projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4 crypto: add symlink encryption
[deliverable/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
27#include <linux/parser.h>
ac27a0ec 28#include <linux/buffer_head.h>
a5694255 29#include <linux/exportfs.h>
ac27a0ec
DK		 30#include <linux/vfs.h>
31#include <linux/random.h>
32#include <linux/mount.h>
33#include <linux/namei.h>
34#include <linux/quotaops.h>
35#include <linux/seq_file.h>
9f6200bb 36#include <linux/proc_fs.h>
3197ebdb 37#include <linux/ctype.h>
1330593e 38#include <linux/log2.h>
717d50e4 39#include <linux/crc16.h>
7abc52c2 40#include <linux/cleancache.h>
ac27a0ec
DK		 41#include <asm/uaccess.h>
42
bfff6873
LC		 43#include <linux/kthread.h>
44#include <linux/freezer.h>
45
3dcf5451 46#include "ext4.h"
4a092d73 47#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 48#include "ext4_jbd2.h"
ac27a0ec
DK		 49#include "xattr.h"
50#include "acl.h"
3661d286 51#include "mballoc.h"
ac27a0ec 52
9bffad1e
TT		 53#define CREATE_TRACE_POINTS
54#include <trace/events/ext4.h>
55
1f109d5a 56static struct proc_dir_entry *ext4_proc_root;
3197ebdb 57static struct kset *ext4_kset;
0b75a840
LC		 58static struct ext4_lazy_init *ext4_li_info;
59static struct mutex ext4_li_mtx;
60static struct ext4_features *ext4_feat;
9c191f70 61static int ext4_mballoc_ready;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 72static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 74static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 75static int ext4_freeze(struct super_block *sb);
152a0836
AV		 76static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
2035e776
TT		 78static inline int ext2_feature_set_ok(struct super_block *sb);
79static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 80static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 81static void ext4_destroy_lazyinit_thread(void);
82static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 83static void ext4_clear_request_list(void);
27dd4385 84static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
ac27a0ec 85
2035e776
TT		 86#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87static struct file_system_type ext2_fs_type = {
88 .owner = THIS_MODULE,
89 .name = "ext2",
90 .mount = ext4_mount,
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
93};
7f78e035 94MODULE_ALIAS_FS("ext2");
fa7614dd 95MODULE_ALIAS("ext2");
2035e776
TT		 96#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97#else
98#define IS_EXT2_SB(sb) (0)
99#endif
100
101
ba69f9ab
JK		 102#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
152a0836 106 .mount = ext4_mount,
ba69f9ab
JK		 107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109};
7f78e035 110MODULE_ALIAS_FS("ext3");
fa7614dd 111MODULE_ALIAS("ext3");
ba69f9ab
JK		 112#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
113#else
114#define IS_EXT3_SB(sb) (0)
115#endif
bd81d8ee 116
d25425f8
DW		 117static int ext4_verify_csum_type(struct super_block *sb,
118 struct ext4_super_block *es)
119{
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
122 return 1;
123
124 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
125}
126
a9c47317
DW		 127static __le32 ext4_superblock_csum(struct super_block *sb,
128 struct ext4_super_block *es)
129{
130 struct ext4_sb_info *sbi = EXT4_SB(sb);
131 int offset = offsetof(struct ext4_super_block, s_checksum);
132 __u32 csum;
133
134 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
135
136 return cpu_to_le32(csum);
137}
138
c197855e
SH		 139static int ext4_superblock_csum_verify(struct super_block *sb,
140 struct ext4_super_block *es)
a9c47317 141{
9aa5d32b 142 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 143 return 1;
144
145 return es->s_checksum == ext4_superblock_csum(sb, es);
146}
147
06db49e6 148void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 149{
06db49e6
TT		 150 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
151
9aa5d32b 152 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 153 return;
154
155 es->s_checksum = ext4_superblock_csum(sb, es);
156}
157
9933fc0a
TT		 158void *ext4_kvmalloc(size_t size, gfp_t flags)
159{
160 void *ret;
161
8be04b93 162 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 163 if (!ret)
164 ret = __vmalloc(size, flags, PAGE_KERNEL);
165 return ret;
166}
167
168void *ext4_kvzalloc(size_t size, gfp_t flags)
169{
170 void *ret;
171
8be04b93 172 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 173 if (!ret)
174 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
175 return ret;
176}
177
8fadc143
AR		 178ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
179 struct ext4_group_desc *bg)
bd81d8ee 180{
3a14589c 181 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 182 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 183 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 184}
185
8fadc143
AR		 186ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
bd81d8ee 188{
5272f837 189 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 191 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 192}
193
8fadc143
AR		 194ext4_fsblk_t ext4_inode_table(struct super_block *sb,
195 struct ext4_group_desc *bg)
bd81d8ee 196{
5272f837 197 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 200}
201
021b65bb
TT		 202__u32 ext4_free_group_clusters(struct super_block *sb,
203 struct ext4_group_desc *bg)
560671a0
AK		 204{
205 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 207 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 208}
209
210__u32 ext4_free_inodes_count(struct super_block *sb,
211 struct ext4_group_desc *bg)
212{
213 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 215 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 216}
217
218__u32 ext4_used_dirs_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220{
221 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 223 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 224}
225
226__u32 ext4_itable_unused_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
228{
229 return le16_to_cpu(bg->bg_itable_unused_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 231 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 232}
233
8fadc143
AR		 234void ext4_block_bitmap_set(struct super_block *sb,
235 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 236{
3a14589c 237 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 238 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 240}
241
8fadc143
AR		 242void ext4_inode_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 244{
5272f837 245 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 248}
249
8fadc143
AR		 250void ext4_inode_table_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 252{
5272f837 253 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 256}
257
021b65bb
TT		 258void ext4_free_group_clusters_set(struct super_block *sb,
259 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 260{
261 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
264}
265
266void ext4_free_inodes_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
268{
269 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
272}
273
274void ext4_used_dirs_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276{
277 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
280}
281
282void ext4_itable_unused_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284{
285 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
288}
289
d3d1faf6 290
1c13d5c0
TT		 291static void __save_error_info(struct super_block *sb, const char *func,
292 unsigned int line)
293{
294 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
295
296 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
297 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
298 es->s_last_error_time = cpu_to_le32(get_seconds());
299 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
300 es->s_last_error_line = cpu_to_le32(line);
301 if (!es->s_first_error_time) {
302 es->s_first_error_time = es->s_last_error_time;
303 strncpy(es->s_first_error_func, func,
304 sizeof(es->s_first_error_func));
305 es->s_first_error_line = cpu_to_le32(line);
306 es->s_first_error_ino = es->s_last_error_ino;
307 es->s_first_error_block = es->s_last_error_block;
308 }
66e61a9e
TT		 309 /*
310 * Start the daily error reporting function if it hasn't been
311 * started already
312 */
313 if (!es->s_error_count)
314 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 315 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 316}
317
318static void save_error_info(struct super_block *sb, const char *func,
319 unsigned int line)
320{
321 __save_error_info(sb, func, line);
322 ext4_commit_super(sb, 1);
323}
324
18aadd47
BJ		 325static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
326{
327 struct super_block *sb = journal->j_private;
328 struct ext4_sb_info *sbi = EXT4_SB(sb);
329 int error = is_journal_aborted(journal);
5d3ee208 330 struct ext4_journal_cb_entry *jce;
18aadd47 331
5d3ee208 332 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 333 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 334 while (!list_empty(&txn->t_private_list)) {
335 jce = list_entry(txn->t_private_list.next,
336 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 337 list_del_init(&jce->jce_list);
338 spin_unlock(&sbi->s_md_lock);
339 jce->jce_func(sb, jce, error);
340 spin_lock(&sbi->s_md_lock);
341 }
342 spin_unlock(&sbi->s_md_lock);
343}
1c13d5c0 344
ac27a0ec
DK		 345/* Deal with the reporting of failure conditions on a filesystem such as
346 * inconsistencies detected or read IO failures.
347 *
348 * On ext2, we can store the error state of the filesystem in the
617ba13b 349 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 350 * write ordering constraints on the superblock which prevent us from
351 * writing it out straight away; and given that the journal is about to
352 * be aborted, we can't rely on the current, or future, transactions to
353 * write out the superblock safely.
354 *
dab291af 355 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 356 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 357 * that error until we've noted it down and cleared it.
358 */
359
617ba13b 360static void ext4_handle_error(struct super_block *sb)
ac27a0ec 361{
ac27a0ec
DK		 362 if (sb->s_flags & MS_RDONLY)
363 return;
364
2b2d6d01 365 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 366 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 367
4ab2f15b 368 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 369 if (journal)
dab291af 370 jbd2_journal_abort(journal, -EIO);
ac27a0ec 371 }
2b2d6d01 372 if (test_opt(sb, ERRORS_RO)) {
b31e1552 373 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 374 /*
375 * Make sure updated value of ->s_mount_flags will be visible
376 * before ->s_flags update
377 */
378 smp_wmb();
ac27a0ec
DK		 379 sb->s_flags |= MS_RDONLY;
380 }
ac27a0ec 381 if (test_opt(sb, ERRORS_PANIC))
617ba13b 382 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 383 sb->s_id);
384}
385
efbed4dc
TT		 386#define ext4_error_ratelimit(sb) \
387 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
388 "EXT4-fs error")
389
12062ddd 390void __ext4_error(struct super_block *sb, const char *function,
c398eda0 391 unsigned int line, const char *fmt, ...)
ac27a0ec 392{
0ff2ea7d 393 struct va_format vaf;
ac27a0ec
DK		 394 va_list args;
395
efbed4dc
TT		 396 if (ext4_error_ratelimit(sb)) {
397 va_start(args, fmt);
398 vaf.fmt = fmt;
399 vaf.va = &args;
400 printk(KERN_CRIT
401 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
402 sb->s_id, function, line, current->comm, &vaf);
403 va_end(args);
404 }
f3fc0210 405 save_error_info(sb, function, line);
617ba13b 406 ext4_handle_error(sb);
ac27a0ec
DK		 407}
408
e7c96e8e
JP		 409void __ext4_error_inode(struct inode *inode, const char *function,
410 unsigned int line, ext4_fsblk_t block,
411 const char *fmt, ...)
273df556
FM		 412{
413 va_list args;
f7c21177 414 struct va_format vaf;
1c13d5c0 415 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 416
1c13d5c0
TT		 417 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
418 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 419 if (ext4_error_ratelimit(inode->i_sb)) {
420 va_start(args, fmt);
421 vaf.fmt = fmt;
422 vaf.va = &args;
423 if (block)
424 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
425 "inode #%lu: block %llu: comm %s: %pV\n",
426 inode->i_sb->s_id, function, line, inode->i_ino,
427 block, current->comm, &vaf);
428 else
429 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
430 "inode #%lu: comm %s: %pV\n",
431 inode->i_sb->s_id, function, line, inode->i_ino,
432 current->comm, &vaf);
433 va_end(args);
434 }
1c13d5c0 435 save_error_info(inode->i_sb, function, line);
273df556
FM		 436 ext4_handle_error(inode->i_sb);
437}
438
e7c96e8e
JP		 439void __ext4_error_file(struct file *file, const char *function,
440 unsigned int line, ext4_fsblk_t block,
441 const char *fmt, ...)
273df556
FM		 442{
443 va_list args;
f7c21177 444 struct va_format vaf;
1c13d5c0 445 struct ext4_super_block *es;
496ad9aa 446 struct inode *inode = file_inode(file);
273df556
FM		 447 char pathname[80], *path;
448
1c13d5c0
TT		 449 es = EXT4_SB(inode->i_sb)->s_es;
450 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc
TT		 451 if (ext4_error_ratelimit(inode->i_sb)) {
452 path = d_path(&(file->f_path), pathname, sizeof(pathname));
453 if (IS_ERR(path))
454 path = "(unknown)";
455 va_start(args, fmt);
456 vaf.fmt = fmt;
457 vaf.va = &args;
458 if (block)
459 printk(KERN_CRIT
460 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
461 "block %llu: comm %s: path %s: %pV\n",
462 inode->i_sb->s_id, function, line, inode->i_ino,
463 block, current->comm, path, &vaf);
464 else
465 printk(KERN_CRIT
466 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
467 "comm %s: path %s: %pV\n",
468 inode->i_sb->s_id, function, line, inode->i_ino,
469 current->comm, path, &vaf);
470 va_end(args);
471 }
1c13d5c0 472 save_error_info(inode->i_sb, function, line);
273df556
FM		 473 ext4_handle_error(inode->i_sb);
474}
475
722887dd
TT		 476const char *ext4_decode_error(struct super_block *sb, int errno,
477 char nbuf[16])
ac27a0ec
DK		 478{
479 char *errstr = NULL;
480
481 switch (errno) {
482 case -EIO:
483 errstr = "IO failure";
484 break;
485 case -ENOMEM:
486 errstr = "Out of memory";
487 break;
488 case -EROFS:
78f1ddbb
TT		 489 if (!sb || (EXT4_SB(sb)->s_journal &&
490 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 491 errstr = "Journal has aborted";
492 else
493 errstr = "Readonly filesystem";
494 break;
495 default:
496 /* If the caller passed in an extra buffer for unknown
497 * errors, textualise them now. Else we just return
498 * NULL. */
499 if (nbuf) {
500 /* Check for truncated error codes... */
501 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
502 errstr = nbuf;
503 }
504 break;
505 }
506
507 return errstr;
508}
509
617ba13b 510/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 511 * automatically and invokes the appropriate error response. */
512
c398eda0
TT		 513void __ext4_std_error(struct super_block *sb, const char *function,
514 unsigned int line, int errno)
ac27a0ec
DK		 515{
516 char nbuf[16];
517 const char *errstr;
518
519 /* Special case: if the error is EROFS, and we're not already
520 * inside a transaction, then there's really no point in logging
521 * an error. */
522 if (errno == -EROFS && journal_current_handle() == NULL &&
523 (sb->s_flags & MS_RDONLY))
524 return;
525
efbed4dc
TT		 526 if (ext4_error_ratelimit(sb)) {
527 errstr = ext4_decode_error(sb, errno, nbuf);
528 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
529 sb->s_id, function, line, errstr);
530 }
ac27a0ec 531
efbed4dc 532 save_error_info(sb, function, line);
617ba13b 533 ext4_handle_error(sb);
ac27a0ec
DK		 534}
535
536/*
617ba13b 537 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 538 * abort function may be used to deal with unrecoverable failures such
539 * as journal IO errors or ENOMEM at a critical moment in log management.
540 *
541 * We unconditionally force the filesystem into an ABORT|READONLY state,
542 * unless the error response on the fs has been set to panic in which
543 * case we take the easy way out and panic immediately.
544 */
545
c67d859e 546void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 547 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 548{
549 va_list args;
550
1c13d5c0 551 save_error_info(sb, function, line);
ac27a0ec 552 va_start(args, fmt);
c398eda0
TT		 553 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
554 function, line);
ac27a0ec
DK		 555 vprintk(fmt, args);
556 printk("\n");
557 va_end(args);
558
1c13d5c0
TT		 559 if ((sb->s_flags & MS_RDONLY) == 0) {
560 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 561 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 562 /*
563 * Make sure updated value of ->s_mount_flags will be visible
564 * before ->s_flags update
565 */
566 smp_wmb();
567 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 568 if (EXT4_SB(sb)->s_journal)
569 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
570 save_error_info(sb, function, line);
571 }
ac27a0ec 572 if (test_opt(sb, ERRORS_PANIC))
617ba13b 573 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 574}
575
e7c96e8e
JP		 576void __ext4_msg(struct super_block *sb,
577 const char *prefix, const char *fmt, ...)
b31e1552 578{
0ff2ea7d 579 struct va_format vaf;
b31e1552
ES		 580 va_list args;
581
efbed4dc
TT		 582 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
583 return;
584
b31e1552 585 va_start(args, fmt);
0ff2ea7d
JP		 586 vaf.fmt = fmt;
587 vaf.va = &args;
588 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 589 va_end(args);
590}
591
12062ddd 592void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 593 unsigned int line, const char *fmt, ...)
ac27a0ec 594{
0ff2ea7d 595 struct va_format vaf;
ac27a0ec
DK		 596 va_list args;
597
efbed4dc
TT		 598 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
599 "EXT4-fs warning"))
600 return;
601
ac27a0ec 602 va_start(args, fmt);
0ff2ea7d
JP		 603 vaf.fmt = fmt;
604 vaf.va = &args;
605 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
606 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 607 va_end(args);
608}
609
e29136f8
TT		 610void __ext4_grp_locked_error(const char *function, unsigned int line,
611 struct super_block *sb, ext4_group_t grp,
612 unsigned long ino, ext4_fsblk_t block,
613 const char *fmt, ...)
5d1b1b3f
AK		 614__releases(bitlock)
615__acquires(bitlock)
616{
0ff2ea7d 617 struct va_format vaf;
5d1b1b3f
AK		 618 va_list args;
619 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
620
1c13d5c0
TT		 621 es->s_last_error_ino = cpu_to_le32(ino);
622 es->s_last_error_block = cpu_to_le64(block);
623 __save_error_info(sb, function, line);
0ff2ea7d 624
efbed4dc
TT		 625 if (ext4_error_ratelimit(sb)) {
626 va_start(args, fmt);
627 vaf.fmt = fmt;
628 vaf.va = &args;
629 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
630 sb->s_id, function, line, grp);
631 if (ino)
632 printk(KERN_CONT "inode %lu: ", ino);
633 if (block)
634 printk(KERN_CONT "block %llu:",
635 (unsigned long long) block);
636 printk(KERN_CONT "%pV\n", &vaf);
637 va_end(args);
638 }
5d1b1b3f
AK		 639
640 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 641 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 642 return;
643 }
1c13d5c0 644
5d1b1b3f
AK		 645 ext4_unlock_group(sb, grp);
646 ext4_handle_error(sb);
647 /*
648 * We only get here in the ERRORS_RO case; relocking the group
649 * may be dangerous, but nothing bad will happen since the
650 * filesystem will have already been marked read/only and the
651 * journal has been aborted. We return 1 as a hint to callers
652 * who might what to use the return value from
25985edc 653 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 654 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
655 * aggressively from the ext4 function in question, with a
656 * more appropriate error code.
657 */
658 ext4_lock_group(sb, grp);
659 return;
660}
661
617ba13b 662void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 663{
617ba13b 664 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 665
617ba13b 666 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 667 return;
668
12062ddd 669 ext4_warning(sb,
ac27a0ec
DK		 670 "updating to rev %d because of new feature flag, "
671 "running e2fsck is recommended",
617ba13b 672 EXT4_DYNAMIC_REV);
ac27a0ec 673
617ba13b
MC		 674 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
675 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
676 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 677 /* leave es->s_feature_*compat flags alone */
678 /* es->s_uuid will be set by e2fsck if empty */
679
680 /*
681 * The rest of the superblock fields should be zero, and if not it
682 * means they are likely already in use, so leave them alone. We
683 * can leave it up to e2fsck to clean up any inconsistencies there.
684 */
685}
686
687/*
688 * Open the external journal device
689 */
b31e1552 690static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 691{
692 struct block_device *bdev;
693 char b[BDEVNAME_SIZE];
694
d4d77629 695 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 696 if (IS_ERR(bdev))
697 goto fail;
698 return bdev;
699
700fail:
b31e1552 701 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 702 __bdevname(dev, b), PTR_ERR(bdev));
703 return NULL;
704}
705
706/*
707 * Release the journal device
708 */
4385bab1 709static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 710{
4385bab1 711 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 712}
713
4385bab1 714static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 715{
716 struct block_device *bdev;
ac27a0ec
DK		 717 bdev = sbi->journal_bdev;
718 if (bdev) {
4385bab1 719 ext4_blkdev_put(bdev);
ac27a0ec
DK		 720 sbi->journal_bdev = NULL;
721 }
ac27a0ec
DK		 722}
723
724static inline struct inode *orphan_list_entry(struct list_head *l)
725{
617ba13b 726 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 727}
728
617ba13b 729static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 730{
731 struct list_head *l;
732
b31e1552
ES		 733 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
734 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 735
736 printk(KERN_ERR "sb_info orphan list:\n");
737 list_for_each(l, &sbi->s_orphan) {
738 struct inode *inode = orphan_list_entry(l);
739 printk(KERN_ERR " "
740 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
741 inode->i_sb->s_id, inode->i_ino, inode,
742 inode->i_mode, inode->i_nlink,
743 NEXT_ORPHAN(inode));
744 }
745}
746
2b2d6d01 747static void ext4_put_super(struct super_block *sb)
ac27a0ec 748{
617ba13b
MC		 749 struct ext4_sb_info *sbi = EXT4_SB(sb);
750 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 751 int i, err;
ac27a0ec 752
857ac889 753 ext4_unregister_li_request(sb);
e0ccfd95
CH		 754 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
755
2e8fa54e 756 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 757 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 758
0390131b
FM		 759 if (sbi->s_journal) {
760 err = jbd2_journal_destroy(sbi->s_journal);
761 sbi->s_journal = NULL;
762 if (err < 0)
c67d859e 763 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 764 }
d4edac31 765
d3922a77 766 ext4_es_unregister_shrinker(sbi);
9105bb14 767 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 768 ext4_release_system_zone(sb);
769 ext4_mb_release(sb);
770 ext4_ext_release(sb);
771 ext4_xattr_put_super(sb);
772
ac27a0ec 773 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 774 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 775 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 776 }
58c5873a 777 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 778 ext4_commit_super(sb, 1);
779
240799cd 780 if (sbi->s_proc) {
66acdcf4 781 remove_proc_entry("options", sbi->s_proc);
9f6200bb 782 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 783 }
3197ebdb 784 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 785
786 for (i = 0; i < sbi->s_gdb_count; i++)
787 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 788 kvfree(sbi->s_group_desc);
789 kvfree(sbi->s_flex_groups);
57042651 790 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 791 percpu_counter_destroy(&sbi->s_freeinodes_counter);
792 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 793 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 794 brelse(sbi->s_sbh);
795#ifdef CONFIG_QUOTA
a2d4a646 796 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 797 kfree(sbi->s_qf_names[i]);
798#endif
799
800 /* Debugging code just in case the in-memory inode orphan list
801 * isn't empty. The on-disk one can be non-empty if we've
802 * detected an error and taken the fs readonly, but the
803 * in-memory list had better be clean by this point. */
804 if (!list_empty(&sbi->s_orphan))
805 dump_orphan_list(sb, sbi);
806 J_ASSERT(list_empty(&sbi->s_orphan));
807
f98393a6 808 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 809 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
810 /*
811 * Invalidate the journal device's buffers. We don't want them
812 * floating about in memory - the physical journal device may
813 * hotswapped, and it breaks the `ro-after' testing code.
814 */
815 sync_blockdev(sbi->journal_bdev);
f98393a6 816 invalidate_bdev(sbi->journal_bdev);
617ba13b 817 ext4_blkdev_remove(sbi);
ac27a0ec 818 }
9c191f70
M		 819 if (sbi->s_mb_cache) {
820 ext4_xattr_destroy_cache(sbi->s_mb_cache);
821 sbi->s_mb_cache = NULL;
822 }
c5e06d10
JL		 823 if (sbi->s_mmp_tsk)
824 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 825 sb->s_fs_info = NULL;
3197ebdb
TT		 826 /*
827 * Now that we are completely done shutting down the
828 * superblock, we need to actually destroy the kobject.
829 */
3197ebdb
TT		 830 kobject_put(&sbi->s_kobj);
831 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 832 if (sbi->s_chksum_driver)
833 crypto_free_shash(sbi->s_chksum_driver);
705895b6 834 kfree(sbi->s_blockgroup_lock);
ac27a0ec 835 kfree(sbi);
ac27a0ec
DK		 836}
837
e18b890b 838static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 839
840/*
841 * Called inside transaction, so use GFP_NOFS
842 */
617ba13b 843static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 844{
617ba13b 845 struct ext4_inode_info *ei;
ac27a0ec 846
e6b4f8da 847 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 848 if (!ei)
849 return NULL;
0b8e58a1 850
ac27a0ec 851 ei->vfs_inode.i_version = 1;
202ee5df 852 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 853 INIT_LIST_HEAD(&ei->i_prealloc_list);
854 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 855 ext4_es_init_tree(&ei->i_es_tree);
856 rwlock_init(&ei->i_es_lock);
edaa53ca 857 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 858 ei->i_es_all_nr = 0;
edaa53ca 859 ei->i_es_shk_nr = 0;
dd475925 860 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 861 ei->i_reserved_data_blocks = 0;
862 ei->i_reserved_meta_blocks = 0;
863 ei->i_allocated_meta_blocks = 0;
9d0be502 864 ei->i_da_metadata_calc_len = 0;
7e731bc9 865 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 866 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 867#ifdef CONFIG_QUOTA
868 ei->i_reserved_quota = 0;
96c7e0d9 869 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 870#endif
8aefcd55 871 ei->jinode = NULL;
2e8fa54e 872 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 873 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 874 ei->i_sync_tid = 0;
875 ei->i_datasync_tid = 0;
f7ad6d2e 876 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 877 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 878 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
b30ab0e0
MH		 879#ifdef CONFIG_EXT4_FS_ENCRYPTION
880 ei->i_encryption_key.mode = EXT4_ENCRYPTION_MODE_INVALID;
881#endif
0b8e58a1 882
ac27a0ec
DK		 883 return &ei->vfs_inode;
884}
885
7ff9c073
TT		 886static int ext4_drop_inode(struct inode *inode)
887{
888 int drop = generic_drop_inode(inode);
889
890 trace_ext4_drop_inode(inode, drop);
891 return drop;
892}
893
fa0d7e3d
NP		 894static void ext4_i_callback(struct rcu_head *head)
895{
896 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 897 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
898}
899
617ba13b 900static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 901{
9f7dd93d 902 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 903 ext4_msg(inode->i_sb, KERN_ERR,
904 "Inode %lu (%p): orphan list check failed!",
905 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 906 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
907 EXT4_I(inode), sizeof(struct ext4_inode_info),
908 true);
909 dump_stack();
910 }
fa0d7e3d 911 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 912}
913
51cc5068 914static void init_once(void *foo)
ac27a0ec 915{
617ba13b 916 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 917
a35afb83 918 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 919 init_rwsem(&ei->xattr_sem);
0e855ac8 920 init_rwsem(&ei->i_data_sem);
a35afb83 921 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 922}
923
e67bc2b3 924static int __init init_inodecache(void)
ac27a0ec 925{
617ba13b
MC		 926 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
927 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 928 0, (SLAB_RECLAIM_ACCOUNT|
929 SLAB_MEM_SPREAD),
20c2df83 930 init_once);
617ba13b 931 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 932 return -ENOMEM;
933 return 0;
934}
935
936static void destroy_inodecache(void)
937{
8c0a8537
KS		 938 /*
939 * Make sure all delayed rcu free inodes are flushed before we
940 * destroy cache.
941 */
942 rcu_barrier();
617ba13b 943 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 944}
945
0930fcc1 946void ext4_clear_inode(struct inode *inode)
ac27a0ec 947{
0930fcc1 948 invalidate_inode_buffers(inode);
dbd5768f 949 clear_inode(inode);
9f754758 950 dquot_drop(inode);
c2ea3fde 951 ext4_discard_preallocations(inode);
51865fda 952 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 953 if (EXT4_I(inode)->jinode) {
954 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
955 EXT4_I(inode)->jinode);
956 jbd2_free_inode(EXT4_I(inode)->jinode);
957 EXT4_I(inode)->jinode = NULL;
958 }
ac27a0ec
DK		 959}
960
1b961ac0 961static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 962 u64 ino, u32 generation)
ac27a0ec 963{
ac27a0ec 964 struct inode *inode;
ac27a0ec 965
617ba13b 966 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 967 return ERR_PTR(-ESTALE);
617ba13b 968 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 969 return ERR_PTR(-ESTALE);
970
971 /* iget isn't really right if the inode is currently unallocated!!
972 *
617ba13b 973 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 974 * deleted, so we should be safe.
975 *
976 * Currently we don't know the generation for parent directory, so
977 * a generation of 0 means "accept any"
978 */
f4bb2981 979 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 980 if (IS_ERR(inode))
981 return ERR_CAST(inode);
982 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 983 iput(inode);
984 return ERR_PTR(-ESTALE);
985 }
1b961ac0
CH		 986
987 return inode;
988}
989
990static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 991 int fh_len, int fh_type)
1b961ac0
CH		 992{
993 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
994 ext4_nfs_get_inode);
995}
996
997static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 998 int fh_len, int fh_type)
1b961ac0
CH		 999{
1000 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1001 ext4_nfs_get_inode);
ac27a0ec
DK		 1002}
1003
c39a7f84
TO		 1004/*
1005 * Try to release metadata pages (indirect blocks, directories) which are
1006 * mapped via the block device. Since these pages could have journal heads
1007 * which would prevent try_to_free_buffers() from freeing them, we must use
1008 * jbd2 layer's try_to_free_buffers() function to release them.
1009 */
0b8e58a1
AD		 1010static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1011 gfp_t wait)
c39a7f84
TO		 1012{
1013 journal_t *journal = EXT4_SB(sb)->s_journal;
1014
1015 WARN_ON(PageChecked(page));
1016 if (!page_has_buffers(page))
1017 return 0;
1018 if (journal)
1019 return jbd2_journal_try_to_free_buffers(journal, page,
1020 wait & ~__GFP_WAIT);
1021 return try_to_free_buffers(page);
1022}
1023
ac27a0ec 1024#ifdef CONFIG_QUOTA
af5bc92d 1025#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1026#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1027
617ba13b
MC		 1028static int ext4_write_dquot(struct dquot *dquot);
1029static int ext4_acquire_dquot(struct dquot *dquot);
1030static int ext4_release_dquot(struct dquot *dquot);
1031static int ext4_mark_dquot_dirty(struct dquot *dquot);
1032static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1033static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1034 struct path *path);
ca0e05e4 1035static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1036static int ext4_quota_on_mount(struct super_block *sb, int type);
1037static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1038 size_t len, loff_t off);
617ba13b 1039static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1040 const char *data, size_t len, loff_t off);
7c319d32
AK		 1041static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1042 unsigned int flags);
1043static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1044
96c7e0d9
JK		 1045static struct dquot **ext4_get_dquots(struct inode *inode)
1046{
1047 return EXT4_I(inode)->i_dquot;
1048}
1049
61e225dc 1050static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1051 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1052 .write_dquot = ext4_write_dquot,
1053 .acquire_dquot = ext4_acquire_dquot,
1054 .release_dquot = ext4_release_dquot,
1055 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1056 .write_info = ext4_write_info,
1057 .alloc_dquot = dquot_alloc,
1058 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1059};
1060
0d54b217 1061static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1062 .quota_on = ext4_quota_on,
ca0e05e4 1063 .quota_off = ext4_quota_off,
287a8095
CH		 1064 .quota_sync = dquot_quota_sync,
1065 .get_info = dquot_get_dqinfo,
1066 .set_info = dquot_set_dqinfo,
1067 .get_dqblk = dquot_get_dqblk,
1068 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1069};
1070#endif
1071
ee9b6d61 1072static const struct super_operations ext4_sops = {
617ba13b
MC		 1073 .alloc_inode = ext4_alloc_inode,
1074 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1075 .write_inode = ext4_write_inode,
1076 .dirty_inode = ext4_dirty_inode,
7ff9c073 1077 .drop_inode = ext4_drop_inode,
0930fcc1 1078 .evict_inode = ext4_evict_inode,
617ba13b 1079 .put_super = ext4_put_super,
617ba13b 1080 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1081 .freeze_fs = ext4_freeze,
1082 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1083 .statfs = ext4_statfs,
1084 .remount_fs = ext4_remount,
617ba13b 1085 .show_options = ext4_show_options,
ac27a0ec 1086#ifdef CONFIG_QUOTA
617ba13b
MC		 1087 .quota_read = ext4_quota_read,
1088 .quota_write = ext4_quota_write,
96c7e0d9 1089 .get_dquots = ext4_get_dquots,
ac27a0ec 1090#endif
c39a7f84 1091 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1092};
1093
39655164 1094static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1095 .fh_to_dentry = ext4_fh_to_dentry,
1096 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1097 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1098};
1099
1100enum {
1101 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1102 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1103 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1104 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1105 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1106 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1107 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1108 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1109 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1110 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1111 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1112 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
923ae0ff 1113 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1449032b 1114 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1115 Opt_lazytime, Opt_nolazytime,
1449032b 1116 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1117 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1118 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1119 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1120 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1121};
1122
a447c093 1123static const match_table_t tokens = {
ac27a0ec
DK		 1124 {Opt_bsd_df, "bsddf"},
1125 {Opt_minix_df, "minixdf"},
1126 {Opt_grpid, "grpid"},
1127 {Opt_grpid, "bsdgroups"},
1128 {Opt_nogrpid, "nogrpid"},
1129 {Opt_nogrpid, "sysvgroups"},
1130 {Opt_resgid, "resgid=%u"},
1131 {Opt_resuid, "resuid=%u"},
1132 {Opt_sb, "sb=%u"},
1133 {Opt_err_cont, "errors=continue"},
1134 {Opt_err_panic, "errors=panic"},
1135 {Opt_err_ro, "errors=remount-ro"},
1136 {Opt_nouid32, "nouid32"},
ac27a0ec 1137 {Opt_debug, "debug"},
72578c33
TT		 1138 {Opt_removed, "oldalloc"},
1139 {Opt_removed, "orlov"},
ac27a0ec
DK		 1140 {Opt_user_xattr, "user_xattr"},
1141 {Opt_nouser_xattr, "nouser_xattr"},
1142 {Opt_acl, "acl"},
1143 {Opt_noacl, "noacl"},
e3bb52ae 1144 {Opt_noload, "norecovery"},
5a916be1 1145 {Opt_noload, "noload"},
72578c33
TT		 1146 {Opt_removed, "nobh"},
1147 {Opt_removed, "bh"},
ac27a0ec 1148 {Opt_commit, "commit=%u"},
30773840
TT		 1149 {Opt_min_batch_time, "min_batch_time=%u"},
1150 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1151 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1152 {Opt_journal_path, "journal_path=%s"},
818d276c 1153 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1154 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1155 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1156 {Opt_abort, "abort"},
1157 {Opt_data_journal, "data=journal"},
1158 {Opt_data_ordered, "data=ordered"},
1159 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1160 {Opt_data_err_abort, "data_err=abort"},
1161 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1162 {Opt_offusrjquota, "usrjquota="},
1163 {Opt_usrjquota, "usrjquota=%s"},
1164 {Opt_offgrpjquota, "grpjquota="},
1165 {Opt_grpjquota, "grpjquota=%s"},
1166 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1167 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1168 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1169 {Opt_grpquota, "grpquota"},
1170 {Opt_noquota, "noquota"},
1171 {Opt_quota, "quota"},
1172 {Opt_usrquota, "usrquota"},
1173 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1174 {Opt_barrier, "barrier"},
1175 {Opt_nobarrier, "nobarrier"},
25ec56b5 1176 {Opt_i_version, "i_version"},
923ae0ff 1177 {Opt_dax, "dax"},
c9de560d 1178 {Opt_stripe, "stripe=%u"},
64769240 1179 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1180 {Opt_lazytime, "lazytime"},
1181 {Opt_nolazytime, "nolazytime"},
dd919b98 1182 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1183 {Opt_removed, "mblk_io_submit"},
1184 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1185 {Opt_block_validity, "block_validity"},
1186 {Opt_noblock_validity, "noblock_validity"},
240799cd 1187 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1188 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1189 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1190 {Opt_auto_da_alloc, "auto_da_alloc"},
1191 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1192 {Opt_dioread_nolock, "dioread_nolock"},
1193 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1194 {Opt_discard, "discard"},
1195 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1196 {Opt_init_itable, "init_itable=%u"},
1197 {Opt_init_itable, "init_itable"},
1198 {Opt_noinit_itable, "noinit_itable"},
df981d03 1199 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1200 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1201 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1202 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1203 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1204 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1205 {Opt_err, NULL},
ac27a0ec
DK		 1206};
1207
617ba13b 1208static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1209{
617ba13b 1210 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1211 char *options = (char *) *data;
1212
1213 if (!options || strncmp(options, "sb=", 3) != 0)
1214 return 1; /* Default location */
0b8e58a1 1215
ac27a0ec 1216 options += 3;
0b8e58a1 1217 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1218 sb_block = simple_strtoul(options, &options, 0);
1219 if (*options && *options != ',') {
4776004f 1220 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1221 (char *) *data);
1222 return 1;
1223 }
1224 if (*options == ',')
1225 options++;
1226 *data = (void *) options;
0b8e58a1 1227
ac27a0ec
DK		 1228 return sb_block;
1229}
1230
b3881f74 1231#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1232static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1233 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1234
56c50f11
DM		 1235#ifdef CONFIG_QUOTA
1236static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1237{
1238 struct ext4_sb_info *sbi = EXT4_SB(sb);
1239 char *qname;
03dafb5f 1240 int ret = -1;
56c50f11
DM		 1241
1242 if (sb_any_quota_loaded(sb) &&
1243 !sbi->s_qf_names[qtype]) {
1244 ext4_msg(sb, KERN_ERR,
1245 "Cannot change journaled "
1246 "quota options when quota turned on");
57f73c2c 1247 return -1;
56c50f11 1248 }
262b4662
JK		 1249 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1250 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1251 "when QUOTA feature is enabled");
1252 return -1;
1253 }
56c50f11
DM		 1254 qname = match_strdup(args);
1255 if (!qname) {
1256 ext4_msg(sb, KERN_ERR,
1257 "Not enough memory for storing quotafile name");
57f73c2c 1258 return -1;
56c50f11 1259 }
03dafb5f
CG		 1260 if (sbi->s_qf_names[qtype]) {
1261 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1262 ret = 1;
1263 else
1264 ext4_msg(sb, KERN_ERR,
1265 "%s quota file already specified",
1266 QTYPE2NAME(qtype));
1267 goto errout;
56c50f11 1268 }
03dafb5f 1269 if (strchr(qname, '/')) {
56c50f11
DM		 1270 ext4_msg(sb, KERN_ERR,
1271 "quotafile must be on filesystem root");
03dafb5f 1272 goto errout;
56c50f11 1273 }
03dafb5f 1274 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1275 set_opt(sb, QUOTA);
56c50f11 1276 return 1;
03dafb5f
CG		 1277errout:
1278 kfree(qname);
1279 return ret;
56c50f11
DM		 1280}
1281
1282static int clear_qf_name(struct super_block *sb, int qtype)
1283{
1284
1285 struct ext4_sb_info *sbi = EXT4_SB(sb);
1286
1287 if (sb_any_quota_loaded(sb) &&
1288 sbi->s_qf_names[qtype]) {
1289 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1290 " when quota turned on");
57f73c2c 1291 return -1;
56c50f11 1292 }
03dafb5f 1293 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1294 sbi->s_qf_names[qtype] = NULL;
1295 return 1;
1296}
1297#endif
1298
26092bf5
TT		 1299#define MOPT_SET 0x0001
1300#define MOPT_CLEAR 0x0002
1301#define MOPT_NOSUPPORT 0x0004
1302#define MOPT_EXPLICIT 0x0008
1303#define MOPT_CLEAR_ERR 0x0010
1304#define MOPT_GTE0 0x0020
ac27a0ec 1305#ifdef CONFIG_QUOTA
26092bf5
TT		 1306#define MOPT_Q 0
1307#define MOPT_QFMT 0x0040
1308#else
1309#define MOPT_Q MOPT_NOSUPPORT
1310#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1311#endif
26092bf5 1312#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1313#define MOPT_NO_EXT2 0x0100
1314#define MOPT_NO_EXT3 0x0200
1315#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1316#define MOPT_STRING 0x0400
26092bf5
TT		 1317
1318static const struct mount_opts {
1319 int token;
1320 int mount_opt;
1321 int flags;
1322} ext4_mount_opts[] = {
1323 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1324 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1325 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1326 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1327 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1328 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1329 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1330 MOPT_EXT4_ONLY | MOPT_SET},
1331 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1332 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1333 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1334 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1335 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1336 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1337 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1338 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1339 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1340 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c
TT		 1341 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1342 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1343 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1344 EXT4_MOUNT_JOURNAL_CHECKSUM),
1345 MOPT_EXT4_ONLY | MOPT_SET},
1346 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1347 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1348 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1349 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1350 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1351 MOPT_NO_EXT2 | MOPT_SET},
1352 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1353 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1354 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1355 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1356 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1357 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1358 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1359 {Opt_commit, 0, MOPT_GTE0},
1360 {Opt_max_batch_time, 0, MOPT_GTE0},
1361 {Opt_min_batch_time, 0, MOPT_GTE0},
1362 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1363 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1364 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1365 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1366 {Opt_resuid, 0, MOPT_GTE0},
1367 {Opt_resgid, 0, MOPT_GTE0},
1368 {Opt_journal_dev, 0, MOPT_GTE0},
ad4eec61 1369 {Opt_journal_path, 0, MOPT_STRING},
0efb3b23 1370 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1371 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1372 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1373 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1374 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1375 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1376 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1377#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1378 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1379 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1380#else
26092bf5
TT		 1381 {Opt_acl, 0, MOPT_NOSUPPORT},
1382 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1383#endif
26092bf5
TT		 1384 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1385 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1386 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1387 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1388 MOPT_SET | MOPT_Q},
1389 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1390 MOPT_SET | MOPT_Q},
1391 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1392 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1393 {Opt_usrjquota, 0, MOPT_Q},
1394 {Opt_grpjquota, 0, MOPT_Q},
1395 {Opt_offusrjquota, 0, MOPT_Q},
1396 {Opt_offgrpjquota, 0, MOPT_Q},
1397 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1398 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1399 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1400 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1401 {Opt_err, 0, 0}
1402};
1403
1404static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1405 substring_t *args, unsigned long *journal_devnum,
1406 unsigned int *journal_ioprio, int is_remount)
1407{
1408 struct ext4_sb_info *sbi = EXT4_SB(sb);
1409 const struct mount_opts *m;
08cefc7a
EB		 1410 kuid_t uid;
1411 kgid_t gid;
26092bf5
TT		 1412 int arg = 0;
1413
57f73c2c
TT		 1414#ifdef CONFIG_QUOTA
1415 if (token == Opt_usrjquota)
1416 return set_qf_name(sb, USRQUOTA, &args[0]);
1417 else if (token == Opt_grpjquota)
1418 return set_qf_name(sb, GRPQUOTA, &args[0]);
1419 else if (token == Opt_offusrjquota)
1420 return clear_qf_name(sb, USRQUOTA);
1421 else if (token == Opt_offgrpjquota)
1422 return clear_qf_name(sb, GRPQUOTA);
1423#endif
26092bf5 1424 switch (token) {
f7048605
TT		 1425 case Opt_noacl:
1426 case Opt_nouser_xattr:
1427 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1428 break;
26092bf5
TT		 1429 case Opt_sb:
1430 return 1; /* handled by get_sb_block() */
1431 case Opt_removed:
5f3633e3 1432 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1433 return 1;
26092bf5
TT		 1434 case Opt_abort:
1435 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1436 return 1;
1437 case Opt_i_version:
1438 sb->s_flags |= MS_I_VERSION;
1439 return 1;
a26f4992
TT		 1440 case Opt_lazytime:
1441 sb->s_flags |= MS_LAZYTIME;
1442 return 1;
1443 case Opt_nolazytime:
1444 sb->s_flags &= ~MS_LAZYTIME;
1445 return 1;
26092bf5
TT		 1446 }
1447
5f3633e3
JK		 1448 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1449 if (token == m->token)
1450 break;
1451
1452 if (m->token == Opt_err) {
1453 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1454 "or missing value", opt);
1455 return -1;
1456 }
1457
8dc0aa8c
TT		 1458 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1459 ext4_msg(sb, KERN_ERR,
1460 "Mount option \"%s\" incompatible with ext2", opt);
1461 return -1;
1462 }
1463 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1464 ext4_msg(sb, KERN_ERR,
1465 "Mount option \"%s\" incompatible with ext3", opt);
1466 return -1;
1467 }
1468
ad4eec61 1469 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1470 return -1;
1471 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1472 return -1;
1473 if (m->flags & MOPT_EXPLICIT)
1474 set_opt2(sb, EXPLICIT_DELALLOC);
1475 if (m->flags & MOPT_CLEAR_ERR)
1476 clear_opt(sb, ERRORS_MASK);
1477 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1478 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1479 "options when quota turned on");
1480 return -1;
1481 }
1482
1483 if (m->flags & MOPT_NOSUPPORT) {
1484 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1485 } else if (token == Opt_commit) {
1486 if (arg == 0)
1487 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1488 sbi->s_commit_interval = HZ * arg;
1489 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1490 sbi->s_max_batch_time = arg;
1491 } else if (token == Opt_min_batch_time) {
1492 sbi->s_min_batch_time = arg;
1493 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1494 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1495 ext4_msg(sb, KERN_ERR,
1496 "EXT4-fs: inode_readahead_blks must be "
1497 "0 or a power of 2 smaller than 2^31");
26092bf5 1498 return -1;
5f3633e3
JK		 1499 }
1500 sbi->s_inode_readahead_blks = arg;
1501 } else if (token == Opt_init_itable) {
1502 set_opt(sb, INIT_INODE_TABLE);
1503 if (!args->from)
1504 arg = EXT4_DEF_LI_WAIT_MULT;
1505 sbi->s_li_wait_mult = arg;
1506 } else if (token == Opt_max_dir_size_kb) {
1507 sbi->s_max_dir_size_kb = arg;
1508 } else if (token == Opt_stripe) {
1509 sbi->s_stripe = arg;
1510 } else if (token == Opt_resuid) {
1511 uid = make_kuid(current_user_ns(), arg);
1512 if (!uid_valid(uid)) {
1513 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1514 return -1;
1515 }
5f3633e3
JK		 1516 sbi->s_resuid = uid;
1517 } else if (token == Opt_resgid) {
1518 gid = make_kgid(current_user_ns(), arg);
1519 if (!gid_valid(gid)) {
1520 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1521 return -1;
1522 }
1523 sbi->s_resgid = gid;
1524 } else if (token == Opt_journal_dev) {
1525 if (is_remount) {
1526 ext4_msg(sb, KERN_ERR,
1527 "Cannot specify journal on remount");
1528 return -1;
1529 }
1530 *journal_devnum = arg;
ad4eec61
ES		 1531 } else if (token == Opt_journal_path) {
1532 char *journal_path;
1533 struct inode *journal_inode;
1534 struct path path;
1535 int error;
1536
1537 if (is_remount) {
1538 ext4_msg(sb, KERN_ERR,
1539 "Cannot specify journal on remount");
1540 return -1;
1541 }
1542 journal_path = match_strdup(&args[0]);
1543 if (!journal_path) {
1544 ext4_msg(sb, KERN_ERR, "error: could not dup "
1545 "journal device string");
1546 return -1;
1547 }
1548
1549 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1550 if (error) {
1551 ext4_msg(sb, KERN_ERR, "error: could not find "
1552 "journal device path: error %d", error);
1553 kfree(journal_path);
1554 return -1;
1555 }
1556
1557 journal_inode = path.dentry->d_inode;
1558 if (!S_ISBLK(journal_inode->i_mode)) {
1559 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1560 "is not a block device", journal_path);
1561 path_put(&path);
1562 kfree(journal_path);
1563 return -1;
1564 }
1565
1566 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1567 path_put(&path);
1568 kfree(journal_path);
5f3633e3
JK		 1569 } else if (token == Opt_journal_ioprio) {
1570 if (arg > 7) {
1571 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1572 " (must be 0-7)");
1573 return -1;
1574 }
1575 *journal_ioprio =
1576 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1577 } else if (m->flags & MOPT_DATAJ) {
1578 if (is_remount) {
1579 if (!sbi->s_journal)
1580 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1581 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1582 ext4_msg(sb, KERN_ERR,
26092bf5 1583 "Cannot change data mode on remount");
26092bf5 1584 return -1;
ac27a0ec 1585 }
26092bf5 1586 } else {
5f3633e3
JK		 1587 clear_opt(sb, DATA_FLAGS);
1588 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1589 }
5f3633e3
JK		 1590#ifdef CONFIG_QUOTA
1591 } else if (m->flags & MOPT_QFMT) {
1592 if (sb_any_quota_loaded(sb) &&
1593 sbi->s_jquota_fmt != m->mount_opt) {
1594 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1595 "quota options when quota turned on");
1596 return -1;
1597 }
262b4662
JK		 1598 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1599 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1600 ext4_msg(sb, KERN_ERR,
1601 "Cannot set journaled quota options "
1602 "when QUOTA feature is enabled");
1603 return -1;
1604 }
5f3633e3 1605 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff
RZ		 1606#endif
1607#ifndef CONFIG_FS_DAX
1608 } else if (token == Opt_dax) {
1609 ext4_msg(sb, KERN_INFO, "dax option not supported");
1610 return -1;
5f3633e3
JK		 1611#endif
1612 } else {
1613 if (!args->from)
1614 arg = 1;
1615 if (m->flags & MOPT_CLEAR)
1616 arg = !arg;
1617 else if (unlikely(!(m->flags & MOPT_SET))) {
1618 ext4_msg(sb, KERN_WARNING,
1619 "buggy handling of option %s", opt);
1620 WARN_ON(1);
1621 return -1;
1622 }
1623 if (arg != 0)
1624 sbi->s_mount_opt |= m->mount_opt;
1625 else
1626 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1627 }
5f3633e3 1628 return 1;
26092bf5
TT		 1629}
1630
1631static int parse_options(char *options, struct super_block *sb,
1632 unsigned long *journal_devnum,
1633 unsigned int *journal_ioprio,
1634 int is_remount)
1635{
1636 struct ext4_sb_info *sbi = EXT4_SB(sb);
1637 char *p;
1638 substring_t args[MAX_OPT_ARGS];
1639 int token;
1640
1641 if (!options)
1642 return 1;
1643
1644 while ((p = strsep(&options, ",")) != NULL) {
1645 if (!*p)
1646 continue;
1647 /*
1648 * Initialize args struct so we know whether arg was
1649 * found; some options take optional arguments.
1650 */
caecd0af 1651 args[0].to = args[0].from = NULL;
26092bf5
TT		 1652 token = match_token(p, tokens, args);
1653 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1654 journal_ioprio, is_remount) < 0)
1655 return 0;
ac27a0ec
DK		 1656 }
1657#ifdef CONFIG_QUOTA
262b4662
JK		 1658 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1659 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1660 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1661 "feature is enabled");
1662 return 0;
1663 }
ac27a0ec 1664 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1665 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1666 clear_opt(sb, USRQUOTA);
ac27a0ec 1667
482a7425 1668 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1669 clear_opt(sb, GRPQUOTA);
ac27a0ec 1670
56c50f11 1671 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1672 ext4_msg(sb, KERN_ERR, "old and new quota "
1673 "format mixing");
ac27a0ec
DK		 1674 return 0;
1675 }
1676
1677 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1678 ext4_msg(sb, KERN_ERR, "journaled quota format "
1679 "not specified");
ac27a0ec
DK		 1680 return 0;
1681 }
ac27a0ec
DK		 1682 }
1683#endif
261cb20c
JK		 1684 if (test_opt(sb, DIOREAD_NOLOCK)) {
1685 int blocksize =
1686 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1687
1688 if (blocksize < PAGE_CACHE_SIZE) {
1689 ext4_msg(sb, KERN_ERR, "can't mount with "
1690 "dioread_nolock if block size != PAGE_SIZE");
1691 return 0;
1692 }
1693 }
d4f76107
JK		 1694 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1695 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1696 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1697 "in data=ordered mode");
1698 return 0;
1699 }
ac27a0ec
DK		 1700 return 1;
1701}
1702
2adf6da8
TT		 1703static inline void ext4_show_quota_options(struct seq_file *seq,
1704 struct super_block *sb)
1705{
1706#if defined(CONFIG_QUOTA)
1707 struct ext4_sb_info *sbi = EXT4_SB(sb);
1708
1709 if (sbi->s_jquota_fmt) {
1710 char *fmtname = "";
1711
1712 switch (sbi->s_jquota_fmt) {
1713 case QFMT_VFS_OLD:
1714 fmtname = "vfsold";
1715 break;
1716 case QFMT_VFS_V0:
1717 fmtname = "vfsv0";
1718 break;
1719 case QFMT_VFS_V1:
1720 fmtname = "vfsv1";
1721 break;
1722 }
1723 seq_printf(seq, ",jqfmt=%s", fmtname);
1724 }
1725
1726 if (sbi->s_qf_names[USRQUOTA])
1727 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1728
1729 if (sbi->s_qf_names[GRPQUOTA])
1730 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1731#endif
1732}
1733
5a916be1
TT		 1734static const char *token2str(int token)
1735{
50df9fd5 1736 const struct match_token *t;
5a916be1
TT		 1737
1738 for (t = tokens; t->token != Opt_err; t++)
1739 if (t->token == token && !strchr(t->pattern, '='))
1740 break;
1741 return t->pattern;
1742}
1743
2adf6da8
TT		 1744/*
1745 * Show an option if
1746 * - it's set to a non-default value OR
1747 * - if the per-sb default is different from the global default
1748 */
66acdcf4
TT		 1749static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1750 int nodefs)
2adf6da8 1751{
2adf6da8
TT		 1752 struct ext4_sb_info *sbi = EXT4_SB(sb);
1753 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1754 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1755 const struct mount_opts *m;
66acdcf4 1756 char sep = nodefs ? '\n' : ',';
2adf6da8 1757
66acdcf4
TT		 1758#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1759#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1760
1761 if (sbi->s_sb_block != 1)
5a916be1
TT		 1762 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1763
1764 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1765 int want_set = m->flags & MOPT_SET;
1766 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1767 (m->flags & MOPT_CLEAR_ERR))
1768 continue;
66acdcf4 1769 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1770 continue; /* skip if same as the default */
1771 if ((want_set &&
1772 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1773 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1774 continue; /* select Opt_noFoo vs Opt_Foo */
1775 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1776 }
5a916be1 1777
08cefc7a 1778 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1779 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1780 SEQ_OPTS_PRINT("resuid=%u",
1781 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1782 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1783 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1784 SEQ_OPTS_PRINT("resgid=%u",
1785 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1786 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1787 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1788 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1789 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1790 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1791 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1792 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1793 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1794 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1795 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1796 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1797 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1798 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1799 if (sb->s_flags & MS_I_VERSION)
5a916be1 1800 SEQ_OPTS_PUTS("i_version");
66acdcf4 1801 if (nodefs || sbi->s_stripe)
5a916be1 1802 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1803 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1804 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1805 SEQ_OPTS_PUTS("data=journal");
1806 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1807 SEQ_OPTS_PUTS("data=ordered");
1808 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1809 SEQ_OPTS_PUTS("data=writeback");
1810 }
66acdcf4
TT		 1811 if (nodefs ||
1812 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1813 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1814 sbi->s_inode_readahead_blks);
2adf6da8 1815
66acdcf4
TT		 1816 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1817 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1818 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1819 if (nodefs || sbi->s_max_dir_size_kb)
1820 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1821
1822 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1823 return 0;
1824}
1825
66acdcf4
TT		 1826static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1827{
1828 return _ext4_show_options(seq, root->d_sb, 0);
1829}
1830
1831static int options_seq_show(struct seq_file *seq, void *offset)
1832{
1833 struct super_block *sb = seq->private;
1834 int rc;
1835
1836 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1837 rc = _ext4_show_options(seq, sb, 1);
1838 seq_puts(seq, "\n");
1839 return rc;
1840}
1841
1842static int options_open_fs(struct inode *inode, struct file *file)
1843{
d9dda78b 1844 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1845}
1846
1847static const struct file_operations ext4_seq_options_fops = {
1848 .owner = THIS_MODULE,
1849 .open = options_open_fs,
1850 .read = seq_read,
1851 .llseek = seq_lseek,
1852 .release = single_release,
1853};
1854
617ba13b 1855static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1856 int read_only)
1857{
617ba13b 1858 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1859 int res = 0;
1860
617ba13b 1861 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1862 ext4_msg(sb, KERN_ERR, "revision level too high, "
1863 "forcing read-only mode");
ac27a0ec
DK		 1864 res = MS_RDONLY;
1865 }
1866 if (read_only)
281b5995 1867 goto done;
617ba13b 1868 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1869 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1870 "running e2fsck is recommended");
c8b459f4 1871 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1872 ext4_msg(sb, KERN_WARNING,
1873 "warning: mounting fs with errors, "
1874 "running e2fsck is recommended");
ed3ce80a 1875 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1876 le16_to_cpu(es->s_mnt_count) >=
1877 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1878 ext4_msg(sb, KERN_WARNING,
1879 "warning: maximal mount count reached, "
1880 "running e2fsck is recommended");
ac27a0ec
DK		 1881 else if (le32_to_cpu(es->s_checkinterval) &&
1882 (le32_to_cpu(es->s_lastcheck) +
1883 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1884 ext4_msg(sb, KERN_WARNING,
1885 "warning: checktime reached, "
1886 "running e2fsck is recommended");
0b8e58a1 1887 if (!sbi->s_journal)
0390131b 1888 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1889 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1890 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1891 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1892 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1893 ext4_update_dynamic_rev(sb);
0390131b
FM		 1894 if (sbi->s_journal)
1895 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1896
e2d67052 1897 ext4_commit_super(sb, 1);
281b5995 1898done:
ac27a0ec 1899 if (test_opt(sb, DEBUG))
a9df9a49 1900 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1901 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1902 sb->s_blocksize,
1903 sbi->s_groups_count,
617ba13b
MC		 1904 EXT4_BLOCKS_PER_GROUP(sb),
1905 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1906 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1907
7abc52c2 1908 cleancache_init_fs(sb);
ac27a0ec
DK		 1909 return res;
1910}
1911
117fff10
TT		 1912int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1913{
1914 struct ext4_sb_info *sbi = EXT4_SB(sb);
1915 struct flex_groups *new_groups;
1916 int size;
1917
1918 if (!sbi->s_log_groups_per_flex)
1919 return 0;
1920
1921 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1922 if (size <= sbi->s_flex_groups_allocated)
1923 return 0;
1924
1925 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1926 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1927 if (!new_groups) {
1928 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1929 size / (int) sizeof(struct flex_groups));
1930 return -ENOMEM;
1931 }
1932
1933 if (sbi->s_flex_groups) {
1934 memcpy(new_groups, sbi->s_flex_groups,
1935 (sbi->s_flex_groups_allocated *
1936 sizeof(struct flex_groups)));
b93b41d4 1937 kvfree(sbi->s_flex_groups);
117fff10
TT		 1938 }
1939 sbi->s_flex_groups = new_groups;
1940 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1941 return 0;
1942}
1943
772cb7c8
JS		 1944static int ext4_fill_flex_info(struct super_block *sb)
1945{
1946 struct ext4_sb_info *sbi = EXT4_SB(sb);
1947 struct ext4_group_desc *gdp = NULL;
772cb7c8 1948 ext4_group_t flex_group;
117fff10 1949 int i, err;
772cb7c8 1950
503358ae 1951 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1952 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1953 sbi->s_log_groups_per_flex = 0;
1954 return 1;
1955 }
1956
117fff10
TT		 1957 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1958 if (err)
9933fc0a 1959 goto failed;
772cb7c8 1960
772cb7c8 1961 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1962 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1963
1964 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1965 atomic_add(ext4_free_inodes_count(sb, gdp),
1966 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 1967 atomic64_add(ext4_free_group_clusters(sb, gdp),
1968 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1969 atomic_add(ext4_used_dirs_count(sb, gdp),
1970 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1971 }
1972
1973 return 1;
1974failed:
1975 return 0;
1976}
1977
feb0ab32
DW		 1978static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1979 struct ext4_group_desc *gdp)
717d50e4 1980{
feb0ab32 1981 int offset;
717d50e4 1982 __u16 crc = 0;
feb0ab32 1983 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1984
9aa5d32b 1985 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 1986 /* Use new metadata_csum algorithm */
d6a77105 1987 __le16 save_csum;
feb0ab32
DW		 1988 __u32 csum32;
1989
d6a77105 1990 save_csum = gdp->bg_checksum;
feb0ab32
DW		 1991 gdp->bg_checksum = 0;
1992 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1993 sizeof(le_group));
1994 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1995 sbi->s_desc_size);
d6a77105 1996 gdp->bg_checksum = save_csum;
feb0ab32
DW		 1997
1998 crc = csum32 & 0xFFFF;
1999 goto out;
717d50e4
AD		 2000 }
2001
feb0ab32 2002 /* old crc16 code */
813d32f9
DW		 2003 if (!(sbi->s_es->s_feature_ro_compat &
2004 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2005 return 0;
2006
feb0ab32
DW		 2007 offset = offsetof(struct ext4_group_desc, bg_checksum);
2008
2009 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2010 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2011 crc = crc16(crc, (__u8 *)gdp, offset);
2012 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2013 /* for checksum of struct ext4_group_desc do the rest...*/
2014 if ((sbi->s_es->s_feature_incompat &
2015 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2016 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2017 crc = crc16(crc, (__u8 *)gdp + offset,
2018 le16_to_cpu(sbi->s_es->s_desc_size) -
2019 offset);
2020
2021out:
717d50e4
AD		 2022 return cpu_to_le16(crc);
2023}
2024
feb0ab32 2025int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2026 struct ext4_group_desc *gdp)
2027{
feb0ab32
DW		 2028 if (ext4_has_group_desc_csum(sb) &&
2029 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2030 block_group, gdp)))
717d50e4
AD		 2031 return 0;
2032
2033 return 1;
2034}
2035
feb0ab32
DW		 2036void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2037 struct ext4_group_desc *gdp)
2038{
2039 if (!ext4_has_group_desc_csum(sb))
2040 return;
2041 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2042}
2043
ac27a0ec 2044/* Called at mount-time, super-block is locked */
bfff6873
LC		 2045static int ext4_check_descriptors(struct super_block *sb,
2046 ext4_group_t *first_not_zeroed)
ac27a0ec 2047{
617ba13b
MC		 2048 struct ext4_sb_info *sbi = EXT4_SB(sb);
2049 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2050 ext4_fsblk_t last_block;
bd81d8ee
LV		 2051 ext4_fsblk_t block_bitmap;
2052 ext4_fsblk_t inode_bitmap;
2053 ext4_fsblk_t inode_table;
ce421581 2054 int flexbg_flag = 0;
bfff6873 2055 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2056
ce421581
JS		 2057 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2058 flexbg_flag = 1;
2059
af5bc92d 2060 ext4_debug("Checking group descriptors");
ac27a0ec 2061
197cd65a
AM		 2062 for (i = 0; i < sbi->s_groups_count; i++) {
2063 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2064
ce421581 2065 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2066 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2067 else
2068 last_block = first_block +
617ba13b 2069 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2070
bfff6873
LC		 2071 if ((grp == sbi->s_groups_count) &&
2072 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2073 grp = i;
2074
8fadc143 2075 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2076 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2077 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2078 "Block bitmap for group %u not in group "
b31e1552 2079 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2080 return 0;
2081 }
8fadc143 2082 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2083 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2084 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2085 "Inode bitmap for group %u not in group "
b31e1552 2086 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2087 return 0;
2088 }
8fadc143 2089 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2090 if (inode_table < first_block ||
2b2d6d01 2091 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2092 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2093 "Inode table for group %u not in group "
b31e1552 2094 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2095 return 0;
2096 }
955ce5f5 2097 ext4_lock_group(sb, i);
feb0ab32 2098 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2099 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2100 "Checksum for group %u failed (%u!=%u)",
2101 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2102 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2103 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2104 ext4_unlock_group(sb, i);
8a266467 2105 return 0;
7ee1ec4c 2106 }
717d50e4 2107 }
955ce5f5 2108 ext4_unlock_group(sb, i);
ce421581
JS		 2109 if (!flexbg_flag)
2110 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2111 }
bfff6873
LC		 2112 if (NULL != first_not_zeroed)
2113 *first_not_zeroed = grp;
ac27a0ec
DK		 2114 return 1;
2115}
2116
617ba13b 2117/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2118 * the superblock) which were deleted from all directories, but held open by
2119 * a process at the time of a crash. We walk the list and try to delete these
2120 * inodes at recovery time (only with a read-write filesystem).
2121 *
2122 * In order to keep the orphan inode chain consistent during traversal (in
2123 * case of crash during recovery), we link each inode into the superblock
2124 * orphan list_head and handle it the same way as an inode deletion during
2125 * normal operation (which journals the operations for us).
2126 *
2127 * We only do an iget() and an iput() on each inode, which is very safe if we
2128 * accidentally point at an in-use or already deleted inode. The worst that
2129 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2130 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2131 * e2fsck was run on this filesystem, and it must have already done the orphan
2132 * inode cleanup for us, so we can safely abort without any further action.
2133 */
2b2d6d01
TT		 2134static void ext4_orphan_cleanup(struct super_block *sb,
2135 struct ext4_super_block *es)
ac27a0ec
DK		 2136{
2137 unsigned int s_flags = sb->s_flags;
2138 int nr_orphans = 0, nr_truncates = 0;
2139#ifdef CONFIG_QUOTA
2140 int i;
2141#endif
2142 if (!es->s_last_orphan) {
2143 jbd_debug(4, "no orphan inodes to clean up\n");
2144 return;
2145 }
2146
a8f48a95 2147 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2148 ext4_msg(sb, KERN_ERR, "write access "
2149 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2150 return;
2151 }
2152
d39195c3
AG		 2153 /* Check if feature set would not allow a r/w mount */
2154 if (!ext4_feature_set_ok(sb, 0)) {
2155 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2156 "unknown ROCOMPAT features");
2157 return;
2158 }
2159
617ba13b 2160 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2161 /* don't clear list on RO mount w/ errors */
2162 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2163 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2164 "clearing orphan list.\n");
c25f9bc6
ES		 2165 es->s_last_orphan = 0;
2166 }
ac27a0ec
DK		 2167 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2168 return;
2169 }
2170
2171 if (s_flags & MS_RDONLY) {
b31e1552 2172 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2173 sb->s_flags &= ~MS_RDONLY;
2174 }
2175#ifdef CONFIG_QUOTA
2176 /* Needed for iput() to work correctly and not trash data */
2177 sb->s_flags |= MS_ACTIVE;
2178 /* Turn on quotas so that they are updated correctly */
a2d4a646 2179 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2180 if (EXT4_SB(sb)->s_qf_names[i]) {
2181 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2182 if (ret < 0)
b31e1552
ES		 2183 ext4_msg(sb, KERN_ERR,
2184 "Cannot turn on journaled "
2185 "quota: error %d", ret);
ac27a0ec
DK		 2186 }
2187 }
2188#endif
2189
2190 while (es->s_last_orphan) {
2191 struct inode *inode;
2192
97bd42b9
JB		 2193 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2194 if (IS_ERR(inode)) {
ac27a0ec
DK		 2195 es->s_last_orphan = 0;
2196 break;
2197 }
2198
617ba13b 2199 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2200 dquot_initialize(inode);
ac27a0ec 2201 if (inode->i_nlink) {
566370a2
PT		 2202 if (test_opt(sb, DEBUG))
2203 ext4_msg(sb, KERN_DEBUG,
2204 "%s: truncating inode %lu to %lld bytes",
2205 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2206 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2207 inode->i_ino, inode->i_size);
721e3eba 2208 mutex_lock(&inode->i_mutex);
55f252c9 2209 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2210 ext4_truncate(inode);
721e3eba 2211 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2212 nr_truncates++;
2213 } else {
566370a2
PT		 2214 if (test_opt(sb, DEBUG))
2215 ext4_msg(sb, KERN_DEBUG,
2216 "%s: deleting unreferenced inode %lu",
2217 __func__, inode->i_ino);
ac27a0ec
DK		 2218 jbd_debug(2, "deleting unreferenced inode %lu\n",
2219 inode->i_ino);
2220 nr_orphans++;
2221 }
2222 iput(inode); /* The delete magic happens here! */
2223 }
2224
2b2d6d01 2225#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2226
2227 if (nr_orphans)
b31e1552
ES		 2228 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2229 PLURAL(nr_orphans));
ac27a0ec 2230 if (nr_truncates)
b31e1552
ES		 2231 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2232 PLURAL(nr_truncates));
ac27a0ec
DK		 2233#ifdef CONFIG_QUOTA
2234 /* Turn quotas off */
a2d4a646 2235 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2236 if (sb_dqopt(sb)->files[i])
287a8095 2237 dquot_quota_off(sb, i);
ac27a0ec
DK		 2238 }
2239#endif
2240 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2241}
0b8e58a1 2242
cd2291a4
ES		 2243/*
2244 * Maximal extent format file size.
2245 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2246 * extent format containers, within a sector_t, and within i_blocks
2247 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2248 * so that won't be a limiting factor.
2249 *
f17722f9
LC		 2250 * However there is other limiting factor. We do store extents in the form
2251 * of starting block and length, hence the resulting length of the extent
2252 * covering maximum file size must fit into on-disk format containers as
2253 * well. Given that length is always by 1 unit bigger than max unit (because
2254 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2255 *
cd2291a4
ES		 2256 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2257 */
f287a1a5 2258static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2259{
2260 loff_t res;
2261 loff_t upper_limit = MAX_LFS_FILESIZE;
2262
2263 /* small i_blocks in vfs inode? */
f287a1a5 2264 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2265 /*
90c699a9 2266 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2267 * i_block represent total blocks in 512 bytes
2268 * 32 == size of vfs inode i_blocks * 8
2269 */
2270 upper_limit = (1LL << 32) - 1;
2271
2272 /* total blocks in file system block size */
2273 upper_limit >>= (blkbits - 9);
2274 upper_limit <<= blkbits;
2275 }
2276
f17722f9
LC		 2277 /*
2278 * 32-bit extent-start container, ee_block. We lower the maxbytes
2279 * by one fs block, so ee_len can cover the extent of maximum file
2280 * size
2281 */
2282 res = (1LL << 32) - 1;
cd2291a4 2283 res <<= blkbits;
cd2291a4
ES		 2284
2285 /* Sanity check against vm- & vfs- imposed limits */
2286 if (res > upper_limit)
2287 res = upper_limit;
2288
2289 return res;
2290}
ac27a0ec 2291
ac27a0ec 2292/*
cd2291a4 2293 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2294 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2295 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2296 */
f287a1a5 2297static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2298{
617ba13b 2299 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2300 int meta_blocks;
2301 loff_t upper_limit;
0b8e58a1
AD		 2302 /* This is calculated to be the largest file size for a dense, block
2303 * mapped file such that the file's total number of 512-byte sectors,
2304 * including data and all indirect blocks, does not exceed (2^48 - 1).
2305 *
2306 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2307 * number of 512-byte sectors of the file.
0fc1b451
AK		 2308 */
2309
f287a1a5 2310 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2311 /*
90c699a9 2312 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2313 * the inode i_block field represents total file blocks in
2314 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2315 */
2316 upper_limit = (1LL << 32) - 1;
2317
2318 /* total blocks in file system block size */
2319 upper_limit >>= (bits - 9);
2320
2321 } else {
8180a562
AK		 2322 /*
2323 * We use 48 bit ext4_inode i_blocks
2324 * With EXT4_HUGE_FILE_FL set the i_blocks
2325 * represent total number of blocks in
2326 * file system block size
2327 */
0fc1b451
AK		 2328 upper_limit = (1LL << 48) - 1;
2329
0fc1b451
AK		 2330 }
2331
2332 /* indirect blocks */
2333 meta_blocks = 1;
2334 /* double indirect blocks */
2335 meta_blocks += 1 + (1LL << (bits-2));
2336 /* tripple indirect blocks */
2337 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2338
2339 upper_limit -= meta_blocks;
2340 upper_limit <<= bits;
ac27a0ec
DK		 2341
2342 res += 1LL << (bits-2);
2343 res += 1LL << (2*(bits-2));
2344 res += 1LL << (3*(bits-2));
2345 res <<= bits;
2346 if (res > upper_limit)
2347 res = upper_limit;
0fc1b451
AK		 2348
2349 if (res > MAX_LFS_FILESIZE)
2350 res = MAX_LFS_FILESIZE;
2351
ac27a0ec
DK		 2352 return res;
2353}
2354
617ba13b 2355static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2356 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2357{
617ba13b 2358 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2359 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2360 int has_super = 0;
2361
2362 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2363
617ba13b 2364 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2365 nr < first_meta_bg)
70bbb3e0 2366 return logical_sb_block + nr + 1;
ac27a0ec 2367 bg = sbi->s_desc_per_block * nr;
617ba13b 2368 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2369 has_super = 1;
0b8e58a1 2370
bd63f6b0
DW		 2371 /*
2372 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2373 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2374 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2375 * compensate.
2376 */
2377 if (sb->s_blocksize == 1024 && nr == 0 &&
2378 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2379 has_super++;
2380
617ba13b 2381 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2382}
2383
c9de560d
AT		 2384/**
2385 * ext4_get_stripe_size: Get the stripe size.
2386 * @sbi: In memory super block info
2387 *
2388 * If we have specified it via mount option, then
2389 * use the mount option value. If the value specified at mount time is
2390 * greater than the blocks per group use the super block value.
2391 * If the super block value is greater than blocks per group return 0.
2392 * Allocator needs it be less than blocks per group.
2393 *
2394 */
2395static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2396{
2397 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2398 unsigned long stripe_width =
2399 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2400 int ret;
c9de560d
AT		 2401
2402 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2403 ret = sbi->s_stripe;
2404 else if (stripe_width <= sbi->s_blocks_per_group)
2405 ret = stripe_width;
2406 else if (stride <= sbi->s_blocks_per_group)
2407 ret = stride;
2408 else
2409 ret = 0;
c9de560d 2410
3eb08658
DE		 2411 /*
2412 * If the stripe width is 1, this makes no sense and
2413 * we set it to 0 to turn off stripe handling code.
2414 */
2415 if (ret <= 1)
2416 ret = 0;
c9de560d 2417
3eb08658 2418 return ret;
c9de560d 2419}
ac27a0ec 2420
3197ebdb
TT		 2421/* sysfs supprt */
2422
2423struct ext4_attr {
2424 struct attribute attr;
2425 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2426 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb 2427 const char *, size_t);
f2d50a65
JK		 2428 union {
2429 int offset;
2430 int deprecated_val;
2431 } u;
3197ebdb
TT		 2432};
2433
27dd4385
LC		 2434static int parse_strtoull(const char *buf,
2435 unsigned long long max, unsigned long long *value)
2436{
2437 int ret;
2438
2439 ret = kstrtoull(skip_spaces(buf), 0, value);
2440 if (!ret && *value > max)
2441 ret = -EINVAL;
2442 return ret;
2443}
2444
3197ebdb
TT		 2445static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2446 struct ext4_sb_info *sbi,
2447 char *buf)
2448{
2449 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2450 (s64) EXT4_C2B(sbi,
2451 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2452}
2453
2454static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2455 struct ext4_sb_info *sbi, char *buf)
2456{
2457 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2458
f613dfcb
TT		 2459 if (!sb->s_bdev->bd_part)
2460 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2461 return snprintf(buf, PAGE_SIZE, "%lu\n",
2462 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2463 sbi->s_sectors_written_start) >> 1);
2464}
2465
2466static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2467 struct ext4_sb_info *sbi, char *buf)
2468{
2469 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2470
f613dfcb
TT		 2471 if (!sb->s_bdev->bd_part)
2472 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2473 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2474 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2475 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2476 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2477}
2478
2479static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2480 struct ext4_sb_info *sbi,
2481 const char *buf, size_t count)
2482{
2483 unsigned long t;
e1091b15 2484 int ret;
3197ebdb 2485
e1091b15
LC		 2486 ret = kstrtoul(skip_spaces(buf), 0, &t);
2487 if (ret)
2488 return ret;
3197ebdb 2489
e1091b15 2490 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2491 return -EINVAL;
2492
2493 sbi->s_inode_readahead_blks = t;
2494 return count;
2495}
2496
2497static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2498 struct ext4_sb_info *sbi, char *buf)
3197ebdb 2499{
f2d50a65 2500 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb
TT		 2501
2502 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2503}
2504
2505static ssize_t sbi_ui_store(struct ext4_attr *a,
2506 struct ext4_sb_info *sbi,
2507 const char *buf, size_t count)
2508{
f2d50a65 2509 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb 2510 unsigned long t;
e1091b15 2511 int ret;
3197ebdb 2512
e1091b15
LC		 2513 ret = kstrtoul(skip_spaces(buf), 0, &t);
2514 if (ret)
2515 return ret;
3197ebdb
TT		 2516 *ui = t;
2517 return count;
2518}
2519
52c198c6
LC		 2520static ssize_t es_ui_show(struct ext4_attr *a,
2521 struct ext4_sb_info *sbi, char *buf)
2522{
2523
2524 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2525 a->u.offset);
2526
2527 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2528}
2529
27dd4385
LC		 2530static ssize_t reserved_clusters_show(struct ext4_attr *a,
2531 struct ext4_sb_info *sbi, char *buf)
2532{
2533 return snprintf(buf, PAGE_SIZE, "%llu\n",
2534 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2535}
2536
2537static ssize_t reserved_clusters_store(struct ext4_attr *a,
2538 struct ext4_sb_info *sbi,
2539 const char *buf, size_t count)
2540{
2541 unsigned long long val;
2542 int ret;
2543
2544 if (parse_strtoull(buf, -1ULL, &val))
2545 return -EINVAL;
2546 ret = ext4_reserve_clusters(sbi, val);
2547
2548 return ret ? ret : count;
2549}
2550
2c0544b2
TT		 2551static ssize_t trigger_test_error(struct ext4_attr *a,
2552 struct ext4_sb_info *sbi,
2553 const char *buf, size_t count)
2554{
2555 int len = count;
2556
2557 if (!capable(CAP_SYS_ADMIN))
2558 return -EPERM;
2559
2560 if (len && buf[len-1] == '\n')
2561 len--;
2562
2563 if (len)
2564 ext4_error(sbi->s_sb, "%.*s", len, buf);
2565 return count;
2566}
2567
f2d50a65
JK		 2568static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2569 struct ext4_sb_info *sbi, char *buf)
2570{
2571 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2572}
2573
3197ebdb
TT		 2574#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2575static struct ext4_attr ext4_attr_##_name = { \
2576 .attr = {.name = __stringify(_name), .mode = _mode }, \
2577 .show = _show, \
2578 .store = _store, \
f2d50a65
JK		 2579 .u = { \
2580 .offset = offsetof(struct ext4_sb_info, _elname),\
2581 }, \
3197ebdb 2582}
52c198c6
LC		 2583
2584#define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2585static struct ext4_attr ext4_attr_##_name = { \
2586 .attr = {.name = __stringify(_name), .mode = _mode }, \
2587 .show = _show, \
2588 .store = _store, \
2589 .u = { \
2590 .offset = offsetof(struct ext4_super_block, _elname), \
2591 }, \
2592}
2593
3197ebdb
TT		 2594#define EXT4_ATTR(name, mode, show, store) \
2595static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2596
857ac889 2597#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2598#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2599#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
52c198c6
LC		 2600
2601#define EXT4_RO_ATTR_ES_UI(name, elname) \
2602 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
3197ebdb
TT		 2603#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2604 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
52c198c6 2605
3197ebdb 2606#define ATTR_LIST(name) &ext4_attr_##name.attr
f2d50a65
JK		 2607#define EXT4_DEPRECATED_ATTR(_name, _val) \
2608static struct ext4_attr ext4_attr_##_name = { \
2609 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2610 .show = sbi_deprecated_show, \
2611 .u = { \
2612 .deprecated_val = _val, \
2613 }, \
2614}
3197ebdb
TT		 2615
2616EXT4_RO_ATTR(delayed_allocation_blocks);
2617EXT4_RO_ATTR(session_write_kbytes);
2618EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2619EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2620EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2621 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2622EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2623EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2624EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2625EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2626EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2627EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2628EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
f2d50a65 2629EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
67a5da56 2630EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2631EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
efbed4dc
TT		 2632EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2633EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2634EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2635EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2636EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2637EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
52c198c6
LC		 2638EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2639EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2640EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
3197ebdb
TT		 2641
2642static struct attribute *ext4_attrs[] = {
2643 ATTR_LIST(delayed_allocation_blocks),
2644 ATTR_LIST(session_write_kbytes),
2645 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2646 ATTR_LIST(reserved_clusters),
3197ebdb 2647 ATTR_LIST(inode_readahead_blks),
11013911 2648 ATTR_LIST(inode_goal),
3197ebdb
TT		 2649 ATTR_LIST(mb_stats),
2650 ATTR_LIST(mb_max_to_scan),
2651 ATTR_LIST(mb_min_to_scan),
2652 ATTR_LIST(mb_order2_req),
2653 ATTR_LIST(mb_stream_req),
2654 ATTR_LIST(mb_group_prealloc),
55138e0b 2655 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2656 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2657 ATTR_LIST(trigger_fs_error),
efbed4dc
TT		 2658 ATTR_LIST(err_ratelimit_interval_ms),
2659 ATTR_LIST(err_ratelimit_burst),
2660 ATTR_LIST(warning_ratelimit_interval_ms),
2661 ATTR_LIST(warning_ratelimit_burst),
2662 ATTR_LIST(msg_ratelimit_interval_ms),
2663 ATTR_LIST(msg_ratelimit_burst),
52c198c6
LC		 2664 ATTR_LIST(errors_count),
2665 ATTR_LIST(first_error_time),
2666 ATTR_LIST(last_error_time),
3197ebdb
TT		 2667 NULL,
2668};
2669
857ac889
LC		 2670/* Features this copy of ext4 supports */
2671EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2672EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2673EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2674
2675static struct attribute *ext4_feat_attrs[] = {
2676 ATTR_LIST(lazy_itable_init),
27ee40df 2677 ATTR_LIST(batched_discard),
5e7bbef1 2678 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2679 NULL,
2680};
2681
3197ebdb
TT		 2682static ssize_t ext4_attr_show(struct kobject *kobj,
2683 struct attribute *attr, char *buf)
2684{
2685 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2686 s_kobj);
2687 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2688
2689 return a->show ? a->show(a, sbi, buf) : 0;
2690}
2691
2692static ssize_t ext4_attr_store(struct kobject *kobj,
2693 struct attribute *attr,
2694 const char *buf, size_t len)
2695{
2696 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2697 s_kobj);
2698 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2699
2700 return a->store ? a->store(a, sbi, buf, len) : 0;
2701}
2702
2703static void ext4_sb_release(struct kobject *kobj)
2704{
2705 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2706 s_kobj);
2707 complete(&sbi->s_kobj_unregister);
2708}
2709
52cf25d0 2710static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2711 .show = ext4_attr_show,
2712 .store = ext4_attr_store,
2713};
2714
2715static struct kobj_type ext4_ktype = {
2716 .default_attrs = ext4_attrs,
2717 .sysfs_ops = &ext4_attr_ops,
2718 .release = ext4_sb_release,
2719};
2720
857ac889
LC		 2721static void ext4_feat_release(struct kobject *kobj)
2722{
2723 complete(&ext4_feat->f_kobj_unregister);
2724}
2725
c7f72543
LC		 2726static ssize_t ext4_feat_show(struct kobject *kobj,
2727 struct attribute *attr, char *buf)
2728{
2729 return snprintf(buf, PAGE_SIZE, "supported\n");
2730}
2731
2732/*
2733 * We can not use ext4_attr_show/store because it relies on the kobject
2734 * being embedded in the ext4_sb_info structure which is definitely not
2735 * true in this case.
2736 */
2737static const struct sysfs_ops ext4_feat_ops = {
2738 .show = ext4_feat_show,
2739 .store = NULL,
2740};
2741
857ac889
LC		 2742static struct kobj_type ext4_feat_ktype = {
2743 .default_attrs = ext4_feat_attrs,
c7f72543 2744 .sysfs_ops = &ext4_feat_ops,
857ac889
LC		 2745 .release = ext4_feat_release,
2746};
2747
a13fb1a4
ES		 2748/*
2749 * Check whether this filesystem can be mounted based on
2750 * the features present and the RDONLY/RDWR mount requested.
2751 * Returns 1 if this filesystem can be mounted as requested,
2752 * 0 if it cannot be.
2753 */
2754static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2755{
2756 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2757 ext4_msg(sb, KERN_ERR,
2758 "Couldn't mount because of "
2759 "unsupported optional features (%x)",
2760 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2761 ~EXT4_FEATURE_INCOMPAT_SUPP));
2762 return 0;
2763 }
2764
2765 if (readonly)
2766 return 1;
2767
2cb5cc8b
DW		 2768 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2769 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2770 sb->s_flags |= MS_RDONLY;
2771 return 1;
2772 }
2773
a13fb1a4
ES		 2774 /* Check that feature set is OK for a read-write mount */
2775 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2776 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2777 "unsupported optional features (%x)",
2778 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2779 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2780 return 0;
2781 }
2782 /*
2783 * Large file size enabled file system can only be mounted
2784 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2785 */
2786 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2787 if (sizeof(blkcnt_t) < sizeof(u64)) {
2788 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2789 "cannot be mounted RDWR without "
2790 "CONFIG_LBDAF");
2791 return 0;
2792 }
2793 }
bab08ab9
TT		 2794 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2795 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2796 ext4_msg(sb, KERN_ERR,
2797 "Can't support bigalloc feature without "
2798 "extents feature\n");
2799 return 0;
2800 }
7c319d32
AK		 2801
2802#ifndef CONFIG_QUOTA
2803 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2804 !readonly) {
2805 ext4_msg(sb, KERN_ERR,
2806 "Filesystem with quota feature cannot be mounted RDWR "
2807 "without CONFIG_QUOTA");
2808 return 0;
2809 }
2810#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2811 return 1;
2812}
2813
66e61a9e
TT		 2814/*
2815 * This function is called once a day if we have errors logged
2816 * on the file system
2817 */
2818static void print_daily_error_info(unsigned long arg)
2819{
2820 struct super_block *sb = (struct super_block *) arg;
2821 struct ext4_sb_info *sbi;
2822 struct ext4_super_block *es;
2823
2824 sbi = EXT4_SB(sb);
2825 es = sbi->s_es;
2826
2827 if (es->s_error_count)
ae0f78de
TT		 2828 /* fsck newer than v1.41.13 is needed to clean this condition. */
2829 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2830 le32_to_cpu(es->s_error_count));
2831 if (es->s_first_error_time) {
ae0f78de 2832 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2833 sb->s_id, le32_to_cpu(es->s_first_error_time),
2834 (int) sizeof(es->s_first_error_func),
2835 es->s_first_error_func,
2836 le32_to_cpu(es->s_first_error_line));
2837 if (es->s_first_error_ino)
2838 printk(": inode %u",
2839 le32_to_cpu(es->s_first_error_ino));
2840 if (es->s_first_error_block)
2841 printk(": block %llu", (unsigned long long)
2842 le64_to_cpu(es->s_first_error_block));
2843 printk("\n");
2844 }
2845 if (es->s_last_error_time) {
ae0f78de 2846 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2847 sb->s_id, le32_to_cpu(es->s_last_error_time),
2848 (int) sizeof(es->s_last_error_func),
2849 es->s_last_error_func,
2850 le32_to_cpu(es->s_last_error_line));
2851 if (es->s_last_error_ino)
2852 printk(": inode %u",
2853 le32_to_cpu(es->s_last_error_ino));
2854 if (es->s_last_error_block)
2855 printk(": block %llu", (unsigned long long)
2856 le64_to_cpu(es->s_last_error_block));
2857 printk("\n");
2858 }
2859 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2860}
2861
bfff6873
LC		 2862/* Find next suitable group and run ext4_init_inode_table */
2863static int ext4_run_li_request(struct ext4_li_request *elr)
2864{
2865 struct ext4_group_desc *gdp = NULL;
2866 ext4_group_t group, ngroups;
2867 struct super_block *sb;
2868 unsigned long timeout = 0;
2869 int ret = 0;
2870
2871 sb = elr->lr_super;
2872 ngroups = EXT4_SB(sb)->s_groups_count;
2873
8e8ad8a5 2874 sb_start_write(sb);
bfff6873
LC		 2875 for (group = elr->lr_next_group; group < ngroups; group++) {
2876 gdp = ext4_get_group_desc(sb, group, NULL);
2877 if (!gdp) {
2878 ret = 1;
2879 break;
2880 }
2881
2882 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2883 break;
2884 }
2885
7f511862 2886 if (group >= ngroups)
bfff6873
LC		 2887 ret = 1;
2888
2889 if (!ret) {
2890 timeout = jiffies;
2891 ret = ext4_init_inode_table(sb, group,
2892 elr->lr_timeout ? 0 : 1);
2893 if (elr->lr_timeout == 0) {
51ce6511
LC		 2894 timeout = (jiffies - timeout) *
2895 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2896 elr->lr_timeout = timeout;
2897 }
2898 elr->lr_next_sched = jiffies + elr->lr_timeout;
2899 elr->lr_next_group = group + 1;
2900 }
8e8ad8a5 2901 sb_end_write(sb);
bfff6873
LC		 2902
2903 return ret;
2904}
2905
2906/*
2907 * Remove lr_request from the list_request and free the
4ed5c033 2908 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2909 */
2910static void ext4_remove_li_request(struct ext4_li_request *elr)
2911{
2912 struct ext4_sb_info *sbi;
2913
2914 if (!elr)
2915 return;
2916
2917 sbi = elr->lr_sbi;
2918
2919 list_del(&elr->lr_request);
2920 sbi->s_li_request = NULL;
2921 kfree(elr);
2922}
2923
2924static void ext4_unregister_li_request(struct super_block *sb)
2925{
1bb933fb
LC		 2926 mutex_lock(&ext4_li_mtx);
2927 if (!ext4_li_info) {
2928 mutex_unlock(&ext4_li_mtx);
bfff6873 2929 return;
1bb933fb 2930 }
bfff6873
LC		 2931
2932 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2933 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2934 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2935 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2936}
2937
8f1f7453
ES		 2938static struct task_struct *ext4_lazyinit_task;
2939
bfff6873
LC		 2940/*
2941 * This is the function where ext4lazyinit thread lives. It walks
2942 * through the request list searching for next scheduled filesystem.
2943 * When such a fs is found, run the lazy initialization request
2944 * (ext4_rn_li_request) and keep track of the time spend in this
2945 * function. Based on that time we compute next schedule time of
2946 * the request. When walking through the list is complete, compute
2947 * next waking time and put itself into sleep.
2948 */
2949static int ext4_lazyinit_thread(void *arg)
2950{
2951 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2952 struct list_head *pos, *n;
2953 struct ext4_li_request *elr;
4ed5c033 2954 unsigned long next_wakeup, cur;
bfff6873
LC		 2955
2956 BUG_ON(NULL == eli);
2957
bfff6873
LC		 2958cont_thread:
2959 while (true) {
2960 next_wakeup = MAX_JIFFY_OFFSET;
2961
2962 mutex_lock(&eli->li_list_mtx);
2963 if (list_empty(&eli->li_request_list)) {
2964 mutex_unlock(&eli->li_list_mtx);
2965 goto exit_thread;
2966 }
2967
2968 list_for_each_safe(pos, n, &eli->li_request_list) {
2969 elr = list_entry(pos, struct ext4_li_request,
2970 lr_request);
2971
b2c78cd0
TT		 2972 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2973 if (ext4_run_li_request(elr) != 0) {
2974 /* error, remove the lazy_init job */
2975 ext4_remove_li_request(elr);
2976 continue;
2977 }
bfff6873
LC		 2978 }
2979
2980 if (time_before(elr->lr_next_sched, next_wakeup))
2981 next_wakeup = elr->lr_next_sched;
2982 }
2983 mutex_unlock(&eli->li_list_mtx);
2984
a0acae0e 2985 try_to_freeze();
bfff6873 2986
4ed5c033
LC		 2987 cur = jiffies;
2988 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2989 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2990 cond_resched();
2991 continue;
2992 }
2993
4ed5c033
LC		 2994 schedule_timeout_interruptible(next_wakeup - cur);
2995
8f1f7453
ES		 2996 if (kthread_should_stop()) {
2997 ext4_clear_request_list();
2998 goto exit_thread;
2999 }
bfff6873
LC		 3000 }
3001
3002exit_thread:
3003 /*
3004 * It looks like the request list is empty, but we need
3005 * to check it under the li_list_mtx lock, to prevent any
3006 * additions into it, and of course we should lock ext4_li_mtx
3007 * to atomically free the list and ext4_li_info, because at
3008 * this point another ext4 filesystem could be registering
3009 * new one.
3010 */
3011 mutex_lock(&ext4_li_mtx);
3012 mutex_lock(&eli->li_list_mtx);
3013 if (!list_empty(&eli->li_request_list)) {
3014 mutex_unlock(&eli->li_list_mtx);
3015 mutex_unlock(&ext4_li_mtx);
3016 goto cont_thread;
3017 }
3018 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3019 kfree(ext4_li_info);
3020 ext4_li_info = NULL;
3021 mutex_unlock(&ext4_li_mtx);
3022
3023 return 0;
3024}
3025
3026static void ext4_clear_request_list(void)
3027{
3028 struct list_head *pos, *n;
3029 struct ext4_li_request *elr;
3030
3031 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3032 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3033 elr = list_entry(pos, struct ext4_li_request,
3034 lr_request);
3035 ext4_remove_li_request(elr);
3036 }
3037 mutex_unlock(&ext4_li_info->li_list_mtx);
3038}
3039
3040static int ext4_run_lazyinit_thread(void)
3041{
8f1f7453
ES		 3042 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3043 ext4_li_info, "ext4lazyinit");
3044 if (IS_ERR(ext4_lazyinit_task)) {
3045 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3046 ext4_clear_request_list();
bfff6873
LC		 3047 kfree(ext4_li_info);
3048 ext4_li_info = NULL;
92b97816 3049 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3050 "initialization thread\n",
3051 err);
3052 return err;
3053 }
3054 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3055 return 0;
3056}
3057
3058/*
3059 * Check whether it make sense to run itable init. thread or not.
3060 * If there is at least one uninitialized inode table, return
3061 * corresponding group number, else the loop goes through all
3062 * groups and return total number of groups.
3063 */
3064static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3065{
3066 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3067 struct ext4_group_desc *gdp = NULL;
3068
3069 for (group = 0; group < ngroups; group++) {
3070 gdp = ext4_get_group_desc(sb, group, NULL);
3071 if (!gdp)
3072 continue;
3073
3074 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3075 break;
3076 }
3077
3078 return group;
3079}
3080
3081static int ext4_li_info_new(void)
3082{
3083 struct ext4_lazy_init *eli = NULL;
3084
3085 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3086 if (!eli)
3087 return -ENOMEM;
3088
bfff6873
LC		 3089 INIT_LIST_HEAD(&eli->li_request_list);
3090 mutex_init(&eli->li_list_mtx);
3091
bfff6873
LC		 3092 eli->li_state |= EXT4_LAZYINIT_QUIT;
3093
3094 ext4_li_info = eli;
3095
3096 return 0;
3097}
3098
3099static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3100 ext4_group_t start)
3101{
3102 struct ext4_sb_info *sbi = EXT4_SB(sb);
3103 struct ext4_li_request *elr;
bfff6873
LC		 3104
3105 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3106 if (!elr)
3107 return NULL;
3108
3109 elr->lr_super = sb;
3110 elr->lr_sbi = sbi;
3111 elr->lr_next_group = start;
3112
3113 /*
3114 * Randomize first schedule time of the request to
3115 * spread the inode table initialization requests
3116 * better.
3117 */
dd1f723b
TT		 3118 elr->lr_next_sched = jiffies + (prandom_u32() %
3119 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3120 return elr;
3121}
3122
7f511862
TT		 3123int ext4_register_li_request(struct super_block *sb,
3124 ext4_group_t first_not_zeroed)
bfff6873
LC		 3125{
3126 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3127 struct ext4_li_request *elr = NULL;
bfff6873 3128 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3129 int ret = 0;
bfff6873 3130
7f511862 3131 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3132 if (sbi->s_li_request != NULL) {
3133 /*
3134 * Reset timeout so it can be computed again, because
3135 * s_li_wait_mult might have changed.
3136 */
3137 sbi->s_li_request->lr_timeout = 0;
7f511862 3138 goto out;
51ce6511 3139 }
bfff6873
LC		 3140
3141 if (first_not_zeroed == ngroups ||
3142 (sb->s_flags & MS_RDONLY) ||
55ff3840 3143 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3144 goto out;
bfff6873
LC		 3145
3146 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3147 if (!elr) {
3148 ret = -ENOMEM;
3149 goto out;
3150 }
bfff6873
LC		 3151
3152 if (NULL == ext4_li_info) {
3153 ret = ext4_li_info_new();
3154 if (ret)
3155 goto out;
3156 }
3157
3158 mutex_lock(&ext4_li_info->li_list_mtx);
3159 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3160 mutex_unlock(&ext4_li_info->li_list_mtx);
3161
3162 sbi->s_li_request = elr;
46e4690b
TM		 3163 /*
3164 * set elr to NULL here since it has been inserted to
3165 * the request_list and the removal and free of it is
3166 * handled by ext4_clear_request_list from now on.
3167 */
3168 elr = NULL;
bfff6873
LC		 3169
3170 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3171 ret = ext4_run_lazyinit_thread();
3172 if (ret)
3173 goto out;
3174 }
bfff6873 3175out:
beed5ecb
NK		 3176 mutex_unlock(&ext4_li_mtx);
3177 if (ret)
bfff6873 3178 kfree(elr);
bfff6873
LC		 3179 return ret;
3180}
3181
3182/*
3183 * We do not need to lock anything since this is called on
3184 * module unload.
3185 */
3186static void ext4_destroy_lazyinit_thread(void)
3187{
3188 /*
3189 * If thread exited earlier
3190 * there's nothing to be done.
3191 */
8f1f7453 3192 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3193 return;
3194
8f1f7453 3195 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3196}
3197
25ed6e8a
DW		 3198static int set_journal_csum_feature_set(struct super_block *sb)
3199{
3200 int ret = 1;
3201 int compat, incompat;
3202 struct ext4_sb_info *sbi = EXT4_SB(sb);
3203
9aa5d32b 3204 if (ext4_has_metadata_csum(sb)) {
db9ee220 3205 /* journal checksum v3 */
25ed6e8a 3206 compat = 0;
db9ee220 3207 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3208 } else {
3209 /* journal checksum v1 */
3210 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3211 incompat = 0;
3212 }
3213
feb8c6d3
DW		 3214 jbd2_journal_clear_features(sbi->s_journal,
3215 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3216 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3217 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3218 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3219 ret = jbd2_journal_set_features(sbi->s_journal,
3220 compat, 0,
3221 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3222 incompat);
3223 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3224 ret = jbd2_journal_set_features(sbi->s_journal,
3225 compat, 0,
3226 incompat);
3227 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3228 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3229 } else {
feb8c6d3
DW		 3230 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3231 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3232 }
3233
3234 return ret;
3235}
3236
952fc18e
TT		 3237/*
3238 * Note: calculating the overhead so we can be compatible with
3239 * historical BSD practice is quite difficult in the face of
3240 * clusters/bigalloc. This is because multiple metadata blocks from
3241 * different block group can end up in the same allocation cluster.
3242 * Calculating the exact overhead in the face of clustered allocation
3243 * requires either O(all block bitmaps) in memory or O(number of block
3244 * groups**2) in time. We will still calculate the superblock for
3245 * older file systems --- and if we come across with a bigalloc file
3246 * system with zero in s_overhead_clusters the estimate will be close to
3247 * correct especially for very large cluster sizes --- but for newer
3248 * file systems, it's better to calculate this figure once at mkfs
3249 * time, and store it in the superblock. If the superblock value is
3250 * present (even for non-bigalloc file systems), we will use it.
3251 */
3252static int count_overhead(struct super_block *sb, ext4_group_t grp,
3253 char *buf)
3254{
3255 struct ext4_sb_info *sbi = EXT4_SB(sb);
3256 struct ext4_group_desc *gdp;
3257 ext4_fsblk_t first_block, last_block, b;
3258 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3259 int s, j, count = 0;
3260
0548bbb8
TT		 3261 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3262 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3263 sbi->s_itb_per_group + 2);
3264
952fc18e
TT		 3265 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3266 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3267 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3268 for (i = 0; i < ngroups; i++) {
3269 gdp = ext4_get_group_desc(sb, i, NULL);
3270 b = ext4_block_bitmap(sb, gdp);
3271 if (b >= first_block && b <= last_block) {
3272 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3273 count++;
3274 }
3275 b = ext4_inode_bitmap(sb, gdp);
3276 if (b >= first_block && b <= last_block) {
3277 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3278 count++;
3279 }
3280 b = ext4_inode_table(sb, gdp);
3281 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3282 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3283 int c = EXT4_B2C(sbi, b - first_block);
3284 ext4_set_bit(c, buf);
3285 count++;
3286 }
3287 if (i != grp)
3288 continue;
3289 s = 0;
3290 if (ext4_bg_has_super(sb, grp)) {
3291 ext4_set_bit(s++, buf);
3292 count++;
3293 }
3294 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3295 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3296 count++;
3297 }
3298 }
3299 if (!count)
3300 return 0;
3301 return EXT4_CLUSTERS_PER_GROUP(sb) -
3302 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3303}
3304
3305/*
3306 * Compute the overhead and stash it in sbi->s_overhead
3307 */
3308int ext4_calculate_overhead(struct super_block *sb)
3309{
3310 struct ext4_sb_info *sbi = EXT4_SB(sb);
3311 struct ext4_super_block *es = sbi->s_es;
3312 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3313 ext4_fsblk_t overhead = 0;
4fdb5543 3314 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3315
952fc18e
TT		 3316 if (!buf)
3317 return -ENOMEM;
3318
3319 /*
3320 * Compute the overhead (FS structures). This is constant
3321 * for a given filesystem unless the number of block groups
3322 * changes so we cache the previous value until it does.
3323 */
3324
3325 /*
3326 * All of the blocks before first_data_block are overhead
3327 */
3328 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3329
3330 /*
3331 * Add the overhead found in each block group
3332 */
3333 for (i = 0; i < ngroups; i++) {
3334 int blks;
3335
3336 blks = count_overhead(sb, i, buf);
3337 overhead += blks;
3338 if (blks)
3339 memset(buf, 0, PAGE_SIZE);
3340 cond_resched();
3341 }
b003b524
ES		 3342 /* Add the internal journal blocks as well */
3343 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3344 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3345
952fc18e
TT		 3346 sbi->s_overhead = overhead;
3347 smp_wmb();
3348 free_page((unsigned long) buf);
3349 return 0;
3350}
3351
27dd4385 3352
30fac0f7 3353static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
27dd4385
LC		 3354{
3355 ext4_fsblk_t resv_clusters;
3356
30fac0f7
JK		 3357 /*
3358 * There's no need to reserve anything when we aren't using extents.
3359 * The space estimates are exact, there are no unwritten extents,
3360 * hole punching doesn't need new metadata... This is needed especially
3361 * to keep ext2/3 backward compatibility.
3362 */
3363 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3364 return 0;
27dd4385
LC		 3365 /*
3366 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3367 * This should cover the situations where we can not afford to run
3368 * out of space like for example punch hole, or converting
556615dc 3369 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3370 * allocation would require 1, or 2 blocks, higher numbers are
3371 * very rare.
3372 */
30fac0f7
JK		 3373 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3374 EXT4_SB(sb)->s_cluster_bits;
27dd4385
LC		 3375
3376 do_div(resv_clusters, 50);
3377 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3378
3379 return resv_clusters;
3380}
3381
3382
3383static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3384{
3385 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3386 sbi->s_cluster_bits;
3387
3388 if (count >= clusters)
3389 return -EINVAL;
3390
3391 atomic64_set(&sbi->s_resv_clusters, count);
3392 return 0;
3393}
3394
2b2d6d01 3395static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3396{
d4c402d9 3397 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3398 struct buffer_head *bh;
617ba13b
MC		 3399 struct ext4_super_block *es = NULL;
3400 struct ext4_sb_info *sbi;
3401 ext4_fsblk_t block;
3402 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3403 ext4_fsblk_t logical_sb_block;
ac27a0ec 3404 unsigned long offset = 0;
ac27a0ec
DK		 3405 unsigned long journal_devnum = 0;
3406 unsigned long def_mount_opts;
3407 struct inode *root;
9f6200bb 3408 char *cp;
0390131b 3409 const char *descr;
dcc7dae3 3410 int ret = -ENOMEM;
281b5995 3411 int blocksize, clustersize;
4ec11028
TT		 3412 unsigned int db_count;
3413 unsigned int i;
281b5995 3414 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3415 __u64 blocks_count;
07aa2ea1 3416 int err = 0;
b3881f74 3417 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3418 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3419
3420 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3421 if (!sbi)
dcc7dae3 3422 goto out_free_orig;
705895b6
PE		 3423
3424 sbi->s_blockgroup_lock =
3425 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3426 if (!sbi->s_blockgroup_lock) {
3427 kfree(sbi);
dcc7dae3 3428 goto out_free_orig;
705895b6 3429 }
ac27a0ec 3430 sb->s_fs_info = sbi;
2c0544b2 3431 sbi->s_sb = sb;
240799cd 3432 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3433 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3434 if (sb->s_bdev->bd_part)
3435 sbi->s_sectors_written_start =
3436 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
b30ab0e0
MH		 3437#ifdef CONFIG_EXT4_FS_ENCRYPTION
3438 /* Modes of operations for file and directory encryption. */
3439 sbi->s_file_encryption_mode = EXT4_ENCRYPTION_MODE_AES_256_XTS;
3440 sbi->s_dir_encryption_mode = EXT4_ENCRYPTION_MODE_INVALID;
3441#endif
ac27a0ec 3442
9f6200bb
TT		 3443 /* Cleanup superblock name */
3444 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3445 *cp = '!';
3446
07aa2ea1 3447 /* -EINVAL is default */
dcc7dae3 3448 ret = -EINVAL;
617ba13b 3449 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3450 if (!blocksize) {
b31e1552 3451 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3452 goto out_fail;
3453 }
3454
3455 /*
617ba13b 3456 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3457 * block sizes. We need to calculate the offset from buffer start.
3458 */
617ba13b 3459 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3460 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3461 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3462 } else {
70bbb3e0 3463 logical_sb_block = sb_block;
ac27a0ec
DK		 3464 }
3465
a8ac900b 3466 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3467 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3468 goto out_fail;
3469 }
3470 /*
3471 * Note: s_es must be initialized as soon as possible because
617ba13b 3472 * some ext4 macro-instructions depend on its value
ac27a0ec 3473 */
2716b802 3474 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3475 sbi->s_es = es;
3476 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3477 if (sb->s_magic != EXT4_SUPER_MAGIC)
3478 goto cantfind_ext4;
afc32f7e 3479 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3480
feb0ab32
DW		 3481 /* Warn if metadata_csum and gdt_csum are both set. */
3482 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3483 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3484 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
363307e6 3485 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3486 "redundant flags; please run fsck.");
3487
d25425f8
DW		 3488 /* Check for a known checksum algorithm */
3489 if (!ext4_verify_csum_type(sb, es)) {
3490 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3491 "unknown checksum algorithm.");
3492 silent = 1;
3493 goto cantfind_ext4;
3494 }
3495
0441984a
DW		 3496 /* Load the checksum driver */
3497 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3498 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3499 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3500 if (IS_ERR(sbi->s_chksum_driver)) {
3501 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3502 ret = PTR_ERR(sbi->s_chksum_driver);
3503 sbi->s_chksum_driver = NULL;
3504 goto failed_mount;
3505 }
3506 }
3507
a9c47317
DW		 3508 /* Check superblock checksum */
3509 if (!ext4_superblock_csum_verify(sb, es)) {
3510 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3511 "invalid superblock checksum. Run e2fsck?");
3512 silent = 1;
3513 goto cantfind_ext4;
3514 }
3515
3516 /* Precompute checksum seed for all metadata */
9aa5d32b 3517 if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3518 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3519 sizeof(es->s_uuid));
3520
ac27a0ec
DK		 3521 /* Set defaults before we parse the mount options */
3522 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3523 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3524 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3525 set_opt(sb, DEBUG);
87f26807 3526 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3527 set_opt(sb, GRPID);
617ba13b 3528 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3529 set_opt(sb, NO_UID32);
ea663336 3530 /* xattr user namespace & acls are now defaulted on */
ea663336 3531 set_opt(sb, XATTR_USER);
03010a33 3532#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3533 set_opt(sb, POSIX_ACL);
2e7842b8 3534#endif
98c1a759
DW		 3535 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3536 if (ext4_has_metadata_csum(sb))
3537 set_opt(sb, JOURNAL_CHECKSUM);
3538
617ba13b 3539 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3540 set_opt(sb, JOURNAL_DATA);
617ba13b 3541 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3542 set_opt(sb, ORDERED_DATA);
617ba13b 3543 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3544 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3545
3546 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3547 set_opt(sb, ERRORS_PANIC);
bb4f397a 3548 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3549 set_opt(sb, ERRORS_CONT);
bb4f397a 3550 else
fd8c37ec 3551 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3552 /* block_validity enabled by default; disable with noblock_validity */
3553 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3554 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3555 set_opt(sb, DISCARD);
ac27a0ec 3556
08cefc7a
EB		 3557 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3558 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3559 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3560 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3561 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3562
8b67f04a 3563 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3564 set_opt(sb, BARRIER);
ac27a0ec 3565
dd919b98
AK		 3566 /*
3567 * enable delayed allocation by default
3568 * Use -o nodelalloc to turn it off
3569 */
bc0b75f7 3570 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3571 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3572 set_opt(sb, DELALLOC);
dd919b98 3573
51ce6511
LC		 3574 /*
3575 * set default s_li_wait_mult for lazyinit, for the case there is
3576 * no mount option specified.
3577 */
3578 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3579
8b67f04a 3580 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3581 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3582 ext4_msg(sb, KERN_WARNING,
3583 "failed to parse options in superblock: %s",
3584 sbi->s_es->s_mount_opts);
3585 }
5a916be1 3586 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3587 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3588 &journal_ioprio, 0))
ac27a0ec
DK		 3589 goto failed_mount;
3590
56889787
TT		 3591 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3592 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3593 "with data=journal disables delayed "
3594 "allocation and O_DIRECT support!\n");
3595 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3596 ext4_msg(sb, KERN_ERR, "can't mount with "
3597 "both data=journal and delalloc");
3598 goto failed_mount;
3599 }
3600 if (test_opt(sb, DIOREAD_NOLOCK)) {
3601 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3602 "both data=journal and dioread_nolock");
56889787
TT		 3603 goto failed_mount;
3604 }
923ae0ff
RZ		 3605 if (test_opt(sb, DAX)) {
3606 ext4_msg(sb, KERN_ERR, "can't mount with "
3607 "both data=journal and dax");
3608 goto failed_mount;
3609 }
56889787
TT		 3610 if (test_opt(sb, DELALLOC))
3611 clear_opt(sb, DELALLOC);
3612 }
3613
ac27a0ec 3614 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3615 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3616
617ba13b
MC		 3617 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3618 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3619 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3620 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3621 ext4_msg(sb, KERN_WARNING,
3622 "feature flags set on rev 0 fs, "
3623 "running e2fsck is recommended");
469108ff 3624
ed3654eb
TT		 3625 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3626 set_opt2(sb, HURD_COMPAT);
3627 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3628 EXT4_FEATURE_INCOMPAT_64BIT)) {
3629 ext4_msg(sb, KERN_ERR,
3630 "The Hurd can't support 64-bit file systems");
3631 goto failed_mount;
3632 }
3633 }
3634
2035e776
TT		 3635 if (IS_EXT2_SB(sb)) {
3636 if (ext2_feature_set_ok(sb))
3637 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3638 "using the ext4 subsystem");
3639 else {
3640 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3641 "to feature incompatibilities");
3642 goto failed_mount;
3643 }
3644 }
3645
3646 if (IS_EXT3_SB(sb)) {
3647 if (ext3_feature_set_ok(sb))
3648 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3649 "using the ext4 subsystem");
3650 else {
3651 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3652 "to feature incompatibilities");
3653 goto failed_mount;
3654 }
3655 }
3656
ac27a0ec
DK		 3657 /*
3658 * Check feature flags regardless of the revision level, since we
3659 * previously didn't change the revision level when setting the flags,
3660 * so there is a chance incompat flags are set on a rev 0 filesystem.
3661 */
a13fb1a4 3662 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3663 goto failed_mount;
a13fb1a4 3664
261cb20c 3665 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3666 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3667 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3668 ext4_msg(sb, KERN_ERR,
3669 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3670 goto failed_mount;
3671 }
3672
923ae0ff
RZ		 3673 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3674 if (blocksize != PAGE_SIZE) {
3675 ext4_msg(sb, KERN_ERR,
3676 "error: unsupported blocksize for dax");
3677 goto failed_mount;
3678 }
3679 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3680 ext4_msg(sb, KERN_ERR,
3681 "error: device does not support dax");
3682 goto failed_mount;
3683 }
3684 }
3685
ac27a0ec 3686 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3687 /* Validate the filesystem blocksize */
3688 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3689 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3690 blocksize);
ac27a0ec
DK		 3691 goto failed_mount;
3692 }
3693
2b2d6d01 3694 brelse(bh);
70bbb3e0
AM		 3695 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3696 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3697 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3698 if (!bh) {
b31e1552
ES		 3699 ext4_msg(sb, KERN_ERR,
3700 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3701 goto failed_mount;
3702 }
2716b802 3703 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3704 sbi->s_es = es;
617ba13b 3705 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3706 ext4_msg(sb, KERN_ERR,
3707 "Magic mismatch, very weird!");
ac27a0ec
DK		 3708 goto failed_mount;
3709 }
3710 }
3711
a13fb1a4
ES		 3712 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3713 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3714 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3715 has_huge_files);
3716 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3717
617ba13b
MC		 3718 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3719 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3720 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3721 } else {
3722 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3723 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3724 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3725 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3726 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3727 ext4_msg(sb, KERN_ERR,
3728 "unsupported inode size: %d",
2b2d6d01 3729 sbi->s_inode_size);
ac27a0ec
DK		 3730 goto failed_mount;
3731 }
ef7f3835
KS		 3732 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3733 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3734 }
0b8e58a1 3735
0d1ee42f
AR		 3736 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3737 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3738 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3739 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3740 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3741 ext4_msg(sb, KERN_ERR,
3742 "unsupported descriptor size %lu",
0d1ee42f
AR		 3743 sbi->s_desc_size);
3744 goto failed_mount;
3745 }
3746 } else
3747 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3748
ac27a0ec 3749 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3750 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3751 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3752 goto cantfind_ext4;
0b8e58a1 3753
617ba13b 3754 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3755 if (sbi->s_inodes_per_block == 0)
617ba13b 3756 goto cantfind_ext4;
ac27a0ec
DK		 3757 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3758 sbi->s_inodes_per_block;
0d1ee42f 3759 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3760 sbi->s_sbh = bh;
3761 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3762 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3763 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3764
2b2d6d01 3765 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3766 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3767 sbi->s_def_hash_version = es->s_def_hash_version;
23301410
TT		 3768 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3769 i = le32_to_cpu(es->s_flags);
3770 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3771 sbi->s_hash_unsigned = 3;
3772 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3773#ifdef __CHAR_UNSIGNED__
23301410
TT		 3774 if (!(sb->s_flags & MS_RDONLY))
3775 es->s_flags |=
3776 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3777 sbi->s_hash_unsigned = 3;
f99b2589 3778#else
23301410
TT		 3779 if (!(sb->s_flags & MS_RDONLY))
3780 es->s_flags |=
3781 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3782#endif
23301410 3783 }
f99b2589 3784 }
ac27a0ec 3785
281b5995
TT		 3786 /* Handle clustersize */
3787 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3788 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3789 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3790 if (has_bigalloc) {
3791 if (clustersize < blocksize) {
3792 ext4_msg(sb, KERN_ERR,
3793 "cluster size (%d) smaller than "
3794 "block size (%d)", clustersize, blocksize);
3795 goto failed_mount;
3796 }
3797 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3798 le32_to_cpu(es->s_log_block_size);
3799 sbi->s_clusters_per_group =
3800 le32_to_cpu(es->s_clusters_per_group);
3801 if (sbi->s_clusters_per_group > blocksize * 8) {
3802 ext4_msg(sb, KERN_ERR,
3803 "#clusters per group too big: %lu",
3804 sbi->s_clusters_per_group);
3805 goto failed_mount;
3806 }
3807 if (sbi->s_blocks_per_group !=
3808 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3809 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3810 "clusters per group (%lu) inconsistent",
3811 sbi->s_blocks_per_group,
3812 sbi->s_clusters_per_group);
3813 goto failed_mount;
3814 }
3815 } else {
3816 if (clustersize != blocksize) {
3817 ext4_warning(sb, "fragment/cluster size (%d) != "
3818 "block size (%d)", clustersize,
3819 blocksize);
3820 clustersize = blocksize;
3821 }
3822 if (sbi->s_blocks_per_group > blocksize * 8) {
3823 ext4_msg(sb, KERN_ERR,
3824 "#blocks per group too big: %lu",
3825 sbi->s_blocks_per_group);
3826 goto failed_mount;
3827 }
3828 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3829 sbi->s_cluster_bits = 0;
ac27a0ec 3830 }
281b5995
TT		 3831 sbi->s_cluster_ratio = clustersize / blocksize;
3832
ac27a0ec 3833 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3834 ext4_msg(sb, KERN_ERR,
3835 "#inodes per group too big: %lu",
2b2d6d01 3836 sbi->s_inodes_per_group);
ac27a0ec
DK		 3837 goto failed_mount;
3838 }
3839
960fd856
TT		 3840 /* Do we have standard group size of clustersize * 8 blocks ? */
3841 if (sbi->s_blocks_per_group == clustersize << 3)
3842 set_opt2(sb, STD_GROUP_SIZE);
3843
bf43d84b
ES		 3844 /*
3845 * Test whether we have more sectors than will fit in sector_t,
3846 * and whether the max offset is addressable by the page cache.
3847 */
5a9ae68a 3848 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3849 ext4_blocks_count(es));
5a9ae68a 3850 if (err) {
b31e1552 3851 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3852 " too large to mount safely on this system");
ac27a0ec 3853 if (sizeof(sector_t) < 8)
90c699a9 3854 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3855 goto failed_mount;
3856 }
3857
617ba13b
MC		 3858 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3859 goto cantfind_ext4;
e7c95593 3860
0f2ddca6
FTN		 3861 /* check blocks count against device size */
3862 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3863 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3864 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3865 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3866 ext4_blocks_count(es), blocks_count);
3867 goto failed_mount;
3868 }
3869
0b8e58a1
AD		 3870 /*
3871 * It makes no sense for the first data block to be beyond the end
3872 * of the filesystem.
3873 */
3874 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3875 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3876 "block %u is beyond end of filesystem (%llu)",
3877 le32_to_cpu(es->s_first_data_block),
3878 ext4_blocks_count(es));
e7c95593
ES		 3879 goto failed_mount;
3880 }
bd81d8ee
LV		 3881 blocks_count = (ext4_blocks_count(es) -
3882 le32_to_cpu(es->s_first_data_block) +
3883 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3884 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3885 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3886 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3887 "(block count %llu, first data block %u, "
b31e1552 3888 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3889 ext4_blocks_count(es),
3890 le32_to_cpu(es->s_first_data_block),
3891 EXT4_BLOCKS_PER_GROUP(sb));
3892 goto failed_mount;
3893 }
bd81d8ee 3894 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3895 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3896 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3897 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3898 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3899 sbi->s_group_desc = ext4_kvmalloc(db_count *
3900 sizeof(struct buffer_head *),
3901 GFP_KERNEL);
ac27a0ec 3902 if (sbi->s_group_desc == NULL) {
b31e1552 3903 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3904 ret = -ENOMEM;
ac27a0ec
DK		 3905 goto failed_mount;
3906 }
3907
9f6200bb
TT		 3908 if (ext4_proc_root)
3909 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3910
66acdcf4
TT		 3911 if (sbi->s_proc)
3912 proc_create_data("options", S_IRUGO, sbi->s_proc,
3913 &ext4_seq_options_fops, sb);
3914
705895b6 3915 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3916
3917 for (i = 0; i < db_count; i++) {
70bbb3e0 3918 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3919 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3920 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3921 ext4_msg(sb, KERN_ERR,
3922 "can't read group descriptor %d", i);
ac27a0ec
DK		 3923 db_count = i;
3924 goto failed_mount2;
3925 }
3926 }
bfff6873 3927 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3928 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
f9ae9cf5 3929 goto failed_mount2;
ac27a0ec 3930 }
772cb7c8 3931
f9ae9cf5 3932 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3933 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3934 spin_lock_init(&sbi->s_next_gen_lock);
3935
04ecddb7
JM		 3936 setup_timer(&sbi->s_err_report, print_daily_error_info,
3937 (unsigned long) sb);
04496411 3938
a75ae78f 3939 /* Register extent status tree shrinker */
eb68d0e2 3940 if (ext4_es_register_shrinker(sbi))
ce7e010a 3941 goto failed_mount3;
ce7e010a 3942
c9de560d 3943 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3944 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3945
f9ae9cf5
TT		 3946 /*
3947 * set up enough so that it can read an inode
3948 */
f6e63f90 3949 sb->s_op = &ext4_sops;
617ba13b
MC		 3950 sb->s_export_op = &ext4_export_ops;
3951 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3952#ifdef CONFIG_QUOTA
617ba13b 3953 sb->dq_op = &ext4_quota_operations;
262b4662 3954 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
1fa5efe3 3955 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3956 else
3957 sb->s_qcop = &ext4_qctl_operations;
96c7e0d9 3958 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
ac27a0ec 3959#endif
f2fa2ffc
AK		 3960 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3961
ac27a0ec 3962 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3963 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3964
3965 sb->s_root = NULL;
3966
3967 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3968 EXT4_HAS_INCOMPAT_FEATURE(sb,
3969 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3970
c5e06d10
JL		 3971 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3972 !(sb->s_flags & MS_RDONLY))
3973 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3974 goto failed_mount3a;
c5e06d10 3975
ac27a0ec
DK		 3976 /*
3977 * The first inode we look at is the journal inode. Don't try
3978 * root first: it may be modified in the journal!
3979 */
3980 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3981 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3982 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3983 goto failed_mount3a;
0390131b
FM		 3984 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3985 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3986 ext4_msg(sb, KERN_ERR, "required journal recovery "
3987 "suppressed and not mounted read-only");
744692dc 3988 goto failed_mount_wq;
ac27a0ec 3989 } else {
fd8c37ec 3990 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3991 sbi->s_journal = NULL;
3992 needs_recovery = 0;
3993 goto no_journal;
ac27a0ec
DK		 3994 }
3995
f32aaf2d 3996 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3997 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3998 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3999 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4000 goto failed_mount_wq;
eb40a09c
JS		 4001 }
4002
25ed6e8a
DW		 4003 if (!set_journal_csum_feature_set(sb)) {
4004 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4005 "feature set");
4006 goto failed_mount_wq;
d4da6c9c 4007 }
818d276c 4008
ac27a0ec
DK		 4009 /* We have now updated the journal if required, so we can
4010 * validate the data journaling mode. */
4011 switch (test_opt(sb, DATA_FLAGS)) {
4012 case 0:
4013 /* No mode set, assume a default based on the journal
63f57933
AM		 4014 * capabilities: ORDERED_DATA if the journal can
4015 * cope, else JOURNAL_DATA
4016 */
dab291af
MC		 4017 if (jbd2_journal_check_available_features
4018 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 4019 set_opt(sb, ORDERED_DATA);
ac27a0ec 4020 else
fd8c37ec 4021 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 4022 break;
4023
617ba13b
MC		 4024 case EXT4_MOUNT_ORDERED_DATA:
4025 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4026 if (!jbd2_journal_check_available_features
4027 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4028 ext4_msg(sb, KERN_ERR, "Journal does not support "
4029 "requested data journaling mode");
744692dc 4030 goto failed_mount_wq;
ac27a0ec
DK		 4031 }
4032 default:
4033 break;
4034 }
b3881f74 4035 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4036
18aadd47
BJ		 4037 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4038
ce7e010a 4039no_journal:
9c191f70
M		 4040 if (ext4_mballoc_ready) {
4041 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4042 if (!sbi->s_mb_cache) {
4043 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4044 goto failed_mount_wq;
4045 }
4046 }
4047
952fc18e
TT		 4048 /*
4049 * Get the # of file system overhead blocks from the
4050 * superblock if present.
4051 */
4052 if (es->s_overhead_clusters)
4053 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4054 else {
07aa2ea1
LC		 4055 err = ext4_calculate_overhead(sb);
4056 if (err)
952fc18e
TT		 4057 goto failed_mount_wq;
4058 }
4059
fd89d5f2
TH		 4060 /*
4061 * The maximum number of concurrent works can be high and
4062 * concurrency isn't really necessary. Limit it to 1.
4063 */
2e8fa54e
JK		 4064 EXT4_SB(sb)->rsv_conversion_wq =
4065 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4066 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4067 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4068 ret = -ENOMEM;
2e8fa54e
JK		 4069 goto failed_mount4;
4070 }
4071
ac27a0ec 4072 /*
dab291af 4073 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4074 * so we can safely mount the rest of the filesystem now.
4075 */
4076
1d1fe1ee
DH		 4077 root = ext4_iget(sb, EXT4_ROOT_INO);
4078 if (IS_ERR(root)) {
b31e1552 4079 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4080 ret = PTR_ERR(root);
32a9bb57 4081 root = NULL;
ac27a0ec
DK		 4082 goto failed_mount4;
4083 }
4084 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4085 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4086 iput(root);
ac27a0ec
DK		 4087 goto failed_mount4;
4088 }
48fde701 4089 sb->s_root = d_make_root(root);
1d1fe1ee 4090 if (!sb->s_root) {
b31e1552 4091 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4092 ret = -ENOMEM;
4093 goto failed_mount4;
4094 }
ac27a0ec 4095
7e84b621
ES		 4096 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4097 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4098
4099 /* determine the minimum size of new large inodes, if present */
4100 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4101 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4102 EXT4_GOOD_OLD_INODE_SIZE;
4103 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4104 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4105 if (sbi->s_want_extra_isize <
4106 le16_to_cpu(es->s_want_extra_isize))
4107 sbi->s_want_extra_isize =
4108 le16_to_cpu(es->s_want_extra_isize);
4109 if (sbi->s_want_extra_isize <
4110 le16_to_cpu(es->s_min_extra_isize))
4111 sbi->s_want_extra_isize =
4112 le16_to_cpu(es->s_min_extra_isize);
4113 }
4114 }
4115 /* Check if enough inode space is available */
4116 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4117 sbi->s_inode_size) {
4118 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4119 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4120 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4121 "available");
ef7f3835
KS		 4122 }
4123
30fac0f7 4124 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
27dd4385
LC		 4125 if (err) {
4126 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
30fac0f7 4127 "reserved pool", ext4_calculate_resv_clusters(sb));
f9ae9cf5 4128 goto failed_mount4a;
27dd4385
LC		 4129 }
4130
6fd058f7
TT		 4131 err = ext4_setup_system_zone(sb);
4132 if (err) {
b31e1552 4133 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4134 "zone (%d)", err);
f9ae9cf5
TT		 4135 goto failed_mount4a;
4136 }
4137
4138 ext4_ext_init(sb);
4139 err = ext4_mb_init(sb);
4140 if (err) {
4141 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4142 err);
dcf2d804 4143 goto failed_mount5;
c2774d84
AK		 4144 }
4145
d5e03cbb
TT		 4146 block = ext4_count_free_clusters(sb);
4147 ext4_free_blocks_count_set(sbi->s_es,
4148 EXT4_C2B(sbi, block));
908c7f19
TH		 4149 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4150 GFP_KERNEL);
d5e03cbb
TT		 4151 if (!err) {
4152 unsigned long freei = ext4_count_free_inodes(sb);
4153 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4154 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4155 GFP_KERNEL);
d5e03cbb
TT		 4156 }
4157 if (!err)
4158 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4159 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4160 if (!err)
908c7f19
TH		 4161 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4162 GFP_KERNEL);
d5e03cbb
TT		 4163 if (err) {
4164 ext4_msg(sb, KERN_ERR, "insufficient memory");
4165 goto failed_mount6;
4166 }
4167
4168 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4169 if (!ext4_fill_flex_info(sb)) {
4170 ext4_msg(sb, KERN_ERR,
4171 "unable to initialize "
4172 "flex_bg meta info!");
4173 goto failed_mount6;
4174 }
4175
bfff6873
LC		 4176 err = ext4_register_li_request(sb, first_not_zeroed);
4177 if (err)
dcf2d804 4178 goto failed_mount6;
bfff6873 4179
3197ebdb
TT		 4180 sbi->s_kobj.kset = ext4_kset;
4181 init_completion(&sbi->s_kobj_unregister);
4182 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4183 "%s", sb->s_id);
dcf2d804
TM		 4184 if (err)
4185 goto failed_mount7;
3197ebdb 4186
9b2ff357
JK		 4187#ifdef CONFIG_QUOTA
4188 /* Enable quota usage during mount. */
4189 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4190 !(sb->s_flags & MS_RDONLY)) {
4191 err = ext4_enable_quotas(sb);
4192 if (err)
4193 goto failed_mount8;
4194 }
4195#endif /* CONFIG_QUOTA */
4196
617ba13b
MC		 4197 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4198 ext4_orphan_cleanup(sb, es);
4199 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4200 if (needs_recovery) {
b31e1552 4201 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4202 ext4_mark_recovery_complete(sb, es);
4203 }
4204 if (EXT4_SB(sb)->s_journal) {
4205 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4206 descr = " journalled data mode";
4207 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4208 descr = " ordered data mode";
4209 else
4210 descr = " writeback data mode";
4211 } else
4212 descr = "out journal";
4213
79add3a3
LC		 4214 if (test_opt(sb, DISCARD)) {
4215 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4216 if (!blk_queue_discard(q))
4217 ext4_msg(sb, KERN_WARNING,
4218 "mounting with \"discard\" option, but "
4219 "the device does not support discard");
4220 }
4221
d4c402d9 4222 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4223 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4224 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4225
66e61a9e
TT		 4226 if (es->s_error_count)
4227 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4228
efbed4dc
TT		 4229 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4230 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4231 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4232 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4233
d4c402d9 4234 kfree(orig_data);
ac27a0ec
DK		 4235 return 0;
4236
617ba13b 4237cantfind_ext4:
ac27a0ec 4238 if (!silent)
b31e1552 4239 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4240 goto failed_mount;
4241
72ba7450
TT		 4242#ifdef CONFIG_QUOTA
4243failed_mount8:
4244 kobject_del(&sbi->s_kobj);
4245#endif
dcf2d804
TM		 4246failed_mount7:
4247 ext4_unregister_li_request(sb);
4248failed_mount6:
f9ae9cf5 4249 ext4_mb_release(sb);
d5e03cbb 4250 if (sbi->s_flex_groups)
b93b41d4 4251 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4252 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4253 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4254 percpu_counter_destroy(&sbi->s_dirs_counter);
4255 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4256failed_mount5:
f9ae9cf5
TT		 4257 ext4_ext_release(sb);
4258 ext4_release_system_zone(sb);
4259failed_mount4a:
94bf608a 4260 dput(sb->s_root);
32a9bb57 4261 sb->s_root = NULL;
94bf608a 4262failed_mount4:
b31e1552 4263 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4264 if (EXT4_SB(sb)->rsv_conversion_wq)
4265 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4266failed_mount_wq:
0390131b
FM		 4267 if (sbi->s_journal) {
4268 jbd2_journal_destroy(sbi->s_journal);
4269 sbi->s_journal = NULL;
4270 }
50460fe8 4271failed_mount3a:
d3922a77 4272 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4273failed_mount3:
9105bb14 4274 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4275 if (sbi->s_mmp_tsk)
4276 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4277failed_mount2:
4278 for (i = 0; i < db_count; i++)
4279 brelse(sbi->s_group_desc[i]);
b93b41d4 4280 kvfree(sbi->s_group_desc);
ac27a0ec 4281failed_mount:
0441984a
DW		 4282 if (sbi->s_chksum_driver)
4283 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4284 if (sbi->s_proc) {
66acdcf4 4285 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4286 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4287 }
ac27a0ec 4288#ifdef CONFIG_QUOTA
a2d4a646 4289 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4290 kfree(sbi->s_qf_names[i]);
4291#endif
617ba13b 4292 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4293 brelse(bh);
4294out_fail:
4295 sb->s_fs_info = NULL;
f6830165 4296 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4297 kfree(sbi);
dcc7dae3 4298out_free_orig:
d4c402d9 4299 kfree(orig_data);
07aa2ea1 4300 return err ? err : ret;
ac27a0ec
DK		 4301}
4302
4303/*
4304 * Setup any per-fs journal parameters now. We'll do this both on
4305 * initial mount, once the journal has been initialised but before we've
4306 * done any recovery; and again on any subsequent remount.
4307 */
617ba13b 4308static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4309{
617ba13b 4310 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4311
30773840
TT		 4312 journal->j_commit_interval = sbi->s_commit_interval;
4313 journal->j_min_batch_time = sbi->s_min_batch_time;
4314 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4315
a931da6a 4316 write_lock(&journal->j_state_lock);
ac27a0ec 4317 if (test_opt(sb, BARRIER))
dab291af 4318 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4319 else
dab291af 4320 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4321 if (test_opt(sb, DATA_ERR_ABORT))
4322 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4323 else
4324 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4325 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4326}
4327
617ba13b 4328static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4329 unsigned int journal_inum)
4330{
4331 struct inode *journal_inode;
4332 journal_t *journal;
4333
0390131b
FM		 4334 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4335
ac27a0ec
DK		 4336 /* First, test for the existence of a valid inode on disk. Bad
4337 * things happen if we iget() an unused inode, as the subsequent
4338 * iput() will try to delete it. */
4339
1d1fe1ee
DH		 4340 journal_inode = ext4_iget(sb, journal_inum);
4341 if (IS_ERR(journal_inode)) {
b31e1552 4342 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4343 return NULL;
4344 }
4345 if (!journal_inode->i_nlink) {
4346 make_bad_inode(journal_inode);
4347 iput(journal_inode);
b31e1552 4348 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4349 return NULL;
4350 }
4351
e5f8eab8 4352 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4353 journal_inode, journal_inode->i_size);
1d1fe1ee 4354 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4355 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4356 iput(journal_inode);
4357 return NULL;
4358 }
4359
dab291af 4360 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4361 if (!journal) {
b31e1552 4362 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4363 iput(journal_inode);
4364 return NULL;
4365 }
4366 journal->j_private = sb;
617ba13b 4367 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4368 return journal;
4369}
4370
617ba13b 4371static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4372 dev_t j_dev)
4373{
2b2d6d01 4374 struct buffer_head *bh;
ac27a0ec 4375 journal_t *journal;
617ba13b
MC		 4376 ext4_fsblk_t start;
4377 ext4_fsblk_t len;
ac27a0ec 4378 int hblock, blocksize;
617ba13b 4379 ext4_fsblk_t sb_block;
ac27a0ec 4380 unsigned long offset;
2b2d6d01 4381 struct ext4_super_block *es;
ac27a0ec
DK		 4382 struct block_device *bdev;
4383
0390131b
FM		 4384 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4385
b31e1552 4386 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4387 if (bdev == NULL)
4388 return NULL;
4389
ac27a0ec 4390 blocksize = sb->s_blocksize;
e1defc4f 4391 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4392 if (blocksize < hblock) {
b31e1552
ES		 4393 ext4_msg(sb, KERN_ERR,
4394 "blocksize too small for journal device");
ac27a0ec
DK		 4395 goto out_bdev;
4396 }
4397
617ba13b
MC		 4398 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4399 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4400 set_blocksize(bdev, blocksize);
4401 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4402 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4403 "external journal");
ac27a0ec
DK		 4404 goto out_bdev;
4405 }
4406
2716b802 4407 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4408 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4409 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4410 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4411 ext4_msg(sb, KERN_ERR, "external journal has "
4412 "bad superblock");
ac27a0ec
DK		 4413 brelse(bh);
4414 goto out_bdev;
4415 }
4416
df4763be
DW		 4417 if ((le32_to_cpu(es->s_feature_ro_compat) &
4418 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4419 es->s_checksum != ext4_superblock_csum(sb, es)) {
4420 ext4_msg(sb, KERN_ERR, "external journal has "
4421 "corrupt superblock");
4422 brelse(bh);
4423 goto out_bdev;
4424 }
4425
617ba13b 4426 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4427 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4428 brelse(bh);
4429 goto out_bdev;
4430 }
4431
bd81d8ee 4432 len = ext4_blocks_count(es);
ac27a0ec
DK		 4433 start = sb_block + 1;
4434 brelse(bh); /* we're done with the superblock */
4435
dab291af 4436 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4437 start, len, blocksize);
4438 if (!journal) {
b31e1552 4439 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4440 goto out_bdev;
4441 }
4442 journal->j_private = sb;
9f203507 4443 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4444 wait_on_buffer(journal->j_sb_buffer);
4445 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4446 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4447 goto out_journal;
4448 }
4449 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4450 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4451 "user (unsupported) - %d",
ac27a0ec
DK		 4452 be32_to_cpu(journal->j_superblock->s_nr_users));
4453 goto out_journal;
4454 }
617ba13b
MC		 4455 EXT4_SB(sb)->journal_bdev = bdev;
4456 ext4_init_journal_params(sb, journal);
ac27a0ec 4457 return journal;
0b8e58a1 4458
ac27a0ec 4459out_journal:
dab291af 4460 jbd2_journal_destroy(journal);
ac27a0ec 4461out_bdev:
617ba13b 4462 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4463 return NULL;
4464}
4465
617ba13b
MC		 4466static int ext4_load_journal(struct super_block *sb,
4467 struct ext4_super_block *es,
ac27a0ec
DK		 4468 unsigned long journal_devnum)
4469{
4470 journal_t *journal;
4471 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4472 dev_t journal_dev;
4473 int err = 0;
4474 int really_read_only;
4475
0390131b
FM		 4476 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4477
ac27a0ec
DK		 4478 if (journal_devnum &&
4479 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4480 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4481 "numbers have changed");
ac27a0ec
DK		 4482 journal_dev = new_decode_dev(journal_devnum);
4483 } else
4484 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4485
4486 really_read_only = bdev_read_only(sb->s_bdev);
4487
4488 /*
4489 * Are we loading a blank journal or performing recovery after a
4490 * crash? For recovery, we need to check in advance whether we
4491 * can get read-write access to the device.
4492 */
617ba13b 4493 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4494 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4495 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4496 "required on readonly filesystem");
ac27a0ec 4497 if (really_read_only) {
b31e1552
ES		 4498 ext4_msg(sb, KERN_ERR, "write access "
4499 "unavailable, cannot proceed");
ac27a0ec
DK		 4500 return -EROFS;
4501 }
b31e1552
ES		 4502 ext4_msg(sb, KERN_INFO, "write access will "
4503 "be enabled during recovery");
ac27a0ec
DK		 4504 }
4505 }
4506
4507 if (journal_inum && journal_dev) {
b31e1552
ES		 4508 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4509 "and inode journals!");
ac27a0ec
DK		 4510 return -EINVAL;
4511 }
4512
4513 if (journal_inum) {
617ba13b 4514 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4515 return -EINVAL;
4516 } else {
617ba13b 4517 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4518 return -EINVAL;
4519 }
4520
90576c0b 4521 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4522 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4523
617ba13b 4524 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4525 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4526 if (!err) {
4527 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4528 if (save)
4529 memcpy(save, ((char *) es) +
4530 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4531 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4532 if (save)
4533 memcpy(((char *) es) + EXT4_S_ERR_START,
4534 save, EXT4_S_ERR_LEN);
4535 kfree(save);
4536 }
ac27a0ec
DK		 4537
4538 if (err) {
b31e1552 4539 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4540 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4541 return err;
4542 }
4543
617ba13b
MC		 4544 EXT4_SB(sb)->s_journal = journal;
4545 ext4_clear_journal_err(sb, es);
ac27a0ec 4546
c41303ce 4547 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4548 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4549 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4550
4551 /* Make sure we flush the recovery flag to disk. */
e2d67052 4552 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4553 }
4554
4555 return 0;
4556}
4557
e2d67052 4558static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4559{
e2d67052 4560 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4561 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4562 int error = 0;
ac27a0ec 4563
08439fec 4564 if (!sbh)
c4be0c1d 4565 return error;
914258bf
TT		 4566 if (buffer_write_io_error(sbh)) {
4567 /*
4568 * Oh, dear. A previous attempt to write the
4569 * superblock failed. This could happen because the
4570 * USB device was yanked out. Or it could happen to
4571 * be a transient write error and maybe the block will
4572 * be remapped. Nothing we can do but to retry the
4573 * write and hope for the best.
4574 */
b31e1552
ES		 4575 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4576 "superblock detected");
914258bf
TT		 4577 clear_buffer_write_io_error(sbh);
4578 set_buffer_uptodate(sbh);
4579 }
71290b36
TT		 4580 /*
4581 * If the file system is mounted read-only, don't update the
4582 * superblock write time. This avoids updating the superblock
4583 * write time when we are mounting the root file system
4584 * read/only but we need to replay the journal; at that point,
4585 * for people who are east of GMT and who make their clock
4586 * tick in localtime for Windows bug-for-bug compatibility,
4587 * the clock is set in the future, and this will cause e2fsck
4588 * to complain and force a full file system check.
4589 */
4590 if (!(sb->s_flags & MS_RDONLY))
4591 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4592 if (sb->s_bdev->bd_part)
4593 es->s_kbytes_written =
4594 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4595 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4596 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4597 else
4598 es->s_kbytes_written =
4599 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4600 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4601 ext4_free_blocks_count_set(es,
57042651
TT		 4602 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4603 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4604 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4605 es->s_free_inodes_count =
4606 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4607 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4608 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4609 ext4_superblock_csum_set(sb);
ac27a0ec 4610 mark_buffer_dirty(sbh);
914258bf 4611 if (sync) {
c4be0c1d
TS		 4612 error = sync_dirty_buffer(sbh);
4613 if (error)
4614 return error;
4615
4616 error = buffer_write_io_error(sbh);
4617 if (error) {
b31e1552
ES		 4618 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4619 "superblock");
914258bf
TT		 4620 clear_buffer_write_io_error(sbh);
4621 set_buffer_uptodate(sbh);
4622 }
4623 }
c4be0c1d 4624 return error;
ac27a0ec
DK		 4625}
4626
ac27a0ec
DK		 4627/*
4628 * Have we just finished recovery? If so, and if we are mounting (or
4629 * remounting) the filesystem readonly, then we will end up with a
4630 * consistent fs on disk. Record that fact.
4631 */
2b2d6d01
TT		 4632static void ext4_mark_recovery_complete(struct super_block *sb,
4633 struct ext4_super_block *es)
ac27a0ec 4634{
617ba13b 4635 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4636
0390131b
FM		 4637 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4638 BUG_ON(journal != NULL);
4639 return;
4640 }
dab291af 4641 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4642 if (jbd2_journal_flush(journal) < 0)
4643 goto out;
4644
617ba13b 4645 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4646 sb->s_flags & MS_RDONLY) {
617ba13b 4647 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4648 ext4_commit_super(sb, 1);
ac27a0ec 4649 }
7ffe1ea8
HK		 4650
4651out:
dab291af 4652 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4653}
4654
4655/*
4656 * If we are mounting (or read-write remounting) a filesystem whose journal
4657 * has recorded an error from a previous lifetime, move that error to the
4658 * main filesystem now.
4659 */
2b2d6d01
TT		 4660static void ext4_clear_journal_err(struct super_block *sb,
4661 struct ext4_super_block *es)
ac27a0ec
DK		 4662{
4663 journal_t *journal;
4664 int j_errno;
4665 const char *errstr;
4666
0390131b
FM		 4667 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4668
617ba13b 4669 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4670
4671 /*
4672 * Now check for any error status which may have been recorded in the
617ba13b 4673 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4674 */
4675
dab291af 4676 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4677 if (j_errno) {
4678 char nbuf[16];
4679
617ba13b 4680 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4681 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4682 "from previous mount: %s", errstr);
12062ddd 4683 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4684
617ba13b
MC		 4685 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4686 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4687 ext4_commit_super(sb, 1);
ac27a0ec 4688
dab291af 4689 jbd2_journal_clear_err(journal);
d796c52e 4690 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4691 }
4692}
4693
4694/*
4695 * Force the running and committing transactions to commit,
4696 * and wait on the commit.
4697 */
617ba13b 4698int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4699{
4700 journal_t *journal;
ac27a0ec
DK		 4701
4702 if (sb->s_flags & MS_RDONLY)
4703 return 0;
4704
617ba13b 4705 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4706 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4707}
4708
617ba13b 4709static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4710{
14ce0cb4 4711 int ret = 0;
9eddacf9 4712 tid_t target;
06a407f1 4713 bool needs_barrier = false;
8d5d02e6 4714 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4715
9bffad1e 4716 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4717 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4718 /*
4719 * Writeback quota in non-journalled quota case - journalled quota has
4720 * no dirty dquots
4721 */
4722 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4723 /*
4724 * Data writeback is possible w/o journal transaction, so barrier must
4725 * being sent at the end of the function. But we can skip it if
4726 * transaction_commit will do it for us.
4727 */
bda32530
TT		 4728 if (sbi->s_journal) {
4729 target = jbd2_get_latest_transaction(sbi->s_journal);
4730 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4731 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4732 needs_barrier = true;
4733
4734 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4735 if (wait)
4736 ret = jbd2_log_wait_commit(sbi->s_journal,
4737 target);
4738 }
4739 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4740 needs_barrier = true;
06a407f1
DM		 4741 if (needs_barrier) {
4742 int err;
4743 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4744 if (!ret)
4745 ret = err;
0390131b 4746 }
06a407f1
DM		 4747
4748 return ret;
4749}
4750
ac27a0ec
DK		 4751/*
4752 * LVM calls this function before a (read-only) snapshot is created. This
4753 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4754 *
4755 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4756 * state independently. It relies on upper layer to stop all data & metadata
4757 * modifications.
ac27a0ec 4758 */
c4be0c1d 4759static int ext4_freeze(struct super_block *sb)
ac27a0ec 4760{
c4be0c1d
TS		 4761 int error = 0;
4762 journal_t *journal;
ac27a0ec 4763
9ca92389
TT		 4764 if (sb->s_flags & MS_RDONLY)
4765 return 0;
ac27a0ec 4766
9ca92389 4767 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4768
bb044576
TT		 4769 if (journal) {
4770 /* Now we set up the journal barrier. */
4771 jbd2_journal_lock_updates(journal);
ac27a0ec 4772
bb044576
TT		 4773 /*
4774 * Don't clear the needs_recovery flag if we failed to
4775 * flush the journal.
4776 */
4777 error = jbd2_journal_flush(journal);
4778 if (error < 0)
4779 goto out;
4780 }
9ca92389
TT		 4781
4782 /* Journal blocked and flushed, clear needs_recovery flag. */
4783 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4784 error = ext4_commit_super(sb, 1);
6b0310fb 4785out:
bb044576
TT		 4786 if (journal)
4787 /* we rely on upper layer to stop further updates */
4788 jbd2_journal_unlock_updates(journal);
6b0310fb 4789 return error;
ac27a0ec
DK		 4790}
4791
4792/*
4793 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4794 * flag here, even though the filesystem is not technically dirty yet.
4795 */
c4be0c1d 4796static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4797{
9ca92389
TT		 4798 if (sb->s_flags & MS_RDONLY)
4799 return 0;
4800
9ca92389
TT		 4801 /* Reset the needs_recovery flag before the fs is unlocked. */
4802 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4803 ext4_commit_super(sb, 1);
c4be0c1d 4804 return 0;
ac27a0ec
DK		 4805}
4806
673c6100
TT		 4807/*
4808 * Structure to save mount options for ext4_remount's benefit
4809 */
4810struct ext4_mount_options {
4811 unsigned long s_mount_opt;
a2595b8a 4812 unsigned long s_mount_opt2;
08cefc7a
EB		 4813 kuid_t s_resuid;
4814 kgid_t s_resgid;
673c6100
TT		 4815 unsigned long s_commit_interval;
4816 u32 s_min_batch_time, s_max_batch_time;
4817#ifdef CONFIG_QUOTA
4818 int s_jquota_fmt;
a2d4a646 4819 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4820#endif
4821};
4822
2b2d6d01 4823static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4824{
2b2d6d01 4825 struct ext4_super_block *es;
617ba13b 4826 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4827 unsigned long old_sb_flags;
617ba13b 4828 struct ext4_mount_options old_opts;
c79d967d 4829 int enable_quota = 0;
8a266467 4830 ext4_group_t g;
b3881f74 4831 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4832 int err = 0;
ac27a0ec 4833#ifdef CONFIG_QUOTA
03dafb5f 4834 int i, j;
ac27a0ec 4835#endif
d4c402d9 4836 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4837
4838 /* Store the original options */
4839 old_sb_flags = sb->s_flags;
4840 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4841 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4842 old_opts.s_resuid = sbi->s_resuid;
4843 old_opts.s_resgid = sbi->s_resgid;
4844 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4845 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4846 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4847#ifdef CONFIG_QUOTA
4848 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4849 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4850 if (sbi->s_qf_names[i]) {
4851 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4852 GFP_KERNEL);
4853 if (!old_opts.s_qf_names[i]) {
4854 for (j = 0; j < i; j++)
4855 kfree(old_opts.s_qf_names[j]);
3e36a163 4856 kfree(orig_data);
03dafb5f
CG		 4857 return -ENOMEM;
4858 }
4859 } else
4860 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4861#endif
b3881f74
TT		 4862 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4863 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4864
661aa520 4865 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4866 err = -EINVAL;
4867 goto restore_opts;
4868 }
4869
6b992ff2 4870 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4871 test_opt(sb, JOURNAL_CHECKSUM)) {
4872 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4873 "during remount not supported; ignoring");
4874 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4875 }
4876
6ae6514b
PS		 4877 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4878 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4879 ext4_msg(sb, KERN_ERR, "can't mount with "
4880 "both data=journal and delalloc");
4881 err = -EINVAL;
4882 goto restore_opts;
4883 }
4884 if (test_opt(sb, DIOREAD_NOLOCK)) {
4885 ext4_msg(sb, KERN_ERR, "can't mount with "
4886 "both data=journal and dioread_nolock");
4887 err = -EINVAL;
4888 goto restore_opts;
4889 }
923ae0ff
RZ		 4890 if (test_opt(sb, DAX)) {
4891 ext4_msg(sb, KERN_ERR, "can't mount with "
4892 "both data=journal and dax");
4893 err = -EINVAL;
4894 goto restore_opts;
4895 }
4896 }
4897
4898 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4899 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4900 "dax flag with busy inodes while remounting");
4901 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4902 }
4903
4ab2f15b 4904 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4905 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4906
4907 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4908 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4909
4910 es = sbi->s_es;
4911
b3881f74 4912 if (sbi->s_journal) {
0390131b 4913 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4914 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4915 }
ac27a0ec 4916
661aa520 4917 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4918 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4919 err = -EROFS;
4920 goto restore_opts;
4921 }
4922
4923 if (*flags & MS_RDONLY) {
38c03b34
TT		 4924 err = sync_filesystem(sb);
4925 if (err < 0)
4926 goto restore_opts;
0f0dd62f
CH		 4927 err = dquot_suspend(sb, -1);
4928 if (err < 0)
c79d967d 4929 goto restore_opts;
c79d967d 4930
ac27a0ec
DK		 4931 /*
4932 * First of all, the unconditional stuff we have to do
4933 * to disable replay of the journal when we next remount
4934 */
4935 sb->s_flags |= MS_RDONLY;
4936
4937 /*
4938 * OK, test if we are remounting a valid rw partition
4939 * readonly, and if so set the rdonly flag and then
4940 * mark the partition as valid again.
4941 */
617ba13b
MC		 4942 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4943 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4944 es->s_state = cpu_to_le16(sbi->s_mount_state);
4945
a63c9eb2 4946 if (sbi->s_journal)
0390131b 4947 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4948 } else {
a13fb1a4 4949 /* Make sure we can mount this feature set readwrite */
2cb5cc8b
DW		 4950 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4951 EXT4_FEATURE_RO_COMPAT_READONLY) ||
4952 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4953 err = -EROFS;
4954 goto restore_opts;
4955 }
8a266467
TT		 4956 /*
4957 * Make sure the group descriptor checksums
0b8e58a1 4958 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4959 */
4960 for (g = 0; g < sbi->s_groups_count; g++) {
4961 struct ext4_group_desc *gdp =
4962 ext4_get_group_desc(sb, g, NULL);
4963
feb0ab32 4964 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4965 ext4_msg(sb, KERN_ERR,
4966 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4967 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4968 le16_to_cpu(gdp->bg_checksum));
4969 err = -EINVAL;
4970 goto restore_opts;
4971 }
4972 }
4973
ead6596b
ES		 4974 /*
4975 * If we have an unprocessed orphan list hanging
4976 * around from a previously readonly bdev mount,
4977 * require a full umount/remount for now.
4978 */
4979 if (es->s_last_orphan) {
b31e1552 4980 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4981 "remount RDWR because of unprocessed "
4982 "orphan inode list. Please "
b31e1552 4983 "umount/remount instead");
ead6596b
ES		 4984 err = -EINVAL;
4985 goto restore_opts;
4986 }
4987
ac27a0ec
DK		 4988 /*
4989 * Mounting a RDONLY partition read-write, so reread
4990 * and store the current valid flag. (It may have
4991 * been changed by e2fsck since we originally mounted
4992 * the partition.)
4993 */
0390131b
FM		 4994 if (sbi->s_journal)
4995 ext4_clear_journal_err(sb, es);
ac27a0ec 4996 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4997 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4998 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4999 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5000 EXT4_FEATURE_INCOMPAT_MMP))
5001 if (ext4_multi_mount_protect(sb,
5002 le64_to_cpu(es->s_mmp_block))) {
5003 err = -EROFS;
5004 goto restore_opts;
5005 }
c79d967d 5006 enable_quota = 1;
ac27a0ec
DK		 5007 }
5008 }
bfff6873
LC		 5009
5010 /*
5011 * Reinitialize lazy itable initialization thread based on
5012 * current settings
5013 */
5014 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5015 ext4_unregister_li_request(sb);
5016 else {
5017 ext4_group_t first_not_zeroed;
5018 first_not_zeroed = ext4_has_uninit_itable(sb);
5019 ext4_register_li_request(sb, first_not_zeroed);
5020 }
5021
6fd058f7 5022 ext4_setup_system_zone(sb);
d096ad0f 5023 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5024 ext4_commit_super(sb, 1);
0390131b 5025
ac27a0ec
DK		 5026#ifdef CONFIG_QUOTA
5027 /* Release old quota file names */
a2d4a646 5028 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5029 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5030 if (enable_quota) {
5031 if (sb_any_quota_suspended(sb))
5032 dquot_resume(sb, -1);
5033 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5034 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5035 err = ext4_enable_quotas(sb);
07724f98 5036 if (err)
7c319d32 5037 goto restore_opts;
7c319d32
AK		 5038 }
5039 }
ac27a0ec 5040#endif
d4c402d9 5041
a26f4992 5042 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5043 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5044 kfree(orig_data);
ac27a0ec 5045 return 0;
0b8e58a1 5046
ac27a0ec
DK		 5047restore_opts:
5048 sb->s_flags = old_sb_flags;
5049 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5050 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5051 sbi->s_resuid = old_opts.s_resuid;
5052 sbi->s_resgid = old_opts.s_resgid;
5053 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5054 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5055 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5056#ifdef CONFIG_QUOTA
5057 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5058 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5059 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5060 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5061 }
5062#endif
d4c402d9 5063 kfree(orig_data);
ac27a0ec
DK		 5064 return err;
5065}
5066
2b2d6d01 5067static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5068{
5069 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5070 struct ext4_sb_info *sbi = EXT4_SB(sb);
5071 struct ext4_super_block *es = sbi->s_es;
27dd4385 5072 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5073 u64 fsid;
d02a9391 5074 s64 bfree;
27dd4385 5075 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5076
952fc18e
TT		 5077 if (!test_opt(sb, MINIX_DF))
5078 overhead = sbi->s_overhead;
ac27a0ec 5079
617ba13b 5080 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5081 buf->f_bsize = sb->s_blocksize;
b72f78cb 5082 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5083 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5084 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5085 /* prevent underflow in case that few free space is available */
57042651 5086 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5087 buf->f_bavail = buf->f_bfree -
5088 (ext4_r_blocks_count(es) + resv_blocks);
5089 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5090 buf->f_bavail = 0;
5091 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5092 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5093 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5094 fsid = le64_to_cpup((void *)es->s_uuid) ^
5095 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5096 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5097 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5098
ac27a0ec
DK		 5099 return 0;
5100}
5101
0b8e58a1
AD		 5102/* Helper function for writing quotas on sync - we need to start transaction
5103 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5104 * Process 1 Process 2
617ba13b 5105 * ext4_create() quota_sync()
a269eb18 5106 * jbd2_journal_start() write_dquot()
871a2931 5107 * dquot_initialize() down(dqio_mutex)
dab291af 5108 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5109 *
5110 */
5111
5112#ifdef CONFIG_QUOTA
5113
5114static inline struct inode *dquot_to_inode(struct dquot *dquot)
5115{
4c376dca 5116 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5117}
5118
617ba13b 5119static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5120{
5121 int ret, err;
5122 handle_t *handle;
5123 struct inode *inode;
5124
5125 inode = dquot_to_inode(dquot);
9924a92a 5126 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5127 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5128 if (IS_ERR(handle))
5129 return PTR_ERR(handle);
5130 ret = dquot_commit(dquot);
617ba13b 5131 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5132 if (!ret)
5133 ret = err;
5134 return ret;
5135}
5136
617ba13b 5137static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5138{
5139 int ret, err;
5140 handle_t *handle;
5141
9924a92a 5142 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5143 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5144 if (IS_ERR(handle))
5145 return PTR_ERR(handle);
5146 ret = dquot_acquire(dquot);
617ba13b 5147 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5148 if (!ret)
5149 ret = err;
5150 return ret;
5151}
5152
617ba13b 5153static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5154{
5155 int ret, err;
5156 handle_t *handle;
5157
9924a92a 5158 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5159 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5160 if (IS_ERR(handle)) {
5161 /* Release dquot anyway to avoid endless cycle in dqput() */
5162 dquot_release(dquot);
ac27a0ec 5163 return PTR_ERR(handle);
9c3013e9 5164 }
ac27a0ec 5165 ret = dquot_release(dquot);
617ba13b 5166 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5167 if (!ret)
5168 ret = err;
5169 return ret;
5170}
5171
617ba13b 5172static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5173{
262b4662
JK		 5174 struct super_block *sb = dquot->dq_sb;
5175 struct ext4_sb_info *sbi = EXT4_SB(sb);
5176
2c8be6b2 5177 /* Are we journaling quotas? */
262b4662
JK		 5178 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5179 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5180 dquot_mark_dquot_dirty(dquot);
617ba13b 5181 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5182 } else {
5183 return dquot_mark_dquot_dirty(dquot);
5184 }
5185}
5186
617ba13b 5187static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5188{
5189 int ret, err;
5190 handle_t *handle;
5191
5192 /* Data block + inode block */
9924a92a 5193 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5194 if (IS_ERR(handle))
5195 return PTR_ERR(handle);
5196 ret = dquot_commit_info(sb, type);
617ba13b 5197 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5198 if (!ret)
5199 ret = err;
5200 return ret;
5201}
5202
5203/*
5204 * Turn on quotas during mount time - we need to find
5205 * the quota file and such...
5206 */
617ba13b 5207static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5208{
287a8095
CH		 5209 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5210 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5211}
5212
5213/*
5214 * Standard function to be called on quota_on
5215 */
617ba13b 5216static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5217 struct path *path)
ac27a0ec
DK		 5218{
5219 int err;
ac27a0ec
DK		 5220
5221 if (!test_opt(sb, QUOTA))
5222 return -EINVAL;
0623543b 5223
ac27a0ec 5224 /* Quotafile not on the same filesystem? */
d8c9584e 5225 if (path->dentry->d_sb != sb)
ac27a0ec 5226 return -EXDEV;
0623543b
JK		 5227 /* Journaling quota? */
5228 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5229 /* Quotafile not in fs root? */
f00c9e44 5230 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5231 ext4_msg(sb, KERN_WARNING,
5232 "Quota file not on filesystem root. "
5233 "Journaled quota will not work");
2b2d6d01 5234 }
0623543b
JK		 5235
5236 /*
5237 * When we journal data on quota file, we have to flush journal to see
5238 * all updates to the file when we bypass pagecache...
5239 */
0390131b 5240 if (EXT4_SB(sb)->s_journal &&
f00c9e44 5241 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 5242 /*
5243 * We don't need to lock updates but journal_flush() could
5244 * otherwise be livelocked...
5245 */
5246 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5247 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5248 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5249 if (err)
7ffe1ea8 5250 return err;
0623543b
JK		 5251 }
5252
f00c9e44 5253 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5254}
5255
7c319d32
AK		 5256static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5257 unsigned int flags)
5258{
5259 int err;
5260 struct inode *qf_inode;
a2d4a646 5261 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5262 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5263 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5264 };
5265
5266 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5267
5268 if (!qf_inums[type])
5269 return -EPERM;
5270
5271 qf_inode = ext4_iget(sb, qf_inums[type]);
5272 if (IS_ERR(qf_inode)) {
5273 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5274 return PTR_ERR(qf_inode);
5275 }
5276
bcb13850
JK		 5277 /* Don't account quota for quota files to avoid recursion */
5278 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5279 err = dquot_enable(qf_inode, type, format_id, flags);
5280 iput(qf_inode);
5281
5282 return err;
5283}
5284
5285/* Enable usage tracking for all quota types. */
5286static int ext4_enable_quotas(struct super_block *sb)
5287{
5288 int type, err = 0;
a2d4a646 5289 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5290 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5291 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5292 };
5293
5294 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5295 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5296 if (qf_inums[type]) {
5297 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5298 DQUOT_USAGE_ENABLED);
5299 if (err) {
5300 ext4_warning(sb,
72ba7450
TT		 5301 "Failed to enable quota tracking "
5302 "(type=%d, err=%d). Please run "
5303 "e2fsck to fix.", type, err);
7c319d32
AK		 5304 return err;
5305 }
5306 }
5307 }
5308 return 0;
5309}
5310
ca0e05e4
DM		 5311static int ext4_quota_off(struct super_block *sb, int type)
5312{
21f97697
JK		 5313 struct inode *inode = sb_dqopt(sb)->files[type];
5314 handle_t *handle;
5315
87009d86
DM		 5316 /* Force all delayed allocation blocks to be allocated.
5317 * Caller already holds s_umount sem */
5318 if (test_opt(sb, DELALLOC))
ca0e05e4 5319 sync_filesystem(sb);
ca0e05e4 5320
0b268590
AG		 5321 if (!inode)
5322 goto out;
5323
21f97697
JK		 5324 /* Update modification times of quota files when userspace can
5325 * start looking at them */
9924a92a 5326 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5327 if (IS_ERR(handle))
5328 goto out;
5329 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5330 ext4_mark_inode_dirty(handle, inode);
5331 ext4_journal_stop(handle);
5332
5333out:
ca0e05e4
DM		 5334 return dquot_quota_off(sb, type);
5335}
5336
ac27a0ec
DK		 5337/* Read data from quotafile - avoid pagecache and such because we cannot afford
5338 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5339 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5340 * we don't have to be afraid of races */
617ba13b 5341static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5342 size_t len, loff_t off)
5343{
5344 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5345 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5346 int offset = off & (sb->s_blocksize - 1);
5347 int tocopy;
5348 size_t toread;
5349 struct buffer_head *bh;
5350 loff_t i_size = i_size_read(inode);
5351
5352 if (off > i_size)
5353 return 0;
5354 if (off+len > i_size)
5355 len = i_size-off;
5356 toread = len;
5357 while (toread > 0) {
5358 tocopy = sb->s_blocksize - offset < toread ?
5359 sb->s_blocksize - offset : toread;
1c215028
TT		 5360 bh = ext4_bread(NULL, inode, blk, 0);
5361 if (IS_ERR(bh))
5362 return PTR_ERR(bh);
ac27a0ec
DK		 5363 if (!bh) /* A hole? */
5364 memset(data, 0, tocopy);
5365 else
5366 memcpy(data, bh->b_data+offset, tocopy);
5367 brelse(bh);
5368 offset = 0;
5369 toread -= tocopy;
5370 data += tocopy;
5371 blk++;
5372 }
5373 return len;
5374}
5375
5376/* Write to quotafile (we know the transaction is already started and has
5377 * enough credits) */
617ba13b 5378static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5379 const char *data, size_t len, loff_t off)
5380{
5381 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5382 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5383 int err, offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5384 struct buffer_head *bh;
5385 handle_t *handle = journal_current_handle();
5386
0390131b 5387 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5388 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5389 " cancelled because transaction is not started",
9c3013e9
JK		 5390 (unsigned long long)off, (unsigned long long)len);
5391 return -EIO;
5392 }
67eeb568
DM		 5393 /*
5394 * Since we account only one data block in transaction credits,
5395 * then it is impossible to cross a block boundary.
5396 */
5397 if (sb->s_blocksize - offset < len) {
5398 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5399 " cancelled because not block aligned",
5400 (unsigned long long)off, (unsigned long long)len);
5401 return -EIO;
5402 }
5403
1c215028
TT		 5404 bh = ext4_bread(handle, inode, blk, 1);
5405 if (IS_ERR(bh))
5406 return PTR_ERR(bh);
67eeb568
DM		 5407 if (!bh)
5408 goto out;
5d601255 5409 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5410 err = ext4_journal_get_write_access(handle, bh);
5411 if (err) {
5412 brelse(bh);
1c215028 5413 return err;
ac27a0ec 5414 }
67eeb568
DM		 5415 lock_buffer(bh);
5416 memcpy(bh->b_data+offset, data, len);
5417 flush_dcache_page(bh->b_page);
5418 unlock_buffer(bh);
62d2b5f2 5419 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5420 brelse(bh);
ac27a0ec 5421out:
67eeb568
DM		 5422 if (inode->i_size < off + len) {
5423 i_size_write(inode, off + len);
617ba13b 5424 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5425 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5426 }
67eeb568 5427 return len;
ac27a0ec
DK		 5428}
5429
5430#endif
5431
152a0836
AV		 5432static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5433 const char *dev_name, void *data)
ac27a0ec 5434{
152a0836 5435 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5436}
5437
37f328eb 5438#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5439static inline void register_as_ext2(void)
5440{
5441 int err = register_filesystem(&ext2_fs_type);
5442 if (err)
5443 printk(KERN_WARNING
5444 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5445}
5446
5447static inline void unregister_as_ext2(void)
5448{
5449 unregister_filesystem(&ext2_fs_type);
5450}
2035e776
TT		 5451
5452static inline int ext2_feature_set_ok(struct super_block *sb)
5453{
5454 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5455 return 0;
5456 if (sb->s_flags & MS_RDONLY)
5457 return 1;
5458 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5459 return 0;
5460 return 1;
5461}
24b58424
TT		 5462#else
5463static inline void register_as_ext2(void) { }
5464static inline void unregister_as_ext2(void) { }
2035e776 5465static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5466#endif
5467
37f328eb 5468#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5469static inline void register_as_ext3(void)
5470{
5471 int err = register_filesystem(&ext3_fs_type);
5472 if (err)
5473 printk(KERN_WARNING
5474 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5475}
5476
5477static inline void unregister_as_ext3(void)
5478{
5479 unregister_filesystem(&ext3_fs_type);
5480}
2035e776
TT		 5481
5482static inline int ext3_feature_set_ok(struct super_block *sb)
5483{
5484 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5485 return 0;
5486 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5487 return 0;
5488 if (sb->s_flags & MS_RDONLY)
5489 return 1;
5490 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5491 return 0;
5492 return 1;
5493}
24b58424
TT		 5494#else
5495static inline void register_as_ext3(void) { }
5496static inline void unregister_as_ext3(void) { }
2035e776 5497static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5498#endif
5499
03010a33
TT		 5500static struct file_system_type ext4_fs_type = {
5501 .owner = THIS_MODULE,
5502 .name = "ext4",
152a0836 5503 .mount = ext4_mount,
03010a33
TT		 5504 .kill_sb = kill_block_super,
5505 .fs_flags = FS_REQUIRES_DEV,
5506};
7f78e035 5507MODULE_ALIAS_FS("ext4");
03010a33 5508
8f021222 5509static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5510{
5511 struct ext4_features *ef;
5512 int ret = -ENOMEM;
5513
5514 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5515 if (!ef)
5516 goto out;
5517
5518 ef->f_kobj.kset = ext4_kset;
5519 init_completion(&ef->f_kobj_unregister);
5520 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5521 "features");
5522 if (ret) {
5523 kfree(ef);
5524 goto out;
5525 }
5526
5527 ext4_feat = ef;
5528 ret = 0;
5529out:
5530 return ret;
5531}
5532
8f021222
LC		 5533static void ext4_exit_feat_adverts(void)
5534{
5535 kobject_put(&ext4_feat->f_kobj);
5536 wait_for_completion(&ext4_feat->f_kobj_unregister);
5537 kfree(ext4_feat);
5538}
5539
e9e3bcec
ES		 5540/* Shared across all ext4 file systems */
5541wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5542struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5543
5dabfc78 5544static int __init ext4_init_fs(void)
ac27a0ec 5545{
e9e3bcec 5546 int i, err;
c9de560d 5547
07c0c5d8
AV		 5548 ext4_li_info = NULL;
5549 mutex_init(&ext4_li_mtx);
5550
9a4c8019 5551 /* Build-time check for flags consistency */
12e9b892 5552 ext4_check_flag_values();
e9e3bcec
ES		 5553
5554 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5555 mutex_init(&ext4__aio_mutex[i]);
5556 init_waitqueue_head(&ext4__ioend_wq[i]);
5557 }
5558
51865fda 5559 err = ext4_init_es();
6fd058f7
TT		 5560 if (err)
5561 return err;
51865fda
ZL		 5562
5563 err = ext4_init_pageio();
5564 if (err)
5565 goto out7;
5566
5dabfc78 5567 err = ext4_init_system_zone();
bd2d0210 5568 if (err)
d44651d0 5569 goto out6;
3197ebdb 5570 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5571 if (!ext4_kset) {
5572 err = -ENOMEM;
dd68314c 5573 goto out5;
0e376b1e 5574 }
d44651d0 5575 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5576
5577 err = ext4_init_feat_adverts();
dd68314c
TT		 5578 if (err)
5579 goto out4;
857ac889 5580
5dabfc78 5581 err = ext4_init_mballoc();
c9de560d
AT		 5582 if (err)
5583 goto out2;
9c191f70
M		 5584 else
5585 ext4_mballoc_ready = 1;
ac27a0ec
DK		 5586 err = init_inodecache();
5587 if (err)
5588 goto out1;
24b58424 5589 register_as_ext3();
2035e776 5590 register_as_ext2();
03010a33 5591 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5592 if (err)
5593 goto out;
bfff6873 5594
ac27a0ec
DK		 5595 return 0;
5596out:
24b58424
TT		 5597 unregister_as_ext2();
5598 unregister_as_ext3();
ac27a0ec
DK		 5599 destroy_inodecache();
5600out1:
9c191f70 5601 ext4_mballoc_ready = 0;
5dabfc78 5602 ext4_exit_mballoc();
9c191f70 5603out2:
8f021222 5604 ext4_exit_feat_adverts();
dd68314c 5605out4:
d44651d0
FJ		 5606 if (ext4_proc_root)
5607 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5608 kset_unregister(ext4_kset);
d44651d0 5609out5:
5dabfc78 5610 ext4_exit_system_zone();
d44651d0 5611out6:
5dabfc78 5612 ext4_exit_pageio();
51865fda
ZL		 5613out7:
5614 ext4_exit_es();
5615
ac27a0ec
DK		 5616 return err;
5617}
5618
5dabfc78 5619static void __exit ext4_exit_fs(void)
ac27a0ec 5620{
bfff6873 5621 ext4_destroy_lazyinit_thread();
24b58424
TT		 5622 unregister_as_ext2();
5623 unregister_as_ext3();
03010a33 5624 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5625 destroy_inodecache();
5dabfc78 5626 ext4_exit_mballoc();
8f021222 5627 ext4_exit_feat_adverts();
9f6200bb 5628 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5629 kset_unregister(ext4_kset);
5dabfc78
TT		 5630 ext4_exit_system_zone();
5631 ext4_exit_pageio();
dd12ed14 5632 ext4_exit_es();
ac27a0ec
DK		 5633}
5634
5635MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5636MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5637MODULE_LICENSE("GPL");
5dabfc78
TT		 5638module_init(ext4_init_fs)
5639module_exit(ext4_exit_fs)
Linux kernel - Deliverables
This page took 2.260728 seconds and 5 git commands to generate.