]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: silence warning in ext4_writepages()
[thirdparty/kernel/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
4a092d73 48#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
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]);
1705
1706 if (test_opt(sb, USRQUOTA))
1707 seq_puts(seq, ",usrquota");
1708
1709 if (test_opt(sb, GRPQUOTA))
1710 seq_puts(seq, ",grpquota");
1711#endif
1712}
1713
5a916be1
TT		 1714static const char *token2str(int token)
1715{
50df9fd5 1716 const struct match_token *t;
5a916be1
TT		 1717
1718 for (t = tokens; t->token != Opt_err; t++)
1719 if (t->token == token && !strchr(t->pattern, '='))
1720 break;
1721 return t->pattern;
1722}
1723
2adf6da8
TT		 1724/*
1725 * Show an option if
1726 * - it's set to a non-default value OR
1727 * - if the per-sb default is different from the global default
1728 */
66acdcf4
TT		 1729static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1730 int nodefs)
2adf6da8 1731{
2adf6da8
TT		 1732 struct ext4_sb_info *sbi = EXT4_SB(sb);
1733 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1734 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1735 const struct mount_opts *m;
66acdcf4 1736 char sep = nodefs ? '\n' : ',';
2adf6da8 1737
66acdcf4
TT		 1738#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1739#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1740
1741 if (sbi->s_sb_block != 1)
5a916be1
TT		 1742 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1743
1744 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1745 int want_set = m->flags & MOPT_SET;
1746 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1747 (m->flags & MOPT_CLEAR_ERR))
1748 continue;
66acdcf4 1749 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1750 continue; /* skip if same as the default */
1751 if ((want_set &&
1752 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1753 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1754 continue; /* select Opt_noFoo vs Opt_Foo */
1755 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1756 }
5a916be1 1757
08cefc7a 1758 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1759 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1760 SEQ_OPTS_PRINT("resuid=%u",
1761 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1762 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1763 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1764 SEQ_OPTS_PRINT("resgid=%u",
1765 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1766 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1767 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1768 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1769 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1770 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1771 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1772 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1773 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1774 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1775 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1776 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1777 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1778 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1779 if (sb->s_flags & MS_I_VERSION)
5a916be1 1780 SEQ_OPTS_PUTS("i_version");
66acdcf4 1781 if (nodefs || sbi->s_stripe)
5a916be1 1782 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1783 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1784 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1785 SEQ_OPTS_PUTS("data=journal");
1786 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1787 SEQ_OPTS_PUTS("data=ordered");
1788 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1789 SEQ_OPTS_PUTS("data=writeback");
1790 }
66acdcf4
TT		 1791 if (nodefs ||
1792 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1793 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1794 sbi->s_inode_readahead_blks);
2adf6da8 1795
66acdcf4
TT		 1796 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1797 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1798 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1799 if (nodefs || sbi->s_max_dir_size_kb)
1800 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1801
1802 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1803 return 0;
1804}
1805
66acdcf4
TT		 1806static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1807{
1808 return _ext4_show_options(seq, root->d_sb, 0);
1809}
1810
1811static int options_seq_show(struct seq_file *seq, void *offset)
1812{
1813 struct super_block *sb = seq->private;
1814 int rc;
1815
1816 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1817 rc = _ext4_show_options(seq, sb, 1);
1818 seq_puts(seq, "\n");
1819 return rc;
1820}
1821
1822static int options_open_fs(struct inode *inode, struct file *file)
1823{
d9dda78b 1824 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1825}
1826
1827static const struct file_operations ext4_seq_options_fops = {
1828 .owner = THIS_MODULE,
1829 .open = options_open_fs,
1830 .read = seq_read,
1831 .llseek = seq_lseek,
1832 .release = single_release,
1833};
1834
617ba13b 1835static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1836 int read_only)
1837{
617ba13b 1838 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1839 int res = 0;
1840
617ba13b 1841 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1842 ext4_msg(sb, KERN_ERR, "revision level too high, "
1843 "forcing read-only mode");
ac27a0ec
DK		 1844 res = MS_RDONLY;
1845 }
1846 if (read_only)
281b5995 1847 goto done;
617ba13b 1848 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1849 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1850 "running e2fsck is recommended");
617ba13b 1851 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1852 ext4_msg(sb, KERN_WARNING,
1853 "warning: mounting fs with errors, "
1854 "running e2fsck is recommended");
ed3ce80a 1855 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1856 le16_to_cpu(es->s_mnt_count) >=
1857 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1858 ext4_msg(sb, KERN_WARNING,
1859 "warning: maximal mount count reached, "
1860 "running e2fsck is recommended");
ac27a0ec
DK		 1861 else if (le32_to_cpu(es->s_checkinterval) &&
1862 (le32_to_cpu(es->s_lastcheck) +
1863 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1864 ext4_msg(sb, KERN_WARNING,
1865 "warning: checktime reached, "
1866 "running e2fsck is recommended");
0b8e58a1 1867 if (!sbi->s_journal)
0390131b 1868 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1869 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1870 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1871 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1872 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1873 ext4_update_dynamic_rev(sb);
0390131b
FM		 1874 if (sbi->s_journal)
1875 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1876
e2d67052 1877 ext4_commit_super(sb, 1);
281b5995 1878done:
ac27a0ec 1879 if (test_opt(sb, DEBUG))
a9df9a49 1880 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1881 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1882 sb->s_blocksize,
1883 sbi->s_groups_count,
617ba13b
MC		 1884 EXT4_BLOCKS_PER_GROUP(sb),
1885 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1886 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1887
7abc52c2 1888 cleancache_init_fs(sb);
ac27a0ec
DK		 1889 return res;
1890}
1891
117fff10
TT		 1892int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1893{
1894 struct ext4_sb_info *sbi = EXT4_SB(sb);
1895 struct flex_groups *new_groups;
1896 int size;
1897
1898 if (!sbi->s_log_groups_per_flex)
1899 return 0;
1900
1901 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1902 if (size <= sbi->s_flex_groups_allocated)
1903 return 0;
1904
1905 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1906 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1907 if (!new_groups) {
1908 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1909 size / (int) sizeof(struct flex_groups));
1910 return -ENOMEM;
1911 }
1912
1913 if (sbi->s_flex_groups) {
1914 memcpy(new_groups, sbi->s_flex_groups,
1915 (sbi->s_flex_groups_allocated *
1916 sizeof(struct flex_groups)));
1917 ext4_kvfree(sbi->s_flex_groups);
1918 }
1919 sbi->s_flex_groups = new_groups;
1920 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1921 return 0;
1922}
1923
772cb7c8
JS		 1924static int ext4_fill_flex_info(struct super_block *sb)
1925{
1926 struct ext4_sb_info *sbi = EXT4_SB(sb);
1927 struct ext4_group_desc *gdp = NULL;
772cb7c8 1928 ext4_group_t flex_group;
117fff10 1929 int i, err;
772cb7c8 1930
503358ae 1931 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1932 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1933 sbi->s_log_groups_per_flex = 0;
1934 return 1;
1935 }
1936
117fff10
TT		 1937 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1938 if (err)
9933fc0a 1939 goto failed;
772cb7c8 1940
772cb7c8 1941 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1942 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1943
1944 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1945 atomic_add(ext4_free_inodes_count(sb, gdp),
1946 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 1947 atomic64_add(ext4_free_group_clusters(sb, gdp),
1948 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1949 atomic_add(ext4_used_dirs_count(sb, gdp),
1950 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1951 }
1952
1953 return 1;
1954failed:
1955 return 0;
1956}
1957
feb0ab32
DW		 1958static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1959 struct ext4_group_desc *gdp)
717d50e4 1960{
feb0ab32 1961 int offset;
717d50e4 1962 __u16 crc = 0;
feb0ab32 1963 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1964
feb0ab32
DW		 1965 if ((sbi->s_es->s_feature_ro_compat &
1966 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1967 /* Use new metadata_csum algorithm */
d6a77105 1968 __le16 save_csum;
feb0ab32
DW		 1969 __u32 csum32;
1970
d6a77105 1971 save_csum = gdp->bg_checksum;
feb0ab32
DW		 1972 gdp->bg_checksum = 0;
1973 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1974 sizeof(le_group));
1975 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1976 sbi->s_desc_size);
d6a77105 1977 gdp->bg_checksum = save_csum;
feb0ab32
DW		 1978
1979 crc = csum32 & 0xFFFF;
1980 goto out;
717d50e4
AD		 1981 }
1982
feb0ab32
DW		 1983 /* old crc16 code */
1984 offset = offsetof(struct ext4_group_desc, bg_checksum);
1985
1986 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1987 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1988 crc = crc16(crc, (__u8 *)gdp, offset);
1989 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1990 /* for checksum of struct ext4_group_desc do the rest...*/
1991 if ((sbi->s_es->s_feature_incompat &
1992 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1993 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1994 crc = crc16(crc, (__u8 *)gdp + offset,
1995 le16_to_cpu(sbi->s_es->s_desc_size) -
1996 offset);
1997
1998out:
717d50e4
AD		 1999 return cpu_to_le16(crc);
2000}
2001
feb0ab32 2002int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2003 struct ext4_group_desc *gdp)
2004{
feb0ab32
DW		 2005 if (ext4_has_group_desc_csum(sb) &&
2006 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2007 block_group, gdp)))
717d50e4
AD		 2008 return 0;
2009
2010 return 1;
2011}
2012
feb0ab32
DW		 2013void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2014 struct ext4_group_desc *gdp)
2015{
2016 if (!ext4_has_group_desc_csum(sb))
2017 return;
2018 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2019}
2020
ac27a0ec 2021/* Called at mount-time, super-block is locked */
bfff6873
LC		 2022static int ext4_check_descriptors(struct super_block *sb,
2023 ext4_group_t *first_not_zeroed)
ac27a0ec 2024{
617ba13b
MC		 2025 struct ext4_sb_info *sbi = EXT4_SB(sb);
2026 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2027 ext4_fsblk_t last_block;
bd81d8ee
LV		 2028 ext4_fsblk_t block_bitmap;
2029 ext4_fsblk_t inode_bitmap;
2030 ext4_fsblk_t inode_table;
ce421581 2031 int flexbg_flag = 0;
bfff6873 2032 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2033
ce421581
JS		 2034 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2035 flexbg_flag = 1;
2036
af5bc92d 2037 ext4_debug("Checking group descriptors");
ac27a0ec 2038
197cd65a
AM		 2039 for (i = 0; i < sbi->s_groups_count; i++) {
2040 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2041
ce421581 2042 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2043 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2044 else
2045 last_block = first_block +
617ba13b 2046 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2047
bfff6873
LC		 2048 if ((grp == sbi->s_groups_count) &&
2049 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2050 grp = i;
2051
8fadc143 2052 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2053 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2054 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2055 "Block bitmap for group %u not in group "
b31e1552 2056 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2057 return 0;
2058 }
8fadc143 2059 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2060 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2061 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2062 "Inode bitmap for group %u not in group "
b31e1552 2063 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2064 return 0;
2065 }
8fadc143 2066 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2067 if (inode_table < first_block ||
2b2d6d01 2068 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2069 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2070 "Inode table for group %u not in group "
b31e1552 2071 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2072 return 0;
2073 }
955ce5f5 2074 ext4_lock_group(sb, i);
feb0ab32 2075 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2076 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2077 "Checksum for group %u failed (%u!=%u)",
2078 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2079 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2080 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2081 ext4_unlock_group(sb, i);
8a266467 2082 return 0;
7ee1ec4c 2083 }
717d50e4 2084 }
955ce5f5 2085 ext4_unlock_group(sb, i);
ce421581
JS		 2086 if (!flexbg_flag)
2087 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2088 }
bfff6873
LC		 2089 if (NULL != first_not_zeroed)
2090 *first_not_zeroed = grp;
ac27a0ec 2091
5dee5437
TT		 2092 ext4_free_blocks_count_set(sbi->s_es,
2093 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2094 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2095 return 1;
2096}
2097
617ba13b 2098/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2099 * the superblock) which were deleted from all directories, but held open by
2100 * a process at the time of a crash. We walk the list and try to delete these
2101 * inodes at recovery time (only with a read-write filesystem).
2102 *
2103 * In order to keep the orphan inode chain consistent during traversal (in
2104 * case of crash during recovery), we link each inode into the superblock
2105 * orphan list_head and handle it the same way as an inode deletion during
2106 * normal operation (which journals the operations for us).
2107 *
2108 * We only do an iget() and an iput() on each inode, which is very safe if we
2109 * accidentally point at an in-use or already deleted inode. The worst that
2110 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2111 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2112 * e2fsck was run on this filesystem, and it must have already done the orphan
2113 * inode cleanup for us, so we can safely abort without any further action.
2114 */
2b2d6d01
TT		 2115static void ext4_orphan_cleanup(struct super_block *sb,
2116 struct ext4_super_block *es)
ac27a0ec
DK		 2117{
2118 unsigned int s_flags = sb->s_flags;
2119 int nr_orphans = 0, nr_truncates = 0;
2120#ifdef CONFIG_QUOTA
2121 int i;
2122#endif
2123 if (!es->s_last_orphan) {
2124 jbd_debug(4, "no orphan inodes to clean up\n");
2125 return;
2126 }
2127
a8f48a95 2128 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2129 ext4_msg(sb, KERN_ERR, "write access "
2130 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2131 return;
2132 }
2133
d39195c3
AG		 2134 /* Check if feature set would not allow a r/w mount */
2135 if (!ext4_feature_set_ok(sb, 0)) {
2136 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2137 "unknown ROCOMPAT features");
2138 return;
2139 }
2140
617ba13b 2141 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2142 /* don't clear list on RO mount w/ errors */
2143 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2144 jbd_debug(1, "Errors on filesystem, "
2145 "clearing orphan list.\n");
c25f9bc6
ES		 2146 es->s_last_orphan = 0;
2147 }
ac27a0ec
DK		 2148 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2149 return;
2150 }
2151
2152 if (s_flags & MS_RDONLY) {
b31e1552 2153 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2154 sb->s_flags &= ~MS_RDONLY;
2155 }
2156#ifdef CONFIG_QUOTA
2157 /* Needed for iput() to work correctly and not trash data */
2158 sb->s_flags |= MS_ACTIVE;
2159 /* Turn on quotas so that they are updated correctly */
2160 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2161 if (EXT4_SB(sb)->s_qf_names[i]) {
2162 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2163 if (ret < 0)
b31e1552
ES		 2164 ext4_msg(sb, KERN_ERR,
2165 "Cannot turn on journaled "
2166 "quota: error %d", ret);
ac27a0ec
DK		 2167 }
2168 }
2169#endif
2170
2171 while (es->s_last_orphan) {
2172 struct inode *inode;
2173
97bd42b9
JB		 2174 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2175 if (IS_ERR(inode)) {
ac27a0ec
DK		 2176 es->s_last_orphan = 0;
2177 break;
2178 }
2179
617ba13b 2180 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2181 dquot_initialize(inode);
ac27a0ec 2182 if (inode->i_nlink) {
566370a2
PT		 2183 if (test_opt(sb, DEBUG))
2184 ext4_msg(sb, KERN_DEBUG,
2185 "%s: truncating inode %lu to %lld bytes",
2186 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2187 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2188 inode->i_ino, inode->i_size);
721e3eba 2189 mutex_lock(&inode->i_mutex);
55f252c9 2190 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2191 ext4_truncate(inode);
721e3eba 2192 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2193 nr_truncates++;
2194 } else {
566370a2
PT		 2195 if (test_opt(sb, DEBUG))
2196 ext4_msg(sb, KERN_DEBUG,
2197 "%s: deleting unreferenced inode %lu",
2198 __func__, inode->i_ino);
ac27a0ec
DK		 2199 jbd_debug(2, "deleting unreferenced inode %lu\n",
2200 inode->i_ino);
2201 nr_orphans++;
2202 }
2203 iput(inode); /* The delete magic happens here! */
2204 }
2205
2b2d6d01 2206#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2207
2208 if (nr_orphans)
b31e1552
ES		 2209 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2210 PLURAL(nr_orphans));
ac27a0ec 2211 if (nr_truncates)
b31e1552
ES		 2212 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2213 PLURAL(nr_truncates));
ac27a0ec
DK		 2214#ifdef CONFIG_QUOTA
2215 /* Turn quotas off */
2216 for (i = 0; i < MAXQUOTAS; i++) {
2217 if (sb_dqopt(sb)->files[i])
287a8095 2218 dquot_quota_off(sb, i);
ac27a0ec
DK		 2219 }
2220#endif
2221 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2222}
0b8e58a1 2223
cd2291a4
ES		 2224/*
2225 * Maximal extent format file size.
2226 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2227 * extent format containers, within a sector_t, and within i_blocks
2228 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2229 * so that won't be a limiting factor.
2230 *
f17722f9
LC		 2231 * However there is other limiting factor. We do store extents in the form
2232 * of starting block and length, hence the resulting length of the extent
2233 * covering maximum file size must fit into on-disk format containers as
2234 * well. Given that length is always by 1 unit bigger than max unit (because
2235 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2236 *
cd2291a4
ES		 2237 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2238 */
f287a1a5 2239static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2240{
2241 loff_t res;
2242 loff_t upper_limit = MAX_LFS_FILESIZE;
2243
2244 /* small i_blocks in vfs inode? */
f287a1a5 2245 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2246 /*
90c699a9 2247 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2248 * i_block represent total blocks in 512 bytes
2249 * 32 == size of vfs inode i_blocks * 8
2250 */
2251 upper_limit = (1LL << 32) - 1;
2252
2253 /* total blocks in file system block size */
2254 upper_limit >>= (blkbits - 9);
2255 upper_limit <<= blkbits;
2256 }
2257
f17722f9
LC		 2258 /*
2259 * 32-bit extent-start container, ee_block. We lower the maxbytes
2260 * by one fs block, so ee_len can cover the extent of maximum file
2261 * size
2262 */
2263 res = (1LL << 32) - 1;
cd2291a4 2264 res <<= blkbits;
cd2291a4
ES		 2265
2266 /* Sanity check against vm- & vfs- imposed limits */
2267 if (res > upper_limit)
2268 res = upper_limit;
2269
2270 return res;
2271}
ac27a0ec 2272
ac27a0ec 2273/*
cd2291a4 2274 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2275 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2276 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2277 */
f287a1a5 2278static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2279{
617ba13b 2280 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2281 int meta_blocks;
2282 loff_t upper_limit;
0b8e58a1
AD		 2283 /* This is calculated to be the largest file size for a dense, block
2284 * mapped file such that the file's total number of 512-byte sectors,
2285 * including data and all indirect blocks, does not exceed (2^48 - 1).
2286 *
2287 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2288 * number of 512-byte sectors of the file.
0fc1b451
AK		 2289 */
2290
f287a1a5 2291 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2292 /*
90c699a9 2293 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2294 * the inode i_block field represents total file blocks in
2295 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2296 */
2297 upper_limit = (1LL << 32) - 1;
2298
2299 /* total blocks in file system block size */
2300 upper_limit >>= (bits - 9);
2301
2302 } else {
8180a562
AK		 2303 /*
2304 * We use 48 bit ext4_inode i_blocks
2305 * With EXT4_HUGE_FILE_FL set the i_blocks
2306 * represent total number of blocks in
2307 * file system block size
2308 */
0fc1b451
AK		 2309 upper_limit = (1LL << 48) - 1;
2310
0fc1b451
AK		 2311 }
2312
2313 /* indirect blocks */
2314 meta_blocks = 1;
2315 /* double indirect blocks */
2316 meta_blocks += 1 + (1LL << (bits-2));
2317 /* tripple indirect blocks */
2318 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2319
2320 upper_limit -= meta_blocks;
2321 upper_limit <<= bits;
ac27a0ec
DK		 2322
2323 res += 1LL << (bits-2);
2324 res += 1LL << (2*(bits-2));
2325 res += 1LL << (3*(bits-2));
2326 res <<= bits;
2327 if (res > upper_limit)
2328 res = upper_limit;
0fc1b451
AK		 2329
2330 if (res > MAX_LFS_FILESIZE)
2331 res = MAX_LFS_FILESIZE;
2332
ac27a0ec
DK		 2333 return res;
2334}
2335
617ba13b 2336static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2337 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2338{
617ba13b 2339 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2340 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2341 int has_super = 0;
2342
2343 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2344
617ba13b 2345 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2346 nr < first_meta_bg)
70bbb3e0 2347 return logical_sb_block + nr + 1;
ac27a0ec 2348 bg = sbi->s_desc_per_block * nr;
617ba13b 2349 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2350 has_super = 1;
0b8e58a1 2351
617ba13b 2352 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2353}
2354
c9de560d
AT		 2355/**
2356 * ext4_get_stripe_size: Get the stripe size.
2357 * @sbi: In memory super block info
2358 *
2359 * If we have specified it via mount option, then
2360 * use the mount option value. If the value specified at mount time is
2361 * greater than the blocks per group use the super block value.
2362 * If the super block value is greater than blocks per group return 0.
2363 * Allocator needs it be less than blocks per group.
2364 *
2365 */
2366static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2367{
2368 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2369 unsigned long stripe_width =
2370 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2371 int ret;
c9de560d
AT		 2372
2373 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2374 ret = sbi->s_stripe;
2375 else if (stripe_width <= sbi->s_blocks_per_group)
2376 ret = stripe_width;
2377 else if (stride <= sbi->s_blocks_per_group)
2378 ret = stride;
2379 else
2380 ret = 0;
c9de560d 2381
3eb08658
DE		 2382 /*
2383 * If the stripe width is 1, this makes no sense and
2384 * we set it to 0 to turn off stripe handling code.
2385 */
2386 if (ret <= 1)
2387 ret = 0;
c9de560d 2388
3eb08658 2389 return ret;
c9de560d 2390}
ac27a0ec 2391
3197ebdb
TT		 2392/* sysfs supprt */
2393
2394struct ext4_attr {
2395 struct attribute attr;
2396 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2397 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb 2398 const char *, size_t);
f2d50a65
JK		 2399 union {
2400 int offset;
2401 int deprecated_val;
2402 } u;
3197ebdb
TT		 2403};
2404
27dd4385
LC		 2405static int parse_strtoull(const char *buf,
2406 unsigned long long max, unsigned long long *value)
2407{
2408 int ret;
2409
2410 ret = kstrtoull(skip_spaces(buf), 0, value);
2411 if (!ret && *value > max)
2412 ret = -EINVAL;
2413 return ret;
2414}
2415
3197ebdb
TT		 2416static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2417 struct ext4_sb_info *sbi,
2418 char *buf)
2419{
2420 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2421 (s64) EXT4_C2B(sbi,
2422 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2423}
2424
2425static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2426 struct ext4_sb_info *sbi, char *buf)
2427{
2428 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2429
f613dfcb
TT		 2430 if (!sb->s_bdev->bd_part)
2431 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2432 return snprintf(buf, PAGE_SIZE, "%lu\n",
2433 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2434 sbi->s_sectors_written_start) >> 1);
2435}
2436
2437static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2438 struct ext4_sb_info *sbi, char *buf)
2439{
2440 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2441
f613dfcb
TT		 2442 if (!sb->s_bdev->bd_part)
2443 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2444 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2445 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2446 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2447 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2448}
2449
2450static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2451 struct ext4_sb_info *sbi,
2452 const char *buf, size_t count)
2453{
2454 unsigned long t;
e1091b15 2455 int ret;
3197ebdb 2456
e1091b15
LC		 2457 ret = kstrtoul(skip_spaces(buf), 0, &t);
2458 if (ret)
2459 return ret;
3197ebdb 2460
e1091b15 2461 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2462 return -EINVAL;
2463
2464 sbi->s_inode_readahead_blks = t;
2465 return count;
2466}
2467
2468static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2469 struct ext4_sb_info *sbi, char *buf)
3197ebdb 2470{
f2d50a65 2471 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb
TT		 2472
2473 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2474}
2475
2476static ssize_t sbi_ui_store(struct ext4_attr *a,
2477 struct ext4_sb_info *sbi,
2478 const char *buf, size_t count)
2479{
f2d50a65 2480 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb 2481 unsigned long t;
e1091b15 2482 int ret;
3197ebdb 2483
e1091b15
LC		 2484 ret = kstrtoul(skip_spaces(buf), 0, &t);
2485 if (ret)
2486 return ret;
3197ebdb
TT		 2487 *ui = t;
2488 return count;
2489}
2490
27dd4385
LC		 2491static ssize_t reserved_clusters_show(struct ext4_attr *a,
2492 struct ext4_sb_info *sbi, char *buf)
2493{
2494 return snprintf(buf, PAGE_SIZE, "%llu\n",
2495 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2496}
2497
2498static ssize_t reserved_clusters_store(struct ext4_attr *a,
2499 struct ext4_sb_info *sbi,
2500 const char *buf, size_t count)
2501{
2502 unsigned long long val;
2503 int ret;
2504
2505 if (parse_strtoull(buf, -1ULL, &val))
2506 return -EINVAL;
2507 ret = ext4_reserve_clusters(sbi, val);
2508
2509 return ret ? ret : count;
2510}
2511
2c0544b2
TT		 2512static ssize_t trigger_test_error(struct ext4_attr *a,
2513 struct ext4_sb_info *sbi,
2514 const char *buf, size_t count)
2515{
2516 int len = count;
2517
2518 if (!capable(CAP_SYS_ADMIN))
2519 return -EPERM;
2520
2521 if (len && buf[len-1] == '\n')
2522 len--;
2523
2524 if (len)
2525 ext4_error(sbi->s_sb, "%.*s", len, buf);
2526 return count;
2527}
2528
f2d50a65
JK		 2529static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2530 struct ext4_sb_info *sbi, char *buf)
2531{
2532 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2533}
2534
3197ebdb
TT		 2535#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2536static struct ext4_attr ext4_attr_##_name = { \
2537 .attr = {.name = __stringify(_name), .mode = _mode }, \
2538 .show = _show, \
2539 .store = _store, \
f2d50a65
JK		 2540 .u = { \
2541 .offset = offsetof(struct ext4_sb_info, _elname),\
2542 }, \
3197ebdb
TT		 2543}
2544#define EXT4_ATTR(name, mode, show, store) \
2545static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2546
857ac889 2547#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2548#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2549#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2550#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2551 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2552#define ATTR_LIST(name) &ext4_attr_##name.attr
f2d50a65
JK		 2553#define EXT4_DEPRECATED_ATTR(_name, _val) \
2554static struct ext4_attr ext4_attr_##_name = { \
2555 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2556 .show = sbi_deprecated_show, \
2557 .u = { \
2558 .deprecated_val = _val, \
2559 }, \
2560}
3197ebdb
TT		 2561
2562EXT4_RO_ATTR(delayed_allocation_blocks);
2563EXT4_RO_ATTR(session_write_kbytes);
2564EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2565EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2566EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2567 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2568EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2569EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2570EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2571EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2572EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2573EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2574EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
f2d50a65 2575EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
67a5da56 2576EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2577EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2578
2579static struct attribute *ext4_attrs[] = {
2580 ATTR_LIST(delayed_allocation_blocks),
2581 ATTR_LIST(session_write_kbytes),
2582 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2583 ATTR_LIST(reserved_clusters),
3197ebdb 2584 ATTR_LIST(inode_readahead_blks),
11013911 2585 ATTR_LIST(inode_goal),
3197ebdb
TT		 2586 ATTR_LIST(mb_stats),
2587 ATTR_LIST(mb_max_to_scan),
2588 ATTR_LIST(mb_min_to_scan),
2589 ATTR_LIST(mb_order2_req),
2590 ATTR_LIST(mb_stream_req),
2591 ATTR_LIST(mb_group_prealloc),
55138e0b 2592 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2593 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2594 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2595 NULL,
2596};
2597
857ac889
LC		 2598/* Features this copy of ext4 supports */
2599EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2600EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2601EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2602
2603static struct attribute *ext4_feat_attrs[] = {
2604 ATTR_LIST(lazy_itable_init),
27ee40df 2605 ATTR_LIST(batched_discard),
5e7bbef1 2606 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2607 NULL,
2608};
2609
3197ebdb
TT		 2610static ssize_t ext4_attr_show(struct kobject *kobj,
2611 struct attribute *attr, char *buf)
2612{
2613 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2614 s_kobj);
2615 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2616
2617 return a->show ? a->show(a, sbi, buf) : 0;
2618}
2619
2620static ssize_t ext4_attr_store(struct kobject *kobj,
2621 struct attribute *attr,
2622 const char *buf, size_t len)
2623{
2624 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2625 s_kobj);
2626 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2627
2628 return a->store ? a->store(a, sbi, buf, len) : 0;
2629}
2630
2631static void ext4_sb_release(struct kobject *kobj)
2632{
2633 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2634 s_kobj);
2635 complete(&sbi->s_kobj_unregister);
2636}
2637
52cf25d0 2638static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2639 .show = ext4_attr_show,
2640 .store = ext4_attr_store,
2641};
2642
2643static struct kobj_type ext4_ktype = {
2644 .default_attrs = ext4_attrs,
2645 .sysfs_ops = &ext4_attr_ops,
2646 .release = ext4_sb_release,
2647};
2648
857ac889
LC		 2649static void ext4_feat_release(struct kobject *kobj)
2650{
2651 complete(&ext4_feat->f_kobj_unregister);
2652}
2653
2654static struct kobj_type ext4_feat_ktype = {
2655 .default_attrs = ext4_feat_attrs,
2656 .sysfs_ops = &ext4_attr_ops,
2657 .release = ext4_feat_release,
2658};
2659
a13fb1a4
ES		 2660/*
2661 * Check whether this filesystem can be mounted based on
2662 * the features present and the RDONLY/RDWR mount requested.
2663 * Returns 1 if this filesystem can be mounted as requested,
2664 * 0 if it cannot be.
2665 */
2666static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2667{
2668 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2669 ext4_msg(sb, KERN_ERR,
2670 "Couldn't mount because of "
2671 "unsupported optional features (%x)",
2672 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2673 ~EXT4_FEATURE_INCOMPAT_SUPP));
2674 return 0;
2675 }
2676
2677 if (readonly)
2678 return 1;
2679
2680 /* Check that feature set is OK for a read-write mount */
2681 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2682 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2683 "unsupported optional features (%x)",
2684 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2685 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2686 return 0;
2687 }
2688 /*
2689 * Large file size enabled file system can only be mounted
2690 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2691 */
2692 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2693 if (sizeof(blkcnt_t) < sizeof(u64)) {
2694 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2695 "cannot be mounted RDWR without "
2696 "CONFIG_LBDAF");
2697 return 0;
2698 }
2699 }
bab08ab9
TT		 2700 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2701 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2702 ext4_msg(sb, KERN_ERR,
2703 "Can't support bigalloc feature without "
2704 "extents feature\n");
2705 return 0;
2706 }
7c319d32
AK		 2707
2708#ifndef CONFIG_QUOTA
2709 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2710 !readonly) {
2711 ext4_msg(sb, KERN_ERR,
2712 "Filesystem with quota feature cannot be mounted RDWR "
2713 "without CONFIG_QUOTA");
2714 return 0;
2715 }
2716#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2717 return 1;
2718}
2719
66e61a9e
TT		 2720/*
2721 * This function is called once a day if we have errors logged
2722 * on the file system
2723 */
2724static void print_daily_error_info(unsigned long arg)
2725{
2726 struct super_block *sb = (struct super_block *) arg;
2727 struct ext4_sb_info *sbi;
2728 struct ext4_super_block *es;
2729
2730 sbi = EXT4_SB(sb);
2731 es = sbi->s_es;
2732
2733 if (es->s_error_count)
2734 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2735 le32_to_cpu(es->s_error_count));
2736 if (es->s_first_error_time) {
2737 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2738 sb->s_id, le32_to_cpu(es->s_first_error_time),
2739 (int) sizeof(es->s_first_error_func),
2740 es->s_first_error_func,
2741 le32_to_cpu(es->s_first_error_line));
2742 if (es->s_first_error_ino)
2743 printk(": inode %u",
2744 le32_to_cpu(es->s_first_error_ino));
2745 if (es->s_first_error_block)
2746 printk(": block %llu", (unsigned long long)
2747 le64_to_cpu(es->s_first_error_block));
2748 printk("\n");
2749 }
2750 if (es->s_last_error_time) {
2751 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2752 sb->s_id, le32_to_cpu(es->s_last_error_time),
2753 (int) sizeof(es->s_last_error_func),
2754 es->s_last_error_func,
2755 le32_to_cpu(es->s_last_error_line));
2756 if (es->s_last_error_ino)
2757 printk(": inode %u",
2758 le32_to_cpu(es->s_last_error_ino));
2759 if (es->s_last_error_block)
2760 printk(": block %llu", (unsigned long long)
2761 le64_to_cpu(es->s_last_error_block));
2762 printk("\n");
2763 }
2764 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2765}
2766
bfff6873
LC		 2767/* Find next suitable group and run ext4_init_inode_table */
2768static int ext4_run_li_request(struct ext4_li_request *elr)
2769{
2770 struct ext4_group_desc *gdp = NULL;
2771 ext4_group_t group, ngroups;
2772 struct super_block *sb;
2773 unsigned long timeout = 0;
2774 int ret = 0;
2775
2776 sb = elr->lr_super;
2777 ngroups = EXT4_SB(sb)->s_groups_count;
2778
8e8ad8a5 2779 sb_start_write(sb);
bfff6873
LC		 2780 for (group = elr->lr_next_group; group < ngroups; group++) {
2781 gdp = ext4_get_group_desc(sb, group, NULL);
2782 if (!gdp) {
2783 ret = 1;
2784 break;
2785 }
2786
2787 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2788 break;
2789 }
2790
7f511862 2791 if (group >= ngroups)
bfff6873
LC		 2792 ret = 1;
2793
2794 if (!ret) {
2795 timeout = jiffies;
2796 ret = ext4_init_inode_table(sb, group,
2797 elr->lr_timeout ? 0 : 1);
2798 if (elr->lr_timeout == 0) {
51ce6511
LC		 2799 timeout = (jiffies - timeout) *
2800 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2801 elr->lr_timeout = timeout;
2802 }
2803 elr->lr_next_sched = jiffies + elr->lr_timeout;
2804 elr->lr_next_group = group + 1;
2805 }
8e8ad8a5 2806 sb_end_write(sb);
bfff6873
LC		 2807
2808 return ret;
2809}
2810
2811/*
2812 * Remove lr_request from the list_request and free the
4ed5c033 2813 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2814 */
2815static void ext4_remove_li_request(struct ext4_li_request *elr)
2816{
2817 struct ext4_sb_info *sbi;
2818
2819 if (!elr)
2820 return;
2821
2822 sbi = elr->lr_sbi;
2823
2824 list_del(&elr->lr_request);
2825 sbi->s_li_request = NULL;
2826 kfree(elr);
2827}
2828
2829static void ext4_unregister_li_request(struct super_block *sb)
2830{
1bb933fb
LC		 2831 mutex_lock(&ext4_li_mtx);
2832 if (!ext4_li_info) {
2833 mutex_unlock(&ext4_li_mtx);
bfff6873 2834 return;
1bb933fb 2835 }
bfff6873
LC		 2836
2837 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2838 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2839 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2840 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2841}
2842
8f1f7453
ES		 2843static struct task_struct *ext4_lazyinit_task;
2844
bfff6873
LC		 2845/*
2846 * This is the function where ext4lazyinit thread lives. It walks
2847 * through the request list searching for next scheduled filesystem.
2848 * When such a fs is found, run the lazy initialization request
2849 * (ext4_rn_li_request) and keep track of the time spend in this
2850 * function. Based on that time we compute next schedule time of
2851 * the request. When walking through the list is complete, compute
2852 * next waking time and put itself into sleep.
2853 */
2854static int ext4_lazyinit_thread(void *arg)
2855{
2856 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2857 struct list_head *pos, *n;
2858 struct ext4_li_request *elr;
4ed5c033 2859 unsigned long next_wakeup, cur;
bfff6873
LC		 2860
2861 BUG_ON(NULL == eli);
2862
bfff6873
LC		 2863cont_thread:
2864 while (true) {
2865 next_wakeup = MAX_JIFFY_OFFSET;
2866
2867 mutex_lock(&eli->li_list_mtx);
2868 if (list_empty(&eli->li_request_list)) {
2869 mutex_unlock(&eli->li_list_mtx);
2870 goto exit_thread;
2871 }
2872
2873 list_for_each_safe(pos, n, &eli->li_request_list) {
2874 elr = list_entry(pos, struct ext4_li_request,
2875 lr_request);
2876
b2c78cd0
TT		 2877 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2878 if (ext4_run_li_request(elr) != 0) {
2879 /* error, remove the lazy_init job */
2880 ext4_remove_li_request(elr);
2881 continue;
2882 }
bfff6873
LC		 2883 }
2884
2885 if (time_before(elr->lr_next_sched, next_wakeup))
2886 next_wakeup = elr->lr_next_sched;
2887 }
2888 mutex_unlock(&eli->li_list_mtx);
2889
a0acae0e 2890 try_to_freeze();
bfff6873 2891
4ed5c033
LC		 2892 cur = jiffies;
2893 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2894 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2895 cond_resched();
2896 continue;
2897 }
2898
4ed5c033
LC		 2899 schedule_timeout_interruptible(next_wakeup - cur);
2900
8f1f7453
ES		 2901 if (kthread_should_stop()) {
2902 ext4_clear_request_list();
2903 goto exit_thread;
2904 }
bfff6873
LC		 2905 }
2906
2907exit_thread:
2908 /*
2909 * It looks like the request list is empty, but we need
2910 * to check it under the li_list_mtx lock, to prevent any
2911 * additions into it, and of course we should lock ext4_li_mtx
2912 * to atomically free the list and ext4_li_info, because at
2913 * this point another ext4 filesystem could be registering
2914 * new one.
2915 */
2916 mutex_lock(&ext4_li_mtx);
2917 mutex_lock(&eli->li_list_mtx);
2918 if (!list_empty(&eli->li_request_list)) {
2919 mutex_unlock(&eli->li_list_mtx);
2920 mutex_unlock(&ext4_li_mtx);
2921 goto cont_thread;
2922 }
2923 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2924 kfree(ext4_li_info);
2925 ext4_li_info = NULL;
2926 mutex_unlock(&ext4_li_mtx);
2927
2928 return 0;
2929}
2930
2931static void ext4_clear_request_list(void)
2932{
2933 struct list_head *pos, *n;
2934 struct ext4_li_request *elr;
2935
2936 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2937 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2938 elr = list_entry(pos, struct ext4_li_request,
2939 lr_request);
2940 ext4_remove_li_request(elr);
2941 }
2942 mutex_unlock(&ext4_li_info->li_list_mtx);
2943}
2944
2945static int ext4_run_lazyinit_thread(void)
2946{
8f1f7453
ES		 2947 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2948 ext4_li_info, "ext4lazyinit");
2949 if (IS_ERR(ext4_lazyinit_task)) {
2950 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2951 ext4_clear_request_list();
bfff6873
LC		 2952 kfree(ext4_li_info);
2953 ext4_li_info = NULL;
92b97816 2954 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2955 "initialization thread\n",
2956 err);
2957 return err;
2958 }
2959 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2960 return 0;
2961}
2962
2963/*
2964 * Check whether it make sense to run itable init. thread or not.
2965 * If there is at least one uninitialized inode table, return
2966 * corresponding group number, else the loop goes through all
2967 * groups and return total number of groups.
2968 */
2969static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2970{
2971 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2972 struct ext4_group_desc *gdp = NULL;
2973
2974 for (group = 0; group < ngroups; group++) {
2975 gdp = ext4_get_group_desc(sb, group, NULL);
2976 if (!gdp)
2977 continue;
2978
2979 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2980 break;
2981 }
2982
2983 return group;
2984}
2985
2986static int ext4_li_info_new(void)
2987{
2988 struct ext4_lazy_init *eli = NULL;
2989
2990 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2991 if (!eli)
2992 return -ENOMEM;
2993
bfff6873
LC		 2994 INIT_LIST_HEAD(&eli->li_request_list);
2995 mutex_init(&eli->li_list_mtx);
2996
bfff6873
LC		 2997 eli->li_state |= EXT4_LAZYINIT_QUIT;
2998
2999 ext4_li_info = eli;
3000
3001 return 0;
3002}
3003
3004static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3005 ext4_group_t start)
3006{
3007 struct ext4_sb_info *sbi = EXT4_SB(sb);
3008 struct ext4_li_request *elr;
3009 unsigned long rnd;
3010
3011 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3012 if (!elr)
3013 return NULL;
3014
3015 elr->lr_super = sb;
3016 elr->lr_sbi = sbi;
3017 elr->lr_next_group = start;
3018
3019 /*
3020 * Randomize first schedule time of the request to
3021 * spread the inode table initialization requests
3022 * better.
3023 */
3024 get_random_bytes(&rnd, sizeof(rnd));
3025 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3026 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3027
3028 return elr;
3029}
3030
7f511862
TT		 3031int ext4_register_li_request(struct super_block *sb,
3032 ext4_group_t first_not_zeroed)
bfff6873
LC		 3033{
3034 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3035 struct ext4_li_request *elr = NULL;
bfff6873 3036 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3037 int ret = 0;
bfff6873 3038
7f511862 3039 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3040 if (sbi->s_li_request != NULL) {
3041 /*
3042 * Reset timeout so it can be computed again, because
3043 * s_li_wait_mult might have changed.
3044 */
3045 sbi->s_li_request->lr_timeout = 0;
7f511862 3046 goto out;
51ce6511 3047 }
bfff6873
LC		 3048
3049 if (first_not_zeroed == ngroups ||
3050 (sb->s_flags & MS_RDONLY) ||
55ff3840 3051 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3052 goto out;
bfff6873
LC		 3053
3054 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3055 if (!elr) {
3056 ret = -ENOMEM;
3057 goto out;
3058 }
bfff6873
LC		 3059
3060 if (NULL == ext4_li_info) {
3061 ret = ext4_li_info_new();
3062 if (ret)
3063 goto out;
3064 }
3065
3066 mutex_lock(&ext4_li_info->li_list_mtx);
3067 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3068 mutex_unlock(&ext4_li_info->li_list_mtx);
3069
3070 sbi->s_li_request = elr;
46e4690b
TM		 3071 /*
3072 * set elr to NULL here since it has been inserted to
3073 * the request_list and the removal and free of it is
3074 * handled by ext4_clear_request_list from now on.
3075 */
3076 elr = NULL;
bfff6873
LC		 3077
3078 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3079 ret = ext4_run_lazyinit_thread();
3080 if (ret)
3081 goto out;
3082 }
bfff6873 3083out:
beed5ecb
NK		 3084 mutex_unlock(&ext4_li_mtx);
3085 if (ret)
bfff6873 3086 kfree(elr);
bfff6873
LC		 3087 return ret;
3088}
3089
3090/*
3091 * We do not need to lock anything since this is called on
3092 * module unload.
3093 */
3094static void ext4_destroy_lazyinit_thread(void)
3095{
3096 /*
3097 * If thread exited earlier
3098 * there's nothing to be done.
3099 */
8f1f7453 3100 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3101 return;
3102
8f1f7453 3103 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3104}
3105
25ed6e8a
DW		 3106static int set_journal_csum_feature_set(struct super_block *sb)
3107{
3108 int ret = 1;
3109 int compat, incompat;
3110 struct ext4_sb_info *sbi = EXT4_SB(sb);
3111
3112 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3113 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3114 /* journal checksum v2 */
3115 compat = 0;
3116 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3117 } else {
3118 /* journal checksum v1 */
3119 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3120 incompat = 0;
3121 }
3122
3123 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3124 ret = jbd2_journal_set_features(sbi->s_journal,
3125 compat, 0,
3126 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3127 incompat);
3128 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3129 ret = jbd2_journal_set_features(sbi->s_journal,
3130 compat, 0,
3131 incompat);
3132 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3133 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3134 } else {
3135 jbd2_journal_clear_features(sbi->s_journal,
3136 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3137 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3138 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3139 }
3140
3141 return ret;
3142}
3143
952fc18e
TT		 3144/*
3145 * Note: calculating the overhead so we can be compatible with
3146 * historical BSD practice is quite difficult in the face of
3147 * clusters/bigalloc. This is because multiple metadata blocks from
3148 * different block group can end up in the same allocation cluster.
3149 * Calculating the exact overhead in the face of clustered allocation
3150 * requires either O(all block bitmaps) in memory or O(number of block
3151 * groups**2) in time. We will still calculate the superblock for
3152 * older file systems --- and if we come across with a bigalloc file
3153 * system with zero in s_overhead_clusters the estimate will be close to
3154 * correct especially for very large cluster sizes --- but for newer
3155 * file systems, it's better to calculate this figure once at mkfs
3156 * time, and store it in the superblock. If the superblock value is
3157 * present (even for non-bigalloc file systems), we will use it.
3158 */
3159static int count_overhead(struct super_block *sb, ext4_group_t grp,
3160 char *buf)
3161{
3162 struct ext4_sb_info *sbi = EXT4_SB(sb);
3163 struct ext4_group_desc *gdp;
3164 ext4_fsblk_t first_block, last_block, b;
3165 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3166 int s, j, count = 0;
3167
0548bbb8
TT		 3168 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3169 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3170 sbi->s_itb_per_group + 2);
3171
952fc18e
TT		 3172 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3173 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3174 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3175 for (i = 0; i < ngroups; i++) {
3176 gdp = ext4_get_group_desc(sb, i, NULL);
3177 b = ext4_block_bitmap(sb, gdp);
3178 if (b >= first_block && b <= last_block) {
3179 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3180 count++;
3181 }
3182 b = ext4_inode_bitmap(sb, gdp);
3183 if (b >= first_block && b <= last_block) {
3184 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3185 count++;
3186 }
3187 b = ext4_inode_table(sb, gdp);
3188 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3189 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3190 int c = EXT4_B2C(sbi, b - first_block);
3191 ext4_set_bit(c, buf);
3192 count++;
3193 }
3194 if (i != grp)
3195 continue;
3196 s = 0;
3197 if (ext4_bg_has_super(sb, grp)) {
3198 ext4_set_bit(s++, buf);
3199 count++;
3200 }
3201 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3202 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3203 count++;
3204 }
3205 }
3206 if (!count)
3207 return 0;
3208 return EXT4_CLUSTERS_PER_GROUP(sb) -
3209 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3210}
3211
3212/*
3213 * Compute the overhead and stash it in sbi->s_overhead
3214 */
3215int ext4_calculate_overhead(struct super_block *sb)
3216{
3217 struct ext4_sb_info *sbi = EXT4_SB(sb);
3218 struct ext4_super_block *es = sbi->s_es;
3219 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3220 ext4_fsblk_t overhead = 0;
3221 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3222
952fc18e
TT		 3223 if (!buf)
3224 return -ENOMEM;
3225
3226 /*
3227 * Compute the overhead (FS structures). This is constant
3228 * for a given filesystem unless the number of block groups
3229 * changes so we cache the previous value until it does.
3230 */
3231
3232 /*
3233 * All of the blocks before first_data_block are overhead
3234 */
3235 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3236
3237 /*
3238 * Add the overhead found in each block group
3239 */
3240 for (i = 0; i < ngroups; i++) {
3241 int blks;
3242
3243 blks = count_overhead(sb, i, buf);
3244 overhead += blks;
3245 if (blks)
3246 memset(buf, 0, PAGE_SIZE);
3247 cond_resched();
3248 }
0875a2b4
ES		 3249 /* Add the journal blocks as well */
3250 if (sbi->s_journal)
810da240 3251 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3252
952fc18e
TT		 3253 sbi->s_overhead = overhead;
3254 smp_wmb();
3255 free_page((unsigned long) buf);
3256 return 0;
3257}
3258
27dd4385
LC		 3259
3260static ext4_fsblk_t ext4_calculate_resv_clusters(struct ext4_sb_info *sbi)
3261{
3262 ext4_fsblk_t resv_clusters;
3263
3264 /*
3265 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3266 * This should cover the situations where we can not afford to run
3267 * out of space like for example punch hole, or converting
3268 * uninitialized extents in delalloc path. In most cases such
3269 * allocation would require 1, or 2 blocks, higher numbers are
3270 * very rare.
3271 */
3272 resv_clusters = ext4_blocks_count(sbi->s_es) >> sbi->s_cluster_bits;
3273
3274 do_div(resv_clusters, 50);
3275 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3276
3277 return resv_clusters;
3278}
3279
3280
3281static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3282{
3283 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3284 sbi->s_cluster_bits;
3285
3286 if (count >= clusters)
3287 return -EINVAL;
3288
3289 atomic64_set(&sbi->s_resv_clusters, count);
3290 return 0;
3291}
3292
2b2d6d01 3293static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3294{
d4c402d9 3295 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3296 struct buffer_head *bh;
617ba13b
MC		 3297 struct ext4_super_block *es = NULL;
3298 struct ext4_sb_info *sbi;
3299 ext4_fsblk_t block;
3300 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3301 ext4_fsblk_t logical_sb_block;
ac27a0ec 3302 unsigned long offset = 0;
ac27a0ec
DK		 3303 unsigned long journal_devnum = 0;
3304 unsigned long def_mount_opts;
3305 struct inode *root;
9f6200bb 3306 char *cp;
0390131b 3307 const char *descr;
dcc7dae3 3308 int ret = -ENOMEM;
281b5995 3309 int blocksize, clustersize;
4ec11028
TT		 3310 unsigned int db_count;
3311 unsigned int i;
281b5995 3312 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3313 __u64 blocks_count;
07aa2ea1 3314 int err = 0;
b3881f74 3315 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3316 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3317
3318 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3319 if (!sbi)
dcc7dae3 3320 goto out_free_orig;
705895b6
PE		 3321
3322 sbi->s_blockgroup_lock =
3323 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3324 if (!sbi->s_blockgroup_lock) {
3325 kfree(sbi);
dcc7dae3 3326 goto out_free_orig;
705895b6 3327 }
ac27a0ec 3328 sb->s_fs_info = sbi;
2c0544b2 3329 sbi->s_sb = sb;
240799cd 3330 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3331 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3332 if (sb->s_bdev->bd_part)
3333 sbi->s_sectors_written_start =
3334 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3335
9f6200bb
TT		 3336 /* Cleanup superblock name */
3337 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3338 *cp = '!';
3339
07aa2ea1 3340 /* -EINVAL is default */
dcc7dae3 3341 ret = -EINVAL;
617ba13b 3342 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3343 if (!blocksize) {
b31e1552 3344 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3345 goto out_fail;
3346 }
3347
3348 /*
617ba13b 3349 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3350 * block sizes. We need to calculate the offset from buffer start.
3351 */
617ba13b 3352 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3353 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3354 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3355 } else {
70bbb3e0 3356 logical_sb_block = sb_block;
ac27a0ec
DK		 3357 }
3358
70bbb3e0 3359 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3360 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3361 goto out_fail;
3362 }
3363 /*
3364 * Note: s_es must be initialized as soon as possible because
617ba13b 3365 * some ext4 macro-instructions depend on its value
ac27a0ec 3366 */
2716b802 3367 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3368 sbi->s_es = es;
3369 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3370 if (sb->s_magic != EXT4_SUPER_MAGIC)
3371 goto cantfind_ext4;
afc32f7e 3372 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3373
feb0ab32
DW		 3374 /* Warn if metadata_csum and gdt_csum are both set. */
3375 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3376 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3377 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3378 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3379 "redundant flags; please run fsck.");
3380
d25425f8
DW		 3381 /* Check for a known checksum algorithm */
3382 if (!ext4_verify_csum_type(sb, es)) {
3383 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3384 "unknown checksum algorithm.");
3385 silent = 1;
3386 goto cantfind_ext4;
3387 }
3388
0441984a
DW		 3389 /* Load the checksum driver */
3390 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3391 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3392 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3393 if (IS_ERR(sbi->s_chksum_driver)) {
3394 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3395 ret = PTR_ERR(sbi->s_chksum_driver);
3396 sbi->s_chksum_driver = NULL;
3397 goto failed_mount;
3398 }
3399 }
3400
a9c47317
DW		 3401 /* Check superblock checksum */
3402 if (!ext4_superblock_csum_verify(sb, es)) {
3403 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3404 "invalid superblock checksum. Run e2fsck?");
3405 silent = 1;
3406 goto cantfind_ext4;
3407 }
3408
3409 /* Precompute checksum seed for all metadata */
3410 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3411 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3412 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3413 sizeof(es->s_uuid));
3414
ac27a0ec
DK		 3415 /* Set defaults before we parse the mount options */
3416 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3417 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3418 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3419 set_opt(sb, DEBUG);
87f26807 3420 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3421 set_opt(sb, GRPID);
617ba13b 3422 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3423 set_opt(sb, NO_UID32);
ea663336 3424 /* xattr user namespace & acls are now defaulted on */
ea663336 3425 set_opt(sb, XATTR_USER);
03010a33 3426#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3427 set_opt(sb, POSIX_ACL);
2e7842b8 3428#endif
617ba13b 3429 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3430 set_opt(sb, JOURNAL_DATA);
617ba13b 3431 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3432 set_opt(sb, ORDERED_DATA);
617ba13b 3433 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3434 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3435
3436 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3437 set_opt(sb, ERRORS_PANIC);
bb4f397a 3438 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3439 set_opt(sb, ERRORS_CONT);
bb4f397a 3440 else
fd8c37ec 3441 set_opt(sb, ERRORS_RO);
8b67f04a 3442 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3443 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3444 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3445 set_opt(sb, DISCARD);
ac27a0ec 3446
08cefc7a
EB		 3447 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3448 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3449 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3450 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3451 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3452
8b67f04a 3453 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3454 set_opt(sb, BARRIER);
ac27a0ec 3455
dd919b98
AK		 3456 /*
3457 * enable delayed allocation by default
3458 * Use -o nodelalloc to turn it off
3459 */
bc0b75f7 3460 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3461 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3462 set_opt(sb, DELALLOC);
dd919b98 3463
51ce6511
LC		 3464 /*
3465 * set default s_li_wait_mult for lazyinit, for the case there is
3466 * no mount option specified.
3467 */
3468 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3469
8b67f04a 3470 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3471 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3472 ext4_msg(sb, KERN_WARNING,
3473 "failed to parse options in superblock: %s",
3474 sbi->s_es->s_mount_opts);
3475 }
5a916be1 3476 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3477 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3478 &journal_ioprio, 0))
ac27a0ec
DK		 3479 goto failed_mount;
3480
56889787
TT		 3481 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3482 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3483 "with data=journal disables delayed "
3484 "allocation and O_DIRECT support!\n");
3485 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3486 ext4_msg(sb, KERN_ERR, "can't mount with "
3487 "both data=journal and delalloc");
3488 goto failed_mount;
3489 }
3490 if (test_opt(sb, DIOREAD_NOLOCK)) {
3491 ext4_msg(sb, KERN_ERR, "can't mount with "
3492 "both data=journal and delalloc");
3493 goto failed_mount;
3494 }
3495 if (test_opt(sb, DELALLOC))
3496 clear_opt(sb, DELALLOC);
3497 }
3498
ac27a0ec 3499 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3500 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3501
617ba13b
MC		 3502 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3503 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3504 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3505 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3506 ext4_msg(sb, KERN_WARNING,
3507 "feature flags set on rev 0 fs, "
3508 "running e2fsck is recommended");
469108ff 3509
2035e776
TT		 3510 if (IS_EXT2_SB(sb)) {
3511 if (ext2_feature_set_ok(sb))
3512 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3513 "using the ext4 subsystem");
3514 else {
3515 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3516 "to feature incompatibilities");
3517 goto failed_mount;
3518 }
3519 }
3520
3521 if (IS_EXT3_SB(sb)) {
3522 if (ext3_feature_set_ok(sb))
3523 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3524 "using the ext4 subsystem");
3525 else {
3526 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3527 "to feature incompatibilities");
3528 goto failed_mount;
3529 }
3530 }
3531
ac27a0ec
DK		 3532 /*
3533 * Check feature flags regardless of the revision level, since we
3534 * previously didn't change the revision level when setting the flags,
3535 * so there is a chance incompat flags are set on a rev 0 filesystem.
3536 */
a13fb1a4 3537 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3538 goto failed_mount;
a13fb1a4 3539
261cb20c 3540 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3541 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3542 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3543 ext4_msg(sb, KERN_ERR,
3544 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3545 goto failed_mount;
3546 }
3547
ac27a0ec 3548 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3549 /* Validate the filesystem blocksize */
3550 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3551 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3552 blocksize);
ac27a0ec
DK		 3553 goto failed_mount;
3554 }
3555
2b2d6d01 3556 brelse(bh);
70bbb3e0
AM		 3557 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3558 offset = do_div(logical_sb_block, blocksize);
3559 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3560 if (!bh) {
b31e1552
ES		 3561 ext4_msg(sb, KERN_ERR,
3562 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3563 goto failed_mount;
3564 }
2716b802 3565 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3566 sbi->s_es = es;
617ba13b 3567 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3568 ext4_msg(sb, KERN_ERR,
3569 "Magic mismatch, very weird!");
ac27a0ec
DK		 3570 goto failed_mount;
3571 }
3572 }
3573
a13fb1a4
ES		 3574 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3575 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3576 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3577 has_huge_files);
3578 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3579
617ba13b
MC		 3580 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3581 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3582 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3583 } else {
3584 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3585 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3586 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3587 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3588 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3589 ext4_msg(sb, KERN_ERR,
3590 "unsupported inode size: %d",
2b2d6d01 3591 sbi->s_inode_size);
ac27a0ec
DK		 3592 goto failed_mount;
3593 }
ef7f3835
KS		 3594 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3595 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3596 }
0b8e58a1 3597
0d1ee42f
AR		 3598 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3599 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3600 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3601 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3602 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3603 ext4_msg(sb, KERN_ERR,
3604 "unsupported descriptor size %lu",
0d1ee42f
AR		 3605 sbi->s_desc_size);
3606 goto failed_mount;
3607 }
3608 } else
3609 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3610
ac27a0ec 3611 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3612 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3613 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3614 goto cantfind_ext4;
0b8e58a1 3615
617ba13b 3616 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3617 if (sbi->s_inodes_per_block == 0)
617ba13b 3618 goto cantfind_ext4;
ac27a0ec
DK		 3619 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3620 sbi->s_inodes_per_block;
0d1ee42f 3621 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3622 sbi->s_sbh = bh;
3623 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3624 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3625 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3626
68911009
LC		 3627 /* Do we have standard group size of blocksize * 8 blocks ? */
3628 if (sbi->s_blocks_per_group == blocksize << 3)
3629 set_opt2(sb, STD_GROUP_SIZE);
3630
2b2d6d01 3631 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3632 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3633 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3634 i = le32_to_cpu(es->s_flags);
3635 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3636 sbi->s_hash_unsigned = 3;
3637 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3638#ifdef __CHAR_UNSIGNED__
3639 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3640 sbi->s_hash_unsigned = 3;
3641#else
3642 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3643#endif
f99b2589 3644 }
ac27a0ec 3645
281b5995
TT		 3646 /* Handle clustersize */
3647 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3648 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3649 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3650 if (has_bigalloc) {
3651 if (clustersize < blocksize) {
3652 ext4_msg(sb, KERN_ERR,
3653 "cluster size (%d) smaller than "
3654 "block size (%d)", clustersize, blocksize);
3655 goto failed_mount;
3656 }
3657 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3658 le32_to_cpu(es->s_log_block_size);
3659 sbi->s_clusters_per_group =
3660 le32_to_cpu(es->s_clusters_per_group);
3661 if (sbi->s_clusters_per_group > blocksize * 8) {
3662 ext4_msg(sb, KERN_ERR,
3663 "#clusters per group too big: %lu",
3664 sbi->s_clusters_per_group);
3665 goto failed_mount;
3666 }
3667 if (sbi->s_blocks_per_group !=
3668 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3669 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3670 "clusters per group (%lu) inconsistent",
3671 sbi->s_blocks_per_group,
3672 sbi->s_clusters_per_group);
3673 goto failed_mount;
3674 }
3675 } else {
3676 if (clustersize != blocksize) {
3677 ext4_warning(sb, "fragment/cluster size (%d) != "
3678 "block size (%d)", clustersize,
3679 blocksize);
3680 clustersize = blocksize;
3681 }
3682 if (sbi->s_blocks_per_group > blocksize * 8) {
3683 ext4_msg(sb, KERN_ERR,
3684 "#blocks per group too big: %lu",
3685 sbi->s_blocks_per_group);
3686 goto failed_mount;
3687 }
3688 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3689 sbi->s_cluster_bits = 0;
ac27a0ec 3690 }
281b5995
TT		 3691 sbi->s_cluster_ratio = clustersize / blocksize;
3692
ac27a0ec 3693 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3694 ext4_msg(sb, KERN_ERR,
3695 "#inodes per group too big: %lu",
2b2d6d01 3696 sbi->s_inodes_per_group);
ac27a0ec
DK		 3697 goto failed_mount;
3698 }
3699
bf43d84b
ES		 3700 /*
3701 * Test whether we have more sectors than will fit in sector_t,
3702 * and whether the max offset is addressable by the page cache.
3703 */
5a9ae68a 3704 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3705 ext4_blocks_count(es));
5a9ae68a 3706 if (err) {
b31e1552 3707 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3708 " too large to mount safely on this system");
ac27a0ec 3709 if (sizeof(sector_t) < 8)
90c699a9 3710 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3711 goto failed_mount;
3712 }
3713
617ba13b
MC		 3714 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3715 goto cantfind_ext4;
e7c95593 3716
0f2ddca6
FTN		 3717 /* check blocks count against device size */
3718 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3719 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3720 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3721 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3722 ext4_blocks_count(es), blocks_count);
3723 goto failed_mount;
3724 }
3725
0b8e58a1
AD		 3726 /*
3727 * It makes no sense for the first data block to be beyond the end
3728 * of the filesystem.
3729 */
3730 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3731 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3732 "block %u is beyond end of filesystem (%llu)",
3733 le32_to_cpu(es->s_first_data_block),
3734 ext4_blocks_count(es));
e7c95593
ES		 3735 goto failed_mount;
3736 }
bd81d8ee
LV		 3737 blocks_count = (ext4_blocks_count(es) -
3738 le32_to_cpu(es->s_first_data_block) +
3739 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3740 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3741 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3742 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3743 "(block count %llu, first data block %u, "
b31e1552 3744 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3745 ext4_blocks_count(es),
3746 le32_to_cpu(es->s_first_data_block),
3747 EXT4_BLOCKS_PER_GROUP(sb));
3748 goto failed_mount;
3749 }
bd81d8ee 3750 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3751 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3752 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3753 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3754 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3755 sbi->s_group_desc = ext4_kvmalloc(db_count *
3756 sizeof(struct buffer_head *),
3757 GFP_KERNEL);
ac27a0ec 3758 if (sbi->s_group_desc == NULL) {
b31e1552 3759 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3760 ret = -ENOMEM;
ac27a0ec
DK		 3761 goto failed_mount;
3762 }
3763
9f6200bb
TT		 3764 if (ext4_proc_root)
3765 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3766
66acdcf4
TT		 3767 if (sbi->s_proc)
3768 proc_create_data("options", S_IRUGO, sbi->s_proc,
3769 &ext4_seq_options_fops, sb);
3770
705895b6 3771 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3772
3773 for (i = 0; i < db_count; i++) {
70bbb3e0 3774 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3775 sbi->s_group_desc[i] = sb_bread(sb, block);
3776 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3777 ext4_msg(sb, KERN_ERR,
3778 "can't read group descriptor %d", i);
ac27a0ec
DK		 3779 db_count = i;
3780 goto failed_mount2;
3781 }
3782 }
bfff6873 3783 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3784 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3785 goto failed_mount2;
3786 }
772cb7c8
JS		 3787 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3788 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3789 ext4_msg(sb, KERN_ERR,
3790 "unable to initialize "
3791 "flex_bg meta info!");
772cb7c8
JS		 3792 goto failed_mount2;
3793 }
3794
ac27a0ec
DK		 3795 sbi->s_gdb_count = db_count;
3796 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3797 spin_lock_init(&sbi->s_next_gen_lock);
3798
04496411
TM		 3799 init_timer(&sbi->s_err_report);
3800 sbi->s_err_report.function = print_daily_error_info;
3801 sbi->s_err_report.data = (unsigned long) sb;
3802
a75ae78f 3803 /* Register extent status tree shrinker */
d3922a77 3804 ext4_es_register_shrinker(sbi);
a75ae78f 3805
57042651 3806 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3807 ext4_count_free_clusters(sb));
ce7e010a
TT		 3808 if (!err) {
3809 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3810 ext4_count_free_inodes(sb));
3811 }
3812 if (!err) {
3813 err = percpu_counter_init(&sbi->s_dirs_counter,
3814 ext4_count_dirs(sb));
3815 }
3816 if (!err) {
57042651 3817 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a 3818 }
1ac6466f
TT		 3819 if (!err) {
3820 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3821 }
ce7e010a
TT		 3822 if (err) {
3823 ext4_msg(sb, KERN_ERR, "insufficient memory");
3824 goto failed_mount3;
3825 }
3826
c9de560d 3827 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3828 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3829
ac27a0ec
DK		 3830 /*
3831 * set up enough so that it can read an inode
3832 */
9ca92389
TT		 3833 if (!test_opt(sb, NOLOAD) &&
3834 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3835 sb->s_op = &ext4_sops;
3836 else
3837 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3838 sb->s_export_op = &ext4_export_ops;
3839 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3840#ifdef CONFIG_QUOTA
617ba13b 3841 sb->dq_op = &ext4_quota_operations;
262b4662 3842 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
7c319d32 3843 sb->s_qcop = &ext4_qctl_sysfile_operations;
262b4662
JK		 3844 else
3845 sb->s_qcop = &ext4_qctl_operations;
ac27a0ec 3846#endif
f2fa2ffc
AK		 3847 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3848
ac27a0ec 3849 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3850 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3851
3852 sb->s_root = NULL;
3853
3854 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3855 EXT4_HAS_INCOMPAT_FEATURE(sb,
3856 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3857
c5e06d10
JL		 3858 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3859 !(sb->s_flags & MS_RDONLY))
3860 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3861 goto failed_mount3;
3862
ac27a0ec
DK		 3863 /*
3864 * The first inode we look at is the journal inode. Don't try
3865 * root first: it may be modified in the journal!
3866 */
3867 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3868 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3869 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3870 goto failed_mount3;
0390131b
FM		 3871 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3872 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3873 ext4_msg(sb, KERN_ERR, "required journal recovery "
3874 "suppressed and not mounted read-only");
744692dc 3875 goto failed_mount_wq;
ac27a0ec 3876 } else {
fd8c37ec 3877 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3878 sbi->s_journal = NULL;
3879 needs_recovery = 0;
3880 goto no_journal;
ac27a0ec
DK		 3881 }
3882
f32aaf2d 3883 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3884 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3885 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3886 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3887 goto failed_mount_wq;
eb40a09c
JS		 3888 }
3889
25ed6e8a
DW		 3890 if (!set_journal_csum_feature_set(sb)) {
3891 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3892 "feature set");
3893 goto failed_mount_wq;
d4da6c9c 3894 }
818d276c 3895
ac27a0ec
DK		 3896 /* We have now updated the journal if required, so we can
3897 * validate the data journaling mode. */
3898 switch (test_opt(sb, DATA_FLAGS)) {
3899 case 0:
3900 /* No mode set, assume a default based on the journal
63f57933
AM		 3901 * capabilities: ORDERED_DATA if the journal can
3902 * cope, else JOURNAL_DATA
3903 */
dab291af
MC		 3904 if (jbd2_journal_check_available_features
3905 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3906 set_opt(sb, ORDERED_DATA);
ac27a0ec 3907 else
fd8c37ec 3908 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3909 break;
3910
617ba13b
MC		 3911 case EXT4_MOUNT_ORDERED_DATA:
3912 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3913 if (!jbd2_journal_check_available_features
3914 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3915 ext4_msg(sb, KERN_ERR, "Journal does not support "
3916 "requested data journaling mode");
744692dc 3917 goto failed_mount_wq;
ac27a0ec
DK		 3918 }
3919 default:
3920 break;
3921 }
b3881f74 3922 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3923
18aadd47
BJ		 3924 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3925
ce7e010a
TT		 3926 /*
3927 * The journal may have updated the bg summary counts, so we
3928 * need to update the global counters.
3929 */
57042651 3930 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3931 ext4_count_free_clusters(sb));
ce7e010a
TT		 3932 percpu_counter_set(&sbi->s_freeinodes_counter,
3933 ext4_count_free_inodes(sb));
3934 percpu_counter_set(&sbi->s_dirs_counter,
3935 ext4_count_dirs(sb));
57042651 3936 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3937
ce7e010a 3938no_journal:
952fc18e
TT		 3939 /*
3940 * Get the # of file system overhead blocks from the
3941 * superblock if present.
3942 */
3943 if (es->s_overhead_clusters)
3944 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3945 else {
07aa2ea1
LC		 3946 err = ext4_calculate_overhead(sb);
3947 if (err)
952fc18e
TT		 3948 goto failed_mount_wq;
3949 }
3950
fd89d5f2
TH		 3951 /*
3952 * The maximum number of concurrent works can be high and
3953 * concurrency isn't really necessary. Limit it to 1.
3954 */
2e8fa54e
JK		 3955 EXT4_SB(sb)->rsv_conversion_wq =
3956 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3957 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3958 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 3959 ret = -ENOMEM;
2e8fa54e
JK		 3960 goto failed_mount4;
3961 }
3962
3963 EXT4_SB(sb)->unrsv_conversion_wq =
3964 alloc_workqueue("ext4-unrsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3965 if (!EXT4_SB(sb)->unrsv_conversion_wq) {
3966 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3967 ret = -ENOMEM;
3968 goto failed_mount4;
4c0425ff
MC		 3969 }
3970
ac27a0ec 3971 /*
dab291af 3972 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3973 * so we can safely mount the rest of the filesystem now.
3974 */
3975
1d1fe1ee
DH		 3976 root = ext4_iget(sb, EXT4_ROOT_INO);
3977 if (IS_ERR(root)) {
b31e1552 3978 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3979 ret = PTR_ERR(root);
32a9bb57 3980 root = NULL;
ac27a0ec
DK		 3981 goto failed_mount4;
3982 }
3983 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3984 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3985 iput(root);
ac27a0ec
DK		 3986 goto failed_mount4;
3987 }
48fde701 3988 sb->s_root = d_make_root(root);
1d1fe1ee 3989 if (!sb->s_root) {
b31e1552 3990 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3991 ret = -ENOMEM;
3992 goto failed_mount4;
3993 }
ac27a0ec 3994
7e84b621
ES		 3995 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3996 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3997
3998 /* determine the minimum size of new large inodes, if present */
3999 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4000 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4001 EXT4_GOOD_OLD_INODE_SIZE;
4002 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4003 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4004 if (sbi->s_want_extra_isize <
4005 le16_to_cpu(es->s_want_extra_isize))
4006 sbi->s_want_extra_isize =
4007 le16_to_cpu(es->s_want_extra_isize);
4008 if (sbi->s_want_extra_isize <
4009 le16_to_cpu(es->s_min_extra_isize))
4010 sbi->s_want_extra_isize =
4011 le16_to_cpu(es->s_min_extra_isize);
4012 }
4013 }
4014 /* Check if enough inode space is available */
4015 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4016 sbi->s_inode_size) {
4017 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4018 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4019 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4020 "available");
ef7f3835
KS		 4021 }
4022
27dd4385
LC		 4023 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sbi));
4024 if (err) {
4025 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4026 "reserved pool", ext4_calculate_resv_clusters(sbi));
4027 goto failed_mount4a;
4028 }
4029
6fd058f7
TT		 4030 err = ext4_setup_system_zone(sb);
4031 if (err) {
b31e1552 4032 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4033 "zone (%d)", err);
94bf608a 4034 goto failed_mount4a;
6fd058f7
TT		 4035 }
4036
c2774d84 4037 ext4_ext_init(sb);
9d99012f 4038 err = ext4_mb_init(sb);
c2774d84 4039 if (err) {
421f91d2 4040 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 4041 err);
dcf2d804 4042 goto failed_mount5;
c2774d84
AK		 4043 }
4044
bfff6873
LC		 4045 err = ext4_register_li_request(sb, first_not_zeroed);
4046 if (err)
dcf2d804 4047 goto failed_mount6;
bfff6873 4048
3197ebdb
TT		 4049 sbi->s_kobj.kset = ext4_kset;
4050 init_completion(&sbi->s_kobj_unregister);
4051 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4052 "%s", sb->s_id);
dcf2d804
TM		 4053 if (err)
4054 goto failed_mount7;
3197ebdb 4055
9b2ff357
JK		 4056#ifdef CONFIG_QUOTA
4057 /* Enable quota usage during mount. */
4058 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4059 !(sb->s_flags & MS_RDONLY)) {
4060 err = ext4_enable_quotas(sb);
4061 if (err)
4062 goto failed_mount8;
4063 }
4064#endif /* CONFIG_QUOTA */
4065
617ba13b
MC		 4066 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4067 ext4_orphan_cleanup(sb, es);
4068 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4069 if (needs_recovery) {
b31e1552 4070 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4071 ext4_mark_recovery_complete(sb, es);
4072 }
4073 if (EXT4_SB(sb)->s_journal) {
4074 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4075 descr = " journalled data mode";
4076 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4077 descr = " ordered data mode";
4078 else
4079 descr = " writeback data mode";
4080 } else
4081 descr = "out journal";
4082
79add3a3
LC		 4083 if (test_opt(sb, DISCARD)) {
4084 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4085 if (!blk_queue_discard(q))
4086 ext4_msg(sb, KERN_WARNING,
4087 "mounting with \"discard\" option, but "
4088 "the device does not support discard");
4089 }
4090
d4c402d9 4091 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4092 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4093 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4094
66e61a9e
TT		 4095 if (es->s_error_count)
4096 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4097
d4c402d9 4098 kfree(orig_data);
ac27a0ec
DK		 4099 return 0;
4100
617ba13b 4101cantfind_ext4:
ac27a0ec 4102 if (!silent)
b31e1552 4103 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4104 goto failed_mount;
4105
72ba7450
TT		 4106#ifdef CONFIG_QUOTA
4107failed_mount8:
4108 kobject_del(&sbi->s_kobj);
4109#endif
dcf2d804
TM		 4110failed_mount7:
4111 ext4_unregister_li_request(sb);
4112failed_mount6:
dcf2d804 4113 ext4_mb_release(sb);
94bf608a
AV		 4114failed_mount5:
4115 ext4_ext_release(sb);
dcf2d804 4116 ext4_release_system_zone(sb);
94bf608a
AV		 4117failed_mount4a:
4118 dput(sb->s_root);
32a9bb57 4119 sb->s_root = NULL;
94bf608a 4120failed_mount4:
b31e1552 4121 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4122 if (EXT4_SB(sb)->rsv_conversion_wq)
4123 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4124 if (EXT4_SB(sb)->unrsv_conversion_wq)
4125 destroy_workqueue(EXT4_SB(sb)->unrsv_conversion_wq);
4c0425ff 4126failed_mount_wq:
0390131b
FM		 4127 if (sbi->s_journal) {
4128 jbd2_journal_destroy(sbi->s_journal);
4129 sbi->s_journal = NULL;
4130 }
ac27a0ec 4131failed_mount3:
d3922a77 4132 ext4_es_unregister_shrinker(sbi);
04496411 4133 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4134 if (sbi->s_flex_groups)
4135 ext4_kvfree(sbi->s_flex_groups);
57042651 4136 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4137 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4138 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4139 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 4140 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
c5e06d10
JL		 4141 if (sbi->s_mmp_tsk)
4142 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4143failed_mount2:
4144 for (i = 0; i < db_count; i++)
4145 brelse(sbi->s_group_desc[i]);
f18a5f21 4146 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4147failed_mount:
0441984a
DW		 4148 if (sbi->s_chksum_driver)
4149 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4150 if (sbi->s_proc) {
66acdcf4 4151 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4152 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4153 }
ac27a0ec
DK		 4154#ifdef CONFIG_QUOTA
4155 for (i = 0; i < MAXQUOTAS; i++)
4156 kfree(sbi->s_qf_names[i]);
4157#endif
617ba13b 4158 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4159 brelse(bh);
4160out_fail:
4161 sb->s_fs_info = NULL;
f6830165 4162 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4163 kfree(sbi);
dcc7dae3 4164out_free_orig:
d4c402d9 4165 kfree(orig_data);
07aa2ea1 4166 return err ? err : ret;
ac27a0ec
DK		 4167}
4168
4169/*
4170 * Setup any per-fs journal parameters now. We'll do this both on
4171 * initial mount, once the journal has been initialised but before we've
4172 * done any recovery; and again on any subsequent remount.
4173 */
617ba13b 4174static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4175{
617ba13b 4176 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4177
30773840
TT		 4178 journal->j_commit_interval = sbi->s_commit_interval;
4179 journal->j_min_batch_time = sbi->s_min_batch_time;
4180 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4181
a931da6a 4182 write_lock(&journal->j_state_lock);
ac27a0ec 4183 if (test_opt(sb, BARRIER))
dab291af 4184 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4185 else
dab291af 4186 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4187 if (test_opt(sb, DATA_ERR_ABORT))
4188 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4189 else
4190 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4191 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4192}
4193
617ba13b 4194static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4195 unsigned int journal_inum)
4196{
4197 struct inode *journal_inode;
4198 journal_t *journal;
4199
0390131b
FM		 4200 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4201
ac27a0ec
DK		 4202 /* First, test for the existence of a valid inode on disk. Bad
4203 * things happen if we iget() an unused inode, as the subsequent
4204 * iput() will try to delete it. */
4205
1d1fe1ee
DH		 4206 journal_inode = ext4_iget(sb, journal_inum);
4207 if (IS_ERR(journal_inode)) {
b31e1552 4208 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4209 return NULL;
4210 }
4211 if (!journal_inode->i_nlink) {
4212 make_bad_inode(journal_inode);
4213 iput(journal_inode);
b31e1552 4214 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4215 return NULL;
4216 }
4217
e5f8eab8 4218 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4219 journal_inode, journal_inode->i_size);
1d1fe1ee 4220 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4221 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4222 iput(journal_inode);
4223 return NULL;
4224 }
4225
dab291af 4226 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4227 if (!journal) {
b31e1552 4228 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4229 iput(journal_inode);
4230 return NULL;
4231 }
4232 journal->j_private = sb;
617ba13b 4233 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4234 return journal;
4235}
4236
617ba13b 4237static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4238 dev_t j_dev)
4239{
2b2d6d01 4240 struct buffer_head *bh;
ac27a0ec 4241 journal_t *journal;
617ba13b
MC		 4242 ext4_fsblk_t start;
4243 ext4_fsblk_t len;
ac27a0ec 4244 int hblock, blocksize;
617ba13b 4245 ext4_fsblk_t sb_block;
ac27a0ec 4246 unsigned long offset;
2b2d6d01 4247 struct ext4_super_block *es;
ac27a0ec
DK		 4248 struct block_device *bdev;
4249
0390131b
FM		 4250 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4251
b31e1552 4252 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4253 if (bdev == NULL)
4254 return NULL;
4255
ac27a0ec 4256 blocksize = sb->s_blocksize;
e1defc4f 4257 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4258 if (blocksize < hblock) {
b31e1552
ES		 4259 ext4_msg(sb, KERN_ERR,
4260 "blocksize too small for journal device");
ac27a0ec
DK		 4261 goto out_bdev;
4262 }
4263
617ba13b
MC		 4264 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4265 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4266 set_blocksize(bdev, blocksize);
4267 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4268 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4269 "external journal");
ac27a0ec
DK		 4270 goto out_bdev;
4271 }
4272
2716b802 4273 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4274 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4275 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4276 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4277 ext4_msg(sb, KERN_ERR, "external journal has "
4278 "bad superblock");
ac27a0ec
DK		 4279 brelse(bh);
4280 goto out_bdev;
4281 }
4282
617ba13b 4283 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4284 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4285 brelse(bh);
4286 goto out_bdev;
4287 }
4288
bd81d8ee 4289 len = ext4_blocks_count(es);
ac27a0ec
DK		 4290 start = sb_block + 1;
4291 brelse(bh); /* we're done with the superblock */
4292
dab291af 4293 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4294 start, len, blocksize);
4295 if (!journal) {
b31e1552 4296 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4297 goto out_bdev;
4298 }
4299 journal->j_private = sb;
9f203507 4300 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4301 wait_on_buffer(journal->j_sb_buffer);
4302 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4303 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4304 goto out_journal;
4305 }
4306 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4307 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4308 "user (unsupported) - %d",
ac27a0ec
DK		 4309 be32_to_cpu(journal->j_superblock->s_nr_users));
4310 goto out_journal;
4311 }
617ba13b
MC		 4312 EXT4_SB(sb)->journal_bdev = bdev;
4313 ext4_init_journal_params(sb, journal);
ac27a0ec 4314 return journal;
0b8e58a1 4315
ac27a0ec 4316out_journal:
dab291af 4317 jbd2_journal_destroy(journal);
ac27a0ec 4318out_bdev:
617ba13b 4319 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4320 return NULL;
4321}
4322
617ba13b
MC		 4323static int ext4_load_journal(struct super_block *sb,
4324 struct ext4_super_block *es,
ac27a0ec
DK		 4325 unsigned long journal_devnum)
4326{
4327 journal_t *journal;
4328 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4329 dev_t journal_dev;
4330 int err = 0;
4331 int really_read_only;
4332
0390131b
FM		 4333 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4334
ac27a0ec
DK		 4335 if (journal_devnum &&
4336 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4337 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4338 "numbers have changed");
ac27a0ec
DK		 4339 journal_dev = new_decode_dev(journal_devnum);
4340 } else
4341 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4342
4343 really_read_only = bdev_read_only(sb->s_bdev);
4344
4345 /*
4346 * Are we loading a blank journal or performing recovery after a
4347 * crash? For recovery, we need to check in advance whether we
4348 * can get read-write access to the device.
4349 */
617ba13b 4350 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4351 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4352 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4353 "required on readonly filesystem");
ac27a0ec 4354 if (really_read_only) {
b31e1552
ES		 4355 ext4_msg(sb, KERN_ERR, "write access "
4356 "unavailable, cannot proceed");
ac27a0ec
DK		 4357 return -EROFS;
4358 }
b31e1552
ES		 4359 ext4_msg(sb, KERN_INFO, "write access will "
4360 "be enabled during recovery");
ac27a0ec
DK		 4361 }
4362 }
4363
4364 if (journal_inum && journal_dev) {
b31e1552
ES		 4365 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4366 "and inode journals!");
ac27a0ec
DK		 4367 return -EINVAL;
4368 }
4369
4370 if (journal_inum) {
617ba13b 4371 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4372 return -EINVAL;
4373 } else {
617ba13b 4374 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4375 return -EINVAL;
4376 }
4377
90576c0b 4378 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4379 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4380
617ba13b 4381 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4382 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4383 if (!err) {
4384 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4385 if (save)
4386 memcpy(save, ((char *) es) +
4387 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4388 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4389 if (save)
4390 memcpy(((char *) es) + EXT4_S_ERR_START,
4391 save, EXT4_S_ERR_LEN);
4392 kfree(save);
4393 }
ac27a0ec
DK		 4394
4395 if (err) {
b31e1552 4396 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4397 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4398 return err;
4399 }
4400
617ba13b
MC		 4401 EXT4_SB(sb)->s_journal = journal;
4402 ext4_clear_journal_err(sb, es);
ac27a0ec 4403
c41303ce 4404 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4405 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4406 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4407
4408 /* Make sure we flush the recovery flag to disk. */
e2d67052 4409 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4410 }
4411
4412 return 0;
4413}
4414
e2d67052 4415static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4416{
e2d67052 4417 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4418 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4419 int error = 0;
ac27a0ec 4420
7c2e7087 4421 if (!sbh || block_device_ejected(sb))
c4be0c1d 4422 return error;
914258bf
TT		 4423 if (buffer_write_io_error(sbh)) {
4424 /*
4425 * Oh, dear. A previous attempt to write the
4426 * superblock failed. This could happen because the
4427 * USB device was yanked out. Or it could happen to
4428 * be a transient write error and maybe the block will
4429 * be remapped. Nothing we can do but to retry the
4430 * write and hope for the best.
4431 */
b31e1552
ES		 4432 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4433 "superblock detected");
914258bf
TT		 4434 clear_buffer_write_io_error(sbh);
4435 set_buffer_uptodate(sbh);
4436 }
71290b36
TT		 4437 /*
4438 * If the file system is mounted read-only, don't update the
4439 * superblock write time. This avoids updating the superblock
4440 * write time when we are mounting the root file system
4441 * read/only but we need to replay the journal; at that point,
4442 * for people who are east of GMT and who make their clock
4443 * tick in localtime for Windows bug-for-bug compatibility,
4444 * the clock is set in the future, and this will cause e2fsck
4445 * to complain and force a full file system check.
4446 */
4447 if (!(sb->s_flags & MS_RDONLY))
4448 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4449 if (sb->s_bdev->bd_part)
4450 es->s_kbytes_written =
4451 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4452 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4453 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4454 else
4455 es->s_kbytes_written =
4456 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4457 ext4_free_blocks_count_set(es,
4458 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4459 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4460 es->s_free_inodes_count =
4461 cpu_to_le32(percpu_counter_sum_positive(
4462 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4463 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4464 ext4_superblock_csum_set(sb);
ac27a0ec 4465 mark_buffer_dirty(sbh);
914258bf 4466 if (sync) {
c4be0c1d
TS		 4467 error = sync_dirty_buffer(sbh);
4468 if (error)
4469 return error;
4470
4471 error = buffer_write_io_error(sbh);
4472 if (error) {
b31e1552
ES		 4473 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4474 "superblock");
914258bf
TT		 4475 clear_buffer_write_io_error(sbh);
4476 set_buffer_uptodate(sbh);
4477 }
4478 }
c4be0c1d 4479 return error;
ac27a0ec
DK		 4480}
4481
ac27a0ec
DK		 4482/*
4483 * Have we just finished recovery? If so, and if we are mounting (or
4484 * remounting) the filesystem readonly, then we will end up with a
4485 * consistent fs on disk. Record that fact.
4486 */
2b2d6d01
TT		 4487static void ext4_mark_recovery_complete(struct super_block *sb,
4488 struct ext4_super_block *es)
ac27a0ec 4489{
617ba13b 4490 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4491
0390131b
FM		 4492 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4493 BUG_ON(journal != NULL);
4494 return;
4495 }
dab291af 4496 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4497 if (jbd2_journal_flush(journal) < 0)
4498 goto out;
4499
617ba13b 4500 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4501 sb->s_flags & MS_RDONLY) {
617ba13b 4502 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4503 ext4_commit_super(sb, 1);
ac27a0ec 4504 }
7ffe1ea8
HK		 4505
4506out:
dab291af 4507 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4508}
4509
4510/*
4511 * If we are mounting (or read-write remounting) a filesystem whose journal
4512 * has recorded an error from a previous lifetime, move that error to the
4513 * main filesystem now.
4514 */
2b2d6d01
TT		 4515static void ext4_clear_journal_err(struct super_block *sb,
4516 struct ext4_super_block *es)
ac27a0ec
DK		 4517{
4518 journal_t *journal;
4519 int j_errno;
4520 const char *errstr;
4521
0390131b
FM		 4522 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4523
617ba13b 4524 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4525
4526 /*
4527 * Now check for any error status which may have been recorded in the
617ba13b 4528 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4529 */
4530
dab291af 4531 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4532 if (j_errno) {
4533 char nbuf[16];
4534
617ba13b 4535 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4536 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4537 "from previous mount: %s", errstr);
12062ddd 4538 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4539
617ba13b
MC		 4540 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4541 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4542 ext4_commit_super(sb, 1);
ac27a0ec 4543
dab291af 4544 jbd2_journal_clear_err(journal);
d796c52e 4545 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4546 }
4547}
4548
4549/*
4550 * Force the running and committing transactions to commit,
4551 * and wait on the commit.
4552 */
617ba13b 4553int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4554{
4555 journal_t *journal;
ac27a0ec
DK		 4556
4557 if (sb->s_flags & MS_RDONLY)
4558 return 0;
4559
617ba13b 4560 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4561 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4562}
4563
617ba13b 4564static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4565{
14ce0cb4 4566 int ret = 0;
9eddacf9 4567 tid_t target;
06a407f1 4568 bool needs_barrier = false;
8d5d02e6 4569 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4570
9bffad1e 4571 trace_ext4_sync_fs(sb, wait);
2e8fa54e
JK		 4572 flush_workqueue(sbi->rsv_conversion_wq);
4573 flush_workqueue(sbi->unrsv_conversion_wq);
a1177825
JK		 4574 /*
4575 * Writeback quota in non-journalled quota case - journalled quota has
4576 * no dirty dquots
4577 */
4578 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4579 /*
4580 * Data writeback is possible w/o journal transaction, so barrier must
4581 * being sent at the end of the function. But we can skip it if
4582 * transaction_commit will do it for us.
4583 */
4584 target = jbd2_get_latest_transaction(sbi->s_journal);
4585 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4586 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4587 needs_barrier = true;
4588
8d5d02e6 4589 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4590 if (wait)
06a407f1
DM		 4591 ret = jbd2_log_wait_commit(sbi->s_journal, target);
4592 }
4593 if (needs_barrier) {
4594 int err;
4595 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4596 if (!ret)
4597 ret = err;
0390131b 4598 }
06a407f1
DM		 4599
4600 return ret;
4601}
4602
4603static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4604{
4605 int ret = 0;
4606
4607 trace_ext4_sync_fs(sb, wait);
4608 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4609 flush_workqueue(EXT4_SB(sb)->unrsv_conversion_wq);
4610 dquot_writeback_dquots(sb, -1);
4611 if (wait && test_opt(sb, BARRIER))
4612 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4613
14ce0cb4 4614 return ret;
ac27a0ec
DK		 4615}
4616
4617/*
4618 * LVM calls this function before a (read-only) snapshot is created. This
4619 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4620 *
4621 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4622 * state independently. It relies on upper layer to stop all data & metadata
4623 * modifications.
ac27a0ec 4624 */
c4be0c1d 4625static int ext4_freeze(struct super_block *sb)
ac27a0ec 4626{
c4be0c1d
TS		 4627 int error = 0;
4628 journal_t *journal;
ac27a0ec 4629
9ca92389
TT		 4630 if (sb->s_flags & MS_RDONLY)
4631 return 0;
ac27a0ec 4632
9ca92389 4633 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4634
9ca92389
TT		 4635 /* Now we set up the journal barrier. */
4636 jbd2_journal_lock_updates(journal);
ac27a0ec 4637
9ca92389
TT		 4638 /*
4639 * Don't clear the needs_recovery flag if we failed to flush
4640 * the journal.
4641 */
4642 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4643 if (error < 0)
4644 goto out;
9ca92389
TT		 4645
4646 /* Journal blocked and flushed, clear needs_recovery flag. */
4647 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4648 error = ext4_commit_super(sb, 1);
6b0310fb 4649out:
8e8ad8a5 4650 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4651 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4652 return error;
ac27a0ec
DK		 4653}
4654
4655/*
4656 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4657 * flag here, even though the filesystem is not technically dirty yet.
4658 */
c4be0c1d 4659static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4660{
9ca92389
TT		 4661 if (sb->s_flags & MS_RDONLY)
4662 return 0;
4663
9ca92389
TT		 4664 /* Reset the needs_recovery flag before the fs is unlocked. */
4665 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4666 ext4_commit_super(sb, 1);
c4be0c1d 4667 return 0;
ac27a0ec
DK		 4668}
4669
673c6100
TT		 4670/*
4671 * Structure to save mount options for ext4_remount's benefit
4672 */
4673struct ext4_mount_options {
4674 unsigned long s_mount_opt;
a2595b8a 4675 unsigned long s_mount_opt2;
08cefc7a
EB		 4676 kuid_t s_resuid;
4677 kgid_t s_resgid;
673c6100
TT		 4678 unsigned long s_commit_interval;
4679 u32 s_min_batch_time, s_max_batch_time;
4680#ifdef CONFIG_QUOTA
4681 int s_jquota_fmt;
4682 char *s_qf_names[MAXQUOTAS];
4683#endif
4684};
4685
2b2d6d01 4686static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4687{
2b2d6d01 4688 struct ext4_super_block *es;
617ba13b 4689 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4690 unsigned long old_sb_flags;
617ba13b 4691 struct ext4_mount_options old_opts;
c79d967d 4692 int enable_quota = 0;
8a266467 4693 ext4_group_t g;
b3881f74 4694 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4695 int err = 0;
ac27a0ec 4696#ifdef CONFIG_QUOTA
03dafb5f 4697 int i, j;
ac27a0ec 4698#endif
d4c402d9 4699 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4700
4701 /* Store the original options */
4702 old_sb_flags = sb->s_flags;
4703 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4704 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4705 old_opts.s_resuid = sbi->s_resuid;
4706 old_opts.s_resgid = sbi->s_resgid;
4707 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4708 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4709 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4710#ifdef CONFIG_QUOTA
4711 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4712 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f
CG		 4713 if (sbi->s_qf_names[i]) {
4714 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4715 GFP_KERNEL);
4716 if (!old_opts.s_qf_names[i]) {
4717 for (j = 0; j < i; j++)
4718 kfree(old_opts.s_qf_names[j]);
3e36a163 4719 kfree(orig_data);
03dafb5f
CG		 4720 return -ENOMEM;
4721 }
4722 } else
4723 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4724#endif
b3881f74
TT		 4725 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4726 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4727
4728 /*
4729 * Allow the "check" option to be passed as a remount option.
4730 */
661aa520 4731 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4732 err = -EINVAL;
4733 goto restore_opts;
4734 }
4735
4ab2f15b 4736 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4737 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4738
4739 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4740 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4741
4742 es = sbi->s_es;
4743
b3881f74 4744 if (sbi->s_journal) {
0390131b 4745 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4746 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4747 }
ac27a0ec 4748
661aa520 4749 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4750 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4751 err = -EROFS;
4752 goto restore_opts;
4753 }
4754
4755 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4756 err = dquot_suspend(sb, -1);
4757 if (err < 0)
c79d967d 4758 goto restore_opts;
c79d967d 4759
ac27a0ec
DK		 4760 /*
4761 * First of all, the unconditional stuff we have to do
4762 * to disable replay of the journal when we next remount
4763 */
4764 sb->s_flags |= MS_RDONLY;
4765
4766 /*
4767 * OK, test if we are remounting a valid rw partition
4768 * readonly, and if so set the rdonly flag and then
4769 * mark the partition as valid again.
4770 */
617ba13b
MC		 4771 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4772 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4773 es->s_state = cpu_to_le16(sbi->s_mount_state);
4774
a63c9eb2 4775 if (sbi->s_journal)
0390131b 4776 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4777 } else {
a13fb1a4
ES		 4778 /* Make sure we can mount this feature set readwrite */
4779 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4780 err = -EROFS;
4781 goto restore_opts;
4782 }
8a266467
TT		 4783 /*
4784 * Make sure the group descriptor checksums
0b8e58a1 4785 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4786 */
4787 for (g = 0; g < sbi->s_groups_count; g++) {
4788 struct ext4_group_desc *gdp =
4789 ext4_get_group_desc(sb, g, NULL);
4790
feb0ab32 4791 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4792 ext4_msg(sb, KERN_ERR,
4793 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4794 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4795 le16_to_cpu(gdp->bg_checksum));
4796 err = -EINVAL;
4797 goto restore_opts;
4798 }
4799 }
4800
ead6596b
ES		 4801 /*
4802 * If we have an unprocessed orphan list hanging
4803 * around from a previously readonly bdev mount,
4804 * require a full umount/remount for now.
4805 */
4806 if (es->s_last_orphan) {
b31e1552 4807 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4808 "remount RDWR because of unprocessed "
4809 "orphan inode list. Please "
b31e1552 4810 "umount/remount instead");
ead6596b
ES		 4811 err = -EINVAL;
4812 goto restore_opts;
4813 }
4814
ac27a0ec
DK		 4815 /*
4816 * Mounting a RDONLY partition read-write, so reread
4817 * and store the current valid flag. (It may have
4818 * been changed by e2fsck since we originally mounted
4819 * the partition.)
4820 */
0390131b
FM		 4821 if (sbi->s_journal)
4822 ext4_clear_journal_err(sb, es);
ac27a0ec 4823 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4824 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4825 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4826 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4827 EXT4_FEATURE_INCOMPAT_MMP))
4828 if (ext4_multi_mount_protect(sb,
4829 le64_to_cpu(es->s_mmp_block))) {
4830 err = -EROFS;
4831 goto restore_opts;
4832 }
c79d967d 4833 enable_quota = 1;
ac27a0ec
DK		 4834 }
4835 }
bfff6873
LC		 4836
4837 /*
4838 * Reinitialize lazy itable initialization thread based on
4839 * current settings
4840 */
4841 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4842 ext4_unregister_li_request(sb);
4843 else {
4844 ext4_group_t first_not_zeroed;
4845 first_not_zeroed = ext4_has_uninit_itable(sb);
4846 ext4_register_li_request(sb, first_not_zeroed);
4847 }
4848
6fd058f7 4849 ext4_setup_system_zone(sb);
d096ad0f 4850 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4851 ext4_commit_super(sb, 1);
0390131b 4852
ac27a0ec
DK		 4853#ifdef CONFIG_QUOTA
4854 /* Release old quota file names */
4855 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f 4856 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4857 if (enable_quota) {
4858 if (sb_any_quota_suspended(sb))
4859 dquot_resume(sb, -1);
4860 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4861 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4862 err = ext4_enable_quotas(sb);
07724f98 4863 if (err)
7c319d32 4864 goto restore_opts;
7c319d32
AK		 4865 }
4866 }
ac27a0ec 4867#endif
d4c402d9
CW		 4868
4869 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4870 kfree(orig_data);
ac27a0ec 4871 return 0;
0b8e58a1 4872
ac27a0ec
DK		 4873restore_opts:
4874 sb->s_flags = old_sb_flags;
4875 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4876 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4877 sbi->s_resuid = old_opts.s_resuid;
4878 sbi->s_resgid = old_opts.s_resgid;
4879 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4880 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4881 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4882#ifdef CONFIG_QUOTA
4883 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4884 for (i = 0; i < MAXQUOTAS; i++) {
03dafb5f 4885 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4886 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4887 }
4888#endif
d4c402d9 4889 kfree(orig_data);
ac27a0ec
DK		 4890 return err;
4891}
4892
2b2d6d01 4893static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4894{
4895 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4896 struct ext4_sb_info *sbi = EXT4_SB(sb);
4897 struct ext4_super_block *es = sbi->s_es;
27dd4385 4898 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 4899 u64 fsid;
d02a9391 4900 s64 bfree;
27dd4385 4901 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 4902
952fc18e
TT		 4903 if (!test_opt(sb, MINIX_DF))
4904 overhead = sbi->s_overhead;
ac27a0ec 4905
617ba13b 4906 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4907 buf->f_bsize = sb->s_blocksize;
b72f78cb 4908 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4909 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4910 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4911 /* prevent underflow in case that few free space is available */
57042651 4912 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 4913 buf->f_bavail = buf->f_bfree -
4914 (ext4_r_blocks_count(es) + resv_blocks);
4915 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 4916 buf->f_bavail = 0;
4917 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4918 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4919 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4920 fsid = le64_to_cpup((void *)es->s_uuid) ^
4921 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4922 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4923 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4924
ac27a0ec
DK		 4925 return 0;
4926}
4927
0b8e58a1
AD		 4928/* Helper function for writing quotas on sync - we need to start transaction
4929 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4930 * Process 1 Process 2
617ba13b 4931 * ext4_create() quota_sync()
a269eb18 4932 * jbd2_journal_start() write_dquot()
871a2931 4933 * dquot_initialize() down(dqio_mutex)
dab291af 4934 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4935 *
4936 */
4937
4938#ifdef CONFIG_QUOTA
4939
4940static inline struct inode *dquot_to_inode(struct dquot *dquot)
4941{
4c376dca 4942 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4943}
4944
617ba13b 4945static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4946{
4947 int ret, err;
4948 handle_t *handle;
4949 struct inode *inode;
4950
4951 inode = dquot_to_inode(dquot);
9924a92a 4952 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 4953 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4954 if (IS_ERR(handle))
4955 return PTR_ERR(handle);
4956 ret = dquot_commit(dquot);
617ba13b 4957 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4958 if (!ret)
4959 ret = err;
4960 return ret;
4961}
4962
617ba13b 4963static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4964{
4965 int ret, err;
4966 handle_t *handle;
4967
9924a92a 4968 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4969 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4970 if (IS_ERR(handle))
4971 return PTR_ERR(handle);
4972 ret = dquot_acquire(dquot);
617ba13b 4973 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4974 if (!ret)
4975 ret = err;
4976 return ret;
4977}
4978
617ba13b 4979static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4980{
4981 int ret, err;
4982 handle_t *handle;
4983
9924a92a 4984 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4985 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4986 if (IS_ERR(handle)) {
4987 /* Release dquot anyway to avoid endless cycle in dqput() */
4988 dquot_release(dquot);
ac27a0ec 4989 return PTR_ERR(handle);
9c3013e9 4990 }
ac27a0ec 4991 ret = dquot_release(dquot);
617ba13b 4992 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4993 if (!ret)
4994 ret = err;
4995 return ret;
4996}
4997
617ba13b 4998static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4999{
262b4662
JK		 5000 struct super_block *sb = dquot->dq_sb;
5001 struct ext4_sb_info *sbi = EXT4_SB(sb);
5002
2c8be6b2 5003 /* Are we journaling quotas? */
262b4662
JK		 5004 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5005 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5006 dquot_mark_dquot_dirty(dquot);
617ba13b 5007 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5008 } else {
5009 return dquot_mark_dquot_dirty(dquot);
5010 }
5011}
5012
617ba13b 5013static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5014{
5015 int ret, err;
5016 handle_t *handle;
5017
5018 /* Data block + inode block */
9924a92a 5019 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5020 if (IS_ERR(handle))
5021 return PTR_ERR(handle);
5022 ret = dquot_commit_info(sb, type);
617ba13b 5023 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5024 if (!ret)
5025 ret = err;
5026 return ret;
5027}
5028
5029/*
5030 * Turn on quotas during mount time - we need to find
5031 * the quota file and such...
5032 */
617ba13b 5033static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5034{
287a8095
CH		 5035 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5036 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5037}
5038
5039/*
5040 * Standard function to be called on quota_on
5041 */
617ba13b 5042static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5043 struct path *path)
ac27a0ec
DK		 5044{
5045 int err;
ac27a0ec
DK		 5046
5047 if (!test_opt(sb, QUOTA))
5048 return -EINVAL;
0623543b 5049
ac27a0ec 5050 /* Quotafile not on the same filesystem? */
d8c9584e 5051 if (path->dentry->d_sb != sb)
ac27a0ec 5052 return -EXDEV;
0623543b
JK		 5053 /* Journaling quota? */
5054 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5055 /* Quotafile not in fs root? */
f00c9e44 5056 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5057 ext4_msg(sb, KERN_WARNING,
5058 "Quota file not on filesystem root. "
5059 "Journaled quota will not work");
2b2d6d01 5060 }
0623543b
JK		 5061
5062 /*
5063 * When we journal data on quota file, we have to flush journal to see
5064 * all updates to the file when we bypass pagecache...
5065 */
0390131b 5066 if (EXT4_SB(sb)->s_journal &&
f00c9e44 5067 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 5068 /*
5069 * We don't need to lock updates but journal_flush() could
5070 * otherwise be livelocked...
5071 */
5072 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5073 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5074 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5075 if (err)
7ffe1ea8 5076 return err;
0623543b
JK		 5077 }
5078
f00c9e44 5079 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5080}
5081
7c319d32
AK		 5082static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5083 unsigned int flags)
5084{
5085 int err;
5086 struct inode *qf_inode;
5087 unsigned long qf_inums[MAXQUOTAS] = {
5088 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5089 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5090 };
5091
5092 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5093
5094 if (!qf_inums[type])
5095 return -EPERM;
5096
5097 qf_inode = ext4_iget(sb, qf_inums[type]);
5098 if (IS_ERR(qf_inode)) {
5099 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5100 return PTR_ERR(qf_inode);
5101 }
5102
bcb13850
JK		 5103 /* Don't account quota for quota files to avoid recursion */
5104 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5105 err = dquot_enable(qf_inode, type, format_id, flags);
5106 iput(qf_inode);
5107
5108 return err;
5109}
5110
5111/* Enable usage tracking for all quota types. */
5112static int ext4_enable_quotas(struct super_block *sb)
5113{
5114 int type, err = 0;
5115 unsigned long qf_inums[MAXQUOTAS] = {
5116 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5117 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5118 };
5119
5120 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5121 for (type = 0; type < MAXQUOTAS; type++) {
5122 if (qf_inums[type]) {
5123 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5124 DQUOT_USAGE_ENABLED);
5125 if (err) {
5126 ext4_warning(sb,
72ba7450
TT		 5127 "Failed to enable quota tracking "
5128 "(type=%d, err=%d). Please run "
5129 "e2fsck to fix.", type, err);
7c319d32
AK		 5130 return err;
5131 }
5132 }
5133 }
5134 return 0;
5135}
5136
5137/*
5138 * quota_on function that is used when QUOTA feature is set.
5139 */
5140static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5141 int format_id)
5142{
5143 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5144 return -EINVAL;
5145
5146 /*
5147 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5148 */
5149 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5150}
5151
ca0e05e4
DM		 5152static int ext4_quota_off(struct super_block *sb, int type)
5153{
21f97697
JK		 5154 struct inode *inode = sb_dqopt(sb)->files[type];
5155 handle_t *handle;
5156
87009d86
DM		 5157 /* Force all delayed allocation blocks to be allocated.
5158 * Caller already holds s_umount sem */
5159 if (test_opt(sb, DELALLOC))
ca0e05e4 5160 sync_filesystem(sb);
ca0e05e4 5161
0b268590
AG		 5162 if (!inode)
5163 goto out;
5164
21f97697
JK		 5165 /* Update modification times of quota files when userspace can
5166 * start looking at them */
9924a92a 5167 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5168 if (IS_ERR(handle))
5169 goto out;
5170 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5171 ext4_mark_inode_dirty(handle, inode);
5172 ext4_journal_stop(handle);
5173
5174out:
ca0e05e4
DM		 5175 return dquot_quota_off(sb, type);
5176}
5177
7c319d32
AK		 5178/*
5179 * quota_off function that is used when QUOTA feature is set.
5180 */
5181static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5182{
5183 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5184 return -EINVAL;
5185
5186 /* Disable only the limits. */
5187 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5188}
5189
ac27a0ec
DK		 5190/* Read data from quotafile - avoid pagecache and such because we cannot afford
5191 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5192 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5193 * we don't have to be afraid of races */
617ba13b 5194static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5195 size_t len, loff_t off)
5196{
5197 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5198 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5199 int err = 0;
5200 int offset = off & (sb->s_blocksize - 1);
5201 int tocopy;
5202 size_t toread;
5203 struct buffer_head *bh;
5204 loff_t i_size = i_size_read(inode);
5205
5206 if (off > i_size)
5207 return 0;
5208 if (off+len > i_size)
5209 len = i_size-off;
5210 toread = len;
5211 while (toread > 0) {
5212 tocopy = sb->s_blocksize - offset < toread ?
5213 sb->s_blocksize - offset : toread;
617ba13b 5214 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5215 if (err)
5216 return err;
5217 if (!bh) /* A hole? */
5218 memset(data, 0, tocopy);
5219 else
5220 memcpy(data, bh->b_data+offset, tocopy);
5221 brelse(bh);
5222 offset = 0;
5223 toread -= tocopy;
5224 data += tocopy;
5225 blk++;
5226 }
5227 return len;
5228}
5229
5230/* Write to quotafile (we know the transaction is already started and has
5231 * enough credits) */
617ba13b 5232static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5233 const char *data, size_t len, loff_t off)
5234{
5235 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5236 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5237 int err = 0;
5238 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5239 struct buffer_head *bh;
5240 handle_t *handle = journal_current_handle();
5241
0390131b 5242 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5243 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5244 " cancelled because transaction is not started",
9c3013e9
JK		 5245 (unsigned long long)off, (unsigned long long)len);
5246 return -EIO;
5247 }
67eeb568
DM		 5248 /*
5249 * Since we account only one data block in transaction credits,
5250 * then it is impossible to cross a block boundary.
5251 */
5252 if (sb->s_blocksize - offset < len) {
5253 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5254 " cancelled because not block aligned",
5255 (unsigned long long)off, (unsigned long long)len);
5256 return -EIO;
5257 }
5258
67eeb568
DM		 5259 bh = ext4_bread(handle, inode, blk, 1, &err);
5260 if (!bh)
5261 goto out;
62d2b5f2
JK		 5262 err = ext4_journal_get_write_access(handle, bh);
5263 if (err) {
5264 brelse(bh);
5265 goto out;
ac27a0ec 5266 }
67eeb568
DM		 5267 lock_buffer(bh);
5268 memcpy(bh->b_data+offset, data, len);
5269 flush_dcache_page(bh->b_page);
5270 unlock_buffer(bh);
62d2b5f2 5271 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5272 brelse(bh);
ac27a0ec 5273out:
0b7f7cef 5274 if (err)
ac27a0ec 5275 return err;
67eeb568
DM		 5276 if (inode->i_size < off + len) {
5277 i_size_write(inode, off + len);
617ba13b 5278 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5279 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5280 }
67eeb568 5281 return len;
ac27a0ec
DK		 5282}
5283
5284#endif
5285
152a0836
AV		 5286static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5287 const char *dev_name, void *data)
ac27a0ec 5288{
152a0836 5289 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5290}
5291
37f328eb 5292#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5293static inline void register_as_ext2(void)
5294{
5295 int err = register_filesystem(&ext2_fs_type);
5296 if (err)
5297 printk(KERN_WARNING
5298 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5299}
5300
5301static inline void unregister_as_ext2(void)
5302{
5303 unregister_filesystem(&ext2_fs_type);
5304}
2035e776
TT		 5305
5306static inline int ext2_feature_set_ok(struct super_block *sb)
5307{
5308 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5309 return 0;
5310 if (sb->s_flags & MS_RDONLY)
5311 return 1;
5312 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5313 return 0;
5314 return 1;
5315}
24b58424
TT		 5316#else
5317static inline void register_as_ext2(void) { }
5318static inline void unregister_as_ext2(void) { }
2035e776 5319static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5320#endif
5321
37f328eb 5322#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5323static inline void register_as_ext3(void)
5324{
5325 int err = register_filesystem(&ext3_fs_type);
5326 if (err)
5327 printk(KERN_WARNING
5328 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5329}
5330
5331static inline void unregister_as_ext3(void)
5332{
5333 unregister_filesystem(&ext3_fs_type);
5334}
2035e776
TT		 5335
5336static inline int ext3_feature_set_ok(struct super_block *sb)
5337{
5338 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5339 return 0;
5340 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5341 return 0;
5342 if (sb->s_flags & MS_RDONLY)
5343 return 1;
5344 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5345 return 0;
5346 return 1;
5347}
24b58424
TT		 5348#else
5349static inline void register_as_ext3(void) { }
5350static inline void unregister_as_ext3(void) { }
2035e776 5351static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5352#endif
5353
03010a33
TT		 5354static struct file_system_type ext4_fs_type = {
5355 .owner = THIS_MODULE,
5356 .name = "ext4",
152a0836 5357 .mount = ext4_mount,
03010a33
TT		 5358 .kill_sb = kill_block_super,
5359 .fs_flags = FS_REQUIRES_DEV,
5360};
7f78e035 5361MODULE_ALIAS_FS("ext4");
03010a33 5362
8f021222 5363static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5364{
5365 struct ext4_features *ef;
5366 int ret = -ENOMEM;
5367
5368 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5369 if (!ef)
5370 goto out;
5371
5372 ef->f_kobj.kset = ext4_kset;
5373 init_completion(&ef->f_kobj_unregister);
5374 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5375 "features");
5376 if (ret) {
5377 kfree(ef);
5378 goto out;
5379 }
5380
5381 ext4_feat = ef;
5382 ret = 0;
5383out:
5384 return ret;
5385}
5386
8f021222
LC		 5387static void ext4_exit_feat_adverts(void)
5388{
5389 kobject_put(&ext4_feat->f_kobj);
5390 wait_for_completion(&ext4_feat->f_kobj_unregister);
5391 kfree(ext4_feat);
5392}
5393
e9e3bcec
ES		 5394/* Shared across all ext4 file systems */
5395wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5396struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5397
5dabfc78 5398static int __init ext4_init_fs(void)
ac27a0ec 5399{
e9e3bcec 5400 int i, err;
c9de560d 5401
07c0c5d8
AV		 5402 ext4_li_info = NULL;
5403 mutex_init(&ext4_li_mtx);
5404
9a4c8019 5405 /* Build-time check for flags consistency */
12e9b892 5406 ext4_check_flag_values();
e9e3bcec
ES		 5407
5408 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5409 mutex_init(&ext4__aio_mutex[i]);
5410 init_waitqueue_head(&ext4__ioend_wq[i]);
5411 }
5412
51865fda 5413 err = ext4_init_es();
6fd058f7
TT		 5414 if (err)
5415 return err;
51865fda
ZL		 5416
5417 err = ext4_init_pageio();
5418 if (err)
5419 goto out7;
5420
5dabfc78 5421 err = ext4_init_system_zone();
bd2d0210 5422 if (err)
d44651d0 5423 goto out6;
3197ebdb 5424 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5425 if (!ext4_kset) {
5426 err = -ENOMEM;
dd68314c 5427 goto out5;
0e376b1e 5428 }
d44651d0 5429 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5430
5431 err = ext4_init_feat_adverts();
dd68314c
TT		 5432 if (err)
5433 goto out4;
857ac889 5434
5dabfc78 5435 err = ext4_init_mballoc();
ac27a0ec 5436 if (err)
6fd058f7 5437 goto out3;
c9de560d 5438
5dabfc78 5439 err = ext4_init_xattr();
c9de560d
AT		 5440 if (err)
5441 goto out2;
ac27a0ec
DK		 5442 err = init_inodecache();
5443 if (err)
5444 goto out1;
24b58424 5445 register_as_ext3();
2035e776 5446 register_as_ext2();
03010a33 5447 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5448 if (err)
5449 goto out;
bfff6873 5450
ac27a0ec
DK		 5451 return 0;
5452out:
24b58424
TT		 5453 unregister_as_ext2();
5454 unregister_as_ext3();
ac27a0ec
DK		 5455 destroy_inodecache();
5456out1:
5dabfc78 5457 ext4_exit_xattr();
c9de560d 5458out2:
5dabfc78 5459 ext4_exit_mballoc();
6fd058f7 5460out3:
8f021222 5461 ext4_exit_feat_adverts();
dd68314c 5462out4:
d44651d0
FJ		 5463 if (ext4_proc_root)
5464 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5465 kset_unregister(ext4_kset);
d44651d0 5466out5:
5dabfc78 5467 ext4_exit_system_zone();
d44651d0 5468out6:
5dabfc78 5469 ext4_exit_pageio();
51865fda
ZL		 5470out7:
5471 ext4_exit_es();
5472
ac27a0ec
DK		 5473 return err;
5474}
5475
5dabfc78 5476static void __exit ext4_exit_fs(void)
ac27a0ec 5477{
bfff6873 5478 ext4_destroy_lazyinit_thread();
24b58424
TT		 5479 unregister_as_ext2();
5480 unregister_as_ext3();
03010a33 5481 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5482 destroy_inodecache();
5dabfc78
TT		 5483 ext4_exit_xattr();
5484 ext4_exit_mballoc();
8f021222 5485 ext4_exit_feat_adverts();
9f6200bb 5486 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5487 kset_unregister(ext4_kset);
5dabfc78
TT		 5488 ext4_exit_system_zone();
5489 ext4_exit_pageio();
ac27a0ec
DK		 5490}
5491
5492MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5493MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5494MODULE_LICENSE("GPL");
5dabfc78
TT		 5495module_init(ext4_init_fs)
5496module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository