]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: silence a format string false positive
[mirror_ubuntu-artful-kernel.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
66114cad 27#include <linux/backing-dev.h>
ac27a0ec 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;
9c191f70 62static int ext4_mballoc_ready;
9f6200bb 63
617ba13b 64static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 65 unsigned long journal_devnum);
2adf6da8 66static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 67static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 68static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
617ba13b 72static int ext4_sync_fs(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
c197855e
SH		 140static int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
a9c47317 142{
9aa5d32b 143 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 144 return 1;
145
146 return es->s_checksum == ext4_superblock_csum(sb, es);
147}
148
06db49e6 149void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 150{
06db49e6
TT		 151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152
9aa5d32b 153 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 154 return;
155
156 es->s_checksum = ext4_superblock_csum(sb, es);
157}
158
9933fc0a
TT		 159void *ext4_kvmalloc(size_t size, gfp_t flags)
160{
161 void *ret;
162
8be04b93 163 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 164 if (!ret)
165 ret = __vmalloc(size, flags, PAGE_KERNEL);
166 return ret;
167}
168
169void *ext4_kvzalloc(size_t size, gfp_t flags)
170{
171 void *ret;
172
8be04b93 173 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 174 if (!ret)
175 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
176 return ret;
177}
178
8fadc143
AR		 179ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
180 struct ext4_group_desc *bg)
bd81d8ee 181{
3a14589c 182 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 183 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 184 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 185}
186
8fadc143
AR		 187ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
188 struct ext4_group_desc *bg)
bd81d8ee 189{
5272f837 190 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 191 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 192 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 193}
194
8fadc143
AR		 195ext4_fsblk_t ext4_inode_table(struct super_block *sb,
196 struct ext4_group_desc *bg)
bd81d8ee 197{
5272f837 198 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 199 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 200 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 201}
202
021b65bb
TT		 203__u32 ext4_free_group_clusters(struct super_block *sb,
204 struct ext4_group_desc *bg)
560671a0
AK		 205{
206 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 208 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 209}
210
211__u32 ext4_free_inodes_count(struct super_block *sb,
212 struct ext4_group_desc *bg)
213{
214 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 216 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 217}
218
219__u32 ext4_used_dirs_count(struct super_block *sb,
220 struct ext4_group_desc *bg)
221{
222 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 224 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 225}
226
227__u32 ext4_itable_unused_count(struct super_block *sb,
228 struct ext4_group_desc *bg)
229{
230 return le16_to_cpu(bg->bg_itable_unused_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 232 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 233}
234
8fadc143
AR		 235void ext4_block_bitmap_set(struct super_block *sb,
236 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 237{
3a14589c 238 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 239 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
240 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 241}
242
8fadc143
AR		 243void ext4_inode_bitmap_set(struct super_block *sb,
244 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 245{
5272f837 246 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 247 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
248 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 249}
250
8fadc143
AR		 251void ext4_inode_table_set(struct super_block *sb,
252 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 253{
5272f837 254 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 255 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
256 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 257}
258
021b65bb
TT		 259void ext4_free_group_clusters_set(struct super_block *sb,
260 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 261{
262 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
265}
266
267void ext4_free_inodes_set(struct super_block *sb,
268 struct ext4_group_desc *bg, __u32 count)
269{
270 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
273}
274
275void ext4_used_dirs_set(struct super_block *sb,
276 struct ext4_group_desc *bg, __u32 count)
277{
278 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
281}
282
283void ext4_itable_unused_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
285{
286 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
289}
290
d3d1faf6 291
1c13d5c0
TT		 292static void __save_error_info(struct super_block *sb, const char *func,
293 unsigned int line)
294{
295 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
296
297 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 298 if (bdev_read_only(sb->s_bdev))
299 return;
1c13d5c0
TT		 300 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
301 es->s_last_error_time = cpu_to_le32(get_seconds());
302 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
303 es->s_last_error_line = cpu_to_le32(line);
304 if (!es->s_first_error_time) {
305 es->s_first_error_time = es->s_last_error_time;
306 strncpy(es->s_first_error_func, func,
307 sizeof(es->s_first_error_func));
308 es->s_first_error_line = cpu_to_le32(line);
309 es->s_first_error_ino = es->s_last_error_ino;
310 es->s_first_error_block = es->s_last_error_block;
311 }
66e61a9e
TT		 312 /*
313 * Start the daily error reporting function if it hasn't been
314 * started already
315 */
316 if (!es->s_error_count)
317 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 318 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 319}
320
321static void save_error_info(struct super_block *sb, const char *func,
322 unsigned int line)
323{
324 __save_error_info(sb, func, line);
325 ext4_commit_super(sb, 1);
326}
327
18aadd47
BJ		 328static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
329{
330 struct super_block *sb = journal->j_private;
331 struct ext4_sb_info *sbi = EXT4_SB(sb);
332 int error = is_journal_aborted(journal);
5d3ee208 333 struct ext4_journal_cb_entry *jce;
18aadd47 334
5d3ee208 335 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 336 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 337 while (!list_empty(&txn->t_private_list)) {
338 jce = list_entry(txn->t_private_list.next,
339 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 340 list_del_init(&jce->jce_list);
341 spin_unlock(&sbi->s_md_lock);
342 jce->jce_func(sb, jce, error);
343 spin_lock(&sbi->s_md_lock);
344 }
345 spin_unlock(&sbi->s_md_lock);
346}
1c13d5c0 347
ac27a0ec
DK		 348/* Deal with the reporting of failure conditions on a filesystem such as
349 * inconsistencies detected or read IO failures.
350 *
351 * On ext2, we can store the error state of the filesystem in the
617ba13b 352 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 353 * write ordering constraints on the superblock which prevent us from
354 * writing it out straight away; and given that the journal is about to
355 * be aborted, we can't rely on the current, or future, transactions to
356 * write out the superblock safely.
357 *
dab291af 358 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 359 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 360 * that error until we've noted it down and cleared it.
361 */
362
617ba13b 363static void ext4_handle_error(struct super_block *sb)
ac27a0ec 364{
ac27a0ec
DK		 365 if (sb->s_flags & MS_RDONLY)
366 return;
367
2b2d6d01 368 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 369 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 370
4ab2f15b 371 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 372 if (journal)
dab291af 373 jbd2_journal_abort(journal, -EIO);
ac27a0ec 374 }
2b2d6d01 375 if (test_opt(sb, ERRORS_RO)) {
b31e1552 376 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 377 /*
378 * Make sure updated value of ->s_mount_flags will be visible
379 * before ->s_flags update
380 */
381 smp_wmb();
ac27a0ec
DK		 382 sb->s_flags |= MS_RDONLY;
383 }
ac27a0ec 384 if (test_opt(sb, ERRORS_PANIC))
617ba13b 385 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 386 sb->s_id);
387}
388
efbed4dc
TT		 389#define ext4_error_ratelimit(sb) \
390 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
391 "EXT4-fs error")
392
12062ddd 393void __ext4_error(struct super_block *sb, const char *function,
c398eda0 394 unsigned int line, const char *fmt, ...)
ac27a0ec 395{
0ff2ea7d 396 struct va_format vaf;
ac27a0ec
DK		 397 va_list args;
398
efbed4dc
TT		 399 if (ext4_error_ratelimit(sb)) {
400 va_start(args, fmt);
401 vaf.fmt = fmt;
402 vaf.va = &args;
403 printk(KERN_CRIT
404 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
405 sb->s_id, function, line, current->comm, &vaf);
406 va_end(args);
407 }
f3fc0210 408 save_error_info(sb, function, line);
617ba13b 409 ext4_handle_error(sb);
ac27a0ec
DK		 410}
411
e7c96e8e
JP		 412void __ext4_error_inode(struct inode *inode, const char *function,
413 unsigned int line, ext4_fsblk_t block,
414 const char *fmt, ...)
273df556
FM		 415{
416 va_list args;
f7c21177 417 struct va_format vaf;
1c13d5c0 418 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 419
1c13d5c0
TT		 420 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
421 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 422 if (ext4_error_ratelimit(inode->i_sb)) {
423 va_start(args, fmt);
424 vaf.fmt = fmt;
425 vaf.va = &args;
426 if (block)
427 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
428 "inode #%lu: block %llu: comm %s: %pV\n",
429 inode->i_sb->s_id, function, line, inode->i_ino,
430 block, current->comm, &vaf);
431 else
432 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
433 "inode #%lu: comm %s: %pV\n",
434 inode->i_sb->s_id, function, line, inode->i_ino,
435 current->comm, &vaf);
436 va_end(args);
437 }
1c13d5c0 438 save_error_info(inode->i_sb, function, line);
273df556
FM		 439 ext4_handle_error(inode->i_sb);
440}
441
e7c96e8e
JP		 442void __ext4_error_file(struct file *file, const char *function,
443 unsigned int line, ext4_fsblk_t block,
444 const char *fmt, ...)
273df556
FM		 445{
446 va_list args;
f7c21177 447 struct va_format vaf;
1c13d5c0 448 struct ext4_super_block *es;
496ad9aa 449 struct inode *inode = file_inode(file);
273df556
FM		 450 char pathname[80], *path;
451
1c13d5c0
TT		 452 es = EXT4_SB(inode->i_sb)->s_es;
453 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 454 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 455 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 456 if (IS_ERR(path))
457 path = "(unknown)";
458 va_start(args, fmt);
459 vaf.fmt = fmt;
460 vaf.va = &args;
461 if (block)
462 printk(KERN_CRIT
463 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
464 "block %llu: comm %s: path %s: %pV\n",
465 inode->i_sb->s_id, function, line, inode->i_ino,
466 block, current->comm, path, &vaf);
467 else
468 printk(KERN_CRIT
469 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
470 "comm %s: path %s: %pV\n",
471 inode->i_sb->s_id, function, line, inode->i_ino,
472 current->comm, path, &vaf);
473 va_end(args);
474 }
1c13d5c0 475 save_error_info(inode->i_sb, function, line);
273df556
FM		 476 ext4_handle_error(inode->i_sb);
477}
478
722887dd
TT		 479const char *ext4_decode_error(struct super_block *sb, int errno,
480 char nbuf[16])
ac27a0ec
DK		 481{
482 char *errstr = NULL;
483
484 switch (errno) {
485 case -EIO:
486 errstr = "IO failure";
487 break;
488 case -ENOMEM:
489 errstr = "Out of memory";
490 break;
491 case -EROFS:
78f1ddbb
TT		 492 if (!sb || (EXT4_SB(sb)->s_journal &&
493 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 494 errstr = "Journal has aborted";
495 else
496 errstr = "Readonly filesystem";
497 break;
498 default:
499 /* If the caller passed in an extra buffer for unknown
500 * errors, textualise them now. Else we just return
501 * NULL. */
502 if (nbuf) {
503 /* Check for truncated error codes... */
504 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
505 errstr = nbuf;
506 }
507 break;
508 }
509
510 return errstr;
511}
512
617ba13b 513/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 514 * automatically and invokes the appropriate error response. */
515
c398eda0
TT		 516void __ext4_std_error(struct super_block *sb, const char *function,
517 unsigned int line, int errno)
ac27a0ec
DK		 518{
519 char nbuf[16];
520 const char *errstr;
521
522 /* Special case: if the error is EROFS, and we're not already
523 * inside a transaction, then there's really no point in logging
524 * an error. */
525 if (errno == -EROFS && journal_current_handle() == NULL &&
526 (sb->s_flags & MS_RDONLY))
527 return;
528
efbed4dc
TT		 529 if (ext4_error_ratelimit(sb)) {
530 errstr = ext4_decode_error(sb, errno, nbuf);
531 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
532 sb->s_id, function, line, errstr);
533 }
ac27a0ec 534
efbed4dc 535 save_error_info(sb, function, line);
617ba13b 536 ext4_handle_error(sb);
ac27a0ec
DK		 537}
538
539/*
617ba13b 540 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 541 * abort function may be used to deal with unrecoverable failures such
542 * as journal IO errors or ENOMEM at a critical moment in log management.
543 *
544 * We unconditionally force the filesystem into an ABORT|READONLY state,
545 * unless the error response on the fs has been set to panic in which
546 * case we take the easy way out and panic immediately.
547 */
548
c67d859e 549void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 550 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 551{
552 va_list args;
553
1c13d5c0 554 save_error_info(sb, function, line);
ac27a0ec 555 va_start(args, fmt);
c398eda0
TT		 556 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
557 function, line);
ac27a0ec
DK		 558 vprintk(fmt, args);
559 printk("\n");
560 va_end(args);
561
1c13d5c0
TT		 562 if ((sb->s_flags & MS_RDONLY) == 0) {
563 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 564 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 565 /*
566 * Make sure updated value of ->s_mount_flags will be visible
567 * before ->s_flags update
568 */
569 smp_wmb();
570 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 571 if (EXT4_SB(sb)->s_journal)
572 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
573 save_error_info(sb, function, line);
574 }
ac27a0ec 575 if (test_opt(sb, ERRORS_PANIC))
617ba13b 576 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 577}
578
e7c96e8e
JP		 579void __ext4_msg(struct super_block *sb,
580 const char *prefix, const char *fmt, ...)
b31e1552 581{
0ff2ea7d 582 struct va_format vaf;
b31e1552
ES		 583 va_list args;
584
efbed4dc
TT		 585 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
586 return;
587
b31e1552 588 va_start(args, fmt);
0ff2ea7d
JP		 589 vaf.fmt = fmt;
590 vaf.va = &args;
591 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 592 va_end(args);
593}
594
b03a2f7e
AD		 595#define ext4_warning_ratelimit(sb) \
596 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
597 "EXT4-fs warning")
598
12062ddd 599void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 600 unsigned int line, const char *fmt, ...)
ac27a0ec 601{
0ff2ea7d 602 struct va_format vaf;
ac27a0ec
DK		 603 va_list args;
604
b03a2f7e 605 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 606 return;
607
ac27a0ec 608 va_start(args, fmt);
0ff2ea7d
JP		 609 vaf.fmt = fmt;
610 vaf.va = &args;
611 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
612 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 613 va_end(args);
614}
615
b03a2f7e
AD		 616void __ext4_warning_inode(const struct inode *inode, const char *function,
617 unsigned int line, const char *fmt, ...)
618{
619 struct va_format vaf;
620 va_list args;
621
622 if (!ext4_warning_ratelimit(inode->i_sb))
623 return;
624
625 va_start(args, fmt);
626 vaf.fmt = fmt;
627 vaf.va = &args;
628 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
629 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
630 function, line, inode->i_ino, current->comm, &vaf);
631 va_end(args);
632}
633
e29136f8
TT		 634void __ext4_grp_locked_error(const char *function, unsigned int line,
635 struct super_block *sb, ext4_group_t grp,
636 unsigned long ino, ext4_fsblk_t block,
637 const char *fmt, ...)
5d1b1b3f
AK		 638__releases(bitlock)
639__acquires(bitlock)
640{
0ff2ea7d 641 struct va_format vaf;
5d1b1b3f
AK		 642 va_list args;
643 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
644
1c13d5c0
TT		 645 es->s_last_error_ino = cpu_to_le32(ino);
646 es->s_last_error_block = cpu_to_le64(block);
647 __save_error_info(sb, function, line);
0ff2ea7d 648
efbed4dc
TT		 649 if (ext4_error_ratelimit(sb)) {
650 va_start(args, fmt);
651 vaf.fmt = fmt;
652 vaf.va = &args;
653 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
654 sb->s_id, function, line, grp);
655 if (ino)
656 printk(KERN_CONT "inode %lu: ", ino);
657 if (block)
658 printk(KERN_CONT "block %llu:",
659 (unsigned long long) block);
660 printk(KERN_CONT "%pV\n", &vaf);
661 va_end(args);
662 }
5d1b1b3f
AK		 663
664 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 665 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 666 return;
667 }
1c13d5c0 668
5d1b1b3f
AK		 669 ext4_unlock_group(sb, grp);
670 ext4_handle_error(sb);
671 /*
672 * We only get here in the ERRORS_RO case; relocking the group
673 * may be dangerous, but nothing bad will happen since the
674 * filesystem will have already been marked read/only and the
675 * journal has been aborted. We return 1 as a hint to callers
676 * who might what to use the return value from
25985edc 677 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 678 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
679 * aggressively from the ext4 function in question, with a
680 * more appropriate error code.
681 */
682 ext4_lock_group(sb, grp);
683 return;
684}
685
617ba13b 686void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 687{
617ba13b 688 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 689
617ba13b 690 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 691 return;
692
12062ddd 693 ext4_warning(sb,
ac27a0ec
DK		 694 "updating to rev %d because of new feature flag, "
695 "running e2fsck is recommended",
617ba13b 696 EXT4_DYNAMIC_REV);
ac27a0ec 697
617ba13b
MC		 698 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
699 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
700 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 701 /* leave es->s_feature_*compat flags alone */
702 /* es->s_uuid will be set by e2fsck if empty */
703
704 /*
705 * The rest of the superblock fields should be zero, and if not it
706 * means they are likely already in use, so leave them alone. We
707 * can leave it up to e2fsck to clean up any inconsistencies there.
708 */
709}
710
711/*
712 * Open the external journal device
713 */
b31e1552 714static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 715{
716 struct block_device *bdev;
717 char b[BDEVNAME_SIZE];
718
d4d77629 719 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 720 if (IS_ERR(bdev))
721 goto fail;
722 return bdev;
723
724fail:
b31e1552 725 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 726 __bdevname(dev, b), PTR_ERR(bdev));
727 return NULL;
728}
729
730/*
731 * Release the journal device
732 */
4385bab1 733static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 734{
4385bab1 735 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 736}
737
4385bab1 738static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 739{
740 struct block_device *bdev;
ac27a0ec
DK		 741 bdev = sbi->journal_bdev;
742 if (bdev) {
4385bab1 743 ext4_blkdev_put(bdev);
ac27a0ec
DK		 744 sbi->journal_bdev = NULL;
745 }
ac27a0ec
DK		 746}
747
748static inline struct inode *orphan_list_entry(struct list_head *l)
749{
617ba13b 750 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 751}
752
617ba13b 753static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 754{
755 struct list_head *l;
756
b31e1552
ES		 757 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
758 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 759
760 printk(KERN_ERR "sb_info orphan list:\n");
761 list_for_each(l, &sbi->s_orphan) {
762 struct inode *inode = orphan_list_entry(l);
763 printk(KERN_ERR " "
764 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
765 inode->i_sb->s_id, inode->i_ino, inode,
766 inode->i_mode, inode->i_nlink,
767 NEXT_ORPHAN(inode));
768 }
769}
770
2b2d6d01 771static void ext4_put_super(struct super_block *sb)
ac27a0ec 772{
617ba13b
MC		 773 struct ext4_sb_info *sbi = EXT4_SB(sb);
774 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 775 int i, err;
ac27a0ec 776
857ac889 777 ext4_unregister_li_request(sb);
e0ccfd95
CH		 778 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
779
2e8fa54e 780 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 781 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 782
0390131b
FM		 783 if (sbi->s_journal) {
784 err = jbd2_journal_destroy(sbi->s_journal);
785 sbi->s_journal = NULL;
786 if (err < 0)
c67d859e 787 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 788 }
d4edac31 789
d3922a77 790 ext4_es_unregister_shrinker(sbi);
9105bb14 791 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 792 ext4_release_system_zone(sb);
793 ext4_mb_release(sb);
794 ext4_ext_release(sb);
795 ext4_xattr_put_super(sb);
796
ac27a0ec 797 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 798 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 799 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 800 }
58c5873a 801 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 802 ext4_commit_super(sb, 1);
803
240799cd 804 if (sbi->s_proc) {
66acdcf4 805 remove_proc_entry("options", sbi->s_proc);
9f6200bb 806 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 807 }
3197ebdb 808 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 809
810 for (i = 0; i < sbi->s_gdb_count; i++)
811 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 812 kvfree(sbi->s_group_desc);
813 kvfree(sbi->s_flex_groups);
57042651 814 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 815 percpu_counter_destroy(&sbi->s_freeinodes_counter);
816 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 817 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 818 brelse(sbi->s_sbh);
819#ifdef CONFIG_QUOTA
a2d4a646 820 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 821 kfree(sbi->s_qf_names[i]);
822#endif
823
824 /* Debugging code just in case the in-memory inode orphan list
825 * isn't empty. The on-disk one can be non-empty if we've
826 * detected an error and taken the fs readonly, but the
827 * in-memory list had better be clean by this point. */
828 if (!list_empty(&sbi->s_orphan))
829 dump_orphan_list(sb, sbi);
830 J_ASSERT(list_empty(&sbi->s_orphan));
831
89d96a6f 832 sync_blockdev(sb->s_bdev);
f98393a6 833 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 834 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
835 /*
836 * Invalidate the journal device's buffers. We don't want them
837 * floating about in memory - the physical journal device may
838 * hotswapped, and it breaks the `ro-after' testing code.
839 */
840 sync_blockdev(sbi->journal_bdev);
f98393a6 841 invalidate_bdev(sbi->journal_bdev);
617ba13b 842 ext4_blkdev_remove(sbi);
ac27a0ec 843 }
9c191f70
M		 844 if (sbi->s_mb_cache) {
845 ext4_xattr_destroy_cache(sbi->s_mb_cache);
846 sbi->s_mb_cache = NULL;
847 }
c5e06d10
JL		 848 if (sbi->s_mmp_tsk)
849 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 850 sb->s_fs_info = NULL;
3197ebdb
TT		 851 /*
852 * Now that we are completely done shutting down the
853 * superblock, we need to actually destroy the kobject.
854 */
3197ebdb
TT		 855 kobject_put(&sbi->s_kobj);
856 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 857 if (sbi->s_chksum_driver)
858 crypto_free_shash(sbi->s_chksum_driver);
705895b6 859 kfree(sbi->s_blockgroup_lock);
ac27a0ec 860 kfree(sbi);
ac27a0ec
DK		 861}
862
e18b890b 863static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 864
865/*
866 * Called inside transaction, so use GFP_NOFS
867 */
617ba13b 868static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 869{
617ba13b 870 struct ext4_inode_info *ei;
ac27a0ec 871
e6b4f8da 872 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 873 if (!ei)
874 return NULL;
0b8e58a1 875
ac27a0ec 876 ei->vfs_inode.i_version = 1;
202ee5df 877 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 878 INIT_LIST_HEAD(&ei->i_prealloc_list);
879 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 880 ext4_es_init_tree(&ei->i_es_tree);
881 rwlock_init(&ei->i_es_lock);
edaa53ca 882 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 883 ei->i_es_all_nr = 0;
edaa53ca 884 ei->i_es_shk_nr = 0;
dd475925 885 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 886 ei->i_reserved_data_blocks = 0;
887 ei->i_reserved_meta_blocks = 0;
888 ei->i_allocated_meta_blocks = 0;
9d0be502 889 ei->i_da_metadata_calc_len = 0;
7e731bc9 890 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 891 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 892#ifdef CONFIG_QUOTA
893 ei->i_reserved_quota = 0;
96c7e0d9 894 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 895#endif
8aefcd55 896 ei->jinode = NULL;
2e8fa54e 897 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 898 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 899 ei->i_sync_tid = 0;
900 ei->i_datasync_tid = 0;
f7ad6d2e 901 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 902 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 903 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
b30ab0e0 904#ifdef CONFIG_EXT4_FS_ENCRYPTION
b7236e21 905 ei->i_crypt_info = NULL;
b30ab0e0 906#endif
ac27a0ec
DK		 907 return &ei->vfs_inode;
908}
909
7ff9c073
TT		 910static int ext4_drop_inode(struct inode *inode)
911{
912 int drop = generic_drop_inode(inode);
913
914 trace_ext4_drop_inode(inode, drop);
915 return drop;
916}
917
fa0d7e3d
NP		 918static void ext4_i_callback(struct rcu_head *head)
919{
920 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 921 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
922}
923
617ba13b 924static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 925{
9f7dd93d 926 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 927 ext4_msg(inode->i_sb, KERN_ERR,
928 "Inode %lu (%p): orphan list check failed!",
929 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 930 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
931 EXT4_I(inode), sizeof(struct ext4_inode_info),
932 true);
933 dump_stack();
934 }
fa0d7e3d 935 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 936}
937
51cc5068 938static void init_once(void *foo)
ac27a0ec 939{
617ba13b 940 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 941
a35afb83 942 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 943 init_rwsem(&ei->xattr_sem);
0e855ac8 944 init_rwsem(&ei->i_data_sem);
a35afb83 945 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 946}
947
e67bc2b3 948static int __init init_inodecache(void)
ac27a0ec 949{
617ba13b
MC		 950 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
951 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 952 0, (SLAB_RECLAIM_ACCOUNT|
953 SLAB_MEM_SPREAD),
20c2df83 954 init_once);
617ba13b 955 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 956 return -ENOMEM;
957 return 0;
958}
959
960static void destroy_inodecache(void)
961{
8c0a8537
KS		 962 /*
963 * Make sure all delayed rcu free inodes are flushed before we
964 * destroy cache.
965 */
966 rcu_barrier();
617ba13b 967 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 968}
969
0930fcc1 970void ext4_clear_inode(struct inode *inode)
ac27a0ec 971{
0930fcc1 972 invalidate_inode_buffers(inode);
dbd5768f 973 clear_inode(inode);
9f754758 974 dquot_drop(inode);
c2ea3fde 975 ext4_discard_preallocations(inode);
51865fda 976 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 977 if (EXT4_I(inode)->jinode) {
978 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
979 EXT4_I(inode)->jinode);
980 jbd2_free_inode(EXT4_I(inode)->jinode);
981 EXT4_I(inode)->jinode = NULL;
982 }
b7236e21
TT		 983#ifdef CONFIG_EXT4_FS_ENCRYPTION
984 if (EXT4_I(inode)->i_crypt_info)
c936e1ec 985 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
b7236e21 986#endif
ac27a0ec
DK		 987}
988
1b961ac0 989static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 990 u64 ino, u32 generation)
ac27a0ec 991{
ac27a0ec 992 struct inode *inode;
ac27a0ec 993
617ba13b 994 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 995 return ERR_PTR(-ESTALE);
617ba13b 996 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 997 return ERR_PTR(-ESTALE);
998
999 /* iget isn't really right if the inode is currently unallocated!!
1000 *
617ba13b 1001 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1002 * deleted, so we should be safe.
1003 *
1004 * Currently we don't know the generation for parent directory, so
1005 * a generation of 0 means "accept any"
1006 */
f4bb2981 1007 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1008 if (IS_ERR(inode))
1009 return ERR_CAST(inode);
1010 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1011 iput(inode);
1012 return ERR_PTR(-ESTALE);
1013 }
1b961ac0
CH		 1014
1015 return inode;
1016}
1017
1018static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1019 int fh_len, int fh_type)
1b961ac0
CH		 1020{
1021 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1022 ext4_nfs_get_inode);
1023}
1024
1025static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1026 int fh_len, int fh_type)
1b961ac0
CH		 1027{
1028 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1029 ext4_nfs_get_inode);
ac27a0ec
DK		 1030}
1031
c39a7f84
TO		 1032/*
1033 * Try to release metadata pages (indirect blocks, directories) which are
1034 * mapped via the block device. Since these pages could have journal heads
1035 * which would prevent try_to_free_buffers() from freeing them, we must use
1036 * jbd2 layer's try_to_free_buffers() function to release them.
1037 */
0b8e58a1
AD		 1038static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1039 gfp_t wait)
c39a7f84
TO		 1040{
1041 journal_t *journal = EXT4_SB(sb)->s_journal;
1042
1043 WARN_ON(PageChecked(page));
1044 if (!page_has_buffers(page))
1045 return 0;
1046 if (journal)
1047 return jbd2_journal_try_to_free_buffers(journal, page,
1048 wait & ~__GFP_WAIT);
1049 return try_to_free_buffers(page);
1050}
1051
ac27a0ec 1052#ifdef CONFIG_QUOTA
af5bc92d 1053#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1054#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1055
617ba13b
MC		 1056static int ext4_write_dquot(struct dquot *dquot);
1057static int ext4_acquire_dquot(struct dquot *dquot);
1058static int ext4_release_dquot(struct dquot *dquot);
1059static int ext4_mark_dquot_dirty(struct dquot *dquot);
1060static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1061static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1062 struct path *path);
ca0e05e4 1063static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1064static int ext4_quota_on_mount(struct super_block *sb, int type);
1065static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1066 size_t len, loff_t off);
617ba13b 1067static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1068 const char *data, size_t len, loff_t off);
7c319d32
AK		 1069static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1070 unsigned int flags);
1071static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1072
96c7e0d9
JK		 1073static struct dquot **ext4_get_dquots(struct inode *inode)
1074{
1075 return EXT4_I(inode)->i_dquot;
1076}
1077
61e225dc 1078static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1079 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1080 .write_dquot = ext4_write_dquot,
1081 .acquire_dquot = ext4_acquire_dquot,
1082 .release_dquot = ext4_release_dquot,
1083 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1084 .write_info = ext4_write_info,
1085 .alloc_dquot = dquot_alloc,
1086 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1087};
1088
0d54b217 1089static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1090 .quota_on = ext4_quota_on,
ca0e05e4 1091 .quota_off = ext4_quota_off,
287a8095 1092 .quota_sync = dquot_quota_sync,
0a240339 1093 .get_state = dquot_get_state,
287a8095
CH		 1094 .set_info = dquot_set_dqinfo,
1095 .get_dqblk = dquot_get_dqblk,
1096 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1097};
1098#endif
1099
ee9b6d61 1100static const struct super_operations ext4_sops = {
617ba13b
MC		 1101 .alloc_inode = ext4_alloc_inode,
1102 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1103 .write_inode = ext4_write_inode,
1104 .dirty_inode = ext4_dirty_inode,
7ff9c073 1105 .drop_inode = ext4_drop_inode,
0930fcc1 1106 .evict_inode = ext4_evict_inode,
617ba13b 1107 .put_super = ext4_put_super,
617ba13b 1108 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1109 .freeze_fs = ext4_freeze,
1110 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1111 .statfs = ext4_statfs,
1112 .remount_fs = ext4_remount,
617ba13b 1113 .show_options = ext4_show_options,
ac27a0ec 1114#ifdef CONFIG_QUOTA
617ba13b
MC		 1115 .quota_read = ext4_quota_read,
1116 .quota_write = ext4_quota_write,
96c7e0d9 1117 .get_dquots = ext4_get_dquots,
ac27a0ec 1118#endif
c39a7f84 1119 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1120};
1121
39655164 1122static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1123 .fh_to_dentry = ext4_fh_to_dentry,
1124 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1125 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1126};
1127
1128enum {
1129 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1130 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1131 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1132 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1133 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1134 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1135 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1136 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1137 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1138 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1139 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1140 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
923ae0ff 1141 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1449032b 1142 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1143 Opt_lazytime, Opt_nolazytime,
1449032b 1144 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1145 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1146 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1147 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1148 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1149};
1150
a447c093 1151static const match_table_t tokens = {
ac27a0ec
DK		 1152 {Opt_bsd_df, "bsddf"},
1153 {Opt_minix_df, "minixdf"},
1154 {Opt_grpid, "grpid"},
1155 {Opt_grpid, "bsdgroups"},
1156 {Opt_nogrpid, "nogrpid"},
1157 {Opt_nogrpid, "sysvgroups"},
1158 {Opt_resgid, "resgid=%u"},
1159 {Opt_resuid, "resuid=%u"},
1160 {Opt_sb, "sb=%u"},
1161 {Opt_err_cont, "errors=continue"},
1162 {Opt_err_panic, "errors=panic"},
1163 {Opt_err_ro, "errors=remount-ro"},
1164 {Opt_nouid32, "nouid32"},
ac27a0ec 1165 {Opt_debug, "debug"},
72578c33
TT		 1166 {Opt_removed, "oldalloc"},
1167 {Opt_removed, "orlov"},
ac27a0ec
DK		 1168 {Opt_user_xattr, "user_xattr"},
1169 {Opt_nouser_xattr, "nouser_xattr"},
1170 {Opt_acl, "acl"},
1171 {Opt_noacl, "noacl"},
e3bb52ae 1172 {Opt_noload, "norecovery"},
5a916be1 1173 {Opt_noload, "noload"},
72578c33
TT		 1174 {Opt_removed, "nobh"},
1175 {Opt_removed, "bh"},
ac27a0ec 1176 {Opt_commit, "commit=%u"},
30773840
TT		 1177 {Opt_min_batch_time, "min_batch_time=%u"},
1178 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1179 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1180 {Opt_journal_path, "journal_path=%s"},
818d276c 1181 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1182 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 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"},
923ae0ff 1205 {Opt_dax, "dax"},
c9de560d 1206 {Opt_stripe, "stripe=%u"},
64769240 1207 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1208 {Opt_lazytime, "lazytime"},
1209 {Opt_nolazytime, "nolazytime"},
dd919b98 1210 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1211 {Opt_removed, "mblk_io_submit"},
1212 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1213 {Opt_block_validity, "block_validity"},
1214 {Opt_noblock_validity, "noblock_validity"},
240799cd 1215 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1216 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1217 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1218 {Opt_auto_da_alloc, "auto_da_alloc"},
1219 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1220 {Opt_dioread_nolock, "dioread_nolock"},
1221 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1222 {Opt_discard, "discard"},
1223 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1224 {Opt_init_itable, "init_itable=%u"},
1225 {Opt_init_itable, "init_itable"},
1226 {Opt_noinit_itable, "noinit_itable"},
df981d03 1227 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1228 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1229 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1230 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1231 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1232 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1233 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1234 {Opt_err, NULL},
ac27a0ec
DK		 1235};
1236
617ba13b 1237static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1238{
617ba13b 1239 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1240 char *options = (char *) *data;
1241
1242 if (!options || strncmp(options, "sb=", 3) != 0)
1243 return 1; /* Default location */
0b8e58a1 1244
ac27a0ec 1245 options += 3;
0b8e58a1 1246 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1247 sb_block = simple_strtoul(options, &options, 0);
1248 if (*options && *options != ',') {
4776004f 1249 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1250 (char *) *data);
1251 return 1;
1252 }
1253 if (*options == ',')
1254 options++;
1255 *data = (void *) options;
0b8e58a1 1256
ac27a0ec
DK		 1257 return sb_block;
1258}
1259
b3881f74 1260#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1261static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1262 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1263
56c50f11
DM		 1264#ifdef CONFIG_QUOTA
1265static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1266{
1267 struct ext4_sb_info *sbi = EXT4_SB(sb);
1268 char *qname;
03dafb5f 1269 int ret = -1;
56c50f11
DM		 1270
1271 if (sb_any_quota_loaded(sb) &&
1272 !sbi->s_qf_names[qtype]) {
1273 ext4_msg(sb, KERN_ERR,
1274 "Cannot change journaled "
1275 "quota options when quota turned on");
57f73c2c 1276 return -1;
56c50f11 1277 }
262b4662
JK		 1278 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1279 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1280 "when QUOTA feature is enabled");
1281 return -1;
1282 }
56c50f11
DM		 1283 qname = match_strdup(args);
1284 if (!qname) {
1285 ext4_msg(sb, KERN_ERR,
1286 "Not enough memory for storing quotafile name");
57f73c2c 1287 return -1;
56c50f11 1288 }
03dafb5f
CG		 1289 if (sbi->s_qf_names[qtype]) {
1290 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1291 ret = 1;
1292 else
1293 ext4_msg(sb, KERN_ERR,
1294 "%s quota file already specified",
1295 QTYPE2NAME(qtype));
1296 goto errout;
56c50f11 1297 }
03dafb5f 1298 if (strchr(qname, '/')) {
56c50f11
DM		 1299 ext4_msg(sb, KERN_ERR,
1300 "quotafile must be on filesystem root");
03dafb5f 1301 goto errout;
56c50f11 1302 }
03dafb5f 1303 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1304 set_opt(sb, QUOTA);
56c50f11 1305 return 1;
03dafb5f
CG		 1306errout:
1307 kfree(qname);
1308 return ret;
56c50f11
DM		 1309}
1310
1311static int clear_qf_name(struct super_block *sb, int qtype)
1312{
1313
1314 struct ext4_sb_info *sbi = EXT4_SB(sb);
1315
1316 if (sb_any_quota_loaded(sb) &&
1317 sbi->s_qf_names[qtype]) {
1318 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1319 " when quota turned on");
57f73c2c 1320 return -1;
56c50f11 1321 }
03dafb5f 1322 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1323 sbi->s_qf_names[qtype] = NULL;
1324 return 1;
1325}
1326#endif
1327
26092bf5
TT		 1328#define MOPT_SET 0x0001
1329#define MOPT_CLEAR 0x0002
1330#define MOPT_NOSUPPORT 0x0004
1331#define MOPT_EXPLICIT 0x0008
1332#define MOPT_CLEAR_ERR 0x0010
1333#define MOPT_GTE0 0x0020
ac27a0ec 1334#ifdef CONFIG_QUOTA
26092bf5
TT		 1335#define MOPT_Q 0
1336#define MOPT_QFMT 0x0040
1337#else
1338#define MOPT_Q MOPT_NOSUPPORT
1339#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1340#endif
26092bf5 1341#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1342#define MOPT_NO_EXT2 0x0100
1343#define MOPT_NO_EXT3 0x0200
1344#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1345#define MOPT_STRING 0x0400
26092bf5
TT		 1346
1347static const struct mount_opts {
1348 int token;
1349 int mount_opt;
1350 int flags;
1351} ext4_mount_opts[] = {
1352 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1353 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1354 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1355 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1356 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1357 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1358 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1359 MOPT_EXT4_ONLY | MOPT_SET},
1360 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1361 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1362 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1363 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1364 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1365 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1366 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1367 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1368 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1369 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c
TT		 1370 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1371 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1372 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1373 EXT4_MOUNT_JOURNAL_CHECKSUM),
1374 MOPT_EXT4_ONLY | MOPT_SET},
1375 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1376 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1377 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1378 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1379 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1380 MOPT_NO_EXT2 | MOPT_SET},
1381 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1382 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1383 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1384 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1385 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1386 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1387 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1388 {Opt_commit, 0, MOPT_GTE0},
1389 {Opt_max_batch_time, 0, MOPT_GTE0},
1390 {Opt_min_batch_time, 0, MOPT_GTE0},
1391 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1392 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1393 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1394 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1395 {Opt_resuid, 0, MOPT_GTE0},
1396 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1397 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1398 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1399 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1400 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1401 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1402 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1403 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1404 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1405 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1406#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1407 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1408 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1409#else
26092bf5
TT		 1410 {Opt_acl, 0, MOPT_NOSUPPORT},
1411 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1412#endif
26092bf5
TT		 1413 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1414 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1415 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1416 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1417 MOPT_SET | MOPT_Q},
1418 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1419 MOPT_SET | MOPT_Q},
1420 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1421 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1422 {Opt_usrjquota, 0, MOPT_Q},
1423 {Opt_grpjquota, 0, MOPT_Q},
1424 {Opt_offusrjquota, 0, MOPT_Q},
1425 {Opt_offgrpjquota, 0, MOPT_Q},
1426 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1427 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1428 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1429 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1430 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1431 {Opt_err, 0, 0}
1432};
1433
1434static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1435 substring_t *args, unsigned long *journal_devnum,
1436 unsigned int *journal_ioprio, int is_remount)
1437{
1438 struct ext4_sb_info *sbi = EXT4_SB(sb);
1439 const struct mount_opts *m;
08cefc7a
EB		 1440 kuid_t uid;
1441 kgid_t gid;
26092bf5
TT		 1442 int arg = 0;
1443
57f73c2c
TT		 1444#ifdef CONFIG_QUOTA
1445 if (token == Opt_usrjquota)
1446 return set_qf_name(sb, USRQUOTA, &args[0]);
1447 else if (token == Opt_grpjquota)
1448 return set_qf_name(sb, GRPQUOTA, &args[0]);
1449 else if (token == Opt_offusrjquota)
1450 return clear_qf_name(sb, USRQUOTA);
1451 else if (token == Opt_offgrpjquota)
1452 return clear_qf_name(sb, GRPQUOTA);
1453#endif
26092bf5 1454 switch (token) {
f7048605
TT		 1455 case Opt_noacl:
1456 case Opt_nouser_xattr:
1457 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1458 break;
26092bf5
TT		 1459 case Opt_sb:
1460 return 1; /* handled by get_sb_block() */
1461 case Opt_removed:
5f3633e3 1462 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1463 return 1;
26092bf5
TT		 1464 case Opt_abort:
1465 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1466 return 1;
1467 case Opt_i_version:
1468 sb->s_flags |= MS_I_VERSION;
1469 return 1;
a26f4992
TT		 1470 case Opt_lazytime:
1471 sb->s_flags |= MS_LAZYTIME;
1472 return 1;
1473 case Opt_nolazytime:
1474 sb->s_flags &= ~MS_LAZYTIME;
1475 return 1;
26092bf5
TT		 1476 }
1477
5f3633e3
JK		 1478 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1479 if (token == m->token)
1480 break;
1481
1482 if (m->token == Opt_err) {
1483 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1484 "or missing value", opt);
1485 return -1;
1486 }
1487
8dc0aa8c
TT		 1488 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1489 ext4_msg(sb, KERN_ERR,
1490 "Mount option \"%s\" incompatible with ext2", opt);
1491 return -1;
1492 }
1493 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1494 ext4_msg(sb, KERN_ERR,
1495 "Mount option \"%s\" incompatible with ext3", opt);
1496 return -1;
1497 }
1498
ad4eec61 1499 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1500 return -1;
1501 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1502 return -1;
1503 if (m->flags & MOPT_EXPLICIT)
1504 set_opt2(sb, EXPLICIT_DELALLOC);
1505 if (m->flags & MOPT_CLEAR_ERR)
1506 clear_opt(sb, ERRORS_MASK);
1507 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1508 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1509 "options when quota turned on");
1510 return -1;
1511 }
1512
1513 if (m->flags & MOPT_NOSUPPORT) {
1514 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1515 } else if (token == Opt_commit) {
1516 if (arg == 0)
1517 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1518 sbi->s_commit_interval = HZ * arg;
1519 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1520 sbi->s_max_batch_time = arg;
1521 } else if (token == Opt_min_batch_time) {
1522 sbi->s_min_batch_time = arg;
1523 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1524 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1525 ext4_msg(sb, KERN_ERR,
1526 "EXT4-fs: inode_readahead_blks must be "
1527 "0 or a power of 2 smaller than 2^31");
26092bf5 1528 return -1;
5f3633e3
JK		 1529 }
1530 sbi->s_inode_readahead_blks = arg;
1531 } else if (token == Opt_init_itable) {
1532 set_opt(sb, INIT_INODE_TABLE);
1533 if (!args->from)
1534 arg = EXT4_DEF_LI_WAIT_MULT;
1535 sbi->s_li_wait_mult = arg;
1536 } else if (token == Opt_max_dir_size_kb) {
1537 sbi->s_max_dir_size_kb = arg;
1538 } else if (token == Opt_stripe) {
1539 sbi->s_stripe = arg;
1540 } else if (token == Opt_resuid) {
1541 uid = make_kuid(current_user_ns(), arg);
1542 if (!uid_valid(uid)) {
1543 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1544 return -1;
1545 }
5f3633e3
JK		 1546 sbi->s_resuid = uid;
1547 } else if (token == Opt_resgid) {
1548 gid = make_kgid(current_user_ns(), arg);
1549 if (!gid_valid(gid)) {
1550 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1551 return -1;
1552 }
1553 sbi->s_resgid = gid;
1554 } else if (token == Opt_journal_dev) {
1555 if (is_remount) {
1556 ext4_msg(sb, KERN_ERR,
1557 "Cannot specify journal on remount");
1558 return -1;
1559 }
1560 *journal_devnum = arg;
ad4eec61
ES		 1561 } else if (token == Opt_journal_path) {
1562 char *journal_path;
1563 struct inode *journal_inode;
1564 struct path path;
1565 int error;
1566
1567 if (is_remount) {
1568 ext4_msg(sb, KERN_ERR,
1569 "Cannot specify journal on remount");
1570 return -1;
1571 }
1572 journal_path = match_strdup(&args[0]);
1573 if (!journal_path) {
1574 ext4_msg(sb, KERN_ERR, "error: could not dup "
1575 "journal device string");
1576 return -1;
1577 }
1578
1579 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1580 if (error) {
1581 ext4_msg(sb, KERN_ERR, "error: could not find "
1582 "journal device path: error %d", error);
1583 kfree(journal_path);
1584 return -1;
1585 }
1586
2b0143b5 1587 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1588 if (!S_ISBLK(journal_inode->i_mode)) {
1589 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1590 "is not a block device", journal_path);
1591 path_put(&path);
1592 kfree(journal_path);
1593 return -1;
1594 }
1595
1596 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1597 path_put(&path);
1598 kfree(journal_path);
5f3633e3
JK		 1599 } else if (token == Opt_journal_ioprio) {
1600 if (arg > 7) {
1601 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1602 " (must be 0-7)");
1603 return -1;
1604 }
1605 *journal_ioprio =
1606 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1607 } else if (token == Opt_test_dummy_encryption) {
1608#ifdef CONFIG_EXT4_FS_ENCRYPTION
1609 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1610 ext4_msg(sb, KERN_WARNING,
1611 "Test dummy encryption mode enabled");
1612#else
1613 ext4_msg(sb, KERN_WARNING,
1614 "Test dummy encryption mount option ignored");
1615#endif
5f3633e3
JK		 1616 } else if (m->flags & MOPT_DATAJ) {
1617 if (is_remount) {
1618 if (!sbi->s_journal)
1619 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1620 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1621 ext4_msg(sb, KERN_ERR,
26092bf5 1622 "Cannot change data mode on remount");
26092bf5 1623 return -1;
ac27a0ec 1624 }
26092bf5 1625 } else {
5f3633e3
JK		 1626 clear_opt(sb, DATA_FLAGS);
1627 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1628 }
5f3633e3
JK		 1629#ifdef CONFIG_QUOTA
1630 } else if (m->flags & MOPT_QFMT) {
1631 if (sb_any_quota_loaded(sb) &&
1632 sbi->s_jquota_fmt != m->mount_opt) {
1633 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1634 "quota options when quota turned on");
1635 return -1;
1636 }
262b4662
JK		 1637 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1638 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1639 ext4_msg(sb, KERN_ERR,
1640 "Cannot set journaled quota options "
1641 "when QUOTA feature is enabled");
1642 return -1;
1643 }
5f3633e3 1644 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff
RZ		 1645#endif
1646#ifndef CONFIG_FS_DAX
1647 } else if (token == Opt_dax) {
1648 ext4_msg(sb, KERN_INFO, "dax option not supported");
1649 return -1;
5f3633e3
JK		 1650#endif
1651 } else {
1652 if (!args->from)
1653 arg = 1;
1654 if (m->flags & MOPT_CLEAR)
1655 arg = !arg;
1656 else if (unlikely(!(m->flags & MOPT_SET))) {
1657 ext4_msg(sb, KERN_WARNING,
1658 "buggy handling of option %s", opt);
1659 WARN_ON(1);
1660 return -1;
1661 }
1662 if (arg != 0)
1663 sbi->s_mount_opt |= m->mount_opt;
1664 else
1665 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1666 }
5f3633e3 1667 return 1;
26092bf5
TT		 1668}
1669
1670static int parse_options(char *options, struct super_block *sb,
1671 unsigned long *journal_devnum,
1672 unsigned int *journal_ioprio,
1673 int is_remount)
1674{
1675 struct ext4_sb_info *sbi = EXT4_SB(sb);
1676 char *p;
1677 substring_t args[MAX_OPT_ARGS];
1678 int token;
1679
1680 if (!options)
1681 return 1;
1682
1683 while ((p = strsep(&options, ",")) != NULL) {
1684 if (!*p)
1685 continue;
1686 /*
1687 * Initialize args struct so we know whether arg was
1688 * found; some options take optional arguments.
1689 */
caecd0af 1690 args[0].to = args[0].from = NULL;
26092bf5
TT		 1691 token = match_token(p, tokens, args);
1692 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1693 journal_ioprio, is_remount) < 0)
1694 return 0;
ac27a0ec
DK		 1695 }
1696#ifdef CONFIG_QUOTA
262b4662
JK		 1697 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1698 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1699 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1700 "feature is enabled");
1701 return 0;
1702 }
ac27a0ec 1703 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1704 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1705 clear_opt(sb, USRQUOTA);
ac27a0ec 1706
482a7425 1707 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1708 clear_opt(sb, GRPQUOTA);
ac27a0ec 1709
56c50f11 1710 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1711 ext4_msg(sb, KERN_ERR, "old and new quota "
1712 "format mixing");
ac27a0ec
DK		 1713 return 0;
1714 }
1715
1716 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1717 ext4_msg(sb, KERN_ERR, "journaled quota format "
1718 "not specified");
ac27a0ec
DK		 1719 return 0;
1720 }
ac27a0ec
DK		 1721 }
1722#endif
261cb20c
JK		 1723 if (test_opt(sb, DIOREAD_NOLOCK)) {
1724 int blocksize =
1725 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1726
1727 if (blocksize < PAGE_CACHE_SIZE) {
1728 ext4_msg(sb, KERN_ERR, "can't mount with "
1729 "dioread_nolock if block size != PAGE_SIZE");
1730 return 0;
1731 }
1732 }
d4f76107
JK		 1733 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1734 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1735 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1736 "in data=ordered mode");
1737 return 0;
1738 }
ac27a0ec
DK		 1739 return 1;
1740}
1741
2adf6da8
TT		 1742static inline void ext4_show_quota_options(struct seq_file *seq,
1743 struct super_block *sb)
1744{
1745#if defined(CONFIG_QUOTA)
1746 struct ext4_sb_info *sbi = EXT4_SB(sb);
1747
1748 if (sbi->s_jquota_fmt) {
1749 char *fmtname = "";
1750
1751 switch (sbi->s_jquota_fmt) {
1752 case QFMT_VFS_OLD:
1753 fmtname = "vfsold";
1754 break;
1755 case QFMT_VFS_V0:
1756 fmtname = "vfsv0";
1757 break;
1758 case QFMT_VFS_V1:
1759 fmtname = "vfsv1";
1760 break;
1761 }
1762 seq_printf(seq, ",jqfmt=%s", fmtname);
1763 }
1764
1765 if (sbi->s_qf_names[USRQUOTA])
1766 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1767
1768 if (sbi->s_qf_names[GRPQUOTA])
1769 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1770#endif
1771}
1772
5a916be1
TT		 1773static const char *token2str(int token)
1774{
50df9fd5 1775 const struct match_token *t;
5a916be1
TT		 1776
1777 for (t = tokens; t->token != Opt_err; t++)
1778 if (t->token == token && !strchr(t->pattern, '='))
1779 break;
1780 return t->pattern;
1781}
1782
2adf6da8
TT		 1783/*
1784 * Show an option if
1785 * - it's set to a non-default value OR
1786 * - if the per-sb default is different from the global default
1787 */
66acdcf4
TT		 1788static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1789 int nodefs)
2adf6da8 1790{
2adf6da8
TT		 1791 struct ext4_sb_info *sbi = EXT4_SB(sb);
1792 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1793 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1794 const struct mount_opts *m;
66acdcf4 1795 char sep = nodefs ? '\n' : ',';
2adf6da8 1796
66acdcf4
TT		 1797#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1798#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1799
1800 if (sbi->s_sb_block != 1)
5a916be1
TT		 1801 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1802
1803 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1804 int want_set = m->flags & MOPT_SET;
1805 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1806 (m->flags & MOPT_CLEAR_ERR))
1807 continue;
66acdcf4 1808 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1809 continue; /* skip if same as the default */
1810 if ((want_set &&
1811 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1812 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1813 continue; /* select Opt_noFoo vs Opt_Foo */
1814 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1815 }
5a916be1 1816
08cefc7a 1817 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1818 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1819 SEQ_OPTS_PRINT("resuid=%u",
1820 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1821 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1822 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1823 SEQ_OPTS_PRINT("resgid=%u",
1824 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1825 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1826 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1827 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1828 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1829 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1830 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1831 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1832 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1833 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1834 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1835 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1836 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1837 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1838 if (sb->s_flags & MS_I_VERSION)
5a916be1 1839 SEQ_OPTS_PUTS("i_version");
66acdcf4 1840 if (nodefs || sbi->s_stripe)
5a916be1 1841 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1842 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1843 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1844 SEQ_OPTS_PUTS("data=journal");
1845 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1846 SEQ_OPTS_PUTS("data=ordered");
1847 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1848 SEQ_OPTS_PUTS("data=writeback");
1849 }
66acdcf4
TT		 1850 if (nodefs ||
1851 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1852 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1853 sbi->s_inode_readahead_blks);
2adf6da8 1854
66acdcf4
TT		 1855 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1856 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1857 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1858 if (nodefs || sbi->s_max_dir_size_kb)
1859 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1860
1861 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1862 return 0;
1863}
1864
66acdcf4
TT		 1865static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1866{
1867 return _ext4_show_options(seq, root->d_sb, 0);
1868}
1869
1870static int options_seq_show(struct seq_file *seq, void *offset)
1871{
1872 struct super_block *sb = seq->private;
1873 int rc;
1874
1875 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1876 rc = _ext4_show_options(seq, sb, 1);
1877 seq_puts(seq, "\n");
1878 return rc;
1879}
1880
1881static int options_open_fs(struct inode *inode, struct file *file)
1882{
d9dda78b 1883 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1884}
1885
1886static const struct file_operations ext4_seq_options_fops = {
1887 .owner = THIS_MODULE,
1888 .open = options_open_fs,
1889 .read = seq_read,
1890 .llseek = seq_lseek,
1891 .release = single_release,
1892};
1893
617ba13b 1894static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1895 int read_only)
1896{
617ba13b 1897 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1898 int res = 0;
1899
617ba13b 1900 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1901 ext4_msg(sb, KERN_ERR, "revision level too high, "
1902 "forcing read-only mode");
ac27a0ec
DK		 1903 res = MS_RDONLY;
1904 }
1905 if (read_only)
281b5995 1906 goto done;
617ba13b 1907 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1908 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1909 "running e2fsck is recommended");
c8b459f4 1910 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1911 ext4_msg(sb, KERN_WARNING,
1912 "warning: mounting fs with errors, "
1913 "running e2fsck is recommended");
ed3ce80a 1914 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1915 le16_to_cpu(es->s_mnt_count) >=
1916 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1917 ext4_msg(sb, KERN_WARNING,
1918 "warning: maximal mount count reached, "
1919 "running e2fsck is recommended");
ac27a0ec
DK		 1920 else if (le32_to_cpu(es->s_checkinterval) &&
1921 (le32_to_cpu(es->s_lastcheck) +
1922 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1923 ext4_msg(sb, KERN_WARNING,
1924 "warning: checktime reached, "
1925 "running e2fsck is recommended");
0b8e58a1 1926 if (!sbi->s_journal)
0390131b 1927 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1928 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1929 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1930 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1931 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1932 ext4_update_dynamic_rev(sb);
0390131b
FM		 1933 if (sbi->s_journal)
1934 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1935
e2d67052 1936 ext4_commit_super(sb, 1);
281b5995 1937done:
ac27a0ec 1938 if (test_opt(sb, DEBUG))
a9df9a49 1939 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1940 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1941 sb->s_blocksize,
1942 sbi->s_groups_count,
617ba13b
MC		 1943 EXT4_BLOCKS_PER_GROUP(sb),
1944 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1945 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1946
7abc52c2 1947 cleancache_init_fs(sb);
ac27a0ec
DK		 1948 return res;
1949}
1950
117fff10
TT		 1951int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1952{
1953 struct ext4_sb_info *sbi = EXT4_SB(sb);
1954 struct flex_groups *new_groups;
1955 int size;
1956
1957 if (!sbi->s_log_groups_per_flex)
1958 return 0;
1959
1960 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1961 if (size <= sbi->s_flex_groups_allocated)
1962 return 0;
1963
1964 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1965 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1966 if (!new_groups) {
1967 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1968 size / (int) sizeof(struct flex_groups));
1969 return -ENOMEM;
1970 }
1971
1972 if (sbi->s_flex_groups) {
1973 memcpy(new_groups, sbi->s_flex_groups,
1974 (sbi->s_flex_groups_allocated *
1975 sizeof(struct flex_groups)));
b93b41d4 1976 kvfree(sbi->s_flex_groups);
117fff10
TT		 1977 }
1978 sbi->s_flex_groups = new_groups;
1979 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1980 return 0;
1981}
1982
772cb7c8
JS		 1983static int ext4_fill_flex_info(struct super_block *sb)
1984{
1985 struct ext4_sb_info *sbi = EXT4_SB(sb);
1986 struct ext4_group_desc *gdp = NULL;
772cb7c8 1987 ext4_group_t flex_group;
117fff10 1988 int i, err;
772cb7c8 1989
503358ae 1990 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1991 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1992 sbi->s_log_groups_per_flex = 0;
1993 return 1;
1994 }
1995
117fff10
TT		 1996 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1997 if (err)
9933fc0a 1998 goto failed;
772cb7c8 1999
772cb7c8 2000 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2001 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2002
2003 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2004 atomic_add(ext4_free_inodes_count(sb, gdp),
2005 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2006 atomic64_add(ext4_free_group_clusters(sb, gdp),
2007 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2008 atomic_add(ext4_used_dirs_count(sb, gdp),
2009 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2010 }
2011
2012 return 1;
2013failed:
2014 return 0;
2015}
2016
feb0ab32
DW		 2017static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2018 struct ext4_group_desc *gdp)
717d50e4 2019{
feb0ab32 2020 int offset;
717d50e4 2021 __u16 crc = 0;
feb0ab32 2022 __le32 le_group = cpu_to_le32(block_group);
717d50e4 2023
9aa5d32b 2024 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2025 /* Use new metadata_csum algorithm */
d6a77105 2026 __le16 save_csum;
feb0ab32
DW		 2027 __u32 csum32;
2028
d6a77105 2029 save_csum = gdp->bg_checksum;
feb0ab32
DW		 2030 gdp->bg_checksum = 0;
2031 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2032 sizeof(le_group));
2033 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2034 sbi->s_desc_size);
d6a77105 2035 gdp->bg_checksum = save_csum;
feb0ab32
DW		 2036
2037 crc = csum32 & 0xFFFF;
2038 goto out;
717d50e4
AD		 2039 }
2040
feb0ab32 2041 /* old crc16 code */
813d32f9
DW		 2042 if (!(sbi->s_es->s_feature_ro_compat &
2043 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2044 return 0;
2045
feb0ab32
DW		 2046 offset = offsetof(struct ext4_group_desc, bg_checksum);
2047
2048 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2049 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2050 crc = crc16(crc, (__u8 *)gdp, offset);
2051 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2052 /* for checksum of struct ext4_group_desc do the rest...*/
2053 if ((sbi->s_es->s_feature_incompat &
2054 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2055 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2056 crc = crc16(crc, (__u8 *)gdp + offset,
2057 le16_to_cpu(sbi->s_es->s_desc_size) -
2058 offset);
2059
2060out:
717d50e4
AD		 2061 return cpu_to_le16(crc);
2062}
2063
feb0ab32 2064int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2065 struct ext4_group_desc *gdp)
2066{
feb0ab32
DW		 2067 if (ext4_has_group_desc_csum(sb) &&
2068 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2069 block_group, gdp)))
717d50e4
AD		 2070 return 0;
2071
2072 return 1;
2073}
2074
feb0ab32
DW		 2075void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2076 struct ext4_group_desc *gdp)
2077{
2078 if (!ext4_has_group_desc_csum(sb))
2079 return;
2080 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2081}
2082
ac27a0ec 2083/* Called at mount-time, super-block is locked */
bfff6873
LC		 2084static int ext4_check_descriptors(struct super_block *sb,
2085 ext4_group_t *first_not_zeroed)
ac27a0ec 2086{
617ba13b
MC		 2087 struct ext4_sb_info *sbi = EXT4_SB(sb);
2088 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2089 ext4_fsblk_t last_block;
bd81d8ee
LV		 2090 ext4_fsblk_t block_bitmap;
2091 ext4_fsblk_t inode_bitmap;
2092 ext4_fsblk_t inode_table;
ce421581 2093 int flexbg_flag = 0;
bfff6873 2094 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2095
ce421581
JS		 2096 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2097 flexbg_flag = 1;
2098
af5bc92d 2099 ext4_debug("Checking group descriptors");
ac27a0ec 2100
197cd65a
AM		 2101 for (i = 0; i < sbi->s_groups_count; i++) {
2102 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2103
ce421581 2104 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2105 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2106 else
2107 last_block = first_block +
617ba13b 2108 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2109
bfff6873
LC		 2110 if ((grp == sbi->s_groups_count) &&
2111 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2112 grp = i;
2113
8fadc143 2114 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2115 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2116 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2117 "Block bitmap for group %u not in group "
b31e1552 2118 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2119 return 0;
2120 }
8fadc143 2121 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2122 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2123 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2124 "Inode bitmap for group %u not in group "
b31e1552 2125 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2126 return 0;
2127 }
8fadc143 2128 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2129 if (inode_table < first_block ||
2b2d6d01 2130 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2131 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2132 "Inode table for group %u not in group "
b31e1552 2133 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2134 return 0;
2135 }
955ce5f5 2136 ext4_lock_group(sb, i);
feb0ab32 2137 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2138 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2139 "Checksum for group %u failed (%u!=%u)",
2140 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2141 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2142 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2143 ext4_unlock_group(sb, i);
8a266467 2144 return 0;
7ee1ec4c 2145 }
717d50e4 2146 }
955ce5f5 2147 ext4_unlock_group(sb, i);
ce421581
JS		 2148 if (!flexbg_flag)
2149 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2150 }
bfff6873
LC		 2151 if (NULL != first_not_zeroed)
2152 *first_not_zeroed = grp;
ac27a0ec
DK		 2153 return 1;
2154}
2155
617ba13b 2156/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2157 * the superblock) which were deleted from all directories, but held open by
2158 * a process at the time of a crash. We walk the list and try to delete these
2159 * inodes at recovery time (only with a read-write filesystem).
2160 *
2161 * In order to keep the orphan inode chain consistent during traversal (in
2162 * case of crash during recovery), we link each inode into the superblock
2163 * orphan list_head and handle it the same way as an inode deletion during
2164 * normal operation (which journals the operations for us).
2165 *
2166 * We only do an iget() and an iput() on each inode, which is very safe if we
2167 * accidentally point at an in-use or already deleted inode. The worst that
2168 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2169 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2170 * e2fsck was run on this filesystem, and it must have already done the orphan
2171 * inode cleanup for us, so we can safely abort without any further action.
2172 */
2b2d6d01
TT		 2173static void ext4_orphan_cleanup(struct super_block *sb,
2174 struct ext4_super_block *es)
ac27a0ec
DK		 2175{
2176 unsigned int s_flags = sb->s_flags;
2177 int nr_orphans = 0, nr_truncates = 0;
2178#ifdef CONFIG_QUOTA
2179 int i;
2180#endif
2181 if (!es->s_last_orphan) {
2182 jbd_debug(4, "no orphan inodes to clean up\n");
2183 return;
2184 }
2185
a8f48a95 2186 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2187 ext4_msg(sb, KERN_ERR, "write access "
2188 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2189 return;
2190 }
2191
d39195c3
AG		 2192 /* Check if feature set would not allow a r/w mount */
2193 if (!ext4_feature_set_ok(sb, 0)) {
2194 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2195 "unknown ROCOMPAT features");
2196 return;
2197 }
2198
617ba13b 2199 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2200 /* don't clear list on RO mount w/ errors */
2201 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2202 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2203 "clearing orphan list.\n");
c25f9bc6
ES		 2204 es->s_last_orphan = 0;
2205 }
ac27a0ec
DK		 2206 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2207 return;
2208 }
2209
2210 if (s_flags & MS_RDONLY) {
b31e1552 2211 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2212 sb->s_flags &= ~MS_RDONLY;
2213 }
2214#ifdef CONFIG_QUOTA
2215 /* Needed for iput() to work correctly and not trash data */
2216 sb->s_flags |= MS_ACTIVE;
2217 /* Turn on quotas so that they are updated correctly */
a2d4a646 2218 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2219 if (EXT4_SB(sb)->s_qf_names[i]) {
2220 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2221 if (ret < 0)
b31e1552
ES		 2222 ext4_msg(sb, KERN_ERR,
2223 "Cannot turn on journaled "
2224 "quota: error %d", ret);
ac27a0ec
DK		 2225 }
2226 }
2227#endif
2228
2229 while (es->s_last_orphan) {
2230 struct inode *inode;
2231
97bd42b9
JB		 2232 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2233 if (IS_ERR(inode)) {
ac27a0ec
DK		 2234 es->s_last_orphan = 0;
2235 break;
2236 }
2237
617ba13b 2238 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2239 dquot_initialize(inode);
ac27a0ec 2240 if (inode->i_nlink) {
566370a2
PT		 2241 if (test_opt(sb, DEBUG))
2242 ext4_msg(sb, KERN_DEBUG,
2243 "%s: truncating inode %lu to %lld bytes",
2244 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2245 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2246 inode->i_ino, inode->i_size);
721e3eba 2247 mutex_lock(&inode->i_mutex);
55f252c9 2248 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2249 ext4_truncate(inode);
721e3eba 2250 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2251 nr_truncates++;
2252 } else {
566370a2
PT		 2253 if (test_opt(sb, DEBUG))
2254 ext4_msg(sb, KERN_DEBUG,
2255 "%s: deleting unreferenced inode %lu",
2256 __func__, inode->i_ino);
ac27a0ec
DK		 2257 jbd_debug(2, "deleting unreferenced inode %lu\n",
2258 inode->i_ino);
2259 nr_orphans++;
2260 }
2261 iput(inode); /* The delete magic happens here! */
2262 }
2263
2b2d6d01 2264#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2265
2266 if (nr_orphans)
b31e1552
ES		 2267 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2268 PLURAL(nr_orphans));
ac27a0ec 2269 if (nr_truncates)
b31e1552
ES		 2270 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2271 PLURAL(nr_truncates));
ac27a0ec
DK		 2272#ifdef CONFIG_QUOTA
2273 /* Turn quotas off */
a2d4a646 2274 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2275 if (sb_dqopt(sb)->files[i])
287a8095 2276 dquot_quota_off(sb, i);
ac27a0ec
DK		 2277 }
2278#endif
2279 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2280}
0b8e58a1 2281
cd2291a4
ES		 2282/*
2283 * Maximal extent format file size.
2284 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2285 * extent format containers, within a sector_t, and within i_blocks
2286 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2287 * so that won't be a limiting factor.
2288 *
f17722f9
LC		 2289 * However there is other limiting factor. We do store extents in the form
2290 * of starting block and length, hence the resulting length of the extent
2291 * covering maximum file size must fit into on-disk format containers as
2292 * well. Given that length is always by 1 unit bigger than max unit (because
2293 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2294 *
cd2291a4
ES		 2295 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2296 */
f287a1a5 2297static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2298{
2299 loff_t res;
2300 loff_t upper_limit = MAX_LFS_FILESIZE;
2301
2302 /* small i_blocks in vfs inode? */
f287a1a5 2303 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2304 /*
90c699a9 2305 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2306 * i_block represent total blocks in 512 bytes
2307 * 32 == size of vfs inode i_blocks * 8
2308 */
2309 upper_limit = (1LL << 32) - 1;
2310
2311 /* total blocks in file system block size */
2312 upper_limit >>= (blkbits - 9);
2313 upper_limit <<= blkbits;
2314 }
2315
f17722f9
LC		 2316 /*
2317 * 32-bit extent-start container, ee_block. We lower the maxbytes
2318 * by one fs block, so ee_len can cover the extent of maximum file
2319 * size
2320 */
2321 res = (1LL << 32) - 1;
cd2291a4 2322 res <<= blkbits;
cd2291a4
ES		 2323
2324 /* Sanity check against vm- & vfs- imposed limits */
2325 if (res > upper_limit)
2326 res = upper_limit;
2327
2328 return res;
2329}
ac27a0ec 2330
ac27a0ec 2331/*
cd2291a4 2332 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2333 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2334 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2335 */
f287a1a5 2336static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2337{
617ba13b 2338 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2339 int meta_blocks;
2340 loff_t upper_limit;
0b8e58a1
AD		 2341 /* This is calculated to be the largest file size for a dense, block
2342 * mapped file such that the file's total number of 512-byte sectors,
2343 * including data and all indirect blocks, does not exceed (2^48 - 1).
2344 *
2345 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2346 * number of 512-byte sectors of the file.
0fc1b451
AK		 2347 */
2348
f287a1a5 2349 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2350 /*
90c699a9 2351 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2352 * the inode i_block field represents total file blocks in
2353 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2354 */
2355 upper_limit = (1LL << 32) - 1;
2356
2357 /* total blocks in file system block size */
2358 upper_limit >>= (bits - 9);
2359
2360 } else {
8180a562
AK		 2361 /*
2362 * We use 48 bit ext4_inode i_blocks
2363 * With EXT4_HUGE_FILE_FL set the i_blocks
2364 * represent total number of blocks in
2365 * file system block size
2366 */
0fc1b451
AK		 2367 upper_limit = (1LL << 48) - 1;
2368
0fc1b451
AK		 2369 }
2370
2371 /* indirect blocks */
2372 meta_blocks = 1;
2373 /* double indirect blocks */
2374 meta_blocks += 1 + (1LL << (bits-2));
2375 /* tripple indirect blocks */
2376 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2377
2378 upper_limit -= meta_blocks;
2379 upper_limit <<= bits;
ac27a0ec
DK		 2380
2381 res += 1LL << (bits-2);
2382 res += 1LL << (2*(bits-2));
2383 res += 1LL << (3*(bits-2));
2384 res <<= bits;
2385 if (res > upper_limit)
2386 res = upper_limit;
0fc1b451
AK		 2387
2388 if (res > MAX_LFS_FILESIZE)
2389 res = MAX_LFS_FILESIZE;
2390
ac27a0ec
DK		 2391 return res;
2392}
2393
617ba13b 2394static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2395 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2396{
617ba13b 2397 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2398 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2399 int has_super = 0;
2400
2401 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2402
617ba13b 2403 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2404 nr < first_meta_bg)
70bbb3e0 2405 return logical_sb_block + nr + 1;
ac27a0ec 2406 bg = sbi->s_desc_per_block * nr;
617ba13b 2407 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2408 has_super = 1;
0b8e58a1 2409
bd63f6b0
DW		 2410 /*
2411 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2412 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2413 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2414 * compensate.
2415 */
2416 if (sb->s_blocksize == 1024 && nr == 0 &&
2417 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2418 has_super++;
2419
617ba13b 2420 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2421}
2422
c9de560d
AT		 2423/**
2424 * ext4_get_stripe_size: Get the stripe size.
2425 * @sbi: In memory super block info
2426 *
2427 * If we have specified it via mount option, then
2428 * use the mount option value. If the value specified at mount time is
2429 * greater than the blocks per group use the super block value.
2430 * If the super block value is greater than blocks per group return 0.
2431 * Allocator needs it be less than blocks per group.
2432 *
2433 */
2434static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2435{
2436 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2437 unsigned long stripe_width =
2438 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2439 int ret;
c9de560d
AT		 2440
2441 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2442 ret = sbi->s_stripe;
2443 else if (stripe_width <= sbi->s_blocks_per_group)
2444 ret = stripe_width;
2445 else if (stride <= sbi->s_blocks_per_group)
2446 ret = stride;
2447 else
2448 ret = 0;
c9de560d 2449
3eb08658
DE		 2450 /*
2451 * If the stripe width is 1, this makes no sense and
2452 * we set it to 0 to turn off stripe handling code.
2453 */
2454 if (ret <= 1)
2455 ret = 0;
c9de560d 2456
3eb08658 2457 return ret;
c9de560d 2458}
ac27a0ec 2459
3197ebdb
TT		 2460/* sysfs supprt */
2461
2462struct ext4_attr {
2463 struct attribute attr;
2464 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2465 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb 2466 const char *, size_t);
f2d50a65
JK		 2467 union {
2468 int offset;
2469 int deprecated_val;
2470 } u;
3197ebdb
TT		 2471};
2472
27dd4385
LC		 2473static int parse_strtoull(const char *buf,
2474 unsigned long long max, unsigned long long *value)
2475{
2476 int ret;
2477
2478 ret = kstrtoull(skip_spaces(buf), 0, value);
2479 if (!ret && *value > max)
2480 ret = -EINVAL;
2481 return ret;
2482}
2483
3197ebdb
TT		 2484static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2485 struct ext4_sb_info *sbi,
2486 char *buf)
2487{
2488 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2489 (s64) EXT4_C2B(sbi,
2490 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2491}
2492
2493static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2494 struct ext4_sb_info *sbi, char *buf)
2495{
2496 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2497
f613dfcb
TT		 2498 if (!sb->s_bdev->bd_part)
2499 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2500 return snprintf(buf, PAGE_SIZE, "%lu\n",
2501 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2502 sbi->s_sectors_written_start) >> 1);
2503}
2504
2505static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2506 struct ext4_sb_info *sbi, char *buf)
2507{
2508 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2509
f613dfcb
TT		 2510 if (!sb->s_bdev->bd_part)
2511 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2512 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2513 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2514 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2515 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2516}
2517
2518static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2519 struct ext4_sb_info *sbi,
2520 const char *buf, size_t count)
2521{
2522 unsigned long t;
e1091b15 2523 int ret;
3197ebdb 2524
e1091b15
LC		 2525 ret = kstrtoul(skip_spaces(buf), 0, &t);
2526 if (ret)
2527 return ret;
3197ebdb 2528
e1091b15 2529 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2530 return -EINVAL;
2531
2532 sbi->s_inode_readahead_blks = t;
2533 return count;
2534}
2535
2536static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2537 struct ext4_sb_info *sbi, char *buf)
3197ebdb 2538{
f2d50a65 2539 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb
TT		 2540
2541 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2542}
2543
2544static ssize_t sbi_ui_store(struct ext4_attr *a,
2545 struct ext4_sb_info *sbi,
2546 const char *buf, size_t count)
2547{
f2d50a65 2548 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb 2549 unsigned long t;
e1091b15 2550 int ret;
3197ebdb 2551
e1091b15
LC		 2552 ret = kstrtoul(skip_spaces(buf), 0, &t);
2553 if (ret)
2554 return ret;
3197ebdb
TT		 2555 *ui = t;
2556 return count;
2557}
2558
52c198c6
LC		 2559static ssize_t es_ui_show(struct ext4_attr *a,
2560 struct ext4_sb_info *sbi, char *buf)
2561{
2562
2563 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2564 a->u.offset);
2565
2566 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2567}
2568
27dd4385
LC		 2569static ssize_t reserved_clusters_show(struct ext4_attr *a,
2570 struct ext4_sb_info *sbi, char *buf)
2571{
2572 return snprintf(buf, PAGE_SIZE, "%llu\n",
2573 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2574}
2575
2576static ssize_t reserved_clusters_store(struct ext4_attr *a,
2577 struct ext4_sb_info *sbi,
2578 const char *buf, size_t count)
2579{
2580 unsigned long long val;
2581 int ret;
2582
2583 if (parse_strtoull(buf, -1ULL, &val))
2584 return -EINVAL;
2585 ret = ext4_reserve_clusters(sbi, val);
2586
2587 return ret ? ret : count;
2588}
2589
2c0544b2
TT		 2590static ssize_t trigger_test_error(struct ext4_attr *a,
2591 struct ext4_sb_info *sbi,
2592 const char *buf, size_t count)
2593{
2594 int len = count;
2595
2596 if (!capable(CAP_SYS_ADMIN))
2597 return -EPERM;
2598
2599 if (len && buf[len-1] == '\n')
2600 len--;
2601
2602 if (len)
2603 ext4_error(sbi->s_sb, "%.*s", len, buf);
2604 return count;
2605}
2606
f2d50a65
JK		 2607static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2608 struct ext4_sb_info *sbi, char *buf)
2609{
2610 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2611}
2612
3197ebdb
TT		 2613#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2614static struct ext4_attr ext4_attr_##_name = { \
2615 .attr = {.name = __stringify(_name), .mode = _mode }, \
2616 .show = _show, \
2617 .store = _store, \
f2d50a65
JK		 2618 .u = { \
2619 .offset = offsetof(struct ext4_sb_info, _elname),\
2620 }, \
3197ebdb 2621}
52c198c6
LC		 2622
2623#define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2624static struct ext4_attr ext4_attr_##_name = { \
2625 .attr = {.name = __stringify(_name), .mode = _mode }, \
2626 .show = _show, \
2627 .store = _store, \
2628 .u = { \
2629 .offset = offsetof(struct ext4_super_block, _elname), \
2630 }, \
2631}
2632
3197ebdb
TT		 2633#define EXT4_ATTR(name, mode, show, store) \
2634static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2635
857ac889 2636#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2637#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2638#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
52c198c6
LC		 2639
2640#define EXT4_RO_ATTR_ES_UI(name, elname) \
2641 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
3197ebdb
TT		 2642#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2643 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
52c198c6 2644
3197ebdb 2645#define ATTR_LIST(name) &ext4_attr_##name.attr
f2d50a65
JK		 2646#define EXT4_DEPRECATED_ATTR(_name, _val) \
2647static struct ext4_attr ext4_attr_##_name = { \
2648 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2649 .show = sbi_deprecated_show, \
2650 .u = { \
2651 .deprecated_val = _val, \
2652 }, \
2653}
3197ebdb
TT		 2654
2655EXT4_RO_ATTR(delayed_allocation_blocks);
2656EXT4_RO_ATTR(session_write_kbytes);
2657EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2658EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2659EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2660 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2661EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2662EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2663EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2664EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2665EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2666EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2667EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
f2d50a65 2668EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
67a5da56 2669EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2670EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
efbed4dc
TT		 2671EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2672EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2673EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2674EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2675EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2676EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
52c198c6
LC		 2677EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2678EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2679EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
3197ebdb
TT		 2680
2681static struct attribute *ext4_attrs[] = {
2682 ATTR_LIST(delayed_allocation_blocks),
2683 ATTR_LIST(session_write_kbytes),
2684 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2685 ATTR_LIST(reserved_clusters),
3197ebdb 2686 ATTR_LIST(inode_readahead_blks),
11013911 2687 ATTR_LIST(inode_goal),
3197ebdb
TT		 2688 ATTR_LIST(mb_stats),
2689 ATTR_LIST(mb_max_to_scan),
2690 ATTR_LIST(mb_min_to_scan),
2691 ATTR_LIST(mb_order2_req),
2692 ATTR_LIST(mb_stream_req),
2693 ATTR_LIST(mb_group_prealloc),
55138e0b 2694 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2695 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2696 ATTR_LIST(trigger_fs_error),
efbed4dc
TT		 2697 ATTR_LIST(err_ratelimit_interval_ms),
2698 ATTR_LIST(err_ratelimit_burst),
2699 ATTR_LIST(warning_ratelimit_interval_ms),
2700 ATTR_LIST(warning_ratelimit_burst),
2701 ATTR_LIST(msg_ratelimit_interval_ms),
2702 ATTR_LIST(msg_ratelimit_burst),
52c198c6
LC		 2703 ATTR_LIST(errors_count),
2704 ATTR_LIST(first_error_time),
2705 ATTR_LIST(last_error_time),
3197ebdb
TT		 2706 NULL,
2707};
2708
857ac889
LC		 2709/* Features this copy of ext4 supports */
2710EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2711EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2712EXT4_INFO_ATTR(meta_bg_resize);
6ddb2447 2713EXT4_INFO_ATTR(encryption);
857ac889
LC		 2714
2715static struct attribute *ext4_feat_attrs[] = {
2716 ATTR_LIST(lazy_itable_init),
27ee40df 2717 ATTR_LIST(batched_discard),
5e7bbef1 2718 ATTR_LIST(meta_bg_resize),
6ddb2447 2719 ATTR_LIST(encryption),
857ac889
LC		 2720 NULL,
2721};
2722
3197ebdb
TT		 2723static ssize_t ext4_attr_show(struct kobject *kobj,
2724 struct attribute *attr, char *buf)
2725{
2726 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2727 s_kobj);
2728 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2729
2730 return a->show ? a->show(a, sbi, buf) : 0;
2731}
2732
2733static ssize_t ext4_attr_store(struct kobject *kobj,
2734 struct attribute *attr,
2735 const char *buf, size_t len)
2736{
2737 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2738 s_kobj);
2739 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2740
2741 return a->store ? a->store(a, sbi, buf, len) : 0;
2742}
2743
2744static void ext4_sb_release(struct kobject *kobj)
2745{
2746 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2747 s_kobj);
2748 complete(&sbi->s_kobj_unregister);
2749}
2750
52cf25d0 2751static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2752 .show = ext4_attr_show,
2753 .store = ext4_attr_store,
2754};
2755
2756static struct kobj_type ext4_ktype = {
2757 .default_attrs = ext4_attrs,
2758 .sysfs_ops = &ext4_attr_ops,
2759 .release = ext4_sb_release,
2760};
2761
857ac889
LC		 2762static void ext4_feat_release(struct kobject *kobj)
2763{
2764 complete(&ext4_feat->f_kobj_unregister);
2765}
2766
c7f72543
LC		 2767static ssize_t ext4_feat_show(struct kobject *kobj,
2768 struct attribute *attr, char *buf)
2769{
2770 return snprintf(buf, PAGE_SIZE, "supported\n");
2771}
2772
2773/*
2774 * We can not use ext4_attr_show/store because it relies on the kobject
2775 * being embedded in the ext4_sb_info structure which is definitely not
2776 * true in this case.
2777 */
2778static const struct sysfs_ops ext4_feat_ops = {
2779 .show = ext4_feat_show,
2780 .store = NULL,
2781};
2782
857ac889
LC		 2783static struct kobj_type ext4_feat_ktype = {
2784 .default_attrs = ext4_feat_attrs,
c7f72543 2785 .sysfs_ops = &ext4_feat_ops,
857ac889
LC		 2786 .release = ext4_feat_release,
2787};
2788
a13fb1a4
ES		 2789/*
2790 * Check whether this filesystem can be mounted based on
2791 * the features present and the RDONLY/RDWR mount requested.
2792 * Returns 1 if this filesystem can be mounted as requested,
2793 * 0 if it cannot be.
2794 */
2795static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2796{
2797 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2798 ext4_msg(sb, KERN_ERR,
2799 "Couldn't mount because of "
2800 "unsupported optional features (%x)",
2801 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2802 ~EXT4_FEATURE_INCOMPAT_SUPP));
2803 return 0;
2804 }
2805
2806 if (readonly)
2807 return 1;
2808
2cb5cc8b
DW		 2809 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2810 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2811 sb->s_flags |= MS_RDONLY;
2812 return 1;
2813 }
2814
a13fb1a4
ES		 2815 /* Check that feature set is OK for a read-write mount */
2816 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2817 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2818 "unsupported optional features (%x)",
2819 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2820 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2821 return 0;
2822 }
2823 /*
2824 * Large file size enabled file system can only be mounted
2825 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2826 */
2827 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2828 if (sizeof(blkcnt_t) < sizeof(u64)) {
2829 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2830 "cannot be mounted RDWR without "
2831 "CONFIG_LBDAF");
2832 return 0;
2833 }
2834 }
bab08ab9
TT		 2835 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2836 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2837 ext4_msg(sb, KERN_ERR,
2838 "Can't support bigalloc feature without "
2839 "extents feature\n");
2840 return 0;
2841 }
7c319d32
AK		 2842
2843#ifndef CONFIG_QUOTA
2844 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2845 !readonly) {
2846 ext4_msg(sb, KERN_ERR,
2847 "Filesystem with quota feature cannot be mounted RDWR "
2848 "without CONFIG_QUOTA");
2849 return 0;
2850 }
2851#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2852 return 1;
2853}
2854
66e61a9e
TT		 2855/*
2856 * This function is called once a day if we have errors logged
2857 * on the file system
2858 */
2859static void print_daily_error_info(unsigned long arg)
2860{
2861 struct super_block *sb = (struct super_block *) arg;
2862 struct ext4_sb_info *sbi;
2863 struct ext4_super_block *es;
2864
2865 sbi = EXT4_SB(sb);
2866 es = sbi->s_es;
2867
2868 if (es->s_error_count)
ae0f78de
TT		 2869 /* fsck newer than v1.41.13 is needed to clean this condition. */
2870 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2871 le32_to_cpu(es->s_error_count));
2872 if (es->s_first_error_time) {
ae0f78de 2873 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2874 sb->s_id, le32_to_cpu(es->s_first_error_time),
2875 (int) sizeof(es->s_first_error_func),
2876 es->s_first_error_func,
2877 le32_to_cpu(es->s_first_error_line));
2878 if (es->s_first_error_ino)
2879 printk(": inode %u",
2880 le32_to_cpu(es->s_first_error_ino));
2881 if (es->s_first_error_block)
2882 printk(": block %llu", (unsigned long long)
2883 le64_to_cpu(es->s_first_error_block));
2884 printk("\n");
2885 }
2886 if (es->s_last_error_time) {
ae0f78de 2887 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2888 sb->s_id, le32_to_cpu(es->s_last_error_time),
2889 (int) sizeof(es->s_last_error_func),
2890 es->s_last_error_func,
2891 le32_to_cpu(es->s_last_error_line));
2892 if (es->s_last_error_ino)
2893 printk(": inode %u",
2894 le32_to_cpu(es->s_last_error_ino));
2895 if (es->s_last_error_block)
2896 printk(": block %llu", (unsigned long long)
2897 le64_to_cpu(es->s_last_error_block));
2898 printk("\n");
2899 }
2900 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2901}
2902
bfff6873
LC		 2903/* Find next suitable group and run ext4_init_inode_table */
2904static int ext4_run_li_request(struct ext4_li_request *elr)
2905{
2906 struct ext4_group_desc *gdp = NULL;
2907 ext4_group_t group, ngroups;
2908 struct super_block *sb;
2909 unsigned long timeout = 0;
2910 int ret = 0;
2911
2912 sb = elr->lr_super;
2913 ngroups = EXT4_SB(sb)->s_groups_count;
2914
8e8ad8a5 2915 sb_start_write(sb);
bfff6873
LC		 2916 for (group = elr->lr_next_group; group < ngroups; group++) {
2917 gdp = ext4_get_group_desc(sb, group, NULL);
2918 if (!gdp) {
2919 ret = 1;
2920 break;
2921 }
2922
2923 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2924 break;
2925 }
2926
7f511862 2927 if (group >= ngroups)
bfff6873
LC		 2928 ret = 1;
2929
2930 if (!ret) {
2931 timeout = jiffies;
2932 ret = ext4_init_inode_table(sb, group,
2933 elr->lr_timeout ? 0 : 1);
2934 if (elr->lr_timeout == 0) {
51ce6511
LC		 2935 timeout = (jiffies - timeout) *
2936 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2937 elr->lr_timeout = timeout;
2938 }
2939 elr->lr_next_sched = jiffies + elr->lr_timeout;
2940 elr->lr_next_group = group + 1;
2941 }
8e8ad8a5 2942 sb_end_write(sb);
bfff6873
LC		 2943
2944 return ret;
2945}
2946
2947/*
2948 * Remove lr_request from the list_request and free the
4ed5c033 2949 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2950 */
2951static void ext4_remove_li_request(struct ext4_li_request *elr)
2952{
2953 struct ext4_sb_info *sbi;
2954
2955 if (!elr)
2956 return;
2957
2958 sbi = elr->lr_sbi;
2959
2960 list_del(&elr->lr_request);
2961 sbi->s_li_request = NULL;
2962 kfree(elr);
2963}
2964
2965static void ext4_unregister_li_request(struct super_block *sb)
2966{
1bb933fb
LC		 2967 mutex_lock(&ext4_li_mtx);
2968 if (!ext4_li_info) {
2969 mutex_unlock(&ext4_li_mtx);
bfff6873 2970 return;
1bb933fb 2971 }
bfff6873
LC		 2972
2973 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2974 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2975 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2976 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2977}
2978
8f1f7453
ES		 2979static struct task_struct *ext4_lazyinit_task;
2980
bfff6873
LC		 2981/*
2982 * This is the function where ext4lazyinit thread lives. It walks
2983 * through the request list searching for next scheduled filesystem.
2984 * When such a fs is found, run the lazy initialization request
2985 * (ext4_rn_li_request) and keep track of the time spend in this
2986 * function. Based on that time we compute next schedule time of
2987 * the request. When walking through the list is complete, compute
2988 * next waking time and put itself into sleep.
2989 */
2990static int ext4_lazyinit_thread(void *arg)
2991{
2992 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2993 struct list_head *pos, *n;
2994 struct ext4_li_request *elr;
4ed5c033 2995 unsigned long next_wakeup, cur;
bfff6873
LC		 2996
2997 BUG_ON(NULL == eli);
2998
bfff6873
LC		 2999cont_thread:
3000 while (true) {
3001 next_wakeup = MAX_JIFFY_OFFSET;
3002
3003 mutex_lock(&eli->li_list_mtx);
3004 if (list_empty(&eli->li_request_list)) {
3005 mutex_unlock(&eli->li_list_mtx);
3006 goto exit_thread;
3007 }
3008
3009 list_for_each_safe(pos, n, &eli->li_request_list) {
3010 elr = list_entry(pos, struct ext4_li_request,
3011 lr_request);
3012
b2c78cd0
TT		 3013 if (time_after_eq(jiffies, elr->lr_next_sched)) {
3014 if (ext4_run_li_request(elr) != 0) {
3015 /* error, remove the lazy_init job */
3016 ext4_remove_li_request(elr);
3017 continue;
3018 }
bfff6873
LC		 3019 }
3020
3021 if (time_before(elr->lr_next_sched, next_wakeup))
3022 next_wakeup = elr->lr_next_sched;
3023 }
3024 mutex_unlock(&eli->li_list_mtx);
3025
a0acae0e 3026 try_to_freeze();
bfff6873 3027
4ed5c033
LC		 3028 cur = jiffies;
3029 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3030 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3031 cond_resched();
3032 continue;
3033 }
3034
4ed5c033
LC		 3035 schedule_timeout_interruptible(next_wakeup - cur);
3036
8f1f7453
ES		 3037 if (kthread_should_stop()) {
3038 ext4_clear_request_list();
3039 goto exit_thread;
3040 }
bfff6873
LC		 3041 }
3042
3043exit_thread:
3044 /*
3045 * It looks like the request list is empty, but we need
3046 * to check it under the li_list_mtx lock, to prevent any
3047 * additions into it, and of course we should lock ext4_li_mtx
3048 * to atomically free the list and ext4_li_info, because at
3049 * this point another ext4 filesystem could be registering
3050 * new one.
3051 */
3052 mutex_lock(&ext4_li_mtx);
3053 mutex_lock(&eli->li_list_mtx);
3054 if (!list_empty(&eli->li_request_list)) {
3055 mutex_unlock(&eli->li_list_mtx);
3056 mutex_unlock(&ext4_li_mtx);
3057 goto cont_thread;
3058 }
3059 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3060 kfree(ext4_li_info);
3061 ext4_li_info = NULL;
3062 mutex_unlock(&ext4_li_mtx);
3063
3064 return 0;
3065}
3066
3067static void ext4_clear_request_list(void)
3068{
3069 struct list_head *pos, *n;
3070 struct ext4_li_request *elr;
3071
3072 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3073 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3074 elr = list_entry(pos, struct ext4_li_request,
3075 lr_request);
3076 ext4_remove_li_request(elr);
3077 }
3078 mutex_unlock(&ext4_li_info->li_list_mtx);
3079}
3080
3081static int ext4_run_lazyinit_thread(void)
3082{
8f1f7453
ES		 3083 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3084 ext4_li_info, "ext4lazyinit");
3085 if (IS_ERR(ext4_lazyinit_task)) {
3086 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3087 ext4_clear_request_list();
bfff6873
LC		 3088 kfree(ext4_li_info);
3089 ext4_li_info = NULL;
92b97816 3090 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3091 "initialization thread\n",
3092 err);
3093 return err;
3094 }
3095 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3096 return 0;
3097}
3098
3099/*
3100 * Check whether it make sense to run itable init. thread or not.
3101 * If there is at least one uninitialized inode table, return
3102 * corresponding group number, else the loop goes through all
3103 * groups and return total number of groups.
3104 */
3105static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3106{
3107 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3108 struct ext4_group_desc *gdp = NULL;
3109
3110 for (group = 0; group < ngroups; group++) {
3111 gdp = ext4_get_group_desc(sb, group, NULL);
3112 if (!gdp)
3113 continue;
3114
3115 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3116 break;
3117 }
3118
3119 return group;
3120}
3121
3122static int ext4_li_info_new(void)
3123{
3124 struct ext4_lazy_init *eli = NULL;
3125
3126 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3127 if (!eli)
3128 return -ENOMEM;
3129
bfff6873
LC		 3130 INIT_LIST_HEAD(&eli->li_request_list);
3131 mutex_init(&eli->li_list_mtx);
3132
bfff6873
LC		 3133 eli->li_state |= EXT4_LAZYINIT_QUIT;
3134
3135 ext4_li_info = eli;
3136
3137 return 0;
3138}
3139
3140static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3141 ext4_group_t start)
3142{
3143 struct ext4_sb_info *sbi = EXT4_SB(sb);
3144 struct ext4_li_request *elr;
bfff6873
LC		 3145
3146 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3147 if (!elr)
3148 return NULL;
3149
3150 elr->lr_super = sb;
3151 elr->lr_sbi = sbi;
3152 elr->lr_next_group = start;
3153
3154 /*
3155 * Randomize first schedule time of the request to
3156 * spread the inode table initialization requests
3157 * better.
3158 */
dd1f723b
TT		 3159 elr->lr_next_sched = jiffies + (prandom_u32() %
3160 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3161 return elr;
3162}
3163
7f511862
TT		 3164int ext4_register_li_request(struct super_block *sb,
3165 ext4_group_t first_not_zeroed)
bfff6873
LC		 3166{
3167 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3168 struct ext4_li_request *elr = NULL;
bfff6873 3169 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3170 int ret = 0;
bfff6873 3171
7f511862 3172 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3173 if (sbi->s_li_request != NULL) {
3174 /*
3175 * Reset timeout so it can be computed again, because
3176 * s_li_wait_mult might have changed.
3177 */
3178 sbi->s_li_request->lr_timeout = 0;
7f511862 3179 goto out;
51ce6511 3180 }
bfff6873
LC		 3181
3182 if (first_not_zeroed == ngroups ||
3183 (sb->s_flags & MS_RDONLY) ||
55ff3840 3184 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3185 goto out;
bfff6873
LC		 3186
3187 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3188 if (!elr) {
3189 ret = -ENOMEM;
3190 goto out;
3191 }
bfff6873
LC		 3192
3193 if (NULL == ext4_li_info) {
3194 ret = ext4_li_info_new();
3195 if (ret)
3196 goto out;
3197 }
3198
3199 mutex_lock(&ext4_li_info->li_list_mtx);
3200 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3201 mutex_unlock(&ext4_li_info->li_list_mtx);
3202
3203 sbi->s_li_request = elr;
46e4690b
TM		 3204 /*
3205 * set elr to NULL here since it has been inserted to
3206 * the request_list and the removal and free of it is
3207 * handled by ext4_clear_request_list from now on.
3208 */
3209 elr = NULL;
bfff6873
LC		 3210
3211 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3212 ret = ext4_run_lazyinit_thread();
3213 if (ret)
3214 goto out;
3215 }
bfff6873 3216out:
beed5ecb
NK		 3217 mutex_unlock(&ext4_li_mtx);
3218 if (ret)
bfff6873 3219 kfree(elr);
bfff6873
LC		 3220 return ret;
3221}
3222
3223/*
3224 * We do not need to lock anything since this is called on
3225 * module unload.
3226 */
3227static void ext4_destroy_lazyinit_thread(void)
3228{
3229 /*
3230 * If thread exited earlier
3231 * there's nothing to be done.
3232 */
8f1f7453 3233 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3234 return;
3235
8f1f7453 3236 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3237}
3238
25ed6e8a
DW		 3239static int set_journal_csum_feature_set(struct super_block *sb)
3240{
3241 int ret = 1;
3242 int compat, incompat;
3243 struct ext4_sb_info *sbi = EXT4_SB(sb);
3244
9aa5d32b 3245 if (ext4_has_metadata_csum(sb)) {
db9ee220 3246 /* journal checksum v3 */
25ed6e8a 3247 compat = 0;
db9ee220 3248 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3249 } else {
3250 /* journal checksum v1 */
3251 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3252 incompat = 0;
3253 }
3254
feb8c6d3
DW		 3255 jbd2_journal_clear_features(sbi->s_journal,
3256 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3257 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3258 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3259 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3260 ret = jbd2_journal_set_features(sbi->s_journal,
3261 compat, 0,
3262 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3263 incompat);
3264 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3265 ret = jbd2_journal_set_features(sbi->s_journal,
3266 compat, 0,
3267 incompat);
3268 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3269 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3270 } else {
feb8c6d3
DW		 3271 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3272 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3273 }
3274
3275 return ret;
3276}
3277
952fc18e
TT		 3278/*
3279 * Note: calculating the overhead so we can be compatible with
3280 * historical BSD practice is quite difficult in the face of
3281 * clusters/bigalloc. This is because multiple metadata blocks from
3282 * different block group can end up in the same allocation cluster.
3283 * Calculating the exact overhead in the face of clustered allocation
3284 * requires either O(all block bitmaps) in memory or O(number of block
3285 * groups**2) in time. We will still calculate the superblock for
3286 * older file systems --- and if we come across with a bigalloc file
3287 * system with zero in s_overhead_clusters the estimate will be close to
3288 * correct especially for very large cluster sizes --- but for newer
3289 * file systems, it's better to calculate this figure once at mkfs
3290 * time, and store it in the superblock. If the superblock value is
3291 * present (even for non-bigalloc file systems), we will use it.
3292 */
3293static int count_overhead(struct super_block *sb, ext4_group_t grp,
3294 char *buf)
3295{
3296 struct ext4_sb_info *sbi = EXT4_SB(sb);
3297 struct ext4_group_desc *gdp;
3298 ext4_fsblk_t first_block, last_block, b;
3299 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3300 int s, j, count = 0;
3301
0548bbb8
TT		 3302 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3303 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3304 sbi->s_itb_per_group + 2);
3305
952fc18e
TT		 3306 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3307 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3308 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3309 for (i = 0; i < ngroups; i++) {
3310 gdp = ext4_get_group_desc(sb, i, NULL);
3311 b = ext4_block_bitmap(sb, gdp);
3312 if (b >= first_block && b <= last_block) {
3313 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3314 count++;
3315 }
3316 b = ext4_inode_bitmap(sb, gdp);
3317 if (b >= first_block && b <= last_block) {
3318 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3319 count++;
3320 }
3321 b = ext4_inode_table(sb, gdp);
3322 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3323 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3324 int c = EXT4_B2C(sbi, b - first_block);
3325 ext4_set_bit(c, buf);
3326 count++;
3327 }
3328 if (i != grp)
3329 continue;
3330 s = 0;
3331 if (ext4_bg_has_super(sb, grp)) {
3332 ext4_set_bit(s++, buf);
3333 count++;
3334 }
3335 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3336 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3337 count++;
3338 }
3339 }
3340 if (!count)
3341 return 0;
3342 return EXT4_CLUSTERS_PER_GROUP(sb) -
3343 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3344}
3345
3346/*
3347 * Compute the overhead and stash it in sbi->s_overhead
3348 */
3349int ext4_calculate_overhead(struct super_block *sb)
3350{
3351 struct ext4_sb_info *sbi = EXT4_SB(sb);
3352 struct ext4_super_block *es = sbi->s_es;
3353 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3354 ext4_fsblk_t overhead = 0;
4fdb5543 3355 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3356
952fc18e
TT		 3357 if (!buf)
3358 return -ENOMEM;
3359
3360 /*
3361 * Compute the overhead (FS structures). This is constant
3362 * for a given filesystem unless the number of block groups
3363 * changes so we cache the previous value until it does.
3364 */
3365
3366 /*
3367 * All of the blocks before first_data_block are overhead
3368 */
3369 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3370
3371 /*
3372 * Add the overhead found in each block group
3373 */
3374 for (i = 0; i < ngroups; i++) {
3375 int blks;
3376
3377 blks = count_overhead(sb, i, buf);
3378 overhead += blks;
3379 if (blks)
3380 memset(buf, 0, PAGE_SIZE);
3381 cond_resched();
3382 }
b003b524
ES		 3383 /* Add the internal journal blocks as well */
3384 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3385 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3386
952fc18e
TT		 3387 sbi->s_overhead = overhead;
3388 smp_wmb();
3389 free_page((unsigned long) buf);
3390 return 0;
3391}
3392
27dd4385 3393
30fac0f7 3394static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
27dd4385
LC		 3395{
3396 ext4_fsblk_t resv_clusters;
3397
30fac0f7
JK		 3398 /*
3399 * There's no need to reserve anything when we aren't using extents.
3400 * The space estimates are exact, there are no unwritten extents,
3401 * hole punching doesn't need new metadata... This is needed especially
3402 * to keep ext2/3 backward compatibility.
3403 */
3404 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3405 return 0;
27dd4385
LC		 3406 /*
3407 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3408 * This should cover the situations where we can not afford to run
3409 * out of space like for example punch hole, or converting
556615dc 3410 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3411 * allocation would require 1, or 2 blocks, higher numbers are
3412 * very rare.
3413 */
30fac0f7
JK		 3414 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3415 EXT4_SB(sb)->s_cluster_bits;
27dd4385
LC		 3416
3417 do_div(resv_clusters, 50);
3418 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3419
3420 return resv_clusters;
3421}
3422
3423
3424static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3425{
3426 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3427 sbi->s_cluster_bits;
3428
3429 if (count >= clusters)
3430 return -EINVAL;
3431
3432 atomic64_set(&sbi->s_resv_clusters, count);
3433 return 0;
3434}
3435
2b2d6d01 3436static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3437{
d4c402d9 3438 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3439 struct buffer_head *bh;
617ba13b
MC		 3440 struct ext4_super_block *es = NULL;
3441 struct ext4_sb_info *sbi;
3442 ext4_fsblk_t block;
3443 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3444 ext4_fsblk_t logical_sb_block;
ac27a0ec 3445 unsigned long offset = 0;
ac27a0ec
DK		 3446 unsigned long journal_devnum = 0;
3447 unsigned long def_mount_opts;
3448 struct inode *root;
0390131b 3449 const char *descr;
dcc7dae3 3450 int ret = -ENOMEM;
281b5995 3451 int blocksize, clustersize;
4ec11028
TT		 3452 unsigned int db_count;
3453 unsigned int i;
281b5995 3454 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3455 __u64 blocks_count;
07aa2ea1 3456 int err = 0;
b3881f74 3457 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3458 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3459
3460 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3461 if (!sbi)
dcc7dae3 3462 goto out_free_orig;
705895b6
PE		 3463
3464 sbi->s_blockgroup_lock =
3465 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3466 if (!sbi->s_blockgroup_lock) {
3467 kfree(sbi);
dcc7dae3 3468 goto out_free_orig;
705895b6 3469 }
ac27a0ec 3470 sb->s_fs_info = sbi;
2c0544b2 3471 sbi->s_sb = sb;
240799cd 3472 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3473 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3474 if (sb->s_bdev->bd_part)
3475 sbi->s_sectors_written_start =
3476 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3477
9f6200bb 3478 /* Cleanup superblock name */
ec3904dc 3479 strreplace(sb->s_id, '/', '!');
9f6200bb 3480
07aa2ea1 3481 /* -EINVAL is default */
dcc7dae3 3482 ret = -EINVAL;
617ba13b 3483 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3484 if (!blocksize) {
b31e1552 3485 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3486 goto out_fail;
3487 }
3488
3489 /*
617ba13b 3490 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3491 * block sizes. We need to calculate the offset from buffer start.
3492 */
617ba13b 3493 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3494 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3495 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3496 } else {
70bbb3e0 3497 logical_sb_block = sb_block;
ac27a0ec
DK		 3498 }
3499
a8ac900b 3500 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3501 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3502 goto out_fail;
3503 }
3504 /*
3505 * Note: s_es must be initialized as soon as possible because
617ba13b 3506 * some ext4 macro-instructions depend on its value
ac27a0ec 3507 */
2716b802 3508 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3509 sbi->s_es = es;
3510 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3511 if (sb->s_magic != EXT4_SUPER_MAGIC)
3512 goto cantfind_ext4;
afc32f7e 3513 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3514
feb0ab32
DW		 3515 /* Warn if metadata_csum and gdt_csum are both set. */
3516 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3517 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3518 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
363307e6 3519 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3520 "redundant flags; please run fsck.");
3521
d25425f8
DW		 3522 /* Check for a known checksum algorithm */
3523 if (!ext4_verify_csum_type(sb, es)) {
3524 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3525 "unknown checksum algorithm.");
3526 silent = 1;
3527 goto cantfind_ext4;
3528 }
3529
0441984a
DW		 3530 /* Load the checksum driver */
3531 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3532 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3533 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3534 if (IS_ERR(sbi->s_chksum_driver)) {
3535 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3536 ret = PTR_ERR(sbi->s_chksum_driver);
3537 sbi->s_chksum_driver = NULL;
3538 goto failed_mount;
3539 }
3540 }
3541
a9c47317
DW		 3542 /* Check superblock checksum */
3543 if (!ext4_superblock_csum_verify(sb, es)) {
3544 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3545 "invalid superblock checksum. Run e2fsck?");
3546 silent = 1;
3547 goto cantfind_ext4;
3548 }
3549
3550 /* Precompute checksum seed for all metadata */
9aa5d32b 3551 if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3552 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3553 sizeof(es->s_uuid));
3554
ac27a0ec
DK		 3555 /* Set defaults before we parse the mount options */
3556 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3557 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3558 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3559 set_opt(sb, DEBUG);
87f26807 3560 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3561 set_opt(sb, GRPID);
617ba13b 3562 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3563 set_opt(sb, NO_UID32);
ea663336 3564 /* xattr user namespace & acls are now defaulted on */
ea663336 3565 set_opt(sb, XATTR_USER);
03010a33 3566#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3567 set_opt(sb, POSIX_ACL);
2e7842b8 3568#endif
98c1a759
DW		 3569 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3570 if (ext4_has_metadata_csum(sb))
3571 set_opt(sb, JOURNAL_CHECKSUM);
3572
617ba13b 3573 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3574 set_opt(sb, JOURNAL_DATA);
617ba13b 3575 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3576 set_opt(sb, ORDERED_DATA);
617ba13b 3577 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3578 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3579
3580 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3581 set_opt(sb, ERRORS_PANIC);
bb4f397a 3582 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3583 set_opt(sb, ERRORS_CONT);
bb4f397a 3584 else
fd8c37ec 3585 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3586 /* block_validity enabled by default; disable with noblock_validity */
3587 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3588 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3589 set_opt(sb, DISCARD);
ac27a0ec 3590
08cefc7a
EB		 3591 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3592 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3593 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3594 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3595 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3596
8b67f04a 3597 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3598 set_opt(sb, BARRIER);
ac27a0ec 3599
dd919b98
AK		 3600 /*
3601 * enable delayed allocation by default
3602 * Use -o nodelalloc to turn it off
3603 */
bc0b75f7 3604 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3605 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3606 set_opt(sb, DELALLOC);
dd919b98 3607
51ce6511
LC		 3608 /*
3609 * set default s_li_wait_mult for lazyinit, for the case there is
3610 * no mount option specified.
3611 */
3612 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3613
8b67f04a 3614 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3615 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3616 ext4_msg(sb, KERN_WARNING,
3617 "failed to parse options in superblock: %s",
3618 sbi->s_es->s_mount_opts);
3619 }
5a916be1 3620 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3621 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3622 &journal_ioprio, 0))
ac27a0ec
DK		 3623 goto failed_mount;
3624
56889787
TT		 3625 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3626 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3627 "with data=journal disables delayed "
3628 "allocation and O_DIRECT support!\n");
3629 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3630 ext4_msg(sb, KERN_ERR, "can't mount with "
3631 "both data=journal and delalloc");
3632 goto failed_mount;
3633 }
3634 if (test_opt(sb, DIOREAD_NOLOCK)) {
3635 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3636 "both data=journal and dioread_nolock");
56889787
TT		 3637 goto failed_mount;
3638 }
923ae0ff
RZ		 3639 if (test_opt(sb, DAX)) {
3640 ext4_msg(sb, KERN_ERR, "can't mount with "
3641 "both data=journal and dax");
3642 goto failed_mount;
3643 }
56889787
TT		 3644 if (test_opt(sb, DELALLOC))
3645 clear_opt(sb, DELALLOC);
001e4a87
TH		 3646 } else {
3647 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3648 }
3649
ac27a0ec 3650 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3651 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3652
617ba13b
MC		 3653 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3654 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3655 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3656 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3657 ext4_msg(sb, KERN_WARNING,
3658 "feature flags set on rev 0 fs, "
3659 "running e2fsck is recommended");
469108ff 3660
ed3654eb
TT		 3661 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3662 set_opt2(sb, HURD_COMPAT);
3663 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3664 EXT4_FEATURE_INCOMPAT_64BIT)) {
3665 ext4_msg(sb, KERN_ERR,
3666 "The Hurd can't support 64-bit file systems");
3667 goto failed_mount;
3668 }
3669 }
3670
2035e776
TT		 3671 if (IS_EXT2_SB(sb)) {
3672 if (ext2_feature_set_ok(sb))
3673 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3674 "using the ext4 subsystem");
3675 else {
3676 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3677 "to feature incompatibilities");
3678 goto failed_mount;
3679 }
3680 }
3681
3682 if (IS_EXT3_SB(sb)) {
3683 if (ext3_feature_set_ok(sb))
3684 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3685 "using the ext4 subsystem");
3686 else {
3687 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3688 "to feature incompatibilities");
3689 goto failed_mount;
3690 }
3691 }
3692
ac27a0ec
DK		 3693 /*
3694 * Check feature flags regardless of the revision level, since we
3695 * previously didn't change the revision level when setting the flags,
3696 * so there is a chance incompat flags are set on a rev 0 filesystem.
3697 */
a13fb1a4 3698 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3699 goto failed_mount;
a13fb1a4 3700
261cb20c 3701 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3702 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3703 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3704 ext4_msg(sb, KERN_ERR,
3705 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3706 goto failed_mount;
3707 }
3708
923ae0ff
RZ		 3709 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3710 if (blocksize != PAGE_SIZE) {
3711 ext4_msg(sb, KERN_ERR,
3712 "error: unsupported blocksize for dax");
3713 goto failed_mount;
3714 }
3715 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3716 ext4_msg(sb, KERN_ERR,
3717 "error: device does not support dax");
3718 goto failed_mount;
3719 }
3720 }
3721
6ddb2447
TT		 3722 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3723 es->s_encryption_level) {
3724 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3725 es->s_encryption_level);
3726 goto failed_mount;
3727 }
3728
ac27a0ec 3729 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3730 /* Validate the filesystem blocksize */
3731 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3732 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3733 blocksize);
ac27a0ec
DK		 3734 goto failed_mount;
3735 }
3736
2b2d6d01 3737 brelse(bh);
70bbb3e0
AM		 3738 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3739 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3740 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3741 if (!bh) {
b31e1552
ES		 3742 ext4_msg(sb, KERN_ERR,
3743 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3744 goto failed_mount;
3745 }
2716b802 3746 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3747 sbi->s_es = es;
617ba13b 3748 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3749 ext4_msg(sb, KERN_ERR,
3750 "Magic mismatch, very weird!");
ac27a0ec
DK		 3751 goto failed_mount;
3752 }
3753 }
3754
a13fb1a4
ES		 3755 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3756 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3757 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3758 has_huge_files);
3759 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3760
617ba13b
MC		 3761 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3762 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3763 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3764 } else {
3765 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3766 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3767 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3768 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3769 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3770 ext4_msg(sb, KERN_ERR,
3771 "unsupported inode size: %d",
2b2d6d01 3772 sbi->s_inode_size);
ac27a0ec
DK		 3773 goto failed_mount;
3774 }
ef7f3835
KS		 3775 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3776 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3777 }
0b8e58a1 3778
0d1ee42f
AR		 3779 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3780 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3781 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3782 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3783 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3784 ext4_msg(sb, KERN_ERR,
3785 "unsupported descriptor size %lu",
0d1ee42f
AR		 3786 sbi->s_desc_size);
3787 goto failed_mount;
3788 }
3789 } else
3790 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3791
ac27a0ec 3792 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3793 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3794 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3795 goto cantfind_ext4;
0b8e58a1 3796
617ba13b 3797 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3798 if (sbi->s_inodes_per_block == 0)
617ba13b 3799 goto cantfind_ext4;
ac27a0ec
DK		 3800 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3801 sbi->s_inodes_per_block;
0d1ee42f 3802 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3803 sbi->s_sbh = bh;
3804 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3805 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3806 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3807
2b2d6d01 3808 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3809 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3810 sbi->s_def_hash_version = es->s_def_hash_version;
23301410
TT		 3811 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3812 i = le32_to_cpu(es->s_flags);
3813 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3814 sbi->s_hash_unsigned = 3;
3815 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3816#ifdef __CHAR_UNSIGNED__
23301410
TT		 3817 if (!(sb->s_flags & MS_RDONLY))
3818 es->s_flags |=
3819 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3820 sbi->s_hash_unsigned = 3;
f99b2589 3821#else
23301410
TT		 3822 if (!(sb->s_flags & MS_RDONLY))
3823 es->s_flags |=
3824 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3825#endif
23301410 3826 }
f99b2589 3827 }
ac27a0ec 3828
281b5995
TT		 3829 /* Handle clustersize */
3830 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3831 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3832 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3833 if (has_bigalloc) {
3834 if (clustersize < blocksize) {
3835 ext4_msg(sb, KERN_ERR,
3836 "cluster size (%d) smaller than "
3837 "block size (%d)", clustersize, blocksize);
3838 goto failed_mount;
3839 }
3840 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3841 le32_to_cpu(es->s_log_block_size);
3842 sbi->s_clusters_per_group =
3843 le32_to_cpu(es->s_clusters_per_group);
3844 if (sbi->s_clusters_per_group > blocksize * 8) {
3845 ext4_msg(sb, KERN_ERR,
3846 "#clusters per group too big: %lu",
3847 sbi->s_clusters_per_group);
3848 goto failed_mount;
3849 }
3850 if (sbi->s_blocks_per_group !=
3851 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3852 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3853 "clusters per group (%lu) inconsistent",
3854 sbi->s_blocks_per_group,
3855 sbi->s_clusters_per_group);
3856 goto failed_mount;
3857 }
3858 } else {
3859 if (clustersize != blocksize) {
3860 ext4_warning(sb, "fragment/cluster size (%d) != "
3861 "block size (%d)", clustersize,
3862 blocksize);
3863 clustersize = blocksize;
3864 }
3865 if (sbi->s_blocks_per_group > blocksize * 8) {
3866 ext4_msg(sb, KERN_ERR,
3867 "#blocks per group too big: %lu",
3868 sbi->s_blocks_per_group);
3869 goto failed_mount;
3870 }
3871 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3872 sbi->s_cluster_bits = 0;
ac27a0ec 3873 }
281b5995
TT		 3874 sbi->s_cluster_ratio = clustersize / blocksize;
3875
ac27a0ec 3876 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3877 ext4_msg(sb, KERN_ERR,
3878 "#inodes per group too big: %lu",
2b2d6d01 3879 sbi->s_inodes_per_group);
ac27a0ec
DK		 3880 goto failed_mount;
3881 }
3882
960fd856
TT		 3883 /* Do we have standard group size of clustersize * 8 blocks ? */
3884 if (sbi->s_blocks_per_group == clustersize << 3)
3885 set_opt2(sb, STD_GROUP_SIZE);
3886
bf43d84b
ES		 3887 /*
3888 * Test whether we have more sectors than will fit in sector_t,
3889 * and whether the max offset is addressable by the page cache.
3890 */
5a9ae68a 3891 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3892 ext4_blocks_count(es));
5a9ae68a 3893 if (err) {
b31e1552 3894 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3895 " too large to mount safely on this system");
ac27a0ec 3896 if (sizeof(sector_t) < 8)
90c699a9 3897 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3898 goto failed_mount;
3899 }
3900
617ba13b
MC		 3901 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3902 goto cantfind_ext4;
e7c95593 3903
0f2ddca6
FTN		 3904 /* check blocks count against device size */
3905 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3906 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3907 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3908 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3909 ext4_blocks_count(es), blocks_count);
3910 goto failed_mount;
3911 }
3912
0b8e58a1
AD		 3913 /*
3914 * It makes no sense for the first data block to be beyond the end
3915 * of the filesystem.
3916 */
3917 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3918 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3919 "block %u is beyond end of filesystem (%llu)",
3920 le32_to_cpu(es->s_first_data_block),
3921 ext4_blocks_count(es));
e7c95593
ES		 3922 goto failed_mount;
3923 }
bd81d8ee
LV		 3924 blocks_count = (ext4_blocks_count(es) -
3925 le32_to_cpu(es->s_first_data_block) +
3926 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3927 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3928 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3929 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3930 "(block count %llu, first data block %u, "
b31e1552 3931 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3932 ext4_blocks_count(es),
3933 le32_to_cpu(es->s_first_data_block),
3934 EXT4_BLOCKS_PER_GROUP(sb));
3935 goto failed_mount;
3936 }
bd81d8ee 3937 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3938 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3939 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3940 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3941 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3942 sbi->s_group_desc = ext4_kvmalloc(db_count *
3943 sizeof(struct buffer_head *),
3944 GFP_KERNEL);
ac27a0ec 3945 if (sbi->s_group_desc == NULL) {
b31e1552 3946 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3947 ret = -ENOMEM;
ac27a0ec
DK		 3948 goto failed_mount;
3949 }
3950
9f6200bb
TT		 3951 if (ext4_proc_root)
3952 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3953
66acdcf4
TT		 3954 if (sbi->s_proc)
3955 proc_create_data("options", S_IRUGO, sbi->s_proc,
3956 &ext4_seq_options_fops, sb);
3957
705895b6 3958 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3959
3960 for (i = 0; i < db_count; i++) {
70bbb3e0 3961 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3962 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3963 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3964 ext4_msg(sb, KERN_ERR,
3965 "can't read group descriptor %d", i);
ac27a0ec
DK		 3966 db_count = i;
3967 goto failed_mount2;
3968 }
3969 }
bfff6873 3970 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3971 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
f9ae9cf5 3972 goto failed_mount2;
ac27a0ec 3973 }
772cb7c8 3974
f9ae9cf5 3975 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3976 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3977 spin_lock_init(&sbi->s_next_gen_lock);
3978
04ecddb7
JM		 3979 setup_timer(&sbi->s_err_report, print_daily_error_info,
3980 (unsigned long) sb);
04496411 3981
a75ae78f 3982 /* Register extent status tree shrinker */
eb68d0e2 3983 if (ext4_es_register_shrinker(sbi))
ce7e010a 3984 goto failed_mount3;
ce7e010a 3985
c9de560d 3986 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3987 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3988
f9ae9cf5
TT		 3989 /*
3990 * set up enough so that it can read an inode
3991 */
f6e63f90 3992 sb->s_op = &ext4_sops;
617ba13b
MC		 3993 sb->s_export_op = &ext4_export_ops;
3994 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3995#ifdef CONFIG_QUOTA
617ba13b 3996 sb->dq_op = &ext4_quota_operations;
262b4662 3997 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
1fa5efe3 3998 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3999 else
4000 sb->s_qcop = &ext4_qctl_operations;
96c7e0d9 4001 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
ac27a0ec 4002#endif
f2fa2ffc
AK		 4003 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4004
ac27a0ec 4005 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4006 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 4007
4008 sb->s_root = NULL;
4009
4010 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 4011 EXT4_HAS_INCOMPAT_FEATURE(sb,
4012 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 4013
c5e06d10
JL		 4014 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
4015 !(sb->s_flags & MS_RDONLY))
4016 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4017 goto failed_mount3a;
c5e06d10 4018
ac27a0ec
DK		 4019 /*
4020 * The first inode we look at is the journal inode. Don't try
4021 * root first: it may be modified in the journal!
4022 */
4023 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 4024 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4025 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 4026 goto failed_mount3a;
0390131b
FM		 4027 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4028 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 4029 ext4_msg(sb, KERN_ERR, "required journal recovery "
4030 "suppressed and not mounted read-only");
744692dc 4031 goto failed_mount_wq;
ac27a0ec 4032 } else {
fd8c37ec 4033 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4034 sbi->s_journal = NULL;
4035 needs_recovery = 0;
4036 goto no_journal;
ac27a0ec
DK		 4037 }
4038
f32aaf2d 4039 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 4040 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4041 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4042 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4043 goto failed_mount_wq;
eb40a09c
JS		 4044 }
4045
25ed6e8a
DW		 4046 if (!set_journal_csum_feature_set(sb)) {
4047 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4048 "feature set");
4049 goto failed_mount_wq;
d4da6c9c 4050 }
818d276c 4051
ac27a0ec
DK		 4052 /* We have now updated the journal if required, so we can
4053 * validate the data journaling mode. */
4054 switch (test_opt(sb, DATA_FLAGS)) {
4055 case 0:
4056 /* No mode set, assume a default based on the journal
63f57933
AM		 4057 * capabilities: ORDERED_DATA if the journal can
4058 * cope, else JOURNAL_DATA
4059 */
dab291af
MC		 4060 if (jbd2_journal_check_available_features
4061 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 4062 set_opt(sb, ORDERED_DATA);
ac27a0ec 4063 else
fd8c37ec 4064 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 4065 break;
4066
617ba13b
MC		 4067 case EXT4_MOUNT_ORDERED_DATA:
4068 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4069 if (!jbd2_journal_check_available_features
4070 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4071 ext4_msg(sb, KERN_ERR, "Journal does not support "
4072 "requested data journaling mode");
744692dc 4073 goto failed_mount_wq;
ac27a0ec
DK		 4074 }
4075 default:
4076 break;
4077 }
b3881f74 4078 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4079
18aadd47
BJ		 4080 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4081
ce7e010a 4082no_journal:
9c191f70
M		 4083 if (ext4_mballoc_ready) {
4084 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4085 if (!sbi->s_mb_cache) {
4086 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4087 goto failed_mount_wq;
4088 }
4089 }
4090
1cb767cd
TT		 4091 if ((DUMMY_ENCRYPTION_ENABLED(sbi) ||
4092 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) &&
4093 (blocksize != PAGE_CACHE_SIZE)) {
4094 ext4_msg(sb, KERN_ERR,
4095 "Unsupported blocksize for fs encryption");
4096 goto failed_mount_wq;
4097 }
4098
4099 if (DUMMY_ENCRYPTION_ENABLED(sbi) &&
6ddb2447
TT		 4100 !(sb->s_flags & MS_RDONLY) &&
4101 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
4102 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
4103 ext4_commit_super(sb, 1);
4104 }
4105
952fc18e
TT		 4106 /*
4107 * Get the # of file system overhead blocks from the
4108 * superblock if present.
4109 */
4110 if (es->s_overhead_clusters)
4111 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4112 else {
07aa2ea1
LC		 4113 err = ext4_calculate_overhead(sb);
4114 if (err)
952fc18e
TT		 4115 goto failed_mount_wq;
4116 }
4117
fd89d5f2
TH		 4118 /*
4119 * The maximum number of concurrent works can be high and
4120 * concurrency isn't really necessary. Limit it to 1.
4121 */
2e8fa54e
JK		 4122 EXT4_SB(sb)->rsv_conversion_wq =
4123 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4124 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4125 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4126 ret = -ENOMEM;
2e8fa54e
JK		 4127 goto failed_mount4;
4128 }
4129
ac27a0ec 4130 /*
dab291af 4131 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4132 * so we can safely mount the rest of the filesystem now.
4133 */
4134
1d1fe1ee
DH		 4135 root = ext4_iget(sb, EXT4_ROOT_INO);
4136 if (IS_ERR(root)) {
b31e1552 4137 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4138 ret = PTR_ERR(root);
32a9bb57 4139 root = NULL;
ac27a0ec
DK		 4140 goto failed_mount4;
4141 }
4142 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4143 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4144 iput(root);
ac27a0ec
DK		 4145 goto failed_mount4;
4146 }
48fde701 4147 sb->s_root = d_make_root(root);
1d1fe1ee 4148 if (!sb->s_root) {
b31e1552 4149 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4150 ret = -ENOMEM;
4151 goto failed_mount4;
4152 }
ac27a0ec 4153
7e84b621
ES		 4154 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4155 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4156
4157 /* determine the minimum size of new large inodes, if present */
4158 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4159 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4160 EXT4_GOOD_OLD_INODE_SIZE;
4161 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4162 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4163 if (sbi->s_want_extra_isize <
4164 le16_to_cpu(es->s_want_extra_isize))
4165 sbi->s_want_extra_isize =
4166 le16_to_cpu(es->s_want_extra_isize);
4167 if (sbi->s_want_extra_isize <
4168 le16_to_cpu(es->s_min_extra_isize))
4169 sbi->s_want_extra_isize =
4170 le16_to_cpu(es->s_min_extra_isize);
4171 }
4172 }
4173 /* Check if enough inode space is available */
4174 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4175 sbi->s_inode_size) {
4176 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4177 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4178 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4179 "available");
ef7f3835
KS		 4180 }
4181
30fac0f7 4182 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
27dd4385
LC		 4183 if (err) {
4184 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
30fac0f7 4185 "reserved pool", ext4_calculate_resv_clusters(sb));
f9ae9cf5 4186 goto failed_mount4a;
27dd4385
LC		 4187 }
4188
6fd058f7
TT		 4189 err = ext4_setup_system_zone(sb);
4190 if (err) {
b31e1552 4191 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4192 "zone (%d)", err);
f9ae9cf5
TT		 4193 goto failed_mount4a;
4194 }
4195
4196 ext4_ext_init(sb);
4197 err = ext4_mb_init(sb);
4198 if (err) {
4199 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4200 err);
dcf2d804 4201 goto failed_mount5;
c2774d84
AK		 4202 }
4203
d5e03cbb
TT		 4204 block = ext4_count_free_clusters(sb);
4205 ext4_free_blocks_count_set(sbi->s_es,
4206 EXT4_C2B(sbi, block));
908c7f19
TH		 4207 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4208 GFP_KERNEL);
d5e03cbb
TT		 4209 if (!err) {
4210 unsigned long freei = ext4_count_free_inodes(sb);
4211 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4212 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4213 GFP_KERNEL);
d5e03cbb
TT		 4214 }
4215 if (!err)
4216 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4217 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4218 if (!err)
908c7f19
TH		 4219 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4220 GFP_KERNEL);
d5e03cbb
TT		 4221 if (err) {
4222 ext4_msg(sb, KERN_ERR, "insufficient memory");
4223 goto failed_mount6;
4224 }
4225
4226 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4227 if (!ext4_fill_flex_info(sb)) {
4228 ext4_msg(sb, KERN_ERR,
4229 "unable to initialize "
4230 "flex_bg meta info!");
4231 goto failed_mount6;
4232 }
4233
bfff6873
LC		 4234 err = ext4_register_li_request(sb, first_not_zeroed);
4235 if (err)
dcf2d804 4236 goto failed_mount6;
bfff6873 4237
3197ebdb
TT		 4238 sbi->s_kobj.kset = ext4_kset;
4239 init_completion(&sbi->s_kobj_unregister);
4240 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4241 "%s", sb->s_id);
dcf2d804
TM		 4242 if (err)
4243 goto failed_mount7;
3197ebdb 4244
9b2ff357
JK		 4245#ifdef CONFIG_QUOTA
4246 /* Enable quota usage during mount. */
4247 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4248 !(sb->s_flags & MS_RDONLY)) {
4249 err = ext4_enable_quotas(sb);
4250 if (err)
4251 goto failed_mount8;
4252 }
4253#endif /* CONFIG_QUOTA */
4254
617ba13b
MC		 4255 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4256 ext4_orphan_cleanup(sb, es);
4257 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4258 if (needs_recovery) {
b31e1552 4259 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4260 ext4_mark_recovery_complete(sb, es);
4261 }
4262 if (EXT4_SB(sb)->s_journal) {
4263 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4264 descr = " journalled data mode";
4265 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4266 descr = " ordered data mode";
4267 else
4268 descr = " writeback data mode";
4269 } else
4270 descr = "out journal";
4271
79add3a3
LC		 4272 if (test_opt(sb, DISCARD)) {
4273 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4274 if (!blk_queue_discard(q))
4275 ext4_msg(sb, KERN_WARNING,
4276 "mounting with \"discard\" option, but "
4277 "the device does not support discard");
4278 }
4279
d4c402d9 4280 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4281 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4282 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4283
66e61a9e
TT		 4284 if (es->s_error_count)
4285 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4286
efbed4dc
TT		 4287 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4288 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4289 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4290 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4291
d4c402d9 4292 kfree(orig_data);
ac27a0ec
DK		 4293 return 0;
4294
617ba13b 4295cantfind_ext4:
ac27a0ec 4296 if (!silent)
b31e1552 4297 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4298 goto failed_mount;
4299
72ba7450
TT		 4300#ifdef CONFIG_QUOTA
4301failed_mount8:
4302 kobject_del(&sbi->s_kobj);
4303#endif
dcf2d804
TM		 4304failed_mount7:
4305 ext4_unregister_li_request(sb);
4306failed_mount6:
f9ae9cf5 4307 ext4_mb_release(sb);
d5e03cbb 4308 if (sbi->s_flex_groups)
b93b41d4 4309 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4310 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4311 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4312 percpu_counter_destroy(&sbi->s_dirs_counter);
4313 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4314failed_mount5:
f9ae9cf5
TT		 4315 ext4_ext_release(sb);
4316 ext4_release_system_zone(sb);
4317failed_mount4a:
94bf608a 4318 dput(sb->s_root);
32a9bb57 4319 sb->s_root = NULL;
94bf608a 4320failed_mount4:
b31e1552 4321 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4322 if (EXT4_SB(sb)->rsv_conversion_wq)
4323 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4324failed_mount_wq:
0390131b
FM		 4325 if (sbi->s_journal) {
4326 jbd2_journal_destroy(sbi->s_journal);
4327 sbi->s_journal = NULL;
4328 }
50460fe8 4329failed_mount3a:
d3922a77 4330 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4331failed_mount3:
9105bb14 4332 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4333 if (sbi->s_mmp_tsk)
4334 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4335failed_mount2:
4336 for (i = 0; i < db_count; i++)
4337 brelse(sbi->s_group_desc[i]);
b93b41d4 4338 kvfree(sbi->s_group_desc);
ac27a0ec 4339failed_mount:
0441984a
DW		 4340 if (sbi->s_chksum_driver)
4341 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4342 if (sbi->s_proc) {
66acdcf4 4343 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4344 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4345 }
ac27a0ec 4346#ifdef CONFIG_QUOTA
a2d4a646 4347 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4348 kfree(sbi->s_qf_names[i]);
4349#endif
617ba13b 4350 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4351 brelse(bh);
4352out_fail:
4353 sb->s_fs_info = NULL;
f6830165 4354 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4355 kfree(sbi);
dcc7dae3 4356out_free_orig:
d4c402d9 4357 kfree(orig_data);
07aa2ea1 4358 return err ? err : ret;
ac27a0ec
DK		 4359}
4360
4361/*
4362 * Setup any per-fs journal parameters now. We'll do this both on
4363 * initial mount, once the journal has been initialised but before we've
4364 * done any recovery; and again on any subsequent remount.
4365 */
617ba13b 4366static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4367{
617ba13b 4368 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4369
30773840
TT		 4370 journal->j_commit_interval = sbi->s_commit_interval;
4371 journal->j_min_batch_time = sbi->s_min_batch_time;
4372 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4373
a931da6a 4374 write_lock(&journal->j_state_lock);
ac27a0ec 4375 if (test_opt(sb, BARRIER))
dab291af 4376 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4377 else
dab291af 4378 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4379 if (test_opt(sb, DATA_ERR_ABORT))
4380 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4381 else
4382 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4383 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4384}
4385
617ba13b 4386static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4387 unsigned int journal_inum)
4388{
4389 struct inode *journal_inode;
4390 journal_t *journal;
4391
0390131b
FM		 4392 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4393
ac27a0ec
DK		 4394 /* First, test for the existence of a valid inode on disk. Bad
4395 * things happen if we iget() an unused inode, as the subsequent
4396 * iput() will try to delete it. */
4397
1d1fe1ee
DH		 4398 journal_inode = ext4_iget(sb, journal_inum);
4399 if (IS_ERR(journal_inode)) {
b31e1552 4400 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4401 return NULL;
4402 }
4403 if (!journal_inode->i_nlink) {
4404 make_bad_inode(journal_inode);
4405 iput(journal_inode);
b31e1552 4406 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4407 return NULL;
4408 }
4409
e5f8eab8 4410 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4411 journal_inode, journal_inode->i_size);
1d1fe1ee 4412 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4413 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4414 iput(journal_inode);
4415 return NULL;
4416 }
4417
dab291af 4418 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4419 if (!journal) {
b31e1552 4420 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4421 iput(journal_inode);
4422 return NULL;
4423 }
4424 journal->j_private = sb;
617ba13b 4425 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4426 return journal;
4427}
4428
617ba13b 4429static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4430 dev_t j_dev)
4431{
2b2d6d01 4432 struct buffer_head *bh;
ac27a0ec 4433 journal_t *journal;
617ba13b
MC		 4434 ext4_fsblk_t start;
4435 ext4_fsblk_t len;
ac27a0ec 4436 int hblock, blocksize;
617ba13b 4437 ext4_fsblk_t sb_block;
ac27a0ec 4438 unsigned long offset;
2b2d6d01 4439 struct ext4_super_block *es;
ac27a0ec
DK		 4440 struct block_device *bdev;
4441
0390131b
FM		 4442 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4443
b31e1552 4444 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4445 if (bdev == NULL)
4446 return NULL;
4447
ac27a0ec 4448 blocksize = sb->s_blocksize;
e1defc4f 4449 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4450 if (blocksize < hblock) {
b31e1552
ES		 4451 ext4_msg(sb, KERN_ERR,
4452 "blocksize too small for journal device");
ac27a0ec
DK		 4453 goto out_bdev;
4454 }
4455
617ba13b
MC		 4456 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4457 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4458 set_blocksize(bdev, blocksize);
4459 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4460 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4461 "external journal");
ac27a0ec
DK		 4462 goto out_bdev;
4463 }
4464
2716b802 4465 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4466 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4467 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4468 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4469 ext4_msg(sb, KERN_ERR, "external journal has "
4470 "bad superblock");
ac27a0ec
DK		 4471 brelse(bh);
4472 goto out_bdev;
4473 }
4474
df4763be
DW		 4475 if ((le32_to_cpu(es->s_feature_ro_compat) &
4476 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4477 es->s_checksum != ext4_superblock_csum(sb, es)) {
4478 ext4_msg(sb, KERN_ERR, "external journal has "
4479 "corrupt superblock");
4480 brelse(bh);
4481 goto out_bdev;
4482 }
4483
617ba13b 4484 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4485 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4486 brelse(bh);
4487 goto out_bdev;
4488 }
4489
bd81d8ee 4490 len = ext4_blocks_count(es);
ac27a0ec
DK		 4491 start = sb_block + 1;
4492 brelse(bh); /* we're done with the superblock */
4493
dab291af 4494 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4495 start, len, blocksize);
4496 if (!journal) {
b31e1552 4497 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4498 goto out_bdev;
4499 }
4500 journal->j_private = sb;
9f203507 4501 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4502 wait_on_buffer(journal->j_sb_buffer);
4503 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4504 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4505 goto out_journal;
4506 }
4507 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4508 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4509 "user (unsupported) - %d",
ac27a0ec
DK		 4510 be32_to_cpu(journal->j_superblock->s_nr_users));
4511 goto out_journal;
4512 }
617ba13b
MC		 4513 EXT4_SB(sb)->journal_bdev = bdev;
4514 ext4_init_journal_params(sb, journal);
ac27a0ec 4515 return journal;
0b8e58a1 4516
ac27a0ec 4517out_journal:
dab291af 4518 jbd2_journal_destroy(journal);
ac27a0ec 4519out_bdev:
617ba13b 4520 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4521 return NULL;
4522}
4523
617ba13b
MC		 4524static int ext4_load_journal(struct super_block *sb,
4525 struct ext4_super_block *es,
ac27a0ec
DK		 4526 unsigned long journal_devnum)
4527{
4528 journal_t *journal;
4529 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4530 dev_t journal_dev;
4531 int err = 0;
4532 int really_read_only;
4533
0390131b
FM		 4534 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4535
ac27a0ec
DK		 4536 if (journal_devnum &&
4537 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4538 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4539 "numbers have changed");
ac27a0ec
DK		 4540 journal_dev = new_decode_dev(journal_devnum);
4541 } else
4542 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4543
4544 really_read_only = bdev_read_only(sb->s_bdev);
4545
4546 /*
4547 * Are we loading a blank journal or performing recovery after a
4548 * crash? For recovery, we need to check in advance whether we
4549 * can get read-write access to the device.
4550 */
617ba13b 4551 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4552 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4553 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4554 "required on readonly filesystem");
ac27a0ec 4555 if (really_read_only) {
b31e1552
ES		 4556 ext4_msg(sb, KERN_ERR, "write access "
4557 "unavailable, cannot proceed");
ac27a0ec
DK		 4558 return -EROFS;
4559 }
b31e1552
ES		 4560 ext4_msg(sb, KERN_INFO, "write access will "
4561 "be enabled during recovery");
ac27a0ec
DK		 4562 }
4563 }
4564
4565 if (journal_inum && journal_dev) {
b31e1552
ES		 4566 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4567 "and inode journals!");
ac27a0ec
DK		 4568 return -EINVAL;
4569 }
4570
4571 if (journal_inum) {
617ba13b 4572 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4573 return -EINVAL;
4574 } else {
617ba13b 4575 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4576 return -EINVAL;
4577 }
4578
90576c0b 4579 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4580 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4581
617ba13b 4582 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4583 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4584 if (!err) {
4585 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4586 if (save)
4587 memcpy(save, ((char *) es) +
4588 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4589 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4590 if (save)
4591 memcpy(((char *) es) + EXT4_S_ERR_START,
4592 save, EXT4_S_ERR_LEN);
4593 kfree(save);
4594 }
ac27a0ec
DK		 4595
4596 if (err) {
b31e1552 4597 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4598 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4599 return err;
4600 }
4601
617ba13b
MC		 4602 EXT4_SB(sb)->s_journal = journal;
4603 ext4_clear_journal_err(sb, es);
ac27a0ec 4604
c41303ce 4605 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4606 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4607 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4608
4609 /* Make sure we flush the recovery flag to disk. */
e2d67052 4610 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4611 }
4612
4613 return 0;
4614}
4615
e2d67052 4616static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4617{
e2d67052 4618 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4619 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4620 int error = 0;
ac27a0ec 4621
08439fec 4622 if (!sbh)
c4be0c1d 4623 return error;
914258bf
TT		 4624 if (buffer_write_io_error(sbh)) {
4625 /*
4626 * Oh, dear. A previous attempt to write the
4627 * superblock failed. This could happen because the
4628 * USB device was yanked out. Or it could happen to
4629 * be a transient write error and maybe the block will
4630 * be remapped. Nothing we can do but to retry the
4631 * write and hope for the best.
4632 */
b31e1552
ES		 4633 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4634 "superblock detected");
914258bf
TT		 4635 clear_buffer_write_io_error(sbh);
4636 set_buffer_uptodate(sbh);
4637 }
71290b36
TT		 4638 /*
4639 * If the file system is mounted read-only, don't update the
4640 * superblock write time. This avoids updating the superblock
4641 * write time when we are mounting the root file system
4642 * read/only but we need to replay the journal; at that point,
4643 * for people who are east of GMT and who make their clock
4644 * tick in localtime for Windows bug-for-bug compatibility,
4645 * the clock is set in the future, and this will cause e2fsck
4646 * to complain and force a full file system check.
4647 */
4648 if (!(sb->s_flags & MS_RDONLY))
4649 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4650 if (sb->s_bdev->bd_part)
4651 es->s_kbytes_written =
4652 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4653 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4654 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4655 else
4656 es->s_kbytes_written =
4657 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4658 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4659 ext4_free_blocks_count_set(es,
57042651
TT		 4660 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4661 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4662 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4663 es->s_free_inodes_count =
4664 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4665 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4666 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4667 ext4_superblock_csum_set(sb);
ac27a0ec 4668 mark_buffer_dirty(sbh);
914258bf 4669 if (sync) {
564bc402
DJ		 4670 error = __sync_dirty_buffer(sbh,
4671 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
c4be0c1d
TS		 4672 if (error)
4673 return error;
4674
4675 error = buffer_write_io_error(sbh);
4676 if (error) {
b31e1552
ES		 4677 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4678 "superblock");
914258bf
TT		 4679 clear_buffer_write_io_error(sbh);
4680 set_buffer_uptodate(sbh);
4681 }
4682 }
c4be0c1d 4683 return error;
ac27a0ec
DK		 4684}
4685
ac27a0ec
DK		 4686/*
4687 * Have we just finished recovery? If so, and if we are mounting (or
4688 * remounting) the filesystem readonly, then we will end up with a
4689 * consistent fs on disk. Record that fact.
4690 */
2b2d6d01
TT		 4691static void ext4_mark_recovery_complete(struct super_block *sb,
4692 struct ext4_super_block *es)
ac27a0ec 4693{
617ba13b 4694 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4695
0390131b
FM		 4696 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4697 BUG_ON(journal != NULL);
4698 return;
4699 }
dab291af 4700 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4701 if (jbd2_journal_flush(journal) < 0)
4702 goto out;
4703
617ba13b 4704 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4705 sb->s_flags & MS_RDONLY) {
617ba13b 4706 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4707 ext4_commit_super(sb, 1);
ac27a0ec 4708 }
7ffe1ea8
HK		 4709
4710out:
dab291af 4711 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4712}
4713
4714/*
4715 * If we are mounting (or read-write remounting) a filesystem whose journal
4716 * has recorded an error from a previous lifetime, move that error to the
4717 * main filesystem now.
4718 */
2b2d6d01
TT		 4719static void ext4_clear_journal_err(struct super_block *sb,
4720 struct ext4_super_block *es)
ac27a0ec
DK		 4721{
4722 journal_t *journal;
4723 int j_errno;
4724 const char *errstr;
4725
0390131b
FM		 4726 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4727
617ba13b 4728 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4729
4730 /*
4731 * Now check for any error status which may have been recorded in the
617ba13b 4732 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4733 */
4734
dab291af 4735 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4736 if (j_errno) {
4737 char nbuf[16];
4738
617ba13b 4739 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4740 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4741 "from previous mount: %s", errstr);
12062ddd 4742 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4743
617ba13b
MC		 4744 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4745 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4746 ext4_commit_super(sb, 1);
ac27a0ec 4747
dab291af 4748 jbd2_journal_clear_err(journal);
d796c52e 4749 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4750 }
4751}
4752
4753/*
4754 * Force the running and committing transactions to commit,
4755 * and wait on the commit.
4756 */
617ba13b 4757int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4758{
4759 journal_t *journal;
ac27a0ec
DK		 4760
4761 if (sb->s_flags & MS_RDONLY)
4762 return 0;
4763
617ba13b 4764 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4765 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4766}
4767
617ba13b 4768static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4769{
14ce0cb4 4770 int ret = 0;
9eddacf9 4771 tid_t target;
06a407f1 4772 bool needs_barrier = false;
8d5d02e6 4773 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4774
9bffad1e 4775 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4776 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4777 /*
4778 * Writeback quota in non-journalled quota case - journalled quota has
4779 * no dirty dquots
4780 */
4781 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4782 /*
4783 * Data writeback is possible w/o journal transaction, so barrier must
4784 * being sent at the end of the function. But we can skip it if
4785 * transaction_commit will do it for us.
4786 */
bda32530
TT		 4787 if (sbi->s_journal) {
4788 target = jbd2_get_latest_transaction(sbi->s_journal);
4789 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4790 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4791 needs_barrier = true;
4792
4793 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4794 if (wait)
4795 ret = jbd2_log_wait_commit(sbi->s_journal,
4796 target);
4797 }
4798 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4799 needs_barrier = true;
06a407f1
DM		 4800 if (needs_barrier) {
4801 int err;
4802 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4803 if (!ret)
4804 ret = err;
0390131b 4805 }
06a407f1
DM		 4806
4807 return ret;
4808}
4809
ac27a0ec
DK		 4810/*
4811 * LVM calls this function before a (read-only) snapshot is created. This
4812 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4813 *
4814 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4815 * state independently. It relies on upper layer to stop all data & metadata
4816 * modifications.
ac27a0ec 4817 */
c4be0c1d 4818static int ext4_freeze(struct super_block *sb)
ac27a0ec 4819{
c4be0c1d
TS		 4820 int error = 0;
4821 journal_t *journal;
ac27a0ec 4822
9ca92389
TT		 4823 if (sb->s_flags & MS_RDONLY)
4824 return 0;
ac27a0ec 4825
9ca92389 4826 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4827
bb044576
TT		 4828 if (journal) {
4829 /* Now we set up the journal barrier. */
4830 jbd2_journal_lock_updates(journal);
ac27a0ec 4831
bb044576
TT		 4832 /*
4833 * Don't clear the needs_recovery flag if we failed to
4834 * flush the journal.
4835 */
4836 error = jbd2_journal_flush(journal);
4837 if (error < 0)
4838 goto out;
c642dc9e
ES		 4839
4840 /* Journal blocked and flushed, clear needs_recovery flag. */
4841 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
bb044576 4842 }
9ca92389 4843
9ca92389 4844 error = ext4_commit_super(sb, 1);
6b0310fb 4845out:
bb044576
TT		 4846 if (journal)
4847 /* we rely on upper layer to stop further updates */
4848 jbd2_journal_unlock_updates(journal);
6b0310fb 4849 return error;
ac27a0ec
DK		 4850}
4851
4852/*
4853 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4854 * flag here, even though the filesystem is not technically dirty yet.
4855 */
c4be0c1d 4856static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4857{
9ca92389
TT		 4858 if (sb->s_flags & MS_RDONLY)
4859 return 0;
4860
c642dc9e
ES		 4861 if (EXT4_SB(sb)->s_journal) {
4862 /* Reset the needs_recovery flag before the fs is unlocked. */
4863 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4864 }
4865
9ca92389 4866 ext4_commit_super(sb, 1);
c4be0c1d 4867 return 0;
ac27a0ec
DK		 4868}
4869
673c6100
TT		 4870/*
4871 * Structure to save mount options for ext4_remount's benefit
4872 */
4873struct ext4_mount_options {
4874 unsigned long s_mount_opt;
a2595b8a 4875 unsigned long s_mount_opt2;
08cefc7a
EB		 4876 kuid_t s_resuid;
4877 kgid_t s_resgid;
673c6100
TT		 4878 unsigned long s_commit_interval;
4879 u32 s_min_batch_time, s_max_batch_time;
4880#ifdef CONFIG_QUOTA
4881 int s_jquota_fmt;
a2d4a646 4882 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4883#endif
4884};
4885
2b2d6d01 4886static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4887{
2b2d6d01 4888 struct ext4_super_block *es;
617ba13b 4889 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4890 unsigned long old_sb_flags;
617ba13b 4891 struct ext4_mount_options old_opts;
c79d967d 4892 int enable_quota = 0;
8a266467 4893 ext4_group_t g;
b3881f74 4894 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4895 int err = 0;
ac27a0ec 4896#ifdef CONFIG_QUOTA
03dafb5f 4897 int i, j;
ac27a0ec 4898#endif
d4c402d9 4899 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4900
4901 /* Store the original options */
4902 old_sb_flags = sb->s_flags;
4903 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4904 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4905 old_opts.s_resuid = sbi->s_resuid;
4906 old_opts.s_resgid = sbi->s_resgid;
4907 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4908 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4909 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4910#ifdef CONFIG_QUOTA
4911 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4912 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4913 if (sbi->s_qf_names[i]) {
4914 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4915 GFP_KERNEL);
4916 if (!old_opts.s_qf_names[i]) {
4917 for (j = 0; j < i; j++)
4918 kfree(old_opts.s_qf_names[j]);
3e36a163 4919 kfree(orig_data);
03dafb5f
CG		 4920 return -ENOMEM;
4921 }
4922 } else
4923 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4924#endif
b3881f74
TT		 4925 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4926 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4927
661aa520 4928 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4929 err = -EINVAL;
4930 goto restore_opts;
4931 }
4932
6b992ff2 4933 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4934 test_opt(sb, JOURNAL_CHECKSUM)) {
4935 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4936 "during remount not supported; ignoring");
4937 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4938 }
4939
6ae6514b
PS		 4940 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4941 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4942 ext4_msg(sb, KERN_ERR, "can't mount with "
4943 "both data=journal and delalloc");
4944 err = -EINVAL;
4945 goto restore_opts;
4946 }
4947 if (test_opt(sb, DIOREAD_NOLOCK)) {
4948 ext4_msg(sb, KERN_ERR, "can't mount with "
4949 "both data=journal and dioread_nolock");
4950 err = -EINVAL;
4951 goto restore_opts;
4952 }
923ae0ff
RZ		 4953 if (test_opt(sb, DAX)) {
4954 ext4_msg(sb, KERN_ERR, "can't mount with "
4955 "both data=journal and dax");
4956 err = -EINVAL;
4957 goto restore_opts;
4958 }
4959 }
4960
4961 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4962 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4963 "dax flag with busy inodes while remounting");
4964 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4965 }
4966
4ab2f15b 4967 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4968 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4969
4970 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4971 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4972
4973 es = sbi->s_es;
4974
b3881f74 4975 if (sbi->s_journal) {
0390131b 4976 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4977 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4978 }
ac27a0ec 4979
a2fd66d0
TT		 4980 if (*flags & MS_LAZYTIME)
4981 sb->s_flags |= MS_LAZYTIME;
4982
661aa520 4983 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4984 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4985 err = -EROFS;
4986 goto restore_opts;
4987 }
4988
4989 if (*flags & MS_RDONLY) {
38c03b34
TT		 4990 err = sync_filesystem(sb);
4991 if (err < 0)
4992 goto restore_opts;
0f0dd62f
CH		 4993 err = dquot_suspend(sb, -1);
4994 if (err < 0)
c79d967d 4995 goto restore_opts;
c79d967d 4996
ac27a0ec
DK		 4997 /*
4998 * First of all, the unconditional stuff we have to do
4999 * to disable replay of the journal when we next remount
5000 */
5001 sb->s_flags |= MS_RDONLY;
5002
5003 /*
5004 * OK, test if we are remounting a valid rw partition
5005 * readonly, and if so set the rdonly flag and then
5006 * mark the partition as valid again.
5007 */
617ba13b
MC		 5008 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5009 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5010 es->s_state = cpu_to_le16(sbi->s_mount_state);
5011
a63c9eb2 5012 if (sbi->s_journal)
0390131b 5013 ext4_mark_recovery_complete(sb, es);
ac27a0ec 5014 } else {
a13fb1a4 5015 /* Make sure we can mount this feature set readwrite */
2cb5cc8b
DW		 5016 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5017 EXT4_FEATURE_RO_COMPAT_READONLY) ||
5018 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5019 err = -EROFS;
5020 goto restore_opts;
5021 }
8a266467
TT		 5022 /*
5023 * Make sure the group descriptor checksums
0b8e58a1 5024 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5025 */
5026 for (g = 0; g < sbi->s_groups_count; g++) {
5027 struct ext4_group_desc *gdp =
5028 ext4_get_group_desc(sb, g, NULL);
5029
feb0ab32 5030 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5031 ext4_msg(sb, KERN_ERR,
5032 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 5033 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
5034 le16_to_cpu(gdp->bg_checksum));
5035 err = -EINVAL;
5036 goto restore_opts;
5037 }
5038 }
5039
ead6596b
ES		 5040 /*
5041 * If we have an unprocessed orphan list hanging
5042 * around from a previously readonly bdev mount,
5043 * require a full umount/remount for now.
5044 */
5045 if (es->s_last_orphan) {
b31e1552 5046 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5047 "remount RDWR because of unprocessed "
5048 "orphan inode list. Please "
b31e1552 5049 "umount/remount instead");
ead6596b
ES		 5050 err = -EINVAL;
5051 goto restore_opts;
5052 }
5053
ac27a0ec
DK		 5054 /*
5055 * Mounting a RDONLY partition read-write, so reread
5056 * and store the current valid flag. (It may have
5057 * been changed by e2fsck since we originally mounted
5058 * the partition.)
5059 */
0390131b
FM		 5060 if (sbi->s_journal)
5061 ext4_clear_journal_err(sb, es);
ac27a0ec 5062 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 5063 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 5064 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 5065 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5066 EXT4_FEATURE_INCOMPAT_MMP))
5067 if (ext4_multi_mount_protect(sb,
5068 le64_to_cpu(es->s_mmp_block))) {
5069 err = -EROFS;
5070 goto restore_opts;
5071 }
c79d967d 5072 enable_quota = 1;
ac27a0ec
DK		 5073 }
5074 }
bfff6873
LC		 5075
5076 /*
5077 * Reinitialize lazy itable initialization thread based on
5078 * current settings
5079 */
5080 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5081 ext4_unregister_li_request(sb);
5082 else {
5083 ext4_group_t first_not_zeroed;
5084 first_not_zeroed = ext4_has_uninit_itable(sb);
5085 ext4_register_li_request(sb, first_not_zeroed);
5086 }
5087
6fd058f7 5088 ext4_setup_system_zone(sb);
d096ad0f 5089 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5090 ext4_commit_super(sb, 1);
0390131b 5091
ac27a0ec
DK		 5092#ifdef CONFIG_QUOTA
5093 /* Release old quota file names */
a2d4a646 5094 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5095 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5096 if (enable_quota) {
5097 if (sb_any_quota_suspended(sb))
5098 dquot_resume(sb, -1);
5099 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5100 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5101 err = ext4_enable_quotas(sb);
07724f98 5102 if (err)
7c319d32 5103 goto restore_opts;
7c319d32
AK		 5104 }
5105 }
ac27a0ec 5106#endif
d4c402d9 5107
a26f4992 5108 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5109 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5110 kfree(orig_data);
ac27a0ec 5111 return 0;
0b8e58a1 5112
ac27a0ec
DK		 5113restore_opts:
5114 sb->s_flags = old_sb_flags;
5115 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5116 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5117 sbi->s_resuid = old_opts.s_resuid;
5118 sbi->s_resgid = old_opts.s_resgid;
5119 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5120 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5121 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5122#ifdef CONFIG_QUOTA
5123 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5124 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5125 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5126 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5127 }
5128#endif
d4c402d9 5129 kfree(orig_data);
ac27a0ec
DK		 5130 return err;
5131}
5132
2b2d6d01 5133static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5134{
5135 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5136 struct ext4_sb_info *sbi = EXT4_SB(sb);
5137 struct ext4_super_block *es = sbi->s_es;
27dd4385 5138 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5139 u64 fsid;
d02a9391 5140 s64 bfree;
27dd4385 5141 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5142
952fc18e
TT		 5143 if (!test_opt(sb, MINIX_DF))
5144 overhead = sbi->s_overhead;
ac27a0ec 5145
617ba13b 5146 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5147 buf->f_bsize = sb->s_blocksize;
b72f78cb 5148 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5149 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5150 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5151 /* prevent underflow in case that few free space is available */
57042651 5152 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5153 buf->f_bavail = buf->f_bfree -
5154 (ext4_r_blocks_count(es) + resv_blocks);
5155 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5156 buf->f_bavail = 0;
5157 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5158 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5159 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5160 fsid = le64_to_cpup((void *)es->s_uuid) ^
5161 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5162 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5163 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5164
ac27a0ec
DK		 5165 return 0;
5166}
5167
0b8e58a1
AD		 5168/* Helper function for writing quotas on sync - we need to start transaction
5169 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5170 * Process 1 Process 2
617ba13b 5171 * ext4_create() quota_sync()
a269eb18 5172 * jbd2_journal_start() write_dquot()
871a2931 5173 * dquot_initialize() down(dqio_mutex)
dab291af 5174 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5175 *
5176 */
5177
5178#ifdef CONFIG_QUOTA
5179
5180static inline struct inode *dquot_to_inode(struct dquot *dquot)
5181{
4c376dca 5182 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5183}
5184
617ba13b 5185static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5186{
5187 int ret, err;
5188 handle_t *handle;
5189 struct inode *inode;
5190
5191 inode = dquot_to_inode(dquot);
9924a92a 5192 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5193 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5194 if (IS_ERR(handle))
5195 return PTR_ERR(handle);
5196 ret = dquot_commit(dquot);
617ba13b 5197 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5198 if (!ret)
5199 ret = err;
5200 return ret;
5201}
5202
617ba13b 5203static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5204{
5205 int ret, err;
5206 handle_t *handle;
5207
9924a92a 5208 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5209 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5210 if (IS_ERR(handle))
5211 return PTR_ERR(handle);
5212 ret = dquot_acquire(dquot);
617ba13b 5213 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5214 if (!ret)
5215 ret = err;
5216 return ret;
5217}
5218
617ba13b 5219static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5220{
5221 int ret, err;
5222 handle_t *handle;
5223
9924a92a 5224 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5225 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5226 if (IS_ERR(handle)) {
5227 /* Release dquot anyway to avoid endless cycle in dqput() */
5228 dquot_release(dquot);
ac27a0ec 5229 return PTR_ERR(handle);
9c3013e9 5230 }
ac27a0ec 5231 ret = dquot_release(dquot);
617ba13b 5232 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5233 if (!ret)
5234 ret = err;
5235 return ret;
5236}
5237
617ba13b 5238static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5239{
262b4662
JK		 5240 struct super_block *sb = dquot->dq_sb;
5241 struct ext4_sb_info *sbi = EXT4_SB(sb);
5242
2c8be6b2 5243 /* Are we journaling quotas? */
262b4662
JK		 5244 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5245 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5246 dquot_mark_dquot_dirty(dquot);
617ba13b 5247 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5248 } else {
5249 return dquot_mark_dquot_dirty(dquot);
5250 }
5251}
5252
617ba13b 5253static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5254{
5255 int ret, err;
5256 handle_t *handle;
5257
5258 /* Data block + inode block */
2b0143b5 5259 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5260 if (IS_ERR(handle))
5261 return PTR_ERR(handle);
5262 ret = dquot_commit_info(sb, type);
617ba13b 5263 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5264 if (!ret)
5265 ret = err;
5266 return ret;
5267}
5268
5269/*
5270 * Turn on quotas during mount time - we need to find
5271 * the quota file and such...
5272 */
617ba13b 5273static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5274{
287a8095
CH		 5275 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5276 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5277}
5278
5279/*
5280 * Standard function to be called on quota_on
5281 */
617ba13b 5282static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5283 struct path *path)
ac27a0ec
DK		 5284{
5285 int err;
ac27a0ec
DK		 5286
5287 if (!test_opt(sb, QUOTA))
5288 return -EINVAL;
0623543b 5289
ac27a0ec 5290 /* Quotafile not on the same filesystem? */
d8c9584e 5291 if (path->dentry->d_sb != sb)
ac27a0ec 5292 return -EXDEV;
0623543b
JK		 5293 /* Journaling quota? */
5294 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5295 /* Quotafile not in fs root? */
f00c9e44 5296 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5297 ext4_msg(sb, KERN_WARNING,
5298 "Quota file not on filesystem root. "
5299 "Journaled quota will not work");
2b2d6d01 5300 }
0623543b
JK		 5301
5302 /*
5303 * When we journal data on quota file, we have to flush journal to see
5304 * all updates to the file when we bypass pagecache...
5305 */
0390131b 5306 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5307 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5308 /*
5309 * We don't need to lock updates but journal_flush() could
5310 * otherwise be livelocked...
5311 */
5312 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5313 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5314 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5315 if (err)
7ffe1ea8 5316 return err;
0623543b
JK		 5317 }
5318
f00c9e44 5319 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5320}
5321
7c319d32
AK		 5322static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5323 unsigned int flags)
5324{
5325 int err;
5326 struct inode *qf_inode;
a2d4a646 5327 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5328 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5329 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5330 };
5331
5332 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5333
5334 if (!qf_inums[type])
5335 return -EPERM;
5336
5337 qf_inode = ext4_iget(sb, qf_inums[type]);
5338 if (IS_ERR(qf_inode)) {
5339 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5340 return PTR_ERR(qf_inode);
5341 }
5342
bcb13850
JK		 5343 /* Don't account quota for quota files to avoid recursion */
5344 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5345 err = dquot_enable(qf_inode, type, format_id, flags);
5346 iput(qf_inode);
5347
5348 return err;
5349}
5350
5351/* Enable usage tracking for all quota types. */
5352static int ext4_enable_quotas(struct super_block *sb)
5353{
5354 int type, err = 0;
a2d4a646 5355 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5356 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5357 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5358 };
5359
5360 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5361 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5362 if (qf_inums[type]) {
5363 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5364 DQUOT_USAGE_ENABLED);
5365 if (err) {
5366 ext4_warning(sb,
72ba7450
TT		 5367 "Failed to enable quota tracking "
5368 "(type=%d, err=%d). Please run "
5369 "e2fsck to fix.", type, err);
7c319d32
AK		 5370 return err;
5371 }
5372 }
5373 }
5374 return 0;
5375}
5376
ca0e05e4
DM		 5377static int ext4_quota_off(struct super_block *sb, int type)
5378{
21f97697
JK		 5379 struct inode *inode = sb_dqopt(sb)->files[type];
5380 handle_t *handle;
5381
87009d86
DM		 5382 /* Force all delayed allocation blocks to be allocated.
5383 * Caller already holds s_umount sem */
5384 if (test_opt(sb, DELALLOC))
ca0e05e4 5385 sync_filesystem(sb);
ca0e05e4 5386
0b268590
AG		 5387 if (!inode)
5388 goto out;
5389
21f97697
JK		 5390 /* Update modification times of quota files when userspace can
5391 * start looking at them */
9924a92a 5392 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5393 if (IS_ERR(handle))
5394 goto out;
5395 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5396 ext4_mark_inode_dirty(handle, inode);
5397 ext4_journal_stop(handle);
5398
5399out:
ca0e05e4
DM		 5400 return dquot_quota_off(sb, type);
5401}
5402
ac27a0ec
DK		 5403/* Read data from quotafile - avoid pagecache and such because we cannot afford
5404 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5405 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5406 * we don't have to be afraid of races */
617ba13b 5407static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5408 size_t len, loff_t off)
5409{
5410 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5411 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5412 int offset = off & (sb->s_blocksize - 1);
5413 int tocopy;
5414 size_t toread;
5415 struct buffer_head *bh;
5416 loff_t i_size = i_size_read(inode);
5417
5418 if (off > i_size)
5419 return 0;
5420 if (off+len > i_size)
5421 len = i_size-off;
5422 toread = len;
5423 while (toread > 0) {
5424 tocopy = sb->s_blocksize - offset < toread ?
5425 sb->s_blocksize - offset : toread;
1c215028
TT		 5426 bh = ext4_bread(NULL, inode, blk, 0);
5427 if (IS_ERR(bh))
5428 return PTR_ERR(bh);
ac27a0ec
DK		 5429 if (!bh) /* A hole? */
5430 memset(data, 0, tocopy);
5431 else
5432 memcpy(data, bh->b_data+offset, tocopy);
5433 brelse(bh);
5434 offset = 0;
5435 toread -= tocopy;
5436 data += tocopy;
5437 blk++;
5438 }
5439 return len;
5440}
5441
5442/* Write to quotafile (we know the transaction is already started and has
5443 * enough credits) */
617ba13b 5444static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5445 const char *data, size_t len, loff_t off)
5446{
5447 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5448 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5449 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5450 int retries = 0;
ac27a0ec
DK		 5451 struct buffer_head *bh;
5452 handle_t *handle = journal_current_handle();
5453
0390131b 5454 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5455 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5456 " cancelled because transaction is not started",
9c3013e9
JK		 5457 (unsigned long long)off, (unsigned long long)len);
5458 return -EIO;
5459 }
67eeb568
DM		 5460 /*
5461 * Since we account only one data block in transaction credits,
5462 * then it is impossible to cross a block boundary.
5463 */
5464 if (sb->s_blocksize - offset < len) {
5465 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5466 " cancelled because not block aligned",
5467 (unsigned long long)off, (unsigned long long)len);
5468 return -EIO;
5469 }
5470
c5e298ae
TT		 5471 do {
5472 bh = ext4_bread(handle, inode, blk,
5473 EXT4_GET_BLOCKS_CREATE |
5474 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5475 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5476 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5477 if (IS_ERR(bh))
5478 return PTR_ERR(bh);
67eeb568
DM		 5479 if (!bh)
5480 goto out;
5d601255 5481 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5482 err = ext4_journal_get_write_access(handle, bh);
5483 if (err) {
5484 brelse(bh);
1c215028 5485 return err;
ac27a0ec 5486 }
67eeb568
DM		 5487 lock_buffer(bh);
5488 memcpy(bh->b_data+offset, data, len);
5489 flush_dcache_page(bh->b_page);
5490 unlock_buffer(bh);
62d2b5f2 5491 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5492 brelse(bh);
ac27a0ec 5493out:
67eeb568
DM		 5494 if (inode->i_size < off + len) {
5495 i_size_write(inode, off + len);
617ba13b 5496 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5497 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5498 }
67eeb568 5499 return len;
ac27a0ec
DK		 5500}
5501
5502#endif
5503
152a0836
AV		 5504static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5505 const char *dev_name, void *data)
ac27a0ec 5506{
152a0836 5507 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5508}
5509
37f328eb 5510#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5511static inline void register_as_ext2(void)
5512{
5513 int err = register_filesystem(&ext2_fs_type);
5514 if (err)
5515 printk(KERN_WARNING
5516 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5517}
5518
5519static inline void unregister_as_ext2(void)
5520{
5521 unregister_filesystem(&ext2_fs_type);
5522}
2035e776
TT		 5523
5524static inline int ext2_feature_set_ok(struct super_block *sb)
5525{
5526 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5527 return 0;
5528 if (sb->s_flags & MS_RDONLY)
5529 return 1;
5530 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5531 return 0;
5532 return 1;
5533}
24b58424
TT		 5534#else
5535static inline void register_as_ext2(void) { }
5536static inline void unregister_as_ext2(void) { }
2035e776 5537static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5538#endif
5539
37f328eb 5540#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5541static inline void register_as_ext3(void)
5542{
5543 int err = register_filesystem(&ext3_fs_type);
5544 if (err)
5545 printk(KERN_WARNING
5546 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5547}
5548
5549static inline void unregister_as_ext3(void)
5550{
5551 unregister_filesystem(&ext3_fs_type);
5552}
2035e776
TT		 5553
5554static inline int ext3_feature_set_ok(struct super_block *sb)
5555{
5556 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5557 return 0;
5558 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5559 return 0;
5560 if (sb->s_flags & MS_RDONLY)
5561 return 1;
5562 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5563 return 0;
5564 return 1;
5565}
24b58424
TT		 5566#else
5567static inline void register_as_ext3(void) { }
5568static inline void unregister_as_ext3(void) { }
2035e776 5569static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5570#endif
5571
03010a33
TT		 5572static struct file_system_type ext4_fs_type = {
5573 .owner = THIS_MODULE,
5574 .name = "ext4",
152a0836 5575 .mount = ext4_mount,
03010a33
TT		 5576 .kill_sb = kill_block_super,
5577 .fs_flags = FS_REQUIRES_DEV,
5578};
7f78e035 5579MODULE_ALIAS_FS("ext4");
03010a33 5580
8f021222 5581static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5582{
5583 struct ext4_features *ef;
5584 int ret = -ENOMEM;
5585
5586 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5587 if (!ef)
5588 goto out;
5589
5590 ef->f_kobj.kset = ext4_kset;
5591 init_completion(&ef->f_kobj_unregister);
5592 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5593 "features");
5594 if (ret) {
5595 kfree(ef);
5596 goto out;
5597 }
5598
5599 ext4_feat = ef;
5600 ret = 0;
5601out:
5602 return ret;
5603}
5604
8f021222
LC		 5605static void ext4_exit_feat_adverts(void)
5606{
5607 kobject_put(&ext4_feat->f_kobj);
5608 wait_for_completion(&ext4_feat->f_kobj_unregister);
5609 kfree(ext4_feat);
5610}
5611
e9e3bcec
ES		 5612/* Shared across all ext4 file systems */
5613wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5614struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5615
5dabfc78 5616static int __init ext4_init_fs(void)
ac27a0ec 5617{
e9e3bcec 5618 int i, err;
c9de560d 5619
07c0c5d8
AV		 5620 ext4_li_info = NULL;
5621 mutex_init(&ext4_li_mtx);
5622
9a4c8019 5623 /* Build-time check for flags consistency */
12e9b892 5624 ext4_check_flag_values();
e9e3bcec
ES		 5625
5626 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5627 mutex_init(&ext4__aio_mutex[i]);
5628 init_waitqueue_head(&ext4__ioend_wq[i]);
5629 }
5630
51865fda 5631 err = ext4_init_es();
6fd058f7
TT		 5632 if (err)
5633 return err;
51865fda
ZL		 5634
5635 err = ext4_init_pageio();
5636 if (err)
5637 goto out7;
5638
5dabfc78 5639 err = ext4_init_system_zone();
bd2d0210 5640 if (err)
d44651d0 5641 goto out6;
3197ebdb 5642 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5643 if (!ext4_kset) {
5644 err = -ENOMEM;
dd68314c 5645 goto out5;
0e376b1e 5646 }
d44651d0 5647 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5648
5649 err = ext4_init_feat_adverts();
dd68314c
TT		 5650 if (err)
5651 goto out4;
857ac889 5652
5dabfc78 5653 err = ext4_init_mballoc();
c9de560d
AT		 5654 if (err)
5655 goto out2;
9c191f70
M		 5656 else
5657 ext4_mballoc_ready = 1;
ac27a0ec
DK		 5658 err = init_inodecache();
5659 if (err)
5660 goto out1;
24b58424 5661 register_as_ext3();
2035e776 5662 register_as_ext2();
03010a33 5663 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5664 if (err)
5665 goto out;
bfff6873 5666
ac27a0ec
DK		 5667 return 0;
5668out:
24b58424
TT		 5669 unregister_as_ext2();
5670 unregister_as_ext3();
ac27a0ec
DK		 5671 destroy_inodecache();
5672out1:
9c191f70 5673 ext4_mballoc_ready = 0;
5dabfc78 5674 ext4_exit_mballoc();
9c191f70 5675out2:
8f021222 5676 ext4_exit_feat_adverts();
dd68314c 5677out4:
d44651d0
FJ		 5678 if (ext4_proc_root)
5679 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5680 kset_unregister(ext4_kset);
d44651d0 5681out5:
5dabfc78 5682 ext4_exit_system_zone();
d44651d0 5683out6:
5dabfc78 5684 ext4_exit_pageio();
51865fda
ZL		 5685out7:
5686 ext4_exit_es();
5687
ac27a0ec
DK		 5688 return err;
5689}
5690
5dabfc78 5691static void __exit ext4_exit_fs(void)
ac27a0ec 5692{
e298e73b 5693 ext4_exit_crypto();
bfff6873 5694 ext4_destroy_lazyinit_thread();
24b58424
TT		 5695 unregister_as_ext2();
5696 unregister_as_ext3();
03010a33 5697 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5698 destroy_inodecache();
5dabfc78 5699 ext4_exit_mballoc();
8f021222 5700 ext4_exit_feat_adverts();
9f6200bb 5701 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5702 kset_unregister(ext4_kset);
5dabfc78
TT		 5703 ext4_exit_system_zone();
5704 ext4_exit_pageio();
dd12ed14 5705 ext4_exit_es();
ac27a0ec
DK		 5706}
5707
5708MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5709MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5710MODULE_LICENSE("GPL");
5dabfc78
TT		 5711module_init(ext4_init_fs)
5712module_exit(ext4_exit_fs)
Ubuntu Artful kernel mirror