projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[PATCH] fix calculation of meta_bg descriptor backups
[GitHub/LineageOS/android_kernel_motorola_exynos9610.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
66114cad 27#include <linux/backing-dev.h>
ac27a0ec 28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
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
0b75a840
LC		 56static struct ext4_lazy_init *ext4_li_info;
57static struct mutex ext4_li_mtx;
9c191f70 58static int ext4_mballoc_ready;
e294a537 59static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 60
617ba13b 61static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 62 unsigned long journal_devnum);
2adf6da8 63static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 64static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 65static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
617ba13b 69static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 70static int ext4_remount(struct super_block *sb, int *flags, char *data);
71static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 72static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 73static int ext4_freeze(struct super_block *sb);
152a0836
AV		 74static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
75 const char *dev_name, void *data);
2035e776
TT		 76static inline int ext2_feature_set_ok(struct super_block *sb);
77static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 78static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 79static void ext4_destroy_lazyinit_thread(void);
80static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 81static void ext4_clear_request_list(void);
ac27a0ec 82
c290ea01 83#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 84static struct file_system_type ext2_fs_type = {
85 .owner = THIS_MODULE,
86 .name = "ext2",
87 .mount = ext4_mount,
88 .kill_sb = kill_block_super,
89 .fs_flags = FS_REQUIRES_DEV,
90};
7f78e035 91MODULE_ALIAS_FS("ext2");
fa7614dd 92MODULE_ALIAS("ext2");
2035e776
TT		 93#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
94#else
95#define IS_EXT2_SB(sb) (0)
96#endif
97
98
ba69f9ab
JK		 99static struct file_system_type ext3_fs_type = {
100 .owner = THIS_MODULE,
101 .name = "ext3",
152a0836 102 .mount = ext4_mount,
ba69f9ab
JK		 103 .kill_sb = kill_block_super,
104 .fs_flags = FS_REQUIRES_DEV,
105};
7f78e035 106MODULE_ALIAS_FS("ext3");
fa7614dd 107MODULE_ALIAS("ext3");
ba69f9ab 108#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 109
d25425f8
DW		 110static int ext4_verify_csum_type(struct super_block *sb,
111 struct ext4_super_block *es)
112{
e2b911c5 113 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 114 return 1;
115
116 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
117}
118
a9c47317
DW		 119static __le32 ext4_superblock_csum(struct super_block *sb,
120 struct ext4_super_block *es)
121{
122 struct ext4_sb_info *sbi = EXT4_SB(sb);
123 int offset = offsetof(struct ext4_super_block, s_checksum);
124 __u32 csum;
125
126 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
127
128 return cpu_to_le32(csum);
129}
130
c197855e
SH		 131static int ext4_superblock_csum_verify(struct super_block *sb,
132 struct ext4_super_block *es)
a9c47317 133{
9aa5d32b 134 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 135 return 1;
136
137 return es->s_checksum == ext4_superblock_csum(sb, es);
138}
139
06db49e6 140void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 141{
06db49e6
TT		 142 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
143
9aa5d32b 144 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 145 return;
146
147 es->s_checksum = ext4_superblock_csum(sb, es);
148}
149
9933fc0a
TT		 150void *ext4_kvmalloc(size_t size, gfp_t flags)
151{
152 void *ret;
153
8be04b93 154 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 155 if (!ret)
156 ret = __vmalloc(size, flags, PAGE_KERNEL);
157 return ret;
158}
159
160void *ext4_kvzalloc(size_t size, gfp_t flags)
161{
162 void *ret;
163
8be04b93 164 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 165 if (!ret)
166 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
167 return ret;
168}
169
8fadc143
AR		 170ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
171 struct ext4_group_desc *bg)
bd81d8ee 172{
3a14589c 173 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 174 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 175 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 176}
177
8fadc143
AR		 178ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
179 struct ext4_group_desc *bg)
bd81d8ee 180{
5272f837 181 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 182 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 183 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 184}
185
8fadc143
AR		 186ext4_fsblk_t ext4_inode_table(struct super_block *sb,
187 struct ext4_group_desc *bg)
bd81d8ee 188{
5272f837 189 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 191 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 192}
193
021b65bb
TT		 194__u32 ext4_free_group_clusters(struct super_block *sb,
195 struct ext4_group_desc *bg)
560671a0
AK		 196{
197 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 199 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 200}
201
202__u32 ext4_free_inodes_count(struct super_block *sb,
203 struct ext4_group_desc *bg)
204{
205 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 207 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 208}
209
210__u32 ext4_used_dirs_count(struct super_block *sb,
211 struct ext4_group_desc *bg)
212{
213 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 215 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 216}
217
218__u32 ext4_itable_unused_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220{
221 return le16_to_cpu(bg->bg_itable_unused_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 223 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 224}
225
8fadc143
AR		 226void ext4_block_bitmap_set(struct super_block *sb,
227 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 228{
3a14589c 229 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 230 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
231 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 232}
233
8fadc143
AR		 234void ext4_inode_bitmap_set(struct super_block *sb,
235 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 236{
5272f837 237 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 238 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 240}
241
8fadc143
AR		 242void ext4_inode_table_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 244{
5272f837 245 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 248}
249
021b65bb
TT		 250void ext4_free_group_clusters_set(struct super_block *sb,
251 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 252{
253 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
256}
257
258void ext4_free_inodes_set(struct super_block *sb,
259 struct ext4_group_desc *bg, __u32 count)
260{
261 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
264}
265
266void ext4_used_dirs_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
268{
269 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
272}
273
274void ext4_itable_unused_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276{
277 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
280}
281
d3d1faf6 282
1c13d5c0
TT		 283static void __save_error_info(struct super_block *sb, const char *func,
284 unsigned int line)
285{
286 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
287
288 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 289 if (bdev_read_only(sb->s_bdev))
290 return;
1c13d5c0
TT		 291 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
292 es->s_last_error_time = cpu_to_le32(get_seconds());
293 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
294 es->s_last_error_line = cpu_to_le32(line);
295 if (!es->s_first_error_time) {
296 es->s_first_error_time = es->s_last_error_time;
297 strncpy(es->s_first_error_func, func,
298 sizeof(es->s_first_error_func));
299 es->s_first_error_line = cpu_to_le32(line);
300 es->s_first_error_ino = es->s_last_error_ino;
301 es->s_first_error_block = es->s_last_error_block;
302 }
66e61a9e
TT		 303 /*
304 * Start the daily error reporting function if it hasn't been
305 * started already
306 */
307 if (!es->s_error_count)
308 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 309 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 310}
311
312static void save_error_info(struct super_block *sb, const char *func,
313 unsigned int line)
314{
315 __save_error_info(sb, func, line);
316 ext4_commit_super(sb, 1);
317}
318
bdfe0cbd
TT		 319/*
320 * The del_gendisk() function uninitializes the disk-specific data
321 * structures, including the bdi structure, without telling anyone
322 * else. Once this happens, any attempt to call mark_buffer_dirty()
323 * (for example, by ext4_commit_super), will cause a kernel OOPS.
324 * This is a kludge to prevent these oops until we can put in a proper
325 * hook in del_gendisk() to inform the VFS and file system layers.
326 */
327static int block_device_ejected(struct super_block *sb)
328{
329 struct inode *bd_inode = sb->s_bdev->bd_inode;
330 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
331
332 return bdi->dev == NULL;
333}
334
18aadd47
BJ		 335static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
336{
337 struct super_block *sb = journal->j_private;
338 struct ext4_sb_info *sbi = EXT4_SB(sb);
339 int error = is_journal_aborted(journal);
5d3ee208 340 struct ext4_journal_cb_entry *jce;
18aadd47 341
5d3ee208 342 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 343 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 344 while (!list_empty(&txn->t_private_list)) {
345 jce = list_entry(txn->t_private_list.next,
346 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 347 list_del_init(&jce->jce_list);
348 spin_unlock(&sbi->s_md_lock);
349 jce->jce_func(sb, jce, error);
350 spin_lock(&sbi->s_md_lock);
351 }
352 spin_unlock(&sbi->s_md_lock);
353}
1c13d5c0 354
ac27a0ec
DK		 355/* Deal with the reporting of failure conditions on a filesystem such as
356 * inconsistencies detected or read IO failures.
357 *
358 * On ext2, we can store the error state of the filesystem in the
617ba13b 359 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 360 * write ordering constraints on the superblock which prevent us from
361 * writing it out straight away; and given that the journal is about to
362 * be aborted, we can't rely on the current, or future, transactions to
363 * write out the superblock safely.
364 *
dab291af 365 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 366 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 367 * that error until we've noted it down and cleared it.
368 */
369
617ba13b 370static void ext4_handle_error(struct super_block *sb)
ac27a0ec 371{
ac27a0ec
DK		 372 if (sb->s_flags & MS_RDONLY)
373 return;
374
2b2d6d01 375 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 376 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 377
4ab2f15b 378 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 379 if (journal)
dab291af 380 jbd2_journal_abort(journal, -EIO);
ac27a0ec 381 }
2b2d6d01 382 if (test_opt(sb, ERRORS_RO)) {
b31e1552 383 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 384 /*
385 * Make sure updated value of ->s_mount_flags will be visible
386 * before ->s_flags update
387 */
388 smp_wmb();
ac27a0ec
DK		 389 sb->s_flags |= MS_RDONLY;
390 }
ac27a0ec 391 if (test_opt(sb, ERRORS_PANIC))
617ba13b 392 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 393 sb->s_id);
394}
395
efbed4dc
TT		 396#define ext4_error_ratelimit(sb) \
397 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
398 "EXT4-fs error")
399
12062ddd 400void __ext4_error(struct super_block *sb, const char *function,
c398eda0 401 unsigned int line, const char *fmt, ...)
ac27a0ec 402{
0ff2ea7d 403 struct va_format vaf;
ac27a0ec
DK		 404 va_list args;
405
efbed4dc
TT		 406 if (ext4_error_ratelimit(sb)) {
407 va_start(args, fmt);
408 vaf.fmt = fmt;
409 vaf.va = &args;
410 printk(KERN_CRIT
411 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
412 sb->s_id, function, line, current->comm, &vaf);
413 va_end(args);
414 }
f3fc0210 415 save_error_info(sb, function, line);
617ba13b 416 ext4_handle_error(sb);
ac27a0ec
DK		 417}
418
e7c96e8e
JP		 419void __ext4_error_inode(struct inode *inode, const char *function,
420 unsigned int line, ext4_fsblk_t block,
421 const char *fmt, ...)
273df556
FM		 422{
423 va_list args;
f7c21177 424 struct va_format vaf;
1c13d5c0 425 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 426
1c13d5c0
TT		 427 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
428 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 429 if (ext4_error_ratelimit(inode->i_sb)) {
430 va_start(args, fmt);
431 vaf.fmt = fmt;
432 vaf.va = &args;
433 if (block)
434 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
435 "inode #%lu: block %llu: comm %s: %pV\n",
436 inode->i_sb->s_id, function, line, inode->i_ino,
437 block, current->comm, &vaf);
438 else
439 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
440 "inode #%lu: comm %s: %pV\n",
441 inode->i_sb->s_id, function, line, inode->i_ino,
442 current->comm, &vaf);
443 va_end(args);
444 }
1c13d5c0 445 save_error_info(inode->i_sb, function, line);
273df556
FM		 446 ext4_handle_error(inode->i_sb);
447}
448
e7c96e8e
JP		 449void __ext4_error_file(struct file *file, const char *function,
450 unsigned int line, ext4_fsblk_t block,
451 const char *fmt, ...)
273df556
FM		 452{
453 va_list args;
f7c21177 454 struct va_format vaf;
1c13d5c0 455 struct ext4_super_block *es;
496ad9aa 456 struct inode *inode = file_inode(file);
273df556
FM		 457 char pathname[80], *path;
458
1c13d5c0
TT		 459 es = EXT4_SB(inode->i_sb)->s_es;
460 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 461 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 462 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 463 if (IS_ERR(path))
464 path = "(unknown)";
465 va_start(args, fmt);
466 vaf.fmt = fmt;
467 vaf.va = &args;
468 if (block)
469 printk(KERN_CRIT
470 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
471 "block %llu: comm %s: path %s: %pV\n",
472 inode->i_sb->s_id, function, line, inode->i_ino,
473 block, current->comm, path, &vaf);
474 else
475 printk(KERN_CRIT
476 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
477 "comm %s: path %s: %pV\n",
478 inode->i_sb->s_id, function, line, inode->i_ino,
479 current->comm, path, &vaf);
480 va_end(args);
481 }
1c13d5c0 482 save_error_info(inode->i_sb, function, line);
273df556
FM		 483 ext4_handle_error(inode->i_sb);
484}
485
722887dd
TT		 486const char *ext4_decode_error(struct super_block *sb, int errno,
487 char nbuf[16])
ac27a0ec
DK		 488{
489 char *errstr = NULL;
490
491 switch (errno) {
6a797d27
DW		 492 case -EFSCORRUPTED:
493 errstr = "Corrupt filesystem";
494 break;
495 case -EFSBADCRC:
496 errstr = "Filesystem failed CRC";
497 break;
ac27a0ec
DK		 498 case -EIO:
499 errstr = "IO failure";
500 break;
501 case -ENOMEM:
502 errstr = "Out of memory";
503 break;
504 case -EROFS:
78f1ddbb
TT		 505 if (!sb || (EXT4_SB(sb)->s_journal &&
506 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 507 errstr = "Journal has aborted";
508 else
509 errstr = "Readonly filesystem";
510 break;
511 default:
512 /* If the caller passed in an extra buffer for unknown
513 * errors, textualise them now. Else we just return
514 * NULL. */
515 if (nbuf) {
516 /* Check for truncated error codes... */
517 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
518 errstr = nbuf;
519 }
520 break;
521 }
522
523 return errstr;
524}
525
617ba13b 526/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 527 * automatically and invokes the appropriate error response. */
528
c398eda0
TT		 529void __ext4_std_error(struct super_block *sb, const char *function,
530 unsigned int line, int errno)
ac27a0ec
DK		 531{
532 char nbuf[16];
533 const char *errstr;
534
535 /* Special case: if the error is EROFS, and we're not already
536 * inside a transaction, then there's really no point in logging
537 * an error. */
538 if (errno == -EROFS && journal_current_handle() == NULL &&
539 (sb->s_flags & MS_RDONLY))
540 return;
541
efbed4dc
TT		 542 if (ext4_error_ratelimit(sb)) {
543 errstr = ext4_decode_error(sb, errno, nbuf);
544 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
545 sb->s_id, function, line, errstr);
546 }
ac27a0ec 547
efbed4dc 548 save_error_info(sb, function, line);
617ba13b 549 ext4_handle_error(sb);
ac27a0ec
DK		 550}
551
552/*
617ba13b 553 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 554 * abort function may be used to deal with unrecoverable failures such
555 * as journal IO errors or ENOMEM at a critical moment in log management.
556 *
557 * We unconditionally force the filesystem into an ABORT|READONLY state,
558 * unless the error response on the fs has been set to panic in which
559 * case we take the easy way out and panic immediately.
560 */
561
c67d859e 562void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 563 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 564{
565 va_list args;
566
1c13d5c0 567 save_error_info(sb, function, line);
ac27a0ec 568 va_start(args, fmt);
c398eda0
TT		 569 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
570 function, line);
ac27a0ec
DK		 571 vprintk(fmt, args);
572 printk("\n");
573 va_end(args);
574
1c13d5c0
TT		 575 if ((sb->s_flags & MS_RDONLY) == 0) {
576 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 577 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 578 /*
579 * Make sure updated value of ->s_mount_flags will be visible
580 * before ->s_flags update
581 */
582 smp_wmb();
583 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 584 if (EXT4_SB(sb)->s_journal)
585 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
586 save_error_info(sb, function, line);
587 }
ac27a0ec 588 if (test_opt(sb, ERRORS_PANIC))
617ba13b 589 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 590}
591
e7c96e8e
JP		 592void __ext4_msg(struct super_block *sb,
593 const char *prefix, const char *fmt, ...)
b31e1552 594{
0ff2ea7d 595 struct va_format vaf;
b31e1552
ES		 596 va_list args;
597
efbed4dc
TT		 598 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
599 return;
600
b31e1552 601 va_start(args, fmt);
0ff2ea7d
JP		 602 vaf.fmt = fmt;
603 vaf.va = &args;
604 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 605 va_end(args);
606}
607
b03a2f7e
AD		 608#define ext4_warning_ratelimit(sb) \
609 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
610 "EXT4-fs warning")
611
12062ddd 612void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 613 unsigned int line, const char *fmt, ...)
ac27a0ec 614{
0ff2ea7d 615 struct va_format vaf;
ac27a0ec
DK		 616 va_list args;
617
b03a2f7e 618 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 619 return;
620
ac27a0ec 621 va_start(args, fmt);
0ff2ea7d
JP		 622 vaf.fmt = fmt;
623 vaf.va = &args;
624 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
625 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 626 va_end(args);
627}
628
b03a2f7e
AD		 629void __ext4_warning_inode(const struct inode *inode, const char *function,
630 unsigned int line, const char *fmt, ...)
631{
632 struct va_format vaf;
633 va_list args;
634
635 if (!ext4_warning_ratelimit(inode->i_sb))
636 return;
637
638 va_start(args, fmt);
639 vaf.fmt = fmt;
640 vaf.va = &args;
641 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
642 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
643 function, line, inode->i_ino, current->comm, &vaf);
644 va_end(args);
645}
646
e29136f8
TT		 647void __ext4_grp_locked_error(const char *function, unsigned int line,
648 struct super_block *sb, ext4_group_t grp,
649 unsigned long ino, ext4_fsblk_t block,
650 const char *fmt, ...)
5d1b1b3f
AK		 651__releases(bitlock)
652__acquires(bitlock)
653{
0ff2ea7d 654 struct va_format vaf;
5d1b1b3f
AK		 655 va_list args;
656 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
657
1c13d5c0
TT		 658 es->s_last_error_ino = cpu_to_le32(ino);
659 es->s_last_error_block = cpu_to_le64(block);
660 __save_error_info(sb, function, line);
0ff2ea7d 661
efbed4dc
TT		 662 if (ext4_error_ratelimit(sb)) {
663 va_start(args, fmt);
664 vaf.fmt = fmt;
665 vaf.va = &args;
666 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
667 sb->s_id, function, line, grp);
668 if (ino)
669 printk(KERN_CONT "inode %lu: ", ino);
670 if (block)
671 printk(KERN_CONT "block %llu:",
672 (unsigned long long) block);
673 printk(KERN_CONT "%pV\n", &vaf);
674 va_end(args);
675 }
5d1b1b3f
AK		 676
677 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 678 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 679 return;
680 }
1c13d5c0 681
5d1b1b3f
AK		 682 ext4_unlock_group(sb, grp);
683 ext4_handle_error(sb);
684 /*
685 * We only get here in the ERRORS_RO case; relocking the group
686 * may be dangerous, but nothing bad will happen since the
687 * filesystem will have already been marked read/only and the
688 * journal has been aborted. We return 1 as a hint to callers
689 * who might what to use the return value from
25985edc 690 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 691 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
692 * aggressively from the ext4 function in question, with a
693 * more appropriate error code.
694 */
695 ext4_lock_group(sb, grp);
696 return;
697}
698
617ba13b 699void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 700{
617ba13b 701 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 702
617ba13b 703 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 704 return;
705
12062ddd 706 ext4_warning(sb,
ac27a0ec
DK		 707 "updating to rev %d because of new feature flag, "
708 "running e2fsck is recommended",
617ba13b 709 EXT4_DYNAMIC_REV);
ac27a0ec 710
617ba13b
MC		 711 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
712 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
713 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 714 /* leave es->s_feature_*compat flags alone */
715 /* es->s_uuid will be set by e2fsck if empty */
716
717 /*
718 * The rest of the superblock fields should be zero, and if not it
719 * means they are likely already in use, so leave them alone. We
720 * can leave it up to e2fsck to clean up any inconsistencies there.
721 */
722}
723
724/*
725 * Open the external journal device
726 */
b31e1552 727static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 728{
729 struct block_device *bdev;
730 char b[BDEVNAME_SIZE];
731
d4d77629 732 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 733 if (IS_ERR(bdev))
734 goto fail;
735 return bdev;
736
737fail:
b31e1552 738 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 739 __bdevname(dev, b), PTR_ERR(bdev));
740 return NULL;
741}
742
743/*
744 * Release the journal device
745 */
4385bab1 746static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 747{
4385bab1 748 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 749}
750
4385bab1 751static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 752{
753 struct block_device *bdev;
ac27a0ec
DK		 754 bdev = sbi->journal_bdev;
755 if (bdev) {
4385bab1 756 ext4_blkdev_put(bdev);
ac27a0ec
DK		 757 sbi->journal_bdev = NULL;
758 }
ac27a0ec
DK		 759}
760
761static inline struct inode *orphan_list_entry(struct list_head *l)
762{
617ba13b 763 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 764}
765
617ba13b 766static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 767{
768 struct list_head *l;
769
b31e1552
ES		 770 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
771 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 772
773 printk(KERN_ERR "sb_info orphan list:\n");
774 list_for_each(l, &sbi->s_orphan) {
775 struct inode *inode = orphan_list_entry(l);
776 printk(KERN_ERR " "
777 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
778 inode->i_sb->s_id, inode->i_ino, inode,
779 inode->i_mode, inode->i_nlink,
780 NEXT_ORPHAN(inode));
781 }
782}
783
2b2d6d01 784static void ext4_put_super(struct super_block *sb)
ac27a0ec 785{
617ba13b
MC		 786 struct ext4_sb_info *sbi = EXT4_SB(sb);
787 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 788 int i, err;
ac27a0ec 789
857ac889 790 ext4_unregister_li_request(sb);
e0ccfd95
CH		 791 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
792
2e8fa54e 793 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 794 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 795
0390131b
FM		 796 if (sbi->s_journal) {
797 err = jbd2_journal_destroy(sbi->s_journal);
798 sbi->s_journal = NULL;
799 if (err < 0)
c67d859e 800 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 801 }
d4edac31 802
ebd173be 803 ext4_unregister_sysfs(sb);
d3922a77 804 ext4_es_unregister_shrinker(sbi);
9105bb14 805 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 806 ext4_release_system_zone(sb);
807 ext4_mb_release(sb);
808 ext4_ext_release(sb);
809 ext4_xattr_put_super(sb);
810
ac27a0ec 811 if (!(sb->s_flags & MS_RDONLY)) {
e2b911c5 812 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 813 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 814 }
58c5873a 815 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 816 ext4_commit_super(sb, 1);
817
ac27a0ec
DK		 818 for (i = 0; i < sbi->s_gdb_count; i++)
819 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 820 kvfree(sbi->s_group_desc);
821 kvfree(sbi->s_flex_groups);
57042651 822 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 823 percpu_counter_destroy(&sbi->s_freeinodes_counter);
824 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 825 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 826 brelse(sbi->s_sbh);
827#ifdef CONFIG_QUOTA
a2d4a646 828 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 829 kfree(sbi->s_qf_names[i]);
830#endif
831
832 /* Debugging code just in case the in-memory inode orphan list
833 * isn't empty. The on-disk one can be non-empty if we've
834 * detected an error and taken the fs readonly, but the
835 * in-memory list had better be clean by this point. */
836 if (!list_empty(&sbi->s_orphan))
837 dump_orphan_list(sb, sbi);
838 J_ASSERT(list_empty(&sbi->s_orphan));
839
89d96a6f 840 sync_blockdev(sb->s_bdev);
f98393a6 841 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 842 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
843 /*
844 * Invalidate the journal device's buffers. We don't want them
845 * floating about in memory - the physical journal device may
846 * hotswapped, and it breaks the `ro-after' testing code.
847 */
848 sync_blockdev(sbi->journal_bdev);
f98393a6 849 invalidate_bdev(sbi->journal_bdev);
617ba13b 850 ext4_blkdev_remove(sbi);
ac27a0ec 851 }
9c191f70
M		 852 if (sbi->s_mb_cache) {
853 ext4_xattr_destroy_cache(sbi->s_mb_cache);
854 sbi->s_mb_cache = NULL;
855 }
c5e06d10
JL		 856 if (sbi->s_mmp_tsk)
857 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 858 sb->s_fs_info = NULL;
3197ebdb
TT		 859 /*
860 * Now that we are completely done shutting down the
861 * superblock, we need to actually destroy the kobject.
862 */
3197ebdb
TT		 863 kobject_put(&sbi->s_kobj);
864 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 865 if (sbi->s_chksum_driver)
866 crypto_free_shash(sbi->s_chksum_driver);
705895b6 867 kfree(sbi->s_blockgroup_lock);
ac27a0ec 868 kfree(sbi);
ac27a0ec
DK		 869}
870
e18b890b 871static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 872
873/*
874 * Called inside transaction, so use GFP_NOFS
875 */
617ba13b 876static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 877{
617ba13b 878 struct ext4_inode_info *ei;
ac27a0ec 879
e6b4f8da 880 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 881 if (!ei)
882 return NULL;
0b8e58a1 883
ac27a0ec 884 ei->vfs_inode.i_version = 1;
202ee5df 885 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 886 INIT_LIST_HEAD(&ei->i_prealloc_list);
887 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 888 ext4_es_init_tree(&ei->i_es_tree);
889 rwlock_init(&ei->i_es_lock);
edaa53ca 890 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 891 ei->i_es_all_nr = 0;
edaa53ca 892 ei->i_es_shk_nr = 0;
dd475925 893 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 894 ei->i_reserved_data_blocks = 0;
895 ei->i_reserved_meta_blocks = 0;
896 ei->i_allocated_meta_blocks = 0;
9d0be502 897 ei->i_da_metadata_calc_len = 0;
7e731bc9 898 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 899 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 900#ifdef CONFIG_QUOTA
901 ei->i_reserved_quota = 0;
96c7e0d9 902 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 903#endif
8aefcd55 904 ei->jinode = NULL;
2e8fa54e 905 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 906 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 907 ei->i_sync_tid = 0;
908 ei->i_datasync_tid = 0;
f7ad6d2e 909 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 910 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 911 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
b30ab0e0 912#ifdef CONFIG_EXT4_FS_ENCRYPTION
b7236e21 913 ei->i_crypt_info = NULL;
b30ab0e0 914#endif
ac27a0ec
DK		 915 return &ei->vfs_inode;
916}
917
7ff9c073
TT		 918static int ext4_drop_inode(struct inode *inode)
919{
920 int drop = generic_drop_inode(inode);
921
922 trace_ext4_drop_inode(inode, drop);
923 return drop;
924}
925
fa0d7e3d
NP		 926static void ext4_i_callback(struct rcu_head *head)
927{
928 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 929 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
930}
931
617ba13b 932static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 933{
9f7dd93d 934 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 935 ext4_msg(inode->i_sb, KERN_ERR,
936 "Inode %lu (%p): orphan list check failed!",
937 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 938 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
939 EXT4_I(inode), sizeof(struct ext4_inode_info),
940 true);
941 dump_stack();
942 }
fa0d7e3d 943 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 944}
945
51cc5068 946static void init_once(void *foo)
ac27a0ec 947{
617ba13b 948 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 949
a35afb83 950 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 951 init_rwsem(&ei->xattr_sem);
0e855ac8 952 init_rwsem(&ei->i_data_sem);
a35afb83 953 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 954}
955
e67bc2b3 956static int __init init_inodecache(void)
ac27a0ec 957{
617ba13b
MC		 958 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
959 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 960 0, (SLAB_RECLAIM_ACCOUNT|
961 SLAB_MEM_SPREAD),
20c2df83 962 init_once);
617ba13b 963 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 964 return -ENOMEM;
965 return 0;
966}
967
968static void destroy_inodecache(void)
969{
8c0a8537
KS		 970 /*
971 * Make sure all delayed rcu free inodes are flushed before we
972 * destroy cache.
973 */
974 rcu_barrier();
617ba13b 975 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 976}
977
0930fcc1 978void ext4_clear_inode(struct inode *inode)
ac27a0ec 979{
0930fcc1 980 invalidate_inode_buffers(inode);
dbd5768f 981 clear_inode(inode);
9f754758 982 dquot_drop(inode);
c2ea3fde 983 ext4_discard_preallocations(inode);
51865fda 984 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 985 if (EXT4_I(inode)->jinode) {
986 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
987 EXT4_I(inode)->jinode);
988 jbd2_free_inode(EXT4_I(inode)->jinode);
989 EXT4_I(inode)->jinode = NULL;
990 }
b7236e21
TT		 991#ifdef CONFIG_EXT4_FS_ENCRYPTION
992 if (EXT4_I(inode)->i_crypt_info)
c936e1ec 993 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
b7236e21 994#endif
ac27a0ec
DK		 995}
996
1b961ac0 997static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 998 u64 ino, u32 generation)
ac27a0ec 999{
ac27a0ec 1000 struct inode *inode;
ac27a0ec 1001
617ba13b 1002 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1003 return ERR_PTR(-ESTALE);
617ba13b 1004 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1005 return ERR_PTR(-ESTALE);
1006
1007 /* iget isn't really right if the inode is currently unallocated!!
1008 *
617ba13b 1009 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1010 * deleted, so we should be safe.
1011 *
1012 * Currently we don't know the generation for parent directory, so
1013 * a generation of 0 means "accept any"
1014 */
f4bb2981 1015 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1016 if (IS_ERR(inode))
1017 return ERR_CAST(inode);
1018 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1019 iput(inode);
1020 return ERR_PTR(-ESTALE);
1021 }
1b961ac0
CH		 1022
1023 return inode;
1024}
1025
1026static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1027 int fh_len, int fh_type)
1b961ac0
CH		 1028{
1029 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1030 ext4_nfs_get_inode);
1031}
1032
1033static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1034 int fh_len, int fh_type)
1b961ac0
CH		 1035{
1036 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1037 ext4_nfs_get_inode);
ac27a0ec
DK		 1038}
1039
c39a7f84
TO		 1040/*
1041 * Try to release metadata pages (indirect blocks, directories) which are
1042 * mapped via the block device. Since these pages could have journal heads
1043 * which would prevent try_to_free_buffers() from freeing them, we must use
1044 * jbd2 layer's try_to_free_buffers() function to release them.
1045 */
0b8e58a1
AD		 1046static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1047 gfp_t wait)
c39a7f84
TO		 1048{
1049 journal_t *journal = EXT4_SB(sb)->s_journal;
1050
1051 WARN_ON(PageChecked(page));
1052 if (!page_has_buffers(page))
1053 return 0;
1054 if (journal)
1055 return jbd2_journal_try_to_free_buffers(journal, page,
1056 wait & ~__GFP_WAIT);
1057 return try_to_free_buffers(page);
1058}
1059
ac27a0ec 1060#ifdef CONFIG_QUOTA
af5bc92d 1061#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1062#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1063
617ba13b
MC		 1064static int ext4_write_dquot(struct dquot *dquot);
1065static int ext4_acquire_dquot(struct dquot *dquot);
1066static int ext4_release_dquot(struct dquot *dquot);
1067static int ext4_mark_dquot_dirty(struct dquot *dquot);
1068static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1069static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1070 struct path *path);
ca0e05e4 1071static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1072static int ext4_quota_on_mount(struct super_block *sb, int type);
1073static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1074 size_t len, loff_t off);
617ba13b 1075static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1076 const char *data, size_t len, loff_t off);
7c319d32
AK		 1077static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1078 unsigned int flags);
1079static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1080
96c7e0d9
JK		 1081static struct dquot **ext4_get_dquots(struct inode *inode)
1082{
1083 return EXT4_I(inode)->i_dquot;
1084}
1085
61e225dc 1086static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1087 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1088 .write_dquot = ext4_write_dquot,
1089 .acquire_dquot = ext4_acquire_dquot,
1090 .release_dquot = ext4_release_dquot,
1091 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1092 .write_info = ext4_write_info,
1093 .alloc_dquot = dquot_alloc,
1094 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1095};
1096
0d54b217 1097static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1098 .quota_on = ext4_quota_on,
ca0e05e4 1099 .quota_off = ext4_quota_off,
287a8095 1100 .quota_sync = dquot_quota_sync,
0a240339 1101 .get_state = dquot_get_state,
287a8095
CH		 1102 .set_info = dquot_set_dqinfo,
1103 .get_dqblk = dquot_get_dqblk,
1104 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1105};
1106#endif
1107
ee9b6d61 1108static const struct super_operations ext4_sops = {
617ba13b
MC		 1109 .alloc_inode = ext4_alloc_inode,
1110 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1111 .write_inode = ext4_write_inode,
1112 .dirty_inode = ext4_dirty_inode,
7ff9c073 1113 .drop_inode = ext4_drop_inode,
0930fcc1 1114 .evict_inode = ext4_evict_inode,
617ba13b 1115 .put_super = ext4_put_super,
617ba13b 1116 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1117 .freeze_fs = ext4_freeze,
1118 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1119 .statfs = ext4_statfs,
1120 .remount_fs = ext4_remount,
617ba13b 1121 .show_options = ext4_show_options,
ac27a0ec 1122#ifdef CONFIG_QUOTA
617ba13b
MC		 1123 .quota_read = ext4_quota_read,
1124 .quota_write = ext4_quota_write,
96c7e0d9 1125 .get_dquots = ext4_get_dquots,
ac27a0ec 1126#endif
c39a7f84 1127 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1128};
1129
39655164 1130static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1131 .fh_to_dentry = ext4_fh_to_dentry,
1132 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1133 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1134};
1135
1136enum {
1137 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1138 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1139 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1140 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1141 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1142 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1143 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1144 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1145 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1146 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1147 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1148 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
923ae0ff 1149 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1449032b 1150 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1151 Opt_lazytime, Opt_nolazytime,
1449032b 1152 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1153 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1154 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1155 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1156 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1157};
1158
a447c093 1159static const match_table_t tokens = {
ac27a0ec
DK		 1160 {Opt_bsd_df, "bsddf"},
1161 {Opt_minix_df, "minixdf"},
1162 {Opt_grpid, "grpid"},
1163 {Opt_grpid, "bsdgroups"},
1164 {Opt_nogrpid, "nogrpid"},
1165 {Opt_nogrpid, "sysvgroups"},
1166 {Opt_resgid, "resgid=%u"},
1167 {Opt_resuid, "resuid=%u"},
1168 {Opt_sb, "sb=%u"},
1169 {Opt_err_cont, "errors=continue"},
1170 {Opt_err_panic, "errors=panic"},
1171 {Opt_err_ro, "errors=remount-ro"},
1172 {Opt_nouid32, "nouid32"},
ac27a0ec 1173 {Opt_debug, "debug"},
72578c33
TT		 1174 {Opt_removed, "oldalloc"},
1175 {Opt_removed, "orlov"},
ac27a0ec
DK		 1176 {Opt_user_xattr, "user_xattr"},
1177 {Opt_nouser_xattr, "nouser_xattr"},
1178 {Opt_acl, "acl"},
1179 {Opt_noacl, "noacl"},
e3bb52ae 1180 {Opt_noload, "norecovery"},
5a916be1 1181 {Opt_noload, "noload"},
72578c33
TT		 1182 {Opt_removed, "nobh"},
1183 {Opt_removed, "bh"},
ac27a0ec 1184 {Opt_commit, "commit=%u"},
30773840
TT		 1185 {Opt_min_batch_time, "min_batch_time=%u"},
1186 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1187 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1188 {Opt_journal_path, "journal_path=%s"},
818d276c 1189 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1190 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1191 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1192 {Opt_abort, "abort"},
1193 {Opt_data_journal, "data=journal"},
1194 {Opt_data_ordered, "data=ordered"},
1195 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1196 {Opt_data_err_abort, "data_err=abort"},
1197 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1198 {Opt_offusrjquota, "usrjquota="},
1199 {Opt_usrjquota, "usrjquota=%s"},
1200 {Opt_offgrpjquota, "grpjquota="},
1201 {Opt_grpjquota, "grpjquota=%s"},
1202 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1203 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1204 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1205 {Opt_grpquota, "grpquota"},
1206 {Opt_noquota, "noquota"},
1207 {Opt_quota, "quota"},
1208 {Opt_usrquota, "usrquota"},
1209 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1210 {Opt_barrier, "barrier"},
1211 {Opt_nobarrier, "nobarrier"},
25ec56b5 1212 {Opt_i_version, "i_version"},
923ae0ff 1213 {Opt_dax, "dax"},
c9de560d 1214 {Opt_stripe, "stripe=%u"},
64769240 1215 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1216 {Opt_lazytime, "lazytime"},
1217 {Opt_nolazytime, "nolazytime"},
dd919b98 1218 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1219 {Opt_removed, "mblk_io_submit"},
1220 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1221 {Opt_block_validity, "block_validity"},
1222 {Opt_noblock_validity, "noblock_validity"},
240799cd 1223 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1224 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1225 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1226 {Opt_auto_da_alloc, "auto_da_alloc"},
1227 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1228 {Opt_dioread_nolock, "dioread_nolock"},
1229 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1230 {Opt_discard, "discard"},
1231 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1232 {Opt_init_itable, "init_itable=%u"},
1233 {Opt_init_itable, "init_itable"},
1234 {Opt_noinit_itable, "noinit_itable"},
df981d03 1235 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1236 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1237 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1238 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1239 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1240 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1241 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1242 {Opt_err, NULL},
ac27a0ec
DK		 1243};
1244
617ba13b 1245static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1246{
617ba13b 1247 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1248 char *options = (char *) *data;
1249
1250 if (!options || strncmp(options, "sb=", 3) != 0)
1251 return 1; /* Default location */
0b8e58a1 1252
ac27a0ec 1253 options += 3;
0b8e58a1 1254 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1255 sb_block = simple_strtoul(options, &options, 0);
1256 if (*options && *options != ',') {
4776004f 1257 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1258 (char *) *data);
1259 return 1;
1260 }
1261 if (*options == ',')
1262 options++;
1263 *data = (void *) options;
0b8e58a1 1264
ac27a0ec
DK		 1265 return sb_block;
1266}
1267
b3881f74 1268#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1269static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1270 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1271
56c50f11
DM		 1272#ifdef CONFIG_QUOTA
1273static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1274{
1275 struct ext4_sb_info *sbi = EXT4_SB(sb);
1276 char *qname;
03dafb5f 1277 int ret = -1;
56c50f11
DM		 1278
1279 if (sb_any_quota_loaded(sb) &&
1280 !sbi->s_qf_names[qtype]) {
1281 ext4_msg(sb, KERN_ERR,
1282 "Cannot change journaled "
1283 "quota options when quota turned on");
57f73c2c 1284 return -1;
56c50f11 1285 }
e2b911c5 1286 if (ext4_has_feature_quota(sb)) {
262b4662
JK		 1287 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1288 "when QUOTA feature is enabled");
1289 return -1;
1290 }
56c50f11
DM		 1291 qname = match_strdup(args);
1292 if (!qname) {
1293 ext4_msg(sb, KERN_ERR,
1294 "Not enough memory for storing quotafile name");
57f73c2c 1295 return -1;
56c50f11 1296 }
03dafb5f
CG		 1297 if (sbi->s_qf_names[qtype]) {
1298 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1299 ret = 1;
1300 else
1301 ext4_msg(sb, KERN_ERR,
1302 "%s quota file already specified",
1303 QTYPE2NAME(qtype));
1304 goto errout;
56c50f11 1305 }
03dafb5f 1306 if (strchr(qname, '/')) {
56c50f11
DM		 1307 ext4_msg(sb, KERN_ERR,
1308 "quotafile must be on filesystem root");
03dafb5f 1309 goto errout;
56c50f11 1310 }
03dafb5f 1311 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1312 set_opt(sb, QUOTA);
56c50f11 1313 return 1;
03dafb5f
CG		 1314errout:
1315 kfree(qname);
1316 return ret;
56c50f11
DM		 1317}
1318
1319static int clear_qf_name(struct super_block *sb, int qtype)
1320{
1321
1322 struct ext4_sb_info *sbi = EXT4_SB(sb);
1323
1324 if (sb_any_quota_loaded(sb) &&
1325 sbi->s_qf_names[qtype]) {
1326 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1327 " when quota turned on");
57f73c2c 1328 return -1;
56c50f11 1329 }
03dafb5f 1330 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1331 sbi->s_qf_names[qtype] = NULL;
1332 return 1;
1333}
1334#endif
1335
26092bf5
TT		 1336#define MOPT_SET 0x0001
1337#define MOPT_CLEAR 0x0002
1338#define MOPT_NOSUPPORT 0x0004
1339#define MOPT_EXPLICIT 0x0008
1340#define MOPT_CLEAR_ERR 0x0010
1341#define MOPT_GTE0 0x0020
ac27a0ec 1342#ifdef CONFIG_QUOTA
26092bf5
TT		 1343#define MOPT_Q 0
1344#define MOPT_QFMT 0x0040
1345#else
1346#define MOPT_Q MOPT_NOSUPPORT
1347#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1348#endif
26092bf5 1349#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1350#define MOPT_NO_EXT2 0x0100
1351#define MOPT_NO_EXT3 0x0200
1352#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1353#define MOPT_STRING 0x0400
26092bf5
TT		 1354
1355static const struct mount_opts {
1356 int token;
1357 int mount_opt;
1358 int flags;
1359} ext4_mount_opts[] = {
1360 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1361 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1362 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1363 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1364 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1365 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1366 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1367 MOPT_EXT4_ONLY | MOPT_SET},
1368 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1369 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1370 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1371 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1372 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1373 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1374 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1375 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1376 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1377 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c
TT		 1378 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1379 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1380 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1381 EXT4_MOUNT_JOURNAL_CHECKSUM),
1382 MOPT_EXT4_ONLY | MOPT_SET},
1383 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1384 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1385 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1386 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1387 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1388 MOPT_NO_EXT2 | MOPT_SET},
1389 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1390 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1391 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1392 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1393 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1394 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1395 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1396 {Opt_commit, 0, MOPT_GTE0},
1397 {Opt_max_batch_time, 0, MOPT_GTE0},
1398 {Opt_min_batch_time, 0, MOPT_GTE0},
1399 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1400 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1401 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1402 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1403 {Opt_resuid, 0, MOPT_GTE0},
1404 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1405 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1406 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1407 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1408 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1409 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1410 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1411 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1412 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1413 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1414#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1415 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1416 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1417#else
26092bf5
TT		 1418 {Opt_acl, 0, MOPT_NOSUPPORT},
1419 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1420#endif
26092bf5
TT		 1421 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1422 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1423 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1424 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1425 MOPT_SET | MOPT_Q},
1426 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1427 MOPT_SET | MOPT_Q},
1428 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1429 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1430 {Opt_usrjquota, 0, MOPT_Q},
1431 {Opt_grpjquota, 0, MOPT_Q},
1432 {Opt_offusrjquota, 0, MOPT_Q},
1433 {Opt_offgrpjquota, 0, MOPT_Q},
1434 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1435 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1436 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1437 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1438 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1439 {Opt_err, 0, 0}
1440};
1441
1442static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1443 substring_t *args, unsigned long *journal_devnum,
1444 unsigned int *journal_ioprio, int is_remount)
1445{
1446 struct ext4_sb_info *sbi = EXT4_SB(sb);
1447 const struct mount_opts *m;
08cefc7a
EB		 1448 kuid_t uid;
1449 kgid_t gid;
26092bf5
TT		 1450 int arg = 0;
1451
57f73c2c
TT		 1452#ifdef CONFIG_QUOTA
1453 if (token == Opt_usrjquota)
1454 return set_qf_name(sb, USRQUOTA, &args[0]);
1455 else if (token == Opt_grpjquota)
1456 return set_qf_name(sb, GRPQUOTA, &args[0]);
1457 else if (token == Opt_offusrjquota)
1458 return clear_qf_name(sb, USRQUOTA);
1459 else if (token == Opt_offgrpjquota)
1460 return clear_qf_name(sb, GRPQUOTA);
1461#endif
26092bf5 1462 switch (token) {
f7048605
TT		 1463 case Opt_noacl:
1464 case Opt_nouser_xattr:
1465 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1466 break;
26092bf5
TT		 1467 case Opt_sb:
1468 return 1; /* handled by get_sb_block() */
1469 case Opt_removed:
5f3633e3 1470 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1471 return 1;
26092bf5
TT		 1472 case Opt_abort:
1473 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1474 return 1;
1475 case Opt_i_version:
1476 sb->s_flags |= MS_I_VERSION;
1477 return 1;
a26f4992
TT		 1478 case Opt_lazytime:
1479 sb->s_flags |= MS_LAZYTIME;
1480 return 1;
1481 case Opt_nolazytime:
1482 sb->s_flags &= ~MS_LAZYTIME;
1483 return 1;
26092bf5
TT		 1484 }
1485
5f3633e3
JK		 1486 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1487 if (token == m->token)
1488 break;
1489
1490 if (m->token == Opt_err) {
1491 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1492 "or missing value", opt);
1493 return -1;
1494 }
1495
8dc0aa8c
TT		 1496 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1497 ext4_msg(sb, KERN_ERR,
1498 "Mount option \"%s\" incompatible with ext2", opt);
1499 return -1;
1500 }
1501 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1502 ext4_msg(sb, KERN_ERR,
1503 "Mount option \"%s\" incompatible with ext3", opt);
1504 return -1;
1505 }
1506
ad4eec61 1507 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1508 return -1;
1509 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1510 return -1;
1511 if (m->flags & MOPT_EXPLICIT)
1512 set_opt2(sb, EXPLICIT_DELALLOC);
1513 if (m->flags & MOPT_CLEAR_ERR)
1514 clear_opt(sb, ERRORS_MASK);
1515 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1516 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1517 "options when quota turned on");
1518 return -1;
1519 }
1520
1521 if (m->flags & MOPT_NOSUPPORT) {
1522 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1523 } else if (token == Opt_commit) {
1524 if (arg == 0)
1525 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1526 sbi->s_commit_interval = HZ * arg;
1527 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1528 sbi->s_max_batch_time = arg;
1529 } else if (token == Opt_min_batch_time) {
1530 sbi->s_min_batch_time = arg;
1531 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1532 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1533 ext4_msg(sb, KERN_ERR,
1534 "EXT4-fs: inode_readahead_blks must be "
1535 "0 or a power of 2 smaller than 2^31");
26092bf5 1536 return -1;
5f3633e3
JK		 1537 }
1538 sbi->s_inode_readahead_blks = arg;
1539 } else if (token == Opt_init_itable) {
1540 set_opt(sb, INIT_INODE_TABLE);
1541 if (!args->from)
1542 arg = EXT4_DEF_LI_WAIT_MULT;
1543 sbi->s_li_wait_mult = arg;
1544 } else if (token == Opt_max_dir_size_kb) {
1545 sbi->s_max_dir_size_kb = arg;
1546 } else if (token == Opt_stripe) {
1547 sbi->s_stripe = arg;
1548 } else if (token == Opt_resuid) {
1549 uid = make_kuid(current_user_ns(), arg);
1550 if (!uid_valid(uid)) {
1551 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1552 return -1;
1553 }
5f3633e3
JK		 1554 sbi->s_resuid = uid;
1555 } else if (token == Opt_resgid) {
1556 gid = make_kgid(current_user_ns(), arg);
1557 if (!gid_valid(gid)) {
1558 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1559 return -1;
1560 }
1561 sbi->s_resgid = gid;
1562 } else if (token == Opt_journal_dev) {
1563 if (is_remount) {
1564 ext4_msg(sb, KERN_ERR,
1565 "Cannot specify journal on remount");
1566 return -1;
1567 }
1568 *journal_devnum = arg;
ad4eec61
ES		 1569 } else if (token == Opt_journal_path) {
1570 char *journal_path;
1571 struct inode *journal_inode;
1572 struct path path;
1573 int error;
1574
1575 if (is_remount) {
1576 ext4_msg(sb, KERN_ERR,
1577 "Cannot specify journal on remount");
1578 return -1;
1579 }
1580 journal_path = match_strdup(&args[0]);
1581 if (!journal_path) {
1582 ext4_msg(sb, KERN_ERR, "error: could not dup "
1583 "journal device string");
1584 return -1;
1585 }
1586
1587 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1588 if (error) {
1589 ext4_msg(sb, KERN_ERR, "error: could not find "
1590 "journal device path: error %d", error);
1591 kfree(journal_path);
1592 return -1;
1593 }
1594
2b0143b5 1595 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1596 if (!S_ISBLK(journal_inode->i_mode)) {
1597 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1598 "is not a block device", journal_path);
1599 path_put(&path);
1600 kfree(journal_path);
1601 return -1;
1602 }
1603
1604 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1605 path_put(&path);
1606 kfree(journal_path);
5f3633e3
JK		 1607 } else if (token == Opt_journal_ioprio) {
1608 if (arg > 7) {
1609 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1610 " (must be 0-7)");
1611 return -1;
1612 }
1613 *journal_ioprio =
1614 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1615 } else if (token == Opt_test_dummy_encryption) {
1616#ifdef CONFIG_EXT4_FS_ENCRYPTION
1617 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1618 ext4_msg(sb, KERN_WARNING,
1619 "Test dummy encryption mode enabled");
1620#else
1621 ext4_msg(sb, KERN_WARNING,
1622 "Test dummy encryption mount option ignored");
1623#endif
5f3633e3
JK		 1624 } else if (m->flags & MOPT_DATAJ) {
1625 if (is_remount) {
1626 if (!sbi->s_journal)
1627 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1628 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1629 ext4_msg(sb, KERN_ERR,
26092bf5 1630 "Cannot change data mode on remount");
26092bf5 1631 return -1;
ac27a0ec 1632 }
26092bf5 1633 } else {
5f3633e3
JK		 1634 clear_opt(sb, DATA_FLAGS);
1635 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1636 }
5f3633e3
JK		 1637#ifdef CONFIG_QUOTA
1638 } else if (m->flags & MOPT_QFMT) {
1639 if (sb_any_quota_loaded(sb) &&
1640 sbi->s_jquota_fmt != m->mount_opt) {
1641 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1642 "quota options when quota turned on");
1643 return -1;
1644 }
e2b911c5 1645 if (ext4_has_feature_quota(sb)) {
262b4662
JK		 1646 ext4_msg(sb, KERN_ERR,
1647 "Cannot set journaled quota options "
1648 "when QUOTA feature is enabled");
1649 return -1;
1650 }
5f3633e3 1651 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff
RZ		 1652#endif
1653#ifndef CONFIG_FS_DAX
1654 } else if (token == Opt_dax) {
1655 ext4_msg(sb, KERN_INFO, "dax option not supported");
1656 return -1;
5f3633e3
JK		 1657#endif
1658 } else {
1659 if (!args->from)
1660 arg = 1;
1661 if (m->flags & MOPT_CLEAR)
1662 arg = !arg;
1663 else if (unlikely(!(m->flags & MOPT_SET))) {
1664 ext4_msg(sb, KERN_WARNING,
1665 "buggy handling of option %s", opt);
1666 WARN_ON(1);
1667 return -1;
1668 }
1669 if (arg != 0)
1670 sbi->s_mount_opt |= m->mount_opt;
1671 else
1672 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1673 }
5f3633e3 1674 return 1;
26092bf5
TT		 1675}
1676
1677static int parse_options(char *options, struct super_block *sb,
1678 unsigned long *journal_devnum,
1679 unsigned int *journal_ioprio,
1680 int is_remount)
1681{
1682 struct ext4_sb_info *sbi = EXT4_SB(sb);
1683 char *p;
1684 substring_t args[MAX_OPT_ARGS];
1685 int token;
1686
1687 if (!options)
1688 return 1;
1689
1690 while ((p = strsep(&options, ",")) != NULL) {
1691 if (!*p)
1692 continue;
1693 /*
1694 * Initialize args struct so we know whether arg was
1695 * found; some options take optional arguments.
1696 */
caecd0af 1697 args[0].to = args[0].from = NULL;
26092bf5
TT		 1698 token = match_token(p, tokens, args);
1699 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1700 journal_ioprio, is_remount) < 0)
1701 return 0;
ac27a0ec
DK		 1702 }
1703#ifdef CONFIG_QUOTA
e2b911c5 1704 if (ext4_has_feature_quota(sb) &&
262b4662
JK		 1705 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1706 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1707 "feature is enabled");
1708 return 0;
1709 }
ac27a0ec 1710 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1711 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1712 clear_opt(sb, USRQUOTA);
ac27a0ec 1713
482a7425 1714 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1715 clear_opt(sb, GRPQUOTA);
ac27a0ec 1716
56c50f11 1717 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1718 ext4_msg(sb, KERN_ERR, "old and new quota "
1719 "format mixing");
ac27a0ec
DK		 1720 return 0;
1721 }
1722
1723 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1724 ext4_msg(sb, KERN_ERR, "journaled quota format "
1725 "not specified");
ac27a0ec
DK		 1726 return 0;
1727 }
ac27a0ec
DK		 1728 }
1729#endif
261cb20c
JK		 1730 if (test_opt(sb, DIOREAD_NOLOCK)) {
1731 int blocksize =
1732 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1733
1734 if (blocksize < PAGE_CACHE_SIZE) {
1735 ext4_msg(sb, KERN_ERR, "can't mount with "
1736 "dioread_nolock if block size != PAGE_SIZE");
1737 return 0;
1738 }
1739 }
d4f76107
JK		 1740 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1741 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1742 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1743 "in data=ordered mode");
1744 return 0;
1745 }
ac27a0ec
DK		 1746 return 1;
1747}
1748
2adf6da8
TT		 1749static inline void ext4_show_quota_options(struct seq_file *seq,
1750 struct super_block *sb)
1751{
1752#if defined(CONFIG_QUOTA)
1753 struct ext4_sb_info *sbi = EXT4_SB(sb);
1754
1755 if (sbi->s_jquota_fmt) {
1756 char *fmtname = "";
1757
1758 switch (sbi->s_jquota_fmt) {
1759 case QFMT_VFS_OLD:
1760 fmtname = "vfsold";
1761 break;
1762 case QFMT_VFS_V0:
1763 fmtname = "vfsv0";
1764 break;
1765 case QFMT_VFS_V1:
1766 fmtname = "vfsv1";
1767 break;
1768 }
1769 seq_printf(seq, ",jqfmt=%s", fmtname);
1770 }
1771
1772 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1773 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1774
1775 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1776 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1777#endif
1778}
1779
5a916be1
TT		 1780static const char *token2str(int token)
1781{
50df9fd5 1782 const struct match_token *t;
5a916be1
TT		 1783
1784 for (t = tokens; t->token != Opt_err; t++)
1785 if (t->token == token && !strchr(t->pattern, '='))
1786 break;
1787 return t->pattern;
1788}
1789
2adf6da8
TT		 1790/*
1791 * Show an option if
1792 * - it's set to a non-default value OR
1793 * - if the per-sb default is different from the global default
1794 */
66acdcf4
TT		 1795static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1796 int nodefs)
2adf6da8 1797{
2adf6da8
TT		 1798 struct ext4_sb_info *sbi = EXT4_SB(sb);
1799 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1800 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1801 const struct mount_opts *m;
66acdcf4 1802 char sep = nodefs ? '\n' : ',';
2adf6da8 1803
66acdcf4
TT		 1804#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1805#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1806
1807 if (sbi->s_sb_block != 1)
5a916be1
TT		 1808 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1809
1810 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1811 int want_set = m->flags & MOPT_SET;
1812 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1813 (m->flags & MOPT_CLEAR_ERR))
1814 continue;
66acdcf4 1815 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1816 continue; /* skip if same as the default */
1817 if ((want_set &&
1818 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1819 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1820 continue; /* select Opt_noFoo vs Opt_Foo */
1821 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1822 }
5a916be1 1823
08cefc7a 1824 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1825 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1826 SEQ_OPTS_PRINT("resuid=%u",
1827 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1828 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1829 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1830 SEQ_OPTS_PRINT("resgid=%u",
1831 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1832 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1833 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1834 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1835 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1836 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1837 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1838 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1839 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1840 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1841 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1842 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1843 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1844 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1845 if (sb->s_flags & MS_I_VERSION)
5a916be1 1846 SEQ_OPTS_PUTS("i_version");
66acdcf4 1847 if (nodefs || sbi->s_stripe)
5a916be1 1848 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1849 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1850 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1851 SEQ_OPTS_PUTS("data=journal");
1852 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1853 SEQ_OPTS_PUTS("data=ordered");
1854 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1855 SEQ_OPTS_PUTS("data=writeback");
1856 }
66acdcf4
TT		 1857 if (nodefs ||
1858 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1859 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1860 sbi->s_inode_readahead_blks);
2adf6da8 1861
66acdcf4
TT		 1862 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1863 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1864 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1865 if (nodefs || sbi->s_max_dir_size_kb)
1866 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1867
1868 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1869 return 0;
1870}
1871
66acdcf4
TT		 1872static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1873{
1874 return _ext4_show_options(seq, root->d_sb, 0);
1875}
1876
ebd173be 1877int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 1878{
1879 struct super_block *sb = seq->private;
1880 int rc;
1881
1882 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1883 rc = _ext4_show_options(seq, sb, 1);
1884 seq_puts(seq, "\n");
1885 return rc;
1886}
1887
617ba13b 1888static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1889 int read_only)
1890{
617ba13b 1891 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1892 int res = 0;
1893
617ba13b 1894 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1895 ext4_msg(sb, KERN_ERR, "revision level too high, "
1896 "forcing read-only mode");
ac27a0ec
DK		 1897 res = MS_RDONLY;
1898 }
1899 if (read_only)
281b5995 1900 goto done;
617ba13b 1901 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1902 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1903 "running e2fsck is recommended");
c8b459f4 1904 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1905 ext4_msg(sb, KERN_WARNING,
1906 "warning: mounting fs with errors, "
1907 "running e2fsck is recommended");
ed3ce80a 1908 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1909 le16_to_cpu(es->s_mnt_count) >=
1910 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1911 ext4_msg(sb, KERN_WARNING,
1912 "warning: maximal mount count reached, "
1913 "running e2fsck is recommended");
ac27a0ec
DK		 1914 else if (le32_to_cpu(es->s_checkinterval) &&
1915 (le32_to_cpu(es->s_lastcheck) +
1916 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1917 ext4_msg(sb, KERN_WARNING,
1918 "warning: checktime reached, "
1919 "running e2fsck is recommended");
0b8e58a1 1920 if (!sbi->s_journal)
0390131b 1921 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1922 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1923 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1924 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1925 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1926 ext4_update_dynamic_rev(sb);
0390131b 1927 if (sbi->s_journal)
e2b911c5 1928 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 1929
e2d67052 1930 ext4_commit_super(sb, 1);
281b5995 1931done:
ac27a0ec 1932 if (test_opt(sb, DEBUG))
a9df9a49 1933 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1934 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1935 sb->s_blocksize,
1936 sbi->s_groups_count,
617ba13b
MC		 1937 EXT4_BLOCKS_PER_GROUP(sb),
1938 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1939 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1940
7abc52c2 1941 cleancache_init_fs(sb);
ac27a0ec
DK		 1942 return res;
1943}
1944
117fff10
TT		 1945int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1946{
1947 struct ext4_sb_info *sbi = EXT4_SB(sb);
1948 struct flex_groups *new_groups;
1949 int size;
1950
1951 if (!sbi->s_log_groups_per_flex)
1952 return 0;
1953
1954 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1955 if (size <= sbi->s_flex_groups_allocated)
1956 return 0;
1957
1958 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1959 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1960 if (!new_groups) {
1961 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1962 size / (int) sizeof(struct flex_groups));
1963 return -ENOMEM;
1964 }
1965
1966 if (sbi->s_flex_groups) {
1967 memcpy(new_groups, sbi->s_flex_groups,
1968 (sbi->s_flex_groups_allocated *
1969 sizeof(struct flex_groups)));
b93b41d4 1970 kvfree(sbi->s_flex_groups);
117fff10
TT		 1971 }
1972 sbi->s_flex_groups = new_groups;
1973 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1974 return 0;
1975}
1976
772cb7c8
JS		 1977static int ext4_fill_flex_info(struct super_block *sb)
1978{
1979 struct ext4_sb_info *sbi = EXT4_SB(sb);
1980 struct ext4_group_desc *gdp = NULL;
772cb7c8 1981 ext4_group_t flex_group;
117fff10 1982 int i, err;
772cb7c8 1983
503358ae 1984 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1985 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1986 sbi->s_log_groups_per_flex = 0;
1987 return 1;
1988 }
1989
117fff10
TT		 1990 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1991 if (err)
9933fc0a 1992 goto failed;
772cb7c8 1993
772cb7c8 1994 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1995 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1996
1997 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1998 atomic_add(ext4_free_inodes_count(sb, gdp),
1999 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2000 atomic64_add(ext4_free_group_clusters(sb, gdp),
2001 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2002 atomic_add(ext4_used_dirs_count(sb, gdp),
2003 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2004 }
2005
2006 return 1;
2007failed:
2008 return 0;
2009}
2010
e2b911c5 2011static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2012 struct ext4_group_desc *gdp)
717d50e4 2013{
feb0ab32 2014 int offset;
717d50e4 2015 __u16 crc = 0;
feb0ab32 2016 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2017 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2018
9aa5d32b 2019 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2020 /* Use new metadata_csum algorithm */
d6a77105 2021 __le16 save_csum;
feb0ab32
DW		 2022 __u32 csum32;
2023
d6a77105 2024 save_csum = gdp->bg_checksum;
feb0ab32
DW		 2025 gdp->bg_checksum = 0;
2026 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2027 sizeof(le_group));
2028 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2029 sbi->s_desc_size);
d6a77105 2030 gdp->bg_checksum = save_csum;
feb0ab32
DW		 2031
2032 crc = csum32 & 0xFFFF;
2033 goto out;
717d50e4
AD		 2034 }
2035
feb0ab32 2036 /* old crc16 code */
e2b911c5 2037 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2038 return 0;
2039
feb0ab32
DW		 2040 offset = offsetof(struct ext4_group_desc, bg_checksum);
2041
2042 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2043 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2044 crc = crc16(crc, (__u8 *)gdp, offset);
2045 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2046 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2047 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2048 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2049 crc = crc16(crc, (__u8 *)gdp + offset,
2050 le16_to_cpu(sbi->s_es->s_desc_size) -
2051 offset);
2052
2053out:
717d50e4
AD		 2054 return cpu_to_le16(crc);
2055}
2056
feb0ab32 2057int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2058 struct ext4_group_desc *gdp)
2059{
feb0ab32 2060 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2061 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2062 return 0;
2063
2064 return 1;
2065}
2066
feb0ab32
DW		 2067void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2068 struct ext4_group_desc *gdp)
2069{
2070 if (!ext4_has_group_desc_csum(sb))
2071 return;
e2b911c5 2072 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2073}
2074
ac27a0ec 2075/* Called at mount-time, super-block is locked */
bfff6873
LC		 2076static int ext4_check_descriptors(struct super_block *sb,
2077 ext4_group_t *first_not_zeroed)
ac27a0ec 2078{
617ba13b
MC		 2079 struct ext4_sb_info *sbi = EXT4_SB(sb);
2080 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2081 ext4_fsblk_t last_block;
bd81d8ee
LV		 2082 ext4_fsblk_t block_bitmap;
2083 ext4_fsblk_t inode_bitmap;
2084 ext4_fsblk_t inode_table;
ce421581 2085 int flexbg_flag = 0;
bfff6873 2086 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2087
e2b911c5 2088 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2089 flexbg_flag = 1;
2090
af5bc92d 2091 ext4_debug("Checking group descriptors");
ac27a0ec 2092
197cd65a
AM		 2093 for (i = 0; i < sbi->s_groups_count; i++) {
2094 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2095
ce421581 2096 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2097 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2098 else
2099 last_block = first_block +
617ba13b 2100 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2101
bfff6873
LC		 2102 if ((grp == sbi->s_groups_count) &&
2103 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2104 grp = i;
2105
8fadc143 2106 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2107 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2108 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2109 "Block bitmap for group %u not in group "
b31e1552 2110 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2111 return 0;
2112 }
8fadc143 2113 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2114 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2115 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2116 "Inode bitmap for group %u not in group "
b31e1552 2117 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2118 return 0;
2119 }
8fadc143 2120 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2121 if (inode_table < first_block ||
2b2d6d01 2122 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2123 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2124 "Inode table for group %u not in group "
b31e1552 2125 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2126 return 0;
2127 }
955ce5f5 2128 ext4_lock_group(sb, i);
feb0ab32 2129 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2130 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2131 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2132 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2133 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2134 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2135 ext4_unlock_group(sb, i);
8a266467 2136 return 0;
7ee1ec4c 2137 }
717d50e4 2138 }
955ce5f5 2139 ext4_unlock_group(sb, i);
ce421581
JS		 2140 if (!flexbg_flag)
2141 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2142 }
bfff6873
LC		 2143 if (NULL != first_not_zeroed)
2144 *first_not_zeroed = grp;
ac27a0ec
DK		 2145 return 1;
2146}
2147
617ba13b 2148/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2149 * the superblock) which were deleted from all directories, but held open by
2150 * a process at the time of a crash. We walk the list and try to delete these
2151 * inodes at recovery time (only with a read-write filesystem).
2152 *
2153 * In order to keep the orphan inode chain consistent during traversal (in
2154 * case of crash during recovery), we link each inode into the superblock
2155 * orphan list_head and handle it the same way as an inode deletion during
2156 * normal operation (which journals the operations for us).
2157 *
2158 * We only do an iget() and an iput() on each inode, which is very safe if we
2159 * accidentally point at an in-use or already deleted inode. The worst that
2160 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2161 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2162 * e2fsck was run on this filesystem, and it must have already done the orphan
2163 * inode cleanup for us, so we can safely abort without any further action.
2164 */
2b2d6d01
TT		 2165static void ext4_orphan_cleanup(struct super_block *sb,
2166 struct ext4_super_block *es)
ac27a0ec
DK		 2167{
2168 unsigned int s_flags = sb->s_flags;
2169 int nr_orphans = 0, nr_truncates = 0;
2170#ifdef CONFIG_QUOTA
2171 int i;
2172#endif
2173 if (!es->s_last_orphan) {
2174 jbd_debug(4, "no orphan inodes to clean up\n");
2175 return;
2176 }
2177
a8f48a95 2178 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2179 ext4_msg(sb, KERN_ERR, "write access "
2180 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2181 return;
2182 }
2183
d39195c3
AG		 2184 /* Check if feature set would not allow a r/w mount */
2185 if (!ext4_feature_set_ok(sb, 0)) {
2186 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2187 "unknown ROCOMPAT features");
2188 return;
2189 }
2190
617ba13b 2191 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2192 /* don't clear list on RO mount w/ errors */
2193 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2194 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2195 "clearing orphan list.\n");
c25f9bc6
ES		 2196 es->s_last_orphan = 0;
2197 }
ac27a0ec
DK		 2198 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2199 return;
2200 }
2201
2202 if (s_flags & MS_RDONLY) {
b31e1552 2203 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2204 sb->s_flags &= ~MS_RDONLY;
2205 }
2206#ifdef CONFIG_QUOTA
2207 /* Needed for iput() to work correctly and not trash data */
2208 sb->s_flags |= MS_ACTIVE;
2209 /* Turn on quotas so that they are updated correctly */
a2d4a646 2210 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2211 if (EXT4_SB(sb)->s_qf_names[i]) {
2212 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2213 if (ret < 0)
b31e1552
ES		 2214 ext4_msg(sb, KERN_ERR,
2215 "Cannot turn on journaled "
2216 "quota: error %d", ret);
ac27a0ec
DK		 2217 }
2218 }
2219#endif
2220
2221 while (es->s_last_orphan) {
2222 struct inode *inode;
2223
97bd42b9
JB		 2224 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2225 if (IS_ERR(inode)) {
ac27a0ec
DK		 2226 es->s_last_orphan = 0;
2227 break;
2228 }
2229
617ba13b 2230 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2231 dquot_initialize(inode);
ac27a0ec 2232 if (inode->i_nlink) {
566370a2
PT		 2233 if (test_opt(sb, DEBUG))
2234 ext4_msg(sb, KERN_DEBUG,
2235 "%s: truncating inode %lu to %lld bytes",
2236 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2237 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2238 inode->i_ino, inode->i_size);
721e3eba 2239 mutex_lock(&inode->i_mutex);
55f252c9 2240 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2241 ext4_truncate(inode);
721e3eba 2242 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2243 nr_truncates++;
2244 } else {
566370a2
PT		 2245 if (test_opt(sb, DEBUG))
2246 ext4_msg(sb, KERN_DEBUG,
2247 "%s: deleting unreferenced inode %lu",
2248 __func__, inode->i_ino);
ac27a0ec
DK		 2249 jbd_debug(2, "deleting unreferenced inode %lu\n",
2250 inode->i_ino);
2251 nr_orphans++;
2252 }
2253 iput(inode); /* The delete magic happens here! */
2254 }
2255
2b2d6d01 2256#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2257
2258 if (nr_orphans)
b31e1552
ES		 2259 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2260 PLURAL(nr_orphans));
ac27a0ec 2261 if (nr_truncates)
b31e1552
ES		 2262 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2263 PLURAL(nr_truncates));
ac27a0ec
DK		 2264#ifdef CONFIG_QUOTA
2265 /* Turn quotas off */
a2d4a646 2266 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2267 if (sb_dqopt(sb)->files[i])
287a8095 2268 dquot_quota_off(sb, i);
ac27a0ec
DK		 2269 }
2270#endif
2271 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2272}
0b8e58a1 2273
cd2291a4
ES		 2274/*
2275 * Maximal extent format file size.
2276 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2277 * extent format containers, within a sector_t, and within i_blocks
2278 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2279 * so that won't be a limiting factor.
2280 *
f17722f9
LC		 2281 * However there is other limiting factor. We do store extents in the form
2282 * of starting block and length, hence the resulting length of the extent
2283 * covering maximum file size must fit into on-disk format containers as
2284 * well. Given that length is always by 1 unit bigger than max unit (because
2285 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2286 *
cd2291a4
ES		 2287 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2288 */
f287a1a5 2289static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2290{
2291 loff_t res;
2292 loff_t upper_limit = MAX_LFS_FILESIZE;
2293
2294 /* small i_blocks in vfs inode? */
f287a1a5 2295 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2296 /*
90c699a9 2297 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2298 * i_block represent total blocks in 512 bytes
2299 * 32 == size of vfs inode i_blocks * 8
2300 */
2301 upper_limit = (1LL << 32) - 1;
2302
2303 /* total blocks in file system block size */
2304 upper_limit >>= (blkbits - 9);
2305 upper_limit <<= blkbits;
2306 }
2307
f17722f9
LC		 2308 /*
2309 * 32-bit extent-start container, ee_block. We lower the maxbytes
2310 * by one fs block, so ee_len can cover the extent of maximum file
2311 * size
2312 */
2313 res = (1LL << 32) - 1;
cd2291a4 2314 res <<= blkbits;
cd2291a4
ES		 2315
2316 /* Sanity check against vm- & vfs- imposed limits */
2317 if (res > upper_limit)
2318 res = upper_limit;
2319
2320 return res;
2321}
ac27a0ec 2322
ac27a0ec 2323/*
cd2291a4 2324 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2325 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2326 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2327 */
f287a1a5 2328static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2329{
617ba13b 2330 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2331 int meta_blocks;
2332 loff_t upper_limit;
0b8e58a1
AD		 2333 /* This is calculated to be the largest file size for a dense, block
2334 * mapped file such that the file's total number of 512-byte sectors,
2335 * including data and all indirect blocks, does not exceed (2^48 - 1).
2336 *
2337 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2338 * number of 512-byte sectors of the file.
0fc1b451
AK		 2339 */
2340
f287a1a5 2341 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2342 /*
90c699a9 2343 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2344 * the inode i_block field represents total file blocks in
2345 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2346 */
2347 upper_limit = (1LL << 32) - 1;
2348
2349 /* total blocks in file system block size */
2350 upper_limit >>= (bits - 9);
2351
2352 } else {
8180a562
AK		 2353 /*
2354 * We use 48 bit ext4_inode i_blocks
2355 * With EXT4_HUGE_FILE_FL set the i_blocks
2356 * represent total number of blocks in
2357 * file system block size
2358 */
0fc1b451
AK		 2359 upper_limit = (1LL << 48) - 1;
2360
0fc1b451
AK		 2361 }
2362
2363 /* indirect blocks */
2364 meta_blocks = 1;
2365 /* double indirect blocks */
2366 meta_blocks += 1 + (1LL << (bits-2));
2367 /* tripple indirect blocks */
2368 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2369
2370 upper_limit -= meta_blocks;
2371 upper_limit <<= bits;
ac27a0ec
DK		 2372
2373 res += 1LL << (bits-2);
2374 res += 1LL << (2*(bits-2));
2375 res += 1LL << (3*(bits-2));
2376 res <<= bits;
2377 if (res > upper_limit)
2378 res = upper_limit;
0fc1b451
AK		 2379
2380 if (res > MAX_LFS_FILESIZE)
2381 res = MAX_LFS_FILESIZE;
2382
ac27a0ec
DK		 2383 return res;
2384}
2385
617ba13b 2386static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2387 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2388{
617ba13b 2389 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2390 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2391 int has_super = 0;
2392
2393 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2394
e2b911c5 2395 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2396 return logical_sb_block + nr + 1;
ac27a0ec 2397 bg = sbi->s_desc_per_block * nr;
617ba13b 2398 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2399 has_super = 1;
0b8e58a1 2400
bd63f6b0
DW		 2401 /*
2402 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2403 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2404 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2405 * compensate.
2406 */
2407 if (sb->s_blocksize == 1024 && nr == 0 &&
2408 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2409 has_super++;
2410
617ba13b 2411 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2412}
2413
c9de560d
AT		 2414/**
2415 * ext4_get_stripe_size: Get the stripe size.
2416 * @sbi: In memory super block info
2417 *
2418 * If we have specified it via mount option, then
2419 * use the mount option value. If the value specified at mount time is
2420 * greater than the blocks per group use the super block value.
2421 * If the super block value is greater than blocks per group return 0.
2422 * Allocator needs it be less than blocks per group.
2423 *
2424 */
2425static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2426{
2427 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2428 unsigned long stripe_width =
2429 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2430 int ret;
c9de560d
AT		 2431
2432 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2433 ret = sbi->s_stripe;
2434 else if (stripe_width <= sbi->s_blocks_per_group)
2435 ret = stripe_width;
2436 else if (stride <= sbi->s_blocks_per_group)
2437 ret = stride;
2438 else
2439 ret = 0;
c9de560d 2440
3eb08658
DE		 2441 /*
2442 * If the stripe width is 1, this makes no sense and
2443 * we set it to 0 to turn off stripe handling code.
2444 */
2445 if (ret <= 1)
2446 ret = 0;
c9de560d 2447
3eb08658 2448 return ret;
c9de560d 2449}
ac27a0ec 2450
a13fb1a4
ES		 2451/*
2452 * Check whether this filesystem can be mounted based on
2453 * the features present and the RDONLY/RDWR mount requested.
2454 * Returns 1 if this filesystem can be mounted as requested,
2455 * 0 if it cannot be.
2456 */
2457static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2458{
e2b911c5 2459 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2460 ext4_msg(sb, KERN_ERR,
2461 "Couldn't mount because of "
2462 "unsupported optional features (%x)",
2463 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2464 ~EXT4_FEATURE_INCOMPAT_SUPP));
2465 return 0;
2466 }
2467
2468 if (readonly)
2469 return 1;
2470
e2b911c5 2471 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b
DW		 2472 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2473 sb->s_flags |= MS_RDONLY;
2474 return 1;
2475 }
2476
a13fb1a4 2477 /* Check that feature set is OK for a read-write mount */
e2b911c5 2478 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2479 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2480 "unsupported optional features (%x)",
2481 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2482 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2483 return 0;
2484 }
2485 /*
2486 * Large file size enabled file system can only be mounted
2487 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2488 */
e2b911c5 2489 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2490 if (sizeof(blkcnt_t) < sizeof(u64)) {
2491 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2492 "cannot be mounted RDWR without "
2493 "CONFIG_LBDAF");
2494 return 0;
2495 }
2496 }
e2b911c5 2497 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2498 ext4_msg(sb, KERN_ERR,
2499 "Can't support bigalloc feature without "
2500 "extents feature\n");
2501 return 0;
2502 }
7c319d32
AK		 2503
2504#ifndef CONFIG_QUOTA
e2b911c5 2505 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2506 ext4_msg(sb, KERN_ERR,
2507 "Filesystem with quota feature cannot be mounted RDWR "
2508 "without CONFIG_QUOTA");
2509 return 0;
2510 }
2511#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2512 return 1;
2513}
2514
66e61a9e
TT		 2515/*
2516 * This function is called once a day if we have errors logged
2517 * on the file system
2518 */
2519static void print_daily_error_info(unsigned long arg)
2520{
2521 struct super_block *sb = (struct super_block *) arg;
2522 struct ext4_sb_info *sbi;
2523 struct ext4_super_block *es;
2524
2525 sbi = EXT4_SB(sb);
2526 es = sbi->s_es;
2527
2528 if (es->s_error_count)
ae0f78de
TT		 2529 /* fsck newer than v1.41.13 is needed to clean this condition. */
2530 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2531 le32_to_cpu(es->s_error_count));
2532 if (es->s_first_error_time) {
ae0f78de 2533 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2534 sb->s_id, le32_to_cpu(es->s_first_error_time),
2535 (int) sizeof(es->s_first_error_func),
2536 es->s_first_error_func,
2537 le32_to_cpu(es->s_first_error_line));
2538 if (es->s_first_error_ino)
2539 printk(": inode %u",
2540 le32_to_cpu(es->s_first_error_ino));
2541 if (es->s_first_error_block)
2542 printk(": block %llu", (unsigned long long)
2543 le64_to_cpu(es->s_first_error_block));
2544 printk("\n");
2545 }
2546 if (es->s_last_error_time) {
ae0f78de 2547 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2548 sb->s_id, le32_to_cpu(es->s_last_error_time),
2549 (int) sizeof(es->s_last_error_func),
2550 es->s_last_error_func,
2551 le32_to_cpu(es->s_last_error_line));
2552 if (es->s_last_error_ino)
2553 printk(": inode %u",
2554 le32_to_cpu(es->s_last_error_ino));
2555 if (es->s_last_error_block)
2556 printk(": block %llu", (unsigned long long)
2557 le64_to_cpu(es->s_last_error_block));
2558 printk("\n");
2559 }
2560 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2561}
2562
bfff6873
LC		 2563/* Find next suitable group and run ext4_init_inode_table */
2564static int ext4_run_li_request(struct ext4_li_request *elr)
2565{
2566 struct ext4_group_desc *gdp = NULL;
2567 ext4_group_t group, ngroups;
2568 struct super_block *sb;
2569 unsigned long timeout = 0;
2570 int ret = 0;
2571
2572 sb = elr->lr_super;
2573 ngroups = EXT4_SB(sb)->s_groups_count;
2574
8e8ad8a5 2575 sb_start_write(sb);
bfff6873
LC		 2576 for (group = elr->lr_next_group; group < ngroups; group++) {
2577 gdp = ext4_get_group_desc(sb, group, NULL);
2578 if (!gdp) {
2579 ret = 1;
2580 break;
2581 }
2582
2583 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2584 break;
2585 }
2586
7f511862 2587 if (group >= ngroups)
bfff6873
LC		 2588 ret = 1;
2589
2590 if (!ret) {
2591 timeout = jiffies;
2592 ret = ext4_init_inode_table(sb, group,
2593 elr->lr_timeout ? 0 : 1);
2594 if (elr->lr_timeout == 0) {
51ce6511
LC		 2595 timeout = (jiffies - timeout) *
2596 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2597 elr->lr_timeout = timeout;
2598 }
2599 elr->lr_next_sched = jiffies + elr->lr_timeout;
2600 elr->lr_next_group = group + 1;
2601 }
8e8ad8a5 2602 sb_end_write(sb);
bfff6873
LC		 2603
2604 return ret;
2605}
2606
2607/*
2608 * Remove lr_request from the list_request and free the
4ed5c033 2609 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2610 */
2611static void ext4_remove_li_request(struct ext4_li_request *elr)
2612{
2613 struct ext4_sb_info *sbi;
2614
2615 if (!elr)
2616 return;
2617
2618 sbi = elr->lr_sbi;
2619
2620 list_del(&elr->lr_request);
2621 sbi->s_li_request = NULL;
2622 kfree(elr);
2623}
2624
2625static void ext4_unregister_li_request(struct super_block *sb)
2626{
1bb933fb
LC		 2627 mutex_lock(&ext4_li_mtx);
2628 if (!ext4_li_info) {
2629 mutex_unlock(&ext4_li_mtx);
bfff6873 2630 return;
1bb933fb 2631 }
bfff6873
LC		 2632
2633 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2634 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2635 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2636 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2637}
2638
8f1f7453
ES		 2639static struct task_struct *ext4_lazyinit_task;
2640
bfff6873
LC		 2641/*
2642 * This is the function where ext4lazyinit thread lives. It walks
2643 * through the request list searching for next scheduled filesystem.
2644 * When such a fs is found, run the lazy initialization request
2645 * (ext4_rn_li_request) and keep track of the time spend in this
2646 * function. Based on that time we compute next schedule time of
2647 * the request. When walking through the list is complete, compute
2648 * next waking time and put itself into sleep.
2649 */
2650static int ext4_lazyinit_thread(void *arg)
2651{
2652 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2653 struct list_head *pos, *n;
2654 struct ext4_li_request *elr;
4ed5c033 2655 unsigned long next_wakeup, cur;
bfff6873
LC		 2656
2657 BUG_ON(NULL == eli);
2658
bfff6873
LC		 2659cont_thread:
2660 while (true) {
2661 next_wakeup = MAX_JIFFY_OFFSET;
2662
2663 mutex_lock(&eli->li_list_mtx);
2664 if (list_empty(&eli->li_request_list)) {
2665 mutex_unlock(&eli->li_list_mtx);
2666 goto exit_thread;
2667 }
2668
2669 list_for_each_safe(pos, n, &eli->li_request_list) {
2670 elr = list_entry(pos, struct ext4_li_request,
2671 lr_request);
2672
b2c78cd0
TT		 2673 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2674 if (ext4_run_li_request(elr) != 0) {
2675 /* error, remove the lazy_init job */
2676 ext4_remove_li_request(elr);
2677 continue;
2678 }
bfff6873
LC		 2679 }
2680
2681 if (time_before(elr->lr_next_sched, next_wakeup))
2682 next_wakeup = elr->lr_next_sched;
2683 }
2684 mutex_unlock(&eli->li_list_mtx);
2685
a0acae0e 2686 try_to_freeze();
bfff6873 2687
4ed5c033
LC		 2688 cur = jiffies;
2689 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2690 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2691 cond_resched();
2692 continue;
2693 }
2694
4ed5c033
LC		 2695 schedule_timeout_interruptible(next_wakeup - cur);
2696
8f1f7453
ES		 2697 if (kthread_should_stop()) {
2698 ext4_clear_request_list();
2699 goto exit_thread;
2700 }
bfff6873
LC		 2701 }
2702
2703exit_thread:
2704 /*
2705 * It looks like the request list is empty, but we need
2706 * to check it under the li_list_mtx lock, to prevent any
2707 * additions into it, and of course we should lock ext4_li_mtx
2708 * to atomically free the list and ext4_li_info, because at
2709 * this point another ext4 filesystem could be registering
2710 * new one.
2711 */
2712 mutex_lock(&ext4_li_mtx);
2713 mutex_lock(&eli->li_list_mtx);
2714 if (!list_empty(&eli->li_request_list)) {
2715 mutex_unlock(&eli->li_list_mtx);
2716 mutex_unlock(&ext4_li_mtx);
2717 goto cont_thread;
2718 }
2719 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2720 kfree(ext4_li_info);
2721 ext4_li_info = NULL;
2722 mutex_unlock(&ext4_li_mtx);
2723
2724 return 0;
2725}
2726
2727static void ext4_clear_request_list(void)
2728{
2729 struct list_head *pos, *n;
2730 struct ext4_li_request *elr;
2731
2732 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2733 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2734 elr = list_entry(pos, struct ext4_li_request,
2735 lr_request);
2736 ext4_remove_li_request(elr);
2737 }
2738 mutex_unlock(&ext4_li_info->li_list_mtx);
2739}
2740
2741static int ext4_run_lazyinit_thread(void)
2742{
8f1f7453
ES		 2743 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2744 ext4_li_info, "ext4lazyinit");
2745 if (IS_ERR(ext4_lazyinit_task)) {
2746 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2747 ext4_clear_request_list();
bfff6873
LC		 2748 kfree(ext4_li_info);
2749 ext4_li_info = NULL;
92b97816 2750 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2751 "initialization thread\n",
2752 err);
2753 return err;
2754 }
2755 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2756 return 0;
2757}
2758
2759/*
2760 * Check whether it make sense to run itable init. thread or not.
2761 * If there is at least one uninitialized inode table, return
2762 * corresponding group number, else the loop goes through all
2763 * groups and return total number of groups.
2764 */
2765static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2766{
2767 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2768 struct ext4_group_desc *gdp = NULL;
2769
2770 for (group = 0; group < ngroups; group++) {
2771 gdp = ext4_get_group_desc(sb, group, NULL);
2772 if (!gdp)
2773 continue;
2774
2775 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2776 break;
2777 }
2778
2779 return group;
2780}
2781
2782static int ext4_li_info_new(void)
2783{
2784 struct ext4_lazy_init *eli = NULL;
2785
2786 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2787 if (!eli)
2788 return -ENOMEM;
2789
bfff6873
LC		 2790 INIT_LIST_HEAD(&eli->li_request_list);
2791 mutex_init(&eli->li_list_mtx);
2792
bfff6873
LC		 2793 eli->li_state |= EXT4_LAZYINIT_QUIT;
2794
2795 ext4_li_info = eli;
2796
2797 return 0;
2798}
2799
2800static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2801 ext4_group_t start)
2802{
2803 struct ext4_sb_info *sbi = EXT4_SB(sb);
2804 struct ext4_li_request *elr;
bfff6873
LC		 2805
2806 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2807 if (!elr)
2808 return NULL;
2809
2810 elr->lr_super = sb;
2811 elr->lr_sbi = sbi;
2812 elr->lr_next_group = start;
2813
2814 /*
2815 * Randomize first schedule time of the request to
2816 * spread the inode table initialization requests
2817 * better.
2818 */
dd1f723b
TT		 2819 elr->lr_next_sched = jiffies + (prandom_u32() %
2820 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 2821 return elr;
2822}
2823
7f511862
TT		 2824int ext4_register_li_request(struct super_block *sb,
2825 ext4_group_t first_not_zeroed)
bfff6873
LC		 2826{
2827 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 2828 struct ext4_li_request *elr = NULL;
bfff6873 2829 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2830 int ret = 0;
bfff6873 2831
7f511862 2832 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 2833 if (sbi->s_li_request != NULL) {
2834 /*
2835 * Reset timeout so it can be computed again, because
2836 * s_li_wait_mult might have changed.
2837 */
2838 sbi->s_li_request->lr_timeout = 0;
7f511862 2839 goto out;
51ce6511 2840 }
bfff6873
LC		 2841
2842 if (first_not_zeroed == ngroups ||
2843 (sb->s_flags & MS_RDONLY) ||
55ff3840 2844 !test_opt(sb, INIT_INODE_TABLE))
7f511862 2845 goto out;
bfff6873
LC		 2846
2847 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 2848 if (!elr) {
2849 ret = -ENOMEM;
2850 goto out;
2851 }
bfff6873
LC		 2852
2853 if (NULL == ext4_li_info) {
2854 ret = ext4_li_info_new();
2855 if (ret)
2856 goto out;
2857 }
2858
2859 mutex_lock(&ext4_li_info->li_list_mtx);
2860 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2861 mutex_unlock(&ext4_li_info->li_list_mtx);
2862
2863 sbi->s_li_request = elr;
46e4690b
TM		 2864 /*
2865 * set elr to NULL here since it has been inserted to
2866 * the request_list and the removal and free of it is
2867 * handled by ext4_clear_request_list from now on.
2868 */
2869 elr = NULL;
bfff6873
LC		 2870
2871 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2872 ret = ext4_run_lazyinit_thread();
2873 if (ret)
2874 goto out;
2875 }
bfff6873 2876out:
beed5ecb
NK		 2877 mutex_unlock(&ext4_li_mtx);
2878 if (ret)
bfff6873 2879 kfree(elr);
bfff6873
LC		 2880 return ret;
2881}
2882
2883/*
2884 * We do not need to lock anything since this is called on
2885 * module unload.
2886 */
2887static void ext4_destroy_lazyinit_thread(void)
2888{
2889 /*
2890 * If thread exited earlier
2891 * there's nothing to be done.
2892 */
8f1f7453 2893 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2894 return;
2895
8f1f7453 2896 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2897}
2898
25ed6e8a
DW		 2899static int set_journal_csum_feature_set(struct super_block *sb)
2900{
2901 int ret = 1;
2902 int compat, incompat;
2903 struct ext4_sb_info *sbi = EXT4_SB(sb);
2904
9aa5d32b 2905 if (ext4_has_metadata_csum(sb)) {
db9ee220 2906 /* journal checksum v3 */
25ed6e8a 2907 compat = 0;
db9ee220 2908 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 2909 } else {
2910 /* journal checksum v1 */
2911 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2912 incompat = 0;
2913 }
2914
feb8c6d3
DW		 2915 jbd2_journal_clear_features(sbi->s_journal,
2916 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2917 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2918 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 2919 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2920 ret = jbd2_journal_set_features(sbi->s_journal,
2921 compat, 0,
2922 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2923 incompat);
2924 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2925 ret = jbd2_journal_set_features(sbi->s_journal,
2926 compat, 0,
2927 incompat);
2928 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2929 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2930 } else {
feb8c6d3
DW		 2931 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2932 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 2933 }
2934
2935 return ret;
2936}
2937
952fc18e
TT		 2938/*
2939 * Note: calculating the overhead so we can be compatible with
2940 * historical BSD practice is quite difficult in the face of
2941 * clusters/bigalloc. This is because multiple metadata blocks from
2942 * different block group can end up in the same allocation cluster.
2943 * Calculating the exact overhead in the face of clustered allocation
2944 * requires either O(all block bitmaps) in memory or O(number of block
2945 * groups**2) in time. We will still calculate the superblock for
2946 * older file systems --- and if we come across with a bigalloc file
2947 * system with zero in s_overhead_clusters the estimate will be close to
2948 * correct especially for very large cluster sizes --- but for newer
2949 * file systems, it's better to calculate this figure once at mkfs
2950 * time, and store it in the superblock. If the superblock value is
2951 * present (even for non-bigalloc file systems), we will use it.
2952 */
2953static int count_overhead(struct super_block *sb, ext4_group_t grp,
2954 char *buf)
2955{
2956 struct ext4_sb_info *sbi = EXT4_SB(sb);
2957 struct ext4_group_desc *gdp;
2958 ext4_fsblk_t first_block, last_block, b;
2959 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
2960 int s, j, count = 0;
2961
e2b911c5 2962 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 2963 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
2964 sbi->s_itb_per_group + 2);
2965
952fc18e
TT		 2966 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
2967 (grp * EXT4_BLOCKS_PER_GROUP(sb));
2968 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
2969 for (i = 0; i < ngroups; i++) {
2970 gdp = ext4_get_group_desc(sb, i, NULL);
2971 b = ext4_block_bitmap(sb, gdp);
2972 if (b >= first_block && b <= last_block) {
2973 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2974 count++;
2975 }
2976 b = ext4_inode_bitmap(sb, gdp);
2977 if (b >= first_block && b <= last_block) {
2978 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2979 count++;
2980 }
2981 b = ext4_inode_table(sb, gdp);
2982 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
2983 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
2984 int c = EXT4_B2C(sbi, b - first_block);
2985 ext4_set_bit(c, buf);
2986 count++;
2987 }
2988 if (i != grp)
2989 continue;
2990 s = 0;
2991 if (ext4_bg_has_super(sb, grp)) {
2992 ext4_set_bit(s++, buf);
2993 count++;
2994 }
2995 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
2996 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
2997 count++;
2998 }
2999 }
3000 if (!count)
3001 return 0;
3002 return EXT4_CLUSTERS_PER_GROUP(sb) -
3003 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3004}
3005
3006/*
3007 * Compute the overhead and stash it in sbi->s_overhead
3008 */
3009int ext4_calculate_overhead(struct super_block *sb)
3010{
3011 struct ext4_sb_info *sbi = EXT4_SB(sb);
3012 struct ext4_super_block *es = sbi->s_es;
3013 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3014 ext4_fsblk_t overhead = 0;
4fdb5543 3015 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3016
952fc18e
TT		 3017 if (!buf)
3018 return -ENOMEM;
3019
3020 /*
3021 * Compute the overhead (FS structures). This is constant
3022 * for a given filesystem unless the number of block groups
3023 * changes so we cache the previous value until it does.
3024 */
3025
3026 /*
3027 * All of the blocks before first_data_block are overhead
3028 */
3029 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3030
3031 /*
3032 * Add the overhead found in each block group
3033 */
3034 for (i = 0; i < ngroups; i++) {
3035 int blks;
3036
3037 blks = count_overhead(sb, i, buf);
3038 overhead += blks;
3039 if (blks)
3040 memset(buf, 0, PAGE_SIZE);
3041 cond_resched();
3042 }
b003b524
ES		 3043 /* Add the internal journal blocks as well */
3044 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3045 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3046
952fc18e
TT		 3047 sbi->s_overhead = overhead;
3048 smp_wmb();
3049 free_page((unsigned long) buf);
3050 return 0;
3051}
3052
b5799018 3053static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3054{
3055 ext4_fsblk_t resv_clusters;
b5799018 3056 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3057
30fac0f7
JK		 3058 /*
3059 * There's no need to reserve anything when we aren't using extents.
3060 * The space estimates are exact, there are no unwritten extents,
3061 * hole punching doesn't need new metadata... This is needed especially
3062 * to keep ext2/3 backward compatibility.
3063 */
e2b911c5 3064 if (!ext4_has_feature_extents(sb))
b5799018 3065 return;
27dd4385
LC		 3066 /*
3067 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3068 * This should cover the situations where we can not afford to run
3069 * out of space like for example punch hole, or converting
556615dc 3070 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3071 * allocation would require 1, or 2 blocks, higher numbers are
3072 * very rare.
3073 */
b5799018
TT		 3074 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3075 sbi->s_cluster_bits);
27dd4385
LC		 3076
3077 do_div(resv_clusters, 50);
3078 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3079
b5799018 3080 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3081}
3082
2b2d6d01 3083static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3084{
d4c402d9 3085 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3086 struct buffer_head *bh;
617ba13b
MC		 3087 struct ext4_super_block *es = NULL;
3088 struct ext4_sb_info *sbi;
3089 ext4_fsblk_t block;
3090 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3091 ext4_fsblk_t logical_sb_block;
ac27a0ec 3092 unsigned long offset = 0;
ac27a0ec
DK		 3093 unsigned long journal_devnum = 0;
3094 unsigned long def_mount_opts;
3095 struct inode *root;
0390131b 3096 const char *descr;
dcc7dae3 3097 int ret = -ENOMEM;
281b5995 3098 int blocksize, clustersize;
4ec11028
TT		 3099 unsigned int db_count;
3100 unsigned int i;
281b5995 3101 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3102 __u64 blocks_count;
07aa2ea1 3103 int err = 0;
b3881f74 3104 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3105 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3106
3107 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3108 if (!sbi)
dcc7dae3 3109 goto out_free_orig;
705895b6
PE		 3110
3111 sbi->s_blockgroup_lock =
3112 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3113 if (!sbi->s_blockgroup_lock) {
3114 kfree(sbi);
dcc7dae3 3115 goto out_free_orig;
705895b6 3116 }
ac27a0ec 3117 sb->s_fs_info = sbi;
2c0544b2 3118 sbi->s_sb = sb;
240799cd 3119 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3120 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3121 if (sb->s_bdev->bd_part)
3122 sbi->s_sectors_written_start =
3123 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3124
9f6200bb 3125 /* Cleanup superblock name */
ec3904dc 3126 strreplace(sb->s_id, '/', '!');
9f6200bb 3127
07aa2ea1 3128 /* -EINVAL is default */
dcc7dae3 3129 ret = -EINVAL;
617ba13b 3130 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3131 if (!blocksize) {
b31e1552 3132 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3133 goto out_fail;
3134 }
3135
3136 /*
617ba13b 3137 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3138 * block sizes. We need to calculate the offset from buffer start.
3139 */
617ba13b 3140 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3141 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3142 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3143 } else {
70bbb3e0 3144 logical_sb_block = sb_block;
ac27a0ec
DK		 3145 }
3146
a8ac900b 3147 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3148 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3149 goto out_fail;
3150 }
3151 /*
3152 * Note: s_es must be initialized as soon as possible because
617ba13b 3153 * some ext4 macro-instructions depend on its value
ac27a0ec 3154 */
2716b802 3155 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3156 sbi->s_es = es;
3157 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3158 if (sb->s_magic != EXT4_SUPER_MAGIC)
3159 goto cantfind_ext4;
afc32f7e 3160 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3161
feb0ab32 3162 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3163 if (ext4_has_feature_metadata_csum(sb) &&
3164 ext4_has_feature_gdt_csum(sb))
363307e6 3165 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3166 "redundant flags; please run fsck.");
3167
d25425f8
DW		 3168 /* Check for a known checksum algorithm */
3169 if (!ext4_verify_csum_type(sb, es)) {
3170 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3171 "unknown checksum algorithm.");
3172 silent = 1;
3173 goto cantfind_ext4;
3174 }
3175
0441984a 3176 /* Load the checksum driver */
e2b911c5 3177 if (ext4_has_feature_metadata_csum(sb)) {
0441984a
DW		 3178 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3179 if (IS_ERR(sbi->s_chksum_driver)) {
3180 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3181 ret = PTR_ERR(sbi->s_chksum_driver);
3182 sbi->s_chksum_driver = NULL;
3183 goto failed_mount;
3184 }
3185 }
3186
a9c47317
DW		 3187 /* Check superblock checksum */
3188 if (!ext4_superblock_csum_verify(sb, es)) {
3189 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3190 "invalid superblock checksum. Run e2fsck?");
3191 silent = 1;
6a797d27 3192 ret = -EFSBADCRC;
a9c47317
DW		 3193 goto cantfind_ext4;
3194 }
3195
3196 /* Precompute checksum seed for all metadata */
e2b911c5 3197 if (ext4_has_feature_csum_seed(sb))
8c81bd8f
DW		 3198 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3199 else if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3200 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3201 sizeof(es->s_uuid));
3202
ac27a0ec
DK		 3203 /* Set defaults before we parse the mount options */
3204 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3205 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3206 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3207 set_opt(sb, DEBUG);
87f26807 3208 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3209 set_opt(sb, GRPID);
617ba13b 3210 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3211 set_opt(sb, NO_UID32);
ea663336 3212 /* xattr user namespace & acls are now defaulted on */
ea663336 3213 set_opt(sb, XATTR_USER);
03010a33 3214#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3215 set_opt(sb, POSIX_ACL);
2e7842b8 3216#endif
98c1a759
DW		 3217 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3218 if (ext4_has_metadata_csum(sb))
3219 set_opt(sb, JOURNAL_CHECKSUM);
3220
617ba13b 3221 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3222 set_opt(sb, JOURNAL_DATA);
617ba13b 3223 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3224 set_opt(sb, ORDERED_DATA);
617ba13b 3225 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3226 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3227
3228 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3229 set_opt(sb, ERRORS_PANIC);
bb4f397a 3230 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3231 set_opt(sb, ERRORS_CONT);
bb4f397a 3232 else
fd8c37ec 3233 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3234 /* block_validity enabled by default; disable with noblock_validity */
3235 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3236 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3237 set_opt(sb, DISCARD);
ac27a0ec 3238
08cefc7a
EB		 3239 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3240 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3241 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3242 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3243 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3244
8b67f04a 3245 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3246 set_opt(sb, BARRIER);
ac27a0ec 3247
dd919b98
AK		 3248 /*
3249 * enable delayed allocation by default
3250 * Use -o nodelalloc to turn it off
3251 */
bc0b75f7 3252 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3253 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3254 set_opt(sb, DELALLOC);
dd919b98 3255
51ce6511
LC		 3256 /*
3257 * set default s_li_wait_mult for lazyinit, for the case there is
3258 * no mount option specified.
3259 */
3260 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3261
8b67f04a 3262 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3263 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3264 ext4_msg(sb, KERN_WARNING,
3265 "failed to parse options in superblock: %s",
3266 sbi->s_es->s_mount_opts);
3267 }
5a916be1 3268 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3269 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3270 &journal_ioprio, 0))
ac27a0ec
DK		 3271 goto failed_mount;
3272
56889787
TT		 3273 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3274 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3275 "with data=journal disables delayed "
3276 "allocation and O_DIRECT support!\n");
3277 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3278 ext4_msg(sb, KERN_ERR, "can't mount with "
3279 "both data=journal and delalloc");
3280 goto failed_mount;
3281 }
3282 if (test_opt(sb, DIOREAD_NOLOCK)) {
3283 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3284 "both data=journal and dioread_nolock");
56889787
TT		 3285 goto failed_mount;
3286 }
923ae0ff
RZ		 3287 if (test_opt(sb, DAX)) {
3288 ext4_msg(sb, KERN_ERR, "can't mount with "
3289 "both data=journal and dax");
3290 goto failed_mount;
3291 }
56889787
TT		 3292 if (test_opt(sb, DELALLOC))
3293 clear_opt(sb, DELALLOC);
001e4a87
TH		 3294 } else {
3295 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3296 }
3297
ac27a0ec 3298 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3299 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3300
617ba13b 3301 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3302 (ext4_has_compat_features(sb) ||
3303 ext4_has_ro_compat_features(sb) ||
3304 ext4_has_incompat_features(sb)))
b31e1552
ES		 3305 ext4_msg(sb, KERN_WARNING,
3306 "feature flags set on rev 0 fs, "
3307 "running e2fsck is recommended");
469108ff 3308
ed3654eb
TT		 3309 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3310 set_opt2(sb, HURD_COMPAT);
e2b911c5 3311 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3312 ext4_msg(sb, KERN_ERR,
3313 "The Hurd can't support 64-bit file systems");
3314 goto failed_mount;
3315 }
3316 }
3317
2035e776
TT		 3318 if (IS_EXT2_SB(sb)) {
3319 if (ext2_feature_set_ok(sb))
3320 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3321 "using the ext4 subsystem");
3322 else {
3323 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3324 "to feature incompatibilities");
3325 goto failed_mount;
3326 }
3327 }
3328
3329 if (IS_EXT3_SB(sb)) {
3330 if (ext3_feature_set_ok(sb))
3331 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3332 "using the ext4 subsystem");
3333 else {
3334 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3335 "to feature incompatibilities");
3336 goto failed_mount;
3337 }
3338 }
3339
ac27a0ec
DK		 3340 /*
3341 * Check feature flags regardless of the revision level, since we
3342 * previously didn't change the revision level when setting the flags,
3343 * so there is a chance incompat flags are set on a rev 0 filesystem.
3344 */
a13fb1a4 3345 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3346 goto failed_mount;
a13fb1a4 3347
261cb20c 3348 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3349 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3350 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3351 ext4_msg(sb, KERN_ERR,
3352 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3353 goto failed_mount;
3354 }
3355
923ae0ff
RZ		 3356 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3357 if (blocksize != PAGE_SIZE) {
3358 ext4_msg(sb, KERN_ERR,
3359 "error: unsupported blocksize for dax");
3360 goto failed_mount;
3361 }
3362 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3363 ext4_msg(sb, KERN_ERR,
3364 "error: device does not support dax");
3365 goto failed_mount;
3366 }
3367 }
3368
e2b911c5 3369 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3370 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3371 es->s_encryption_level);
3372 goto failed_mount;
3373 }
3374
ac27a0ec 3375 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3376 /* Validate the filesystem blocksize */
3377 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3378 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3379 blocksize);
ac27a0ec
DK		 3380 goto failed_mount;
3381 }
3382
2b2d6d01 3383 brelse(bh);
70bbb3e0
AM		 3384 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3385 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3386 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3387 if (!bh) {
b31e1552
ES		 3388 ext4_msg(sb, KERN_ERR,
3389 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3390 goto failed_mount;
3391 }
2716b802 3392 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3393 sbi->s_es = es;
617ba13b 3394 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3395 ext4_msg(sb, KERN_ERR,
3396 "Magic mismatch, very weird!");
ac27a0ec
DK		 3397 goto failed_mount;
3398 }
3399 }
3400
e2b911c5 3401 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3402 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3403 has_huge_files);
3404 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3405
617ba13b
MC		 3406 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3407 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3408 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3409 } else {
3410 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3411 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3412 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3413 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3414 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3415 ext4_msg(sb, KERN_ERR,
3416 "unsupported inode size: %d",
2b2d6d01 3417 sbi->s_inode_size);
ac27a0ec
DK		 3418 goto failed_mount;
3419 }
ef7f3835
KS		 3420 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3421 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3422 }
0b8e58a1 3423
0d1ee42f 3424 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3425 if (ext4_has_feature_64bit(sb)) {
8fadc143 3426 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3427 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3428 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3429 ext4_msg(sb, KERN_ERR,
3430 "unsupported descriptor size %lu",
0d1ee42f
AR		 3431 sbi->s_desc_size);
3432 goto failed_mount;
3433 }
3434 } else
3435 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3436
ac27a0ec 3437 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3438 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3439 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3440 goto cantfind_ext4;
0b8e58a1 3441
617ba13b 3442 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3443 if (sbi->s_inodes_per_block == 0)
617ba13b 3444 goto cantfind_ext4;
ac27a0ec
DK		 3445 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3446 sbi->s_inodes_per_block;
0d1ee42f 3447 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3448 sbi->s_sbh = bh;
3449 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3450 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3451 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3452
2b2d6d01 3453 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3454 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3455 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3456 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3457 i = le32_to_cpu(es->s_flags);
3458 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3459 sbi->s_hash_unsigned = 3;
3460 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3461#ifdef __CHAR_UNSIGNED__
23301410
TT		 3462 if (!(sb->s_flags & MS_RDONLY))
3463 es->s_flags |=
3464 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3465 sbi->s_hash_unsigned = 3;
f99b2589 3466#else
23301410
TT		 3467 if (!(sb->s_flags & MS_RDONLY))
3468 es->s_flags |=
3469 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3470#endif
23301410 3471 }
f99b2589 3472 }
ac27a0ec 3473
281b5995
TT		 3474 /* Handle clustersize */
3475 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3476 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3477 if (has_bigalloc) {
3478 if (clustersize < blocksize) {
3479 ext4_msg(sb, KERN_ERR,
3480 "cluster size (%d) smaller than "
3481 "block size (%d)", clustersize, blocksize);
3482 goto failed_mount;
3483 }
3484 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3485 le32_to_cpu(es->s_log_block_size);
3486 sbi->s_clusters_per_group =
3487 le32_to_cpu(es->s_clusters_per_group);
3488 if (sbi->s_clusters_per_group > blocksize * 8) {
3489 ext4_msg(sb, KERN_ERR,
3490 "#clusters per group too big: %lu",
3491 sbi->s_clusters_per_group);
3492 goto failed_mount;
3493 }
3494 if (sbi->s_blocks_per_group !=
3495 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3496 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3497 "clusters per group (%lu) inconsistent",
3498 sbi->s_blocks_per_group,
3499 sbi->s_clusters_per_group);
3500 goto failed_mount;
3501 }
3502 } else {
3503 if (clustersize != blocksize) {
3504 ext4_warning(sb, "fragment/cluster size (%d) != "
3505 "block size (%d)", clustersize,
3506 blocksize);
3507 clustersize = blocksize;
3508 }
3509 if (sbi->s_blocks_per_group > blocksize * 8) {
3510 ext4_msg(sb, KERN_ERR,
3511 "#blocks per group too big: %lu",
3512 sbi->s_blocks_per_group);
3513 goto failed_mount;
3514 }
3515 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3516 sbi->s_cluster_bits = 0;
ac27a0ec 3517 }
281b5995
TT		 3518 sbi->s_cluster_ratio = clustersize / blocksize;
3519
ac27a0ec 3520 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3521 ext4_msg(sb, KERN_ERR,
3522 "#inodes per group too big: %lu",
2b2d6d01 3523 sbi->s_inodes_per_group);
ac27a0ec
DK		 3524 goto failed_mount;
3525 }
3526
960fd856
TT		 3527 /* Do we have standard group size of clustersize * 8 blocks ? */
3528 if (sbi->s_blocks_per_group == clustersize << 3)
3529 set_opt2(sb, STD_GROUP_SIZE);
3530
bf43d84b
ES		 3531 /*
3532 * Test whether we have more sectors than will fit in sector_t,
3533 * and whether the max offset is addressable by the page cache.
3534 */
5a9ae68a 3535 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3536 ext4_blocks_count(es));
5a9ae68a 3537 if (err) {
b31e1552 3538 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3539 " too large to mount safely on this system");
ac27a0ec 3540 if (sizeof(sector_t) < 8)
90c699a9 3541 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3542 goto failed_mount;
3543 }
3544
617ba13b
MC		 3545 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3546 goto cantfind_ext4;
e7c95593 3547
0f2ddca6
FTN		 3548 /* check blocks count against device size */
3549 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3550 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3551 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3552 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3553 ext4_blocks_count(es), blocks_count);
3554 goto failed_mount;
3555 }
3556
0b8e58a1
AD		 3557 /*
3558 * It makes no sense for the first data block to be beyond the end
3559 * of the filesystem.
3560 */
3561 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3562 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3563 "block %u is beyond end of filesystem (%llu)",
3564 le32_to_cpu(es->s_first_data_block),
3565 ext4_blocks_count(es));
e7c95593
ES		 3566 goto failed_mount;
3567 }
bd81d8ee
LV		 3568 blocks_count = (ext4_blocks_count(es) -
3569 le32_to_cpu(es->s_first_data_block) +
3570 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3571 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3572 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3573 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3574 "(block count %llu, first data block %u, "
b31e1552 3575 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3576 ext4_blocks_count(es),
3577 le32_to_cpu(es->s_first_data_block),
3578 EXT4_BLOCKS_PER_GROUP(sb));
3579 goto failed_mount;
3580 }
bd81d8ee 3581 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3582 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3583 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3584 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3585 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3586 sbi->s_group_desc = ext4_kvmalloc(db_count *
3587 sizeof(struct buffer_head *),
3588 GFP_KERNEL);
ac27a0ec 3589 if (sbi->s_group_desc == NULL) {
b31e1552 3590 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3591 ret = -ENOMEM;
ac27a0ec
DK		 3592 goto failed_mount;
3593 }
3594
705895b6 3595 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3596
3597 for (i = 0; i < db_count; i++) {
70bbb3e0 3598 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3599 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3600 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3601 ext4_msg(sb, KERN_ERR,
3602 "can't read group descriptor %d", i);
ac27a0ec
DK		 3603 db_count = i;
3604 goto failed_mount2;
3605 }
3606 }
bfff6873 3607 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3608 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 3609 ret = -EFSCORRUPTED;
f9ae9cf5 3610 goto failed_mount2;
ac27a0ec 3611 }
772cb7c8 3612
f9ae9cf5 3613 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3614 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3615 spin_lock_init(&sbi->s_next_gen_lock);
3616
04ecddb7
JM		 3617 setup_timer(&sbi->s_err_report, print_daily_error_info,
3618 (unsigned long) sb);
04496411 3619
a75ae78f 3620 /* Register extent status tree shrinker */
eb68d0e2 3621 if (ext4_es_register_shrinker(sbi))
ce7e010a 3622 goto failed_mount3;
ce7e010a 3623
c9de560d 3624 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3625 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3626
f9ae9cf5
TT		 3627 /*
3628 * set up enough so that it can read an inode
3629 */
f6e63f90 3630 sb->s_op = &ext4_sops;
617ba13b
MC		 3631 sb->s_export_op = &ext4_export_ops;
3632 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3633#ifdef CONFIG_QUOTA
617ba13b 3634 sb->dq_op = &ext4_quota_operations;
e2b911c5 3635 if (ext4_has_feature_quota(sb))
1fa5efe3 3636 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3637 else
3638 sb->s_qcop = &ext4_qctl_operations;
96c7e0d9 3639 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
ac27a0ec 3640#endif
f2fa2ffc
AK		 3641 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3642
ac27a0ec 3643 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3644 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3645
3646 sb->s_root = NULL;
3647
3648 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 3649 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 3650
e2b911c5 3651 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
c5e06d10 3652 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3653 goto failed_mount3a;
c5e06d10 3654
ac27a0ec
DK		 3655 /*
3656 * The first inode we look at is the journal inode. Don't try
3657 * root first: it may be modified in the journal!
3658 */
e2b911c5 3659 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
617ba13b 3660 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3661 goto failed_mount3a;
0390131b 3662 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
e2b911c5 3663 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 3664 ext4_msg(sb, KERN_ERR, "required journal recovery "
3665 "suppressed and not mounted read-only");
744692dc 3666 goto failed_mount_wq;
ac27a0ec 3667 } else {
fd8c37ec 3668 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3669 sbi->s_journal = NULL;
3670 needs_recovery = 0;
3671 goto no_journal;
ac27a0ec
DK		 3672 }
3673
e2b911c5 3674 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 3675 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3676 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3677 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3678 goto failed_mount_wq;
eb40a09c
JS		 3679 }
3680
25ed6e8a
DW		 3681 if (!set_journal_csum_feature_set(sb)) {
3682 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3683 "feature set");
3684 goto failed_mount_wq;
d4da6c9c 3685 }
818d276c 3686
ac27a0ec
DK		 3687 /* We have now updated the journal if required, so we can
3688 * validate the data journaling mode. */
3689 switch (test_opt(sb, DATA_FLAGS)) {
3690 case 0:
3691 /* No mode set, assume a default based on the journal
63f57933
AM		 3692 * capabilities: ORDERED_DATA if the journal can
3693 * cope, else JOURNAL_DATA
3694 */
dab291af
MC		 3695 if (jbd2_journal_check_available_features
3696 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3697 set_opt(sb, ORDERED_DATA);
ac27a0ec 3698 else
fd8c37ec 3699 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3700 break;
3701
617ba13b
MC		 3702 case EXT4_MOUNT_ORDERED_DATA:
3703 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3704 if (!jbd2_journal_check_available_features
3705 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3706 ext4_msg(sb, KERN_ERR, "Journal does not support "
3707 "requested data journaling mode");
744692dc 3708 goto failed_mount_wq;
ac27a0ec
DK		 3709 }
3710 default:
3711 break;
3712 }
b3881f74 3713 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3714
18aadd47
BJ		 3715 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3716
ce7e010a 3717no_journal:
9c191f70
M		 3718 if (ext4_mballoc_ready) {
3719 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
3720 if (!sbi->s_mb_cache) {
3721 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3722 goto failed_mount_wq;
3723 }
3724 }
3725
e2b911c5 3726 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
1cb767cd
TT		 3727 (blocksize != PAGE_CACHE_SIZE)) {
3728 ext4_msg(sb, KERN_ERR,
3729 "Unsupported blocksize for fs encryption");
3730 goto failed_mount_wq;
3731 }
3732
e2b911c5
DW		 3733 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3734 !ext4_has_feature_encrypt(sb)) {
3735 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 3736 ext4_commit_super(sb, 1);
3737 }
3738
952fc18e
TT		 3739 /*
3740 * Get the # of file system overhead blocks from the
3741 * superblock if present.
3742 */
3743 if (es->s_overhead_clusters)
3744 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3745 else {
07aa2ea1
LC		 3746 err = ext4_calculate_overhead(sb);
3747 if (err)
952fc18e
TT		 3748 goto failed_mount_wq;
3749 }
3750
fd89d5f2
TH		 3751 /*
3752 * The maximum number of concurrent works can be high and
3753 * concurrency isn't really necessary. Limit it to 1.
3754 */
2e8fa54e
JK		 3755 EXT4_SB(sb)->rsv_conversion_wq =
3756 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3757 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3758 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 3759 ret = -ENOMEM;
2e8fa54e
JK		 3760 goto failed_mount4;
3761 }
3762
ac27a0ec 3763 /*
dab291af 3764 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3765 * so we can safely mount the rest of the filesystem now.
3766 */
3767
1d1fe1ee
DH		 3768 root = ext4_iget(sb, EXT4_ROOT_INO);
3769 if (IS_ERR(root)) {
b31e1552 3770 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3771 ret = PTR_ERR(root);
32a9bb57 3772 root = NULL;
ac27a0ec
DK		 3773 goto failed_mount4;
3774 }
3775 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3776 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3777 iput(root);
ac27a0ec
DK		 3778 goto failed_mount4;
3779 }
48fde701 3780 sb->s_root = d_make_root(root);
1d1fe1ee 3781 if (!sb->s_root) {
b31e1552 3782 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3783 ret = -ENOMEM;
3784 goto failed_mount4;
3785 }
ac27a0ec 3786
7e84b621
ES		 3787 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3788 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3789
3790 /* determine the minimum size of new large inodes, if present */
3791 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3792 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3793 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 3794 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 3795 if (sbi->s_want_extra_isize <
3796 le16_to_cpu(es->s_want_extra_isize))
3797 sbi->s_want_extra_isize =
3798 le16_to_cpu(es->s_want_extra_isize);
3799 if (sbi->s_want_extra_isize <
3800 le16_to_cpu(es->s_min_extra_isize))
3801 sbi->s_want_extra_isize =
3802 le16_to_cpu(es->s_min_extra_isize);
3803 }
3804 }
3805 /* Check if enough inode space is available */
3806 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3807 sbi->s_inode_size) {
3808 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3809 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3810 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3811 "available");
ef7f3835
KS		 3812 }
3813
b5799018 3814 ext4_set_resv_clusters(sb);
27dd4385 3815
6fd058f7
TT		 3816 err = ext4_setup_system_zone(sb);
3817 if (err) {
b31e1552 3818 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3819 "zone (%d)", err);
f9ae9cf5
TT		 3820 goto failed_mount4a;
3821 }
3822
3823 ext4_ext_init(sb);
3824 err = ext4_mb_init(sb);
3825 if (err) {
3826 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3827 err);
dcf2d804 3828 goto failed_mount5;
c2774d84
AK		 3829 }
3830
d5e03cbb
TT		 3831 block = ext4_count_free_clusters(sb);
3832 ext4_free_blocks_count_set(sbi->s_es,
3833 EXT4_C2B(sbi, block));
908c7f19
TH		 3834 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3835 GFP_KERNEL);
d5e03cbb
TT		 3836 if (!err) {
3837 unsigned long freei = ext4_count_free_inodes(sb);
3838 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 3839 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3840 GFP_KERNEL);
d5e03cbb
TT		 3841 }
3842 if (!err)
3843 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 3844 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 3845 if (!err)
908c7f19
TH		 3846 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3847 GFP_KERNEL);
d5e03cbb
TT		 3848 if (err) {
3849 ext4_msg(sb, KERN_ERR, "insufficient memory");
3850 goto failed_mount6;
3851 }
3852
e2b911c5 3853 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 3854 if (!ext4_fill_flex_info(sb)) {
3855 ext4_msg(sb, KERN_ERR,
3856 "unable to initialize "
3857 "flex_bg meta info!");
3858 goto failed_mount6;
3859 }
3860
bfff6873
LC		 3861 err = ext4_register_li_request(sb, first_not_zeroed);
3862 if (err)
dcf2d804 3863 goto failed_mount6;
bfff6873 3864
b5799018 3865 err = ext4_register_sysfs(sb);
dcf2d804
TM		 3866 if (err)
3867 goto failed_mount7;
3197ebdb 3868
9b2ff357
JK		 3869#ifdef CONFIG_QUOTA
3870 /* Enable quota usage during mount. */
e2b911c5 3871 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
9b2ff357
JK		 3872 err = ext4_enable_quotas(sb);
3873 if (err)
3874 goto failed_mount8;
3875 }
3876#endif /* CONFIG_QUOTA */
3877
617ba13b
MC		 3878 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3879 ext4_orphan_cleanup(sb, es);
3880 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3881 if (needs_recovery) {
b31e1552 3882 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3883 ext4_mark_recovery_complete(sb, es);
3884 }
3885 if (EXT4_SB(sb)->s_journal) {
3886 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3887 descr = " journalled data mode";
3888 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3889 descr = " ordered data mode";
3890 else
3891 descr = " writeback data mode";
3892 } else
3893 descr = "out journal";
3894
79add3a3
LC		 3895 if (test_opt(sb, DISCARD)) {
3896 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3897 if (!blk_queue_discard(q))
3898 ext4_msg(sb, KERN_WARNING,
3899 "mounting with \"discard\" option, but "
3900 "the device does not support discard");
3901 }
3902
e294a537
TT		 3903 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3904 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3905 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3906 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3907
66e61a9e
TT		 3908 if (es->s_error_count)
3909 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3910
efbed4dc
TT		 3911 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3912 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
3913 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
3914 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
3915
d4c402d9 3916 kfree(orig_data);
ac27a0ec
DK		 3917 return 0;
3918
617ba13b 3919cantfind_ext4:
ac27a0ec 3920 if (!silent)
b31e1552 3921 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3922 goto failed_mount;
3923
72ba7450
TT		 3924#ifdef CONFIG_QUOTA
3925failed_mount8:
ebd173be 3926 ext4_unregister_sysfs(sb);
72ba7450 3927#endif
dcf2d804
TM		 3928failed_mount7:
3929 ext4_unregister_li_request(sb);
3930failed_mount6:
f9ae9cf5 3931 ext4_mb_release(sb);
d5e03cbb 3932 if (sbi->s_flex_groups)
b93b41d4 3933 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 3934 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3935 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3936 percpu_counter_destroy(&sbi->s_dirs_counter);
3937 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 3938failed_mount5:
f9ae9cf5
TT		 3939 ext4_ext_release(sb);
3940 ext4_release_system_zone(sb);
3941failed_mount4a:
94bf608a 3942 dput(sb->s_root);
32a9bb57 3943 sb->s_root = NULL;
94bf608a 3944failed_mount4:
b31e1552 3945 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 3946 if (EXT4_SB(sb)->rsv_conversion_wq)
3947 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 3948failed_mount_wq:
0390131b
FM		 3949 if (sbi->s_journal) {
3950 jbd2_journal_destroy(sbi->s_journal);
3951 sbi->s_journal = NULL;
3952 }
50460fe8 3953failed_mount3a:
d3922a77 3954 ext4_es_unregister_shrinker(sbi);
eb68d0e2 3955failed_mount3:
9105bb14 3956 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 3957 if (sbi->s_mmp_tsk)
3958 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3959failed_mount2:
3960 for (i = 0; i < db_count; i++)
3961 brelse(sbi->s_group_desc[i]);
b93b41d4 3962 kvfree(sbi->s_group_desc);
ac27a0ec 3963failed_mount:
0441984a
DW		 3964 if (sbi->s_chksum_driver)
3965 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 3966#ifdef CONFIG_QUOTA
a2d4a646 3967 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 3968 kfree(sbi->s_qf_names[i]);
3969#endif
617ba13b 3970 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3971 brelse(bh);
3972out_fail:
3973 sb->s_fs_info = NULL;
f6830165 3974 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3975 kfree(sbi);
dcc7dae3 3976out_free_orig:
d4c402d9 3977 kfree(orig_data);
07aa2ea1 3978 return err ? err : ret;
ac27a0ec
DK		 3979}
3980
3981/*
3982 * Setup any per-fs journal parameters now. We'll do this both on
3983 * initial mount, once the journal has been initialised but before we've
3984 * done any recovery; and again on any subsequent remount.
3985 */
617ba13b 3986static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3987{
617ba13b 3988 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3989
30773840
TT		 3990 journal->j_commit_interval = sbi->s_commit_interval;
3991 journal->j_min_batch_time = sbi->s_min_batch_time;
3992 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3993
a931da6a 3994 write_lock(&journal->j_state_lock);
ac27a0ec 3995 if (test_opt(sb, BARRIER))
dab291af 3996 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3997 else
dab291af 3998 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3999 if (test_opt(sb, DATA_ERR_ABORT))
4000 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4001 else
4002 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4003 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4004}
4005
617ba13b 4006static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4007 unsigned int journal_inum)
4008{
4009 struct inode *journal_inode;
4010 journal_t *journal;
4011
e2b911c5 4012 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4013
ac27a0ec
DK		 4014 /* First, test for the existence of a valid inode on disk. Bad
4015 * things happen if we iget() an unused inode, as the subsequent
4016 * iput() will try to delete it. */
4017
1d1fe1ee
DH		 4018 journal_inode = ext4_iget(sb, journal_inum);
4019 if (IS_ERR(journal_inode)) {
b31e1552 4020 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4021 return NULL;
4022 }
4023 if (!journal_inode->i_nlink) {
4024 make_bad_inode(journal_inode);
4025 iput(journal_inode);
b31e1552 4026 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4027 return NULL;
4028 }
4029
e5f8eab8 4030 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4031 journal_inode, journal_inode->i_size);
1d1fe1ee 4032 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4033 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4034 iput(journal_inode);
4035 return NULL;
4036 }
4037
dab291af 4038 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4039 if (!journal) {
b31e1552 4040 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4041 iput(journal_inode);
4042 return NULL;
4043 }
4044 journal->j_private = sb;
617ba13b 4045 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4046 return journal;
4047}
4048
617ba13b 4049static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4050 dev_t j_dev)
4051{
2b2d6d01 4052 struct buffer_head *bh;
ac27a0ec 4053 journal_t *journal;
617ba13b
MC		 4054 ext4_fsblk_t start;
4055 ext4_fsblk_t len;
ac27a0ec 4056 int hblock, blocksize;
617ba13b 4057 ext4_fsblk_t sb_block;
ac27a0ec 4058 unsigned long offset;
2b2d6d01 4059 struct ext4_super_block *es;
ac27a0ec
DK		 4060 struct block_device *bdev;
4061
e2b911c5 4062 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4063
b31e1552 4064 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4065 if (bdev == NULL)
4066 return NULL;
4067
ac27a0ec 4068 blocksize = sb->s_blocksize;
e1defc4f 4069 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4070 if (blocksize < hblock) {
b31e1552
ES		 4071 ext4_msg(sb, KERN_ERR,
4072 "blocksize too small for journal device");
ac27a0ec
DK		 4073 goto out_bdev;
4074 }
4075
617ba13b
MC		 4076 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4077 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4078 set_blocksize(bdev, blocksize);
4079 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4080 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4081 "external journal");
ac27a0ec
DK		 4082 goto out_bdev;
4083 }
4084
2716b802 4085 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4086 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4087 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4088 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4089 ext4_msg(sb, KERN_ERR, "external journal has "
4090 "bad superblock");
ac27a0ec
DK		 4091 brelse(bh);
4092 goto out_bdev;
4093 }
4094
df4763be
DW		 4095 if ((le32_to_cpu(es->s_feature_ro_compat) &
4096 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4097 es->s_checksum != ext4_superblock_csum(sb, es)) {
4098 ext4_msg(sb, KERN_ERR, "external journal has "
4099 "corrupt superblock");
4100 brelse(bh);
4101 goto out_bdev;
4102 }
4103
617ba13b 4104 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4105 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4106 brelse(bh);
4107 goto out_bdev;
4108 }
4109
bd81d8ee 4110 len = ext4_blocks_count(es);
ac27a0ec
DK		 4111 start = sb_block + 1;
4112 brelse(bh); /* we're done with the superblock */
4113
dab291af 4114 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4115 start, len, blocksize);
4116 if (!journal) {
b31e1552 4117 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4118 goto out_bdev;
4119 }
4120 journal->j_private = sb;
9f203507 4121 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4122 wait_on_buffer(journal->j_sb_buffer);
4123 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4124 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4125 goto out_journal;
4126 }
4127 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4128 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4129 "user (unsupported) - %d",
ac27a0ec
DK		 4130 be32_to_cpu(journal->j_superblock->s_nr_users));
4131 goto out_journal;
4132 }
617ba13b
MC		 4133 EXT4_SB(sb)->journal_bdev = bdev;
4134 ext4_init_journal_params(sb, journal);
ac27a0ec 4135 return journal;
0b8e58a1 4136
ac27a0ec 4137out_journal:
dab291af 4138 jbd2_journal_destroy(journal);
ac27a0ec 4139out_bdev:
617ba13b 4140 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4141 return NULL;
4142}
4143
617ba13b
MC		 4144static int ext4_load_journal(struct super_block *sb,
4145 struct ext4_super_block *es,
ac27a0ec
DK		 4146 unsigned long journal_devnum)
4147{
4148 journal_t *journal;
4149 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4150 dev_t journal_dev;
4151 int err = 0;
4152 int really_read_only;
4153
e2b911c5 4154 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4155
ac27a0ec
DK		 4156 if (journal_devnum &&
4157 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4158 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4159 "numbers have changed");
ac27a0ec
DK		 4160 journal_dev = new_decode_dev(journal_devnum);
4161 } else
4162 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4163
4164 really_read_only = bdev_read_only(sb->s_bdev);
4165
4166 /*
4167 * Are we loading a blank journal or performing recovery after a
4168 * crash? For recovery, we need to check in advance whether we
4169 * can get read-write access to the device.
4170 */
e2b911c5 4171 if (ext4_has_feature_journal_needs_recovery(sb)) {
ac27a0ec 4172 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4173 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4174 "required on readonly filesystem");
ac27a0ec 4175 if (really_read_only) {
b31e1552
ES		 4176 ext4_msg(sb, KERN_ERR, "write access "
4177 "unavailable, cannot proceed");
ac27a0ec
DK		 4178 return -EROFS;
4179 }
b31e1552
ES		 4180 ext4_msg(sb, KERN_INFO, "write access will "
4181 "be enabled during recovery");
ac27a0ec
DK		 4182 }
4183 }
4184
4185 if (journal_inum && journal_dev) {
b31e1552
ES		 4186 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4187 "and inode journals!");
ac27a0ec
DK		 4188 return -EINVAL;
4189 }
4190
4191 if (journal_inum) {
617ba13b 4192 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4193 return -EINVAL;
4194 } else {
617ba13b 4195 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4196 return -EINVAL;
4197 }
4198
90576c0b 4199 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4200 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4201
e2b911c5 4202 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4203 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4204 if (!err) {
4205 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4206 if (save)
4207 memcpy(save, ((char *) es) +
4208 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4209 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4210 if (save)
4211 memcpy(((char *) es) + EXT4_S_ERR_START,
4212 save, EXT4_S_ERR_LEN);
4213 kfree(save);
4214 }
ac27a0ec
DK		 4215
4216 if (err) {
b31e1552 4217 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4218 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4219 return err;
4220 }
4221
617ba13b
MC		 4222 EXT4_SB(sb)->s_journal = journal;
4223 ext4_clear_journal_err(sb, es);
ac27a0ec 4224
c41303ce 4225 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4226 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4227 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4228
4229 /* Make sure we flush the recovery flag to disk. */
e2d67052 4230 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4231 }
4232
4233 return 0;
4234}
4235
e2d67052 4236static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4237{
e2d67052 4238 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4239 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4240 int error = 0;
ac27a0ec 4241
bdfe0cbd 4242 if (!sbh || block_device_ejected(sb))
c4be0c1d 4243 return error;
914258bf
TT		 4244 if (buffer_write_io_error(sbh)) {
4245 /*
4246 * Oh, dear. A previous attempt to write the
4247 * superblock failed. This could happen because the
4248 * USB device was yanked out. Or it could happen to
4249 * be a transient write error and maybe the block will
4250 * be remapped. Nothing we can do but to retry the
4251 * write and hope for the best.
4252 */
b31e1552
ES		 4253 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4254 "superblock detected");
914258bf
TT		 4255 clear_buffer_write_io_error(sbh);
4256 set_buffer_uptodate(sbh);
4257 }
71290b36
TT		 4258 /*
4259 * If the file system is mounted read-only, don't update the
4260 * superblock write time. This avoids updating the superblock
4261 * write time when we are mounting the root file system
4262 * read/only but we need to replay the journal; at that point,
4263 * for people who are east of GMT and who make their clock
4264 * tick in localtime for Windows bug-for-bug compatibility,
4265 * the clock is set in the future, and this will cause e2fsck
4266 * to complain and force a full file system check.
4267 */
4268 if (!(sb->s_flags & MS_RDONLY))
4269 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4270 if (sb->s_bdev->bd_part)
4271 es->s_kbytes_written =
4272 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4273 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4274 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4275 else
4276 es->s_kbytes_written =
4277 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4278 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4279 ext4_free_blocks_count_set(es,
57042651
TT		 4280 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4281 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4282 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4283 es->s_free_inodes_count =
4284 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4285 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4286 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4287 ext4_superblock_csum_set(sb);
ac27a0ec 4288 mark_buffer_dirty(sbh);
914258bf 4289 if (sync) {
564bc402
DJ		 4290 error = __sync_dirty_buffer(sbh,
4291 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
c4be0c1d
TS		 4292 if (error)
4293 return error;
4294
4295 error = buffer_write_io_error(sbh);
4296 if (error) {
b31e1552
ES		 4297 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4298 "superblock");
914258bf
TT		 4299 clear_buffer_write_io_error(sbh);
4300 set_buffer_uptodate(sbh);
4301 }
4302 }
c4be0c1d 4303 return error;
ac27a0ec
DK		 4304}
4305
ac27a0ec
DK		 4306/*
4307 * Have we just finished recovery? If so, and if we are mounting (or
4308 * remounting) the filesystem readonly, then we will end up with a
4309 * consistent fs on disk. Record that fact.
4310 */
2b2d6d01
TT		 4311static void ext4_mark_recovery_complete(struct super_block *sb,
4312 struct ext4_super_block *es)
ac27a0ec 4313{
617ba13b 4314 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4315
e2b911c5 4316 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4317 BUG_ON(journal != NULL);
4318 return;
4319 }
dab291af 4320 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4321 if (jbd2_journal_flush(journal) < 0)
4322 goto out;
4323
e2b911c5 4324 if (ext4_has_feature_journal_needs_recovery(sb) &&
ac27a0ec 4325 sb->s_flags & MS_RDONLY) {
e2b911c5 4326 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4327 ext4_commit_super(sb, 1);
ac27a0ec 4328 }
7ffe1ea8
HK		 4329
4330out:
dab291af 4331 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4332}
4333
4334/*
4335 * If we are mounting (or read-write remounting) a filesystem whose journal
4336 * has recorded an error from a previous lifetime, move that error to the
4337 * main filesystem now.
4338 */
2b2d6d01
TT		 4339static void ext4_clear_journal_err(struct super_block *sb,
4340 struct ext4_super_block *es)
ac27a0ec
DK		 4341{
4342 journal_t *journal;
4343 int j_errno;
4344 const char *errstr;
4345
e2b911c5 4346 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4347
617ba13b 4348 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4349
4350 /*
4351 * Now check for any error status which may have been recorded in the
617ba13b 4352 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4353 */
4354
dab291af 4355 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4356 if (j_errno) {
4357 char nbuf[16];
4358
617ba13b 4359 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4360 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4361 "from previous mount: %s", errstr);
12062ddd 4362 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4363
617ba13b
MC		 4364 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4365 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4366 ext4_commit_super(sb, 1);
ac27a0ec 4367
dab291af 4368 jbd2_journal_clear_err(journal);
d796c52e 4369 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4370 }
4371}
4372
4373/*
4374 * Force the running and committing transactions to commit,
4375 * and wait on the commit.
4376 */
617ba13b 4377int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4378{
4379 journal_t *journal;
ac27a0ec
DK		 4380
4381 if (sb->s_flags & MS_RDONLY)
4382 return 0;
4383
617ba13b 4384 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4385 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4386}
4387
617ba13b 4388static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4389{
14ce0cb4 4390 int ret = 0;
9eddacf9 4391 tid_t target;
06a407f1 4392 bool needs_barrier = false;
8d5d02e6 4393 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4394
9bffad1e 4395 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4396 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4397 /*
4398 * Writeback quota in non-journalled quota case - journalled quota has
4399 * no dirty dquots
4400 */
4401 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4402 /*
4403 * Data writeback is possible w/o journal transaction, so barrier must
4404 * being sent at the end of the function. But we can skip it if
4405 * transaction_commit will do it for us.
4406 */
bda32530
TT		 4407 if (sbi->s_journal) {
4408 target = jbd2_get_latest_transaction(sbi->s_journal);
4409 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4410 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4411 needs_barrier = true;
4412
4413 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4414 if (wait)
4415 ret = jbd2_log_wait_commit(sbi->s_journal,
4416 target);
4417 }
4418 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4419 needs_barrier = true;
06a407f1
DM		 4420 if (needs_barrier) {
4421 int err;
4422 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4423 if (!ret)
4424 ret = err;
0390131b 4425 }
06a407f1
DM		 4426
4427 return ret;
4428}
4429
ac27a0ec
DK		 4430/*
4431 * LVM calls this function before a (read-only) snapshot is created. This
4432 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4433 *
4434 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4435 * state independently. It relies on upper layer to stop all data & metadata
4436 * modifications.
ac27a0ec 4437 */
c4be0c1d 4438static int ext4_freeze(struct super_block *sb)
ac27a0ec 4439{
c4be0c1d
TS		 4440 int error = 0;
4441 journal_t *journal;
ac27a0ec 4442
9ca92389
TT		 4443 if (sb->s_flags & MS_RDONLY)
4444 return 0;
ac27a0ec 4445
9ca92389 4446 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4447
bb044576
TT		 4448 if (journal) {
4449 /* Now we set up the journal barrier. */
4450 jbd2_journal_lock_updates(journal);
ac27a0ec 4451
bb044576
TT		 4452 /*
4453 * Don't clear the needs_recovery flag if we failed to
4454 * flush the journal.
4455 */
4456 error = jbd2_journal_flush(journal);
4457 if (error < 0)
4458 goto out;
c642dc9e
ES		 4459
4460 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4461 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4462 }
9ca92389 4463
9ca92389 4464 error = ext4_commit_super(sb, 1);
6b0310fb 4465out:
bb044576
TT		 4466 if (journal)
4467 /* we rely on upper layer to stop further updates */
4468 jbd2_journal_unlock_updates(journal);
6b0310fb 4469 return error;
ac27a0ec
DK		 4470}
4471
4472/*
4473 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4474 * flag here, even though the filesystem is not technically dirty yet.
4475 */
c4be0c1d 4476static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4477{
9ca92389
TT		 4478 if (sb->s_flags & MS_RDONLY)
4479 return 0;
4480
c642dc9e
ES		 4481 if (EXT4_SB(sb)->s_journal) {
4482 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4483 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4484 }
4485
9ca92389 4486 ext4_commit_super(sb, 1);
c4be0c1d 4487 return 0;
ac27a0ec
DK		 4488}
4489
673c6100
TT		 4490/*
4491 * Structure to save mount options for ext4_remount's benefit
4492 */
4493struct ext4_mount_options {
4494 unsigned long s_mount_opt;
a2595b8a 4495 unsigned long s_mount_opt2;
08cefc7a
EB		 4496 kuid_t s_resuid;
4497 kgid_t s_resgid;
673c6100
TT		 4498 unsigned long s_commit_interval;
4499 u32 s_min_batch_time, s_max_batch_time;
4500#ifdef CONFIG_QUOTA
4501 int s_jquota_fmt;
a2d4a646 4502 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4503#endif
4504};
4505
2b2d6d01 4506static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4507{
2b2d6d01 4508 struct ext4_super_block *es;
617ba13b 4509 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4510 unsigned long old_sb_flags;
617ba13b 4511 struct ext4_mount_options old_opts;
c79d967d 4512 int enable_quota = 0;
8a266467 4513 ext4_group_t g;
b3881f74 4514 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4515 int err = 0;
ac27a0ec 4516#ifdef CONFIG_QUOTA
03dafb5f 4517 int i, j;
ac27a0ec 4518#endif
d4c402d9 4519 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4520
4521 /* Store the original options */
4522 old_sb_flags = sb->s_flags;
4523 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4524 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4525 old_opts.s_resuid = sbi->s_resuid;
4526 old_opts.s_resgid = sbi->s_resgid;
4527 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4528 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4529 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4530#ifdef CONFIG_QUOTA
4531 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4532 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4533 if (sbi->s_qf_names[i]) {
4534 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4535 GFP_KERNEL);
4536 if (!old_opts.s_qf_names[i]) {
4537 for (j = 0; j < i; j++)
4538 kfree(old_opts.s_qf_names[j]);
3e36a163 4539 kfree(orig_data);
03dafb5f
CG		 4540 return -ENOMEM;
4541 }
4542 } else
4543 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4544#endif
b3881f74
TT		 4545 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4546 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4547
661aa520 4548 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4549 err = -EINVAL;
4550 goto restore_opts;
4551 }
4552
6b992ff2 4553 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4554 test_opt(sb, JOURNAL_CHECKSUM)) {
4555 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4556 "during remount not supported; ignoring");
4557 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4558 }
4559
6ae6514b
PS		 4560 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4561 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4562 ext4_msg(sb, KERN_ERR, "can't mount with "
4563 "both data=journal and delalloc");
4564 err = -EINVAL;
4565 goto restore_opts;
4566 }
4567 if (test_opt(sb, DIOREAD_NOLOCK)) {
4568 ext4_msg(sb, KERN_ERR, "can't mount with "
4569 "both data=journal and dioread_nolock");
4570 err = -EINVAL;
4571 goto restore_opts;
4572 }
923ae0ff
RZ		 4573 if (test_opt(sb, DAX)) {
4574 ext4_msg(sb, KERN_ERR, "can't mount with "
4575 "both data=journal and dax");
4576 err = -EINVAL;
4577 goto restore_opts;
4578 }
4579 }
4580
4581 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4582 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4583 "dax flag with busy inodes while remounting");
4584 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4585 }
4586
4ab2f15b 4587 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4588 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4589
4590 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4591 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4592
4593 es = sbi->s_es;
4594
b3881f74 4595 if (sbi->s_journal) {
0390131b 4596 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4597 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4598 }
ac27a0ec 4599
a2fd66d0
TT		 4600 if (*flags & MS_LAZYTIME)
4601 sb->s_flags |= MS_LAZYTIME;
4602
661aa520 4603 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4604 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4605 err = -EROFS;
4606 goto restore_opts;
4607 }
4608
4609 if (*flags & MS_RDONLY) {
38c03b34
TT		 4610 err = sync_filesystem(sb);
4611 if (err < 0)
4612 goto restore_opts;
0f0dd62f
CH		 4613 err = dquot_suspend(sb, -1);
4614 if (err < 0)
c79d967d 4615 goto restore_opts;
c79d967d 4616
ac27a0ec
DK		 4617 /*
4618 * First of all, the unconditional stuff we have to do
4619 * to disable replay of the journal when we next remount
4620 */
4621 sb->s_flags |= MS_RDONLY;
4622
4623 /*
4624 * OK, test if we are remounting a valid rw partition
4625 * readonly, and if so set the rdonly flag and then
4626 * mark the partition as valid again.
4627 */
617ba13b
MC		 4628 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4629 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4630 es->s_state = cpu_to_le16(sbi->s_mount_state);
4631
a63c9eb2 4632 if (sbi->s_journal)
0390131b 4633 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4634 } else {
a13fb1a4 4635 /* Make sure we can mount this feature set readwrite */
e2b911c5 4636 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 4637 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4638 err = -EROFS;
4639 goto restore_opts;
4640 }
8a266467
TT		 4641 /*
4642 * Make sure the group descriptor checksums
0b8e58a1 4643 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4644 */
4645 for (g = 0; g < sbi->s_groups_count; g++) {
4646 struct ext4_group_desc *gdp =
4647 ext4_get_group_desc(sb, g, NULL);
4648
feb0ab32 4649 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4650 ext4_msg(sb, KERN_ERR,
4651 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 4652 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 4653 le16_to_cpu(gdp->bg_checksum));
6a797d27 4654 err = -EFSBADCRC;
8a266467
TT		 4655 goto restore_opts;
4656 }
4657 }
4658
ead6596b
ES		 4659 /*
4660 * If we have an unprocessed orphan list hanging
4661 * around from a previously readonly bdev mount,
4662 * require a full umount/remount for now.
4663 */
4664 if (es->s_last_orphan) {
b31e1552 4665 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4666 "remount RDWR because of unprocessed "
4667 "orphan inode list. Please "
b31e1552 4668 "umount/remount instead");
ead6596b
ES		 4669 err = -EINVAL;
4670 goto restore_opts;
4671 }
4672
ac27a0ec
DK		 4673 /*
4674 * Mounting a RDONLY partition read-write, so reread
4675 * and store the current valid flag. (It may have
4676 * been changed by e2fsck since we originally mounted
4677 * the partition.)
4678 */
0390131b
FM		 4679 if (sbi->s_journal)
4680 ext4_clear_journal_err(sb, es);
ac27a0ec 4681 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4682 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4683 sb->s_flags &= ~MS_RDONLY;
e2b911c5 4684 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 4685 if (ext4_multi_mount_protect(sb,
4686 le64_to_cpu(es->s_mmp_block))) {
4687 err = -EROFS;
4688 goto restore_opts;
4689 }
c79d967d 4690 enable_quota = 1;
ac27a0ec
DK		 4691 }
4692 }
bfff6873
LC		 4693
4694 /*
4695 * Reinitialize lazy itable initialization thread based on
4696 * current settings
4697 */
4698 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4699 ext4_unregister_li_request(sb);
4700 else {
4701 ext4_group_t first_not_zeroed;
4702 first_not_zeroed = ext4_has_uninit_itable(sb);
4703 ext4_register_li_request(sb, first_not_zeroed);
4704 }
4705
6fd058f7 4706 ext4_setup_system_zone(sb);
d096ad0f 4707 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4708 ext4_commit_super(sb, 1);
0390131b 4709
ac27a0ec
DK		 4710#ifdef CONFIG_QUOTA
4711 /* Release old quota file names */
a2d4a646 4712 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 4713 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4714 if (enable_quota) {
4715 if (sb_any_quota_suspended(sb))
4716 dquot_resume(sb, -1);
e2b911c5 4717 else if (ext4_has_feature_quota(sb)) {
7c319d32 4718 err = ext4_enable_quotas(sb);
07724f98 4719 if (err)
7c319d32 4720 goto restore_opts;
7c319d32
AK		 4721 }
4722 }
ac27a0ec 4723#endif
d4c402d9 4724
a26f4992 4725 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 4726 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4727 kfree(orig_data);
ac27a0ec 4728 return 0;
0b8e58a1 4729
ac27a0ec
DK		 4730restore_opts:
4731 sb->s_flags = old_sb_flags;
4732 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4733 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4734 sbi->s_resuid = old_opts.s_resuid;
4735 sbi->s_resgid = old_opts.s_resgid;
4736 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4737 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4738 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4739#ifdef CONFIG_QUOTA
4740 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 4741 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 4742 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4743 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4744 }
4745#endif
d4c402d9 4746 kfree(orig_data);
ac27a0ec
DK		 4747 return err;
4748}
4749
2b2d6d01 4750static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4751{
4752 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4753 struct ext4_sb_info *sbi = EXT4_SB(sb);
4754 struct ext4_super_block *es = sbi->s_es;
27dd4385 4755 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 4756 u64 fsid;
d02a9391 4757 s64 bfree;
27dd4385 4758 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 4759
952fc18e
TT		 4760 if (!test_opt(sb, MINIX_DF))
4761 overhead = sbi->s_overhead;
ac27a0ec 4762
617ba13b 4763 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4764 buf->f_bsize = sb->s_blocksize;
b72f78cb 4765 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4766 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4767 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4768 /* prevent underflow in case that few free space is available */
57042651 4769 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 4770 buf->f_bavail = buf->f_bfree -
4771 (ext4_r_blocks_count(es) + resv_blocks);
4772 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 4773 buf->f_bavail = 0;
4774 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4775 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4776 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4777 fsid = le64_to_cpup((void *)es->s_uuid) ^
4778 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4779 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4780 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4781
ac27a0ec
DK		 4782 return 0;
4783}
4784
0b8e58a1
AD		 4785/* Helper function for writing quotas on sync - we need to start transaction
4786 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4787 * Process 1 Process 2
617ba13b 4788 * ext4_create() quota_sync()
a269eb18 4789 * jbd2_journal_start() write_dquot()
871a2931 4790 * dquot_initialize() down(dqio_mutex)
dab291af 4791 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4792 *
4793 */
4794
4795#ifdef CONFIG_QUOTA
4796
4797static inline struct inode *dquot_to_inode(struct dquot *dquot)
4798{
4c376dca 4799 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4800}
4801
617ba13b 4802static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4803{
4804 int ret, err;
4805 handle_t *handle;
4806 struct inode *inode;
4807
4808 inode = dquot_to_inode(dquot);
9924a92a 4809 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 4810 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4811 if (IS_ERR(handle))
4812 return PTR_ERR(handle);
4813 ret = dquot_commit(dquot);
617ba13b 4814 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4815 if (!ret)
4816 ret = err;
4817 return ret;
4818}
4819
617ba13b 4820static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4821{
4822 int ret, err;
4823 handle_t *handle;
4824
9924a92a 4825 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4826 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4827 if (IS_ERR(handle))
4828 return PTR_ERR(handle);
4829 ret = dquot_acquire(dquot);
617ba13b 4830 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4831 if (!ret)
4832 ret = err;
4833 return ret;
4834}
4835
617ba13b 4836static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4837{
4838 int ret, err;
4839 handle_t *handle;
4840
9924a92a 4841 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4842 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4843 if (IS_ERR(handle)) {
4844 /* Release dquot anyway to avoid endless cycle in dqput() */
4845 dquot_release(dquot);
ac27a0ec 4846 return PTR_ERR(handle);
9c3013e9 4847 }
ac27a0ec 4848 ret = dquot_release(dquot);
617ba13b 4849 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4850 if (!ret)
4851 ret = err;
4852 return ret;
4853}
4854
617ba13b 4855static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4856{
262b4662
JK		 4857 struct super_block *sb = dquot->dq_sb;
4858 struct ext4_sb_info *sbi = EXT4_SB(sb);
4859
2c8be6b2 4860 /* Are we journaling quotas? */
e2b911c5 4861 if (ext4_has_feature_quota(sb) ||
262b4662 4862 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 4863 dquot_mark_dquot_dirty(dquot);
617ba13b 4864 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4865 } else {
4866 return dquot_mark_dquot_dirty(dquot);
4867 }
4868}
4869
617ba13b 4870static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4871{
4872 int ret, err;
4873 handle_t *handle;
4874
4875 /* Data block + inode block */
2b0143b5 4876 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 4877 if (IS_ERR(handle))
4878 return PTR_ERR(handle);
4879 ret = dquot_commit_info(sb, type);
617ba13b 4880 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4881 if (!ret)
4882 ret = err;
4883 return ret;
4884}
4885
4886/*
4887 * Turn on quotas during mount time - we need to find
4888 * the quota file and such...
4889 */
617ba13b 4890static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4891{
287a8095
CH		 4892 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4893 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4894}
4895
4896/*
4897 * Standard function to be called on quota_on
4898 */
617ba13b 4899static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4900 struct path *path)
ac27a0ec
DK		 4901{
4902 int err;
ac27a0ec
DK		 4903
4904 if (!test_opt(sb, QUOTA))
4905 return -EINVAL;
0623543b 4906
ac27a0ec 4907 /* Quotafile not on the same filesystem? */
d8c9584e 4908 if (path->dentry->d_sb != sb)
ac27a0ec 4909 return -EXDEV;
0623543b
JK		 4910 /* Journaling quota? */
4911 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4912 /* Quotafile not in fs root? */
f00c9e44 4913 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4914 ext4_msg(sb, KERN_WARNING,
4915 "Quota file not on filesystem root. "
4916 "Journaled quota will not work");
2b2d6d01 4917 }
0623543b
JK		 4918
4919 /*
4920 * When we journal data on quota file, we have to flush journal to see
4921 * all updates to the file when we bypass pagecache...
4922 */
0390131b 4923 if (EXT4_SB(sb)->s_journal &&
2b0143b5 4924 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 4925 /*
4926 * We don't need to lock updates but journal_flush() could
4927 * otherwise be livelocked...
4928 */
4929 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4930 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4931 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4932 if (err)
7ffe1ea8 4933 return err;
0623543b
JK		 4934 }
4935
f00c9e44 4936 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4937}
4938
7c319d32
AK		 4939static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4940 unsigned int flags)
4941{
4942 int err;
4943 struct inode *qf_inode;
a2d4a646 4944 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 4945 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4946 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4947 };
4948
e2b911c5 4949 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 4950
4951 if (!qf_inums[type])
4952 return -EPERM;
4953
4954 qf_inode = ext4_iget(sb, qf_inums[type]);
4955 if (IS_ERR(qf_inode)) {
4956 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4957 return PTR_ERR(qf_inode);
4958 }
4959
bcb13850
JK		 4960 /* Don't account quota for quota files to avoid recursion */
4961 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 4962 err = dquot_enable(qf_inode, type, format_id, flags);
4963 iput(qf_inode);
4964
4965 return err;
4966}
4967
4968/* Enable usage tracking for all quota types. */
4969static int ext4_enable_quotas(struct super_block *sb)
4970{
4971 int type, err = 0;
a2d4a646 4972 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 4973 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4974 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4975 };
4976
4977 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 4978 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 4979 if (qf_inums[type]) {
4980 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4981 DQUOT_USAGE_ENABLED);
4982 if (err) {
4983 ext4_warning(sb,
72ba7450
TT		 4984 "Failed to enable quota tracking "
4985 "(type=%d, err=%d). Please run "
4986 "e2fsck to fix.", type, err);
7c319d32
AK		 4987 return err;
4988 }
4989 }
4990 }
4991 return 0;
4992}
4993
ca0e05e4
DM		 4994static int ext4_quota_off(struct super_block *sb, int type)
4995{
21f97697
JK		 4996 struct inode *inode = sb_dqopt(sb)->files[type];
4997 handle_t *handle;
4998
87009d86
DM		 4999 /* Force all delayed allocation blocks to be allocated.
5000 * Caller already holds s_umount sem */
5001 if (test_opt(sb, DELALLOC))
ca0e05e4 5002 sync_filesystem(sb);
ca0e05e4 5003
0b268590
AG		 5004 if (!inode)
5005 goto out;
5006
21f97697
JK		 5007 /* Update modification times of quota files when userspace can
5008 * start looking at them */
9924a92a 5009 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5010 if (IS_ERR(handle))
5011 goto out;
5012 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5013 ext4_mark_inode_dirty(handle, inode);
5014 ext4_journal_stop(handle);
5015
5016out:
ca0e05e4
DM		 5017 return dquot_quota_off(sb, type);
5018}
5019
ac27a0ec
DK		 5020/* Read data from quotafile - avoid pagecache and such because we cannot afford
5021 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5022 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5023 * we don't have to be afraid of races */
617ba13b 5024static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5025 size_t len, loff_t off)
5026{
5027 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5028 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5029 int offset = off & (sb->s_blocksize - 1);
5030 int tocopy;
5031 size_t toread;
5032 struct buffer_head *bh;
5033 loff_t i_size = i_size_read(inode);
5034
5035 if (off > i_size)
5036 return 0;
5037 if (off+len > i_size)
5038 len = i_size-off;
5039 toread = len;
5040 while (toread > 0) {
5041 tocopy = sb->s_blocksize - offset < toread ?
5042 sb->s_blocksize - offset : toread;
1c215028
TT		 5043 bh = ext4_bread(NULL, inode, blk, 0);
5044 if (IS_ERR(bh))
5045 return PTR_ERR(bh);
ac27a0ec
DK		 5046 if (!bh) /* A hole? */
5047 memset(data, 0, tocopy);
5048 else
5049 memcpy(data, bh->b_data+offset, tocopy);
5050 brelse(bh);
5051 offset = 0;
5052 toread -= tocopy;
5053 data += tocopy;
5054 blk++;
5055 }
5056 return len;
5057}
5058
5059/* Write to quotafile (we know the transaction is already started and has
5060 * enough credits) */
617ba13b 5061static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5062 const char *data, size_t len, loff_t off)
5063{
5064 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5065 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5066 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5067 int retries = 0;
ac27a0ec
DK		 5068 struct buffer_head *bh;
5069 handle_t *handle = journal_current_handle();
5070
0390131b 5071 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5072 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5073 " cancelled because transaction is not started",
9c3013e9
JK		 5074 (unsigned long long)off, (unsigned long long)len);
5075 return -EIO;
5076 }
67eeb568
DM		 5077 /*
5078 * Since we account only one data block in transaction credits,
5079 * then it is impossible to cross a block boundary.
5080 */
5081 if (sb->s_blocksize - offset < len) {
5082 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5083 " cancelled because not block aligned",
5084 (unsigned long long)off, (unsigned long long)len);
5085 return -EIO;
5086 }
5087
c5e298ae
TT		 5088 do {
5089 bh = ext4_bread(handle, inode, blk,
5090 EXT4_GET_BLOCKS_CREATE |
5091 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5092 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5093 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5094 if (IS_ERR(bh))
5095 return PTR_ERR(bh);
67eeb568
DM		 5096 if (!bh)
5097 goto out;
5d601255 5098 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5099 err = ext4_journal_get_write_access(handle, bh);
5100 if (err) {
5101 brelse(bh);
1c215028 5102 return err;
ac27a0ec 5103 }
67eeb568
DM		 5104 lock_buffer(bh);
5105 memcpy(bh->b_data+offset, data, len);
5106 flush_dcache_page(bh->b_page);
5107 unlock_buffer(bh);
62d2b5f2 5108 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5109 brelse(bh);
ac27a0ec 5110out:
67eeb568
DM		 5111 if (inode->i_size < off + len) {
5112 i_size_write(inode, off + len);
617ba13b 5113 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5114 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5115 }
67eeb568 5116 return len;
ac27a0ec
DK		 5117}
5118
5119#endif
5120
152a0836
AV		 5121static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5122 const char *dev_name, void *data)
ac27a0ec 5123{
152a0836 5124 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5125}
5126
c290ea01 5127#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5128static inline void register_as_ext2(void)
5129{
5130 int err = register_filesystem(&ext2_fs_type);
5131 if (err)
5132 printk(KERN_WARNING
5133 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5134}
5135
5136static inline void unregister_as_ext2(void)
5137{
5138 unregister_filesystem(&ext2_fs_type);
5139}
2035e776
TT		 5140
5141static inline int ext2_feature_set_ok(struct super_block *sb)
5142{
e2b911c5 5143 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776
TT		 5144 return 0;
5145 if (sb->s_flags & MS_RDONLY)
5146 return 1;
e2b911c5 5147 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5148 return 0;
5149 return 1;
5150}
24b58424
TT		 5151#else
5152static inline void register_as_ext2(void) { }
5153static inline void unregister_as_ext2(void) { }
2035e776 5154static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5155#endif
5156
24b58424
TT		 5157static inline void register_as_ext3(void)
5158{
5159 int err = register_filesystem(&ext3_fs_type);
5160 if (err)
5161 printk(KERN_WARNING
5162 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5163}
5164
5165static inline void unregister_as_ext3(void)
5166{
5167 unregister_filesystem(&ext3_fs_type);
5168}
2035e776
TT		 5169
5170static inline int ext3_feature_set_ok(struct super_block *sb)
5171{
e2b911c5 5172 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5173 return 0;
e2b911c5 5174 if (!ext4_has_feature_journal(sb))
2035e776
TT		 5175 return 0;
5176 if (sb->s_flags & MS_RDONLY)
5177 return 1;
e2b911c5 5178 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5179 return 0;
5180 return 1;
5181}
24b58424 5182
03010a33
TT		 5183static struct file_system_type ext4_fs_type = {
5184 .owner = THIS_MODULE,
5185 .name = "ext4",
152a0836 5186 .mount = ext4_mount,
03010a33
TT		 5187 .kill_sb = kill_block_super,
5188 .fs_flags = FS_REQUIRES_DEV,
5189};
7f78e035 5190MODULE_ALIAS_FS("ext4");
03010a33 5191
e9e3bcec
ES		 5192/* Shared across all ext4 file systems */
5193wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5194struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5195
5dabfc78 5196static int __init ext4_init_fs(void)
ac27a0ec 5197{
e9e3bcec 5198 int i, err;
c9de560d 5199
e294a537 5200 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5201 ext4_li_info = NULL;
5202 mutex_init(&ext4_li_mtx);
5203
9a4c8019 5204 /* Build-time check for flags consistency */
12e9b892 5205 ext4_check_flag_values();
e9e3bcec
ES		 5206
5207 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5208 mutex_init(&ext4__aio_mutex[i]);
5209 init_waitqueue_head(&ext4__ioend_wq[i]);
5210 }
5211
51865fda 5212 err = ext4_init_es();
6fd058f7
TT		 5213 if (err)
5214 return err;
51865fda
ZL		 5215
5216 err = ext4_init_pageio();
5217 if (err)
b5799018 5218 goto out5;
51865fda 5219
5dabfc78 5220 err = ext4_init_system_zone();
bd2d0210 5221 if (err)
b5799018 5222 goto out4;
857ac889 5223
b5799018 5224 err = ext4_init_sysfs();
dd68314c 5225 if (err)
b5799018 5226 goto out3;
857ac889 5227
5dabfc78 5228 err = ext4_init_mballoc();
c9de560d
AT		 5229 if (err)
5230 goto out2;
9c191f70
M		 5231 else
5232 ext4_mballoc_ready = 1;
ac27a0ec
DK		 5233 err = init_inodecache();
5234 if (err)
5235 goto out1;
24b58424 5236 register_as_ext3();
2035e776 5237 register_as_ext2();
03010a33 5238 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5239 if (err)
5240 goto out;
bfff6873 5241
ac27a0ec
DK		 5242 return 0;
5243out:
24b58424
TT		 5244 unregister_as_ext2();
5245 unregister_as_ext3();
ac27a0ec
DK		 5246 destroy_inodecache();
5247out1:
9c191f70 5248 ext4_mballoc_ready = 0;
5dabfc78 5249 ext4_exit_mballoc();
9c191f70 5250out2:
b5799018
TT		 5251 ext4_exit_sysfs();
5252out3:
5dabfc78 5253 ext4_exit_system_zone();
b5799018 5254out4:
5dabfc78 5255 ext4_exit_pageio();
b5799018 5256out5:
51865fda
ZL		 5257 ext4_exit_es();
5258
ac27a0ec
DK		 5259 return err;
5260}
5261
5dabfc78 5262static void __exit ext4_exit_fs(void)
ac27a0ec 5263{
e298e73b 5264 ext4_exit_crypto();
bfff6873 5265 ext4_destroy_lazyinit_thread();
24b58424
TT		 5266 unregister_as_ext2();
5267 unregister_as_ext3();
03010a33 5268 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5269 destroy_inodecache();
5dabfc78 5270 ext4_exit_mballoc();
b5799018 5271 ext4_exit_sysfs();
5dabfc78
TT		 5272 ext4_exit_system_zone();
5273 ext4_exit_pageio();
dd12ed14 5274 ext4_exit_es();
ac27a0ec
DK		 5275}
5276
5277MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5278MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5279MODULE_LICENSE("GPL");
5dabfc78
TT		 5280module_init(ext4_init_fs)
5281module_exit(ext4_exit_fs)