projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: make sure group number is bumped after a inode allocation race
[GitHub/LineageOS/android_kernel_motorola_exynos9610.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
JK		 765 flush_workqueue(sbi->unrsv_conversion_wq);
766 flush_workqueue(sbi->rsv_conversion_wq);
767 destroy_workqueue(sbi->unrsv_conversion_wq);
768 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 769
0390131b
FM		 770 if (sbi->s_journal) {
771 err = jbd2_journal_destroy(sbi->s_journal);
772 sbi->s_journal = NULL;
773 if (err < 0)
c67d859e 774 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 775 }
d4edac31 776
d3922a77 777 ext4_es_unregister_shrinker(sbi);
a1c6c569 778 del_timer(&sbi->s_err_report);
d4edac31
JB		 779 ext4_release_system_zone(sb);
780 ext4_mb_release(sb);
781 ext4_ext_release(sb);
782 ext4_xattr_put_super(sb);
783
ac27a0ec 784 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 785 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 786 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 787 }
58c5873a 788 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 789 ext4_commit_super(sb, 1);
790
240799cd 791 if (sbi->s_proc) {
66acdcf4 792 remove_proc_entry("options", sbi->s_proc);
9f6200bb 793 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 794 }
3197ebdb 795 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 796
797 for (i = 0; i < sbi->s_gdb_count; i++)
798 brelse(sbi->s_group_desc[i]);
f18a5f21 799 ext4_kvfree(sbi->s_group_desc);
9933fc0a 800 ext4_kvfree(sbi->s_flex_groups);
57042651 801 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 802 percpu_counter_destroy(&sbi->s_freeinodes_counter);
803 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 804 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 805 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
ac27a0ec
DK		 806 brelse(sbi->s_sbh);
807#ifdef CONFIG_QUOTA
808 for (i = 0; i < MAXQUOTAS; i++)
809 kfree(sbi->s_qf_names[i]);
810#endif
811
812 /* Debugging code just in case the in-memory inode orphan list
813 * isn't empty. The on-disk one can be non-empty if we've
814 * detected an error and taken the fs readonly, but the
815 * in-memory list had better be clean by this point. */
816 if (!list_empty(&sbi->s_orphan))
817 dump_orphan_list(sb, sbi);
818 J_ASSERT(list_empty(&sbi->s_orphan));
819
f98393a6 820 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 821 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
822 /*
823 * Invalidate the journal device's buffers. We don't want them
824 * floating about in memory - the physical journal device may
825 * hotswapped, and it breaks the `ro-after' testing code.
826 */
827 sync_blockdev(sbi->journal_bdev);
f98393a6 828 invalidate_bdev(sbi->journal_bdev);
617ba13b 829 ext4_blkdev_remove(sbi);
ac27a0ec 830 }
c5e06d10
JL		 831 if (sbi->s_mmp_tsk)
832 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 833 sb->s_fs_info = NULL;
3197ebdb
TT		 834 /*
835 * Now that we are completely done shutting down the
836 * superblock, we need to actually destroy the kobject.
837 */
3197ebdb
TT		 838 kobject_put(&sbi->s_kobj);
839 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 840 if (sbi->s_chksum_driver)
841 crypto_free_shash(sbi->s_chksum_driver);
705895b6 842 kfree(sbi->s_blockgroup_lock);
ac27a0ec 843 kfree(sbi);
ac27a0ec
DK		 844}
845
e18b890b 846static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 847
848/*
849 * Called inside transaction, so use GFP_NOFS
850 */
617ba13b 851static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 852{
617ba13b 853 struct ext4_inode_info *ei;
ac27a0ec 854
e6b4f8da 855 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 856 if (!ei)
857 return NULL;
0b8e58a1 858
ac27a0ec 859 ei->vfs_inode.i_version = 1;
c9de560d
AT		 860 INIT_LIST_HEAD(&ei->i_prealloc_list);
861 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 862 ext4_es_init_tree(&ei->i_es_tree);
863 rwlock_init(&ei->i_es_lock);
74cd15cd
ZL		 864 INIT_LIST_HEAD(&ei->i_es_lru);
865 ei->i_es_lru_nr = 0;
d3922a77 866 ei->i_touch_when = 0;
d2a17637
MC		 867 ei->i_reserved_data_blocks = 0;
868 ei->i_reserved_meta_blocks = 0;
869 ei->i_allocated_meta_blocks = 0;
9d0be502 870 ei->i_da_metadata_calc_len = 0;
7e731bc9 871 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 872 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 873#ifdef CONFIG_QUOTA
874 ei->i_reserved_quota = 0;
875#endif
8aefcd55 876 ei->jinode = NULL;
2e8fa54e
JK		 877 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
878 INIT_LIST_HEAD(&ei->i_unrsv_conversion_list);
744692dc 879 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 880 ei->i_sync_tid = 0;
881 ei->i_datasync_tid = 0;
f7ad6d2e 882 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 883 atomic_set(&ei->i_unwritten, 0);
2e8fa54e
JK		 884 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
885 INIT_WORK(&ei->i_unrsv_conversion_work, ext4_end_io_unrsv_work);
0b8e58a1 886
ac27a0ec
DK		 887 return &ei->vfs_inode;
888}
889
7ff9c073
TT		 890static int ext4_drop_inode(struct inode *inode)
891{
892 int drop = generic_drop_inode(inode);
893
894 trace_ext4_drop_inode(inode, drop);
895 return drop;
896}
897
fa0d7e3d
NP		 898static void ext4_i_callback(struct rcu_head *head)
899{
900 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 901 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
902}
903
617ba13b 904static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 905{
9f7dd93d 906 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 907 ext4_msg(inode->i_sb, KERN_ERR,
908 "Inode %lu (%p): orphan list check failed!",
909 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 910 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
911 EXT4_I(inode), sizeof(struct ext4_inode_info),
912 true);
913 dump_stack();
914 }
fa0d7e3d 915 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 916}
917
51cc5068 918static void init_once(void *foo)
ac27a0ec 919{
617ba13b 920 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 921
a35afb83 922 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 923 init_rwsem(&ei->xattr_sem);
0e855ac8 924 init_rwsem(&ei->i_data_sem);
a35afb83 925 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 926}
927
928static int init_inodecache(void)
929{
617ba13b
MC		 930 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
931 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 932 0, (SLAB_RECLAIM_ACCOUNT|
933 SLAB_MEM_SPREAD),
20c2df83 934 init_once);
617ba13b 935 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 936 return -ENOMEM;
937 return 0;
938}
939
940static void destroy_inodecache(void)
941{
8c0a8537
KS		 942 /*
943 * Make sure all delayed rcu free inodes are flushed before we
944 * destroy cache.
945 */
946 rcu_barrier();
617ba13b 947 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 948}
949
0930fcc1 950void ext4_clear_inode(struct inode *inode)
ac27a0ec 951{
0930fcc1 952 invalidate_inode_buffers(inode);
dbd5768f 953 clear_inode(inode);
9f754758 954 dquot_drop(inode);
c2ea3fde 955 ext4_discard_preallocations(inode);
51865fda 956 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
74cd15cd 957 ext4_es_lru_del(inode);
8aefcd55
TT		 958 if (EXT4_I(inode)->jinode) {
959 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
960 EXT4_I(inode)->jinode);
961 jbd2_free_inode(EXT4_I(inode)->jinode);
962 EXT4_I(inode)->jinode = NULL;
963 }
ac27a0ec
DK		 964}
965
1b961ac0 966static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 967 u64 ino, u32 generation)
ac27a0ec 968{
ac27a0ec 969 struct inode *inode;
ac27a0ec 970
617ba13b 971 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 972 return ERR_PTR(-ESTALE);
617ba13b 973 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 974 return ERR_PTR(-ESTALE);
975
976 /* iget isn't really right if the inode is currently unallocated!!
977 *
617ba13b 978 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 979 * deleted, so we should be safe.
980 *
981 * Currently we don't know the generation for parent directory, so
982 * a generation of 0 means "accept any"
983 */
1d1fe1ee
DH		 984 inode = ext4_iget(sb, ino);
985 if (IS_ERR(inode))
986 return ERR_CAST(inode);
987 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 988 iput(inode);
989 return ERR_PTR(-ESTALE);
990 }
1b961ac0
CH		 991
992 return inode;
993}
994
995static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 996 int fh_len, int fh_type)
1b961ac0
CH		 997{
998 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
999 ext4_nfs_get_inode);
1000}
1001
1002static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1003 int fh_len, int fh_type)
1b961ac0
CH		 1004{
1005 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1006 ext4_nfs_get_inode);
ac27a0ec
DK		 1007}
1008
c39a7f84
TO		 1009/*
1010 * Try to release metadata pages (indirect blocks, directories) which are
1011 * mapped via the block device. Since these pages could have journal heads
1012 * which would prevent try_to_free_buffers() from freeing them, we must use
1013 * jbd2 layer's try_to_free_buffers() function to release them.
1014 */
0b8e58a1
AD		 1015static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1016 gfp_t wait)
c39a7f84
TO		 1017{
1018 journal_t *journal = EXT4_SB(sb)->s_journal;
1019
1020 WARN_ON(PageChecked(page));
1021 if (!page_has_buffers(page))
1022 return 0;
1023 if (journal)
1024 return jbd2_journal_try_to_free_buffers(journal, page,
1025 wait & ~__GFP_WAIT);
1026 return try_to_free_buffers(page);
1027}
1028
ac27a0ec 1029#ifdef CONFIG_QUOTA
af5bc92d 1030#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1031#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1032
617ba13b
MC		 1033static int ext4_write_dquot(struct dquot *dquot);
1034static int ext4_acquire_dquot(struct dquot *dquot);
1035static int ext4_release_dquot(struct dquot *dquot);
1036static int ext4_mark_dquot_dirty(struct dquot *dquot);
1037static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1038static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1039 struct path *path);
7c319d32
AK		 1040static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1041 int format_id);
ca0e05e4 1042static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1043static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1044static int ext4_quota_on_mount(struct super_block *sb, int type);
1045static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1046 size_t len, loff_t off);
617ba13b 1047static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1048 const char *data, size_t len, loff_t off);
7c319d32
AK		 1049static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1050 unsigned int flags);
1051static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1052
61e225dc 1053static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1054 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1055 .write_dquot = ext4_write_dquot,
1056 .acquire_dquot = ext4_acquire_dquot,
1057 .release_dquot = ext4_release_dquot,
1058 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1059 .write_info = ext4_write_info,
1060 .alloc_dquot = dquot_alloc,
1061 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1062};
1063
0d54b217 1064static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1065 .quota_on = ext4_quota_on,
ca0e05e4 1066 .quota_off = ext4_quota_off,
287a8095
CH		 1067 .quota_sync = dquot_quota_sync,
1068 .get_info = dquot_get_dqinfo,
1069 .set_info = dquot_set_dqinfo,
1070 .get_dqblk = dquot_get_dqblk,
1071 .set_dqblk = dquot_set_dqblk
ac27a0ec 1072};
7c319d32
AK		 1073
1074static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1075 .quota_on_meta = ext4_quota_on_sysfile,
1076 .quota_off = ext4_quota_off_sysfile,
1077 .quota_sync = dquot_quota_sync,
1078 .get_info = dquot_get_dqinfo,
1079 .set_info = dquot_set_dqinfo,
1080 .get_dqblk = dquot_get_dqblk,
1081 .set_dqblk = dquot_set_dqblk
1082};
ac27a0ec
DK		 1083#endif
1084
ee9b6d61 1085static const struct super_operations ext4_sops = {
617ba13b
MC		 1086 .alloc_inode = ext4_alloc_inode,
1087 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1088 .write_inode = ext4_write_inode,
1089 .dirty_inode = ext4_dirty_inode,
7ff9c073 1090 .drop_inode = ext4_drop_inode,
0930fcc1 1091 .evict_inode = ext4_evict_inode,
617ba13b 1092 .put_super = ext4_put_super,
617ba13b 1093 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1094 .freeze_fs = ext4_freeze,
1095 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1096 .statfs = ext4_statfs,
1097 .remount_fs = ext4_remount,
617ba13b 1098 .show_options = ext4_show_options,
ac27a0ec 1099#ifdef CONFIG_QUOTA
617ba13b
MC		 1100 .quota_read = ext4_quota_read,
1101 .quota_write = ext4_quota_write,
ac27a0ec 1102#endif
c39a7f84 1103 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1104};
1105
9ca92389
TT		 1106static const struct super_operations ext4_nojournal_sops = {
1107 .alloc_inode = ext4_alloc_inode,
1108 .destroy_inode = ext4_destroy_inode,
1109 .write_inode = ext4_write_inode,
1110 .dirty_inode = ext4_dirty_inode,
7ff9c073 1111 .drop_inode = ext4_drop_inode,
0930fcc1 1112 .evict_inode = ext4_evict_inode,
06a407f1 1113 .sync_fs = ext4_sync_fs_nojournal,
9ca92389
TT		 1114 .put_super = ext4_put_super,
1115 .statfs = ext4_statfs,
1116 .remount_fs = ext4_remount,
9ca92389
TT		 1117 .show_options = ext4_show_options,
1118#ifdef CONFIG_QUOTA
1119 .quota_read = ext4_quota_read,
1120 .quota_write = ext4_quota_write,
1121#endif
1122 .bdev_try_to_free_page = bdev_try_to_free_page,
1123};
1124
39655164 1125static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1126 .fh_to_dentry = ext4_fh_to_dentry,
1127 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1128 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1129};
1130
1131enum {
1132 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1133 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1134 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1135 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1136 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1137 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1138 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1139 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1140 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1141 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1142 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1143 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1144 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1145 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1146 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1147 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1148 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1149 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1150 Opt_max_dir_size_kb,
ac27a0ec
DK		 1151};
1152
a447c093 1153static const match_table_t tokens = {
ac27a0ec
DK		 1154 {Opt_bsd_df, "bsddf"},
1155 {Opt_minix_df, "minixdf"},
1156 {Opt_grpid, "grpid"},
1157 {Opt_grpid, "bsdgroups"},
1158 {Opt_nogrpid, "nogrpid"},
1159 {Opt_nogrpid, "sysvgroups"},
1160 {Opt_resgid, "resgid=%u"},
1161 {Opt_resuid, "resuid=%u"},
1162 {Opt_sb, "sb=%u"},
1163 {Opt_err_cont, "errors=continue"},
1164 {Opt_err_panic, "errors=panic"},
1165 {Opt_err_ro, "errors=remount-ro"},
1166 {Opt_nouid32, "nouid32"},
ac27a0ec 1167 {Opt_debug, "debug"},
72578c33
TT		 1168 {Opt_removed, "oldalloc"},
1169 {Opt_removed, "orlov"},
ac27a0ec
DK		 1170 {Opt_user_xattr, "user_xattr"},
1171 {Opt_nouser_xattr, "nouser_xattr"},
1172 {Opt_acl, "acl"},
1173 {Opt_noacl, "noacl"},
e3bb52ae 1174 {Opt_noload, "norecovery"},
5a916be1 1175 {Opt_noload, "noload"},
72578c33
TT		 1176 {Opt_removed, "nobh"},
1177 {Opt_removed, "bh"},
ac27a0ec 1178 {Opt_commit, "commit=%u"},
30773840
TT		 1179 {Opt_min_batch_time, "min_batch_time=%u"},
1180 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1181 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1182 {Opt_journal_checksum, "journal_checksum"},
1183 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1184 {Opt_abort, "abort"},
1185 {Opt_data_journal, "data=journal"},
1186 {Opt_data_ordered, "data=ordered"},
1187 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1188 {Opt_data_err_abort, "data_err=abort"},
1189 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1190 {Opt_offusrjquota, "usrjquota="},
1191 {Opt_usrjquota, "usrjquota=%s"},
1192 {Opt_offgrpjquota, "grpjquota="},
1193 {Opt_grpjquota, "grpjquota=%s"},
1194 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1195 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1196 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1197 {Opt_grpquota, "grpquota"},
1198 {Opt_noquota, "noquota"},
1199 {Opt_quota, "quota"},
1200 {Opt_usrquota, "usrquota"},
1201 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1202 {Opt_barrier, "barrier"},
1203 {Opt_nobarrier, "nobarrier"},
25ec56b5 1204 {Opt_i_version, "i_version"},
c9de560d 1205 {Opt_stripe, "stripe=%u"},
64769240 1206 {Opt_delalloc, "delalloc"},
dd919b98 1207 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1208 {Opt_removed, "mblk_io_submit"},
1209 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1210 {Opt_block_validity, "block_validity"},
1211 {Opt_noblock_validity, "noblock_validity"},
240799cd 1212 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1213 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1214 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1215 {Opt_auto_da_alloc, "auto_da_alloc"},
1216 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1217 {Opt_dioread_nolock, "dioread_nolock"},
1218 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1219 {Opt_discard, "discard"},
1220 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1221 {Opt_init_itable, "init_itable=%u"},
1222 {Opt_init_itable, "init_itable"},
1223 {Opt_noinit_itable, "noinit_itable"},
df981d03 1224 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1225 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1226 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1227 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1228 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1229 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1230 {Opt_err, NULL},
ac27a0ec
DK		 1231};
1232
617ba13b 1233static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1234{
617ba13b 1235 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1236 char *options = (char *) *data;
1237
1238 if (!options || strncmp(options, "sb=", 3) != 0)
1239 return 1; /* Default location */
0b8e58a1 1240
ac27a0ec 1241 options += 3;
0b8e58a1 1242 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1243 sb_block = simple_strtoul(options, &options, 0);
1244 if (*options && *options != ',') {
4776004f 1245 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1246 (char *) *data);
1247 return 1;
1248 }
1249 if (*options == ',')
1250 options++;
1251 *data = (void *) options;
0b8e58a1 1252
ac27a0ec
DK		 1253 return sb_block;
1254}
1255
b3881f74 1256#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1257static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1258 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1259
56c50f11
DM		 1260#ifdef CONFIG_QUOTA
1261static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1262{
1263 struct ext4_sb_info *sbi = EXT4_SB(sb);
1264 char *qname;
03dafb5f 1265 int ret = -1;
56c50f11
DM		 1266
1267 if (sb_any_quota_loaded(sb) &&
1268 !sbi->s_qf_names[qtype]) {
1269 ext4_msg(sb, KERN_ERR,
1270 "Cannot change journaled "
1271 "quota options when quota turned on");
57f73c2c 1272 return -1;
56c50f11 1273 }
262b4662
JK		 1274 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1275 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1276 "when QUOTA feature is enabled");
1277 return -1;
1278 }
56c50f11
DM		 1279 qname = match_strdup(args);
1280 if (!qname) {
1281 ext4_msg(sb, KERN_ERR,
1282 "Not enough memory for storing quotafile name");
57f73c2c 1283 return -1;
56c50f11 1284 }
03dafb5f
CG		 1285 if (sbi->s_qf_names[qtype]) {
1286 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1287 ret = 1;
1288 else
1289 ext4_msg(sb, KERN_ERR,
1290 "%s quota file already specified",
1291 QTYPE2NAME(qtype));
1292 goto errout;
56c50f11 1293 }
03dafb5f 1294 if (strchr(qname, '/')) {
56c50f11
DM		 1295 ext4_msg(sb, KERN_ERR,
1296 "quotafile must be on filesystem root");
03dafb5f 1297 goto errout;
56c50f11 1298 }
03dafb5f 1299 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1300 set_opt(sb, QUOTA);
56c50f11 1301 return 1;
03dafb5f
CG		 1302errout:
1303 kfree(qname);
1304 return ret;
56c50f11
DM		 1305}
1306
1307static int clear_qf_name(struct super_block *sb, int qtype)
1308{
1309
1310 struct ext4_sb_info *sbi = EXT4_SB(sb);
1311
1312 if (sb_any_quota_loaded(sb) &&
1313 sbi->s_qf_names[qtype]) {
1314 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1315 " when quota turned on");
57f73c2c 1316 return -1;
56c50f11 1317 }
03dafb5f 1318 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1319 sbi->s_qf_names[qtype] = NULL;
1320 return 1;
1321}
1322#endif
1323
26092bf5
TT		 1324#define MOPT_SET 0x0001
1325#define MOPT_CLEAR 0x0002
1326#define MOPT_NOSUPPORT 0x0004
1327#define MOPT_EXPLICIT 0x0008
1328#define MOPT_CLEAR_ERR 0x0010
1329#define MOPT_GTE0 0x0020
ac27a0ec 1330#ifdef CONFIG_QUOTA
26092bf5
TT		 1331#define MOPT_Q 0
1332#define MOPT_QFMT 0x0040
1333#else
1334#define MOPT_Q MOPT_NOSUPPORT
1335#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1336#endif
26092bf5 1337#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1338#define MOPT_NO_EXT2 0x0100
1339#define MOPT_NO_EXT3 0x0200
1340#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
26092bf5
TT		 1341
1342static const struct mount_opts {
1343 int token;
1344 int mount_opt;
1345 int flags;
1346} ext4_mount_opts[] = {
1347 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1348 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1349 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1350 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1351 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1352 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1353 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1354 MOPT_EXT4_ONLY | MOPT_SET},
1355 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1356 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1357 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1358 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1359 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1360 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1361 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1362 MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1363 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1364 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1365 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1366 EXT4_MOUNT_JOURNAL_CHECKSUM),
1367 MOPT_EXT4_ONLY | MOPT_SET},
1368 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1369 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1370 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1371 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1372 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1373 MOPT_NO_EXT2 | MOPT_SET},
1374 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1375 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1376 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1377 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1378 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1379 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1380 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1381 {Opt_commit, 0, MOPT_GTE0},
1382 {Opt_max_batch_time, 0, MOPT_GTE0},
1383 {Opt_min_batch_time, 0, MOPT_GTE0},
1384 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1385 {Opt_init_itable, 0, MOPT_GTE0},
1386 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1387 {Opt_resuid, 0, MOPT_GTE0},
1388 {Opt_resgid, 0, MOPT_GTE0},
1389 {Opt_journal_dev, 0, MOPT_GTE0},
1390 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1391 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1392 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1393 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1394 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1395 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1396 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1397#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1398 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1399 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1400#else
26092bf5
TT		 1401 {Opt_acl, 0, MOPT_NOSUPPORT},
1402 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1403#endif
26092bf5
TT		 1404 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1405 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1406 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1407 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1408 MOPT_SET | MOPT_Q},
1409 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1410 MOPT_SET | MOPT_Q},
1411 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1412 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1413 {Opt_usrjquota, 0, MOPT_Q},
1414 {Opt_grpjquota, 0, MOPT_Q},
1415 {Opt_offusrjquota, 0, MOPT_Q},
1416 {Opt_offgrpjquota, 0, MOPT_Q},
1417 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1418 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1419 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1420 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1421 {Opt_err, 0, 0}
1422};
1423
1424static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1425 substring_t *args, unsigned long *journal_devnum,
1426 unsigned int *journal_ioprio, int is_remount)
1427{
1428 struct ext4_sb_info *sbi = EXT4_SB(sb);
1429 const struct mount_opts *m;
08cefc7a
EB		 1430 kuid_t uid;
1431 kgid_t gid;
26092bf5
TT		 1432 int arg = 0;
1433
57f73c2c
TT		 1434#ifdef CONFIG_QUOTA
1435 if (token == Opt_usrjquota)
1436 return set_qf_name(sb, USRQUOTA, &args[0]);
1437 else if (token == Opt_grpjquota)
1438 return set_qf_name(sb, GRPQUOTA, &args[0]);
1439 else if (token == Opt_offusrjquota)
1440 return clear_qf_name(sb, USRQUOTA);
1441 else if (token == Opt_offgrpjquota)
1442 return clear_qf_name(sb, GRPQUOTA);
1443#endif
26092bf5 1444 switch (token) {
f7048605
TT		 1445 case Opt_noacl:
1446 case Opt_nouser_xattr:
1447 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1448 break;
26092bf5
TT		 1449 case Opt_sb:
1450 return 1; /* handled by get_sb_block() */
1451 case Opt_removed:
5f3633e3 1452 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1453 return 1;
26092bf5
TT		 1454 case Opt_abort:
1455 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1456 return 1;
1457 case Opt_i_version:
1458 sb->s_flags |= MS_I_VERSION;
1459 return 1;
26092bf5
TT		 1460 }
1461
5f3633e3
JK		 1462 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1463 if (token == m->token)
1464 break;
1465
1466 if (m->token == Opt_err) {
1467 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1468 "or missing value", opt);
1469 return -1;
1470 }
1471
8dc0aa8c
TT		 1472 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1473 ext4_msg(sb, KERN_ERR,
1474 "Mount option \"%s\" incompatible with ext2", opt);
1475 return -1;
1476 }
1477 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1478 ext4_msg(sb, KERN_ERR,
1479 "Mount option \"%s\" incompatible with ext3", opt);
1480 return -1;
1481 }
1482
5f3633e3
JK		 1483 if (args->from && match_int(args, &arg))
1484 return -1;
1485 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1486 return -1;
1487 if (m->flags & MOPT_EXPLICIT)
1488 set_opt2(sb, EXPLICIT_DELALLOC);
1489 if (m->flags & MOPT_CLEAR_ERR)
1490 clear_opt(sb, ERRORS_MASK);
1491 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1492 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1493 "options when quota turned on");
1494 return -1;
1495 }
1496
1497 if (m->flags & MOPT_NOSUPPORT) {
1498 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1499 } else if (token == Opt_commit) {
1500 if (arg == 0)
1501 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1502 sbi->s_commit_interval = HZ * arg;
1503 } else if (token == Opt_max_batch_time) {
1504 if (arg == 0)
1505 arg = EXT4_DEF_MAX_BATCH_TIME;
1506 sbi->s_max_batch_time = arg;
1507 } else if (token == Opt_min_batch_time) {
1508 sbi->s_min_batch_time = arg;
1509 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1510 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1511 ext4_msg(sb, KERN_ERR,
1512 "EXT4-fs: inode_readahead_blks must be "
1513 "0 or a power of 2 smaller than 2^31");
26092bf5 1514 return -1;
5f3633e3
JK		 1515 }
1516 sbi->s_inode_readahead_blks = arg;
1517 } else if (token == Opt_init_itable) {
1518 set_opt(sb, INIT_INODE_TABLE);
1519 if (!args->from)
1520 arg = EXT4_DEF_LI_WAIT_MULT;
1521 sbi->s_li_wait_mult = arg;
1522 } else if (token == Opt_max_dir_size_kb) {
1523 sbi->s_max_dir_size_kb = arg;
1524 } else if (token == Opt_stripe) {
1525 sbi->s_stripe = arg;
1526 } else if (token == Opt_resuid) {
1527 uid = make_kuid(current_user_ns(), arg);
1528 if (!uid_valid(uid)) {
1529 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1530 return -1;
1531 }
5f3633e3
JK		 1532 sbi->s_resuid = uid;
1533 } else if (token == Opt_resgid) {
1534 gid = make_kgid(current_user_ns(), arg);
1535 if (!gid_valid(gid)) {
1536 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1537 return -1;
1538 }
1539 sbi->s_resgid = gid;
1540 } else if (token == Opt_journal_dev) {
1541 if (is_remount) {
1542 ext4_msg(sb, KERN_ERR,
1543 "Cannot specify journal on remount");
1544 return -1;
1545 }
1546 *journal_devnum = arg;
1547 } else if (token == Opt_journal_ioprio) {
1548 if (arg > 7) {
1549 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1550 " (must be 0-7)");
1551 return -1;
1552 }
1553 *journal_ioprio =
1554 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1555 } else if (m->flags & MOPT_DATAJ) {
1556 if (is_remount) {
1557 if (!sbi->s_journal)
1558 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1559 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1560 ext4_msg(sb, KERN_ERR,
26092bf5 1561 "Cannot change data mode on remount");
26092bf5 1562 return -1;
ac27a0ec 1563 }
26092bf5 1564 } else {
5f3633e3
JK		 1565 clear_opt(sb, DATA_FLAGS);
1566 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1567 }
5f3633e3
JK		 1568#ifdef CONFIG_QUOTA
1569 } else if (m->flags & MOPT_QFMT) {
1570 if (sb_any_quota_loaded(sb) &&
1571 sbi->s_jquota_fmt != m->mount_opt) {
1572 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1573 "quota options when quota turned on");
1574 return -1;
1575 }
262b4662
JK		 1576 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1577 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1578 ext4_msg(sb, KERN_ERR,
1579 "Cannot set journaled quota options "
1580 "when QUOTA feature is enabled");
1581 return -1;
1582 }
5f3633e3
JK		 1583 sbi->s_jquota_fmt = m->mount_opt;
1584#endif
1585 } else {
1586 if (!args->from)
1587 arg = 1;
1588 if (m->flags & MOPT_CLEAR)
1589 arg = !arg;
1590 else if (unlikely(!(m->flags & MOPT_SET))) {
1591 ext4_msg(sb, KERN_WARNING,
1592 "buggy handling of option %s", opt);
1593 WARN_ON(1);
1594 return -1;
1595 }
1596 if (arg != 0)
1597 sbi->s_mount_opt |= m->mount_opt;
1598 else
1599 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1600 }
5f3633e3 1601 return 1;
26092bf5
TT		 1602}
1603
1604static int parse_options(char *options, struct super_block *sb,
1605 unsigned long *journal_devnum,
1606 unsigned int *journal_ioprio,
1607 int is_remount)
1608{
1609 struct ext4_sb_info *sbi = EXT4_SB(sb);
1610 char *p;
1611 substring_t args[MAX_OPT_ARGS];
1612 int token;
1613
1614 if (!options)
1615 return 1;
1616
1617 while ((p = strsep(&options, ",")) != NULL) {
1618 if (!*p)
1619 continue;
1620 /*
1621 * Initialize args struct so we know whether arg was
1622 * found; some options take optional arguments.
1623 */
caecd0af 1624 args[0].to = args[0].from = NULL;
26092bf5
TT		 1625 token = match_token(p, tokens, args);
1626 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1627 journal_ioprio, is_remount) < 0)
1628 return 0;
ac27a0ec
DK		 1629 }
1630#ifdef CONFIG_QUOTA
262b4662
JK		 1631 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1632 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1633 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1634 "feature is enabled");
1635 return 0;
1636 }
ac27a0ec 1637 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1638 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1639 clear_opt(sb, USRQUOTA);
ac27a0ec 1640
482a7425 1641 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1642 clear_opt(sb, GRPQUOTA);
ac27a0ec 1643
56c50f11 1644 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1645 ext4_msg(sb, KERN_ERR, "old and new quota "
1646 "format mixing");
ac27a0ec
DK		 1647 return 0;
1648 }
1649
1650 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1651 ext4_msg(sb, KERN_ERR, "journaled quota format "
1652 "not specified");
ac27a0ec
DK		 1653 return 0;
1654 }
1655 } else {
1656 if (sbi->s_jquota_fmt) {
b31e1552 1657 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1658 "specified with no journaling "
b31e1552 1659 "enabled");
ac27a0ec
DK		 1660 return 0;
1661 }
1662 }
1663#endif
261cb20c
JK		 1664 if (test_opt(sb, DIOREAD_NOLOCK)) {
1665 int blocksize =
1666 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1667
1668 if (blocksize < PAGE_CACHE_SIZE) {
1669 ext4_msg(sb, KERN_ERR, "can't mount with "
1670 "dioread_nolock if block size != PAGE_SIZE");
1671 return 0;
1672 }
1673 }
ac27a0ec
DK		 1674 return 1;
1675}
1676
2adf6da8
TT		 1677static inline void ext4_show_quota_options(struct seq_file *seq,
1678 struct super_block *sb)
1679{
1680#if defined(CONFIG_QUOTA)
1681 struct ext4_sb_info *sbi = EXT4_SB(sb);
1682
1683 if (sbi->s_jquota_fmt) {
1684 char *fmtname = "";
1685
1686 switch (sbi->s_jquota_fmt) {
1687 case QFMT_VFS_OLD:
1688 fmtname = "vfsold";
1689 break;
1690 case QFMT_VFS_V0:
1691 fmtname = "vfsv0";
1692 break;
1693 case QFMT_VFS_V1:
1694 fmtname = "vfsv1";
1695 break;
1696 }
1697 seq_printf(seq, ",jqfmt=%s", fmtname);
1698 }
1699
1700 if (sbi->s_qf_names[USRQUOTA])
1701 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1702
1703 if (sbi->s_qf_names[GRPQUOTA])
1704 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1705#endif
1706}
1707
5a916be1
TT		 1708static const char *token2str(int token)
1709{
50df9fd5 1710 const struct match_token *t;
5a916be1
TT		 1711
1712 for (t = tokens; t->token != Opt_err; t++)
1713 if (t->token == token && !strchr(t->pattern, '='))
1714 break;
1715 return t->pattern;
1716}
1717
2adf6da8
TT		 1718/*
1719 * Show an option if
1720 * - it's set to a non-default value OR
1721 * - if the per-sb default is different from the global default
1722 */
66acdcf4
TT		 1723static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1724 int nodefs)
2adf6da8 1725{
2adf6da8
TT		 1726 struct ext4_sb_info *sbi = EXT4_SB(sb);
1727 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1728 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1729 const struct mount_opts *m;
66acdcf4 1730 char sep = nodefs ? '\n' : ',';
2adf6da8 1731
66acdcf4
TT		 1732#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1733#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1734
1735 if (sbi->s_sb_block != 1)
5a916be1
TT		 1736 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1737
1738 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1739 int want_set = m->flags & MOPT_SET;
1740 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1741 (m->flags & MOPT_CLEAR_ERR))
1742 continue;
66acdcf4 1743 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1744 continue; /* skip if same as the default */
1745 if ((want_set &&
1746 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1747 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1748 continue; /* select Opt_noFoo vs Opt_Foo */
1749 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1750 }
5a916be1 1751
08cefc7a 1752 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1753 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1754 SEQ_OPTS_PRINT("resuid=%u",
1755 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1756 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1757 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1758 SEQ_OPTS_PRINT("resgid=%u",
1759 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1760 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1761 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1762 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1763 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1764 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1765 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1766 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1767 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1768 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1769 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1770 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1771 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1772 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1773 if (sb->s_flags & MS_I_VERSION)
5a916be1 1774 SEQ_OPTS_PUTS("i_version");
66acdcf4 1775 if (nodefs || sbi->s_stripe)
5a916be1 1776 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1777 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1778 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1779 SEQ_OPTS_PUTS("data=journal");
1780 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1781 SEQ_OPTS_PUTS("data=ordered");
1782 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1783 SEQ_OPTS_PUTS("data=writeback");
1784 }
66acdcf4
TT		 1785 if (nodefs ||
1786 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1787 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1788 sbi->s_inode_readahead_blks);
2adf6da8 1789
66acdcf4
TT		 1790 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1791 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1792 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1793 if (nodefs || sbi->s_max_dir_size_kb)
1794 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1795
1796 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1797 return 0;
1798}
1799
66acdcf4
TT		 1800static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1801{
1802 return _ext4_show_options(seq, root->d_sb, 0);
1803}
1804
1805static int options_seq_show(struct seq_file *seq, void *offset)
1806{
1807 struct super_block *sb = seq->private;
1808 int rc;
1809
1810 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1811 rc = _ext4_show_options(seq, sb, 1);
1812 seq_puts(seq, "\n");
1813 return rc;
1814}
1815
1816static int options_open_fs(struct inode *inode, struct file *file)
1817{
d9dda78b 1818 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1819}
1820
1821static const struct file_operations ext4_seq_options_fops = {
1822 .owner = THIS_MODULE,
1823 .open = options_open_fs,
1824 .read = seq_read,
1825 .llseek = seq_lseek,
1826 .release = single_release,
1827};
1828
617ba13b 1829static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1830 int read_only)
1831{
617ba13b 1832 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1833 int res = 0;
1834
617ba13b 1835 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1836 ext4_msg(sb, KERN_ERR, "revision level too high, "
1837 "forcing read-only mode");
ac27a0ec
DK		 1838 res = MS_RDONLY;
1839 }
1840 if (read_only)
281b5995 1841 goto done;
617ba13b 1842 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1843 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1844 "running e2fsck is recommended");
617ba13b 1845 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1846 ext4_msg(sb, KERN_WARNING,
1847 "warning: mounting fs with errors, "
1848 "running e2fsck is recommended");
ed3ce80a 1849 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1850 le16_to_cpu(es->s_mnt_count) >=
1851 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1852 ext4_msg(sb, KERN_WARNING,
1853 "warning: maximal mount count reached, "
1854 "running e2fsck is recommended");
ac27a0ec
DK		 1855 else if (le32_to_cpu(es->s_checkinterval) &&
1856 (le32_to_cpu(es->s_lastcheck) +
1857 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1858 ext4_msg(sb, KERN_WARNING,
1859 "warning: checktime reached, "
1860 "running e2fsck is recommended");
0b8e58a1 1861 if (!sbi->s_journal)
0390131b 1862 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1863 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1864 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1865 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1866 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1867 ext4_update_dynamic_rev(sb);
0390131b
FM		 1868 if (sbi->s_journal)
1869 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1870
e2d67052 1871 ext4_commit_super(sb, 1);
281b5995 1872done:
ac27a0ec 1873 if (test_opt(sb, DEBUG))
a9df9a49 1874 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1875 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1876 sb->s_blocksize,
1877 sbi->s_groups_count,
617ba13b
MC		 1878 EXT4_BLOCKS_PER_GROUP(sb),
1879 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1880 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1881
7abc52c2 1882 cleancache_init_fs(sb);
ac27a0ec
DK		 1883 return res;
1884}
1885
117fff10
TT		 1886int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1887{
1888 struct ext4_sb_info *sbi = EXT4_SB(sb);
1889 struct flex_groups *new_groups;
1890 int size;
1891
1892 if (!sbi->s_log_groups_per_flex)
1893 return 0;
1894
1895 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1896 if (size <= sbi->s_flex_groups_allocated)
1897 return 0;
1898
1899 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1900 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1901 if (!new_groups) {
1902 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1903 size / (int) sizeof(struct flex_groups));
1904 return -ENOMEM;
1905 }
1906
1907 if (sbi->s_flex_groups) {
1908 memcpy(new_groups, sbi->s_flex_groups,
1909 (sbi->s_flex_groups_allocated *
1910 sizeof(struct flex_groups)));
1911 ext4_kvfree(sbi->s_flex_groups);
1912 }
1913 sbi->s_flex_groups = new_groups;
1914 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1915 return 0;
1916}
1917
772cb7c8
JS		 1918static int ext4_fill_flex_info(struct super_block *sb)
1919{
1920 struct ext4_sb_info *sbi = EXT4_SB(sb);
1921 struct ext4_group_desc *gdp = NULL;
772cb7c8 1922 ext4_group_t flex_group;
117fff10 1923 int i, err;
772cb7c8 1924
503358ae 1925 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1926 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1927 sbi->s_log_groups_per_flex = 0;
1928 return 1;
1929 }
1930
117fff10
TT		 1931 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1932 if (err)
9933fc0a 1933 goto failed;
772cb7c8 1934
772cb7c8 1935 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1936 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1937
1938 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1939 atomic_add(ext4_free_inodes_count(sb, gdp),
1940 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 1941 atomic64_add(ext4_free_group_clusters(sb, gdp),
1942 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1943 atomic_add(ext4_used_dirs_count(sb, gdp),
1944 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1945 }
1946
1947 return 1;
1948failed:
1949 return 0;
1950}
1951
feb0ab32
DW		 1952static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1953 struct ext4_group_desc *gdp)
717d50e4 1954{
feb0ab32 1955 int offset;
717d50e4 1956 __u16 crc = 0;
feb0ab32 1957 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1958
feb0ab32
DW		 1959 if ((sbi->s_es->s_feature_ro_compat &
1960 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1961 /* Use new metadata_csum algorithm */
d6a77105 1962 __le16 save_csum;
feb0ab32
DW		 1963 __u32 csum32;
1964
d6a77105 1965 save_csum = gdp->bg_checksum;
feb0ab32
DW		 1966 gdp->bg_checksum = 0;
1967 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1968 sizeof(le_group));
1969 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1970 sbi->s_desc_size);
d6a77105 1971 gdp->bg_checksum = save_csum;
feb0ab32
DW		 1972
1973 crc = csum32 & 0xFFFF;
1974 goto out;
717d50e4
AD		 1975 }
1976
feb0ab32
DW		 1977 /* old crc16 code */
1978 offset = offsetof(struct ext4_group_desc, bg_checksum);
1979
1980 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1981 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1982 crc = crc16(crc, (__u8 *)gdp, offset);
1983 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1984 /* for checksum of struct ext4_group_desc do the rest...*/
1985 if ((sbi->s_es->s_feature_incompat &
1986 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1987 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1988 crc = crc16(crc, (__u8 *)gdp + offset,
1989 le16_to_cpu(sbi->s_es->s_desc_size) -
1990 offset);
1991
1992out:
717d50e4
AD		 1993 return cpu_to_le16(crc);
1994}
1995
feb0ab32 1996int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 1997 struct ext4_group_desc *gdp)
1998{
feb0ab32
DW		 1999 if (ext4_has_group_desc_csum(sb) &&
2000 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2001 block_group, gdp)))
717d50e4
AD		 2002 return 0;
2003
2004 return 1;
2005}
2006
feb0ab32
DW		 2007void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2008 struct ext4_group_desc *gdp)
2009{
2010 if (!ext4_has_group_desc_csum(sb))
2011 return;
2012 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2013}
2014
ac27a0ec 2015/* Called at mount-time, super-block is locked */
bfff6873
LC		 2016static int ext4_check_descriptors(struct super_block *sb,
2017 ext4_group_t *first_not_zeroed)
ac27a0ec 2018{
617ba13b
MC		 2019 struct ext4_sb_info *sbi = EXT4_SB(sb);
2020 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2021 ext4_fsblk_t last_block;
bd81d8ee
LV		 2022 ext4_fsblk_t block_bitmap;
2023 ext4_fsblk_t inode_bitmap;
2024 ext4_fsblk_t inode_table;
ce421581 2025 int flexbg_flag = 0;
bfff6873 2026 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2027
ce421581
JS		 2028 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2029 flexbg_flag = 1;
2030
af5bc92d 2031 ext4_debug("Checking group descriptors");
ac27a0ec 2032
197cd65a
AM		 2033 for (i = 0; i < sbi->s_groups_count; i++) {
2034 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2035
ce421581 2036 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2037 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2038 else
2039 last_block = first_block +
617ba13b 2040 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2041
bfff6873
LC		 2042 if ((grp == sbi->s_groups_count) &&
2043 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2044 grp = i;
2045
8fadc143 2046 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2047 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2048 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2049 "Block bitmap for group %u not in group "
b31e1552 2050 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2051 return 0;
2052 }
8fadc143 2053 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2054 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2055 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2056 "Inode bitmap for group %u not in group "
b31e1552 2057 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2058 return 0;
2059 }
8fadc143 2060 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2061 if (inode_table < first_block ||
2b2d6d01 2062 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2063 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2064 "Inode table for group %u not in group "
b31e1552 2065 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2066 return 0;
2067 }
955ce5f5 2068 ext4_lock_group(sb, i);
feb0ab32 2069 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2070 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2071 "Checksum for group %u failed (%u!=%u)",
2072 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2073 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2074 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2075 ext4_unlock_group(sb, i);
8a266467 2076 return 0;
7ee1ec4c 2077 }
717d50e4 2078 }
955ce5f5 2079 ext4_unlock_group(sb, i);
ce421581
JS		 2080 if (!flexbg_flag)
2081 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2082 }
bfff6873
LC		 2083 if (NULL != first_not_zeroed)
2084 *first_not_zeroed = grp;
ac27a0ec 2085
5dee5437
TT		 2086 ext4_free_blocks_count_set(sbi->s_es,
2087 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2088 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2089 return 1;
2090}
2091
617ba13b 2092/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2093 * the superblock) which were deleted from all directories, but held open by
2094 * a process at the time of a crash. We walk the list and try to delete these
2095 * inodes at recovery time (only with a read-write filesystem).
2096 *
2097 * In order to keep the orphan inode chain consistent during traversal (in
2098 * case of crash during recovery), we link each inode into the superblock
2099 * orphan list_head and handle it the same way as an inode deletion during
2100 * normal operation (which journals the operations for us).
2101 *
2102 * We only do an iget() and an iput() on each inode, which is very safe if we
2103 * accidentally point at an in-use or already deleted inode. The worst that
2104 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2105 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2106 * e2fsck was run on this filesystem, and it must have already done the orphan
2107 * inode cleanup for us, so we can safely abort without any further action.
2108 */
2b2d6d01
TT		 2109static void ext4_orphan_cleanup(struct super_block *sb,
2110 struct ext4_super_block *es)
ac27a0ec
DK		 2111{
2112 unsigned int s_flags = sb->s_flags;
2113 int nr_orphans = 0, nr_truncates = 0;
2114#ifdef CONFIG_QUOTA
2115 int i;
2116#endif
2117 if (!es->s_last_orphan) {
2118 jbd_debug(4, "no orphan inodes to clean up\n");
2119 return;
2120 }
2121
a8f48a95 2122 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2123 ext4_msg(sb, KERN_ERR, "write access "
2124 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2125 return;
2126 }
2127
d39195c3
AG		 2128 /* Check if feature set would not allow a r/w mount */
2129 if (!ext4_feature_set_ok(sb, 0)) {
2130 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2131 "unknown ROCOMPAT features");
2132 return;
2133 }
2134
617ba13b 2135 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2136 /* don't clear list on RO mount w/ errors */
2137 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2138 jbd_debug(1, "Errors on filesystem, "
2139 "clearing orphan list.\n");
c25f9bc6
ES		 2140 es->s_last_orphan = 0;
2141 }
ac27a0ec
DK		 2142 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2143 return;
2144 }
2145
2146 if (s_flags & MS_RDONLY) {
b31e1552 2147 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2148 sb->s_flags &= ~MS_RDONLY;
2149 }
2150#ifdef CONFIG_QUOTA
2151 /* Needed for iput() to work correctly and not trash data */
2152 sb->s_flags |= MS_ACTIVE;
2153 /* Turn on quotas so that they are updated correctly */
2154 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2155 if (EXT4_SB(sb)->s_qf_names[i]) {
2156 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2157 if (ret < 0)
b31e1552
ES		 2158 ext4_msg(sb, KERN_ERR,
2159 "Cannot turn on journaled "
2160 "quota: error %d", ret);
ac27a0ec
DK		 2161 }
2162 }
2163#endif
2164
2165 while (es->s_last_orphan) {
2166 struct inode *inode;
2167
97bd42b9
JB		 2168 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2169 if (IS_ERR(inode)) {
ac27a0ec
DK		 2170 es->s_last_orphan = 0;
2171 break;
2172 }
2173
617ba13b 2174 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2175 dquot_initialize(inode);
ac27a0ec 2176 if (inode->i_nlink) {
566370a2
PT		 2177 if (test_opt(sb, DEBUG))
2178 ext4_msg(sb, KERN_DEBUG,
2179 "%s: truncating inode %lu to %lld bytes",
2180 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2181 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2182 inode->i_ino, inode->i_size);
721e3eba 2183 mutex_lock(&inode->i_mutex);
55f252c9 2184 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2185 ext4_truncate(inode);
721e3eba 2186 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2187 nr_truncates++;
2188 } else {
566370a2
PT		 2189 if (test_opt(sb, DEBUG))
2190 ext4_msg(sb, KERN_DEBUG,
2191 "%s: deleting unreferenced inode %lu",
2192 __func__, inode->i_ino);
ac27a0ec
DK		 2193 jbd_debug(2, "deleting unreferenced inode %lu\n",
2194 inode->i_ino);
2195 nr_orphans++;
2196 }
2197 iput(inode); /* The delete magic happens here! */
2198 }
2199
2b2d6d01 2200#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2201
2202 if (nr_orphans)
b31e1552
ES		 2203 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2204 PLURAL(nr_orphans));
ac27a0ec 2205 if (nr_truncates)
b31e1552
ES		 2206 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2207 PLURAL(nr_truncates));
ac27a0ec
DK		 2208#ifdef CONFIG_QUOTA
2209 /* Turn quotas off */
2210 for (i = 0; i < MAXQUOTAS; i++) {
2211 if (sb_dqopt(sb)->files[i])
287a8095 2212 dquot_quota_off(sb, i);
ac27a0ec
DK		 2213 }
2214#endif
2215 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2216}
0b8e58a1 2217
cd2291a4
ES		 2218/*
2219 * Maximal extent format file size.
2220 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2221 * extent format containers, within a sector_t, and within i_blocks
2222 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2223 * so that won't be a limiting factor.
2224 *
f17722f9
LC		 2225 * However there is other limiting factor. We do store extents in the form
2226 * of starting block and length, hence the resulting length of the extent
2227 * covering maximum file size must fit into on-disk format containers as
2228 * well. Given that length is always by 1 unit bigger than max unit (because
2229 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2230 *
cd2291a4
ES		 2231 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2232 */
f287a1a5 2233static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2234{
2235 loff_t res;
2236 loff_t upper_limit = MAX_LFS_FILESIZE;
2237
2238 /* small i_blocks in vfs inode? */
f287a1a5 2239 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2240 /*
90c699a9 2241 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2242 * i_block represent total blocks in 512 bytes
2243 * 32 == size of vfs inode i_blocks * 8
2244 */
2245 upper_limit = (1LL << 32) - 1;
2246
2247 /* total blocks in file system block size */
2248 upper_limit >>= (blkbits - 9);
2249 upper_limit <<= blkbits;
2250 }
2251
f17722f9
LC		 2252 /*
2253 * 32-bit extent-start container, ee_block. We lower the maxbytes
2254 * by one fs block, so ee_len can cover the extent of maximum file
2255 * size
2256 */
2257 res = (1LL << 32) - 1;
cd2291a4 2258 res <<= blkbits;
cd2291a4
ES		 2259
2260 /* Sanity check against vm- & vfs- imposed limits */
2261 if (res > upper_limit)
2262 res = upper_limit;
2263
2264 return res;
2265}
ac27a0ec 2266
ac27a0ec 2267/*
cd2291a4 2268 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2269 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2270 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2271 */
f287a1a5 2272static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2273{
617ba13b 2274 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2275 int meta_blocks;
2276 loff_t upper_limit;
0b8e58a1
AD		 2277 /* This is calculated to be the largest file size for a dense, block
2278 * mapped file such that the file's total number of 512-byte sectors,
2279 * including data and all indirect blocks, does not exceed (2^48 - 1).
2280 *
2281 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2282 * number of 512-byte sectors of the file.
0fc1b451
AK		 2283 */
2284
f287a1a5 2285 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2286 /*
90c699a9 2287 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2288 * the inode i_block field represents total file blocks in
2289 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2290 */
2291 upper_limit = (1LL << 32) - 1;
2292
2293 /* total blocks in file system block size */
2294 upper_limit >>= (bits - 9);
2295
2296 } else {
8180a562
AK		 2297 /*
2298 * We use 48 bit ext4_inode i_blocks
2299 * With EXT4_HUGE_FILE_FL set the i_blocks
2300 * represent total number of blocks in
2301 * file system block size
2302 */
0fc1b451
AK		 2303 upper_limit = (1LL << 48) - 1;
2304
0fc1b451
AK		 2305 }
2306
2307 /* indirect blocks */
2308 meta_blocks = 1;
2309 /* double indirect blocks */
2310 meta_blocks += 1 + (1LL << (bits-2));
2311 /* tripple indirect blocks */
2312 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2313
2314 upper_limit -= meta_blocks;
2315 upper_limit <<= bits;
ac27a0ec
DK		 2316
2317 res += 1LL << (bits-2);
2318 res += 1LL << (2*(bits-2));
2319 res += 1LL << (3*(bits-2));
2320 res <<= bits;
2321 if (res > upper_limit)
2322 res = upper_limit;
0fc1b451
AK		 2323
2324 if (res > MAX_LFS_FILESIZE)
2325 res = MAX_LFS_FILESIZE;
2326
ac27a0ec
DK		 2327 return res;
2328}
2329
617ba13b 2330static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2331 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2332{
617ba13b 2333 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2334 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2335 int has_super = 0;
2336
2337 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2338
617ba13b 2339 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2340 nr < first_meta_bg)
70bbb3e0 2341 return logical_sb_block + nr + 1;
ac27a0ec 2342 bg = sbi->s_desc_per_block * nr;
617ba13b 2343 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2344 has_super = 1;
0b8e58a1 2345
617ba13b 2346 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2347}
2348
c9de560d
AT		 2349/**
2350 * ext4_get_stripe_size: Get the stripe size.
2351 * @sbi: In memory super block info
2352 *
2353 * If we have specified it via mount option, then
2354 * use the mount option value. If the value specified at mount time is
2355 * greater than the blocks per group use the super block value.
2356 * If the super block value is greater than blocks per group return 0.
2357 * Allocator needs it be less than blocks per group.
2358 *
2359 */
2360static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2361{
2362 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2363 unsigned long stripe_width =
2364 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2365 int ret;
c9de560d
AT		 2366
2367 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2368 ret = sbi->s_stripe;
2369 else if (stripe_width <= sbi->s_blocks_per_group)
2370 ret = stripe_width;
2371 else if (stride <= sbi->s_blocks_per_group)
2372 ret = stride;
2373 else
2374 ret = 0;
c9de560d 2375
3eb08658
DE		 2376 /*
2377 * If the stripe width is 1, this makes no sense and
2378 * we set it to 0 to turn off stripe handling code.
2379 */
2380 if (ret <= 1)
2381 ret = 0;
c9de560d 2382
3eb08658 2383 return ret;
c9de560d 2384}
ac27a0ec 2385
3197ebdb
TT		 2386/* sysfs supprt */
2387
2388struct ext4_attr {
2389 struct attribute attr;
2390 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2391 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb 2392 const char *, size_t);
f2d50a65
JK		 2393 union {
2394 int offset;
2395 int deprecated_val;
2396 } u;
3197ebdb
TT		 2397};
2398
27dd4385
LC		 2399static int parse_strtoull(const char *buf,
2400 unsigned long long max, unsigned long long *value)
2401{
2402 int ret;
2403
2404 ret = kstrtoull(skip_spaces(buf), 0, value);
2405 if (!ret && *value > max)
2406 ret = -EINVAL;
2407 return ret;
2408}
2409
3197ebdb
TT		 2410static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2411 struct ext4_sb_info *sbi,
2412 char *buf)
2413{
2414 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2415 (s64) EXT4_C2B(sbi,
2416 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2417}
2418
2419static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2420 struct ext4_sb_info *sbi, char *buf)
2421{
2422 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2423
f613dfcb
TT		 2424 if (!sb->s_bdev->bd_part)
2425 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2426 return snprintf(buf, PAGE_SIZE, "%lu\n",
2427 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2428 sbi->s_sectors_written_start) >> 1);
2429}
2430
2431static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2432 struct ext4_sb_info *sbi, char *buf)
2433{
2434 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2435
f613dfcb
TT		 2436 if (!sb->s_bdev->bd_part)
2437 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2438 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2439 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2440 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2441 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2442}
2443
2444static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2445 struct ext4_sb_info *sbi,
2446 const char *buf, size_t count)
2447{
2448 unsigned long t;
e1091b15 2449 int ret;
3197ebdb 2450
e1091b15
LC		 2451 ret = kstrtoul(skip_spaces(buf), 0, &t);
2452 if (ret)
2453 return ret;
3197ebdb 2454
e1091b15 2455 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2456 return -EINVAL;
2457
2458 sbi->s_inode_readahead_blks = t;
2459 return count;
2460}
2461
2462static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2463 struct ext4_sb_info *sbi, char *buf)
3197ebdb 2464{
f2d50a65 2465 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb
TT		 2466
2467 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2468}
2469
2470static ssize_t sbi_ui_store(struct ext4_attr *a,
2471 struct ext4_sb_info *sbi,
2472 const char *buf, size_t count)
2473{
f2d50a65 2474 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb 2475 unsigned long t;
e1091b15 2476 int ret;
3197ebdb 2477
e1091b15
LC		 2478 ret = kstrtoul(skip_spaces(buf), 0, &t);
2479 if (ret)
2480 return ret;
3197ebdb
TT		 2481 *ui = t;
2482 return count;
2483}
2484
27dd4385
LC		 2485static ssize_t reserved_clusters_show(struct ext4_attr *a,
2486 struct ext4_sb_info *sbi, char *buf)
2487{
2488 return snprintf(buf, PAGE_SIZE, "%llu\n",
2489 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2490}
2491
2492static ssize_t reserved_clusters_store(struct ext4_attr *a,
2493 struct ext4_sb_info *sbi,
2494 const char *buf, size_t count)
2495{
2496 unsigned long long val;
2497 int ret;
2498
2499 if (parse_strtoull(buf, -1ULL, &val))
2500 return -EINVAL;
2501 ret = ext4_reserve_clusters(sbi, val);
2502
2503 return ret ? ret : count;
2504}
2505
2c0544b2
TT		 2506static ssize_t trigger_test_error(struct ext4_attr *a,
2507 struct ext4_sb_info *sbi,
2508 const char *buf, size_t count)
2509{
2510 int len = count;
2511
2512 if (!capable(CAP_SYS_ADMIN))
2513 return -EPERM;
2514
2515 if (len && buf[len-1] == '\n')
2516 len--;
2517
2518 if (len)
2519 ext4_error(sbi->s_sb, "%.*s", len, buf);
2520 return count;
2521}
2522
f2d50a65
JK		 2523static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2524 struct ext4_sb_info *sbi, char *buf)
2525{
2526 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2527}
2528
3197ebdb
TT		 2529#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2530static struct ext4_attr ext4_attr_##_name = { \
2531 .attr = {.name = __stringify(_name), .mode = _mode }, \
2532 .show = _show, \
2533 .store = _store, \
f2d50a65
JK		 2534 .u = { \
2535 .offset = offsetof(struct ext4_sb_info, _elname),\
2536 }, \
3197ebdb
TT		 2537}
2538#define EXT4_ATTR(name, mode, show, store) \
2539static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2540
857ac889 2541#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2542#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2543#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2544#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2545 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2546#define ATTR_LIST(name) &ext4_attr_##name.attr
f2d50a65
JK		 2547#define EXT4_DEPRECATED_ATTR(_name, _val) \
2548static struct ext4_attr ext4_attr_##_name = { \
2549 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2550 .show = sbi_deprecated_show, \
2551 .u = { \
2552 .deprecated_val = _val, \
2553 }, \
2554}
3197ebdb
TT		 2555
2556EXT4_RO_ATTR(delayed_allocation_blocks);
2557EXT4_RO_ATTR(session_write_kbytes);
2558EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2559EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2560EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2561 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2562EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2563EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2564EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2565EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2566EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2567EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2568EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
f2d50a65 2569EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
67a5da56 2570EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2571EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2572
2573static struct attribute *ext4_attrs[] = {
2574 ATTR_LIST(delayed_allocation_blocks),
2575 ATTR_LIST(session_write_kbytes),
2576 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2577 ATTR_LIST(reserved_clusters),
3197ebdb 2578 ATTR_LIST(inode_readahead_blks),
11013911 2579 ATTR_LIST(inode_goal),
3197ebdb
TT		 2580 ATTR_LIST(mb_stats),
2581 ATTR_LIST(mb_max_to_scan),
2582 ATTR_LIST(mb_min_to_scan),
2583 ATTR_LIST(mb_order2_req),
2584 ATTR_LIST(mb_stream_req),
2585 ATTR_LIST(mb_group_prealloc),
55138e0b 2586 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2587 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2588 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2589 NULL,
2590};
2591
857ac889
LC		 2592/* Features this copy of ext4 supports */
2593EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2594EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2595EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2596
2597static struct attribute *ext4_feat_attrs[] = {
2598 ATTR_LIST(lazy_itable_init),
27ee40df 2599 ATTR_LIST(batched_discard),
5e7bbef1 2600 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2601 NULL,
2602};
2603
3197ebdb
TT		 2604static ssize_t ext4_attr_show(struct kobject *kobj,
2605 struct attribute *attr, char *buf)
2606{
2607 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2608 s_kobj);
2609 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2610
2611 return a->show ? a->show(a, sbi, buf) : 0;
2612}
2613
2614static ssize_t ext4_attr_store(struct kobject *kobj,
2615 struct attribute *attr,
2616 const char *buf, size_t len)
2617{
2618 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2619 s_kobj);
2620 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2621
2622 return a->store ? a->store(a, sbi, buf, len) : 0;
2623}
2624
2625static void ext4_sb_release(struct kobject *kobj)
2626{
2627 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2628 s_kobj);
2629 complete(&sbi->s_kobj_unregister);
2630}
2631
52cf25d0 2632static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2633 .show = ext4_attr_show,
2634 .store = ext4_attr_store,
2635};
2636
2637static struct kobj_type ext4_ktype = {
2638 .default_attrs = ext4_attrs,
2639 .sysfs_ops = &ext4_attr_ops,
2640 .release = ext4_sb_release,
2641};
2642
857ac889
LC		 2643static void ext4_feat_release(struct kobject *kobj)
2644{
2645 complete(&ext4_feat->f_kobj_unregister);
2646}
2647
2648static struct kobj_type ext4_feat_ktype = {
2649 .default_attrs = ext4_feat_attrs,
2650 .sysfs_ops = &ext4_attr_ops,
2651 .release = ext4_feat_release,
2652};
2653
a13fb1a4
ES		 2654/*
2655 * Check whether this filesystem can be mounted based on
2656 * the features present and the RDONLY/RDWR mount requested.
2657 * Returns 1 if this filesystem can be mounted as requested,
2658 * 0 if it cannot be.
2659 */
2660static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2661{
2662 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2663 ext4_msg(sb, KERN_ERR,
2664 "Couldn't mount because of "
2665 "unsupported optional features (%x)",
2666 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2667 ~EXT4_FEATURE_INCOMPAT_SUPP));
2668 return 0;
2669 }
2670
2671 if (readonly)
2672 return 1;
2673
2674 /* Check that feature set is OK for a read-write mount */
2675 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2676 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2677 "unsupported optional features (%x)",
2678 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2679 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2680 return 0;
2681 }
2682 /*
2683 * Large file size enabled file system can only be mounted
2684 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2685 */
2686 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2687 if (sizeof(blkcnt_t) < sizeof(u64)) {
2688 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2689 "cannot be mounted RDWR without "
2690 "CONFIG_LBDAF");
2691 return 0;
2692 }
2693 }
bab08ab9
TT		 2694 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2695 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2696 ext4_msg(sb, KERN_ERR,
2697 "Can't support bigalloc feature without "
2698 "extents feature\n");
2699 return 0;
2700 }
7c319d32
AK		 2701
2702#ifndef CONFIG_QUOTA
2703 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2704 !readonly) {
2705 ext4_msg(sb, KERN_ERR,
2706 "Filesystem with quota feature cannot be mounted RDWR "
2707 "without CONFIG_QUOTA");
2708 return 0;
2709 }
2710#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2711 return 1;
2712}
2713
66e61a9e
TT		 2714/*
2715 * This function is called once a day if we have errors logged
2716 * on the file system
2717 */
2718static void print_daily_error_info(unsigned long arg)
2719{
2720 struct super_block *sb = (struct super_block *) arg;
2721 struct ext4_sb_info *sbi;
2722 struct ext4_super_block *es;
2723
2724 sbi = EXT4_SB(sb);
2725 es = sbi->s_es;
2726
2727 if (es->s_error_count)
2728 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2729 le32_to_cpu(es->s_error_count));
2730 if (es->s_first_error_time) {
2731 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2732 sb->s_id, le32_to_cpu(es->s_first_error_time),
2733 (int) sizeof(es->s_first_error_func),
2734 es->s_first_error_func,
2735 le32_to_cpu(es->s_first_error_line));
2736 if (es->s_first_error_ino)
2737 printk(": inode %u",
2738 le32_to_cpu(es->s_first_error_ino));
2739 if (es->s_first_error_block)
2740 printk(": block %llu", (unsigned long long)
2741 le64_to_cpu(es->s_first_error_block));
2742 printk("\n");
2743 }
2744 if (es->s_last_error_time) {
2745 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2746 sb->s_id, le32_to_cpu(es->s_last_error_time),
2747 (int) sizeof(es->s_last_error_func),
2748 es->s_last_error_func,
2749 le32_to_cpu(es->s_last_error_line));
2750 if (es->s_last_error_ino)
2751 printk(": inode %u",
2752 le32_to_cpu(es->s_last_error_ino));
2753 if (es->s_last_error_block)
2754 printk(": block %llu", (unsigned long long)
2755 le64_to_cpu(es->s_last_error_block));
2756 printk("\n");
2757 }
2758 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2759}
2760
bfff6873
LC		 2761/* Find next suitable group and run ext4_init_inode_table */
2762static int ext4_run_li_request(struct ext4_li_request *elr)
2763{
2764 struct ext4_group_desc *gdp = NULL;
2765 ext4_group_t group, ngroups;
2766 struct super_block *sb;
2767 unsigned long timeout = 0;
2768 int ret = 0;
2769
2770 sb = elr->lr_super;
2771 ngroups = EXT4_SB(sb)->s_groups_count;
2772
8e8ad8a5 2773 sb_start_write(sb);
bfff6873
LC		 2774 for (group = elr->lr_next_group; group < ngroups; group++) {
2775 gdp = ext4_get_group_desc(sb, group, NULL);
2776 if (!gdp) {
2777 ret = 1;
2778 break;
2779 }
2780
2781 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2782 break;
2783 }
2784
7f511862 2785 if (group >= ngroups)
bfff6873
LC		 2786 ret = 1;
2787
2788 if (!ret) {
2789 timeout = jiffies;
2790 ret = ext4_init_inode_table(sb, group,
2791 elr->lr_timeout ? 0 : 1);
2792 if (elr->lr_timeout == 0) {
51ce6511
LC		 2793 timeout = (jiffies - timeout) *
2794 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2795 elr->lr_timeout = timeout;
2796 }
2797 elr->lr_next_sched = jiffies + elr->lr_timeout;
2798 elr->lr_next_group = group + 1;
2799 }
8e8ad8a5 2800 sb_end_write(sb);
bfff6873
LC		 2801
2802 return ret;
2803}
2804
2805/*
2806 * Remove lr_request from the list_request and free the
4ed5c033 2807 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2808 */
2809static void ext4_remove_li_request(struct ext4_li_request *elr)
2810{
2811 struct ext4_sb_info *sbi;
2812
2813 if (!elr)
2814 return;
2815
2816 sbi = elr->lr_sbi;
2817
2818 list_del(&elr->lr_request);
2819 sbi->s_li_request = NULL;
2820 kfree(elr);
2821}
2822
2823static void ext4_unregister_li_request(struct super_block *sb)
2824{
1bb933fb
LC		 2825 mutex_lock(&ext4_li_mtx);
2826 if (!ext4_li_info) {
2827 mutex_unlock(&ext4_li_mtx);
bfff6873 2828 return;
1bb933fb 2829 }
bfff6873
LC		 2830
2831 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2832 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2833 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2834 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2835}
2836
8f1f7453
ES		 2837static struct task_struct *ext4_lazyinit_task;
2838
bfff6873
LC		 2839/*
2840 * This is the function where ext4lazyinit thread lives. It walks
2841 * through the request list searching for next scheduled filesystem.
2842 * When such a fs is found, run the lazy initialization request
2843 * (ext4_rn_li_request) and keep track of the time spend in this
2844 * function. Based on that time we compute next schedule time of
2845 * the request. When walking through the list is complete, compute
2846 * next waking time and put itself into sleep.
2847 */
2848static int ext4_lazyinit_thread(void *arg)
2849{
2850 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2851 struct list_head *pos, *n;
2852 struct ext4_li_request *elr;
4ed5c033 2853 unsigned long next_wakeup, cur;
bfff6873
LC		 2854
2855 BUG_ON(NULL == eli);
2856
bfff6873
LC		 2857cont_thread:
2858 while (true) {
2859 next_wakeup = MAX_JIFFY_OFFSET;
2860
2861 mutex_lock(&eli->li_list_mtx);
2862 if (list_empty(&eli->li_request_list)) {
2863 mutex_unlock(&eli->li_list_mtx);
2864 goto exit_thread;
2865 }
2866
2867 list_for_each_safe(pos, n, &eli->li_request_list) {
2868 elr = list_entry(pos, struct ext4_li_request,
2869 lr_request);
2870
b2c78cd0
TT		 2871 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2872 if (ext4_run_li_request(elr) != 0) {
2873 /* error, remove the lazy_init job */
2874 ext4_remove_li_request(elr);
2875 continue;
2876 }
bfff6873
LC		 2877 }
2878
2879 if (time_before(elr->lr_next_sched, next_wakeup))
2880 next_wakeup = elr->lr_next_sched;
2881 }
2882 mutex_unlock(&eli->li_list_mtx);
2883
a0acae0e 2884 try_to_freeze();
bfff6873 2885
4ed5c033
LC		 2886 cur = jiffies;
2887 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2888 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2889 cond_resched();
2890 continue;
2891 }
2892
4ed5c033
LC		 2893 schedule_timeout_interruptible(next_wakeup - cur);
2894
8f1f7453
ES		 2895 if (kthread_should_stop()) {
2896 ext4_clear_request_list();
2897 goto exit_thread;
2898 }
bfff6873
LC		 2899 }
2900
2901exit_thread:
2902 /*
2903 * It looks like the request list is empty, but we need
2904 * to check it under the li_list_mtx lock, to prevent any
2905 * additions into it, and of course we should lock ext4_li_mtx
2906 * to atomically free the list and ext4_li_info, because at
2907 * this point another ext4 filesystem could be registering
2908 * new one.
2909 */
2910 mutex_lock(&ext4_li_mtx);
2911 mutex_lock(&eli->li_list_mtx);
2912 if (!list_empty(&eli->li_request_list)) {
2913 mutex_unlock(&eli->li_list_mtx);
2914 mutex_unlock(&ext4_li_mtx);
2915 goto cont_thread;
2916 }
2917 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2918 kfree(ext4_li_info);
2919 ext4_li_info = NULL;
2920 mutex_unlock(&ext4_li_mtx);
2921
2922 return 0;
2923}
2924
2925static void ext4_clear_request_list(void)
2926{
2927 struct list_head *pos, *n;
2928 struct ext4_li_request *elr;
2929
2930 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2931 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2932 elr = list_entry(pos, struct ext4_li_request,
2933 lr_request);
2934 ext4_remove_li_request(elr);
2935 }
2936 mutex_unlock(&ext4_li_info->li_list_mtx);
2937}
2938
2939static int ext4_run_lazyinit_thread(void)
2940{
8f1f7453
ES		 2941 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2942 ext4_li_info, "ext4lazyinit");
2943 if (IS_ERR(ext4_lazyinit_task)) {
2944 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2945 ext4_clear_request_list();
bfff6873
LC		 2946 kfree(ext4_li_info);
2947 ext4_li_info = NULL;
92b97816 2948 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2949 "initialization thread\n",
2950 err);
2951 return err;
2952 }
2953 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2954 return 0;
2955}
2956
2957/*
2958 * Check whether it make sense to run itable init. thread or not.
2959 * If there is at least one uninitialized inode table, return
2960 * corresponding group number, else the loop goes through all
2961 * groups and return total number of groups.
2962 */
2963static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2964{
2965 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2966 struct ext4_group_desc *gdp = NULL;
2967
2968 for (group = 0; group < ngroups; group++) {
2969 gdp = ext4_get_group_desc(sb, group, NULL);
2970 if (!gdp)
2971 continue;
2972
2973 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2974 break;
2975 }
2976
2977 return group;
2978}
2979
2980static int ext4_li_info_new(void)
2981{
2982 struct ext4_lazy_init *eli = NULL;
2983
2984 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2985 if (!eli)
2986 return -ENOMEM;
2987
bfff6873
LC		 2988 INIT_LIST_HEAD(&eli->li_request_list);
2989 mutex_init(&eli->li_list_mtx);
2990
bfff6873
LC		 2991 eli->li_state |= EXT4_LAZYINIT_QUIT;
2992
2993 ext4_li_info = eli;
2994
2995 return 0;
2996}
2997
2998static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2999 ext4_group_t start)
3000{
3001 struct ext4_sb_info *sbi = EXT4_SB(sb);
3002 struct ext4_li_request *elr;
3003 unsigned long rnd;
3004
3005 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3006 if (!elr)
3007 return NULL;
3008
3009 elr->lr_super = sb;
3010 elr->lr_sbi = sbi;
3011 elr->lr_next_group = start;
3012
3013 /*
3014 * Randomize first schedule time of the request to
3015 * spread the inode table initialization requests
3016 * better.
3017 */
3018 get_random_bytes(&rnd, sizeof(rnd));
3019 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3020 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3021
3022 return elr;
3023}
3024
7f511862
TT		 3025int ext4_register_li_request(struct super_block *sb,
3026 ext4_group_t first_not_zeroed)
bfff6873
LC		 3027{
3028 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3029 struct ext4_li_request *elr = NULL;
bfff6873 3030 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3031 int ret = 0;
bfff6873 3032
7f511862 3033 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3034 if (sbi->s_li_request != NULL) {
3035 /*
3036 * Reset timeout so it can be computed again, because
3037 * s_li_wait_mult might have changed.
3038 */
3039 sbi->s_li_request->lr_timeout = 0;
7f511862 3040 goto out;
51ce6511 3041 }
bfff6873
LC		 3042
3043 if (first_not_zeroed == ngroups ||
3044 (sb->s_flags & MS_RDONLY) ||
55ff3840 3045 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3046 goto out;
bfff6873
LC		 3047
3048 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3049 if (!elr) {
3050 ret = -ENOMEM;
3051 goto out;
3052 }
bfff6873
LC		 3053
3054 if (NULL == ext4_li_info) {
3055 ret = ext4_li_info_new();
3056 if (ret)
3057 goto out;
3058 }
3059
3060 mutex_lock(&ext4_li_info->li_list_mtx);
3061 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3062 mutex_unlock(&ext4_li_info->li_list_mtx);
3063
3064 sbi->s_li_request = elr;
46e4690b
TM		 3065 /*
3066 * set elr to NULL here since it has been inserted to
3067 * the request_list and the removal and free of it is
3068 * handled by ext4_clear_request_list from now on.
3069 */
3070 elr = NULL;
bfff6873
LC		 3071
3072 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3073 ret = ext4_run_lazyinit_thread();
3074 if (ret)
3075 goto out;
3076 }
bfff6873 3077out:
beed5ecb
NK		 3078 mutex_unlock(&ext4_li_mtx);
3079 if (ret)
bfff6873 3080 kfree(elr);
bfff6873
LC		 3081 return ret;
3082}
3083
3084/*
3085 * We do not need to lock anything since this is called on
3086 * module unload.
3087 */
3088static void ext4_destroy_lazyinit_thread(void)
3089{
3090 /*
3091 * If thread exited earlier
3092 * there's nothing to be done.
3093 */
8f1f7453 3094 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3095 return;
3096
8f1f7453 3097 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3098}
3099
25ed6e8a
DW		 3100static int set_journal_csum_feature_set(struct super_block *sb)
3101{
3102 int ret = 1;
3103 int compat, incompat;
3104 struct ext4_sb_info *sbi = EXT4_SB(sb);
3105
3106 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3107 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3108 /* journal checksum v2 */
3109 compat = 0;
3110 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3111 } else {
3112 /* journal checksum v1 */
3113 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3114 incompat = 0;
3115 }
3116
3117 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3118 ret = jbd2_journal_set_features(sbi->s_journal,
3119 compat, 0,
3120 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3121 incompat);
3122 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3123 ret = jbd2_journal_set_features(sbi->s_journal,
3124 compat, 0,
3125 incompat);
3126 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3127 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3128 } else {
3129 jbd2_journal_clear_features(sbi->s_journal,
3130 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3131 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3132 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3133 }
3134
3135 return ret;
3136}
3137
952fc18e
TT		 3138/*
3139 * Note: calculating the overhead so we can be compatible with
3140 * historical BSD practice is quite difficult in the face of
3141 * clusters/bigalloc. This is because multiple metadata blocks from
3142 * different block group can end up in the same allocation cluster.
3143 * Calculating the exact overhead in the face of clustered allocation
3144 * requires either O(all block bitmaps) in memory or O(number of block
3145 * groups**2) in time. We will still calculate the superblock for
3146 * older file systems --- and if we come across with a bigalloc file
3147 * system with zero in s_overhead_clusters the estimate will be close to
3148 * correct especially for very large cluster sizes --- but for newer
3149 * file systems, it's better to calculate this figure once at mkfs
3150 * time, and store it in the superblock. If the superblock value is
3151 * present (even for non-bigalloc file systems), we will use it.
3152 */
3153static int count_overhead(struct super_block *sb, ext4_group_t grp,
3154 char *buf)
3155{
3156 struct ext4_sb_info *sbi = EXT4_SB(sb);
3157 struct ext4_group_desc *gdp;
3158 ext4_fsblk_t first_block, last_block, b;
3159 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3160 int s, j, count = 0;
3161
0548bbb8
TT		 3162 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3163 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3164 sbi->s_itb_per_group + 2);
3165
952fc18e
TT		 3166 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3167 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3168 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3169 for (i = 0; i < ngroups; i++) {
3170 gdp = ext4_get_group_desc(sb, i, NULL);
3171 b = ext4_block_bitmap(sb, gdp);
3172 if (b >= first_block && b <= last_block) {
3173 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3174 count++;
3175 }
3176 b = ext4_inode_bitmap(sb, gdp);
3177 if (b >= first_block && b <= last_block) {
3178 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3179 count++;
3180 }
3181 b = ext4_inode_table(sb, gdp);
3182 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3183 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3184 int c = EXT4_B2C(sbi, b - first_block);
3185 ext4_set_bit(c, buf);
3186 count++;
3187 }
3188 if (i != grp)
3189 continue;
3190 s = 0;
3191 if (ext4_bg_has_super(sb, grp)) {
3192 ext4_set_bit(s++, buf);
3193 count++;
3194 }
3195 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3196 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3197 count++;
3198 }
3199 }
3200 if (!count)
3201 return 0;
3202 return EXT4_CLUSTERS_PER_GROUP(sb) -
3203 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3204}
3205
3206/*
3207 * Compute the overhead and stash it in sbi->s_overhead
3208 */
3209int ext4_calculate_overhead(struct super_block *sb)
3210{
3211 struct ext4_sb_info *sbi = EXT4_SB(sb);
3212 struct ext4_super_block *es = sbi->s_es;
3213 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3214 ext4_fsblk_t overhead = 0;
3215 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3216
952fc18e
TT		 3217 if (!buf)
3218 return -ENOMEM;
3219
3220 /*
3221 * Compute the overhead (FS structures). This is constant
3222 * for a given filesystem unless the number of block groups
3223 * changes so we cache the previous value until it does.
3224 */
3225
3226 /*
3227 * All of the blocks before first_data_block are overhead
3228 */
3229 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3230
3231 /*
3232 * Add the overhead found in each block group
3233 */
3234 for (i = 0; i < ngroups; i++) {
3235 int blks;
3236
3237 blks = count_overhead(sb, i, buf);
3238 overhead += blks;
3239 if (blks)
3240 memset(buf, 0, PAGE_SIZE);
3241 cond_resched();
3242 }
0875a2b4
ES		 3243 /* Add the journal blocks as well */
3244 if (sbi->s_journal)
810da240 3245 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3246
952fc18e
TT		 3247 sbi->s_overhead = overhead;
3248 smp_wmb();
3249 free_page((unsigned long) buf);
3250 return 0;
3251}
3252
27dd4385
LC		 3253
3254static ext4_fsblk_t ext4_calculate_resv_clusters(struct ext4_sb_info *sbi)
3255{
3256 ext4_fsblk_t resv_clusters;
3257
3258 /*
3259 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3260 * This should cover the situations where we can not afford to run
3261 * out of space like for example punch hole, or converting
3262 * uninitialized extents in delalloc path. In most cases such
3263 * allocation would require 1, or 2 blocks, higher numbers are
3264 * very rare.
3265 */
3266 resv_clusters = ext4_blocks_count(sbi->s_es) >> sbi->s_cluster_bits;
3267
3268 do_div(resv_clusters, 50);
3269 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3270
3271 return resv_clusters;
3272}
3273
3274
3275static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3276{
3277 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3278 sbi->s_cluster_bits;
3279
3280 if (count >= clusters)
3281 return -EINVAL;
3282
3283 atomic64_set(&sbi->s_resv_clusters, count);
3284 return 0;
3285}
3286
2b2d6d01 3287static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3288{
d4c402d9 3289 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3290 struct buffer_head *bh;
617ba13b
MC		 3291 struct ext4_super_block *es = NULL;
3292 struct ext4_sb_info *sbi;
3293 ext4_fsblk_t block;
3294 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3295 ext4_fsblk_t logical_sb_block;
ac27a0ec 3296 unsigned long offset = 0;
ac27a0ec
DK		 3297 unsigned long journal_devnum = 0;
3298 unsigned long def_mount_opts;
3299 struct inode *root;
9f6200bb 3300 char *cp;
0390131b 3301 const char *descr;
dcc7dae3 3302 int ret = -ENOMEM;
281b5995 3303 int blocksize, clustersize;
4ec11028
TT		 3304 unsigned int db_count;
3305 unsigned int i;
281b5995 3306 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3307 __u64 blocks_count;
07aa2ea1 3308 int err = 0;
b3881f74 3309 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3310 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3311
3312 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3313 if (!sbi)
dcc7dae3 3314 goto out_free_orig;
705895b6
PE		 3315
3316 sbi->s_blockgroup_lock =
3317 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3318 if (!sbi->s_blockgroup_lock) {
3319 kfree(sbi);
dcc7dae3 3320 goto out_free_orig;
705895b6 3321 }
ac27a0ec 3322 sb->s_fs_info = sbi;
2c0544b2 3323 sbi->s_sb = sb;
240799cd 3324 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3325 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3326 if (sb->s_bdev->bd_part)
3327 sbi->s_sectors_written_start =
3328 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3329
9f6200bb
TT		 3330 /* Cleanup superblock name */
3331 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3332 *cp = '!';
3333
07aa2ea1 3334 /* -EINVAL is default */
dcc7dae3 3335 ret = -EINVAL;
617ba13b 3336 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3337 if (!blocksize) {
b31e1552 3338 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3339 goto out_fail;
3340 }
3341
3342 /*
617ba13b 3343 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3344 * block sizes. We need to calculate the offset from buffer start.
3345 */
617ba13b 3346 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3347 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3348 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3349 } else {
70bbb3e0 3350 logical_sb_block = sb_block;
ac27a0ec
DK		 3351 }
3352
70bbb3e0 3353 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3354 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3355 goto out_fail;
3356 }
3357 /*
3358 * Note: s_es must be initialized as soon as possible because
617ba13b 3359 * some ext4 macro-instructions depend on its value
ac27a0ec 3360 */
2716b802 3361 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3362 sbi->s_es = es;
3363 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3364 if (sb->s_magic != EXT4_SUPER_MAGIC)
3365 goto cantfind_ext4;
afc32f7e 3366 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3367
feb0ab32
DW		 3368 /* Warn if metadata_csum and gdt_csum are both set. */
3369 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3370 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3371 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3372 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3373 "redundant flags; please run fsck.");
3374
d25425f8
DW		 3375 /* Check for a known checksum algorithm */
3376 if (!ext4_verify_csum_type(sb, es)) {
3377 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3378 "unknown checksum algorithm.");
3379 silent = 1;
3380 goto cantfind_ext4;
3381 }
3382
0441984a
DW		 3383 /* Load the checksum driver */
3384 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3385 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3386 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3387 if (IS_ERR(sbi->s_chksum_driver)) {
3388 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3389 ret = PTR_ERR(sbi->s_chksum_driver);
3390 sbi->s_chksum_driver = NULL;
3391 goto failed_mount;
3392 }
3393 }
3394
a9c47317
DW		 3395 /* Check superblock checksum */
3396 if (!ext4_superblock_csum_verify(sb, es)) {
3397 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3398 "invalid superblock checksum. Run e2fsck?");
3399 silent = 1;
3400 goto cantfind_ext4;
3401 }
3402
3403 /* Precompute checksum seed for all metadata */
3404 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3405 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3406 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3407 sizeof(es->s_uuid));
3408
ac27a0ec
DK		 3409 /* Set defaults before we parse the mount options */
3410 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3411 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3412 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3413 set_opt(sb, DEBUG);
87f26807 3414 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3415 set_opt(sb, GRPID);
617ba13b 3416 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3417 set_opt(sb, NO_UID32);
ea663336 3418 /* xattr user namespace & acls are now defaulted on */
ea663336 3419 set_opt(sb, XATTR_USER);
03010a33 3420#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3421 set_opt(sb, POSIX_ACL);
2e7842b8 3422#endif
617ba13b 3423 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3424 set_opt(sb, JOURNAL_DATA);
617ba13b 3425 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3426 set_opt(sb, ORDERED_DATA);
617ba13b 3427 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3428 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3429
3430 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3431 set_opt(sb, ERRORS_PANIC);
bb4f397a 3432 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3433 set_opt(sb, ERRORS_CONT);
bb4f397a 3434 else
fd8c37ec 3435 set_opt(sb, ERRORS_RO);
8b67f04a 3436 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3437 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3438 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3439 set_opt(sb, DISCARD);
ac27a0ec 3440
08cefc7a
EB		 3441 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3442 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3443 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3444 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3445 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3446
8b67f04a 3447 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3448 set_opt(sb, BARRIER);
ac27a0ec 3449
dd919b98
AK		 3450 /*
3451 * enable delayed allocation by default
3452 * Use -o nodelalloc to turn it off
3453 */
bc0b75f7 3454 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3455 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3456 set_opt(sb, DELALLOC);
dd919b98 3457
51ce6511
LC		 3458 /*
3459 * set default s_li_wait_mult for lazyinit, for the case there is
3460 * no mount option specified.
3461 */
3462 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3463
8b67f04a 3464 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3465 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3466 ext4_msg(sb, KERN_WARNING,
3467 "failed to parse options in superblock: %s",
3468 sbi->s_es->s_mount_opts);
3469 }
5a916be1 3470 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3471 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3472 &journal_ioprio, 0))
ac27a0ec
DK		 3473 goto failed_mount;
3474
56889787
TT		 3475 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3476 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3477 "with data=journal disables delayed "
3478 "allocation and O_DIRECT support!\n");
3479 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3480 ext4_msg(sb, KERN_ERR, "can't mount with "
3481 "both data=journal and delalloc");
3482 goto failed_mount;
3483 }
3484 if (test_opt(sb, DIOREAD_NOLOCK)) {
3485 ext4_msg(sb, KERN_ERR, "can't mount with "
3486 "both data=journal and delalloc");
3487 goto failed_mount;
3488 }
3489 if (test_opt(sb, DELALLOC))
3490 clear_opt(sb, DELALLOC);
3491 }
3492
ac27a0ec 3493 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3494 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3495
617ba13b
MC		 3496 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3497 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3498 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3499 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3500 ext4_msg(sb, KERN_WARNING,
3501 "feature flags set on rev 0 fs, "
3502 "running e2fsck is recommended");
469108ff 3503
2035e776
TT		 3504 if (IS_EXT2_SB(sb)) {
3505 if (ext2_feature_set_ok(sb))
3506 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3507 "using the ext4 subsystem");
3508 else {
3509 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3510 "to feature incompatibilities");
3511 goto failed_mount;
3512 }
3513 }
3514
3515 if (IS_EXT3_SB(sb)) {
3516 if (ext3_feature_set_ok(sb))
3517 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3518 "using the ext4 subsystem");
3519 else {
3520 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3521 "to feature incompatibilities");
3522 goto failed_mount;
3523 }
3524 }
3525
ac27a0ec
DK		 3526 /*
3527 * Check feature flags regardless of the revision level, since we
3528 * previously didn't change the revision level when setting the flags,
3529 * so there is a chance incompat flags are set on a rev 0 filesystem.
3530 */
a13fb1a4 3531 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3532 goto failed_mount;
a13fb1a4 3533
261cb20c 3534 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3535 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3536 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3537 ext4_msg(sb, KERN_ERR,
3538 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3539 goto failed_mount;
3540 }
3541
ac27a0ec 3542 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3543 /* Validate the filesystem blocksize */
3544 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3545 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3546 blocksize);
ac27a0ec
DK		 3547 goto failed_mount;
3548 }
3549
2b2d6d01 3550 brelse(bh);
70bbb3e0
AM		 3551 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3552 offset = do_div(logical_sb_block, blocksize);
3553 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3554 if (!bh) {
b31e1552
ES		 3555 ext4_msg(sb, KERN_ERR,
3556 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3557 goto failed_mount;
3558 }
2716b802 3559 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3560 sbi->s_es = es;
617ba13b 3561 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3562 ext4_msg(sb, KERN_ERR,
3563 "Magic mismatch, very weird!");
ac27a0ec
DK		 3564 goto failed_mount;
3565 }
3566 }
3567
a13fb1a4
ES		 3568 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3569 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3570 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3571 has_huge_files);
3572 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3573
617ba13b
MC		 3574 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3575 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3576 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3577 } else {
3578 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3579 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3580 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3581 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3582 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3583 ext4_msg(sb, KERN_ERR,
3584 "unsupported inode size: %d",
2b2d6d01 3585 sbi->s_inode_size);
ac27a0ec
DK		 3586 goto failed_mount;
3587 }
ef7f3835
KS		 3588 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3589 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3590 }
0b8e58a1 3591
0d1ee42f
AR		 3592 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3593 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3594 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3595 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3596 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3597 ext4_msg(sb, KERN_ERR,
3598 "unsupported descriptor size %lu",
0d1ee42f
AR		 3599 sbi->s_desc_size);
3600 goto failed_mount;
3601 }
3602 } else
3603 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3604
ac27a0ec 3605 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3606 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3607 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3608 goto cantfind_ext4;
0b8e58a1 3609
617ba13b 3610 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3611 if (sbi->s_inodes_per_block == 0)
617ba13b 3612 goto cantfind_ext4;
ac27a0ec
DK		 3613 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3614 sbi->s_inodes_per_block;
0d1ee42f 3615 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3616 sbi->s_sbh = bh;
3617 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3618 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3619 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3620
2b2d6d01 3621 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3622 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3623 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3624 i = le32_to_cpu(es->s_flags);
3625 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3626 sbi->s_hash_unsigned = 3;
3627 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3628#ifdef __CHAR_UNSIGNED__
3629 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3630 sbi->s_hash_unsigned = 3;
3631#else
3632 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3633#endif
f99b2589 3634 }
ac27a0ec 3635
281b5995
TT		 3636 /* Handle clustersize */
3637 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3638 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3639 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3640 if (has_bigalloc) {
3641 if (clustersize < blocksize) {
3642 ext4_msg(sb, KERN_ERR,
3643 "cluster size (%d) smaller than "
3644 "block size (%d)", clustersize, blocksize);
3645 goto failed_mount;
3646 }
3647 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3648 le32_to_cpu(es->s_log_block_size);
3649 sbi->s_clusters_per_group =
3650 le32_to_cpu(es->s_clusters_per_group);
3651 if (sbi->s_clusters_per_group > blocksize * 8) {
3652 ext4_msg(sb, KERN_ERR,
3653 "#clusters per group too big: %lu",
3654 sbi->s_clusters_per_group);
3655 goto failed_mount;
3656 }
3657 if (sbi->s_blocks_per_group !=
3658 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3659 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3660 "clusters per group (%lu) inconsistent",
3661 sbi->s_blocks_per_group,
3662 sbi->s_clusters_per_group);
3663 goto failed_mount;
3664 }
3665 } else {
3666 if (clustersize != blocksize) {
3667 ext4_warning(sb, "fragment/cluster size (%d) != "
3668 "block size (%d)", clustersize,
3669 blocksize);
3670 clustersize = blocksize;
3671 }
3672 if (sbi->s_blocks_per_group > blocksize * 8) {
3673 ext4_msg(sb, KERN_ERR,
3674 "#blocks per group too big: %lu",
3675 sbi->s_blocks_per_group);
3676 goto failed_mount;
3677 }
3678 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3679 sbi->s_cluster_bits = 0;
ac27a0ec 3680 }
281b5995
TT		 3681 sbi->s_cluster_ratio = clustersize / blocksize;
3682
ac27a0ec 3683 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3684 ext4_msg(sb, KERN_ERR,
3685 "#inodes per group too big: %lu",
2b2d6d01 3686 sbi->s_inodes_per_group);
ac27a0ec
DK		 3687 goto failed_mount;
3688 }
3689
960fd856
TT		 3690 /* Do we have standard group size of clustersize * 8 blocks ? */
3691 if (sbi->s_blocks_per_group == clustersize << 3)
3692 set_opt2(sb, STD_GROUP_SIZE);
3693
bf43d84b
ES		 3694 /*
3695 * Test whether we have more sectors than will fit in sector_t,
3696 * and whether the max offset is addressable by the page cache.
3697 */
5a9ae68a 3698 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3699 ext4_blocks_count(es));
5a9ae68a 3700 if (err) {
b31e1552 3701 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3702 " too large to mount safely on this system");
ac27a0ec 3703 if (sizeof(sector_t) < 8)
90c699a9 3704 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3705 goto failed_mount;
3706 }
3707
617ba13b
MC		 3708 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3709 goto cantfind_ext4;
e7c95593 3710
0f2ddca6
FTN		 3711 /* check blocks count against device size */
3712 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3713 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3714 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3715 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3716 ext4_blocks_count(es), blocks_count);
3717 goto failed_mount;
3718 }
3719
0b8e58a1
AD		 3720 /*
3721 * It makes no sense for the first data block to be beyond the end
3722 * of the filesystem.
3723 */
3724 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3725 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3726 "block %u is beyond end of filesystem (%llu)",
3727 le32_to_cpu(es->s_first_data_block),
3728 ext4_blocks_count(es));
e7c95593
ES		 3729 goto failed_mount;
3730 }
bd81d8ee
LV		 3731 blocks_count = (ext4_blocks_count(es) -
3732 le32_to_cpu(es->s_first_data_block) +
3733 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3734 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3735 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3736 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3737 "(block count %llu, first data block %u, "
b31e1552 3738 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3739 ext4_blocks_count(es),
3740 le32_to_cpu(es->s_first_data_block),
3741 EXT4_BLOCKS_PER_GROUP(sb));
3742 goto failed_mount;
3743 }
bd81d8ee 3744 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3745 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3746 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3747 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3748 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3749 sbi->s_group_desc = ext4_kvmalloc(db_count *
3750 sizeof(struct buffer_head *),
3751 GFP_KERNEL);
ac27a0ec 3752 if (sbi->s_group_desc == NULL) {
b31e1552 3753 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3754 ret = -ENOMEM;
ac27a0ec
DK		 3755 goto failed_mount;
3756 }
3757
9f6200bb
TT		 3758 if (ext4_proc_root)
3759 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3760
66acdcf4
TT		 3761 if (sbi->s_proc)
3762 proc_create_data("options", S_IRUGO, sbi->s_proc,
3763 &ext4_seq_options_fops, sb);
3764
705895b6 3765 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3766
3767 for (i = 0; i < db_count; i++) {
70bbb3e0 3768 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3769 sbi->s_group_desc[i] = sb_bread(sb, block);
3770 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3771 ext4_msg(sb, KERN_ERR,
3772 "can't read group descriptor %d", i);
ac27a0ec
DK		 3773 db_count = i;
3774 goto failed_mount2;
3775 }
3776 }
bfff6873 3777 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3778 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3779 goto failed_mount2;
3780 }
772cb7c8
JS		 3781 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3782 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3783 ext4_msg(sb, KERN_ERR,
3784 "unable to initialize "
3785 "flex_bg meta info!");
772cb7c8
JS		 3786 goto failed_mount2;
3787 }
3788
ac27a0ec
DK		 3789 sbi->s_gdb_count = db_count;
3790 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3791 spin_lock_init(&sbi->s_next_gen_lock);
3792
04496411
TM		 3793 init_timer(&sbi->s_err_report);
3794 sbi->s_err_report.function = print_daily_error_info;
3795 sbi->s_err_report.data = (unsigned long) sb;
3796
a75ae78f 3797 /* Register extent status tree shrinker */
d3922a77 3798 ext4_es_register_shrinker(sbi);
a75ae78f 3799
57042651 3800 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3801 ext4_count_free_clusters(sb));
ce7e010a
TT		 3802 if (!err) {
3803 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3804 ext4_count_free_inodes(sb));
3805 }
3806 if (!err) {
3807 err = percpu_counter_init(&sbi->s_dirs_counter,
3808 ext4_count_dirs(sb));
3809 }
3810 if (!err) {
57042651 3811 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a 3812 }
1ac6466f
TT		 3813 if (!err) {
3814 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3815 }
ce7e010a
TT		 3816 if (err) {
3817 ext4_msg(sb, KERN_ERR, "insufficient memory");
3818 goto failed_mount3;
3819 }
3820
c9de560d 3821 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3822 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3823
ac27a0ec
DK		 3824 /*
3825 * set up enough so that it can read an inode
3826 */
9ca92389
TT		 3827 if (!test_opt(sb, NOLOAD) &&
3828 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3829 sb->s_op = &ext4_sops;
3830 else
3831 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3832 sb->s_export_op = &ext4_export_ops;
3833 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3834#ifdef CONFIG_QUOTA
617ba13b 3835 sb->dq_op = &ext4_quota_operations;
262b4662 3836 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
7c319d32 3837 sb->s_qcop = &ext4_qctl_sysfile_operations;
262b4662
JK		 3838 else
3839 sb->s_qcop = &ext4_qctl_operations;
ac27a0ec 3840#endif
f2fa2ffc
AK		 3841 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3842
ac27a0ec 3843 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3844 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3845
3846 sb->s_root = NULL;
3847
3848 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3849 EXT4_HAS_INCOMPAT_FEATURE(sb,
3850 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3851
c5e06d10
JL		 3852 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3853 !(sb->s_flags & MS_RDONLY))
3854 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3855 goto failed_mount3;
3856
ac27a0ec
DK		 3857 /*
3858 * The first inode we look at is the journal inode. Don't try
3859 * root first: it may be modified in the journal!
3860 */
3861 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3862 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3863 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3864 goto failed_mount3;
0390131b
FM		 3865 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3866 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3867 ext4_msg(sb, KERN_ERR, "required journal recovery "
3868 "suppressed and not mounted read-only");
744692dc 3869 goto failed_mount_wq;
ac27a0ec 3870 } else {
fd8c37ec 3871 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3872 sbi->s_journal = NULL;
3873 needs_recovery = 0;
3874 goto no_journal;
ac27a0ec
DK		 3875 }
3876
f32aaf2d 3877 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3878 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3879 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3880 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3881 goto failed_mount_wq;
eb40a09c
JS		 3882 }
3883
25ed6e8a
DW		 3884 if (!set_journal_csum_feature_set(sb)) {
3885 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3886 "feature set");
3887 goto failed_mount_wq;
d4da6c9c 3888 }
818d276c 3889
ac27a0ec
DK		 3890 /* We have now updated the journal if required, so we can
3891 * validate the data journaling mode. */
3892 switch (test_opt(sb, DATA_FLAGS)) {
3893 case 0:
3894 /* No mode set, assume a default based on the journal
63f57933
AM		 3895 * capabilities: ORDERED_DATA if the journal can
3896 * cope, else JOURNAL_DATA
3897 */
dab291af
MC		 3898 if (jbd2_journal_check_available_features
3899 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3900 set_opt(sb, ORDERED_DATA);
ac27a0ec 3901 else
fd8c37ec 3902 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3903 break;
3904
617ba13b
MC		 3905 case EXT4_MOUNT_ORDERED_DATA:
3906 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3907 if (!jbd2_journal_check_available_features
3908 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3909 ext4_msg(sb, KERN_ERR, "Journal does not support "
3910 "requested data journaling mode");
744692dc 3911 goto failed_mount_wq;
ac27a0ec
DK		 3912 }
3913 default:
3914 break;
3915 }
b3881f74 3916 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3917
18aadd47
BJ		 3918 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3919
ce7e010a
TT		 3920 /*
3921 * The journal may have updated the bg summary counts, so we
3922 * need to update the global counters.
3923 */
57042651 3924 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3925 ext4_count_free_clusters(sb));
ce7e010a
TT		 3926 percpu_counter_set(&sbi->s_freeinodes_counter,
3927 ext4_count_free_inodes(sb));
3928 percpu_counter_set(&sbi->s_dirs_counter,
3929 ext4_count_dirs(sb));
57042651 3930 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3931
ce7e010a 3932no_journal:
952fc18e
TT		 3933 /*
3934 * Get the # of file system overhead blocks from the
3935 * superblock if present.
3936 */
3937 if (es->s_overhead_clusters)
3938 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3939 else {
07aa2ea1
LC		 3940 err = ext4_calculate_overhead(sb);
3941 if (err)
952fc18e
TT		 3942 goto failed_mount_wq;
3943 }
3944
fd89d5f2
TH		 3945 /*
3946 * The maximum number of concurrent works can be high and
3947 * concurrency isn't really necessary. Limit it to 1.
3948 */
2e8fa54e
JK		 3949 EXT4_SB(sb)->rsv_conversion_wq =
3950 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3951 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3952 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 3953 ret = -ENOMEM;
2e8fa54e
JK		 3954 goto failed_mount4;
3955 }
3956
3957 EXT4_SB(sb)->unrsv_conversion_wq =
3958 alloc_workqueue("ext4-unrsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3959 if (!EXT4_SB(sb)->unrsv_conversion_wq) {
3960 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3961 ret = -ENOMEM;
3962 goto failed_mount4;
4c0425ff
MC		 3963 }
3964
ac27a0ec 3965 /*
dab291af 3966 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3967 * so we can safely mount the rest of the filesystem now.
3968 */
3969
1d1fe1ee
DH		 3970 root = ext4_iget(sb, EXT4_ROOT_INO);
3971 if (IS_ERR(root)) {
b31e1552 3972 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3973 ret = PTR_ERR(root);
32a9bb57 3974 root = NULL;
ac27a0ec
DK		 3975 goto failed_mount4;
3976 }
3977 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3978 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3979 iput(root);
ac27a0ec
DK		 3980 goto failed_mount4;
3981 }
48fde701 3982 sb->s_root = d_make_root(root);
1d1fe1ee 3983 if (!sb->s_root) {
b31e1552 3984 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3985 ret = -ENOMEM;
3986 goto failed_mount4;
3987 }
ac27a0ec 3988
7e84b621
ES		 3989 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3990 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3991
3992 /* determine the minimum size of new large inodes, if present */
3993 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3994 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3995 EXT4_GOOD_OLD_INODE_SIZE;
3996 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3997 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3998 if (sbi->s_want_extra_isize <
3999 le16_to_cpu(es->s_want_extra_isize))
4000 sbi->s_want_extra_isize =
4001 le16_to_cpu(es->s_want_extra_isize);
4002 if (sbi->s_want_extra_isize <
4003 le16_to_cpu(es->s_min_extra_isize))
4004 sbi->s_want_extra_isize =
4005 le16_to_cpu(es->s_min_extra_isize);
4006 }
4007 }
4008 /* Check if enough inode space is available */
4009 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4010 sbi->s_inode_size) {
4011 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4012 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4013 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4014 "available");
ef7f3835
KS		 4015 }
4016
27dd4385
LC		 4017 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sbi));
4018 if (err) {
4019 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4020 "reserved pool", ext4_calculate_resv_clusters(sbi));
4021 goto failed_mount4a;
4022 }
4023
6fd058f7
TT		 4024 err = ext4_setup_system_zone(sb);
4025 if (err) {
b31e1552 4026 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4027 "zone (%d)", err);
94bf608a 4028 goto failed_mount4a;
6fd058f7
TT		 4029 }
4030
c2774d84 4031 ext4_ext_init(sb);
9d99012f 4032 err = ext4_mb_init(sb);
c2774d84 4033 if (err) {
421f91d2 4034 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 4035 err);
dcf2d804 4036 goto failed_mount5;
c2774d84
AK		 4037 }
4038
bfff6873
LC		 4039 err = ext4_register_li_request(sb, first_not_zeroed);
4040 if (err)
dcf2d804 4041 goto failed_mount6;
bfff6873 4042
3197ebdb
TT		 4043 sbi->s_kobj.kset = ext4_kset;
4044 init_completion(&sbi->s_kobj_unregister);
4045 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4046 "%s", sb->s_id);
dcf2d804
TM		 4047 if (err)
4048 goto failed_mount7;
3197ebdb 4049
9b2ff357
JK		 4050#ifdef CONFIG_QUOTA
4051 /* Enable quota usage during mount. */
4052 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4053 !(sb->s_flags & MS_RDONLY)) {
4054 err = ext4_enable_quotas(sb);
4055 if (err)
4056 goto failed_mount8;
4057 }
4058#endif /* CONFIG_QUOTA */
4059
617ba13b
MC		 4060 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4061 ext4_orphan_cleanup(sb, es);
4062 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4063 if (needs_recovery) {
b31e1552 4064 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4065 ext4_mark_recovery_complete(sb, es);
4066 }
4067 if (EXT4_SB(sb)->s_journal) {
4068 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4069 descr = " journalled data mode";
4070 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4071 descr = " ordered data mode";
4072 else
4073 descr = " writeback data mode";
4074 } else
4075 descr = "out journal";
4076
79add3a3
LC		 4077 if (test_opt(sb, DISCARD)) {
4078 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4079 if (!blk_queue_discard(q))
4080 ext4_msg(sb, KERN_WARNING,
4081 "mounting with \"discard\" option, but "
4082 "the device does not support discard");
4083 }
4084
d4c402d9 4085 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4086 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4087 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4088
66e61a9e
TT		 4089 if (es->s_error_count)
4090 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4091
d4c402d9 4092 kfree(orig_data);
ac27a0ec
DK		 4093 return 0;
4094
617ba13b 4095cantfind_ext4:
ac27a0ec 4096 if (!silent)
b31e1552 4097 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4098 goto failed_mount;
4099
72ba7450
TT		 4100#ifdef CONFIG_QUOTA
4101failed_mount8:
4102 kobject_del(&sbi->s_kobj);
4103#endif
dcf2d804
TM		 4104failed_mount7:
4105 ext4_unregister_li_request(sb);
4106failed_mount6:
dcf2d804 4107 ext4_mb_release(sb);
94bf608a
AV		 4108failed_mount5:
4109 ext4_ext_release(sb);
dcf2d804 4110 ext4_release_system_zone(sb);
94bf608a
AV		 4111failed_mount4a:
4112 dput(sb->s_root);
32a9bb57 4113 sb->s_root = NULL;
94bf608a 4114failed_mount4:
b31e1552 4115 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4116 if (EXT4_SB(sb)->rsv_conversion_wq)
4117 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4118 if (EXT4_SB(sb)->unrsv_conversion_wq)
4119 destroy_workqueue(EXT4_SB(sb)->unrsv_conversion_wq);
4c0425ff 4120failed_mount_wq:
0390131b
FM		 4121 if (sbi->s_journal) {
4122 jbd2_journal_destroy(sbi->s_journal);
4123 sbi->s_journal = NULL;
4124 }
ac27a0ec 4125failed_mount3:
d3922a77 4126 ext4_es_unregister_shrinker(sbi);
04496411 4127 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4128 if (sbi->s_flex_groups)
4129 ext4_kvfree(sbi->s_flex_groups);
57042651 4130 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4131 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4132 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4133 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 4134 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
c5e06d10
JL		 4135 if (sbi->s_mmp_tsk)
4136 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4137failed_mount2:
4138 for (i = 0; i < db_count; i++)
4139 brelse(sbi->s_group_desc[i]);
f18a5f21 4140 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4141failed_mount:
0441984a
DW		 4142 if (sbi->s_chksum_driver)
4143 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4144 if (sbi->s_proc) {
66acdcf4 4145 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4146 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4147 }
ac27a0ec
DK		 4148#ifdef CONFIG_QUOTA
4149 for (i = 0; i < MAXQUOTAS; i++)
4150 kfree(sbi->s_qf_names[i]);
4151#endif
617ba13b 4152 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4153 brelse(bh);
4154out_fail:
4155 sb->s_fs_info = NULL;
f6830165 4156 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4157 kfree(sbi);
dcc7dae3 4158out_free_orig:
d4c402d9 4159 kfree(orig_data);
07aa2ea1 4160 return err ? err : ret;
ac27a0ec
DK		 4161}
4162
4163/*
4164 * Setup any per-fs journal parameters now. We'll do this both on
4165 * initial mount, once the journal has been initialised but before we've
4166 * done any recovery; and again on any subsequent remount.
4167 */
617ba13b 4168static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4169{
617ba13b 4170 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4171
30773840
TT		 4172 journal->j_commit_interval = sbi->s_commit_interval;
4173 journal->j_min_batch_time = sbi->s_min_batch_time;
4174 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4175
a931da6a 4176 write_lock(&journal->j_state_lock);
ac27a0ec 4177 if (test_opt(sb, BARRIER))
dab291af 4178 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4179 else
dab291af 4180 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4181 if (test_opt(sb, DATA_ERR_ABORT))
4182 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4183 else
4184 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4185 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4186}
4187
617ba13b 4188static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4189 unsigned int journal_inum)
4190{
4191 struct inode *journal_inode;
4192 journal_t *journal;
4193
0390131b
FM		 4194 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4195
ac27a0ec
DK		 4196 /* First, test for the existence of a valid inode on disk. Bad
4197 * things happen if we iget() an unused inode, as the subsequent
4198 * iput() will try to delete it. */
4199
1d1fe1ee
DH		 4200 journal_inode = ext4_iget(sb, journal_inum);
4201 if (IS_ERR(journal_inode)) {
b31e1552 4202 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4203 return NULL;
4204 }
4205 if (!journal_inode->i_nlink) {
4206 make_bad_inode(journal_inode);
4207 iput(journal_inode);
b31e1552 4208 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4209 return NULL;
4210 }
4211
e5f8eab8 4212 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4213 journal_inode, journal_inode->i_size);
1d1fe1ee 4214 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4215 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4216 iput(journal_inode);
4217 return NULL;
4218 }
4219
dab291af 4220 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4221 if (!journal) {
b31e1552 4222 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4223 iput(journal_inode);
4224 return NULL;
4225 }
4226 journal->j_private = sb;
617ba13b 4227 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4228 return journal;
4229}
4230
617ba13b 4231static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4232 dev_t j_dev)
4233{
2b2d6d01 4234 struct buffer_head *bh;
ac27a0ec 4235 journal_t *journal;
617ba13b
MC		 4236 ext4_fsblk_t start;
4237 ext4_fsblk_t len;
ac27a0ec 4238 int hblock, blocksize;
617ba13b 4239 ext4_fsblk_t sb_block;
ac27a0ec 4240 unsigned long offset;
2b2d6d01 4241 struct ext4_super_block *es;
ac27a0ec
DK		 4242 struct block_device *bdev;
4243
0390131b
FM		 4244 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4245
b31e1552 4246 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4247 if (bdev == NULL)
4248 return NULL;
4249
ac27a0ec 4250 blocksize = sb->s_blocksize;
e1defc4f 4251 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4252 if (blocksize < hblock) {
b31e1552
ES		 4253 ext4_msg(sb, KERN_ERR,
4254 "blocksize too small for journal device");
ac27a0ec
DK		 4255 goto out_bdev;
4256 }
4257
617ba13b
MC		 4258 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4259 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4260 set_blocksize(bdev, blocksize);
4261 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4262 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4263 "external journal");
ac27a0ec
DK		 4264 goto out_bdev;
4265 }
4266
2716b802 4267 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4268 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4269 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4270 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4271 ext4_msg(sb, KERN_ERR, "external journal has "
4272 "bad superblock");
ac27a0ec
DK		 4273 brelse(bh);
4274 goto out_bdev;
4275 }
4276
617ba13b 4277 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4278 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4279 brelse(bh);
4280 goto out_bdev;
4281 }
4282
bd81d8ee 4283 len = ext4_blocks_count(es);
ac27a0ec
DK		 4284 start = sb_block + 1;
4285 brelse(bh); /* we're done with the superblock */
4286
dab291af 4287 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4288 start, len, blocksize);
4289 if (!journal) {
b31e1552 4290 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4291 goto out_bdev;
4292 }
4293 journal->j_private = sb;
9f203507 4294 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4295 wait_on_buffer(journal->j_sb_buffer);
4296 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4297 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4298 goto out_journal;
4299 }
4300 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4301 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4302 "user (unsupported) - %d",
ac27a0ec
DK		 4303 be32_to_cpu(journal->j_superblock->s_nr_users));
4304 goto out_journal;
4305 }
617ba13b
MC		 4306 EXT4_SB(sb)->journal_bdev = bdev;
4307 ext4_init_journal_params(sb, journal);
ac27a0ec 4308 return journal;
0b8e58a1 4309
ac27a0ec 4310out_journal:
dab291af 4311 jbd2_journal_destroy(journal);
ac27a0ec 4312out_bdev:
617ba13b 4313 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4314 return NULL;
4315}
4316
617ba13b
MC		 4317static int ext4_load_journal(struct super_block *sb,
4318 struct ext4_super_block *es,
ac27a0ec
DK		 4319 unsigned long journal_devnum)
4320{
4321 journal_t *journal;
4322 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4323 dev_t journal_dev;
4324 int err = 0;
4325 int really_read_only;
4326
0390131b
FM		 4327 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4328
ac27a0ec
DK		 4329 if (journal_devnum &&
4330 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4331 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4332 "numbers have changed");
ac27a0ec
DK		 4333 journal_dev = new_decode_dev(journal_devnum);
4334 } else
4335 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4336
4337 really_read_only = bdev_read_only(sb->s_bdev);
4338
4339 /*
4340 * Are we loading a blank journal or performing recovery after a
4341 * crash? For recovery, we need to check in advance whether we
4342 * can get read-write access to the device.
4343 */
617ba13b 4344 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4345 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4346 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4347 "required on readonly filesystem");
ac27a0ec 4348 if (really_read_only) {
b31e1552
ES		 4349 ext4_msg(sb, KERN_ERR, "write access "
4350 "unavailable, cannot proceed");
ac27a0ec
DK		 4351 return -EROFS;
4352 }
b31e1552
ES		 4353 ext4_msg(sb, KERN_INFO, "write access will "
4354 "be enabled during recovery");
ac27a0ec
DK		 4355 }
4356 }
4357
4358 if (journal_inum && journal_dev) {
b31e1552
ES		 4359 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4360 "and inode journals!");
ac27a0ec
DK		 4361 return -EINVAL;
4362 }
4363
4364 if (journal_inum) {
617ba13b 4365 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4366 return -EINVAL;
4367 } else {
617ba13b 4368 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4369 return -EINVAL;
4370 }
4371
90576c0b 4372 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4373 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4374
617ba13b 4375 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4376 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4377 if (!err) {
4378 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4379 if (save)
4380 memcpy(save, ((char *) es) +
4381 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4382 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4383 if (save)
4384 memcpy(((char *) es) + EXT4_S_ERR_START,
4385 save, EXT4_S_ERR_LEN);
4386 kfree(save);
4387 }
ac27a0ec
DK		 4388
4389 if (err) {
b31e1552 4390 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4391 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4392 return err;
4393 }
4394
617ba13b
MC		 4395 EXT4_SB(sb)->s_journal = journal;
4396 ext4_clear_journal_err(sb, es);
ac27a0ec 4397
c41303ce 4398 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4399 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4400 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4401
4402 /* Make sure we flush the recovery flag to disk. */
e2d67052 4403 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4404 }
4405
4406 return 0;
4407}
4408
e2d67052 4409static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4410{
e2d67052 4411 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4412 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4413 int error = 0;
ac27a0ec 4414
7c2e7087 4415 if (!sbh || block_device_ejected(sb))
c4be0c1d 4416 return error;
914258bf
TT		 4417 if (buffer_write_io_error(sbh)) {
4418 /*
4419 * Oh, dear. A previous attempt to write the
4420 * superblock failed. This could happen because the
4421 * USB device was yanked out. Or it could happen to
4422 * be a transient write error and maybe the block will
4423 * be remapped. Nothing we can do but to retry the
4424 * write and hope for the best.
4425 */
b31e1552
ES		 4426 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4427 "superblock detected");
914258bf
TT		 4428 clear_buffer_write_io_error(sbh);
4429 set_buffer_uptodate(sbh);
4430 }
71290b36
TT		 4431 /*
4432 * If the file system is mounted read-only, don't update the
4433 * superblock write time. This avoids updating the superblock
4434 * write time when we are mounting the root file system
4435 * read/only but we need to replay the journal; at that point,
4436 * for people who are east of GMT and who make their clock
4437 * tick in localtime for Windows bug-for-bug compatibility,
4438 * the clock is set in the future, and this will cause e2fsck
4439 * to complain and force a full file system check.
4440 */
4441 if (!(sb->s_flags & MS_RDONLY))
4442 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4443 if (sb->s_bdev->bd_part)
4444 es->s_kbytes_written =
4445 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4446 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4447 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4448 else
4449 es->s_kbytes_written =
4450 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4451 ext4_free_blocks_count_set(es,
4452 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4453 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4454 es->s_free_inodes_count =
4455 cpu_to_le32(percpu_counter_sum_positive(
4456 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4457 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4458 ext4_superblock_csum_set(sb);
ac27a0ec 4459 mark_buffer_dirty(sbh);
914258bf 4460 if (sync) {
c4be0c1d
TS		 4461 error = sync_dirty_buffer(sbh);
4462 if (error)
4463 return error;
4464
4465 error = buffer_write_io_error(sbh);
4466 if (error) {
b31e1552
ES		 4467 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4468 "superblock");
914258bf
TT		 4469 clear_buffer_write_io_error(sbh);
4470 set_buffer_uptodate(sbh);
4471 }
4472 }
c4be0c1d 4473 return error;
ac27a0ec
DK		 4474}
4475
ac27a0ec
DK		 4476/*
4477 * Have we just finished recovery? If so, and if we are mounting (or
4478 * remounting) the filesystem readonly, then we will end up with a
4479 * consistent fs on disk. Record that fact.
4480 */
2b2d6d01
TT		 4481static void ext4_mark_recovery_complete(struct super_block *sb,
4482 struct ext4_super_block *es)
ac27a0ec 4483{
617ba13b 4484 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4485
0390131b
FM		 4486 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4487 BUG_ON(journal != NULL);
4488 return;
4489 }
dab291af 4490 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4491 if (jbd2_journal_flush(journal) < 0)
4492 goto out;
4493
617ba13b 4494 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4495 sb->s_flags & MS_RDONLY) {
617ba13b 4496 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4497 ext4_commit_super(sb, 1);
ac27a0ec 4498 }
7ffe1ea8
HK		 4499
4500out:
dab291af 4501 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4502}
4503
4504/*
4505 * If we are mounting (or read-write remounting) a filesystem whose journal
4506 * has recorded an error from a previous lifetime, move that error to the
4507 * main filesystem now.
4508 */
2b2d6d01
TT		 4509static void ext4_clear_journal_err(struct super_block *sb,
4510 struct ext4_super_block *es)
ac27a0ec
DK		 4511{
4512 journal_t *journal;
4513 int j_errno;
4514 const char *errstr;
4515
0390131b
FM		 4516 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4517
617ba13b 4518 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4519
4520 /*
4521 * Now check for any error status which may have been recorded in the
617ba13b 4522 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4523 */
4524
dab291af 4525 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4526 if (j_errno) {
4527 char nbuf[16];
4528
617ba13b 4529 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4530 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4531 "from previous mount: %s", errstr);
12062ddd 4532 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4533
617ba13b
MC		 4534 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4535 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4536 ext4_commit_super(sb, 1);
ac27a0ec 4537
dab291af 4538 jbd2_journal_clear_err(journal);
d796c52e 4539 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4540 }
4541}
4542
4543/*
4544 * Force the running and committing transactions to commit,
4545 * and wait on the commit.
4546 */
617ba13b 4547int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4548{
4549 journal_t *journal;
ac27a0ec
DK		 4550
4551 if (sb->s_flags & MS_RDONLY)
4552 return 0;
4553
617ba13b 4554 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4555 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4556}
4557
617ba13b 4558static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4559{
14ce0cb4 4560 int ret = 0;
9eddacf9 4561 tid_t target;
06a407f1 4562 bool needs_barrier = false;
8d5d02e6 4563 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4564
9bffad1e 4565 trace_ext4_sync_fs(sb, wait);
2e8fa54e
JK		 4566 flush_workqueue(sbi->rsv_conversion_wq);
4567 flush_workqueue(sbi->unrsv_conversion_wq);
a1177825
JK		 4568 /*
4569 * Writeback quota in non-journalled quota case - journalled quota has
4570 * no dirty dquots
4571 */
4572 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4573 /*
4574 * Data writeback is possible w/o journal transaction, so barrier must
4575 * being sent at the end of the function. But we can skip it if
4576 * transaction_commit will do it for us.
4577 */
4578 target = jbd2_get_latest_transaction(sbi->s_journal);
4579 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4580 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4581 needs_barrier = true;
4582
8d5d02e6 4583 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4584 if (wait)
06a407f1
DM		 4585 ret = jbd2_log_wait_commit(sbi->s_journal, target);
4586 }
4587 if (needs_barrier) {
4588 int err;
4589 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4590 if (!ret)
4591 ret = err;
0390131b 4592 }
06a407f1
DM		 4593
4594 return ret;
4595}
4596
4597static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4598{
4599 int ret = 0;
4600
4601 trace_ext4_sync_fs(sb, wait);
4602 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4603 flush_workqueue(EXT4_SB(sb)->unrsv_conversion_wq);
4604 dquot_writeback_dquots(sb, -1);
4605 if (wait && test_opt(sb, BARRIER))
4606 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4607
14ce0cb4 4608 return ret;
ac27a0ec
DK		 4609}
4610
4611/*
4612 * LVM calls this function before a (read-only) snapshot is created. This
4613 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4614 *
4615 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4616 * state independently. It relies on upper layer to stop all data & metadata
4617 * modifications.
ac27a0ec 4618 */
c4be0c1d 4619static int ext4_freeze(struct super_block *sb)
ac27a0ec 4620{
c4be0c1d
TS		 4621 int error = 0;
4622 journal_t *journal;
ac27a0ec 4623
9ca92389
TT		 4624 if (sb->s_flags & MS_RDONLY)
4625 return 0;
ac27a0ec 4626
9ca92389 4627 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4628
9ca92389
TT		 4629 /* Now we set up the journal barrier. */
4630 jbd2_journal_lock_updates(journal);
ac27a0ec 4631
9ca92389
TT		 4632 /*
4633 * Don't clear the needs_recovery flag if we failed to flush
4634 * the journal.
4635 */
4636 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4637 if (error < 0)
4638 goto out;
9ca92389
TT		 4639
4640 /* Journal blocked and flushed, clear needs_recovery flag. */
4641 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4642 error = ext4_commit_super(sb, 1);
6b0310fb 4643out:
8e8ad8a5 4644 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4645 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4646 return error;
ac27a0ec
DK		 4647}
4648
4649/*
4650 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4651 * flag here, even though the filesystem is not technically dirty yet.
4652 */
c4be0c1d 4653static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4654{
9ca92389
TT		 4655 if (sb->s_flags & MS_RDONLY)
4656 return 0;
4657
9ca92389
TT		 4658 /* Reset the needs_recovery flag before the fs is unlocked. */
4659 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4660 ext4_commit_super(sb, 1);
c4be0c1d 4661 return 0;
ac27a0ec
DK		 4662}
4663
673c6100
TT		 4664/*
4665 * Structure to save mount options for ext4_remount's benefit
4666 */
4667struct ext4_mount_options {
4668 unsigned long s_mount_opt;
a2595b8a 4669 unsigned long s_mount_opt2;
08cefc7a
EB		 4670 kuid_t s_resuid;
4671 kgid_t s_resgid;
673c6100
TT		 4672 unsigned long s_commit_interval;
4673 u32 s_min_batch_time, s_max_batch_time;
4674#ifdef CONFIG_QUOTA
4675 int s_jquota_fmt;
4676 char *s_qf_names[MAXQUOTAS];
4677#endif
4678};
4679
2b2d6d01 4680static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4681{
2b2d6d01 4682 struct ext4_super_block *es;
617ba13b 4683 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4684 unsigned long old_sb_flags;
617ba13b 4685 struct ext4_mount_options old_opts;
c79d967d 4686 int enable_quota = 0;
8a266467 4687 ext4_group_t g;
b3881f74 4688 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4689 int err = 0;
ac27a0ec 4690#ifdef CONFIG_QUOTA
03dafb5f 4691 int i, j;
ac27a0ec 4692#endif
d4c402d9 4693 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4694
4695 /* Store the original options */
4696 old_sb_flags = sb->s_flags;
4697 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4698 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4699 old_opts.s_resuid = sbi->s_resuid;
4700 old_opts.s_resgid = sbi->s_resgid;
4701 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4702 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4703 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4704#ifdef CONFIG_QUOTA
4705 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4706 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f
CG		 4707 if (sbi->s_qf_names[i]) {
4708 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4709 GFP_KERNEL);
4710 if (!old_opts.s_qf_names[i]) {
4711 for (j = 0; j < i; j++)
4712 kfree(old_opts.s_qf_names[j]);
3e36a163 4713 kfree(orig_data);
03dafb5f
CG		 4714 return -ENOMEM;
4715 }
4716 } else
4717 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4718#endif
b3881f74
TT		 4719 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4720 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4721
4722 /*
4723 * Allow the "check" option to be passed as a remount option.
4724 */
661aa520 4725 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4726 err = -EINVAL;
4727 goto restore_opts;
4728 }
4729
4ab2f15b 4730 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4731 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4732
4733 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4734 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4735
4736 es = sbi->s_es;
4737
b3881f74 4738 if (sbi->s_journal) {
0390131b 4739 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4740 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4741 }
ac27a0ec 4742
661aa520 4743 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4744 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4745 err = -EROFS;
4746 goto restore_opts;
4747 }
4748
4749 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4750 err = dquot_suspend(sb, -1);
4751 if (err < 0)
c79d967d 4752 goto restore_opts;
c79d967d 4753
ac27a0ec
DK		 4754 /*
4755 * First of all, the unconditional stuff we have to do
4756 * to disable replay of the journal when we next remount
4757 */
4758 sb->s_flags |= MS_RDONLY;
4759
4760 /*
4761 * OK, test if we are remounting a valid rw partition
4762 * readonly, and if so set the rdonly flag and then
4763 * mark the partition as valid again.
4764 */
617ba13b
MC		 4765 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4766 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4767 es->s_state = cpu_to_le16(sbi->s_mount_state);
4768
a63c9eb2 4769 if (sbi->s_journal)
0390131b 4770 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4771 } else {
a13fb1a4
ES		 4772 /* Make sure we can mount this feature set readwrite */
4773 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4774 err = -EROFS;
4775 goto restore_opts;
4776 }
8a266467
TT		 4777 /*
4778 * Make sure the group descriptor checksums
0b8e58a1 4779 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4780 */
4781 for (g = 0; g < sbi->s_groups_count; g++) {
4782 struct ext4_group_desc *gdp =
4783 ext4_get_group_desc(sb, g, NULL);
4784
feb0ab32 4785 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4786 ext4_msg(sb, KERN_ERR,
4787 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4788 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4789 le16_to_cpu(gdp->bg_checksum));
4790 err = -EINVAL;
4791 goto restore_opts;
4792 }
4793 }
4794
ead6596b
ES		 4795 /*
4796 * If we have an unprocessed orphan list hanging
4797 * around from a previously readonly bdev mount,
4798 * require a full umount/remount for now.
4799 */
4800 if (es->s_last_orphan) {
b31e1552 4801 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4802 "remount RDWR because of unprocessed "
4803 "orphan inode list. Please "
b31e1552 4804 "umount/remount instead");
ead6596b
ES		 4805 err = -EINVAL;
4806 goto restore_opts;
4807 }
4808
ac27a0ec
DK		 4809 /*
4810 * Mounting a RDONLY partition read-write, so reread
4811 * and store the current valid flag. (It may have
4812 * been changed by e2fsck since we originally mounted
4813 * the partition.)
4814 */
0390131b
FM		 4815 if (sbi->s_journal)
4816 ext4_clear_journal_err(sb, es);
ac27a0ec 4817 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4818 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4819 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4820 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4821 EXT4_FEATURE_INCOMPAT_MMP))
4822 if (ext4_multi_mount_protect(sb,
4823 le64_to_cpu(es->s_mmp_block))) {
4824 err = -EROFS;
4825 goto restore_opts;
4826 }
c79d967d 4827 enable_quota = 1;
ac27a0ec
DK		 4828 }
4829 }
bfff6873
LC		 4830
4831 /*
4832 * Reinitialize lazy itable initialization thread based on
4833 * current settings
4834 */
4835 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4836 ext4_unregister_li_request(sb);
4837 else {
4838 ext4_group_t first_not_zeroed;
4839 first_not_zeroed = ext4_has_uninit_itable(sb);
4840 ext4_register_li_request(sb, first_not_zeroed);
4841 }
4842
6fd058f7 4843 ext4_setup_system_zone(sb);
d096ad0f 4844 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4845 ext4_commit_super(sb, 1);
0390131b 4846
ac27a0ec
DK		 4847#ifdef CONFIG_QUOTA
4848 /* Release old quota file names */
4849 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f 4850 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4851 if (enable_quota) {
4852 if (sb_any_quota_suspended(sb))
4853 dquot_resume(sb, -1);
4854 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4855 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4856 err = ext4_enable_quotas(sb);
07724f98 4857 if (err)
7c319d32 4858 goto restore_opts;
7c319d32
AK		 4859 }
4860 }
ac27a0ec 4861#endif
d4c402d9
CW		 4862
4863 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4864 kfree(orig_data);
ac27a0ec 4865 return 0;
0b8e58a1 4866
ac27a0ec
DK		 4867restore_opts:
4868 sb->s_flags = old_sb_flags;
4869 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4870 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4871 sbi->s_resuid = old_opts.s_resuid;
4872 sbi->s_resgid = old_opts.s_resgid;
4873 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4874 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4875 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4876#ifdef CONFIG_QUOTA
4877 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4878 for (i = 0; i < MAXQUOTAS; i++) {
03dafb5f 4879 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4880 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4881 }
4882#endif
d4c402d9 4883 kfree(orig_data);
ac27a0ec
DK		 4884 return err;
4885}
4886
2b2d6d01 4887static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4888{
4889 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4890 struct ext4_sb_info *sbi = EXT4_SB(sb);
4891 struct ext4_super_block *es = sbi->s_es;
27dd4385 4892 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 4893 u64 fsid;
d02a9391 4894 s64 bfree;
27dd4385 4895 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 4896
952fc18e
TT		 4897 if (!test_opt(sb, MINIX_DF))
4898 overhead = sbi->s_overhead;
ac27a0ec 4899
617ba13b 4900 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4901 buf->f_bsize = sb->s_blocksize;
b72f78cb 4902 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4903 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4904 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4905 /* prevent underflow in case that few free space is available */
57042651 4906 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 4907 buf->f_bavail = buf->f_bfree -
4908 (ext4_r_blocks_count(es) + resv_blocks);
4909 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 4910 buf->f_bavail = 0;
4911 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4912 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4913 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4914 fsid = le64_to_cpup((void *)es->s_uuid) ^
4915 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4916 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4917 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4918
ac27a0ec
DK		 4919 return 0;
4920}
4921
0b8e58a1
AD		 4922/* Helper function for writing quotas on sync - we need to start transaction
4923 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4924 * Process 1 Process 2
617ba13b 4925 * ext4_create() quota_sync()
a269eb18 4926 * jbd2_journal_start() write_dquot()
871a2931 4927 * dquot_initialize() down(dqio_mutex)
dab291af 4928 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4929 *
4930 */
4931
4932#ifdef CONFIG_QUOTA
4933
4934static inline struct inode *dquot_to_inode(struct dquot *dquot)
4935{
4c376dca 4936 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4937}
4938
617ba13b 4939static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4940{
4941 int ret, err;
4942 handle_t *handle;
4943 struct inode *inode;
4944
4945 inode = dquot_to_inode(dquot);
9924a92a 4946 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 4947 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4948 if (IS_ERR(handle))
4949 return PTR_ERR(handle);
4950 ret = dquot_commit(dquot);
617ba13b 4951 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4952 if (!ret)
4953 ret = err;
4954 return ret;
4955}
4956
617ba13b 4957static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4958{
4959 int ret, err;
4960 handle_t *handle;
4961
9924a92a 4962 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4963 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4964 if (IS_ERR(handle))
4965 return PTR_ERR(handle);
4966 ret = dquot_acquire(dquot);
617ba13b 4967 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4968 if (!ret)
4969 ret = err;
4970 return ret;
4971}
4972
617ba13b 4973static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4974{
4975 int ret, err;
4976 handle_t *handle;
4977
9924a92a 4978 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4979 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4980 if (IS_ERR(handle)) {
4981 /* Release dquot anyway to avoid endless cycle in dqput() */
4982 dquot_release(dquot);
ac27a0ec 4983 return PTR_ERR(handle);
9c3013e9 4984 }
ac27a0ec 4985 ret = dquot_release(dquot);
617ba13b 4986 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4987 if (!ret)
4988 ret = err;
4989 return ret;
4990}
4991
617ba13b 4992static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4993{
262b4662
JK		 4994 struct super_block *sb = dquot->dq_sb;
4995 struct ext4_sb_info *sbi = EXT4_SB(sb);
4996
2c8be6b2 4997 /* Are we journaling quotas? */
262b4662
JK		 4998 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
4999 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5000 dquot_mark_dquot_dirty(dquot);
617ba13b 5001 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5002 } else {
5003 return dquot_mark_dquot_dirty(dquot);
5004 }
5005}
5006
617ba13b 5007static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5008{
5009 int ret, err;
5010 handle_t *handle;
5011
5012 /* Data block + inode block */
9924a92a 5013 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5014 if (IS_ERR(handle))
5015 return PTR_ERR(handle);
5016 ret = dquot_commit_info(sb, type);
617ba13b 5017 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5018 if (!ret)
5019 ret = err;
5020 return ret;
5021}
5022
5023/*
5024 * Turn on quotas during mount time - we need to find
5025 * the quota file and such...
5026 */
617ba13b 5027static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5028{
287a8095
CH		 5029 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5030 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5031}
5032
5033/*
5034 * Standard function to be called on quota_on
5035 */
617ba13b 5036static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5037 struct path *path)
ac27a0ec
DK		 5038{
5039 int err;
ac27a0ec
DK		 5040
5041 if (!test_opt(sb, QUOTA))
5042 return -EINVAL;
0623543b 5043
ac27a0ec 5044 /* Quotafile not on the same filesystem? */
d8c9584e 5045 if (path->dentry->d_sb != sb)
ac27a0ec 5046 return -EXDEV;
0623543b
JK		 5047 /* Journaling quota? */
5048 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5049 /* Quotafile not in fs root? */
f00c9e44 5050 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5051 ext4_msg(sb, KERN_WARNING,
5052 "Quota file not on filesystem root. "
5053 "Journaled quota will not work");
2b2d6d01 5054 }
0623543b
JK		 5055
5056 /*
5057 * When we journal data on quota file, we have to flush journal to see
5058 * all updates to the file when we bypass pagecache...
5059 */
0390131b 5060 if (EXT4_SB(sb)->s_journal &&
f00c9e44 5061 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 5062 /*
5063 * We don't need to lock updates but journal_flush() could
5064 * otherwise be livelocked...
5065 */
5066 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5067 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5068 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5069 if (err)
7ffe1ea8 5070 return err;
0623543b
JK		 5071 }
5072
f00c9e44 5073 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5074}
5075
7c319d32
AK		 5076static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5077 unsigned int flags)
5078{
5079 int err;
5080 struct inode *qf_inode;
5081 unsigned long qf_inums[MAXQUOTAS] = {
5082 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5083 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5084 };
5085
5086 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5087
5088 if (!qf_inums[type])
5089 return -EPERM;
5090
5091 qf_inode = ext4_iget(sb, qf_inums[type]);
5092 if (IS_ERR(qf_inode)) {
5093 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5094 return PTR_ERR(qf_inode);
5095 }
5096
bcb13850
JK		 5097 /* Don't account quota for quota files to avoid recursion */
5098 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5099 err = dquot_enable(qf_inode, type, format_id, flags);
5100 iput(qf_inode);
5101
5102 return err;
5103}
5104
5105/* Enable usage tracking for all quota types. */
5106static int ext4_enable_quotas(struct super_block *sb)
5107{
5108 int type, err = 0;
5109 unsigned long qf_inums[MAXQUOTAS] = {
5110 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5111 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5112 };
5113
5114 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5115 for (type = 0; type < MAXQUOTAS; type++) {
5116 if (qf_inums[type]) {
5117 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5118 DQUOT_USAGE_ENABLED);
5119 if (err) {
5120 ext4_warning(sb,
72ba7450
TT		 5121 "Failed to enable quota tracking "
5122 "(type=%d, err=%d). Please run "
5123 "e2fsck to fix.", type, err);
7c319d32
AK		 5124 return err;
5125 }
5126 }
5127 }
5128 return 0;
5129}
5130
5131/*
5132 * quota_on function that is used when QUOTA feature is set.
5133 */
5134static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5135 int format_id)
5136{
5137 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5138 return -EINVAL;
5139
5140 /*
5141 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5142 */
5143 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5144}
5145
ca0e05e4
DM		 5146static int ext4_quota_off(struct super_block *sb, int type)
5147{
21f97697
JK		 5148 struct inode *inode = sb_dqopt(sb)->files[type];
5149 handle_t *handle;
5150
87009d86
DM		 5151 /* Force all delayed allocation blocks to be allocated.
5152 * Caller already holds s_umount sem */
5153 if (test_opt(sb, DELALLOC))
ca0e05e4 5154 sync_filesystem(sb);
ca0e05e4 5155
0b268590
AG		 5156 if (!inode)
5157 goto out;
5158
21f97697
JK		 5159 /* Update modification times of quota files when userspace can
5160 * start looking at them */
9924a92a 5161 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5162 if (IS_ERR(handle))
5163 goto out;
5164 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5165 ext4_mark_inode_dirty(handle, inode);
5166 ext4_journal_stop(handle);
5167
5168out:
ca0e05e4
DM		 5169 return dquot_quota_off(sb, type);
5170}
5171
7c319d32
AK		 5172/*
5173 * quota_off function that is used when QUOTA feature is set.
5174 */
5175static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5176{
5177 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5178 return -EINVAL;
5179
5180 /* Disable only the limits. */
5181 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5182}
5183
ac27a0ec
DK		 5184/* Read data from quotafile - avoid pagecache and such because we cannot afford
5185 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5186 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5187 * we don't have to be afraid of races */
617ba13b 5188static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5189 size_t len, loff_t off)
5190{
5191 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5192 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5193 int err = 0;
5194 int offset = off & (sb->s_blocksize - 1);
5195 int tocopy;
5196 size_t toread;
5197 struct buffer_head *bh;
5198 loff_t i_size = i_size_read(inode);
5199
5200 if (off > i_size)
5201 return 0;
5202 if (off+len > i_size)
5203 len = i_size-off;
5204 toread = len;
5205 while (toread > 0) {
5206 tocopy = sb->s_blocksize - offset < toread ?
5207 sb->s_blocksize - offset : toread;
617ba13b 5208 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5209 if (err)
5210 return err;
5211 if (!bh) /* A hole? */
5212 memset(data, 0, tocopy);
5213 else
5214 memcpy(data, bh->b_data+offset, tocopy);
5215 brelse(bh);
5216 offset = 0;
5217 toread -= tocopy;
5218 data += tocopy;
5219 blk++;
5220 }
5221 return len;
5222}
5223
5224/* Write to quotafile (we know the transaction is already started and has
5225 * enough credits) */
617ba13b 5226static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5227 const char *data, size_t len, loff_t off)
5228{
5229 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5230 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5231 int err = 0;
5232 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5233 struct buffer_head *bh;
5234 handle_t *handle = journal_current_handle();
5235
0390131b 5236 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5237 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5238 " cancelled because transaction is not started",
9c3013e9
JK		 5239 (unsigned long long)off, (unsigned long long)len);
5240 return -EIO;
5241 }
67eeb568
DM		 5242 /*
5243 * Since we account only one data block in transaction credits,
5244 * then it is impossible to cross a block boundary.
5245 */
5246 if (sb->s_blocksize - offset < len) {
5247 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5248 " cancelled because not block aligned",
5249 (unsigned long long)off, (unsigned long long)len);
5250 return -EIO;
5251 }
5252
67eeb568
DM		 5253 bh = ext4_bread(handle, inode, blk, 1, &err);
5254 if (!bh)
5255 goto out;
62d2b5f2
JK		 5256 err = ext4_journal_get_write_access(handle, bh);
5257 if (err) {
5258 brelse(bh);
5259 goto out;
ac27a0ec 5260 }
67eeb568
DM		 5261 lock_buffer(bh);
5262 memcpy(bh->b_data+offset, data, len);
5263 flush_dcache_page(bh->b_page);
5264 unlock_buffer(bh);
62d2b5f2 5265 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5266 brelse(bh);
ac27a0ec 5267out:
0b7f7cef 5268 if (err)
ac27a0ec 5269 return err;
67eeb568
DM		 5270 if (inode->i_size < off + len) {
5271 i_size_write(inode, off + len);
617ba13b 5272 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5273 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5274 }
67eeb568 5275 return len;
ac27a0ec
DK		 5276}
5277
5278#endif
5279
152a0836
AV		 5280static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5281 const char *dev_name, void *data)
ac27a0ec 5282{
152a0836 5283 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5284}
5285
37f328eb 5286#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5287static inline void register_as_ext2(void)
5288{
5289 int err = register_filesystem(&ext2_fs_type);
5290 if (err)
5291 printk(KERN_WARNING
5292 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5293}
5294
5295static inline void unregister_as_ext2(void)
5296{
5297 unregister_filesystem(&ext2_fs_type);
5298}
2035e776
TT		 5299
5300static inline int ext2_feature_set_ok(struct super_block *sb)
5301{
5302 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5303 return 0;
5304 if (sb->s_flags & MS_RDONLY)
5305 return 1;
5306 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5307 return 0;
5308 return 1;
5309}
24b58424
TT		 5310#else
5311static inline void register_as_ext2(void) { }
5312static inline void unregister_as_ext2(void) { }
2035e776 5313static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5314#endif
5315
37f328eb 5316#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5317static inline void register_as_ext3(void)
5318{
5319 int err = register_filesystem(&ext3_fs_type);
5320 if (err)
5321 printk(KERN_WARNING
5322 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5323}
5324
5325static inline void unregister_as_ext3(void)
5326{
5327 unregister_filesystem(&ext3_fs_type);
5328}
2035e776
TT		 5329
5330static inline int ext3_feature_set_ok(struct super_block *sb)
5331{
5332 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5333 return 0;
5334 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5335 return 0;
5336 if (sb->s_flags & MS_RDONLY)
5337 return 1;
5338 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5339 return 0;
5340 return 1;
5341}
24b58424
TT		 5342#else
5343static inline void register_as_ext3(void) { }
5344static inline void unregister_as_ext3(void) { }
2035e776 5345static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5346#endif
5347
03010a33
TT		 5348static struct file_system_type ext4_fs_type = {
5349 .owner = THIS_MODULE,
5350 .name = "ext4",
152a0836 5351 .mount = ext4_mount,
03010a33
TT		 5352 .kill_sb = kill_block_super,
5353 .fs_flags = FS_REQUIRES_DEV,
5354};
7f78e035 5355MODULE_ALIAS_FS("ext4");
03010a33 5356
8f021222 5357static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5358{
5359 struct ext4_features *ef;
5360 int ret = -ENOMEM;
5361
5362 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5363 if (!ef)
5364 goto out;
5365
5366 ef->f_kobj.kset = ext4_kset;
5367 init_completion(&ef->f_kobj_unregister);
5368 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5369 "features");
5370 if (ret) {
5371 kfree(ef);
5372 goto out;
5373 }
5374
5375 ext4_feat = ef;
5376 ret = 0;
5377out:
5378 return ret;
5379}
5380
8f021222
LC		 5381static void ext4_exit_feat_adverts(void)
5382{
5383 kobject_put(&ext4_feat->f_kobj);
5384 wait_for_completion(&ext4_feat->f_kobj_unregister);
5385 kfree(ext4_feat);
5386}
5387
e9e3bcec
ES		 5388/* Shared across all ext4 file systems */
5389wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5390struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5391
5dabfc78 5392static int __init ext4_init_fs(void)
ac27a0ec 5393{
e9e3bcec 5394 int i, err;
c9de560d 5395
07c0c5d8
AV		 5396 ext4_li_info = NULL;
5397 mutex_init(&ext4_li_mtx);
5398
9a4c8019 5399 /* Build-time check for flags consistency */
12e9b892 5400 ext4_check_flag_values();
e9e3bcec
ES		 5401
5402 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5403 mutex_init(&ext4__aio_mutex[i]);
5404 init_waitqueue_head(&ext4__ioend_wq[i]);
5405 }
5406
51865fda 5407 err = ext4_init_es();
6fd058f7
TT		 5408 if (err)
5409 return err;
51865fda
ZL		 5410
5411 err = ext4_init_pageio();
5412 if (err)
5413 goto out7;
5414
5dabfc78 5415 err = ext4_init_system_zone();
bd2d0210 5416 if (err)
d44651d0 5417 goto out6;
3197ebdb 5418 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5419 if (!ext4_kset) {
5420 err = -ENOMEM;
dd68314c 5421 goto out5;
0e376b1e 5422 }
d44651d0 5423 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5424
5425 err = ext4_init_feat_adverts();
dd68314c
TT		 5426 if (err)
5427 goto out4;
857ac889 5428
5dabfc78 5429 err = ext4_init_mballoc();
ac27a0ec 5430 if (err)
6fd058f7 5431 goto out3;
c9de560d 5432
5dabfc78 5433 err = ext4_init_xattr();
c9de560d
AT		 5434 if (err)
5435 goto out2;
ac27a0ec
DK		 5436 err = init_inodecache();
5437 if (err)
5438 goto out1;
24b58424 5439 register_as_ext3();
2035e776 5440 register_as_ext2();
03010a33 5441 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5442 if (err)
5443 goto out;
bfff6873 5444
ac27a0ec
DK		 5445 return 0;
5446out:
24b58424
TT		 5447 unregister_as_ext2();
5448 unregister_as_ext3();
ac27a0ec
DK		 5449 destroy_inodecache();
5450out1:
5dabfc78 5451 ext4_exit_xattr();
c9de560d 5452out2:
5dabfc78 5453 ext4_exit_mballoc();
6fd058f7 5454out3:
8f021222 5455 ext4_exit_feat_adverts();
dd68314c 5456out4:
d44651d0
FJ		 5457 if (ext4_proc_root)
5458 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5459 kset_unregister(ext4_kset);
d44651d0 5460out5:
5dabfc78 5461 ext4_exit_system_zone();
d44651d0 5462out6:
5dabfc78 5463 ext4_exit_pageio();
51865fda
ZL		 5464out7:
5465 ext4_exit_es();
5466
ac27a0ec
DK		 5467 return err;
5468}
5469
5dabfc78 5470static void __exit ext4_exit_fs(void)
ac27a0ec 5471{
bfff6873 5472 ext4_destroy_lazyinit_thread();
24b58424
TT		 5473 unregister_as_ext2();
5474 unregister_as_ext3();
03010a33 5475 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5476 destroy_inodecache();
5dabfc78
TT		 5477 ext4_exit_xattr();
5478 ext4_exit_mballoc();
8f021222 5479 ext4_exit_feat_adverts();
9f6200bb 5480 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5481 kset_unregister(ext4_kset);
5dabfc78
TT		 5482 ext4_exit_system_zone();
5483 ext4_exit_pageio();
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)