280
f287a1a5 2281 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2282 /*
90c699a9 2283 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2284 * the inode i_block field represents total file blocks in
2285 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2286 */
2287 upper_limit = (1LL << 32) - 1;
2288
2289 /* total blocks in file system block size */
2290 upper_limit >>= (bits - 9);
2291
2292 } else {
8180a562
AK		 2293 /*
2294 * We use 48 bit ext4_inode i_blocks
2295 * With EXT4_HUGE_FILE_FL set the i_blocks
2296 * represent total number of blocks in
2297 * file system block size
2298 */
0fc1b451
AK		 2299 upper_limit = (1LL << 48) - 1;
2300
0fc1b451
AK		 2301 }
2302
2303 /* indirect blocks */
2304 meta_blocks = 1;
2305 /* double indirect blocks */
2306 meta_blocks += 1 + (1LL << (bits-2));
2307 /* tripple indirect blocks */
2308 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2309
2310 upper_limit -= meta_blocks;
2311 upper_limit <<= bits;
ac27a0ec
DK		 2312
2313 res += 1LL << (bits-2);
2314 res += 1LL << (2*(bits-2));
2315 res += 1LL << (3*(bits-2));
2316 res <<= bits;
2317 if (res > upper_limit)
2318 res = upper_limit;
0fc1b451
AK		 2319
2320 if (res > MAX_LFS_FILESIZE)
2321 res = MAX_LFS_FILESIZE;
2322
ac27a0ec
DK		 2323 return res;
2324}
2325
617ba13b 2326static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2327 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2328{
617ba13b 2329 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2330 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2331 int has_super = 0;
2332
2333 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2334
617ba13b 2335 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2336 nr < first_meta_bg)
70bbb3e0 2337 return logical_sb_block + nr + 1;
ac27a0ec 2338 bg = sbi->s_desc_per_block * nr;
617ba13b 2339 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2340 has_super = 1;
0b8e58a1 2341
617ba13b 2342 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2343}
2344
c9de560d
AT		 2345/**
2346 * ext4_get_stripe_size: Get the stripe size.
2347 * @sbi: In memory super block info
2348 *
2349 * If we have specified it via mount option, then
2350 * use the mount option value. If the value specified at mount time is
2351 * greater than the blocks per group use the super block value.
2352 * If the super block value is greater than blocks per group return 0.
2353 * Allocator needs it be less than blocks per group.
2354 *
2355 */
2356static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2357{
2358 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2359 unsigned long stripe_width =
2360 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2361 int ret;
c9de560d
AT		 2362
2363 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2364 ret = sbi->s_stripe;
2365 else if (stripe_width <= sbi->s_blocks_per_group)
2366 ret = stripe_width;
2367 else if (stride <= sbi->s_blocks_per_group)
2368 ret = stride;
2369 else
2370 ret = 0;
c9de560d 2371
3eb08658
DE		 2372 /*
2373 * If the stripe width is 1, this makes no sense and
2374 * we set it to 0 to turn off stripe handling code.
2375 */
2376 if (ret <= 1)
2377 ret = 0;
c9de560d 2378
3eb08658 2379 return ret;
c9de560d 2380}
ac27a0ec 2381
3197ebdb
TT		 2382/* sysfs supprt */
2383
2384struct ext4_attr {
2385 struct attribute attr;
2386 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2387 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb 2388 const char *, size_t);
f2d50a65
JK		 2389 union {
2390 int offset;
2391 int deprecated_val;
2392 } u;
3197ebdb
TT		 2393};
2394
27dd4385
LC		 2395static int parse_strtoull(const char *buf,
2396 unsigned long long max, unsigned long long *value)
2397{
2398 int ret;
2399
2400 ret = kstrtoull(skip_spaces(buf), 0, value);
2401 if (!ret && *value > max)
2402 ret = -EINVAL;
2403 return ret;
2404}
2405
3197ebdb
TT		 2406static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2407 struct ext4_sb_info *sbi,
2408 char *buf)
2409{
2410 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2411 (s64) EXT4_C2B(sbi,
2412 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2413}
2414
2415static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2416 struct ext4_sb_info *sbi, char *buf)
2417{
2418 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2419
f613dfcb
TT		 2420 if (!sb->s_bdev->bd_part)
2421 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2422 return snprintf(buf, PAGE_SIZE, "%lu\n",
2423 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2424 sbi->s_sectors_written_start) >> 1);
2425}
2426
2427static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2428 struct ext4_sb_info *sbi, char *buf)
2429{
2430 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2431
f613dfcb
TT		 2432 if (!sb->s_bdev->bd_part)
2433 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2434 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2435 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2436 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2437 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2438}
2439
2440static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2441 struct ext4_sb_info *sbi,
2442 const char *buf, size_t count)
2443{
2444 unsigned long t;
e1091b15 2445 int ret;
3197ebdb 2446
e1091b15
LC		 2447 ret = kstrtoul(skip_spaces(buf), 0, &t);
2448 if (ret)
2449 return ret;
3197ebdb 2450
e1091b15 2451 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2452 return -EINVAL;
2453
2454 sbi->s_inode_readahead_blks = t;
2455 return count;
2456}
2457
2458static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2459 struct ext4_sb_info *sbi, char *buf)
3197ebdb 2460{
f2d50a65 2461 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb
TT		 2462
2463 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2464}
2465
2466static ssize_t sbi_ui_store(struct ext4_attr *a,
2467 struct ext4_sb_info *sbi,
2468 const char *buf, size_t count)
2469{
f2d50a65 2470 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb 2471 unsigned long t;
e1091b15 2472 int ret;
3197ebdb 2473
e1091b15
LC		 2474 ret = kstrtoul(skip_spaces(buf), 0, &t);
2475 if (ret)
2476 return ret;
3197ebdb
TT		 2477 *ui = t;
2478 return count;
2479}
2480
27dd4385
LC		 2481static ssize_t reserved_clusters_show(struct ext4_attr *a,
2482 struct ext4_sb_info *sbi, char *buf)
2483{
2484 return snprintf(buf, PAGE_SIZE, "%llu\n",
2485 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2486}
2487
2488static ssize_t reserved_clusters_store(struct ext4_attr *a,
2489 struct ext4_sb_info *sbi,
2490 const char *buf, size_t count)
2491{
2492 unsigned long long val;
2493 int ret;
2494
2495 if (parse_strtoull(buf, -1ULL, &val))
2496 return -EINVAL;
2497 ret = ext4_reserve_clusters(sbi, val);
2498
2499 return ret ? ret : count;
2500}
2501
2c0544b2
TT		 2502static ssize_t trigger_test_error(struct ext4_attr *a,
2503 struct ext4_sb_info *sbi,
2504 const char *buf, size_t count)
2505{
2506 int len = count;
2507
2508 if (!capable(CAP_SYS_ADMIN))
2509 return -EPERM;
2510
2511 if (len && buf[len-1] == '\n')
2512 len--;
2513
2514 if (len)
2515 ext4_error(sbi->s_sb, "%.*s", len, buf);
2516 return count;
2517}
2518
f2d50a65
JK		 2519static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2520 struct ext4_sb_info *sbi, char *buf)
2521{
2522 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2523}
2524
3197ebdb
TT		 2525#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2526static struct ext4_attr ext4_attr_##_name = { \
2527 .attr = {.name = __stringify(_name), .mode = _mode }, \
2528 .show = _show, \
2529 .store = _store, \
f2d50a65
JK		 2530 .u = { \
2531 .offset = offsetof(struct ext4_sb_info, _elname),\
2532 }, \
3197ebdb
TT		 2533}
2534#define EXT4_ATTR(name, mode, show, store) \
2535static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2536
857ac889 2537#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2538#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2539#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2540#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2541 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2542#define ATTR_LIST(name) &ext4_attr_##name.attr
f2d50a65
JK		 2543#define EXT4_DEPRECATED_ATTR(_name, _val) \
2544static struct ext4_attr ext4_attr_##_name = { \
2545 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2546 .show = sbi_deprecated_show, \
2547 .u = { \
2548 .deprecated_val = _val, \
2549 }, \
2550}
3197ebdb
TT		 2551
2552EXT4_RO_ATTR(delayed_allocation_blocks);
2553EXT4_RO_ATTR(session_write_kbytes);
2554EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2555EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2556EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2557 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2558EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2559EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2560EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2561EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2562EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2563EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2564EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
f2d50a65 2565EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
67a5da56 2566EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2567EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2568
2569static struct attribute *ext4_attrs[] = {
2570 ATTR_LIST(delayed_allocation_blocks),
2571 ATTR_LIST(session_write_kbytes),
2572 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2573 ATTR_LIST(reserved_clusters),
3197ebdb 2574 ATTR_LIST(inode_readahead_blks),
11013911 2575 ATTR_LIST(inode_goal),
3197ebdb
TT		 2576 ATTR_LIST(mb_stats),
2577 ATTR_LIST(mb_max_to_scan),
2578 ATTR_LIST(mb_min_to_scan),
2579 ATTR_LIST(mb_order2_req),
2580 ATTR_LIST(mb_stream_req),
2581 ATTR_LIST(mb_group_prealloc),
55138e0b 2582 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2583 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2584 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2585 NULL,
2586};
2587
857ac889
LC		 2588/* Features this copy of ext4 supports */
2589EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2590EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2591EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2592
2593static struct attribute *ext4_feat_attrs[] = {
2594 ATTR_LIST(lazy_itable_init),
27ee40df 2595 ATTR_LIST(batched_discard),
5e7bbef1 2596 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2597 NULL,
2598};
2599
3197ebdb
TT		 2600static ssize_t ext4_attr_show(struct kobject *kobj,
2601 struct attribute *attr, char *buf)
2602{
2603 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2604 s_kobj);
2605 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2606
2607 return a->show ? a->show(a, sbi, buf) : 0;
2608}
2609
2610static ssize_t ext4_attr_store(struct kobject *kobj,
2611 struct attribute *attr,
2612 const char *buf, size_t len)
2613{
2614 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2615 s_kobj);
2616 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2617
2618 return a->store ? a->store(a, sbi, buf, len) : 0;
2619}
2620
2621static void ext4_sb_release(struct kobject *kobj)
2622{
2623 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2624 s_kobj);
2625 complete(&sbi->s_kobj_unregister);
2626}
2627
52cf25d0 2628static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2629 .show = ext4_attr_show,
2630 .store = ext4_attr_store,
2631};
2632
2633static struct kobj_type ext4_ktype = {
2634 .default_attrs = ext4_attrs,
2635 .sysfs_ops = &ext4_attr_ops,
2636 .release = ext4_sb_release,
2637};
2638
857ac889
LC		 2639static void ext4_feat_release(struct kobject *kobj)
2640{
2641 complete(&ext4_feat->f_kobj_unregister);
2642}
2643
2644static struct kobj_type ext4_feat_ktype = {
2645 .default_attrs = ext4_feat_attrs,
2646 .sysfs_ops = &ext4_attr_ops,
2647 .release = ext4_feat_release,
2648};
2649
a13fb1a4
ES		 2650/*
2651 * Check whether this filesystem can be mounted based on
2652 * the features present and the RDONLY/RDWR mount requested.
2653 * Returns 1 if this filesystem can be mounted as requested,
2654 * 0 if it cannot be.
2655 */
2656static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2657{
2658 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2659 ext4_msg(sb, KERN_ERR,
2660 "Couldn't mount because of "
2661 "unsupported optional features (%x)",
2662 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2663 ~EXT4_FEATURE_INCOMPAT_SUPP));
2664 return 0;
2665 }
2666
2667 if (readonly)
2668 return 1;
2669
2670 /* Check that feature set is OK for a read-write mount */
2671 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2672 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2673 "unsupported optional features (%x)",
2674 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2675 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2676 return 0;
2677 }
2678 /*
2679 * Large file size enabled file system can only be mounted
2680 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2681 */
2682 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2683 if (sizeof(blkcnt_t) < sizeof(u64)) {
2684 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2685 "cannot be mounted RDWR without "
2686 "CONFIG_LBDAF");
2687 return 0;
2688 }
2689 }
bab08ab9
TT		 2690 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2691 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2692 ext4_msg(sb, KERN_ERR,
2693 "Can't support bigalloc feature without "
2694 "extents feature\n");
2695 return 0;
2696 }
7c319d32
AK		 2697
2698#ifndef CONFIG_QUOTA
2699 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2700 !readonly) {
2701 ext4_msg(sb, KERN_ERR,
2702 "Filesystem with quota feature cannot be mounted RDWR "
2703 "without CONFIG_QUOTA");
2704 return 0;
2705 }
2706#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2707 return 1;
2708}
2709
66e61a9e
TT		 2710/*
2711 * This function is called once a day if we have errors logged
2712 * on the file system
2713 */
2714static void print_daily_error_info(unsigned long arg)
2715{
2716 struct super_block *sb = (struct super_block *) arg;
2717 struct ext4_sb_info *sbi;
2718 struct ext4_super_block *es;
2719
2720 sbi = EXT4_SB(sb);
2721 es = sbi->s_es;
2722
2723 if (es->s_error_count)
2724 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2725 le32_to_cpu(es->s_error_count));
2726 if (es->s_first_error_time) {
2727 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2728 sb->s_id, le32_to_cpu(es->s_first_error_time),
2729 (int) sizeof(es->s_first_error_func),
2730 es->s_first_error_func,
2731 le32_to_cpu(es->s_first_error_line));
2732 if (es->s_first_error_ino)
2733 printk(": inode %u",
2734 le32_to_cpu(es->s_first_error_ino));
2735 if (es->s_first_error_block)
2736 printk(": block %llu", (unsigned long long)
2737 le64_to_cpu(es->s_first_error_block));
2738 printk("\n");
2739 }
2740 if (es->s_last_error_time) {
2741 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2742 sb->s_id, le32_to_cpu(es->s_last_error_time),
2743 (int) sizeof(es->s_last_error_func),
2744 es->s_last_error_func,
2745 le32_to_cpu(es->s_last_error_line));
2746 if (es->s_last_error_ino)
2747 printk(": inode %u",
2748 le32_to_cpu(es->s_last_error_ino));
2749 if (es->s_last_error_block)
2750 printk(": block %llu", (unsigned long long)
2751 le64_to_cpu(es->s_last_error_block));
2752 printk("\n");
2753 }
2754 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2755}
2756
bfff6873
LC		 2757/* Find next suitable group and run ext4_init_inode_table */
2758static int ext4_run_li_request(struct ext4_li_request *elr)
2759{
2760 struct ext4_group_desc *gdp = NULL;
2761 ext4_group_t group, ngroups;
2762 struct super_block *sb;
2763 unsigned long timeout = 0;
2764 int ret = 0;
2765
2766 sb = elr->lr_super;
2767 ngroups = EXT4_SB(sb)->s_groups_count;
2768
8e8ad8a5 2769 sb_start_write(sb);
bfff6873
LC		 2770 for (group = elr->lr_next_group; group < ngroups; group++) {
2771 gdp = ext4_get_group_desc(sb, group, NULL);
2772 if (!gdp) {
2773 ret = 1;
2774 break;
2775 }
2776
2777 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2778 break;
2779 }
2780
7f511862 2781 if (group >= ngroups)
bfff6873
LC		 2782 ret = 1;
2783
2784 if (!ret) {
2785 timeout = jiffies;
2786 ret = ext4_init_inode_table(sb, group,
2787 elr->lr_timeout ? 0 : 1);
2788 if (elr->lr_timeout == 0) {
51ce6511
LC		 2789 timeout = (jiffies - timeout) *
2790 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2791 elr->lr_timeout = timeout;
2792 }
2793 elr->lr_next_sched = jiffies + elr->lr_timeout;
2794 elr->lr_next_group = group + 1;
2795 }
8e8ad8a5 2796 sb_end_write(sb);
bfff6873
LC		 2797
2798 return ret;
2799}
2800
2801/*
2802 * Remove lr_request from the list_request and free the
4ed5c033 2803 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2804 */
2805static void ext4_remove_li_request(struct ext4_li_request *elr)
2806{
2807 struct ext4_sb_info *sbi;
2808
2809 if (!elr)
2810 return;
2811
2812 sbi = elr->lr_sbi;
2813
2814 list_del(&elr->lr_request);
2815 sbi->s_li_request = NULL;
2816 kfree(elr);
2817}
2818
2819static void ext4_unregister_li_request(struct super_block *sb)
2820{
1bb933fb
LC		 2821 mutex_lock(&ext4_li_mtx);
2822 if (!ext4_li_info) {
2823 mutex_unlock(&ext4_li_mtx);
bfff6873 2824 return;
1bb933fb 2825 }
bfff6873
LC		 2826
2827 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2828 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2829 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2830 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2831}
2832
8f1f7453
ES		 2833static struct task_struct *ext4_lazyinit_task;
2834
bfff6873
LC		 2835/*
2836 * This is the function where ext4lazyinit thread lives. It walks
2837 * through the request list searching for next scheduled filesystem.
2838 * When such a fs is found, run the lazy initialization request
2839 * (ext4_rn_li_request) and keep track of the time spend in this
2840 * function. Based on that time we compute next schedule time of
2841 * the request. When walking through the list is complete, compute
2842 * next waking time and put itself into sleep.
2843 */
2844static int ext4_lazyinit_thread(void *arg)
2845{
2846 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2847 struct list_head *pos, *n;
2848 struct ext4_li_request *elr;
4ed5c033 2849 unsigned long next_wakeup, cur;
bfff6873
LC		 2850
2851 BUG_ON(NULL == eli);
2852
bfff6873
LC		 2853cont_thread:
2854 while (true) {
2855 next_wakeup = MAX_JIFFY_OFFSET;
2856
2857 mutex_lock(&eli->li_list_mtx);
2858 if (list_empty(&eli->li_request_list)) {
2859 mutex_unlock(&eli->li_list_mtx);
2860 goto exit_thread;
2861 }
2862
2863 list_for_each_safe(pos, n, &eli->li_request_list) {
2864 elr = list_entry(pos, struct ext4_li_request,
2865 lr_request);
2866
b2c78cd0
TT		 2867 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2868 if (ext4_run_li_request(elr) != 0) {
2869 /* error, remove the lazy_init job */
2870 ext4_remove_li_request(elr);
2871 continue;
2872 }
bfff6873
LC		 2873 }
2874
2875 if (time_before(elr->lr_next_sched, next_wakeup))
2876 next_wakeup = elr->lr_next_sched;
2877 }
2878 mutex_unlock(&eli->li_list_mtx);
2879
a0acae0e 2880 try_to_freeze();
bfff6873 2881
4ed5c033
LC		 2882 cur = jiffies;
2883 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2884 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2885 cond_resched();
2886 continue;
2887 }
2888
4ed5c033
LC		 2889 schedule_timeout_interruptible(next_wakeup - cur);
2890
8f1f7453
ES		 2891 if (kthread_should_stop()) {
2892 ext4_clear_request_list();
2893 goto exit_thread;
2894 }
bfff6873
LC		 2895 }
2896
2897exit_thread:
2898 /*
2899 * It looks like the request list is empty, but we need
2900 * to check it under the li_list_mtx lock, to prevent any
2901 * additions into it, and of course we should lock ext4_li_mtx
2902 * to atomically free the list and ext4_li_info, because at
2903 * this point another ext4 filesystem could be registering
2904 * new one.
2905 */
2906 mutex_lock(&ext4_li_mtx);
2907 mutex_lock(&eli->li_list_mtx);
2908 if (!list_empty(&eli->li_request_list)) {
2909 mutex_unlock(&eli->li_list_mtx);
2910 mutex_unlock(&ext4_li_mtx);
2911 goto cont_thread;
2912 }
2913 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2914 kfree(ext4_li_info);
2915 ext4_li_info = NULL;
2916 mutex_unlock(&ext4_li_mtx);
2917
2918 return 0;
2919}
2920
2921static void ext4_clear_request_list(void)
2922{
2923 struct list_head *pos, *n;
2924 struct ext4_li_request *elr;
2925
2926 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2927 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2928 elr = list_entry(pos, struct ext4_li_request,
2929 lr_request);
2930 ext4_remove_li_request(elr);
2931 }
2932 mutex_unlock(&ext4_li_info->li_list_mtx);
2933}
2934
2935static int ext4_run_lazyinit_thread(void)
2936{
8f1f7453
ES		 2937 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2938 ext4_li_info, "ext4lazyinit");
2939 if (IS_ERR(ext4_lazyinit_task)) {
2940 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2941 ext4_clear_request_list();
bfff6873
LC		 2942 kfree(ext4_li_info);
2943 ext4_li_info = NULL;
92b97816 2944 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2945 "initialization thread\n",
2946 err);
2947 return err;
2948 }
2949 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2950 return 0;
2951}
2952
2953/*
2954 * Check whether it make sense to run itable init. thread or not.
2955 * If there is at least one uninitialized inode table, return
2956 * corresponding group number, else the loop goes through all
2957 * groups and return total number of groups.
2958 */
2959static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2960{
2961 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2962 struct ext4_group_desc *gdp = NULL;
2963
2964 for (group = 0; group < ngroups; group++) {
2965 gdp = ext4_get_group_desc(sb, group, NULL);
2966 if (!gdp)
2967 continue;
2968
2969 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2970 break;
2971 }
2972
2973 return group;
2974}
2975
2976static int ext4_li_info_new(void)
2977{
2978 struct ext4_lazy_init *eli = NULL;
2979
2980 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2981 if (!eli)
2982 return -ENOMEM;
2983
bfff6873
LC		 2984 INIT_LIST_HEAD(&eli->li_request_list);
2985 mutex_init(&eli->li_list_mtx);
2986
bfff6873
LC		 2987 eli->li_state |= EXT4_LAZYINIT_QUIT;
2988
2989 ext4_li_info = eli;
2990
2991 return 0;
2992}
2993
2994static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2995 ext4_group_t start)
2996{
2997 struct ext4_sb_info *sbi = EXT4_SB(sb);
2998 struct ext4_li_request *elr;
2999 unsigned long rnd;
3000
3001 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3002 if (!elr)
3003 return NULL;
3004
3005 elr->lr_super = sb;
3006 elr->lr_sbi = sbi;
3007 elr->lr_next_group = start;
3008
3009 /*
3010 * Randomize first schedule time of the request to
3011 * spread the inode table initialization requests
3012 * better.
3013 */
3014 get_random_bytes(&rnd, sizeof(rnd));
3015 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3016 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3017
3018 return elr;
3019}
3020
7f511862
TT		 3021int ext4_register_li_request(struct super_block *sb,
3022 ext4_group_t first_not_zeroed)
bfff6873
LC		 3023{
3024 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3025 struct ext4_li_request *elr = NULL;
bfff6873 3026 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3027 int ret = 0;
bfff6873 3028
7f511862 3029 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3030 if (sbi->s_li_request != NULL) {
3031 /*
3032 * Reset timeout so it can be computed again, because
3033 * s_li_wait_mult might have changed.
3034 */
3035 sbi->s_li_request->lr_timeout = 0;
7f511862 3036 goto out;
51ce6511 3037 }
bfff6873
LC		 3038
3039 if (first_not_zeroed == ngroups ||
3040 (sb->s_flags & MS_RDONLY) ||
55ff3840 3041 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3042 goto out;
bfff6873
LC		 3043
3044 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3045 if (!elr) {
3046 ret = -ENOMEM;
3047 goto out;
3048 }
bfff6873
LC		 3049
3050 if (NULL == ext4_li_info) {
3051 ret = ext4_li_info_new();
3052 if (ret)
3053 goto out;
3054 }
3055
3056 mutex_lock(&ext4_li_info->li_list_mtx);
3057 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3058 mutex_unlock(&ext4_li_info->li_list_mtx);
3059
3060 sbi->s_li_request = elr;
46e4690b
TM		 3061 /*
3062 * set elr to NULL here since it has been inserted to
3063 * the request_list and the removal and free of it is
3064 * handled by ext4_clear_request_list from now on.
3065 */
3066 elr = NULL;
bfff6873
LC		 3067
3068 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3069 ret = ext4_run_lazyinit_thread();
3070 if (ret)
3071 goto out;
3072 }
bfff6873 3073out:
beed5ecb
NK		 3074 mutex_unlock(&ext4_li_mtx);
3075 if (ret)
bfff6873 3076 kfree(elr);
bfff6873
LC		 3077 return ret;
3078}
3079
3080/*
3081 * We do not need to lock anything since this is called on
3082 * module unload.
3083 */
3084static void ext4_destroy_lazyinit_thread(void)
3085{
3086 /*
3087 * If thread exited earlier
3088 * there's nothing to be done.
3089 */
8f1f7453 3090 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3091 return;
3092
8f1f7453 3093 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3094}
3095
25ed6e8a
DW		 3096static int set_journal_csum_feature_set(struct super_block *sb)
3097{
3098 int ret = 1;
3099 int compat, incompat;
3100 struct ext4_sb_info *sbi = EXT4_SB(sb);
3101
3102 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3103 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3104 /* journal checksum v2 */
3105 compat = 0;
3106 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3107 } else {
3108 /* journal checksum v1 */
3109 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3110 incompat = 0;
3111 }
3112
3113 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3114 ret = jbd2_journal_set_features(sbi->s_journal,
3115 compat, 0,
3116 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3117 incompat);
3118 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3119 ret = jbd2_journal_set_features(sbi->s_journal,
3120 compat, 0,
3121 incompat);
3122 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3123 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3124 } else {
3125 jbd2_journal_clear_features(sbi->s_journal,
3126 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3127 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3128 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3129 }
3130
3131 return ret;
3132}
3133
952fc18e
TT		 3134/*
3135 * Note: calculating the overhead so we can be compatible with
3136 * historical BSD practice is quite difficult in the face of
3137 * clusters/bigalloc. This is because multiple metadata blocks from
3138 * different block group can end up in the same allocation cluster.
3139 * Calculating the exact overhead in the face of clustered allocation
3140 * requires either O(all block bitmaps) in memory or O(number of block
3141 * groups**2) in time. We will still calculate the superblock for
3142 * older file systems --- and if we come across with a bigalloc file
3143 * system with zero in s_overhead_clusters the estimate will be close to
3144 * correct especially for very large cluster sizes --- but for newer
3145 * file systems, it's better to calculate this figure once at mkfs
3146 * time, and store it in the superblock. If the superblock value is
3147 * present (even for non-bigalloc file systems), we will use it.
3148 */
3149static int count_overhead(struct super_block *sb, ext4_group_t grp,
3150 char *buf)
3151{
3152 struct ext4_sb_info *sbi = EXT4_SB(sb);
3153 struct ext4_group_desc *gdp;
3154 ext4_fsblk_t first_block, last_block, b;
3155 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3156 int s, j, count = 0;
3157
0548bbb8
TT		 3158 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3159 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3160 sbi->s_itb_per_group + 2);
3161
952fc18e
TT		 3162 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3163 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3164 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3165 for (i = 0; i < ngroups; i++) {
3166 gdp = ext4_get_group_desc(sb, i, NULL);
3167 b = ext4_block_bitmap(sb, gdp);
3168 if (b >= first_block && b <= last_block) {
3169 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3170 count++;
3171 }
3172 b = ext4_inode_bitmap(sb, gdp);
3173 if (b >= first_block && b <= last_block) {
3174 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3175 count++;
3176 }
3177 b = ext4_inode_table(sb, gdp);
3178 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3179 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3180 int c = EXT4_B2C(sbi, b - first_block);
3181 ext4_set_bit(c, buf);
3182 count++;
3183 }
3184 if (i != grp)
3185 continue;
3186 s = 0;
3187 if (ext4_bg_has_super(sb, grp)) {
3188 ext4_set_bit(s++, buf);
3189 count++;
3190 }
3191 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3192 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3193 count++;
3194 }
3195 }
3196 if (!count)
3197 return 0;
3198 return EXT4_CLUSTERS_PER_GROUP(sb) -
3199 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3200}
3201
3202/*
3203 * Compute the overhead and stash it in sbi->s_overhead
3204 */
3205int ext4_calculate_overhead(struct super_block *sb)
3206{
3207 struct ext4_sb_info *sbi = EXT4_SB(sb);
3208 struct ext4_super_block *es = sbi->s_es;
3209 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3210 ext4_fsblk_t overhead = 0;
3211 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3212
952fc18e
TT		 3213 if (!buf)
3214 return -ENOMEM;
3215
3216 /*
3217 * Compute the overhead (FS structures). This is constant
3218 * for a given filesystem unless the number of block groups
3219 * changes so we cache the previous value until it does.
3220 */
3221
3222 /*
3223 * All of the blocks before first_data_block are overhead
3224 */
3225 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3226
3227 /*
3228 * Add the overhead found in each block group
3229 */
3230 for (i = 0; i < ngroups; i++) {
3231 int blks;
3232
3233 blks = count_overhead(sb, i, buf);
3234 overhead += blks;
3235 if (blks)
3236 memset(buf, 0, PAGE_SIZE);
3237 cond_resched();
3238 }
0875a2b4
ES		 3239 /* Add the journal blocks as well */
3240 if (sbi->s_journal)
810da240 3241 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3242
952fc18e
TT		 3243 sbi->s_overhead = overhead;
3244 smp_wmb();
3245 free_page((unsigned long) buf);
3246 return 0;
3247}
3248
27dd4385
LC		 3249
3250static ext4_fsblk_t ext4_calculate_resv_clusters(struct ext4_sb_info *sbi)
3251{
3252 ext4_fsblk_t resv_clusters;
3253
3254 /*
3255 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3256 * This should cover the situations where we can not afford to run
3257 * out of space like for example punch hole, or converting
3258 * uninitialized extents in delalloc path. In most cases such
3259 * allocation would require 1, or 2 blocks, higher numbers are
3260 * very rare.
3261 */
3262 resv_clusters = ext4_blocks_count(sbi->s_es) >> sbi->s_cluster_bits;
3263
3264 do_div(resv_clusters, 50);
3265 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3266
3267 return resv_clusters;
3268}
3269
3270
3271static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3272{
3273 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3274 sbi->s_cluster_bits;
3275
3276 if (count >= clusters)
3277 return -EINVAL;
3278
3279 atomic64_set(&sbi->s_resv_clusters, count);
3280 return 0;
3281}
3282
2b2d6d01 3283static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3284{
d4c402d9 3285 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3286 struct buffer_head *bh;
617ba13b
MC		 3287 struct ext4_super_block *es = NULL;
3288 struct ext4_sb_info *sbi;
3289 ext4_fsblk_t block;
3290 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3291 ext4_fsblk_t logical_sb_block;
ac27a0ec 3292 unsigned long offset = 0;
ac27a0ec
DK		 3293 unsigned long journal_devnum = 0;
3294 unsigned long def_mount_opts;
3295 struct inode *root;
9f6200bb 3296 char *cp;
0390131b 3297 const char *descr;
dcc7dae3 3298 int ret = -ENOMEM;
281b5995 3299 int blocksize, clustersize;
4ec11028
TT		 3300 unsigned int db_count;
3301 unsigned int i;
281b5995 3302 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3303 __u64 blocks_count;
07aa2ea1 3304 int err = 0;
b3881f74 3305 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3306 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3307
3308 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3309 if (!sbi)
dcc7dae3 3310 goto out_free_orig;
705895b6
PE		 3311
3312 sbi->s_blockgroup_lock =
3313 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3314 if (!sbi->s_blockgroup_lock) {
3315 kfree(sbi);
dcc7dae3 3316 goto out_free_orig;
705895b6 3317 }
ac27a0ec 3318 sb->s_fs_info = sbi;
2c0544b2 3319 sbi->s_sb = sb;
240799cd 3320 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3321 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3322 if (sb->s_bdev->bd_part)
3323 sbi->s_sectors_written_start =
3324 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3325
9f6200bb
TT		 3326 /* Cleanup superblock name */
3327 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3328 *cp = '!';
3329
07aa2ea1 3330 /* -EINVAL is default */
dcc7dae3 3331 ret = -EINVAL;
617ba13b 3332 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3333 if (!blocksize) {
b31e1552 3334 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3335 goto out_fail;
3336 }
3337
3338 /*
617ba13b 3339 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3340 * block sizes. We need to calculate the offset from buffer start.
3341 */
617ba13b 3342 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3343 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3344 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3345 } else {
70bbb3e0 3346 logical_sb_block = sb_block;
ac27a0ec
DK		 3347 }
3348
70bbb3e0 3349 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3350 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3351 goto out_fail;
3352 }
3353 /*
3354 * Note: s_es must be initialized as soon as possible because
617ba13b 3355 * some ext4 macro-instructions depend on its value
ac27a0ec 3356 */
2716b802 3357 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3358 sbi->s_es = es;
3359 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3360 if (sb->s_magic != EXT4_SUPER_MAGIC)
3361 goto cantfind_ext4;
afc32f7e 3362 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3363
feb0ab32
DW		 3364 /* Warn if metadata_csum and gdt_csum are both set. */
3365 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3366 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3367 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3368 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3369 "redundant flags; please run fsck.");
3370
d25425f8
DW		 3371 /* Check for a known checksum algorithm */
3372 if (!ext4_verify_csum_type(sb, es)) {
3373 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3374 "unknown checksum algorithm.");
3375 silent = 1;
3376 goto cantfind_ext4;
3377 }
3378
0441984a
DW		 3379 /* Load the checksum driver */
3380 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3381 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3382 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3383 if (IS_ERR(sbi->s_chksum_driver)) {
3384 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3385 ret = PTR_ERR(sbi->s_chksum_driver);
3386 sbi->s_chksum_driver = NULL;
3387 goto failed_mount;
3388 }
3389 }
3390
a9c47317
DW		 3391 /* Check superblock checksum */
3392 if (!ext4_superblock_csum_verify(sb, es)) {
3393 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3394 "invalid superblock checksum. Run e2fsck?");
3395 silent = 1;
3396 goto cantfind_ext4;
3397 }
3398
3399 /* Precompute checksum seed for all metadata */
3400 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3401 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3402 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3403 sizeof(es->s_uuid));
3404
ac27a0ec
DK		 3405 /* Set defaults before we parse the mount options */
3406 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3407 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3408 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3409 set_opt(sb, DEBUG);
87f26807 3410 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3411 set_opt(sb, GRPID);
617ba13b 3412 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3413 set_opt(sb, NO_UID32);
ea663336 3414 /* xattr user namespace & acls are now defaulted on */
ea663336 3415 set_opt(sb, XATTR_USER);
03010a33 3416#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3417 set_opt(sb, POSIX_ACL);
2e7842b8 3418#endif
617ba13b 3419 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3420 set_opt(sb, JOURNAL_DATA);
617ba13b 3421 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3422 set_opt(sb, ORDERED_DATA);
617ba13b 3423 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3424 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3425
3426 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3427 set_opt(sb, ERRORS_PANIC);
bb4f397a 3428 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3429 set_opt(sb, ERRORS_CONT);
bb4f397a 3430 else
fd8c37ec 3431 set_opt(sb, ERRORS_RO);
8b67f04a 3432 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3433 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3434 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3435 set_opt(sb, DISCARD);
ac27a0ec 3436
08cefc7a
EB		 3437 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3438 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3439 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3440 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3441 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3442
8b67f04a 3443 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3444 set_opt(sb, BARRIER);
ac27a0ec 3445
dd919b98
AK		 3446 /*
3447 * enable delayed allocation by default
3448 * Use -o nodelalloc to turn it off
3449 */
bc0b75f7 3450 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3451 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3452 set_opt(sb, DELALLOC);
dd919b98 3453
51ce6511
LC		 3454 /*
3455 * set default s_li_wait_mult for lazyinit, for the case there is
3456 * no mount option specified.
3457 */
3458 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3459
8b67f04a 3460 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3461 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3462 ext4_msg(sb, KERN_WARNING,
3463 "failed to parse options in superblock: %s",
3464 sbi->s_es->s_mount_opts);
3465 }
5a916be1 3466 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3467 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3468 &journal_ioprio, 0))
ac27a0ec
DK		 3469 goto failed_mount;
3470
56889787
TT		 3471 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3472 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3473 "with data=journal disables delayed "
3474 "allocation and O_DIRECT support!\n");
3475 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3476 ext4_msg(sb, KERN_ERR, "can't mount with "
3477 "both data=journal and delalloc");
3478 goto failed_mount;
3479 }
3480 if (test_opt(sb, DIOREAD_NOLOCK)) {
3481 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3482 "both data=journal and dioread_nolock");
56889787
TT		 3483 goto failed_mount;
3484 }
3485 if (test_opt(sb, DELALLOC))
3486 clear_opt(sb, DELALLOC);
3487 }
3488
ac27a0ec 3489 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3490 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3491
617ba13b
MC		 3492 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3493 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3494 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3495 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3496 ext4_msg(sb, KERN_WARNING,
3497 "feature flags set on rev 0 fs, "
3498 "running e2fsck is recommended");
469108ff 3499
2035e776
TT		 3500 if (IS_EXT2_SB(sb)) {
3501 if (ext2_feature_set_ok(sb))
3502 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3503 "using the ext4 subsystem");
3504 else {
3505 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3506 "to feature incompatibilities");
3507 goto failed_mount;
3508 }
3509 }
3510
3511 if (IS_EXT3_SB(sb)) {
3512 if (ext3_feature_set_ok(sb))
3513 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3514 "using the ext4 subsystem");
3515 else {
3516 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3517 "to feature incompatibilities");
3518 goto failed_mount;
3519 }
3520 }
3521
ac27a0ec
DK		 3522 /*
3523 * Check feature flags regardless of the revision level, since we
3524 * previously didn't change the revision level when setting the flags,
3525 * so there is a chance incompat flags are set on a rev 0 filesystem.
3526 */
a13fb1a4 3527 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3528 goto failed_mount;
a13fb1a4 3529
261cb20c 3530 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3531 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3532 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3533 ext4_msg(sb, KERN_ERR,
3534 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3535 goto failed_mount;
3536 }
3537
ac27a0ec 3538 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3539 /* Validate the filesystem blocksize */
3540 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3541 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3542 blocksize);
ac27a0ec
DK		 3543 goto failed_mount;
3544 }
3545
2b2d6d01 3546 brelse(bh);
70bbb3e0
AM		 3547 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3548 offset = do_div(logical_sb_block, blocksize);
3549 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3550 if (!bh) {
b31e1552
ES		 3551 ext4_msg(sb, KERN_ERR,
3552 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3553 goto failed_mount;
3554 }
2716b802 3555 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3556 sbi->s_es = es;
617ba13b 3557 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3558 ext4_msg(sb, KERN_ERR,
3559 "Magic mismatch, very weird!");
ac27a0ec
DK		 3560 goto failed_mount;
3561 }
3562 }
3563
a13fb1a4
ES		 3564 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3565 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3566 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3567 has_huge_files);
3568 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3569
617ba13b
MC		 3570 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3571 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3572 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3573 } else {
3574 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3575 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3576 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3577 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3578 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3579 ext4_msg(sb, KERN_ERR,
3580 "unsupported inode size: %d",
2b2d6d01 3581 sbi->s_inode_size);
ac27a0ec
DK		 3582 goto failed_mount;
3583 }
ef7f3835
KS		 3584 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3585 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3586 }
0b8e58a1 3587
0d1ee42f
AR		 3588 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3589 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3590 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3591 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3592 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3593 ext4_msg(sb, KERN_ERR,
3594 "unsupported descriptor size %lu",
0d1ee42f
AR		 3595 sbi->s_desc_size);
3596 goto failed_mount;
3597 }
3598 } else
3599 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3600
ac27a0ec 3601 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3602 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3603 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3604 goto cantfind_ext4;
0b8e58a1 3605
617ba13b 3606 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3607 if (sbi->s_inodes_per_block == 0)
617ba13b 3608 goto cantfind_ext4;
ac27a0ec
DK		 3609 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3610 sbi->s_inodes_per_block;
0d1ee42f 3611 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3612 sbi->s_sbh = bh;
3613 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3614 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3615 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3616
2b2d6d01 3617 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3618 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3619 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3620 i = le32_to_cpu(es->s_flags);
3621 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3622 sbi->s_hash_unsigned = 3;
3623 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3624#ifdef __CHAR_UNSIGNED__
3625 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3626 sbi->s_hash_unsigned = 3;
3627#else
3628 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3629#endif
f99b2589 3630 }
ac27a0ec 3631
281b5995
TT		 3632 /* Handle clustersize */
3633 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3634 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3635 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3636 if (has_bigalloc) {
3637 if (clustersize < blocksize) {
3638 ext4_msg(sb, KERN_ERR,
3639 "cluster size (%d) smaller than "
3640 "block size (%d)", clustersize, blocksize);
3641 goto failed_mount;
3642 }
3643 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3644 le32_to_cpu(es->s_log_block_size);
3645 sbi->s_clusters_per_group =
3646 le32_to_cpu(es->s_clusters_per_group);
3647 if (sbi->s_clusters_per_group > blocksize * 8) {
3648 ext4_msg(sb, KERN_ERR,
3649 "#clusters per group too big: %lu",
3650 sbi->s_clusters_per_group);
3651 goto failed_mount;
3652 }
3653 if (sbi->s_blocks_per_group !=
3654 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3655 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3656 "clusters per group (%lu) inconsistent",
3657 sbi->s_blocks_per_group,
3658 sbi->s_clusters_per_group);
3659 goto failed_mount;
3660 }
3661 } else {
3662 if (clustersize != blocksize) {
3663 ext4_warning(sb, "fragment/cluster size (%d) != "
3664 "block size (%d)", clustersize,
3665 blocksize);
3666 clustersize = blocksize;
3667 }
3668 if (sbi->s_blocks_per_group > blocksize * 8) {
3669 ext4_msg(sb, KERN_ERR,
3670 "#blocks per group too big: %lu",
3671 sbi->s_blocks_per_group);
3672 goto failed_mount;
3673 }
3674 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3675 sbi->s_cluster_bits = 0;
ac27a0ec 3676 }
281b5995
TT		 3677 sbi->s_cluster_ratio = clustersize / blocksize;
3678
ac27a0ec 3679 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3680 ext4_msg(sb, KERN_ERR,
3681 "#inodes per group too big: %lu",
2b2d6d01 3682 sbi->s_inodes_per_group);
ac27a0ec
DK		 3683 goto failed_mount;
3684 }
3685
960fd856
TT		 3686 /* Do we have standard group size of clustersize * 8 blocks ? */
3687 if (sbi->s_blocks_per_group == clustersize << 3)
3688 set_opt2(sb, STD_GROUP_SIZE);
3689
bf43d84b
ES		 3690 /*
3691 * Test whether we have more sectors than will fit in sector_t,
3692 * and whether the max offset is addressable by the page cache.
3693 */
5a9ae68a 3694 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3695 ext4_blocks_count(es));
5a9ae68a 3696 if (err) {
b31e1552 3697 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3698 " too large to mount safely on this system");
ac27a0ec 3699 if (sizeof(sector_t) < 8)
90c699a9 3700 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3701 goto failed_mount;
3702 }
3703
617ba13b
MC		 3704 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3705 goto cantfind_ext4;
e7c95593 3706
0f2ddca6
FTN		 3707 /* check blocks count against device size */
3708 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3709 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3710 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3711 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3712 ext4_blocks_count(es), blocks_count);
3713 goto failed_mount;
3714 }
3715
0b8e58a1
AD		 3716 /*
3717 * It makes no sense for the first data block to be beyond the end
3718 * of the filesystem.
3719 */
3720 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3721 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3722 "block %u is beyond end of filesystem (%llu)",
3723 le32_to_cpu(es->s_first_data_block),
3724 ext4_blocks_count(es));
e7c95593
ES		 3725 goto failed_mount;
3726 }
bd81d8ee
LV		 3727 blocks_count = (ext4_blocks_count(es) -
3728 le32_to_cpu(es->s_first_data_block) +
3729 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3730 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3731 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3732 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3733 "(block count %llu, first data block %u, "
b31e1552 3734 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3735 ext4_blocks_count(es),
3736 le32_to_cpu(es->s_first_data_block),
3737 EXT4_BLOCKS_PER_GROUP(sb));
3738 goto failed_mount;
3739 }
bd81d8ee 3740 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3741 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3742 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3743 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3744 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3745 sbi->s_group_desc = ext4_kvmalloc(db_count *
3746 sizeof(struct buffer_head *),
3747 GFP_KERNEL);
ac27a0ec 3748 if (sbi->s_group_desc == NULL) {
b31e1552 3749 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3750 ret = -ENOMEM;
ac27a0ec
DK		 3751 goto failed_mount;
3752 }
3753
9f6200bb
TT		 3754 if (ext4_proc_root)
3755 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3756
66acdcf4
TT		 3757 if (sbi->s_proc)
3758 proc_create_data("options", S_IRUGO, sbi->s_proc,
3759 &ext4_seq_options_fops, sb);
3760
705895b6 3761 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3762
3763 for (i = 0; i < db_count; i++) {
70bbb3e0 3764 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3765 sbi->s_group_desc[i] = sb_bread(sb, block);
3766 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3767 ext4_msg(sb, KERN_ERR,
3768 "can't read group descriptor %d", i);
ac27a0ec
DK		 3769 db_count = i;
3770 goto failed_mount2;
3771 }
3772 }
bfff6873 3773 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3774 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3775 goto failed_mount2;
3776 }
772cb7c8
JS		 3777 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3778 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3779 ext4_msg(sb, KERN_ERR,
3780 "unable to initialize "
3781 "flex_bg meta info!");
772cb7c8
JS		 3782 goto failed_mount2;
3783 }
3784
ac27a0ec
DK		 3785 sbi->s_gdb_count = db_count;
3786 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3787 spin_lock_init(&sbi->s_next_gen_lock);
3788
04496411
TM		 3789 init_timer(&sbi->s_err_report);
3790 sbi->s_err_report.function = print_daily_error_info;
3791 sbi->s_err_report.data = (unsigned long) sb;
3792
a75ae78f 3793 /* Register extent status tree shrinker */
d3922a77 3794 ext4_es_register_shrinker(sbi);
a75ae78f 3795
57042651 3796 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3797 ext4_count_free_clusters(sb));
ce7e010a
TT		 3798 if (!err) {
3799 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3800 ext4_count_free_inodes(sb));
3801 }
3802 if (!err) {
3803 err = percpu_counter_init(&sbi->s_dirs_counter,
3804 ext4_count_dirs(sb));
3805 }
3806 if (!err) {
57042651 3807 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a 3808 }
1ac6466f
TT		 3809 if (!err) {
3810 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3811 }
ce7e010a
TT		 3812 if (err) {
3813 ext4_msg(sb, KERN_ERR, "insufficient memory");
3814 goto failed_mount3;
3815 }
3816
c9de560d 3817 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3818 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3819
ac27a0ec
DK		 3820 /*
3821 * set up enough so that it can read an inode
3822 */
9ca92389
TT		 3823 if (!test_opt(sb, NOLOAD) &&
3824 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3825 sb->s_op = &ext4_sops;
3826 else
3827 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3828 sb->s_export_op = &ext4_export_ops;
3829 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3830#ifdef CONFIG_QUOTA
617ba13b 3831 sb->dq_op = &ext4_quota_operations;
262b4662 3832 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
7c319d32 3833 sb->s_qcop = &ext4_qctl_sysfile_operations;
262b4662
JK		 3834 else
3835 sb->s_qcop = &ext4_qctl_operations;
ac27a0ec 3836#endif
f2fa2ffc
AK		 3837 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3838
ac27a0ec 3839 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3840 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3841
3842 sb->s_root = NULL;
3843
3844 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3845 EXT4_HAS_INCOMPAT_FEATURE(sb,
3846 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3847
c5e06d10
JL		 3848 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3849 !(sb->s_flags & MS_RDONLY))
3850 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3851 goto failed_mount3;
3852
ac27a0ec
DK		 3853 /*
3854 * The first inode we look at is the journal inode. Don't try
3855 * root first: it may be modified in the journal!
3856 */
3857 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3858 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3859 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3860 goto failed_mount3;
0390131b
FM		 3861 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3862 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3863 ext4_msg(sb, KERN_ERR, "required journal recovery "
3864 "suppressed and not mounted read-only");
744692dc 3865 goto failed_mount_wq;
ac27a0ec 3866 } else {
fd8c37ec 3867 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3868 sbi->s_journal = NULL;
3869 needs_recovery = 0;
3870 goto no_journal;
ac27a0ec
DK		 3871 }
3872
f32aaf2d 3873 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3874 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3875 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3876 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3877 goto failed_mount_wq;
eb40a09c
JS		 3878 }
3879
25ed6e8a
DW		 3880 if (!set_journal_csum_feature_set(sb)) {
3881 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3882 "feature set");
3883 goto failed_mount_wq;
d4da6c9c 3884 }
818d276c 3885
ac27a0ec
DK		 3886 /* We have now updated the journal if required, so we can
3887 * validate the data journaling mode. */
3888 switch (test_opt(sb, DATA_FLAGS)) {
3889 case 0:
3890 /* No mode set, assume a default based on the journal
63f57933
AM		 3891 * capabilities: ORDERED_DATA if the journal can
3892 * cope, else JOURNAL_DATA
3893 */
dab291af
MC		 3894 if (jbd2_journal_check_available_features
3895 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3896 set_opt(sb, ORDERED_DATA);
ac27a0ec 3897 else
fd8c37ec 3898 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3899 break;
3900
617ba13b
MC		 3901 case EXT4_MOUNT_ORDERED_DATA:
3902 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3903 if (!jbd2_journal_check_available_features
3904 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3905 ext4_msg(sb, KERN_ERR, "Journal does not support "
3906 "requested data journaling mode");
744692dc 3907 goto failed_mount_wq;
ac27a0ec
DK		 3908 }
3909 default:
3910 break;
3911 }
b3881f74 3912 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3913
18aadd47
BJ		 3914 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3915
ce7e010a
TT		 3916 /*
3917 * The journal may have updated the bg summary counts, so we
3918 * need to update the global counters.
3919 */
57042651 3920 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3921 ext4_count_free_clusters(sb));
ce7e010a
TT		 3922 percpu_counter_set(&sbi->s_freeinodes_counter,
3923 ext4_count_free_inodes(sb));
3924 percpu_counter_set(&sbi->s_dirs_counter,
3925 ext4_count_dirs(sb));
57042651 3926 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3927
ce7e010a 3928no_journal:
952fc18e
TT		 3929 /*
3930 * Get the # of file system overhead blocks from the
3931 * superblock if present.
3932 */
3933 if (es->s_overhead_clusters)
3934 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3935 else {
07aa2ea1
LC		 3936 err = ext4_calculate_overhead(sb);
3937 if (err)
952fc18e
TT		 3938 goto failed_mount_wq;
3939 }
3940
fd89d5f2
TH		 3941 /*
3942 * The maximum number of concurrent works can be high and
3943 * concurrency isn't really necessary. Limit it to 1.
3944 */
2e8fa54e
JK		 3945 EXT4_SB(sb)->rsv_conversion_wq =
3946 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3947 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3948 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 3949 ret = -ENOMEM;
2e8fa54e
JK		 3950 goto failed_mount4;
3951 }
3952
ac27a0ec 3953 /*
dab291af 3954 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3955 * so we can safely mount the rest of the filesystem now.
3956 */
3957
1d1fe1ee
DH		 3958 root = ext4_iget(sb, EXT4_ROOT_INO);
3959 if (IS_ERR(root)) {
b31e1552 3960 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3961 ret = PTR_ERR(root);
32a9bb57 3962 root = NULL;
ac27a0ec
DK		 3963 goto failed_mount4;
3964 }
3965 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3966 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3967 iput(root);
ac27a0ec
DK		 3968 goto failed_mount4;
3969 }
48fde701 3970 sb->s_root = d_make_root(root);
1d1fe1ee 3971 if (!sb->s_root) {
b31e1552 3972 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3973 ret = -ENOMEM;
3974 goto failed_mount4;
3975 }
ac27a0ec 3976
7e84b621
ES		 3977 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3978 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3979
3980 /* determine the minimum size of new large inodes, if present */
3981 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3982 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3983 EXT4_GOOD_OLD_INODE_SIZE;
3984 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3985 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3986 if (sbi->s_want_extra_isize <
3987 le16_to_cpu(es->s_want_extra_isize))
3988 sbi->s_want_extra_isize =
3989 le16_to_cpu(es->s_want_extra_isize);
3990 if (sbi->s_want_extra_isize <
3991 le16_to_cpu(es->s_min_extra_isize))
3992 sbi->s_want_extra_isize =
3993 le16_to_cpu(es->s_min_extra_isize);
3994 }
3995 }
3996 /* Check if enough inode space is available */
3997 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3998 sbi->s_inode_size) {
3999 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4000 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4001 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4002 "available");
ef7f3835
KS		 4003 }
4004
27dd4385
LC		 4005 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sbi));
4006 if (err) {
4007 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4008 "reserved pool", ext4_calculate_resv_clusters(sbi));
4009 goto failed_mount4a;
4010 }
4011
6fd058f7
TT		 4012 err = ext4_setup_system_zone(sb);
4013 if (err) {
b31e1552 4014 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4015 "zone (%d)", err);
94bf608a 4016 goto failed_mount4a;
6fd058f7
TT		 4017 }
4018
c2774d84 4019 ext4_ext_init(sb);
9d99012f 4020 err = ext4_mb_init(sb);
c2774d84 4021 if (err) {
421f91d2 4022 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 4023 err);
dcf2d804 4024 goto failed_mount5;
c2774d84
AK		 4025 }
4026
bfff6873
LC		 4027 err = ext4_register_li_request(sb, first_not_zeroed);
4028 if (err)
dcf2d804 4029 goto failed_mount6;
bfff6873 4030
3197ebdb
TT		 4031 sbi->s_kobj.kset = ext4_kset;
4032 init_completion(&sbi->s_kobj_unregister);
4033 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4034 "%s", sb->s_id);
dcf2d804
TM		 4035 if (err)
4036 goto failed_mount7;
3197ebdb 4037
9b2ff357
JK		 4038#ifdef CONFIG_QUOTA
4039 /* Enable quota usage during mount. */
4040 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4041 !(sb->s_flags & MS_RDONLY)) {
4042 err = ext4_enable_quotas(sb);
4043 if (err)
4044 goto failed_mount8;
4045 }
4046#endif /* CONFIG_QUOTA */
4047
617ba13b
MC		 4048 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4049 ext4_orphan_cleanup(sb, es);
4050 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4051 if (needs_recovery) {
b31e1552 4052 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4053 ext4_mark_recovery_complete(sb, es);
4054 }
4055 if (EXT4_SB(sb)->s_journal) {
4056 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4057 descr = " journalled data mode";
4058 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4059 descr = " ordered data mode";
4060 else
4061 descr = " writeback data mode";
4062 } else
4063 descr = "out journal";
4064
79add3a3
LC		 4065 if (test_opt(sb, DISCARD)) {
4066 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4067 if (!blk_queue_discard(q))
4068 ext4_msg(sb, KERN_WARNING,
4069 "mounting with \"discard\" option, but "
4070 "the device does not support discard");
4071 }
4072
d4c402d9 4073 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4074 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4075 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4076
66e61a9e
TT		 4077 if (es->s_error_count)
4078 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4079
d4c402d9 4080 kfree(orig_data);
ac27a0ec
DK		 4081 return 0;
4082
617ba13b 4083cantfind_ext4:
ac27a0ec 4084 if (!silent)
b31e1552 4085 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4086 goto failed_mount;
4087
72ba7450
TT		 4088#ifdef CONFIG_QUOTA
4089failed_mount8:
4090 kobject_del(&sbi->s_kobj);
4091#endif
dcf2d804
TM		 4092failed_mount7:
4093 ext4_unregister_li_request(sb);
4094failed_mount6:
dcf2d804 4095 ext4_mb_release(sb);
94bf608a
AV		 4096failed_mount5:
4097 ext4_ext_release(sb);
dcf2d804 4098 ext4_release_system_zone(sb);
94bf608a
AV		 4099failed_mount4a:
4100 dput(sb->s_root);
32a9bb57 4101 sb->s_root = NULL;
94bf608a 4102failed_mount4:
b31e1552 4103 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4104 if (EXT4_SB(sb)->rsv_conversion_wq)
4105 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4106failed_mount_wq:
0390131b
FM		 4107 if (sbi->s_journal) {
4108 jbd2_journal_destroy(sbi->s_journal);
4109 sbi->s_journal = NULL;
4110 }
ac27a0ec 4111failed_mount3:
d3922a77 4112 ext4_es_unregister_shrinker(sbi);
04496411 4113 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4114 if (sbi->s_flex_groups)
4115 ext4_kvfree(sbi->s_flex_groups);
57042651 4116 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4117 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4118 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4119 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 4120 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
c5e06d10
JL		 4121 if (sbi->s_mmp_tsk)
4122 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4123failed_mount2:
4124 for (i = 0; i < db_count; i++)
4125 brelse(sbi->s_group_desc[i]);
f18a5f21 4126 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4127failed_mount:
0441984a
DW		 4128 if (sbi->s_chksum_driver)
4129 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4130 if (sbi->s_proc) {
66acdcf4 4131 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4132 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4133 }
ac27a0ec
DK		 4134#ifdef CONFIG_QUOTA
4135 for (i = 0; i < MAXQUOTAS; i++)
4136 kfree(sbi->s_qf_names[i]);
4137#endif
617ba13b 4138 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4139 brelse(bh);
4140out_fail:
4141 sb->s_fs_info = NULL;
f6830165 4142 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4143 kfree(sbi);
dcc7dae3 4144out_free_orig:
d4c402d9 4145 kfree(orig_data);
07aa2ea1 4146 return err ? err : ret;
ac27a0ec
DK		 4147}
4148
4149/*
4150 * Setup any per-fs journal parameters now. We'll do this both on
4151 * initial mount, once the journal has been initialised but before we've
4152 * done any recovery; and again on any subsequent remount.
4153 */
617ba13b 4154static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4155{
617ba13b 4156 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4157
30773840
TT		 4158 journal->j_commit_interval = sbi->s_commit_interval;
4159 journal->j_min_batch_time = sbi->s_min_batch_time;
4160 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4161
a931da6a 4162 write_lock(&journal->j_state_lock);
ac27a0ec 4163 if (test_opt(sb, BARRIER))
dab291af 4164 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4165 else
dab291af 4166 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4167 if (test_opt(sb, DATA_ERR_ABORT))
4168 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4169 else
4170 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4171 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4172}
4173
617ba13b 4174static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4175 unsigned int journal_inum)
4176{
4177 struct inode *journal_inode;
4178 journal_t *journal;
4179
0390131b
FM		 4180 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4181
ac27a0ec
DK		 4182 /* First, test for the existence of a valid inode on disk. Bad
4183 * things happen if we iget() an unused inode, as the subsequent
4184 * iput() will try to delete it. */
4185
1d1fe1ee
DH		 4186 journal_inode = ext4_iget(sb, journal_inum);
4187 if (IS_ERR(journal_inode)) {
b31e1552 4188 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4189 return NULL;
4190 }
4191 if (!journal_inode->i_nlink) {
4192 make_bad_inode(journal_inode);
4193 iput(journal_inode);
b31e1552 4194 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4195 return NULL;
4196 }
4197
e5f8eab8 4198 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4199 journal_inode, journal_inode->i_size);
1d1fe1ee 4200 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4201 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4202 iput(journal_inode);
4203 return NULL;
4204 }
4205
dab291af 4206 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4207 if (!journal) {
b31e1552 4208 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4209 iput(journal_inode);
4210 return NULL;
4211 }
4212 journal->j_private = sb;
617ba13b 4213 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4214 return journal;
4215}
4216
617ba13b 4217static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4218 dev_t j_dev)
4219{
2b2d6d01 4220 struct buffer_head *bh;
ac27a0ec 4221 journal_t *journal;
617ba13b
MC		 4222 ext4_fsblk_t start;
4223 ext4_fsblk_t len;
ac27a0ec 4224 int hblock, blocksize;
617ba13b 4225 ext4_fsblk_t sb_block;
ac27a0ec 4226 unsigned long offset;
2b2d6d01 4227 struct ext4_super_block *es;
ac27a0ec
DK		 4228 struct block_device *bdev;
4229
0390131b
FM		 4230 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4231
b31e1552 4232 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4233 if (bdev == NULL)
4234 return NULL;
4235
ac27a0ec 4236 blocksize = sb->s_blocksize;
e1defc4f 4237 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4238 if (blocksize < hblock) {
b31e1552
ES		 4239 ext4_msg(sb, KERN_ERR,
4240 "blocksize too small for journal device");
ac27a0ec
DK		 4241 goto out_bdev;
4242 }
4243
617ba13b
MC		 4244 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4245 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4246 set_blocksize(bdev, blocksize);
4247 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4248 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4249 "external journal");
ac27a0ec
DK		 4250 goto out_bdev;
4251 }
4252
2716b802 4253 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4254 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4255 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4256 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4257 ext4_msg(sb, KERN_ERR, "external journal has "
4258 "bad superblock");
ac27a0ec
DK		 4259 brelse(bh);
4260 goto out_bdev;
4261 }
4262
617ba13b 4263 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4264 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4265 brelse(bh);
4266 goto out_bdev;
4267 }
4268
bd81d8ee 4269 len = ext4_blocks_count(es);
ac27a0ec
DK		 4270 start = sb_block + 1;
4271 brelse(bh); /* we're done with the superblock */
4272
dab291af 4273 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4274 start, len, blocksize);
4275 if (!journal) {
b31e1552 4276 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4277 goto out_bdev;
4278 }
4279 journal->j_private = sb;
9f203507 4280 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4281 wait_on_buffer(journal->j_sb_buffer);
4282 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4283 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4284 goto out_journal;
4285 }
4286 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4287 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4288 "user (unsupported) - %d",
ac27a0ec
DK		 4289 be32_to_cpu(journal->j_superblock->s_nr_users));
4290 goto out_journal;
4291 }
617ba13b
MC		 4292 EXT4_SB(sb)->journal_bdev = bdev;
4293 ext4_init_journal_params(sb, journal);
ac27a0ec 4294 return journal;
0b8e58a1 4295
ac27a0ec 4296out_journal:
dab291af 4297 jbd2_journal_destroy(journal);
ac27a0ec 4298out_bdev:
617ba13b 4299 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4300 return NULL;
4301}
4302
617ba13b
MC		 4303static int ext4_load_journal(struct super_block *sb,
4304 struct ext4_super_block *es,
ac27a0ec
DK		 4305 unsigned long journal_devnum)
4306{
4307 journal_t *journal;
4308 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4309 dev_t journal_dev;
4310 int err = 0;
4311 int really_read_only;
4312
0390131b
FM		 4313 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4314
ac27a0ec
DK		 4315 if (journal_devnum &&
4316 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4317 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4318 "numbers have changed");
ac27a0ec
DK		 4319 journal_dev = new_decode_dev(journal_devnum);
4320 } else
4321 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4322
4323 really_read_only = bdev_read_only(sb->s_bdev);
4324
4325 /*
4326 * Are we loading a blank journal or performing recovery after a
4327 * crash? For recovery, we need to check in advance whether we
4328 * can get read-write access to the device.
4329 */
617ba13b 4330 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4331 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4332 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4333 "required on readonly filesystem");
ac27a0ec 4334 if (really_read_only) {
b31e1552
ES		 4335 ext4_msg(sb, KERN_ERR, "write access "
4336 "unavailable, cannot proceed");
ac27a0ec
DK		 4337 return -EROFS;
4338 }
b31e1552
ES		 4339 ext4_msg(sb, KERN_INFO, "write access will "
4340 "be enabled during recovery");
ac27a0ec
DK		 4341 }
4342 }
4343
4344 if (journal_inum && journal_dev) {
b31e1552
ES		 4345 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4346 "and inode journals!");
ac27a0ec
DK		 4347 return -EINVAL;
4348 }
4349
4350 if (journal_inum) {
617ba13b 4351 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4352 return -EINVAL;
4353 } else {
617ba13b 4354 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4355 return -EINVAL;
4356 }
4357
90576c0b 4358 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4359 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4360
617ba13b 4361 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4362 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4363 if (!err) {
4364 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4365 if (save)
4366 memcpy(save, ((char *) es) +
4367 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4368 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4369 if (save)
4370 memcpy(((char *) es) + EXT4_S_ERR_START,
4371 save, EXT4_S_ERR_LEN);
4372 kfree(save);
4373 }
ac27a0ec
DK		 4374
4375 if (err) {
b31e1552 4376 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4377 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4378 return err;
4379 }
4380
617ba13b
MC		 4381 EXT4_SB(sb)->s_journal = journal;
4382 ext4_clear_journal_err(sb, es);
ac27a0ec 4383
c41303ce 4384 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4385 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4386 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4387
4388 /* Make sure we flush the recovery flag to disk. */
e2d67052 4389 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4390 }
4391
4392 return 0;
4393}
4394
e2d67052 4395static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4396{
e2d67052 4397 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4398 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4399 int error = 0;
ac27a0ec 4400
7c2e7087 4401 if (!sbh || block_device_ejected(sb))
c4be0c1d 4402 return error;
914258bf
TT		 4403 if (buffer_write_io_error(sbh)) {
4404 /*
4405 * Oh, dear. A previous attempt to write the
4406 * superblock failed. This could happen because the
4407 * USB device was yanked out. Or it could happen to
4408 * be a transient write error and maybe the block will
4409 * be remapped. Nothing we can do but to retry the
4410 * write and hope for the best.
4411 */
b31e1552
ES		 4412 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4413 "superblock detected");
914258bf
TT		 4414 clear_buffer_write_io_error(sbh);
4415 set_buffer_uptodate(sbh);
4416 }
71290b36
TT		 4417 /*
4418 * If the file system is mounted read-only, don't update the
4419 * superblock write time. This avoids updating the superblock
4420 * write time when we are mounting the root file system
4421 * read/only but we need to replay the journal; at that point,
4422 * for people who are east of GMT and who make their clock
4423 * tick in localtime for Windows bug-for-bug compatibility,
4424 * the clock is set in the future, and this will cause e2fsck
4425 * to complain and force a full file system check.
4426 */
4427 if (!(sb->s_flags & MS_RDONLY))
4428 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4429 if (sb->s_bdev->bd_part)
4430 es->s_kbytes_written =
4431 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4432 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4433 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4434 else
4435 es->s_kbytes_written =
4436 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4437 ext4_free_blocks_count_set(es,
4438 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4439 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4440 es->s_free_inodes_count =
4441 cpu_to_le32(percpu_counter_sum_positive(
4442 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4443 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4444 ext4_superblock_csum_set(sb);
ac27a0ec 4445 mark_buffer_dirty(sbh);
914258bf 4446 if (sync) {
c4be0c1d
TS		 4447 error = sync_dirty_buffer(sbh);
4448 if (error)
4449 return error;
4450
4451 error = buffer_write_io_error(sbh);
4452 if (error) {
b31e1552
ES		 4453 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4454 "superblock");
914258bf
TT		 4455 clear_buffer_write_io_error(sbh);
4456 set_buffer_uptodate(sbh);
4457 }
4458 }
c4be0c1d 4459 return error;
ac27a0ec
DK		 4460}
4461
ac27a0ec
DK		 4462/*
4463 * Have we just finished recovery? If so, and if we are mounting (or
4464 * remounting) the filesystem readonly, then we will end up with a
4465 * consistent fs on disk. Record that fact.
4466 */
2b2d6d01
TT		 4467static void ext4_mark_recovery_complete(struct super_block *sb,
4468 struct ext4_super_block *es)
ac27a0ec 4469{
617ba13b 4470 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4471
0390131b
FM		 4472 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4473 BUG_ON(journal != NULL);
4474 return;
4475 }
dab291af 4476 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4477 if (jbd2_journal_flush(journal) < 0)
4478 goto out;
4479
617ba13b 4480 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4481 sb->s_flags & MS_RDONLY) {
617ba13b 4482 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4483 ext4_commit_super(sb, 1);
ac27a0ec 4484 }
7ffe1ea8
HK		 4485
4486out:
dab291af 4487 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4488}
4489
4490/*
4491 * If we are mounting (or read-write remounting) a filesystem whose journal
4492 * has recorded an error from a previous lifetime, move that error to the
4493 * main filesystem now.
4494 */
2b2d6d01
TT		 4495static void ext4_clear_journal_err(struct super_block *sb,
4496 struct ext4_super_block *es)
ac27a0ec
DK		 4497{
4498 journal_t *journal;
4499 int j_errno;
4500 const char *errstr;
4501
0390131b
FM		 4502 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4503
617ba13b 4504 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4505
4506 /*
4507 * Now check for any error status which may have been recorded in the
617ba13b 4508 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4509 */
4510
dab291af 4511 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4512 if (j_errno) {
4513 char nbuf[16];
4514
617ba13b 4515 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4516 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4517 "from previous mount: %s", errstr);
12062ddd 4518 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4519
617ba13b
MC		 4520 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4521 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4522 ext4_commit_super(sb, 1);
ac27a0ec 4523
dab291af 4524 jbd2_journal_clear_err(journal);
d796c52e 4525 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4526 }
4527}
4528
4529/*
4530 * Force the running and committing transactions to commit,
4531 * and wait on the commit.
4532 */
617ba13b 4533int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4534{
4535 journal_t *journal;
ac27a0ec
DK		 4536
4537 if (sb->s_flags & MS_RDONLY)
4538 return 0;
4539
617ba13b 4540 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4541 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4542}
4543
617ba13b 4544static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4545{
14ce0cb4 4546 int ret = 0;
9eddacf9 4547 tid_t target;
06a407f1 4548 bool needs_barrier = false;
8d5d02e6 4549 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4550
9bffad1e 4551 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4552 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4553 /*
4554 * Writeback quota in non-journalled quota case - journalled quota has
4555 * no dirty dquots
4556 */
4557 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4558 /*
4559 * Data writeback is possible w/o journal transaction, so barrier must
4560 * being sent at the end of the function. But we can skip it if
4561 * transaction_commit will do it for us.
4562 */
4563 target = jbd2_get_latest_transaction(sbi->s_journal);
4564 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4565 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4566 needs_barrier = true;
4567
8d5d02e6 4568 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4569 if (wait)
06a407f1
DM		 4570 ret = jbd2_log_wait_commit(sbi->s_journal, target);
4571 }
4572 if (needs_barrier) {
4573 int err;
4574 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4575 if (!ret)
4576 ret = err;
0390131b 4577 }
06a407f1
DM		 4578
4579 return ret;
4580}
4581
4582static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4583{
4584 int ret = 0;
4585
4586 trace_ext4_sync_fs(sb, wait);
4587 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
06a407f1
DM		 4588 dquot_writeback_dquots(sb, -1);
4589 if (wait && test_opt(sb, BARRIER))
4590 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4591
14ce0cb4 4592 return ret;
ac27a0ec
DK		 4593}
4594
4595/*
4596 * LVM calls this function before a (read-only) snapshot is created. This
4597 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4598 *
4599 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4600 * state independently. It relies on upper layer to stop all data & metadata
4601 * modifications.
ac27a0ec 4602 */
c4be0c1d 4603static int ext4_freeze(struct super_block *sb)
ac27a0ec 4604{
c4be0c1d
TS		 4605 int error = 0;
4606 journal_t *journal;
ac27a0ec 4607
9ca92389
TT		 4608 if (sb->s_flags & MS_RDONLY)
4609 return 0;
ac27a0ec 4610
9ca92389 4611 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4612
9ca92389
TT		 4613 /* Now we set up the journal barrier. */
4614 jbd2_journal_lock_updates(journal);
ac27a0ec 4615
9ca92389
TT		 4616 /*
4617 * Don't clear the needs_recovery flag if we failed to flush
4618 * the journal.
4619 */
4620 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4621 if (error < 0)
4622 goto out;
9ca92389
TT		 4623
4624 /* Journal blocked and flushed, clear needs_recovery flag. */
4625 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4626 error = ext4_commit_super(sb, 1);
6b0310fb 4627out:
8e8ad8a5 4628 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4629 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4630 return error;
ac27a0ec
DK		 4631}
4632
4633/*
4634 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4635 * flag here, even though the filesystem is not technically dirty yet.
4636 */
c4be0c1d 4637static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4638{
9ca92389
TT		 4639 if (sb->s_flags & MS_RDONLY)
4640 return 0;
4641
9ca92389
TT		 4642 /* Reset the needs_recovery flag before the fs is unlocked. */
4643 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4644 ext4_commit_super(sb, 1);
c4be0c1d 4645 return 0;
ac27a0ec
DK		 4646}
4647
673c6100
TT		 4648/*
4649 * Structure to save mount options for ext4_remount's benefit
4650 */
4651struct ext4_mount_options {
4652 unsigned long s_mount_opt;
a2595b8a 4653 unsigned long s_mount_opt2;
08cefc7a
EB		 4654 kuid_t s_resuid;
4655 kgid_t s_resgid;
673c6100
TT		 4656 unsigned long s_commit_interval;
4657 u32 s_min_batch_time, s_max_batch_time;
4658#ifdef CONFIG_QUOTA
4659 int s_jquota_fmt;
4660 char *s_qf_names[MAXQUOTAS];
4661#endif
4662};
4663
2b2d6d01 4664static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4665{
2b2d6d01 4666 struct ext4_super_block *es;
617ba13b 4667 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4668 unsigned long old_sb_flags;
617ba13b 4669 struct ext4_mount_options old_opts;
c79d967d 4670 int enable_quota = 0;
8a266467 4671 ext4_group_t g;
b3881f74 4672 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4673 int err = 0;
ac27a0ec 4674#ifdef CONFIG_QUOTA
03dafb5f 4675 int i, j;
ac27a0ec 4676#endif
d4c402d9 4677 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4678
4679 /* Store the original options */
4680 old_sb_flags = sb->s_flags;
4681 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4682 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4683 old_opts.s_resuid = sbi->s_resuid;
4684 old_opts.s_resgid = sbi->s_resgid;
4685 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4686 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4687 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4688#ifdef CONFIG_QUOTA
4689 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4690 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f
CG		 4691 if (sbi->s_qf_names[i]) {
4692 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4693 GFP_KERNEL);
4694 if (!old_opts.s_qf_names[i]) {
4695 for (j = 0; j < i; j++)
4696 kfree(old_opts.s_qf_names[j]);
3e36a163 4697 kfree(orig_data);
03dafb5f
CG		 4698 return -ENOMEM;
4699 }
4700 } else
4701 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4702#endif
b3881f74
TT		 4703 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4704 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4705
4706 /*
4707 * Allow the "check" option to be passed as a remount option.
4708 */
661aa520 4709 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4710 err = -EINVAL;
4711 goto restore_opts;
4712 }
4713
6ae6514b
PS		 4714 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4715 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4716 ext4_msg(sb, KERN_ERR, "can't mount with "
4717 "both data=journal and delalloc");
4718 err = -EINVAL;
4719 goto restore_opts;
4720 }
4721 if (test_opt(sb, DIOREAD_NOLOCK)) {
4722 ext4_msg(sb, KERN_ERR, "can't mount with "
4723 "both data=journal and dioread_nolock");
4724 err = -EINVAL;
4725 goto restore_opts;
4726 }
4727 }
4728
4ab2f15b 4729 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4730 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4731
4732 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4733 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4734
4735 es = sbi->s_es;
4736
b3881f74 4737 if (sbi->s_journal) {
0390131b 4738 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4739 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4740 }
ac27a0ec 4741
661aa520 4742 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4743 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4744 err = -EROFS;
4745 goto restore_opts;
4746 }
4747
4748 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4749 err = dquot_suspend(sb, -1);
4750 if (err < 0)
c79d967d 4751 goto restore_opts;
c79d967d 4752
ac27a0ec
DK		 4753 /*
4754 * First of all, the unconditional stuff we have to do
4755 * to disable replay of the journal when we next remount
4756 */
4757 sb->s_flags |= MS_RDONLY;
4758
4759 /*
4760 * OK, test if we are remounting a valid rw partition
4761 * readonly, and if so set the rdonly flag and then
4762 * mark the partition as valid again.
4763 */
617ba13b
MC		 4764 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4765 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4766 es->s_state = cpu_to_le16(sbi->s_mount_state);
4767
a63c9eb2 4768 if (sbi->s_journal)
0390131b 4769 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4770 } else {
a13fb1a4
ES		 4771 /* Make sure we can mount this feature set readwrite */
4772 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4773 err = -EROFS;
4774 goto restore_opts;
4775 }
8a266467
TT		 4776 /*
4777 * Make sure the group descriptor checksums
0b8e58a1 4778 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4779 */
4780 for (g = 0; g < sbi->s_groups_count; g++) {
4781 struct ext4_group_desc *gdp =
4782 ext4_get_group_desc(sb, g, NULL);
4783
feb0ab32 4784 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4785 ext4_msg(sb, KERN_ERR,
4786 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4787 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4788 le16_to_cpu(gdp->bg_checksum));
4789 err = -EINVAL;
4790 goto restore_opts;
4791 }
4792 }
4793
ead6596b
ES		 4794 /*
4795 * If we have an unprocessed orphan list hanging
4796 * around from a previously readonly bdev mount,
4797 * require a full umount/remount for now.
4798 */
4799 if (es->s_last_orphan) {
b31e1552 4800 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4801 "remount RDWR because of unprocessed "
4802 "orphan inode list. Please "
b31e1552 4803 "umount/remount instead");
ead6596b
ES		 4804 err = -EINVAL;
4805 goto restore_opts;
4806 }
4807
ac27a0ec
DK		 4808 /*
4809 * Mounting a RDONLY partition read-write, so reread
4810 * and store the current valid flag. (It may have
4811 * been changed by e2fsck since we originally mounted
4812 * the partition.)
4813 */
0390131b
FM		 4814 if (sbi->s_journal)
4815 ext4_clear_journal_err(sb, es);
ac27a0ec 4816 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4817 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4818 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4819 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4820 EXT4_FEATURE_INCOMPAT_MMP))
4821 if (ext4_multi_mount_protect(sb,
4822 le64_to_cpu(es->s_mmp_block))) {
4823 err = -EROFS;
4824 goto restore_opts;
4825 }
c79d967d 4826 enable_quota = 1;
ac27a0ec
DK		 4827 }
4828 }
bfff6873
LC		 4829
4830 /*
4831 * Reinitialize lazy itable initialization thread based on
4832 * current settings
4833 */
4834 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4835 ext4_unregister_li_request(sb);
4836 else {
4837 ext4_group_t first_not_zeroed;
4838 first_not_zeroed = ext4_has_uninit_itable(sb);
4839 ext4_register_li_request(sb, first_not_zeroed);
4840 }
4841
6fd058f7 4842 ext4_setup_system_zone(sb);
d096ad0f 4843 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4844 ext4_commit_super(sb, 1);
0390131b 4845
ac27a0ec
DK		 4846#ifdef CONFIG_QUOTA
4847 /* Release old quota file names */
4848 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f 4849 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4850 if (enable_quota) {
4851 if (sb_any_quota_suspended(sb))
4852 dquot_resume(sb, -1);
4853 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4854 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4855 err = ext4_enable_quotas(sb);
07724f98 4856 if (err)
7c319d32 4857 goto restore_opts;
7c319d32
AK		 4858 }
4859 }
ac27a0ec 4860#endif
d4c402d9
CW		 4861
4862 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4863 kfree(orig_data);
ac27a0ec 4864 return 0;
0b8e58a1 4865
ac27a0ec
DK		 4866restore_opts:
4867 sb->s_flags = old_sb_flags;
4868 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4869 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4870 sbi->s_resuid = old_opts.s_resuid;
4871 sbi->s_resgid = old_opts.s_resgid;
4872 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4873 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4874 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4875#ifdef CONFIG_QUOTA
4876 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4877 for (i = 0; i < MAXQUOTAS; i++) {
03dafb5f 4878 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4879 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4880 }
4881#endif
d4c402d9 4882 kfree(orig_data);
ac27a0ec
DK		 4883 return err;
4884}
4885
2b2d6d01 4886static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4887{
4888 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4889 struct ext4_sb_info *sbi = EXT4_SB(sb);
4890 struct ext4_super_block *es = sbi->s_es;
27dd4385 4891 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 4892 u64 fsid;
d02a9391 4893 s64 bfree;
27dd4385 4894 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 4895
952fc18e
TT		 4896 if (!test_opt(sb, MINIX_DF))
4897 overhead = sbi->s_overhead;
ac27a0ec 4898
617ba13b 4899 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4900 buf->f_bsize = sb->s_blocksize;
b72f78cb 4901 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4902 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4903 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4904 /* prevent underflow in case that few free space is available */
57042651 4905 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 4906 buf->f_bavail = buf->f_bfree -
4907 (ext4_r_blocks_count(es) + resv_blocks);
4908 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 4909 buf->f_bavail = 0;
4910 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4911 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4912 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4913 fsid = le64_to_cpup((void *)es->s_uuid) ^
4914 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4915 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4916 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4917
ac27a0ec
DK		 4918 return 0;
4919}
4920
0b8e58a1
AD		 4921/* Helper function for writing quotas on sync - we need to start transaction
4922 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4923 * Process 1 Process 2
617ba13b 4924 * ext4_create() quota_sync()
a269eb18 4925 * jbd2_journal_start() write_dquot()
871a2931 4926 * dquot_initialize() down(dqio_mutex)
dab291af 4927 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4928 *
4929 */
4930
4931#ifdef CONFIG_QUOTA
4932
4933static inline struct inode *dquot_to_inode(struct dquot *dquot)
4934{
4c376dca 4935 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4936}
4937
617ba13b 4938static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4939{
4940 int ret, err;
4941 handle_t *handle;
4942 struct inode *inode;
4943
4944 inode = dquot_to_inode(dquot);
9924a92a 4945 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 4946 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4947 if (IS_ERR(handle))
4948 return PTR_ERR(handle);
4949 ret = dquot_commit(dquot);
617ba13b 4950 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4951 if (!ret)
4952 ret = err;
4953 return ret;
4954}
4955
617ba13b 4956static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4957{
4958 int ret, err;
4959 handle_t *handle;
4960
9924a92a 4961 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4962 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4963 if (IS_ERR(handle))
4964 return PTR_ERR(handle);
4965 ret = dquot_acquire(dquot);
617ba13b 4966 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4967 if (!ret)
4968 ret = err;
4969 return ret;
4970}
4971
617ba13b 4972static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4973{
4974 int ret, err;
4975 handle_t *handle;
4976
9924a92a 4977 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4978 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4979 if (IS_ERR(handle)) {
4980 /* Release dquot anyway to avoid endless cycle in dqput() */
4981 dquot_release(dquot);
ac27a0ec 4982 return PTR_ERR(handle);
9c3013e9 4983 }
ac27a0ec 4984 ret = dquot_release(dquot);
617ba13b 4985 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4986 if (!ret)
4987 ret = err;
4988 return ret;
4989}
4990
617ba13b 4991static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4992{
262b4662
JK		 4993 struct super_block *sb = dquot->dq_sb;
4994 struct ext4_sb_info *sbi = EXT4_SB(sb);
4995
2c8be6b2 4996 /* Are we journaling quotas? */
262b4662
JK		 4997 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4998 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 4999 dquot_mark_dquot_dirty(dquot);
617ba13b 5000 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5001 } else {
5002 return dquot_mark_dquot_dirty(dquot);
5003 }
5004}
5005
617ba13b 5006static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5007{
5008 int ret, err;
5009 handle_t *handle;
5010
5011 /* Data block + inode block */
9924a92a 5012 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5013 if (IS_ERR(handle))
5014 return PTR_ERR(handle);
5015 ret = dquot_commit_info(sb, type);
617ba13b 5016 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5017 if (!ret)
5018 ret = err;
5019 return ret;
5020}
5021
5022/*
5023 * Turn on quotas during mount time - we need to find
5024 * the quota file and such...
5025 */
617ba13b 5026static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5027{
287a8095
CH		 5028 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5029 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5030}
5031
5032/*
5033 * Standard function to be called on quota_on
5034 */
617ba13b 5035static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5036 struct path *path)
ac27a0ec
DK		 5037{
5038 int err;
ac27a0ec
DK		 5039
5040 if (!test_opt(sb, QUOTA))
5041 return -EINVAL;
0623543b 5042
ac27a0ec 5043 /* Quotafile not on the same filesystem? */
d8c9584e 5044 if (path->dentry->d_sb != sb)
ac27a0ec 5045 return -EXDEV;
0623543b
JK		 5046 /* Journaling quota? */
5047 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5048 /* Quotafile not in fs root? */
f00c9e44 5049 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5050 ext4_msg(sb, KERN_WARNING,
5051 "Quota file not on filesystem root. "
5052 "Journaled quota will not work");
2b2d6d01 5053 }
0623543b
JK		 5054
5055 /*
5056 * When we journal data on quota file, we have to flush journal to see
5057 * all updates to the file when we bypass pagecache...
5058 */
0390131b 5059 if (EXT4_SB(sb)->s_journal &&
f00c9e44 5060 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 5061 /*
5062 * We don't need to lock updates but journal_flush() could
5063 * otherwise be livelocked...
5064 */
5065 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5066 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5067 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5068 if (err)
7ffe1ea8 5069 return err;
0623543b
JK		 5070 }
5071
f00c9e44 5072 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5073}
5074
7c319d32
AK		 5075static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5076 unsigned int flags)
5077{
5078 int err;
5079 struct inode *qf_inode;
5080 unsigned long qf_inums[MAXQUOTAS] = {
5081 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5082 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5083 };
5084
5085 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5086
5087 if (!qf_inums[type])
5088 return -EPERM;
5089
5090 qf_inode = ext4_iget(sb, qf_inums[type]);
5091 if (IS_ERR(qf_inode)) {
5092 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5093 return PTR_ERR(qf_inode);
5094 }
5095
bcb13850
JK		 5096 /* Don't account quota for quota files to avoid recursion */
5097 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5098 err = dquot_enable(qf_inode, type, format_id, flags);
5099 iput(qf_inode);
5100
5101 return err;
5102}
5103
5104/* Enable usage tracking for all quota types. */
5105static int ext4_enable_quotas(struct super_block *sb)
5106{
5107 int type, err = 0;
5108 unsigned long qf_inums[MAXQUOTAS] = {
5109 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5110 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5111 };
5112
5113 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5114 for (type = 0; type < MAXQUOTAS; type++) {
5115 if (qf_inums[type]) {
5116 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5117 DQUOT_USAGE_ENABLED);
5118 if (err) {
5119 ext4_warning(sb,
72ba7450
TT		 5120 "Failed to enable quota tracking "
5121 "(type=%d, err=%d). Please run "
5122 "e2fsck to fix.", type, err);
7c319d32
AK		 5123 return err;
5124 }
5125 }
5126 }
5127 return 0;
5128}
5129
5130/*
5131 * quota_on function that is used when QUOTA feature is set.
5132 */
5133static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5134 int format_id)
5135{
5136 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5137 return -EINVAL;
5138
5139 /*
5140 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5141 */
5142 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5143}
5144
ca0e05e4
DM		 5145static int ext4_quota_off(struct super_block *sb, int type)
5146{
21f97697
JK		 5147 struct inode *inode = sb_dqopt(sb)->files[type];
5148 handle_t *handle;
5149
87009d86
DM		 5150 /* Force all delayed allocation blocks to be allocated.
5151 * Caller already holds s_umount sem */
5152 if (test_opt(sb, DELALLOC))
ca0e05e4 5153 sync_filesystem(sb);
ca0e05e4 5154
0b268590
AG		 5155 if (!inode)
5156 goto out;
5157
21f97697
JK		 5158 /* Update modification times of quota files when userspace can
5159 * start looking at them */
9924a92a 5160 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5161 if (IS_ERR(handle))
5162 goto out;
5163 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5164 ext4_mark_inode_dirty(handle, inode);
5165 ext4_journal_stop(handle);
5166
5167out:
ca0e05e4
DM		 5168 return dquot_quota_off(sb, type);
5169}
5170
7c319d32
AK		 5171/*
5172 * quota_off function that is used when QUOTA feature is set.
5173 */
5174static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5175{
5176 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5177 return -EINVAL;
5178
5179 /* Disable only the limits. */
5180 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5181}
5182
ac27a0ec
DK		 5183/* Read data from quotafile - avoid pagecache and such because we cannot afford
5184 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5185 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5186 * we don't have to be afraid of races */
617ba13b 5187static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5188 size_t len, loff_t off)
5189{
5190 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5191 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5192 int err = 0;
5193 int offset = off & (sb->s_blocksize - 1);
5194 int tocopy;
5195 size_t toread;
5196 struct buffer_head *bh;
5197 loff_t i_size = i_size_read(inode);
5198
5199 if (off > i_size)
5200 return 0;
5201 if (off+len > i_size)
5202 len = i_size-off;
5203 toread = len;
5204 while (toread > 0) {
5205 tocopy = sb->s_blocksize - offset < toread ?
5206 sb->s_blocksize - offset : toread;
617ba13b 5207 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5208 if (err)
5209 return err;
5210 if (!bh) /* A hole? */
5211 memset(data, 0, tocopy);
5212 else
5213 memcpy(data, bh->b_data+offset, tocopy);
5214 brelse(bh);
5215 offset = 0;
5216 toread -= tocopy;
5217 data += tocopy;
5218 blk++;
5219 }
5220 return len;
5221}
5222
5223/* Write to quotafile (we know the transaction is already started and has
5224 * enough credits) */
617ba13b 5225static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5226 const char *data, size_t len, loff_t off)
5227{
5228 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5229 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5230 int err = 0;
5231 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5232 struct buffer_head *bh;
5233 handle_t *handle = journal_current_handle();
5234
0390131b 5235 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5236 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5237 " cancelled because transaction is not started",
9c3013e9
JK		 5238 (unsigned long long)off, (unsigned long long)len);
5239 return -EIO;
5240 }
67eeb568
DM		 5241 /*
5242 * Since we account only one data block in transaction credits,
5243 * then it is impossible to cross a block boundary.
5244 */
5245 if (sb->s_blocksize - offset < len) {
5246 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5247 " cancelled because not block aligned",
5248 (unsigned long long)off, (unsigned long long)len);
5249 return -EIO;
5250 }
5251
67eeb568
DM		 5252 bh = ext4_bread(handle, inode, blk, 1, &err);
5253 if (!bh)
5254 goto out;
62d2b5f2
JK		 5255 err = ext4_journal_get_write_access(handle, bh);
5256 if (err) {
5257 brelse(bh);
5258 goto out;
ac27a0ec 5259 }
67eeb568
DM		 5260 lock_buffer(bh);
5261 memcpy(bh->b_data+offset, data, len);
5262 flush_dcache_page(bh->b_page);
5263 unlock_buffer(bh);
62d2b5f2 5264 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5265 brelse(bh);
ac27a0ec 5266out:
0b7f7cef 5267 if (err)
ac27a0ec 5268 return err;
67eeb568
DM		 5269 if (inode->i_size < off + len) {
5270 i_size_write(inode, off + len);
617ba13b 5271 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5272 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5273 }
67eeb568 5274 return len;
ac27a0ec
DK		 5275}
5276
5277#endif
5278
152a0836
AV		 5279static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5280 const char *dev_name, void *data)
ac27a0ec 5281{
152a0836 5282 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5283}
5284
37f328eb 5285#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5286static inline void register_as_ext2(void)
5287{
5288 int err = register_filesystem(&ext2_fs_type);
5289 if (err)
5290 printk(KERN_WARNING
5291 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5292}
5293
5294static inline void unregister_as_ext2(void)
5295{
5296 unregister_filesystem(&ext2_fs_type);
5297}
2035e776
TT		 5298
5299static inline int ext2_feature_set_ok(struct super_block *sb)
5300{
5301 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5302 return 0;
5303 if (sb->s_flags & MS_RDONLY)
5304 return 1;
5305 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5306 return 0;
5307 return 1;
5308}
24b58424
TT		 5309#else
5310static inline void register_as_ext2(void) { }
5311static inline void unregister_as_ext2(void) { }
2035e776 5312static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5313#endif
5314
37f328eb 5315#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5316static inline void register_as_ext3(void)
5317{
5318 int err = register_filesystem(&ext3_fs_type);
5319 if (err)
5320 printk(KERN_WARNING
5321 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5322}
5323
5324static inline void unregister_as_ext3(void)
5325{
5326 unregister_filesystem(&ext3_fs_type);
5327}
2035e776
TT		 5328
5329static inline int ext3_feature_set_ok(struct super_block *sb)
5330{
5331 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5332 return 0;
5333 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5334 return 0;
5335 if (sb->s_flags & MS_RDONLY)
5336 return 1;
5337 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5338 return 0;
5339 return 1;
5340}
24b58424
TT		 5341#else
5342static inline void register_as_ext3(void) { }
5343static inline void unregister_as_ext3(void) { }
2035e776 5344static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5345#endif
5346
03010a33
TT		 5347static struct file_system_type ext4_fs_type = {
5348 .owner = THIS_MODULE,
5349 .name = "ext4",
152a0836 5350 .mount = ext4_mount,
03010a33
TT		 5351 .kill_sb = kill_block_super,
5352 .fs_flags = FS_REQUIRES_DEV,
5353};
7f78e035 5354MODULE_ALIAS_FS("ext4");
03010a33 5355
8f021222 5356static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5357{
5358 struct ext4_features *ef;
5359 int ret = -ENOMEM;
5360
5361 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5362 if (!ef)
5363 goto out;
5364
5365 ef->f_kobj.kset = ext4_kset;
5366 init_completion(&ef->f_kobj_unregister);
5367 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5368 "features");
5369 if (ret) {
5370 kfree(ef);
5371 goto out;
5372 }
5373
5374 ext4_feat = ef;
5375 ret = 0;
5376out:
5377 return ret;
5378}
5379
8f021222
LC		 5380static void ext4_exit_feat_adverts(void)
5381{
5382 kobject_put(&ext4_feat->f_kobj);
5383 wait_for_completion(&ext4_feat->f_kobj_unregister);
5384 kfree(ext4_feat);
5385}
5386
e9e3bcec
ES		 5387/* Shared across all ext4 file systems */
5388wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5389struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5390
5dabfc78 5391static int __init ext4_init_fs(void)
ac27a0ec 5392{
e9e3bcec 5393 int i, err;
c9de560d 5394
07c0c5d8
AV		 5395 ext4_li_info = NULL;
5396 mutex_init(&ext4_li_mtx);
5397
9a4c8019 5398 /* Build-time check for flags consistency */
12e9b892 5399 ext4_check_flag_values();
e9e3bcec
ES		 5400
5401 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5402 mutex_init(&ext4__aio_mutex[i]);
5403 init_waitqueue_head(&ext4__ioend_wq[i]);
5404 }
5405
51865fda 5406 err = ext4_init_es();
6fd058f7
TT		 5407 if (err)
5408 return err;
51865fda
ZL		 5409
5410 err = ext4_init_pageio();
5411 if (err)
5412 goto out7;
5413
5dabfc78 5414 err = ext4_init_system_zone();
bd2d0210 5415 if (err)
d44651d0 5416 goto out6;
3197ebdb 5417 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5418 if (!ext4_kset) {
5419 err = -ENOMEM;
dd68314c 5420 goto out5;
0e376b1e 5421 }
d44651d0 5422 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5423
5424 err = ext4_init_feat_adverts();
dd68314c
TT		 5425 if (err)
5426 goto out4;
857ac889 5427
5dabfc78 5428 err = ext4_init_mballoc();
ac27a0ec 5429 if (err)
6fd058f7 5430 goto out3;
c9de560d 5431
5dabfc78 5432 err = ext4_init_xattr();
c9de560d
AT		 5433 if (err)
5434 goto out2;
ac27a0ec
DK		 5435 err = init_inodecache();
5436 if (err)
5437 goto out1;
24b58424 5438 register_as_ext3();
2035e776 5439 register_as_ext2();
03010a33 5440 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5441 if (err)
5442 goto out;
bfff6873 5443
ac27a0ec
DK		 5444 return 0;
5445out:
24b58424
TT		 5446 unregister_as_ext2();
5447 unregister_as_ext3();
ac27a0ec
DK		 5448 destroy_inodecache();
5449out1:
5dabfc78 5450 ext4_exit_xattr();
c9de560d 5451out2:
5dabfc78 5452 ext4_exit_mballoc();
6fd058f7 5453out3:
8f021222 5454 ext4_exit_feat_adverts();
dd68314c 5455out4:
d44651d0
FJ		 5456 if (ext4_proc_root)
5457 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5458 kset_unregister(ext4_kset);
d44651d0 5459out5:
5dabfc78 5460 ext4_exit_system_zone();
d44651d0 5461out6:
5dabfc78 5462 ext4_exit_pageio();
51865fda
ZL		 5463out7:
5464 ext4_exit_es();
5465
ac27a0ec
DK		 5466 return err;
5467}
5468
5dabfc78 5469static void __exit ext4_exit_fs(void)
ac27a0ec 5470{
bfff6873 5471 ext4_destroy_lazyinit_thread();
24b58424
TT		 5472 unregister_as_ext2();
5473 unregister_as_ext3();
03010a33 5474 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5475 destroy_inodecache();
5dabfc78
TT		 5476 ext4_exit_xattr();
5477 ext4_exit_mballoc();
8f021222 5478 ext4_exit_feat_adverts();
9f6200bb 5479 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5480 kset_unregister(ext4_kset);
5dabfc78
TT		 5481 ext4_exit_system_zone();
5482 ext4_exit_pageio();
dd12ed14 5483 ext4_exit_es();
ac27a0ec
DK		 5484}
5485
5486MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5487MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5488MODULE_LICENSE("GPL");
5dabfc78
TT		 5489module_init(ext4_init_fs)
5490module_exit(ext4_exit_fs)
Linux 6.x block layer and io_uring tree(s)