]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-zesty-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
UBUNTU: SAUCE: (namespace) ext4: Add module parameter to enable user namespace mounts
[mirror_ubuntu-zesty-kernel.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
66114cad 27#include <linux/backing-dev.h>
ac27a0ec 28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
3197ebdb 37#include <linux/ctype.h>
1330593e 38#include <linux/log2.h>
717d50e4 39#include <linux/crc16.h>
7abc52c2 40#include <linux/cleancache.h>
7c0f6ba6 41#include <linux/uaccess.h>
acc94812 42#include <linux/user_namespace.h>
ac27a0ec 43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
4a092d73 48#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
0b75a840
LC		 57static struct ext4_lazy_init *ext4_li_info;
58static struct mutex ext4_li_mtx;
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);
c6cb7e77
EW		 82static struct inode *ext4_get_journal_inode(struct super_block *sb,
83 unsigned int journal_inum);
ac27a0ec 84
e74031fd
JK		 85/*
86 * Lock ordering
87 *
88 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
89 * i_mmap_rwsem (inode->i_mmap_rwsem)!
90 *
91 * page fault path:
92 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
93 * page lock -> i_data_sem (rw)
94 *
95 * buffered write path:
96 * sb_start_write -> i_mutex -> mmap_sem
97 * sb_start_write -> i_mutex -> transaction start -> page lock ->
98 * i_data_sem (rw)
99 *
100 * truncate:
101 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
102 * i_mmap_rwsem (w) -> page lock
103 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
104 * transaction start -> i_data_sem (rw)
105 *
106 * direct IO:
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
108 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
109 * transaction start -> i_data_sem (rw)
110 *
111 * writepages:
112 * transaction start -> page lock(s) -> i_data_sem (rw)
113 */
114
73eb7eae
SF		 115static bool userns_mounts = false;
116module_param(userns_mounts, bool, 0644);
117MODULE_PARM_DESC(userns_mounts, "Allow mounts from unprivileged user namespaces");
118
c290ea01 119#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 120static struct file_system_type ext2_fs_type = {
121 .owner = THIS_MODULE,
122 .name = "ext2",
123 .mount = ext4_mount,
124 .kill_sb = kill_block_super,
acc94812 125 .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT,
2035e776 126};
7f78e035 127MODULE_ALIAS_FS("ext2");
fa7614dd 128MODULE_ALIAS("ext2");
2035e776
TT		 129#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
130#else
131#define IS_EXT2_SB(sb) (0)
132#endif
133
134
ba69f9ab
JK		 135static struct file_system_type ext3_fs_type = {
136 .owner = THIS_MODULE,
137 .name = "ext3",
152a0836 138 .mount = ext4_mount,
ba69f9ab 139 .kill_sb = kill_block_super,
acc94812 140 .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT,
ba69f9ab 141};
7f78e035 142MODULE_ALIAS_FS("ext3");
fa7614dd 143MODULE_ALIAS("ext3");
ba69f9ab 144#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 145
d25425f8
DW		 146static int ext4_verify_csum_type(struct super_block *sb,
147 struct ext4_super_block *es)
148{
e2b911c5 149 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 150 return 1;
151
152 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
153}
154
a9c47317
DW		 155static __le32 ext4_superblock_csum(struct super_block *sb,
156 struct ext4_super_block *es)
157{
158 struct ext4_sb_info *sbi = EXT4_SB(sb);
159 int offset = offsetof(struct ext4_super_block, s_checksum);
160 __u32 csum;
161
162 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
163
164 return cpu_to_le32(csum);
165}
166
c197855e
SH		 167static int ext4_superblock_csum_verify(struct super_block *sb,
168 struct ext4_super_block *es)
a9c47317 169{
9aa5d32b 170 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 171 return 1;
172
173 return es->s_checksum == ext4_superblock_csum(sb, es);
174}
175
06db49e6 176void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 177{
06db49e6
TT		 178 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
179
9aa5d32b 180 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 181 return;
182
183 es->s_checksum = ext4_superblock_csum(sb, es);
184}
185
9933fc0a
TT		 186void *ext4_kvmalloc(size_t size, gfp_t flags)
187{
188 void *ret;
189
8be04b93 190 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 191 if (!ret)
192 ret = __vmalloc(size, flags, PAGE_KERNEL);
193 return ret;
194}
195
196void *ext4_kvzalloc(size_t size, gfp_t flags)
197{
198 void *ret;
199
8be04b93 200 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 201 if (!ret)
202 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
203 return ret;
204}
205
8fadc143
AR		 206ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
207 struct ext4_group_desc *bg)
bd81d8ee 208{
3a14589c 209 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 210 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 211 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 212}
213
8fadc143
AR		 214ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
215 struct ext4_group_desc *bg)
bd81d8ee 216{
5272f837 217 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 218 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 219 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 220}
221
8fadc143
AR		 222ext4_fsblk_t ext4_inode_table(struct super_block *sb,
223 struct ext4_group_desc *bg)
bd81d8ee 224{
5272f837 225 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 226 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 227 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 228}
229
021b65bb
TT		 230__u32 ext4_free_group_clusters(struct super_block *sb,
231 struct ext4_group_desc *bg)
560671a0
AK		 232{
233 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
234 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 235 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 236}
237
238__u32 ext4_free_inodes_count(struct super_block *sb,
239 struct ext4_group_desc *bg)
240{
241 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
242 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 243 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 244}
245
246__u32 ext4_used_dirs_count(struct super_block *sb,
247 struct ext4_group_desc *bg)
248{
249 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
250 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 251 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 252}
253
254__u32 ext4_itable_unused_count(struct super_block *sb,
255 struct ext4_group_desc *bg)
256{
257 return le16_to_cpu(bg->bg_itable_unused_lo) |
258 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 259 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 260}
261
8fadc143
AR		 262void ext4_block_bitmap_set(struct super_block *sb,
263 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 264{
3a14589c 265 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 266 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
267 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 268}
269
8fadc143
AR		 270void ext4_inode_bitmap_set(struct super_block *sb,
271 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 272{
5272f837 273 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 274 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
275 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 276}
277
8fadc143
AR		 278void ext4_inode_table_set(struct super_block *sb,
279 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 280{
5272f837 281 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 282 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
283 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 284}
285
021b65bb
TT		 286void ext4_free_group_clusters_set(struct super_block *sb,
287 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 288{
289 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
290 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
291 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
292}
293
294void ext4_free_inodes_set(struct super_block *sb,
295 struct ext4_group_desc *bg, __u32 count)
296{
297 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
298 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
299 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
300}
301
302void ext4_used_dirs_set(struct super_block *sb,
303 struct ext4_group_desc *bg, __u32 count)
304{
305 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
306 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
307 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
308}
309
310void ext4_itable_unused_set(struct super_block *sb,
311 struct ext4_group_desc *bg, __u32 count)
312{
313 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
314 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
315 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
316}
317
d3d1faf6 318
1c13d5c0
TT		 319static void __save_error_info(struct super_block *sb, const char *func,
320 unsigned int line)
321{
322 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
323
324 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 325 if (bdev_read_only(sb->s_bdev))
326 return;
1c13d5c0
TT		 327 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
328 es->s_last_error_time = cpu_to_le32(get_seconds());
329 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
330 es->s_last_error_line = cpu_to_le32(line);
331 if (!es->s_first_error_time) {
332 es->s_first_error_time = es->s_last_error_time;
333 strncpy(es->s_first_error_func, func,
334 sizeof(es->s_first_error_func));
335 es->s_first_error_line = cpu_to_le32(line);
336 es->s_first_error_ino = es->s_last_error_ino;
337 es->s_first_error_block = es->s_last_error_block;
338 }
66e61a9e
TT		 339 /*
340 * Start the daily error reporting function if it hasn't been
341 * started already
342 */
343 if (!es->s_error_count)
344 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 345 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 346}
347
348static void save_error_info(struct super_block *sb, const char *func,
349 unsigned int line)
350{
351 __save_error_info(sb, func, line);
352 ext4_commit_super(sb, 1);
353}
354
bdfe0cbd
TT		 355/*
356 * The del_gendisk() function uninitializes the disk-specific data
357 * structures, including the bdi structure, without telling anyone
358 * else. Once this happens, any attempt to call mark_buffer_dirty()
359 * (for example, by ext4_commit_super), will cause a kernel OOPS.
360 * This is a kludge to prevent these oops until we can put in a proper
361 * hook in del_gendisk() to inform the VFS and file system layers.
362 */
363static int block_device_ejected(struct super_block *sb)
364{
365 struct inode *bd_inode = sb->s_bdev->bd_inode;
366 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
367
368 return bdi->dev == NULL;
369}
370
18aadd47
BJ		 371static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
372{
373 struct super_block *sb = journal->j_private;
374 struct ext4_sb_info *sbi = EXT4_SB(sb);
375 int error = is_journal_aborted(journal);
5d3ee208 376 struct ext4_journal_cb_entry *jce;
18aadd47 377
5d3ee208 378 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 379 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 380 while (!list_empty(&txn->t_private_list)) {
381 jce = list_entry(txn->t_private_list.next,
382 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 383 list_del_init(&jce->jce_list);
384 spin_unlock(&sbi->s_md_lock);
385 jce->jce_func(sb, jce, error);
386 spin_lock(&sbi->s_md_lock);
387 }
388 spin_unlock(&sbi->s_md_lock);
389}
1c13d5c0 390
ac27a0ec
DK		 391/* Deal with the reporting of failure conditions on a filesystem such as
392 * inconsistencies detected or read IO failures.
393 *
394 * On ext2, we can store the error state of the filesystem in the
617ba13b 395 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 396 * write ordering constraints on the superblock which prevent us from
397 * writing it out straight away; and given that the journal is about to
398 * be aborted, we can't rely on the current, or future, transactions to
399 * write out the superblock safely.
400 *
dab291af 401 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 402 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 403 * that error until we've noted it down and cleared it.
404 */
405
617ba13b 406static void ext4_handle_error(struct super_block *sb)
ac27a0ec 407{
ac27a0ec
DK		 408 if (sb->s_flags & MS_RDONLY)
409 return;
410
2b2d6d01 411 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 412 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 413
4ab2f15b 414 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 415 if (journal)
dab291af 416 jbd2_journal_abort(journal, -EIO);
ac27a0ec 417 }
2b2d6d01 418 if (test_opt(sb, ERRORS_RO)) {
b31e1552 419 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 420 /*
421 * Make sure updated value of ->s_mount_flags will be visible
422 * before ->s_flags update
423 */
424 smp_wmb();
ac27a0ec
DK		 425 sb->s_flags |= MS_RDONLY;
426 }
4327ba52
DJ		 427 if (test_opt(sb, ERRORS_PANIC)) {
428 if (EXT4_SB(sb)->s_journal &&
429 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
430 return;
617ba13b 431 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 432 sb->s_id);
4327ba52 433 }
ac27a0ec
DK		 434}
435
efbed4dc
TT		 436#define ext4_error_ratelimit(sb) \
437 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
438 "EXT4-fs error")
439
12062ddd 440void __ext4_error(struct super_block *sb, const char *function,
c398eda0 441 unsigned int line, const char *fmt, ...)
ac27a0ec 442{
0ff2ea7d 443 struct va_format vaf;
ac27a0ec
DK		 444 va_list args;
445
efbed4dc
TT		 446 if (ext4_error_ratelimit(sb)) {
447 va_start(args, fmt);
448 vaf.fmt = fmt;
449 vaf.va = &args;
450 printk(KERN_CRIT
451 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
452 sb->s_id, function, line, current->comm, &vaf);
453 va_end(args);
454 }
f3fc0210 455 save_error_info(sb, function, line);
617ba13b 456 ext4_handle_error(sb);
ac27a0ec
DK		 457}
458
e7c96e8e
JP		 459void __ext4_error_inode(struct inode *inode, const char *function,
460 unsigned int line, ext4_fsblk_t block,
461 const char *fmt, ...)
273df556
FM		 462{
463 va_list args;
f7c21177 464 struct va_format vaf;
1c13d5c0 465 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 466
1c13d5c0
TT		 467 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
468 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 469 if (ext4_error_ratelimit(inode->i_sb)) {
470 va_start(args, fmt);
471 vaf.fmt = fmt;
472 vaf.va = &args;
473 if (block)
474 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
475 "inode #%lu: block %llu: comm %s: %pV\n",
476 inode->i_sb->s_id, function, line, inode->i_ino,
477 block, current->comm, &vaf);
478 else
479 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
480 "inode #%lu: comm %s: %pV\n",
481 inode->i_sb->s_id, function, line, inode->i_ino,
482 current->comm, &vaf);
483 va_end(args);
484 }
1c13d5c0 485 save_error_info(inode->i_sb, function, line);
273df556
FM		 486 ext4_handle_error(inode->i_sb);
487}
488
e7c96e8e
JP		 489void __ext4_error_file(struct file *file, const char *function,
490 unsigned int line, ext4_fsblk_t block,
491 const char *fmt, ...)
273df556
FM		 492{
493 va_list args;
f7c21177 494 struct va_format vaf;
1c13d5c0 495 struct ext4_super_block *es;
496ad9aa 496 struct inode *inode = file_inode(file);
273df556
FM		 497 char pathname[80], *path;
498
1c13d5c0
TT		 499 es = EXT4_SB(inode->i_sb)->s_es;
500 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 501 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 502 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 503 if (IS_ERR(path))
504 path = "(unknown)";
505 va_start(args, fmt);
506 vaf.fmt = fmt;
507 vaf.va = &args;
508 if (block)
509 printk(KERN_CRIT
510 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
511 "block %llu: comm %s: path %s: %pV\n",
512 inode->i_sb->s_id, function, line, inode->i_ino,
513 block, current->comm, path, &vaf);
514 else
515 printk(KERN_CRIT
516 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
517 "comm %s: path %s: %pV\n",
518 inode->i_sb->s_id, function, line, inode->i_ino,
519 current->comm, path, &vaf);
520 va_end(args);
521 }
1c13d5c0 522 save_error_info(inode->i_sb, function, line);
273df556
FM		 523 ext4_handle_error(inode->i_sb);
524}
525
722887dd
TT		 526const char *ext4_decode_error(struct super_block *sb, int errno,
527 char nbuf[16])
ac27a0ec
DK		 528{
529 char *errstr = NULL;
530
531 switch (errno) {
6a797d27
DW		 532 case -EFSCORRUPTED:
533 errstr = "Corrupt filesystem";
534 break;
535 case -EFSBADCRC:
536 errstr = "Filesystem failed CRC";
537 break;
ac27a0ec
DK		 538 case -EIO:
539 errstr = "IO failure";
540 break;
541 case -ENOMEM:
542 errstr = "Out of memory";
543 break;
544 case -EROFS:
78f1ddbb
TT		 545 if (!sb || (EXT4_SB(sb)->s_journal &&
546 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 547 errstr = "Journal has aborted";
548 else
549 errstr = "Readonly filesystem";
550 break;
551 default:
552 /* If the caller passed in an extra buffer for unknown
553 * errors, textualise them now. Else we just return
554 * NULL. */
555 if (nbuf) {
556 /* Check for truncated error codes... */
557 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
558 errstr = nbuf;
559 }
560 break;
561 }
562
563 return errstr;
564}
565
617ba13b 566/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 567 * automatically and invokes the appropriate error response. */
568
c398eda0
TT		 569void __ext4_std_error(struct super_block *sb, const char *function,
570 unsigned int line, int errno)
ac27a0ec
DK		 571{
572 char nbuf[16];
573 const char *errstr;
574
575 /* Special case: if the error is EROFS, and we're not already
576 * inside a transaction, then there's really no point in logging
577 * an error. */
578 if (errno == -EROFS && journal_current_handle() == NULL &&
579 (sb->s_flags & MS_RDONLY))
580 return;
581
efbed4dc
TT		 582 if (ext4_error_ratelimit(sb)) {
583 errstr = ext4_decode_error(sb, errno, nbuf);
584 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
585 sb->s_id, function, line, errstr);
586 }
ac27a0ec 587
efbed4dc 588 save_error_info(sb, function, line);
617ba13b 589 ext4_handle_error(sb);
ac27a0ec
DK		 590}
591
592/*
617ba13b 593 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 594 * abort function may be used to deal with unrecoverable failures such
595 * as journal IO errors or ENOMEM at a critical moment in log management.
596 *
597 * We unconditionally force the filesystem into an ABORT|READONLY state,
598 * unless the error response on the fs has been set to panic in which
599 * case we take the easy way out and panic immediately.
600 */
601
c67d859e 602void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 603 unsigned int line, const char *fmt, ...)
ac27a0ec 604{
651e1c3b 605 struct va_format vaf;
ac27a0ec
DK		 606 va_list args;
607
1c13d5c0 608 save_error_info(sb, function, line);
ac27a0ec 609 va_start(args, fmt);
651e1c3b
JP		 610 vaf.fmt = fmt;
611 vaf.va = &args;
612 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
613 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 614 va_end(args);
615
1c13d5c0
TT		 616 if ((sb->s_flags & MS_RDONLY) == 0) {
617 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 618 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 619 /*
620 * Make sure updated value of ->s_mount_flags will be visible
621 * before ->s_flags update
622 */
623 smp_wmb();
624 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 625 if (EXT4_SB(sb)->s_journal)
626 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
627 save_error_info(sb, function, line);
628 }
4327ba52
DJ		 629 if (test_opt(sb, ERRORS_PANIC)) {
630 if (EXT4_SB(sb)->s_journal &&
631 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
632 return;
617ba13b 633 panic("EXT4-fs panic from previous error\n");
4327ba52 634 }
ac27a0ec
DK		 635}
636
e7c96e8e
JP		 637void __ext4_msg(struct super_block *sb,
638 const char *prefix, const char *fmt, ...)
b31e1552 639{
0ff2ea7d 640 struct va_format vaf;
b31e1552
ES		 641 va_list args;
642
efbed4dc
TT		 643 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
644 return;
645
b31e1552 646 va_start(args, fmt);
0ff2ea7d
JP		 647 vaf.fmt = fmt;
648 vaf.va = &args;
649 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 650 va_end(args);
651}
652
b03a2f7e
AD		 653#define ext4_warning_ratelimit(sb) \
654 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
655 "EXT4-fs warning")
656
12062ddd 657void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 658 unsigned int line, const char *fmt, ...)
ac27a0ec 659{
0ff2ea7d 660 struct va_format vaf;
ac27a0ec
DK		 661 va_list args;
662
b03a2f7e 663 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 664 return;
665
ac27a0ec 666 va_start(args, fmt);
0ff2ea7d
JP		 667 vaf.fmt = fmt;
668 vaf.va = &args;
669 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
670 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 671 va_end(args);
672}
673
b03a2f7e
AD		 674void __ext4_warning_inode(const struct inode *inode, const char *function,
675 unsigned int line, const char *fmt, ...)
676{
677 struct va_format vaf;
678 va_list args;
679
680 if (!ext4_warning_ratelimit(inode->i_sb))
681 return;
682
683 va_start(args, fmt);
684 vaf.fmt = fmt;
685 vaf.va = &args;
686 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
687 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
688 function, line, inode->i_ino, current->comm, &vaf);
689 va_end(args);
690}
691
e29136f8
TT		 692void __ext4_grp_locked_error(const char *function, unsigned int line,
693 struct super_block *sb, ext4_group_t grp,
694 unsigned long ino, ext4_fsblk_t block,
695 const char *fmt, ...)
5d1b1b3f
AK		 696__releases(bitlock)
697__acquires(bitlock)
698{
0ff2ea7d 699 struct va_format vaf;
5d1b1b3f
AK		 700 va_list args;
701 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
702
1c13d5c0
TT		 703 es->s_last_error_ino = cpu_to_le32(ino);
704 es->s_last_error_block = cpu_to_le64(block);
705 __save_error_info(sb, function, line);
0ff2ea7d 706
efbed4dc
TT		 707 if (ext4_error_ratelimit(sb)) {
708 va_start(args, fmt);
709 vaf.fmt = fmt;
710 vaf.va = &args;
711 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
712 sb->s_id, function, line, grp);
713 if (ino)
714 printk(KERN_CONT "inode %lu: ", ino);
715 if (block)
716 printk(KERN_CONT "block %llu:",
717 (unsigned long long) block);
718 printk(KERN_CONT "%pV\n", &vaf);
719 va_end(args);
720 }
5d1b1b3f
AK		 721
722 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 723 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 724 return;
725 }
1c13d5c0 726
5d1b1b3f
AK		 727 ext4_unlock_group(sb, grp);
728 ext4_handle_error(sb);
729 /*
730 * We only get here in the ERRORS_RO case; relocking the group
731 * may be dangerous, but nothing bad will happen since the
732 * filesystem will have already been marked read/only and the
733 * journal has been aborted. We return 1 as a hint to callers
734 * who might what to use the return value from
25985edc 735 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 736 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
737 * aggressively from the ext4 function in question, with a
738 * more appropriate error code.
739 */
740 ext4_lock_group(sb, grp);
741 return;
742}
743
617ba13b 744void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 745{
617ba13b 746 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 747
617ba13b 748 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 749 return;
750
12062ddd 751 ext4_warning(sb,
ac27a0ec
DK		 752 "updating to rev %d because of new feature flag, "
753 "running e2fsck is recommended",
617ba13b 754 EXT4_DYNAMIC_REV);
ac27a0ec 755
617ba13b
MC		 756 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
757 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
758 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 759 /* leave es->s_feature_*compat flags alone */
760 /* es->s_uuid will be set by e2fsck if empty */
761
762 /*
763 * The rest of the superblock fields should be zero, and if not it
764 * means they are likely already in use, so leave them alone. We
765 * can leave it up to e2fsck to clean up any inconsistencies there.
766 */
767}
768
769/*
770 * Open the external journal device
771 */
b31e1552 772static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 773{
774 struct block_device *bdev;
775 char b[BDEVNAME_SIZE];
776
d4d77629 777 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 778 if (IS_ERR(bdev))
779 goto fail;
780 return bdev;
781
782fail:
b31e1552 783 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 784 __bdevname(dev, b), PTR_ERR(bdev));
785 return NULL;
786}
787
788/*
789 * Release the journal device
790 */
4385bab1 791static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 792{
4385bab1 793 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 794}
795
4385bab1 796static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 797{
798 struct block_device *bdev;
ac27a0ec
DK		 799 bdev = sbi->journal_bdev;
800 if (bdev) {
4385bab1 801 ext4_blkdev_put(bdev);
ac27a0ec
DK		 802 sbi->journal_bdev = NULL;
803 }
ac27a0ec
DK		 804}
805
806static inline struct inode *orphan_list_entry(struct list_head *l)
807{
617ba13b 808 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 809}
810
617ba13b 811static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 812{
813 struct list_head *l;
814
b31e1552
ES		 815 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
816 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 817
818 printk(KERN_ERR "sb_info orphan list:\n");
819 list_for_each(l, &sbi->s_orphan) {
820 struct inode *inode = orphan_list_entry(l);
821 printk(KERN_ERR " "
822 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
823 inode->i_sb->s_id, inode->i_ino, inode,
824 inode->i_mode, inode->i_nlink,
825 NEXT_ORPHAN(inode));
826 }
827}
828
2b2d6d01 829static void ext4_put_super(struct super_block *sb)
ac27a0ec 830{
617ba13b
MC		 831 struct ext4_sb_info *sbi = EXT4_SB(sb);
832 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 833 int i, err;
ac27a0ec 834
857ac889 835 ext4_unregister_li_request(sb);
e0ccfd95
CH		 836 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
837
2e8fa54e 838 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 839 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 840
0390131b
FM		 841 if (sbi->s_journal) {
842 err = jbd2_journal_destroy(sbi->s_journal);
843 sbi->s_journal = NULL;
844 if (err < 0)
c67d859e 845 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 846 }
d4edac31 847
ebd173be 848 ext4_unregister_sysfs(sb);
d3922a77 849 ext4_es_unregister_shrinker(sbi);
9105bb14 850 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 851 ext4_release_system_zone(sb);
852 ext4_mb_release(sb);
853 ext4_ext_release(sb);
d4edac31 854
ac27a0ec 855 if (!(sb->s_flags & MS_RDONLY)) {
e2b911c5 856 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 857 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 858 }
58c5873a 859 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 860 ext4_commit_super(sb, 1);
861
ac27a0ec
DK		 862 for (i = 0; i < sbi->s_gdb_count; i++)
863 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 864 kvfree(sbi->s_group_desc);
865 kvfree(sbi->s_flex_groups);
57042651 866 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 867 percpu_counter_destroy(&sbi->s_freeinodes_counter);
868 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 869 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c8585c6f 870 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec 871#ifdef CONFIG_QUOTA
a2d4a646 872 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 873 kfree(sbi->s_qf_names[i]);
874#endif
875
876 /* Debugging code just in case the in-memory inode orphan list
877 * isn't empty. The on-disk one can be non-empty if we've
878 * detected an error and taken the fs readonly, but the
879 * in-memory list had better be clean by this point. */
880 if (!list_empty(&sbi->s_orphan))
881 dump_orphan_list(sb, sbi);
882 J_ASSERT(list_empty(&sbi->s_orphan));
883
89d96a6f 884 sync_blockdev(sb->s_bdev);
f98393a6 885 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 886 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
887 /*
888 * Invalidate the journal device's buffers. We don't want them
889 * floating about in memory - the physical journal device may
890 * hotswapped, and it breaks the `ro-after' testing code.
891 */
892 sync_blockdev(sbi->journal_bdev);
f98393a6 893 invalidate_bdev(sbi->journal_bdev);
617ba13b 894 ext4_blkdev_remove(sbi);
ac27a0ec 895 }
9c191f70
M		 896 if (sbi->s_mb_cache) {
897 ext4_xattr_destroy_cache(sbi->s_mb_cache);
898 sbi->s_mb_cache = NULL;
899 }
c5e06d10
JL		 900 if (sbi->s_mmp_tsk)
901 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 902 brelse(sbi->s_sbh);
ac27a0ec 903 sb->s_fs_info = NULL;
3197ebdb
TT		 904 /*
905 * Now that we are completely done shutting down the
906 * superblock, we need to actually destroy the kobject.
907 */
3197ebdb
TT		 908 kobject_put(&sbi->s_kobj);
909 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 910 if (sbi->s_chksum_driver)
911 crypto_free_shash(sbi->s_chksum_driver);
705895b6 912 kfree(sbi->s_blockgroup_lock);
ac27a0ec 913 kfree(sbi);
ac27a0ec
DK		 914}
915
e18b890b 916static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 917
918/*
919 * Called inside transaction, so use GFP_NOFS
920 */
617ba13b 921static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 922{
617ba13b 923 struct ext4_inode_info *ei;
ac27a0ec 924
e6b4f8da 925 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 926 if (!ei)
927 return NULL;
0b8e58a1 928
ac27a0ec 929 ei->vfs_inode.i_version = 1;
202ee5df 930 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 931 INIT_LIST_HEAD(&ei->i_prealloc_list);
932 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 933 ext4_es_init_tree(&ei->i_es_tree);
934 rwlock_init(&ei->i_es_lock);
edaa53ca 935 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 936 ei->i_es_all_nr = 0;
edaa53ca 937 ei->i_es_shk_nr = 0;
dd475925 938 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 939 ei->i_reserved_data_blocks = 0;
940 ei->i_reserved_meta_blocks = 0;
941 ei->i_allocated_meta_blocks = 0;
9d0be502 942 ei->i_da_metadata_calc_len = 0;
7e731bc9 943 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 944 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 945#ifdef CONFIG_QUOTA
946 ei->i_reserved_quota = 0;
96c7e0d9 947 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 948#endif
8aefcd55 949 ei->jinode = NULL;
2e8fa54e 950 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 951 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 952 ei->i_sync_tid = 0;
953 ei->i_datasync_tid = 0;
e27f41e1 954 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 955 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 956 return &ei->vfs_inode;
957}
958
7ff9c073
TT		 959static int ext4_drop_inode(struct inode *inode)
960{
961 int drop = generic_drop_inode(inode);
962
963 trace_ext4_drop_inode(inode, drop);
964 return drop;
965}
966
fa0d7e3d
NP		 967static void ext4_i_callback(struct rcu_head *head)
968{
969 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 970 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
971}
972
617ba13b 973static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 974{
9f7dd93d 975 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 976 ext4_msg(inode->i_sb, KERN_ERR,
977 "Inode %lu (%p): orphan list check failed!",
978 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 979 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
980 EXT4_I(inode), sizeof(struct ext4_inode_info),
981 true);
982 dump_stack();
983 }
fa0d7e3d 984 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 985}
986
51cc5068 987static void init_once(void *foo)
ac27a0ec 988{
617ba13b 989 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 990
a35afb83 991 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 992 init_rwsem(&ei->xattr_sem);
0e855ac8 993 init_rwsem(&ei->i_data_sem);
ea3d7209 994 init_rwsem(&ei->i_mmap_sem);
a35afb83 995 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 996}
997
e67bc2b3 998static int __init init_inodecache(void)
ac27a0ec 999{
617ba13b
MC		 1000 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1001 sizeof(struct ext4_inode_info),
ac27a0ec 1002 0, (SLAB_RECLAIM_ACCOUNT|
5d097056 1003 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
20c2df83 1004 init_once);
617ba13b 1005 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1006 return -ENOMEM;
1007 return 0;
1008}
1009
1010static void destroy_inodecache(void)
1011{
8c0a8537
KS		 1012 /*
1013 * Make sure all delayed rcu free inodes are flushed before we
1014 * destroy cache.
1015 */
1016 rcu_barrier();
617ba13b 1017 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1018}
1019
0930fcc1 1020void ext4_clear_inode(struct inode *inode)
ac27a0ec 1021{
0930fcc1 1022 invalidate_inode_buffers(inode);
dbd5768f 1023 clear_inode(inode);
9f754758 1024 dquot_drop(inode);
c2ea3fde 1025 ext4_discard_preallocations(inode);
51865fda 1026 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1027 if (EXT4_I(inode)->jinode) {
1028 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1029 EXT4_I(inode)->jinode);
1030 jbd2_free_inode(EXT4_I(inode)->jinode);
1031 EXT4_I(inode)->jinode = NULL;
1032 }
b7236e21 1033#ifdef CONFIG_EXT4_FS_ENCRYPTION
a7550b30 1034 fscrypt_put_encryption_info(inode, NULL);
b7236e21 1035#endif
ac27a0ec
DK		 1036}
1037
1b961ac0 1038static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1039 u64 ino, u32 generation)
ac27a0ec 1040{
ac27a0ec 1041 struct inode *inode;
ac27a0ec 1042
617ba13b 1043 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1044 return ERR_PTR(-ESTALE);
617ba13b 1045 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1046 return ERR_PTR(-ESTALE);
1047
1048 /* iget isn't really right if the inode is currently unallocated!!
1049 *
617ba13b 1050 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1051 * deleted, so we should be safe.
1052 *
1053 * Currently we don't know the generation for parent directory, so
1054 * a generation of 0 means "accept any"
1055 */
f4bb2981 1056 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1057 if (IS_ERR(inode))
1058 return ERR_CAST(inode);
1059 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1060 iput(inode);
1061 return ERR_PTR(-ESTALE);
1062 }
1b961ac0
CH		 1063
1064 return inode;
1065}
1066
1067static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1068 int fh_len, int fh_type)
1b961ac0
CH		 1069{
1070 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1071 ext4_nfs_get_inode);
1072}
1073
1074static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1075 int fh_len, int fh_type)
1b961ac0
CH		 1076{
1077 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1078 ext4_nfs_get_inode);
ac27a0ec
DK		 1079}
1080
c39a7f84
TO		 1081/*
1082 * Try to release metadata pages (indirect blocks, directories) which are
1083 * mapped via the block device. Since these pages could have journal heads
1084 * which would prevent try_to_free_buffers() from freeing them, we must use
1085 * jbd2 layer's try_to_free_buffers() function to release them.
1086 */
0b8e58a1
AD		 1087static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1088 gfp_t wait)
c39a7f84
TO		 1089{
1090 journal_t *journal = EXT4_SB(sb)->s_journal;
1091
1092 WARN_ON(PageChecked(page));
1093 if (!page_has_buffers(page))
1094 return 0;
1095 if (journal)
1096 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1097 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1098 return try_to_free_buffers(page);
1099}
1100
a7550b30
JK		 1101#ifdef CONFIG_EXT4_FS_ENCRYPTION
1102static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1103{
1104 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1105 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1106}
1107
1108static int ext4_key_prefix(struct inode *inode, u8 **key)
1109{
1110 *key = EXT4_SB(inode->i_sb)->key_prefix;
1111 return EXT4_SB(inode->i_sb)->key_prefix_size;
1112}
1113
1114static int ext4_prepare_context(struct inode *inode)
1115{
1116 return ext4_convert_inline_data(inode);
1117}
1118
1119static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1120 void *fs_data)
1121{
2f8f5e76
EB		 1122 handle_t *handle = fs_data;
1123 int res, res2, retries = 0;
1124
1125 /*
1126 * If a journal handle was specified, then the encryption context is
1127 * being set on a new inode via inheritance and is part of a larger
1128 * transaction to create the inode. Otherwise the encryption context is
1129 * being set on an existing inode in its own transaction. Only in the
1130 * latter case should the "retry on ENOSPC" logic be used.
1131 */
a7550b30 1132
2f8f5e76
EB		 1133 if (handle) {
1134 res = ext4_xattr_set_handle(handle, inode,
1135 EXT4_XATTR_INDEX_ENCRYPTION,
1136 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1137 ctx, len, 0);
a7550b30
JK		 1138 if (!res) {
1139 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1140 ext4_clear_inode_state(inode,
1141 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b
JK		 1142 /*
1143 * Update inode->i_flags - e.g. S_DAX may get disabled
1144 */
1145 ext4_set_inode_flags(inode);
a7550b30
JK		 1146 }
1147 return res;
1148 }
1149
2f8f5e76 1150retry:
a7550b30
JK		 1151 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1152 ext4_jbd2_credits_xattr(inode));
1153 if (IS_ERR(handle))
1154 return PTR_ERR(handle);
1155
2f8f5e76
EB		 1156 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1157 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1158 ctx, len, 0);
a7550b30
JK		 1159 if (!res) {
1160 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
a3caa24b
JK		 1161 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1162 ext4_set_inode_flags(inode);
a7550b30
JK		 1163 res = ext4_mark_inode_dirty(handle, inode);
1164 if (res)
1165 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1166 }
1167 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1168
1169 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1170 goto retry;
a7550b30
JK		 1171 if (!res)
1172 res = res2;
1173 return res;
1174}
1175
1176static int ext4_dummy_context(struct inode *inode)
1177{
1178 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1179}
1180
1181static unsigned ext4_max_namelen(struct inode *inode)
1182{
1183 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1184 EXT4_NAME_LEN;
1185}
1186
1187static struct fscrypt_operations ext4_cryptops = {
1188 .get_context = ext4_get_context,
1189 .key_prefix = ext4_key_prefix,
1190 .prepare_context = ext4_prepare_context,
1191 .set_context = ext4_set_context,
1192 .dummy_context = ext4_dummy_context,
1193 .is_encrypted = ext4_encrypted_inode,
1194 .empty_dir = ext4_empty_dir,
1195 .max_namelen = ext4_max_namelen,
1196};
1197#else
1198static struct fscrypt_operations ext4_cryptops = {
1199 .is_encrypted = ext4_encrypted_inode,
1200};
1201#endif
1202
ac27a0ec 1203#ifdef CONFIG_QUOTA
689c958c
LX		 1204static char *quotatypes[] = INITQFNAMES;
1205#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1206
617ba13b
MC		 1207static int ext4_write_dquot(struct dquot *dquot);
1208static int ext4_acquire_dquot(struct dquot *dquot);
1209static int ext4_release_dquot(struct dquot *dquot);
1210static int ext4_mark_dquot_dirty(struct dquot *dquot);
1211static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1212static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1213 const struct path *path);
ca0e05e4 1214static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1215static int ext4_quota_on_mount(struct super_block *sb, int type);
1216static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1217 size_t len, loff_t off);
617ba13b 1218static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1219 const char *data, size_t len, loff_t off);
7c319d32
AK		 1220static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1221 unsigned int flags);
1222static int ext4_enable_quotas(struct super_block *sb);
8f0e8746 1223static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1224
96c7e0d9
JK		 1225static struct dquot **ext4_get_dquots(struct inode *inode)
1226{
1227 return EXT4_I(inode)->i_dquot;
1228}
1229
61e225dc 1230static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1231 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1232 .write_dquot = ext4_write_dquot,
1233 .acquire_dquot = ext4_acquire_dquot,
1234 .release_dquot = ext4_release_dquot,
1235 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1236 .write_info = ext4_write_info,
1237 .alloc_dquot = dquot_alloc,
1238 .destroy_dquot = dquot_destroy,
040cb378 1239 .get_projid = ext4_get_projid,
8f0e8746 1240 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1241};
1242
0d54b217 1243static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1244 .quota_on = ext4_quota_on,
ca0e05e4 1245 .quota_off = ext4_quota_off,
287a8095 1246 .quota_sync = dquot_quota_sync,
0a240339 1247 .get_state = dquot_get_state,
287a8095
CH		 1248 .set_info = dquot_set_dqinfo,
1249 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1250 .set_dqblk = dquot_set_dqblk,
1251 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1252};
1253#endif
1254
ee9b6d61 1255static const struct super_operations ext4_sops = {
617ba13b
MC		 1256 .alloc_inode = ext4_alloc_inode,
1257 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1258 .write_inode = ext4_write_inode,
1259 .dirty_inode = ext4_dirty_inode,
7ff9c073 1260 .drop_inode = ext4_drop_inode,
0930fcc1 1261 .evict_inode = ext4_evict_inode,
617ba13b 1262 .put_super = ext4_put_super,
617ba13b 1263 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1264 .freeze_fs = ext4_freeze,
1265 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1266 .statfs = ext4_statfs,
1267 .remount_fs = ext4_remount,
617ba13b 1268 .show_options = ext4_show_options,
ac27a0ec 1269#ifdef CONFIG_QUOTA
617ba13b
MC		 1270 .quota_read = ext4_quota_read,
1271 .quota_write = ext4_quota_write,
96c7e0d9 1272 .get_dquots = ext4_get_dquots,
ac27a0ec 1273#endif
c39a7f84 1274 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1275};
1276
39655164 1277static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1278 .fh_to_dentry = ext4_fh_to_dentry,
1279 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1280 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1281};
1282
1283enum {
1284 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1285 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1286 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1287 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1288 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1289 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1290 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1291 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1292 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1293 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1294 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1295 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1296 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1449032b 1297 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1298 Opt_lazytime, Opt_nolazytime,
1449032b 1299 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1300 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1301 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1302 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1303 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1304};
1305
a447c093 1306static const match_table_t tokens = {
ac27a0ec
DK		 1307 {Opt_bsd_df, "bsddf"},
1308 {Opt_minix_df, "minixdf"},
1309 {Opt_grpid, "grpid"},
1310 {Opt_grpid, "bsdgroups"},
1311 {Opt_nogrpid, "nogrpid"},
1312 {Opt_nogrpid, "sysvgroups"},
1313 {Opt_resgid, "resgid=%u"},
1314 {Opt_resuid, "resuid=%u"},
1315 {Opt_sb, "sb=%u"},
1316 {Opt_err_cont, "errors=continue"},
1317 {Opt_err_panic, "errors=panic"},
1318 {Opt_err_ro, "errors=remount-ro"},
1319 {Opt_nouid32, "nouid32"},
ac27a0ec 1320 {Opt_debug, "debug"},
72578c33
TT		 1321 {Opt_removed, "oldalloc"},
1322 {Opt_removed, "orlov"},
ac27a0ec
DK		 1323 {Opt_user_xattr, "user_xattr"},
1324 {Opt_nouser_xattr, "nouser_xattr"},
1325 {Opt_acl, "acl"},
1326 {Opt_noacl, "noacl"},
e3bb52ae 1327 {Opt_noload, "norecovery"},
5a916be1 1328 {Opt_noload, "noload"},
72578c33
TT		 1329 {Opt_removed, "nobh"},
1330 {Opt_removed, "bh"},
ac27a0ec 1331 {Opt_commit, "commit=%u"},
30773840
TT		 1332 {Opt_min_batch_time, "min_batch_time=%u"},
1333 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1334 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1335 {Opt_journal_path, "journal_path=%s"},
818d276c 1336 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1337 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1338 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1339 {Opt_abort, "abort"},
1340 {Opt_data_journal, "data=journal"},
1341 {Opt_data_ordered, "data=ordered"},
1342 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1343 {Opt_data_err_abort, "data_err=abort"},
1344 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1345 {Opt_offusrjquota, "usrjquota="},
1346 {Opt_usrjquota, "usrjquota=%s"},
1347 {Opt_offgrpjquota, "grpjquota="},
1348 {Opt_grpjquota, "grpjquota=%s"},
1349 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1350 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1351 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1352 {Opt_grpquota, "grpquota"},
1353 {Opt_noquota, "noquota"},
1354 {Opt_quota, "quota"},
1355 {Opt_usrquota, "usrquota"},
49da9392 1356 {Opt_prjquota, "prjquota"},
ac27a0ec 1357 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1358 {Opt_barrier, "barrier"},
1359 {Opt_nobarrier, "nobarrier"},
25ec56b5 1360 {Opt_i_version, "i_version"},
923ae0ff 1361 {Opt_dax, "dax"},
c9de560d 1362 {Opt_stripe, "stripe=%u"},
64769240 1363 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1364 {Opt_lazytime, "lazytime"},
1365 {Opt_nolazytime, "nolazytime"},
dd919b98 1366 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1367 {Opt_removed, "mblk_io_submit"},
1368 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1369 {Opt_block_validity, "block_validity"},
1370 {Opt_noblock_validity, "noblock_validity"},
240799cd 1371 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1372 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1373 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1374 {Opt_auto_da_alloc, "auto_da_alloc"},
1375 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1376 {Opt_dioread_nolock, "dioread_nolock"},
1377 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1378 {Opt_discard, "discard"},
1379 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1380 {Opt_init_itable, "init_itable=%u"},
1381 {Opt_init_itable, "init_itable"},
1382 {Opt_noinit_itable, "noinit_itable"},
df981d03 1383 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1384 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1385 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1386 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1387 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1388 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1389 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1390 {Opt_err, NULL},
ac27a0ec
DK		 1391};
1392
617ba13b 1393static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1394{
617ba13b 1395 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1396 char *options = (char *) *data;
1397
1398 if (!options || strncmp(options, "sb=", 3) != 0)
1399 return 1; /* Default location */
0b8e58a1 1400
ac27a0ec 1401 options += 3;
0b8e58a1 1402 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1403 sb_block = simple_strtoul(options, &options, 0);
1404 if (*options && *options != ',') {
4776004f 1405 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1406 (char *) *data);
1407 return 1;
1408 }
1409 if (*options == ',')
1410 options++;
1411 *data = (void *) options;
0b8e58a1 1412
ac27a0ec
DK		 1413 return sb_block;
1414}
1415
b3881f74 1416#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1417static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1418 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1419
56c50f11
DM		 1420#ifdef CONFIG_QUOTA
1421static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1422{
1423 struct ext4_sb_info *sbi = EXT4_SB(sb);
1424 char *qname;
03dafb5f 1425 int ret = -1;
56c50f11
DM		 1426
1427 if (sb_any_quota_loaded(sb) &&
1428 !sbi->s_qf_names[qtype]) {
1429 ext4_msg(sb, KERN_ERR,
1430 "Cannot change journaled "
1431 "quota options when quota turned on");
57f73c2c 1432 return -1;
56c50f11 1433 }
e2b911c5 1434 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1435 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1436 "ignored when QUOTA feature is enabled");
1437 return 1;
262b4662 1438 }
56c50f11
DM		 1439 qname = match_strdup(args);
1440 if (!qname) {
1441 ext4_msg(sb, KERN_ERR,
1442 "Not enough memory for storing quotafile name");
57f73c2c 1443 return -1;
56c50f11 1444 }
03dafb5f
CG		 1445 if (sbi->s_qf_names[qtype]) {
1446 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1447 ret = 1;
1448 else
1449 ext4_msg(sb, KERN_ERR,
1450 "%s quota file already specified",
1451 QTYPE2NAME(qtype));
1452 goto errout;
56c50f11 1453 }
03dafb5f 1454 if (strchr(qname, '/')) {
56c50f11
DM		 1455 ext4_msg(sb, KERN_ERR,
1456 "quotafile must be on filesystem root");
03dafb5f 1457 goto errout;
56c50f11 1458 }
03dafb5f 1459 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1460 set_opt(sb, QUOTA);
56c50f11 1461 return 1;
03dafb5f
CG		 1462errout:
1463 kfree(qname);
1464 return ret;
56c50f11
DM		 1465}
1466
1467static int clear_qf_name(struct super_block *sb, int qtype)
1468{
1469
1470 struct ext4_sb_info *sbi = EXT4_SB(sb);
1471
1472 if (sb_any_quota_loaded(sb) &&
1473 sbi->s_qf_names[qtype]) {
1474 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1475 " when quota turned on");
57f73c2c 1476 return -1;
56c50f11 1477 }
03dafb5f 1478 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1479 sbi->s_qf_names[qtype] = NULL;
1480 return 1;
1481}
1482#endif
1483
26092bf5
TT		 1484#define MOPT_SET 0x0001
1485#define MOPT_CLEAR 0x0002
1486#define MOPT_NOSUPPORT 0x0004
1487#define MOPT_EXPLICIT 0x0008
1488#define MOPT_CLEAR_ERR 0x0010
1489#define MOPT_GTE0 0x0020
ac27a0ec 1490#ifdef CONFIG_QUOTA
26092bf5
TT		 1491#define MOPT_Q 0
1492#define MOPT_QFMT 0x0040
1493#else
1494#define MOPT_Q MOPT_NOSUPPORT
1495#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1496#endif
26092bf5 1497#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1498#define MOPT_NO_EXT2 0x0100
1499#define MOPT_NO_EXT3 0x0200
1500#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1501#define MOPT_STRING 0x0400
26092bf5
TT		 1502
1503static const struct mount_opts {
1504 int token;
1505 int mount_opt;
1506 int flags;
1507} ext4_mount_opts[] = {
1508 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1509 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1510 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1511 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1512 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1513 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1514 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1515 MOPT_EXT4_ONLY | MOPT_SET},
1516 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1517 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1518 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1519 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1520 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1521 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1522 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1523 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1524 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1525 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1526 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1527 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1528 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1529 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1530 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1531 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1532 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1533 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1534 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1535 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1536 MOPT_NO_EXT2},
8dc0aa8c 1537 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1538 MOPT_NO_EXT2},
26092bf5
TT		 1539 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1540 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1541 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1542 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1543 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1544 {Opt_commit, 0, MOPT_GTE0},
1545 {Opt_max_batch_time, 0, MOPT_GTE0},
1546 {Opt_min_batch_time, 0, MOPT_GTE0},
1547 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1548 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1549 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1550 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1551 {Opt_resuid, 0, MOPT_GTE0},
1552 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1553 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1554 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1555 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1556 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1557 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1558 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1559 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1560 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1561 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1562#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1563 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1564 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1565#else
26092bf5
TT		 1566 {Opt_acl, 0, MOPT_NOSUPPORT},
1567 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1568#endif
26092bf5
TT		 1569 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1570 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1571 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1572 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1573 MOPT_SET | MOPT_Q},
1574 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1575 MOPT_SET | MOPT_Q},
49da9392
JK		 1576 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1577 MOPT_SET | MOPT_Q},
26092bf5 1578 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1579 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1580 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1581 {Opt_usrjquota, 0, MOPT_Q},
1582 {Opt_grpjquota, 0, MOPT_Q},
1583 {Opt_offusrjquota, 0, MOPT_Q},
1584 {Opt_offgrpjquota, 0, MOPT_Q},
1585 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1586 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1587 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1588 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1589 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1590 {Opt_err, 0, 0}
1591};
1592
1593static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1594 substring_t *args, unsigned long *journal_devnum,
1595 unsigned int *journal_ioprio, int is_remount)
1596{
1597 struct ext4_sb_info *sbi = EXT4_SB(sb);
1598 const struct mount_opts *m;
08cefc7a
EB		 1599 kuid_t uid;
1600 kgid_t gid;
26092bf5
TT		 1601 int arg = 0;
1602
57f73c2c
TT		 1603#ifdef CONFIG_QUOTA
1604 if (token == Opt_usrjquota)
1605 return set_qf_name(sb, USRQUOTA, &args[0]);
1606 else if (token == Opt_grpjquota)
1607 return set_qf_name(sb, GRPQUOTA, &args[0]);
1608 else if (token == Opt_offusrjquota)
1609 return clear_qf_name(sb, USRQUOTA);
1610 else if (token == Opt_offgrpjquota)
1611 return clear_qf_name(sb, GRPQUOTA);
1612#endif
26092bf5 1613 switch (token) {
f7048605
TT		 1614 case Opt_noacl:
1615 case Opt_nouser_xattr:
1616 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1617 break;
26092bf5
TT		 1618 case Opt_sb:
1619 return 1; /* handled by get_sb_block() */
1620 case Opt_removed:
5f3633e3 1621 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1622 return 1;
26092bf5
TT		 1623 case Opt_abort:
1624 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1625 return 1;
1626 case Opt_i_version:
1627 sb->s_flags |= MS_I_VERSION;
1628 return 1;
a26f4992
TT		 1629 case Opt_lazytime:
1630 sb->s_flags |= MS_LAZYTIME;
1631 return 1;
1632 case Opt_nolazytime:
1633 sb->s_flags &= ~MS_LAZYTIME;
1634 return 1;
26092bf5
TT		 1635 }
1636
5f3633e3
JK		 1637 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1638 if (token == m->token)
1639 break;
1640
1641 if (m->token == Opt_err) {
1642 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1643 "or missing value", opt);
1644 return -1;
1645 }
1646
8dc0aa8c
TT		 1647 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1648 ext4_msg(sb, KERN_ERR,
1649 "Mount option \"%s\" incompatible with ext2", opt);
1650 return -1;
1651 }
1652 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1653 ext4_msg(sb, KERN_ERR,
1654 "Mount option \"%s\" incompatible with ext3", opt);
1655 return -1;
1656 }
1657
acc94812
SF		 1658 if (token == Opt_err_panic && !capable(CAP_SYS_ADMIN)) {
1659 ext4_msg(sb, KERN_ERR,
1660 "Mount option \"%s\" not allowed for unprivileged mounts",
1661 opt);
1662 return -1;
1663 }
1664
ad4eec61 1665 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1666 return -1;
1667 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1668 return -1;
c93cf2d7
DM		 1669 if (m->flags & MOPT_EXPLICIT) {
1670 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1671 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1672 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1673 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1674 } else
1675 return -1;
1676 }
5f3633e3
JK		 1677 if (m->flags & MOPT_CLEAR_ERR)
1678 clear_opt(sb, ERRORS_MASK);
1679 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1680 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1681 "options when quota turned on");
1682 return -1;
1683 }
1684
1685 if (m->flags & MOPT_NOSUPPORT) {
1686 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1687 } else if (token == Opt_commit) {
1688 if (arg == 0)
1689 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1690 sbi->s_commit_interval = HZ * arg;
1691 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1692 sbi->s_max_batch_time = arg;
1693 } else if (token == Opt_min_batch_time) {
1694 sbi->s_min_batch_time = arg;
1695 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1696 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1697 ext4_msg(sb, KERN_ERR,
1698 "EXT4-fs: inode_readahead_blks must be "
1699 "0 or a power of 2 smaller than 2^31");
26092bf5 1700 return -1;
5f3633e3
JK		 1701 }
1702 sbi->s_inode_readahead_blks = arg;
1703 } else if (token == Opt_init_itable) {
1704 set_opt(sb, INIT_INODE_TABLE);
1705 if (!args->from)
1706 arg = EXT4_DEF_LI_WAIT_MULT;
1707 sbi->s_li_wait_mult = arg;
1708 } else if (token == Opt_max_dir_size_kb) {
1709 sbi->s_max_dir_size_kb = arg;
1710 } else if (token == Opt_stripe) {
1711 sbi->s_stripe = arg;
1712 } else if (token == Opt_resuid) {
acc94812 1713 uid = make_kuid(sb->s_user_ns, arg);
5f3633e3
JK		 1714 if (!uid_valid(uid)) {
1715 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1716 return -1;
1717 }
5f3633e3
JK		 1718 sbi->s_resuid = uid;
1719 } else if (token == Opt_resgid) {
acc94812 1720 gid = make_kgid(sb->s_user_ns, arg);
5f3633e3
JK		 1721 if (!gid_valid(gid)) {
1722 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1723 return -1;
1724 }
1725 sbi->s_resgid = gid;
1726 } else if (token == Opt_journal_dev) {
1727 if (is_remount) {
1728 ext4_msg(sb, KERN_ERR,
1729 "Cannot specify journal on remount");
1730 return -1;
1731 }
1732 *journal_devnum = arg;
ad4eec61
ES		 1733 } else if (token == Opt_journal_path) {
1734 char *journal_path;
1735 struct inode *journal_inode;
1736 struct path path;
1737 int error;
1738
1739 if (is_remount) {
1740 ext4_msg(sb, KERN_ERR,
1741 "Cannot specify journal on remount");
1742 return -1;
1743 }
1744 journal_path = match_strdup(&args[0]);
1745 if (!journal_path) {
1746 ext4_msg(sb, KERN_ERR, "error: could not dup "
1747 "journal device string");
1748 return -1;
1749 }
1750
1751 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1752 if (error) {
1753 ext4_msg(sb, KERN_ERR, "error: could not find "
1754 "journal device path: error %d", error);
1755 kfree(journal_path);
1756 return -1;
1757 }
1758
acc94812
SF		 1759 /*
1760 * Refuse access for unprivileged mounts if the user does
1761 * not have rw access to the journal device via the supplied
1762 * path.
1763 */
1764 if (!capable(CAP_SYS_ADMIN) &&
1765 inode_permission(d_inode(path.dentry), MAY_READ|MAY_WRITE)) {
1766 ext4_msg(sb, KERN_ERR,
1767 "error: Insufficient access to journal path %s",
1768 journal_path);
1769 return -1;
1770 }
1771
2b0143b5 1772 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1773 if (!S_ISBLK(journal_inode->i_mode)) {
1774 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1775 "is not a block device", journal_path);
1776 path_put(&path);
1777 kfree(journal_path);
1778 return -1;
1779 }
1780
1781 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1782 path_put(&path);
1783 kfree(journal_path);
5f3633e3
JK		 1784 } else if (token == Opt_journal_ioprio) {
1785 if (arg > 7) {
1786 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1787 " (must be 0-7)");
1788 return -1;
1789 }
1790 *journal_ioprio =
1791 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1792 } else if (token == Opt_test_dummy_encryption) {
1793#ifdef CONFIG_EXT4_FS_ENCRYPTION
1794 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1795 ext4_msg(sb, KERN_WARNING,
1796 "Test dummy encryption mode enabled");
1797#else
1798 ext4_msg(sb, KERN_WARNING,
1799 "Test dummy encryption mount option ignored");
1800#endif
5f3633e3
JK		 1801 } else if (m->flags & MOPT_DATAJ) {
1802 if (is_remount) {
1803 if (!sbi->s_journal)
1804 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1805 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1806 ext4_msg(sb, KERN_ERR,
26092bf5 1807 "Cannot change data mode on remount");
26092bf5 1808 return -1;
ac27a0ec 1809 }
26092bf5 1810 } else {
5f3633e3
JK		 1811 clear_opt(sb, DATA_FLAGS);
1812 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1813 }
5f3633e3
JK		 1814#ifdef CONFIG_QUOTA
1815 } else if (m->flags & MOPT_QFMT) {
1816 if (sb_any_quota_loaded(sb) &&
1817 sbi->s_jquota_fmt != m->mount_opt) {
1818 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1819 "quota options when quota turned on");
1820 return -1;
1821 }
e2b911c5 1822 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1823 ext4_msg(sb, KERN_INFO,
1824 "Quota format mount options ignored "
262b4662 1825 "when QUOTA feature is enabled");
c325a67c 1826 return 1;
262b4662 1827 }
5f3633e3 1828 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1829#endif
923ae0ff 1830 } else if (token == Opt_dax) {
ef83b6e8
DW		 1831#ifdef CONFIG_FS_DAX
1832 ext4_msg(sb, KERN_WARNING,
1833 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1834 sbi->s_mount_opt |= m->mount_opt;
1835#else
923ae0ff
RZ		 1836 ext4_msg(sb, KERN_INFO, "dax option not supported");
1837 return -1;
5f3633e3 1838#endif
7915a861
AN		 1839 } else if (token == Opt_data_err_abort) {
1840 sbi->s_mount_opt |= m->mount_opt;
1841 } else if (token == Opt_data_err_ignore) {
1842 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1843 } else {
1844 if (!args->from)
1845 arg = 1;
1846 if (m->flags & MOPT_CLEAR)
1847 arg = !arg;
1848 else if (unlikely(!(m->flags & MOPT_SET))) {
1849 ext4_msg(sb, KERN_WARNING,
1850 "buggy handling of option %s", opt);
1851 WARN_ON(1);
1852 return -1;
1853 }
1854 if (arg != 0)
1855 sbi->s_mount_opt |= m->mount_opt;
1856 else
1857 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1858 }
5f3633e3 1859 return 1;
26092bf5
TT		 1860}
1861
1862static int parse_options(char *options, struct super_block *sb,
1863 unsigned long *journal_devnum,
1864 unsigned int *journal_ioprio,
1865 int is_remount)
1866{
1867 struct ext4_sb_info *sbi = EXT4_SB(sb);
1868 char *p;
1869 substring_t args[MAX_OPT_ARGS];
1870 int token;
1871
1872 if (!options)
1873 return 1;
1874
1875 while ((p = strsep(&options, ",")) != NULL) {
1876 if (!*p)
1877 continue;
1878 /*
1879 * Initialize args struct so we know whether arg was
1880 * found; some options take optional arguments.
1881 */
caecd0af 1882 args[0].to = args[0].from = NULL;
26092bf5
TT		 1883 token = match_token(p, tokens, args);
1884 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1885 journal_ioprio, is_remount) < 0)
1886 return 0;
ac27a0ec
DK		 1887 }
1888#ifdef CONFIG_QUOTA
49da9392
JK		 1889 /*
1890 * We do the test below only for project quotas. 'usrquota' and
1891 * 'grpquota' mount options are allowed even without quota feature
1892 * to support legacy quotas in quota files.
1893 */
1894 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1895 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1896 "Cannot enable project quota enforcement.");
1897 return 0;
1898 }
1899 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1900 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1901 clear_opt(sb, USRQUOTA);
ac27a0ec 1902
482a7425 1903 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1904 clear_opt(sb, GRPQUOTA);
ac27a0ec 1905
56c50f11 1906 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1907 ext4_msg(sb, KERN_ERR, "old and new quota "
1908 "format mixing");
ac27a0ec
DK		 1909 return 0;
1910 }
1911
1912 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1913 ext4_msg(sb, KERN_ERR, "journaled quota format "
1914 "not specified");
ac27a0ec
DK		 1915 return 0;
1916 }
ac27a0ec
DK		 1917 }
1918#endif
261cb20c
JK		 1919 if (test_opt(sb, DIOREAD_NOLOCK)) {
1920 int blocksize =
1921 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1922
09cbfeaf 1923 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 1924 ext4_msg(sb, KERN_ERR, "can't mount with "
1925 "dioread_nolock if block size != PAGE_SIZE");
1926 return 0;
1927 }
1928 }
ac27a0ec
DK		 1929 return 1;
1930}
1931
2adf6da8
TT		 1932static inline void ext4_show_quota_options(struct seq_file *seq,
1933 struct super_block *sb)
1934{
1935#if defined(CONFIG_QUOTA)
1936 struct ext4_sb_info *sbi = EXT4_SB(sb);
1937
1938 if (sbi->s_jquota_fmt) {
1939 char *fmtname = "";
1940
1941 switch (sbi->s_jquota_fmt) {
1942 case QFMT_VFS_OLD:
1943 fmtname = "vfsold";
1944 break;
1945 case QFMT_VFS_V0:
1946 fmtname = "vfsv0";
1947 break;
1948 case QFMT_VFS_V1:
1949 fmtname = "vfsv1";
1950 break;
1951 }
1952 seq_printf(seq, ",jqfmt=%s", fmtname);
1953 }
1954
1955 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1956 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1957
1958 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1959 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1960#endif
1961}
1962
5a916be1
TT		 1963static const char *token2str(int token)
1964{
50df9fd5 1965 const struct match_token *t;
5a916be1
TT		 1966
1967 for (t = tokens; t->token != Opt_err; t++)
1968 if (t->token == token && !strchr(t->pattern, '='))
1969 break;
1970 return t->pattern;
1971}
1972
2adf6da8
TT		 1973/*
1974 * Show an option if
1975 * - it's set to a non-default value OR
1976 * - if the per-sb default is different from the global default
1977 */
66acdcf4
TT		 1978static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1979 int nodefs)
2adf6da8 1980{
2adf6da8
TT		 1981 struct ext4_sb_info *sbi = EXT4_SB(sb);
1982 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1983 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1984 const struct mount_opts *m;
66acdcf4 1985 char sep = nodefs ? '\n' : ',';
2adf6da8 1986
66acdcf4
TT		 1987#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1988#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1989
1990 if (sbi->s_sb_block != 1)
5a916be1
TT		 1991 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1992
1993 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1994 int want_set = m->flags & MOPT_SET;
1995 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1996 (m->flags & MOPT_CLEAR_ERR))
1997 continue;
66acdcf4 1998 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1999 continue; /* skip if same as the default */
2000 if ((want_set &&
2001 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2002 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2003 continue; /* select Opt_noFoo vs Opt_Foo */
2004 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2005 }
5a916be1 2006
acc94812 2007 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(sb->s_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2008 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a 2009 SEQ_OPTS_PRINT("resuid=%u",
acc94812
SF		 2010 from_kuid_munged(sb->s_user_ns, sbi->s_resuid));
2011 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(sb->s_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2012 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a 2013 SEQ_OPTS_PRINT("resgid=%u",
acc94812 2014 from_kgid_munged(sb->s_user_ns, sbi->s_resgid));
66acdcf4 2015 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2016 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2017 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2018 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2019 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2020 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2021 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2022 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2023 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2024 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2025 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2026 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2027 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 2028 if (sb->s_flags & MS_I_VERSION)
5a916be1 2029 SEQ_OPTS_PUTS("i_version");
66acdcf4 2030 if (nodefs || sbi->s_stripe)
5a916be1 2031 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 2032 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2033 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2034 SEQ_OPTS_PUTS("data=journal");
2035 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2036 SEQ_OPTS_PUTS("data=ordered");
2037 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2038 SEQ_OPTS_PUTS("data=writeback");
2039 }
66acdcf4
TT		 2040 if (nodefs ||
2041 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2042 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2043 sbi->s_inode_readahead_blks);
2adf6da8 2044
66acdcf4
TT		 2045 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2046 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2047 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2048 if (nodefs || sbi->s_max_dir_size_kb)
2049 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2050 if (test_opt(sb, DATA_ERR_ABORT))
2051 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 2052
2053 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2054 return 0;
2055}
2056
66acdcf4
TT		 2057static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2058{
2059 return _ext4_show_options(seq, root->d_sb, 0);
2060}
2061
ebd173be 2062int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2063{
2064 struct super_block *sb = seq->private;
2065 int rc;
2066
2067 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2068 rc = _ext4_show_options(seq, sb, 1);
2069 seq_puts(seq, "\n");
2070 return rc;
2071}
2072
617ba13b 2073static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2074 int read_only)
2075{
617ba13b 2076 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2077 int res = 0;
2078
617ba13b 2079 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2080 ext4_msg(sb, KERN_ERR, "revision level too high, "
2081 "forcing read-only mode");
ac27a0ec
DK		 2082 res = MS_RDONLY;
2083 }
2084 if (read_only)
281b5995 2085 goto done;
617ba13b 2086 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2087 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2088 "running e2fsck is recommended");
c8b459f4 2089 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2090 ext4_msg(sb, KERN_WARNING,
2091 "warning: mounting fs with errors, "
2092 "running e2fsck is recommended");
ed3ce80a 2093 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2094 le16_to_cpu(es->s_mnt_count) >=
2095 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2096 ext4_msg(sb, KERN_WARNING,
2097 "warning: maximal mount count reached, "
2098 "running e2fsck is recommended");
ac27a0ec
DK		 2099 else if (le32_to_cpu(es->s_checkinterval) &&
2100 (le32_to_cpu(es->s_lastcheck) +
2101 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 2102 ext4_msg(sb, KERN_WARNING,
2103 "warning: checktime reached, "
2104 "running e2fsck is recommended");
0b8e58a1 2105 if (!sbi->s_journal)
0390131b 2106 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2107 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2108 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2109 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 2110 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 2111 ext4_update_dynamic_rev(sb);
0390131b 2112 if (sbi->s_journal)
e2b911c5 2113 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2114
e2d67052 2115 ext4_commit_super(sb, 1);
281b5995 2116done:
ac27a0ec 2117 if (test_opt(sb, DEBUG))
a9df9a49 2118 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2119 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2120 sb->s_blocksize,
2121 sbi->s_groups_count,
617ba13b
MC		 2122 EXT4_BLOCKS_PER_GROUP(sb),
2123 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2124 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2125
7abc52c2 2126 cleancache_init_fs(sb);
ac27a0ec
DK		 2127 return res;
2128}
2129
117fff10
TT		 2130int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2131{
2132 struct ext4_sb_info *sbi = EXT4_SB(sb);
2133 struct flex_groups *new_groups;
2134 int size;
2135
2136 if (!sbi->s_log_groups_per_flex)
2137 return 0;
2138
2139 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2140 if (size <= sbi->s_flex_groups_allocated)
2141 return 0;
2142
2143 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2144 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2145 if (!new_groups) {
2146 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2147 size / (int) sizeof(struct flex_groups));
2148 return -ENOMEM;
2149 }
2150
2151 if (sbi->s_flex_groups) {
2152 memcpy(new_groups, sbi->s_flex_groups,
2153 (sbi->s_flex_groups_allocated *
2154 sizeof(struct flex_groups)));
b93b41d4 2155 kvfree(sbi->s_flex_groups);
117fff10
TT		 2156 }
2157 sbi->s_flex_groups = new_groups;
2158 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2159 return 0;
2160}
2161
772cb7c8
JS		 2162static int ext4_fill_flex_info(struct super_block *sb)
2163{
2164 struct ext4_sb_info *sbi = EXT4_SB(sb);
2165 struct ext4_group_desc *gdp = NULL;
772cb7c8 2166 ext4_group_t flex_group;
117fff10 2167 int i, err;
772cb7c8 2168
503358ae 2169 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2170 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2171 sbi->s_log_groups_per_flex = 0;
2172 return 1;
2173 }
2174
117fff10
TT		 2175 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2176 if (err)
9933fc0a 2177 goto failed;
772cb7c8 2178
772cb7c8 2179 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2180 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2181
2182 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2183 atomic_add(ext4_free_inodes_count(sb, gdp),
2184 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2185 atomic64_add(ext4_free_group_clusters(sb, gdp),
2186 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2187 atomic_add(ext4_used_dirs_count(sb, gdp),
2188 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2189 }
2190
2191 return 1;
2192failed:
2193 return 0;
2194}
2195
e2b911c5 2196static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2197 struct ext4_group_desc *gdp)
717d50e4 2198{
b47820ed 2199 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2200 __u16 crc = 0;
feb0ab32 2201 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2202 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2203
9aa5d32b 2204 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2205 /* Use new metadata_csum algorithm */
feb0ab32 2206 __u32 csum32;
b47820ed 2207 __u16 dummy_csum = 0;
feb0ab32 2208
feb0ab32
DW		 2209 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2210 sizeof(le_group));
b47820ed
DJ		 2211 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2212 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2213 sizeof(dummy_csum));
2214 offset += sizeof(dummy_csum);
2215 if (offset < sbi->s_desc_size)
2216 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2217 sbi->s_desc_size - offset);
feb0ab32
DW		 2218
2219 crc = csum32 & 0xFFFF;
2220 goto out;
717d50e4
AD		 2221 }
2222
feb0ab32 2223 /* old crc16 code */
e2b911c5 2224 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2225 return 0;
2226
feb0ab32
DW		 2227 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2228 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2229 crc = crc16(crc, (__u8 *)gdp, offset);
2230 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2231 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2232 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2233 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2234 crc = crc16(crc, (__u8 *)gdp + offset,
2235 le16_to_cpu(sbi->s_es->s_desc_size) -
2236 offset);
2237
2238out:
717d50e4
AD		 2239 return cpu_to_le16(crc);
2240}
2241
feb0ab32 2242int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2243 struct ext4_group_desc *gdp)
2244{
feb0ab32 2245 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2246 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2247 return 0;
2248
2249 return 1;
2250}
2251
feb0ab32
DW		 2252void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2253 struct ext4_group_desc *gdp)
2254{
2255 if (!ext4_has_group_desc_csum(sb))
2256 return;
e2b911c5 2257 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2258}
2259
ac27a0ec 2260/* Called at mount-time, super-block is locked */
bfff6873 2261static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2262 ext4_fsblk_t sb_block,
bfff6873 2263 ext4_group_t *first_not_zeroed)
ac27a0ec 2264{
617ba13b
MC		 2265 struct ext4_sb_info *sbi = EXT4_SB(sb);
2266 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2267 ext4_fsblk_t last_block;
bd81d8ee
LV		 2268 ext4_fsblk_t block_bitmap;
2269 ext4_fsblk_t inode_bitmap;
2270 ext4_fsblk_t inode_table;
ce421581 2271 int flexbg_flag = 0;
bfff6873 2272 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2273
e2b911c5 2274 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2275 flexbg_flag = 1;
2276
af5bc92d 2277 ext4_debug("Checking group descriptors");
ac27a0ec 2278
197cd65a
AM		 2279 for (i = 0; i < sbi->s_groups_count; i++) {
2280 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2281
ce421581 2282 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2283 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2284 else
2285 last_block = first_block +
617ba13b 2286 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2287
bfff6873
LC		 2288 if ((grp == sbi->s_groups_count) &&
2289 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2290 grp = i;
2291
8fadc143 2292 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2293 if (block_bitmap == sb_block) {
2294 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2295 "Block bitmap for group %u overlaps "
2296 "superblock", i);
2297 }
2b2d6d01 2298 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2299 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2300 "Block bitmap for group %u not in group "
b31e1552 2301 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2302 return 0;
2303 }
8fadc143 2304 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2305 if (inode_bitmap == sb_block) {
2306 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2307 "Inode bitmap for group %u overlaps "
2308 "superblock", i);
2309 }
2b2d6d01 2310 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2311 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2312 "Inode bitmap for group %u not in group "
b31e1552 2313 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2314 return 0;
2315 }
8fadc143 2316 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2317 if (inode_table == sb_block) {
2318 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2319 "Inode table for group %u overlaps "
2320 "superblock", i);
2321 }
bd81d8ee 2322 if (inode_table < first_block ||
2b2d6d01 2323 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2324 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2325 "Inode table for group %u not in group "
b31e1552 2326 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2327 return 0;
2328 }
955ce5f5 2329 ext4_lock_group(sb, i);
feb0ab32 2330 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2331 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2332 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2333 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2334 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2335 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2336 ext4_unlock_group(sb, i);
8a266467 2337 return 0;
7ee1ec4c 2338 }
717d50e4 2339 }
955ce5f5 2340 ext4_unlock_group(sb, i);
ce421581
JS		 2341 if (!flexbg_flag)
2342 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2343 }
bfff6873
LC		 2344 if (NULL != first_not_zeroed)
2345 *first_not_zeroed = grp;
ac27a0ec
DK		 2346 return 1;
2347}
2348
617ba13b 2349/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2350 * the superblock) which were deleted from all directories, but held open by
2351 * a process at the time of a crash. We walk the list and try to delete these
2352 * inodes at recovery time (only with a read-write filesystem).
2353 *
2354 * In order to keep the orphan inode chain consistent during traversal (in
2355 * case of crash during recovery), we link each inode into the superblock
2356 * orphan list_head and handle it the same way as an inode deletion during
2357 * normal operation (which journals the operations for us).
2358 *
2359 * We only do an iget() and an iput() on each inode, which is very safe if we
2360 * accidentally point at an in-use or already deleted inode. The worst that
2361 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2362 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2363 * e2fsck was run on this filesystem, and it must have already done the orphan
2364 * inode cleanup for us, so we can safely abort without any further action.
2365 */
2b2d6d01
TT		 2366static void ext4_orphan_cleanup(struct super_block *sb,
2367 struct ext4_super_block *es)
ac27a0ec
DK		 2368{
2369 unsigned int s_flags = sb->s_flags;
2c98eb5e 2370 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec
DK		 2371#ifdef CONFIG_QUOTA
2372 int i;
2373#endif
2374 if (!es->s_last_orphan) {
2375 jbd_debug(4, "no orphan inodes to clean up\n");
2376 return;
2377 }
2378
a8f48a95 2379 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2380 ext4_msg(sb, KERN_ERR, "write access "
2381 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2382 return;
2383 }
2384
d39195c3
AG		 2385 /* Check if feature set would not allow a r/w mount */
2386 if (!ext4_feature_set_ok(sb, 0)) {
2387 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2388 "unknown ROCOMPAT features");
2389 return;
2390 }
2391
617ba13b 2392 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2393 /* don't clear list on RO mount w/ errors */
2394 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2395 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2396 "clearing orphan list.\n");
c25f9bc6
ES		 2397 es->s_last_orphan = 0;
2398 }
ac27a0ec
DK		 2399 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2400 return;
2401 }
2402
2403 if (s_flags & MS_RDONLY) {
b31e1552 2404 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2405 sb->s_flags &= ~MS_RDONLY;
2406 }
2407#ifdef CONFIG_QUOTA
2408 /* Needed for iput() to work correctly and not trash data */
2409 sb->s_flags |= MS_ACTIVE;
2410 /* Turn on quotas so that they are updated correctly */
a2d4a646 2411 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2412 if (EXT4_SB(sb)->s_qf_names[i]) {
2413 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2414 if (ret < 0)
b31e1552
ES		 2415 ext4_msg(sb, KERN_ERR,
2416 "Cannot turn on journaled "
2417 "quota: error %d", ret);
ac27a0ec
DK		 2418 }
2419 }
2420#endif
2421
2422 while (es->s_last_orphan) {
2423 struct inode *inode;
2424
c65d5c6c
VN		 2425 /*
2426 * We may have encountered an error during cleanup; if
2427 * so, skip the rest.
2428 */
2429 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2430 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2431 es->s_last_orphan = 0;
2432 break;
2433 }
2434
97bd42b9
JB		 2435 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2436 if (IS_ERR(inode)) {
ac27a0ec
DK		 2437 es->s_last_orphan = 0;
2438 break;
2439 }
2440
617ba13b 2441 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2442 dquot_initialize(inode);
ac27a0ec 2443 if (inode->i_nlink) {
566370a2
PT		 2444 if (test_opt(sb, DEBUG))
2445 ext4_msg(sb, KERN_DEBUG,
2446 "%s: truncating inode %lu to %lld bytes",
2447 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2448 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2449 inode->i_ino, inode->i_size);
5955102c 2450 inode_lock(inode);
55f252c9 2451 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2452 ret = ext4_truncate(inode);
2453 if (ret)
2454 ext4_std_error(inode->i_sb, ret);
5955102c 2455 inode_unlock(inode);
ac27a0ec
DK		 2456 nr_truncates++;
2457 } else {
566370a2
PT		 2458 if (test_opt(sb, DEBUG))
2459 ext4_msg(sb, KERN_DEBUG,
2460 "%s: deleting unreferenced inode %lu",
2461 __func__, inode->i_ino);
ac27a0ec
DK		 2462 jbd_debug(2, "deleting unreferenced inode %lu\n",
2463 inode->i_ino);
2464 nr_orphans++;
2465 }
2466 iput(inode); /* The delete magic happens here! */
2467 }
2468
2b2d6d01 2469#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2470
2471 if (nr_orphans)
b31e1552
ES		 2472 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2473 PLURAL(nr_orphans));
ac27a0ec 2474 if (nr_truncates)
b31e1552
ES		 2475 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2476 PLURAL(nr_truncates));
ac27a0ec
DK		 2477#ifdef CONFIG_QUOTA
2478 /* Turn quotas off */
a2d4a646 2479 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2480 if (sb_dqopt(sb)->files[i])
287a8095 2481 dquot_quota_off(sb, i);
ac27a0ec
DK		 2482 }
2483#endif
2484 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2485}
0b8e58a1 2486
cd2291a4
ES		 2487/*
2488 * Maximal extent format file size.
2489 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2490 * extent format containers, within a sector_t, and within i_blocks
2491 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2492 * so that won't be a limiting factor.
2493 *
f17722f9
LC		 2494 * However there is other limiting factor. We do store extents in the form
2495 * of starting block and length, hence the resulting length of the extent
2496 * covering maximum file size must fit into on-disk format containers as
2497 * well. Given that length is always by 1 unit bigger than max unit (because
2498 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2499 *
cd2291a4
ES		 2500 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2501 */
f287a1a5 2502static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2503{
2504 loff_t res;
2505 loff_t upper_limit = MAX_LFS_FILESIZE;
2506
2507 /* small i_blocks in vfs inode? */
f287a1a5 2508 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2509 /*
90c699a9 2510 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2511 * i_block represent total blocks in 512 bytes
2512 * 32 == size of vfs inode i_blocks * 8
2513 */
2514 upper_limit = (1LL << 32) - 1;
2515
2516 /* total blocks in file system block size */
2517 upper_limit >>= (blkbits - 9);
2518 upper_limit <<= blkbits;
2519 }
2520
f17722f9
LC		 2521 /*
2522 * 32-bit extent-start container, ee_block. We lower the maxbytes
2523 * by one fs block, so ee_len can cover the extent of maximum file
2524 * size
2525 */
2526 res = (1LL << 32) - 1;
cd2291a4 2527 res <<= blkbits;
cd2291a4
ES		 2528
2529 /* Sanity check against vm- & vfs- imposed limits */
2530 if (res > upper_limit)
2531 res = upper_limit;
2532
2533 return res;
2534}
ac27a0ec 2535
ac27a0ec 2536/*
cd2291a4 2537 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2538 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2539 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2540 */
f287a1a5 2541static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2542{
617ba13b 2543 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2544 int meta_blocks;
2545 loff_t upper_limit;
0b8e58a1
AD		 2546 /* This is calculated to be the largest file size for a dense, block
2547 * mapped file such that the file's total number of 512-byte sectors,
2548 * including data and all indirect blocks, does not exceed (2^48 - 1).
2549 *
2550 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2551 * number of 512-byte sectors of the file.
0fc1b451
AK		 2552 */
2553
f287a1a5 2554 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2555 /*
90c699a9 2556 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2557 * the inode i_block field represents total file blocks in
2558 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2559 */
2560 upper_limit = (1LL << 32) - 1;
2561
2562 /* total blocks in file system block size */
2563 upper_limit >>= (bits - 9);
2564
2565 } else {
8180a562
AK		 2566 /*
2567 * We use 48 bit ext4_inode i_blocks
2568 * With EXT4_HUGE_FILE_FL set the i_blocks
2569 * represent total number of blocks in
2570 * file system block size
2571 */
0fc1b451
AK		 2572 upper_limit = (1LL << 48) - 1;
2573
0fc1b451
AK		 2574 }
2575
2576 /* indirect blocks */
2577 meta_blocks = 1;
2578 /* double indirect blocks */
2579 meta_blocks += 1 + (1LL << (bits-2));
2580 /* tripple indirect blocks */
2581 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2582
2583 upper_limit -= meta_blocks;
2584 upper_limit <<= bits;
ac27a0ec
DK		 2585
2586 res += 1LL << (bits-2);
2587 res += 1LL << (2*(bits-2));
2588 res += 1LL << (3*(bits-2));
2589 res <<= bits;
2590 if (res > upper_limit)
2591 res = upper_limit;
0fc1b451
AK		 2592
2593 if (res > MAX_LFS_FILESIZE)
2594 res = MAX_LFS_FILESIZE;
2595
ac27a0ec
DK		 2596 return res;
2597}
2598
617ba13b 2599static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2600 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2601{
617ba13b 2602 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2603 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2604 int has_super = 0;
2605
2606 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2607
e2b911c5 2608 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2609 return logical_sb_block + nr + 1;
ac27a0ec 2610 bg = sbi->s_desc_per_block * nr;
617ba13b 2611 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2612 has_super = 1;
0b8e58a1 2613
bd63f6b0
DW		 2614 /*
2615 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2616 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2617 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2618 * compensate.
2619 */
2620 if (sb->s_blocksize == 1024 && nr == 0 &&
2621 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2622 has_super++;
2623
617ba13b 2624 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2625}
2626
c9de560d
AT		 2627/**
2628 * ext4_get_stripe_size: Get the stripe size.
2629 * @sbi: In memory super block info
2630 *
2631 * If we have specified it via mount option, then
2632 * use the mount option value. If the value specified at mount time is
2633 * greater than the blocks per group use the super block value.
2634 * If the super block value is greater than blocks per group return 0.
2635 * Allocator needs it be less than blocks per group.
2636 *
2637 */
2638static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2639{
2640 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2641 unsigned long stripe_width =
2642 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2643 int ret;
c9de560d
AT		 2644
2645 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2646 ret = sbi->s_stripe;
2647 else if (stripe_width <= sbi->s_blocks_per_group)
2648 ret = stripe_width;
2649 else if (stride <= sbi->s_blocks_per_group)
2650 ret = stride;
2651 else
2652 ret = 0;
c9de560d 2653
3eb08658
DE		 2654 /*
2655 * If the stripe width is 1, this makes no sense and
2656 * we set it to 0 to turn off stripe handling code.
2657 */
2658 if (ret <= 1)
2659 ret = 0;
c9de560d 2660
3eb08658 2661 return ret;
c9de560d 2662}
ac27a0ec 2663
a13fb1a4
ES		 2664/*
2665 * Check whether this filesystem can be mounted based on
2666 * the features present and the RDONLY/RDWR mount requested.
2667 * Returns 1 if this filesystem can be mounted as requested,
2668 * 0 if it cannot be.
2669 */
2670static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2671{
e2b911c5 2672 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2673 ext4_msg(sb, KERN_ERR,
2674 "Couldn't mount because of "
2675 "unsupported optional features (%x)",
2676 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2677 ~EXT4_FEATURE_INCOMPAT_SUPP));
2678 return 0;
2679 }
2680
2681 if (readonly)
2682 return 1;
2683
e2b911c5 2684 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b
DW		 2685 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2686 sb->s_flags |= MS_RDONLY;
2687 return 1;
2688 }
2689
a13fb1a4 2690 /* Check that feature set is OK for a read-write mount */
e2b911c5 2691 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2692 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2693 "unsupported optional features (%x)",
2694 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2695 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2696 return 0;
2697 }
2698 /*
2699 * Large file size enabled file system can only be mounted
2700 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2701 */
e2b911c5 2702 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2703 if (sizeof(blkcnt_t) < sizeof(u64)) {
2704 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2705 "cannot be mounted RDWR without "
2706 "CONFIG_LBDAF");
2707 return 0;
2708 }
2709 }
e2b911c5 2710 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2711 ext4_msg(sb, KERN_ERR,
2712 "Can't support bigalloc feature without "
2713 "extents feature\n");
2714 return 0;
2715 }
7c319d32
AK		 2716
2717#ifndef CONFIG_QUOTA
e2b911c5 2718 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2719 ext4_msg(sb, KERN_ERR,
2720 "Filesystem with quota feature cannot be mounted RDWR "
2721 "without CONFIG_QUOTA");
2722 return 0;
2723 }
689c958c
LX		 2724 if (ext4_has_feature_project(sb) && !readonly) {
2725 ext4_msg(sb, KERN_ERR,
2726 "Filesystem with project quota feature cannot be mounted RDWR "
2727 "without CONFIG_QUOTA");
2728 return 0;
2729 }
7c319d32 2730#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2731 return 1;
2732}
2733
66e61a9e
TT		 2734/*
2735 * This function is called once a day if we have errors logged
2736 * on the file system
2737 */
2738static void print_daily_error_info(unsigned long arg)
2739{
2740 struct super_block *sb = (struct super_block *) arg;
2741 struct ext4_sb_info *sbi;
2742 struct ext4_super_block *es;
2743
2744 sbi = EXT4_SB(sb);
2745 es = sbi->s_es;
2746
2747 if (es->s_error_count)
ae0f78de
TT		 2748 /* fsck newer than v1.41.13 is needed to clean this condition. */
2749 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2750 le32_to_cpu(es->s_error_count));
2751 if (es->s_first_error_time) {
ae0f78de 2752 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2753 sb->s_id, le32_to_cpu(es->s_first_error_time),
2754 (int) sizeof(es->s_first_error_func),
2755 es->s_first_error_func,
2756 le32_to_cpu(es->s_first_error_line));
2757 if (es->s_first_error_ino)
651e1c3b 2758 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2759 le32_to_cpu(es->s_first_error_ino));
2760 if (es->s_first_error_block)
651e1c3b 2761 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2762 le64_to_cpu(es->s_first_error_block));
651e1c3b 2763 printk(KERN_CONT "\n");
66e61a9e
TT		 2764 }
2765 if (es->s_last_error_time) {
ae0f78de 2766 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2767 sb->s_id, le32_to_cpu(es->s_last_error_time),
2768 (int) sizeof(es->s_last_error_func),
2769 es->s_last_error_func,
2770 le32_to_cpu(es->s_last_error_line));
2771 if (es->s_last_error_ino)
651e1c3b 2772 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2773 le32_to_cpu(es->s_last_error_ino));
2774 if (es->s_last_error_block)
651e1c3b 2775 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2776 le64_to_cpu(es->s_last_error_block));
651e1c3b 2777 printk(KERN_CONT "\n");
66e61a9e
TT		 2778 }
2779 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2780}
2781
bfff6873
LC		 2782/* Find next suitable group and run ext4_init_inode_table */
2783static int ext4_run_li_request(struct ext4_li_request *elr)
2784{
2785 struct ext4_group_desc *gdp = NULL;
2786 ext4_group_t group, ngroups;
2787 struct super_block *sb;
2788 unsigned long timeout = 0;
2789 int ret = 0;
2790
2791 sb = elr->lr_super;
2792 ngroups = EXT4_SB(sb)->s_groups_count;
2793
2794 for (group = elr->lr_next_group; group < ngroups; group++) {
2795 gdp = ext4_get_group_desc(sb, group, NULL);
2796 if (!gdp) {
2797 ret = 1;
2798 break;
2799 }
2800
2801 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2802 break;
2803 }
2804
7f511862 2805 if (group >= ngroups)
bfff6873
LC		 2806 ret = 1;
2807
2808 if (!ret) {
2809 timeout = jiffies;
2810 ret = ext4_init_inode_table(sb, group,
2811 elr->lr_timeout ? 0 : 1);
2812 if (elr->lr_timeout == 0) {
51ce6511
LC		 2813 timeout = (jiffies - timeout) *
2814 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2815 elr->lr_timeout = timeout;
2816 }
2817 elr->lr_next_sched = jiffies + elr->lr_timeout;
2818 elr->lr_next_group = group + 1;
2819 }
bfff6873
LC		 2820 return ret;
2821}
2822
2823/*
2824 * Remove lr_request from the list_request and free the
4ed5c033 2825 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2826 */
2827static void ext4_remove_li_request(struct ext4_li_request *elr)
2828{
2829 struct ext4_sb_info *sbi;
2830
2831 if (!elr)
2832 return;
2833
2834 sbi = elr->lr_sbi;
2835
2836 list_del(&elr->lr_request);
2837 sbi->s_li_request = NULL;
2838 kfree(elr);
2839}
2840
2841static void ext4_unregister_li_request(struct super_block *sb)
2842{
1bb933fb
LC		 2843 mutex_lock(&ext4_li_mtx);
2844 if (!ext4_li_info) {
2845 mutex_unlock(&ext4_li_mtx);
bfff6873 2846 return;
1bb933fb 2847 }
bfff6873
LC		 2848
2849 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2850 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2851 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2852 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2853}
2854
8f1f7453
ES		 2855static struct task_struct *ext4_lazyinit_task;
2856
bfff6873
LC		 2857/*
2858 * This is the function where ext4lazyinit thread lives. It walks
2859 * through the request list searching for next scheduled filesystem.
2860 * When such a fs is found, run the lazy initialization request
2861 * (ext4_rn_li_request) and keep track of the time spend in this
2862 * function. Based on that time we compute next schedule time of
2863 * the request. When walking through the list is complete, compute
2864 * next waking time and put itself into sleep.
2865 */
2866static int ext4_lazyinit_thread(void *arg)
2867{
2868 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2869 struct list_head *pos, *n;
2870 struct ext4_li_request *elr;
4ed5c033 2871 unsigned long next_wakeup, cur;
bfff6873
LC		 2872
2873 BUG_ON(NULL == eli);
2874
bfff6873
LC		 2875cont_thread:
2876 while (true) {
2877 next_wakeup = MAX_JIFFY_OFFSET;
2878
2879 mutex_lock(&eli->li_list_mtx);
2880 if (list_empty(&eli->li_request_list)) {
2881 mutex_unlock(&eli->li_list_mtx);
2882 goto exit_thread;
2883 }
bfff6873 2884 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 2885 int err = 0;
2886 int progress = 0;
bfff6873
LC		 2887 elr = list_entry(pos, struct ext4_li_request,
2888 lr_request);
2889
e22834f0
DM		 2890 if (time_before(jiffies, elr->lr_next_sched)) {
2891 if (time_before(elr->lr_next_sched, next_wakeup))
2892 next_wakeup = elr->lr_next_sched;
2893 continue;
2894 }
2895 if (down_read_trylock(&elr->lr_super->s_umount)) {
2896 if (sb_start_write_trylock(elr->lr_super)) {
2897 progress = 1;
2898 /*
2899 * We hold sb->s_umount, sb can not
2900 * be removed from the list, it is
2901 * now safe to drop li_list_mtx
2902 */
2903 mutex_unlock(&eli->li_list_mtx);
2904 err = ext4_run_li_request(elr);
2905 sb_end_write(elr->lr_super);
2906 mutex_lock(&eli->li_list_mtx);
2907 n = pos->next;
b2c78cd0 2908 }
e22834f0
DM		 2909 up_read((&elr->lr_super->s_umount));
2910 }
2911 /* error, remove the lazy_init job */
2912 if (err) {
2913 ext4_remove_li_request(elr);
2914 continue;
2915 }
2916 if (!progress) {
2917 elr->lr_next_sched = jiffies +
2918 (prandom_u32()
2919 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 2920 }
bfff6873
LC		 2921 if (time_before(elr->lr_next_sched, next_wakeup))
2922 next_wakeup = elr->lr_next_sched;
2923 }
2924 mutex_unlock(&eli->li_list_mtx);
2925
a0acae0e 2926 try_to_freeze();
bfff6873 2927
4ed5c033
LC		 2928 cur = jiffies;
2929 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2930 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2931 cond_resched();
2932 continue;
2933 }
2934
4ed5c033
LC		 2935 schedule_timeout_interruptible(next_wakeup - cur);
2936
8f1f7453
ES		 2937 if (kthread_should_stop()) {
2938 ext4_clear_request_list();
2939 goto exit_thread;
2940 }
bfff6873
LC		 2941 }
2942
2943exit_thread:
2944 /*
2945 * It looks like the request list is empty, but we need
2946 * to check it under the li_list_mtx lock, to prevent any
2947 * additions into it, and of course we should lock ext4_li_mtx
2948 * to atomically free the list and ext4_li_info, because at
2949 * this point another ext4 filesystem could be registering
2950 * new one.
2951 */
2952 mutex_lock(&ext4_li_mtx);
2953 mutex_lock(&eli->li_list_mtx);
2954 if (!list_empty(&eli->li_request_list)) {
2955 mutex_unlock(&eli->li_list_mtx);
2956 mutex_unlock(&ext4_li_mtx);
2957 goto cont_thread;
2958 }
2959 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2960 kfree(ext4_li_info);
2961 ext4_li_info = NULL;
2962 mutex_unlock(&ext4_li_mtx);
2963
2964 return 0;
2965}
2966
2967static void ext4_clear_request_list(void)
2968{
2969 struct list_head *pos, *n;
2970 struct ext4_li_request *elr;
2971
2972 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2973 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2974 elr = list_entry(pos, struct ext4_li_request,
2975 lr_request);
2976 ext4_remove_li_request(elr);
2977 }
2978 mutex_unlock(&ext4_li_info->li_list_mtx);
2979}
2980
2981static int ext4_run_lazyinit_thread(void)
2982{
8f1f7453
ES		 2983 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2984 ext4_li_info, "ext4lazyinit");
2985 if (IS_ERR(ext4_lazyinit_task)) {
2986 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2987 ext4_clear_request_list();
bfff6873
LC		 2988 kfree(ext4_li_info);
2989 ext4_li_info = NULL;
92b97816 2990 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2991 "initialization thread\n",
2992 err);
2993 return err;
2994 }
2995 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2996 return 0;
2997}
2998
2999/*
3000 * Check whether it make sense to run itable init. thread or not.
3001 * If there is at least one uninitialized inode table, return
3002 * corresponding group number, else the loop goes through all
3003 * groups and return total number of groups.
3004 */
3005static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3006{
3007 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3008 struct ext4_group_desc *gdp = NULL;
3009
3010 for (group = 0; group < ngroups; group++) {
3011 gdp = ext4_get_group_desc(sb, group, NULL);
3012 if (!gdp)
3013 continue;
3014
3015 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3016 break;
3017 }
3018
3019 return group;
3020}
3021
3022static int ext4_li_info_new(void)
3023{
3024 struct ext4_lazy_init *eli = NULL;
3025
3026 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3027 if (!eli)
3028 return -ENOMEM;
3029
bfff6873
LC		 3030 INIT_LIST_HEAD(&eli->li_request_list);
3031 mutex_init(&eli->li_list_mtx);
3032
bfff6873
LC		 3033 eli->li_state |= EXT4_LAZYINIT_QUIT;
3034
3035 ext4_li_info = eli;
3036
3037 return 0;
3038}
3039
3040static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3041 ext4_group_t start)
3042{
3043 struct ext4_sb_info *sbi = EXT4_SB(sb);
3044 struct ext4_li_request *elr;
bfff6873
LC		 3045
3046 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3047 if (!elr)
3048 return NULL;
3049
3050 elr->lr_super = sb;
3051 elr->lr_sbi = sbi;
3052 elr->lr_next_group = start;
3053
3054 /*
3055 * Randomize first schedule time of the request to
3056 * spread the inode table initialization requests
3057 * better.
3058 */
dd1f723b
TT		 3059 elr->lr_next_sched = jiffies + (prandom_u32() %
3060 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3061 return elr;
3062}
3063
7f511862
TT		 3064int ext4_register_li_request(struct super_block *sb,
3065 ext4_group_t first_not_zeroed)
bfff6873
LC		 3066{
3067 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3068 struct ext4_li_request *elr = NULL;
bfff6873 3069 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3070 int ret = 0;
bfff6873 3071
7f511862 3072 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3073 if (sbi->s_li_request != NULL) {
3074 /*
3075 * Reset timeout so it can be computed again, because
3076 * s_li_wait_mult might have changed.
3077 */
3078 sbi->s_li_request->lr_timeout = 0;
7f511862 3079 goto out;
51ce6511 3080 }
bfff6873
LC		 3081
3082 if (first_not_zeroed == ngroups ||
3083 (sb->s_flags & MS_RDONLY) ||
55ff3840 3084 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3085 goto out;
bfff6873
LC		 3086
3087 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3088 if (!elr) {
3089 ret = -ENOMEM;
3090 goto out;
3091 }
bfff6873
LC		 3092
3093 if (NULL == ext4_li_info) {
3094 ret = ext4_li_info_new();
3095 if (ret)
3096 goto out;
3097 }
3098
3099 mutex_lock(&ext4_li_info->li_list_mtx);
3100 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3101 mutex_unlock(&ext4_li_info->li_list_mtx);
3102
3103 sbi->s_li_request = elr;
46e4690b
TM		 3104 /*
3105 * set elr to NULL here since it has been inserted to
3106 * the request_list and the removal and free of it is
3107 * handled by ext4_clear_request_list from now on.
3108 */
3109 elr = NULL;
bfff6873
LC		 3110
3111 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3112 ret = ext4_run_lazyinit_thread();
3113 if (ret)
3114 goto out;
3115 }
bfff6873 3116out:
beed5ecb
NK		 3117 mutex_unlock(&ext4_li_mtx);
3118 if (ret)
bfff6873 3119 kfree(elr);
bfff6873
LC		 3120 return ret;
3121}
3122
3123/*
3124 * We do not need to lock anything since this is called on
3125 * module unload.
3126 */
3127static void ext4_destroy_lazyinit_thread(void)
3128{
3129 /*
3130 * If thread exited earlier
3131 * there's nothing to be done.
3132 */
8f1f7453 3133 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3134 return;
3135
8f1f7453 3136 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3137}
3138
25ed6e8a
DW		 3139static int set_journal_csum_feature_set(struct super_block *sb)
3140{
3141 int ret = 1;
3142 int compat, incompat;
3143 struct ext4_sb_info *sbi = EXT4_SB(sb);
3144
9aa5d32b 3145 if (ext4_has_metadata_csum(sb)) {
db9ee220 3146 /* journal checksum v3 */
25ed6e8a 3147 compat = 0;
db9ee220 3148 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3149 } else {
3150 /* journal checksum v1 */
3151 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3152 incompat = 0;
3153 }
3154
feb8c6d3
DW		 3155 jbd2_journal_clear_features(sbi->s_journal,
3156 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3157 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3158 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3159 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3160 ret = jbd2_journal_set_features(sbi->s_journal,
3161 compat, 0,
3162 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3163 incompat);
3164 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3165 ret = jbd2_journal_set_features(sbi->s_journal,
3166 compat, 0,
3167 incompat);
3168 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3169 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3170 } else {
feb8c6d3
DW		 3171 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3172 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3173 }
3174
3175 return ret;
3176}
3177
952fc18e
TT		 3178/*
3179 * Note: calculating the overhead so we can be compatible with
3180 * historical BSD practice is quite difficult in the face of
3181 * clusters/bigalloc. This is because multiple metadata blocks from
3182 * different block group can end up in the same allocation cluster.
3183 * Calculating the exact overhead in the face of clustered allocation
3184 * requires either O(all block bitmaps) in memory or O(number of block
3185 * groups**2) in time. We will still calculate the superblock for
3186 * older file systems --- and if we come across with a bigalloc file
3187 * system with zero in s_overhead_clusters the estimate will be close to
3188 * correct especially for very large cluster sizes --- but for newer
3189 * file systems, it's better to calculate this figure once at mkfs
3190 * time, and store it in the superblock. If the superblock value is
3191 * present (even for non-bigalloc file systems), we will use it.
3192 */
3193static int count_overhead(struct super_block *sb, ext4_group_t grp,
3194 char *buf)
3195{
3196 struct ext4_sb_info *sbi = EXT4_SB(sb);
3197 struct ext4_group_desc *gdp;
3198 ext4_fsblk_t first_block, last_block, b;
3199 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3200 int s, j, count = 0;
3201
e2b911c5 3202 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3203 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3204 sbi->s_itb_per_group + 2);
3205
952fc18e
TT		 3206 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3207 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3208 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3209 for (i = 0; i < ngroups; i++) {
3210 gdp = ext4_get_group_desc(sb, i, NULL);
3211 b = ext4_block_bitmap(sb, gdp);
3212 if (b >= first_block && b <= last_block) {
3213 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3214 count++;
3215 }
3216 b = ext4_inode_bitmap(sb, gdp);
3217 if (b >= first_block && b <= last_block) {
3218 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3219 count++;
3220 }
3221 b = ext4_inode_table(sb, gdp);
3222 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3223 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3224 int c = EXT4_B2C(sbi, b - first_block);
3225 ext4_set_bit(c, buf);
3226 count++;
3227 }
3228 if (i != grp)
3229 continue;
3230 s = 0;
3231 if (ext4_bg_has_super(sb, grp)) {
3232 ext4_set_bit(s++, buf);
3233 count++;
3234 }
c48ae41b
TT		 3235 j = ext4_bg_num_gdb(sb, grp);
3236 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3237 ext4_error(sb, "Invalid number of block group "
3238 "descriptor blocks: %d", j);
3239 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3240 }
c48ae41b
TT		 3241 count += j;
3242 for (; j > 0; j--)
3243 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3244 }
3245 if (!count)
3246 return 0;
3247 return EXT4_CLUSTERS_PER_GROUP(sb) -
3248 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3249}
3250
3251/*
3252 * Compute the overhead and stash it in sbi->s_overhead
3253 */
3254int ext4_calculate_overhead(struct super_block *sb)
3255{
3256 struct ext4_sb_info *sbi = EXT4_SB(sb);
3257 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3258 struct inode *j_inode;
3259 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3260 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3261 ext4_fsblk_t overhead = 0;
4fdb5543 3262 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3263
952fc18e
TT		 3264 if (!buf)
3265 return -ENOMEM;
3266
3267 /*
3268 * Compute the overhead (FS structures). This is constant
3269 * for a given filesystem unless the number of block groups
3270 * changes so we cache the previous value until it does.
3271 */
3272
3273 /*
3274 * All of the blocks before first_data_block are overhead
3275 */
3276 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3277
3278 /*
3279 * Add the overhead found in each block group
3280 */
3281 for (i = 0; i < ngroups; i++) {
3282 int blks;
3283
3284 blks = count_overhead(sb, i, buf);
3285 overhead += blks;
3286 if (blks)
3287 memset(buf, 0, PAGE_SIZE);
3288 cond_resched();
3289 }
3c816ded
EW		 3290
3291 /*
3292 * Add the internal journal blocks whether the journal has been
3293 * loaded or not
3294 */
b003b524 3295 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3296 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3297 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3298 j_inode = ext4_get_journal_inode(sb, j_inum);
3299 if (j_inode) {
3300 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3301 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3302 iput(j_inode);
3303 } else {
3304 ext4_msg(sb, KERN_ERR, "can't get journal size");
3305 }
3306 }
952fc18e
TT		 3307 sbi->s_overhead = overhead;
3308 smp_wmb();
3309 free_page((unsigned long) buf);
3310 return 0;
3311}
3312
b5799018 3313static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3314{
3315 ext4_fsblk_t resv_clusters;
b5799018 3316 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3317
30fac0f7
JK		 3318 /*
3319 * There's no need to reserve anything when we aren't using extents.
3320 * The space estimates are exact, there are no unwritten extents,
3321 * hole punching doesn't need new metadata... This is needed especially
3322 * to keep ext2/3 backward compatibility.
3323 */
e2b911c5 3324 if (!ext4_has_feature_extents(sb))
b5799018 3325 return;
27dd4385
LC		 3326 /*
3327 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3328 * This should cover the situations where we can not afford to run
3329 * out of space like for example punch hole, or converting
556615dc 3330 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3331 * allocation would require 1, or 2 blocks, higher numbers are
3332 * very rare.
3333 */
b5799018
TT		 3334 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3335 sbi->s_cluster_bits);
27dd4385
LC		 3336
3337 do_div(resv_clusters, 50);
3338 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3339
b5799018 3340 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3341}
3342
2b2d6d01 3343static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3344{
d4c402d9 3345 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3346 struct buffer_head *bh;
617ba13b 3347 struct ext4_super_block *es = NULL;
5aee0f8a 3348 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3349 ext4_fsblk_t block;
3350 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3351 ext4_fsblk_t logical_sb_block;
ac27a0ec 3352 unsigned long offset = 0;
ac27a0ec
DK		 3353 unsigned long journal_devnum = 0;
3354 unsigned long def_mount_opts;
3355 struct inode *root;
0390131b 3356 const char *descr;
dcc7dae3 3357 int ret = -ENOMEM;
281b5995 3358 int blocksize, clustersize;
4ec11028
TT		 3359 unsigned int db_count;
3360 unsigned int i;
281b5995 3361 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3362 __u64 blocks_count;
07aa2ea1 3363 int err = 0;
b3881f74 3364 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3365 ext4_group_t first_not_zeroed;
ac27a0ec 3366
5aee0f8a
TT		 3367 if ((data && !orig_data) || !sbi)
3368 goto out_free_base;
705895b6 3369
73eb7eae
SF		 3370 if (!userns_mounts && !capable(CAP_SYS_ADMIN)) {
3371 ret = -EPERM;
3372 goto out_free_base;
3373 }
3374
705895b6
PE		 3375 sbi->s_blockgroup_lock =
3376 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3377 if (!sbi->s_blockgroup_lock)
3378 goto out_free_base;
3379
ac27a0ec 3380 sb->s_fs_info = sbi;
2c0544b2 3381 sbi->s_sb = sb;
240799cd 3382 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3383 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3384 if (sb->s_bdev->bd_part)
3385 sbi->s_sectors_written_start =
3386 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3387
9f6200bb 3388 /* Cleanup superblock name */
ec3904dc 3389 strreplace(sb->s_id, '/', '!');
9f6200bb 3390
07aa2ea1 3391 /* -EINVAL is default */
dcc7dae3 3392 ret = -EINVAL;
617ba13b 3393 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3394 if (!blocksize) {
b31e1552 3395 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3396 goto out_fail;
3397 }
3398
3399 /*
617ba13b 3400 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3401 * block sizes. We need to calculate the offset from buffer start.
3402 */
617ba13b 3403 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3404 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3405 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3406 } else {
70bbb3e0 3407 logical_sb_block = sb_block;
ac27a0ec
DK		 3408 }
3409
a8ac900b 3410 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3411 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3412 goto out_fail;
3413 }
3414 /*
3415 * Note: s_es must be initialized as soon as possible because
617ba13b 3416 * some ext4 macro-instructions depend on its value
ac27a0ec 3417 */
2716b802 3418 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3419 sbi->s_es = es;
3420 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3421 if (sb->s_magic != EXT4_SUPER_MAGIC)
3422 goto cantfind_ext4;
afc32f7e 3423 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3424
feb0ab32 3425 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3426 if (ext4_has_feature_metadata_csum(sb) &&
3427 ext4_has_feature_gdt_csum(sb))
363307e6 3428 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3429 "redundant flags; please run fsck.");
3430
d25425f8
DW		 3431 /* Check for a known checksum algorithm */
3432 if (!ext4_verify_csum_type(sb, es)) {
3433 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3434 "unknown checksum algorithm.");
3435 silent = 1;
3436 goto cantfind_ext4;
3437 }
3438
0441984a 3439 /* Load the checksum driver */
e2b911c5 3440 if (ext4_has_feature_metadata_csum(sb)) {
0441984a
DW		 3441 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3442 if (IS_ERR(sbi->s_chksum_driver)) {
3443 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3444 ret = PTR_ERR(sbi->s_chksum_driver);
3445 sbi->s_chksum_driver = NULL;
3446 goto failed_mount;
3447 }
3448 }
3449
a9c47317
DW		 3450 /* Check superblock checksum */
3451 if (!ext4_superblock_csum_verify(sb, es)) {
3452 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3453 "invalid superblock checksum. Run e2fsck?");
3454 silent = 1;
6a797d27 3455 ret = -EFSBADCRC;
a9c47317
DW		 3456 goto cantfind_ext4;
3457 }
3458
3459 /* Precompute checksum seed for all metadata */
e2b911c5 3460 if (ext4_has_feature_csum_seed(sb))
8c81bd8f
DW		 3461 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3462 else if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3463 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3464 sizeof(es->s_uuid));
3465
ac27a0ec
DK		 3466 /* Set defaults before we parse the mount options */
3467 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3468 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3469 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3470 set_opt(sb, DEBUG);
87f26807 3471 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3472 set_opt(sb, GRPID);
617ba13b 3473 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3474 set_opt(sb, NO_UID32);
ea663336 3475 /* xattr user namespace & acls are now defaulted on */
ea663336 3476 set_opt(sb, XATTR_USER);
03010a33 3477#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3478 set_opt(sb, POSIX_ACL);
2e7842b8 3479#endif
98c1a759
DW		 3480 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3481 if (ext4_has_metadata_csum(sb))
3482 set_opt(sb, JOURNAL_CHECKSUM);
3483
617ba13b 3484 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3485 set_opt(sb, JOURNAL_DATA);
617ba13b 3486 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3487 set_opt(sb, ORDERED_DATA);
617ba13b 3488 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3489 set_opt(sb, WRITEBACK_DATA);
617ba13b 3490
acc94812
SF		 3491 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) {
3492 if (!capable(CAP_SYS_ADMIN))
3493 goto failed_mount;
fd8c37ec 3494 set_opt(sb, ERRORS_PANIC);
acc94812 3495 } else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) {
fd8c37ec 3496 set_opt(sb, ERRORS_CONT);
acc94812 3497 } else {
fd8c37ec 3498 set_opt(sb, ERRORS_RO);
acc94812 3499 }
45f1a9c3
DW		 3500 /* block_validity enabled by default; disable with noblock_validity */
3501 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3502 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3503 set_opt(sb, DISCARD);
ac27a0ec 3504
acc94812
SF		 3505 sbi->s_resuid = make_kuid(sb->s_user_ns, le16_to_cpu(es->s_def_resuid));
3506 if (!uid_valid(sbi->s_resuid))
3507 sbi->s_resuid = make_kuid(sb->s_user_ns, EXT4_DEF_RESUID);
3508 sbi->s_resgid = make_kgid(sb->s_user_ns, le16_to_cpu(es->s_def_resgid));
3509 if (!gid_valid(sbi->s_resgid))
3510 sbi->s_resgid = make_kgid(sb->s_user_ns, EXT4_DEF_RESGID);
30773840
TT		 3511 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3512 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3513 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3514
8b67f04a 3515 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3516 set_opt(sb, BARRIER);
ac27a0ec 3517
dd919b98
AK		 3518 /*
3519 * enable delayed allocation by default
3520 * Use -o nodelalloc to turn it off
3521 */
bc0b75f7 3522 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3523 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3524 set_opt(sb, DELALLOC);
dd919b98 3525
51ce6511
LC		 3526 /*
3527 * set default s_li_wait_mult for lazyinit, for the case there is
3528 * no mount option specified.
3529 */
3530 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3531
5aee0f8a
TT		 3532 if (sbi->s_es->s_mount_opts[0]) {
3533 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3534 sizeof(sbi->s_es->s_mount_opts),
3535 GFP_KERNEL);
3536 if (!s_mount_opts)
3537 goto failed_mount;
3538 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3539 &journal_ioprio, 0)) {
3540 ext4_msg(sb, KERN_WARNING,
3541 "failed to parse options in superblock: %s",
3542 s_mount_opts);
3543 }
3544 kfree(s_mount_opts);
8b67f04a 3545 }
5a916be1 3546 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3547 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3548 &journal_ioprio, 0))
ac27a0ec
DK		 3549 goto failed_mount;
3550
56889787
TT		 3551 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3552 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3553 "with data=journal disables delayed "
3554 "allocation and O_DIRECT support!\n");
3555 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3556 ext4_msg(sb, KERN_ERR, "can't mount with "
3557 "both data=journal and delalloc");
3558 goto failed_mount;
3559 }
3560 if (test_opt(sb, DIOREAD_NOLOCK)) {
3561 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3562 "both data=journal and dioread_nolock");
56889787
TT		 3563 goto failed_mount;
3564 }
923ae0ff
RZ		 3565 if (test_opt(sb, DAX)) {
3566 ext4_msg(sb, KERN_ERR, "can't mount with "
3567 "both data=journal and dax");
3568 goto failed_mount;
3569 }
73b92a2a
SK		 3570 if (ext4_has_feature_encrypt(sb)) {
3571 ext4_msg(sb, KERN_WARNING,
3572 "encrypted files will use data=ordered "
3573 "instead of data journaling mode");
3574 }
56889787
TT		 3575 if (test_opt(sb, DELALLOC))
3576 clear_opt(sb, DELALLOC);
001e4a87
TH		 3577 } else {
3578 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3579 }
3580
ac27a0ec 3581 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3582 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3583
617ba13b 3584 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3585 (ext4_has_compat_features(sb) ||
3586 ext4_has_ro_compat_features(sb) ||
3587 ext4_has_incompat_features(sb)))
b31e1552
ES		 3588 ext4_msg(sb, KERN_WARNING,
3589 "feature flags set on rev 0 fs, "
3590 "running e2fsck is recommended");
469108ff 3591
ed3654eb
TT		 3592 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3593 set_opt2(sb, HURD_COMPAT);
e2b911c5 3594 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3595 ext4_msg(sb, KERN_ERR,
3596 "The Hurd can't support 64-bit file systems");
3597 goto failed_mount;
3598 }
3599 }
3600
2035e776
TT		 3601 if (IS_EXT2_SB(sb)) {
3602 if (ext2_feature_set_ok(sb))
3603 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3604 "using the ext4 subsystem");
3605 else {
3606 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3607 "to feature incompatibilities");
3608 goto failed_mount;
3609 }
3610 }
3611
3612 if (IS_EXT3_SB(sb)) {
3613 if (ext3_feature_set_ok(sb))
3614 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3615 "using the ext4 subsystem");
3616 else {
3617 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3618 "to feature incompatibilities");
3619 goto failed_mount;
3620 }
3621 }
3622
ac27a0ec
DK		 3623 /*
3624 * Check feature flags regardless of the revision level, since we
3625 * previously didn't change the revision level when setting the flags,
3626 * so there is a chance incompat flags are set on a rev 0 filesystem.
3627 */
a13fb1a4 3628 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3629 goto failed_mount;
a13fb1a4 3630
261cb20c 3631 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3632 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3633 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3634 ext4_msg(sb, KERN_ERR,
8cdf3372
TT		 3635 "Unsupported filesystem blocksize %d (%d log_block_size)",
3636 blocksize, le32_to_cpu(es->s_log_block_size));
3637 goto failed_mount;
3638 }
3639 if (le32_to_cpu(es->s_log_block_size) >
3640 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3641 ext4_msg(sb, KERN_ERR,
3642 "Invalid log block size: %u",
3643 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3644 goto failed_mount;
3645 }
3646
5b9554dc
TT		 3647 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3648 ext4_msg(sb, KERN_ERR,
3649 "Number of reserved GDT blocks insanely large: %d",
3650 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3651 goto failed_mount;
3652 }
3653
923ae0ff 3654 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
87eefeb4
TK		 3655 err = bdev_dax_supported(sb, blocksize);
3656 if (err)
923ae0ff 3657 goto failed_mount;
923ae0ff
RZ		 3658 }
3659
e2b911c5 3660 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3661 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3662 es->s_encryption_level);
3663 goto failed_mount;
3664 }
3665
ac27a0ec 3666 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3667 /* Validate the filesystem blocksize */
3668 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3669 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3670 blocksize);
ac27a0ec
DK		 3671 goto failed_mount;
3672 }
3673
2b2d6d01 3674 brelse(bh);
70bbb3e0
AM		 3675 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3676 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3677 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3678 if (!bh) {
b31e1552
ES		 3679 ext4_msg(sb, KERN_ERR,
3680 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3681 goto failed_mount;
3682 }
2716b802 3683 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3684 sbi->s_es = es;
617ba13b 3685 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3686 ext4_msg(sb, KERN_ERR,
3687 "Magic mismatch, very weird!");
ac27a0ec
DK		 3688 goto failed_mount;
3689 }
3690 }
3691
e2b911c5 3692 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3693 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3694 has_huge_files);
3695 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3696
617ba13b
MC		 3697 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3698 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3699 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3700 } else {
3701 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3702 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3703 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3704 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3705 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3706 ext4_msg(sb, KERN_ERR,
3707 "unsupported inode size: %d",
2b2d6d01 3708 sbi->s_inode_size);
ac27a0ec
DK		 3709 goto failed_mount;
3710 }
ef7f3835
KS		 3711 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3712 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3713 }
0b8e58a1 3714
0d1ee42f 3715 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3716 if (ext4_has_feature_64bit(sb)) {
8fadc143 3717 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3718 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3719 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3720 ext4_msg(sb, KERN_ERR,
3721 "unsupported descriptor size %lu",
0d1ee42f
AR		 3722 sbi->s_desc_size);
3723 goto failed_mount;
3724 }
3725 } else
3726 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3727
ac27a0ec 3728 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3729 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 3730
617ba13b 3731 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3732 if (sbi->s_inodes_per_block == 0)
617ba13b 3733 goto cantfind_ext4;
cd6bb35b
TT		 3734 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3735 sbi->s_inodes_per_group > blocksize * 8) {
3736 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3737 sbi->s_blocks_per_group);
3738 goto failed_mount;
3739 }
ac27a0ec
DK		 3740 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3741 sbi->s_inodes_per_block;
0d1ee42f 3742 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3743 sbi->s_sbh = bh;
3744 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3745 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3746 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3747
2b2d6d01 3748 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3749 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3750 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3751 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3752 i = le32_to_cpu(es->s_flags);
3753 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3754 sbi->s_hash_unsigned = 3;
3755 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3756#ifdef __CHAR_UNSIGNED__
23301410
TT		 3757 if (!(sb->s_flags & MS_RDONLY))
3758 es->s_flags |=
3759 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3760 sbi->s_hash_unsigned = 3;
f99b2589 3761#else
23301410
TT		 3762 if (!(sb->s_flags & MS_RDONLY))
3763 es->s_flags |=
3764 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3765#endif
23301410 3766 }
f99b2589 3767 }
ac27a0ec 3768
281b5995
TT		 3769 /* Handle clustersize */
3770 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3771 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3772 if (has_bigalloc) {
3773 if (clustersize < blocksize) {
3774 ext4_msg(sb, KERN_ERR,
3775 "cluster size (%d) smaller than "
3776 "block size (%d)", clustersize, blocksize);
3777 goto failed_mount;
3778 }
8cdf3372
TT		 3779 if (le32_to_cpu(es->s_log_cluster_size) >
3780 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3781 ext4_msg(sb, KERN_ERR,
3782 "Invalid log cluster size: %u",
3783 le32_to_cpu(es->s_log_cluster_size));
3784 goto failed_mount;
3785 }
281b5995
TT		 3786 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3787 le32_to_cpu(es->s_log_block_size);
3788 sbi->s_clusters_per_group =
3789 le32_to_cpu(es->s_clusters_per_group);
3790 if (sbi->s_clusters_per_group > blocksize * 8) {
3791 ext4_msg(sb, KERN_ERR,
3792 "#clusters per group too big: %lu",
3793 sbi->s_clusters_per_group);
3794 goto failed_mount;
3795 }
3796 if (sbi->s_blocks_per_group !=
3797 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3798 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3799 "clusters per group (%lu) inconsistent",
3800 sbi->s_blocks_per_group,
3801 sbi->s_clusters_per_group);
3802 goto failed_mount;
3803 }
3804 } else {
3805 if (clustersize != blocksize) {
3806 ext4_warning(sb, "fragment/cluster size (%d) != "
3807 "block size (%d)", clustersize,
3808 blocksize);
3809 clustersize = blocksize;
3810 }
3811 if (sbi->s_blocks_per_group > blocksize * 8) {
3812 ext4_msg(sb, KERN_ERR,
3813 "#blocks per group too big: %lu",
3814 sbi->s_blocks_per_group);
3815 goto failed_mount;
3816 }
3817 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3818 sbi->s_cluster_bits = 0;
ac27a0ec 3819 }
281b5995
TT		 3820 sbi->s_cluster_ratio = clustersize / blocksize;
3821
960fd856
TT		 3822 /* Do we have standard group size of clustersize * 8 blocks ? */
3823 if (sbi->s_blocks_per_group == clustersize << 3)
3824 set_opt2(sb, STD_GROUP_SIZE);
3825
bf43d84b
ES		 3826 /*
3827 * Test whether we have more sectors than will fit in sector_t,
3828 * and whether the max offset is addressable by the page cache.
3829 */
5a9ae68a 3830 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3831 ext4_blocks_count(es));
5a9ae68a 3832 if (err) {
b31e1552 3833 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3834 " too large to mount safely on this system");
ac27a0ec 3835 if (sizeof(sector_t) < 8)
90c699a9 3836 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3837 goto failed_mount;
3838 }
3839
617ba13b
MC		 3840 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3841 goto cantfind_ext4;
e7c95593 3842
0f2ddca6
FTN		 3843 /* check blocks count against device size */
3844 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3845 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3846 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3847 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3848 ext4_blocks_count(es), blocks_count);
3849 goto failed_mount;
3850 }
3851
0b8e58a1
AD		 3852 /*
3853 * It makes no sense for the first data block to be beyond the end
3854 * of the filesystem.
3855 */
3856 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3857 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3858 "block %u is beyond end of filesystem (%llu)",
3859 le32_to_cpu(es->s_first_data_block),
3860 ext4_blocks_count(es));
e7c95593
ES		 3861 goto failed_mount;
3862 }
bd81d8ee
LV		 3863 blocks_count = (ext4_blocks_count(es) -
3864 le32_to_cpu(es->s_first_data_block) +
3865 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3866 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3867 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3868 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3869 "(block count %llu, first data block %u, "
b31e1552 3870 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3871 ext4_blocks_count(es),
3872 le32_to_cpu(es->s_first_data_block),
3873 EXT4_BLOCKS_PER_GROUP(sb));
3874 goto failed_mount;
3875 }
bd81d8ee 3876 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3877 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3878 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3879 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3880 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd
EG		 3881 if (ext4_has_feature_meta_bg(sb)) {
3882 if (le32_to_cpu(es->s_first_meta_bg) >= db_count) {
3883 ext4_msg(sb, KERN_WARNING,
3884 "first meta block group too large: %u "
3885 "(group descriptor block count %u)",
3886 le32_to_cpu(es->s_first_meta_bg), db_count);
3887 goto failed_mount;
3888 }
3889 }
f18a5f21
TT		 3890 sbi->s_group_desc = ext4_kvmalloc(db_count *
3891 sizeof(struct buffer_head *),
3892 GFP_KERNEL);
ac27a0ec 3893 if (sbi->s_group_desc == NULL) {
b31e1552 3894 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3895 ret = -ENOMEM;
ac27a0ec
DK		 3896 goto failed_mount;
3897 }
3898
705895b6 3899 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3900
3901 for (i = 0; i < db_count; i++) {
70bbb3e0 3902 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3903 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3904 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3905 ext4_msg(sb, KERN_ERR,
3906 "can't read group descriptor %d", i);
ac27a0ec
DK		 3907 db_count = i;
3908 goto failed_mount2;
3909 }
3910 }
829fa70d 3911 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 3912 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 3913 ret = -EFSCORRUPTED;
f9ae9cf5 3914 goto failed_mount2;
ac27a0ec 3915 }
772cb7c8 3916
f9ae9cf5 3917 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3918 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3919 spin_lock_init(&sbi->s_next_gen_lock);
3920
04ecddb7
JM		 3921 setup_timer(&sbi->s_err_report, print_daily_error_info,
3922 (unsigned long) sb);
04496411 3923
a75ae78f 3924 /* Register extent status tree shrinker */
eb68d0e2 3925 if (ext4_es_register_shrinker(sbi))
ce7e010a 3926 goto failed_mount3;
ce7e010a 3927
c9de560d 3928 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3929 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3930
f9ae9cf5
TT		 3931 /*
3932 * set up enough so that it can read an inode
3933 */
f6e63f90 3934 sb->s_op = &ext4_sops;
617ba13b
MC		 3935 sb->s_export_op = &ext4_export_ops;
3936 sb->s_xattr = ext4_xattr_handlers;
a7550b30 3937 sb->s_cop = &ext4_cryptops;
ac27a0ec 3938#ifdef CONFIG_QUOTA
617ba13b 3939 sb->dq_op = &ext4_quota_operations;
e2b911c5 3940 if (ext4_has_feature_quota(sb))
1fa5efe3 3941 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3942 else
3943 sb->s_qcop = &ext4_qctl_operations;
689c958c 3944 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 3945#endif
f2fa2ffc
AK		 3946 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3947
ac27a0ec 3948 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3949 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3950
3951 sb->s_root = NULL;
3952
3953 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 3954 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 3955
e2b911c5 3956 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
c5e06d10 3957 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3958 goto failed_mount3a;
c5e06d10 3959
ac27a0ec
DK		 3960 /*
3961 * The first inode we look at is the journal inode. Don't try
3962 * root first: it may be modified in the journal!
3963 */
e2b911c5 3964 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
617ba13b 3965 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3966 goto failed_mount3a;
0390131b 3967 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
e2b911c5 3968 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 3969 ext4_msg(sb, KERN_ERR, "required journal recovery "
3970 "suppressed and not mounted read-only");
744692dc 3971 goto failed_mount_wq;
ac27a0ec 3972 } else {
1e381f60
DM		 3973 /* Nojournal mode, all journal mount options are illegal */
3974 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3975 ext4_msg(sb, KERN_ERR, "can't mount with "
3976 "journal_checksum, fs mounted w/o journal");
3977 goto failed_mount_wq;
3978 }
3979 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3980 ext4_msg(sb, KERN_ERR, "can't mount with "
3981 "journal_async_commit, fs mounted w/o journal");
3982 goto failed_mount_wq;
3983 }
3984 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3985 ext4_msg(sb, KERN_ERR, "can't mount with "
3986 "commit=%lu, fs mounted w/o journal",
3987 sbi->s_commit_interval / HZ);
3988 goto failed_mount_wq;
3989 }
3990 if (EXT4_MOUNT_DATA_FLAGS &
3991 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3992 ext4_msg(sb, KERN_ERR, "can't mount with "
3993 "data=, fs mounted w/o journal");
3994 goto failed_mount_wq;
3995 }
3996 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3997 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 3998 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3999 sbi->s_journal = NULL;
4000 needs_recovery = 0;
4001 goto no_journal;
ac27a0ec
DK		 4002 }
4003
e2b911c5 4004 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4005 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4006 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4007 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4008 goto failed_mount_wq;
eb40a09c
JS		 4009 }
4010
25ed6e8a
DW		 4011 if (!set_journal_csum_feature_set(sb)) {
4012 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4013 "feature set");
4014 goto failed_mount_wq;
d4da6c9c 4015 }
818d276c 4016
ac27a0ec
DK		 4017 /* We have now updated the journal if required, so we can
4018 * validate the data journaling mode. */
4019 switch (test_opt(sb, DATA_FLAGS)) {
4020 case 0:
4021 /* No mode set, assume a default based on the journal
63f57933
AM		 4022 * capabilities: ORDERED_DATA if the journal can
4023 * cope, else JOURNAL_DATA
4024 */
dab291af
MC		 4025 if (jbd2_journal_check_available_features
4026 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 4027 set_opt(sb, ORDERED_DATA);
ac27a0ec 4028 else
fd8c37ec 4029 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 4030 break;
4031
617ba13b
MC		 4032 case EXT4_MOUNT_ORDERED_DATA:
4033 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4034 if (!jbd2_journal_check_available_features
4035 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4036 ext4_msg(sb, KERN_ERR, "Journal does not support "
4037 "requested data journaling mode");
744692dc 4038 goto failed_mount_wq;
ac27a0ec
DK		 4039 }
4040 default:
4041 break;
4042 }
ab04df78
JK		 4043
4044 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4045 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4046 ext4_msg(sb, KERN_ERR, "can't mount with "
4047 "journal_async_commit in data=ordered mode");
4048 goto failed_mount_wq;
4049 }
4050
b3881f74 4051 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4052
18aadd47
BJ		 4053 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4054
ce7e010a 4055no_journal:
2335d05f
AG		 4056 sbi->s_mb_cache = ext4_xattr_create_cache();
4057 if (!sbi->s_mb_cache) {
4058 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4059 goto failed_mount_wq;
9c191f70
M		 4060 }
4061
e2b911c5 4062 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 4063 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 4064 ext4_msg(sb, KERN_ERR,
4065 "Unsupported blocksize for fs encryption");
4066 goto failed_mount_wq;
4067 }
4068
e2b911c5
DW		 4069 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4070 !ext4_has_feature_encrypt(sb)) {
4071 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4072 ext4_commit_super(sb, 1);
4073 }
4074
952fc18e
TT		 4075 /*
4076 * Get the # of file system overhead blocks from the
4077 * superblock if present.
4078 */
4079 if (es->s_overhead_clusters)
4080 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4081 else {
07aa2ea1
LC		 4082 err = ext4_calculate_overhead(sb);
4083 if (err)
952fc18e
TT		 4084 goto failed_mount_wq;
4085 }
4086
fd89d5f2
TH		 4087 /*
4088 * The maximum number of concurrent works can be high and
4089 * concurrency isn't really necessary. Limit it to 1.
4090 */
2e8fa54e
JK		 4091 EXT4_SB(sb)->rsv_conversion_wq =
4092 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4093 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4094 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4095 ret = -ENOMEM;
2e8fa54e
JK		 4096 goto failed_mount4;
4097 }
4098
ac27a0ec 4099 /*
dab291af 4100 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4101 * so we can safely mount the rest of the filesystem now.
4102 */
4103
1d1fe1ee
DH		 4104 root = ext4_iget(sb, EXT4_ROOT_INO);
4105 if (IS_ERR(root)) {
b31e1552 4106 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4107 ret = PTR_ERR(root);
32a9bb57 4108 root = NULL;
ac27a0ec
DK		 4109 goto failed_mount4;
4110 }
4111 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4112 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4113 iput(root);
ac27a0ec
DK		 4114 goto failed_mount4;
4115 }
48fde701 4116 sb->s_root = d_make_root(root);
1d1fe1ee 4117 if (!sb->s_root) {
b31e1552 4118 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4119 ret = -ENOMEM;
4120 goto failed_mount4;
4121 }
ac27a0ec 4122
7e84b621
ES		 4123 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4124 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4125
4126 /* determine the minimum size of new large inodes, if present */
4127 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4128 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4129 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4130 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4131 if (sbi->s_want_extra_isize <
4132 le16_to_cpu(es->s_want_extra_isize))
4133 sbi->s_want_extra_isize =
4134 le16_to_cpu(es->s_want_extra_isize);
4135 if (sbi->s_want_extra_isize <
4136 le16_to_cpu(es->s_min_extra_isize))
4137 sbi->s_want_extra_isize =
4138 le16_to_cpu(es->s_min_extra_isize);
4139 }
4140 }
4141 /* Check if enough inode space is available */
4142 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4143 sbi->s_inode_size) {
4144 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4145 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4146 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4147 "available");
ef7f3835
KS		 4148 }
4149
b5799018 4150 ext4_set_resv_clusters(sb);
27dd4385 4151
6fd058f7
TT		 4152 err = ext4_setup_system_zone(sb);
4153 if (err) {
b31e1552 4154 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4155 "zone (%d)", err);
f9ae9cf5
TT		 4156 goto failed_mount4a;
4157 }
4158
4159 ext4_ext_init(sb);
4160 err = ext4_mb_init(sb);
4161 if (err) {
4162 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4163 err);
dcf2d804 4164 goto failed_mount5;
c2774d84
AK		 4165 }
4166
d5e03cbb
TT		 4167 block = ext4_count_free_clusters(sb);
4168 ext4_free_blocks_count_set(sbi->s_es,
4169 EXT4_C2B(sbi, block));
908c7f19
TH		 4170 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4171 GFP_KERNEL);
d5e03cbb
TT		 4172 if (!err) {
4173 unsigned long freei = ext4_count_free_inodes(sb);
4174 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4175 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4176 GFP_KERNEL);
d5e03cbb
TT		 4177 }
4178 if (!err)
4179 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4180 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4181 if (!err)
908c7f19
TH		 4182 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4183 GFP_KERNEL);
c8585c6f
DJ		 4184 if (!err)
4185 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4186
d5e03cbb
TT		 4187 if (err) {
4188 ext4_msg(sb, KERN_ERR, "insufficient memory");
4189 goto failed_mount6;
4190 }
4191
e2b911c5 4192 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4193 if (!ext4_fill_flex_info(sb)) {
4194 ext4_msg(sb, KERN_ERR,
4195 "unable to initialize "
4196 "flex_bg meta info!");
4197 goto failed_mount6;
4198 }
4199
bfff6873
LC		 4200 err = ext4_register_li_request(sb, first_not_zeroed);
4201 if (err)
dcf2d804 4202 goto failed_mount6;
bfff6873 4203
b5799018 4204 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4205 if (err)
4206 goto failed_mount7;
3197ebdb 4207
9b2ff357
JK		 4208#ifdef CONFIG_QUOTA
4209 /* Enable quota usage during mount. */
e2b911c5 4210 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
9b2ff357
JK		 4211 err = ext4_enable_quotas(sb);
4212 if (err)
4213 goto failed_mount8;
4214 }
4215#endif /* CONFIG_QUOTA */
4216
617ba13b
MC		 4217 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4218 ext4_orphan_cleanup(sb, es);
4219 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4220 if (needs_recovery) {
b31e1552 4221 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4222 ext4_mark_recovery_complete(sb, es);
4223 }
4224 if (EXT4_SB(sb)->s_journal) {
4225 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4226 descr = " journalled data mode";
4227 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4228 descr = " ordered data mode";
4229 else
4230 descr = " writeback data mode";
4231 } else
4232 descr = "out journal";
4233
79add3a3
LC		 4234 if (test_opt(sb, DISCARD)) {
4235 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4236 if (!blk_queue_discard(q))
4237 ext4_msg(sb, KERN_WARNING,
4238 "mounting with \"discard\" option, but "
4239 "the device does not support discard");
4240 }
4241
e294a537
TT		 4242 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4243 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4244 "Opts: %.*s%s%s", descr,
4245 (int) sizeof(sbi->s_es->s_mount_opts),
4246 sbi->s_es->s_mount_opts,
e294a537 4247 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4248
66e61a9e
TT		 4249 if (es->s_error_count)
4250 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4251
efbed4dc
TT		 4252 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4253 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4254 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4255 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4256
d4c402d9 4257 kfree(orig_data);
a7550b30
JK		 4258#ifdef CONFIG_EXT4_FS_ENCRYPTION
4259 memcpy(sbi->key_prefix, EXT4_KEY_DESC_PREFIX,
4260 EXT4_KEY_DESC_PREFIX_SIZE);
4261 sbi->key_prefix_size = EXT4_KEY_DESC_PREFIX_SIZE;
4262#endif
ac27a0ec
DK		 4263 return 0;
4264
617ba13b 4265cantfind_ext4:
ac27a0ec 4266 if (!silent)
b31e1552 4267 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4268 goto failed_mount;
4269
72ba7450
TT		 4270#ifdef CONFIG_QUOTA
4271failed_mount8:
ebd173be 4272 ext4_unregister_sysfs(sb);
72ba7450 4273#endif
dcf2d804
TM		 4274failed_mount7:
4275 ext4_unregister_li_request(sb);
4276failed_mount6:
f9ae9cf5 4277 ext4_mb_release(sb);
d5e03cbb 4278 if (sbi->s_flex_groups)
b93b41d4 4279 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4280 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4281 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4282 percpu_counter_destroy(&sbi->s_dirs_counter);
4283 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4284failed_mount5:
f9ae9cf5
TT		 4285 ext4_ext_release(sb);
4286 ext4_release_system_zone(sb);
4287failed_mount4a:
94bf608a 4288 dput(sb->s_root);
32a9bb57 4289 sb->s_root = NULL;
94bf608a 4290failed_mount4:
b31e1552 4291 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4292 if (EXT4_SB(sb)->rsv_conversion_wq)
4293 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4294failed_mount_wq:
82939d79
JK		 4295 if (sbi->s_mb_cache) {
4296 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4297 sbi->s_mb_cache = NULL;
4298 }
0390131b
FM		 4299 if (sbi->s_journal) {
4300 jbd2_journal_destroy(sbi->s_journal);
4301 sbi->s_journal = NULL;
4302 }
50460fe8 4303failed_mount3a:
d3922a77 4304 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4305failed_mount3:
9105bb14 4306 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4307 if (sbi->s_mmp_tsk)
4308 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4309failed_mount2:
4310 for (i = 0; i < db_count; i++)
4311 brelse(sbi->s_group_desc[i]);
b93b41d4 4312 kvfree(sbi->s_group_desc);
ac27a0ec 4313failed_mount:
0441984a
DW		 4314 if (sbi->s_chksum_driver)
4315 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4316#ifdef CONFIG_QUOTA
a2d4a646 4317 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4318 kfree(sbi->s_qf_names[i]);
4319#endif
617ba13b 4320 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4321 brelse(bh);
4322out_fail:
acc94812 4323 /* sb->s_user_ns will be put when sb is destroyed */
ac27a0ec 4324 sb->s_fs_info = NULL;
f6830165 4325 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4326out_free_base:
ac27a0ec 4327 kfree(sbi);
d4c402d9 4328 kfree(orig_data);
07aa2ea1 4329 return err ? err : ret;
ac27a0ec
DK		 4330}
4331
4332/*
4333 * Setup any per-fs journal parameters now. We'll do this both on
4334 * initial mount, once the journal has been initialised but before we've
4335 * done any recovery; and again on any subsequent remount.
4336 */
617ba13b 4337static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4338{
617ba13b 4339 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4340
30773840
TT		 4341 journal->j_commit_interval = sbi->s_commit_interval;
4342 journal->j_min_batch_time = sbi->s_min_batch_time;
4343 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4344
a931da6a 4345 write_lock(&journal->j_state_lock);
ac27a0ec 4346 if (test_opt(sb, BARRIER))
dab291af 4347 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4348 else
dab291af 4349 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4350 if (test_opt(sb, DATA_ERR_ABORT))
4351 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4352 else
4353 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4354 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4355}
4356
c6cb7e77
EW		 4357static struct inode *ext4_get_journal_inode(struct super_block *sb,
4358 unsigned int journal_inum)
ac27a0ec
DK		 4359{
4360 struct inode *journal_inode;
ac27a0ec 4361
c6cb7e77
EW		 4362 /*
4363 * Test for the existence of a valid inode on disk. Bad things
4364 * happen if we iget() an unused inode, as the subsequent iput()
4365 * will try to delete it.
4366 */
1d1fe1ee
DH		 4367 journal_inode = ext4_iget(sb, journal_inum);
4368 if (IS_ERR(journal_inode)) {
b31e1552 4369 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4370 return NULL;
4371 }
4372 if (!journal_inode->i_nlink) {
4373 make_bad_inode(journal_inode);
4374 iput(journal_inode);
b31e1552 4375 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4376 return NULL;
4377 }
4378
e5f8eab8 4379 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4380 journal_inode, journal_inode->i_size);
1d1fe1ee 4381 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4382 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4383 iput(journal_inode);
4384 return NULL;
4385 }
c6cb7e77
EW		 4386 return journal_inode;
4387}
4388
4389static journal_t *ext4_get_journal(struct super_block *sb,
4390 unsigned int journal_inum)
4391{
4392 struct inode *journal_inode;
4393 journal_t *journal;
4394
4395 BUG_ON(!ext4_has_feature_journal(sb));
4396
4397 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4398 if (!journal_inode)
4399 return NULL;
ac27a0ec 4400
dab291af 4401 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4402 if (!journal) {
b31e1552 4403 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4404 iput(journal_inode);
4405 return NULL;
4406 }
4407 journal->j_private = sb;
617ba13b 4408 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4409 return journal;
4410}
4411
617ba13b 4412static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4413 dev_t j_dev)
4414{
2b2d6d01 4415 struct buffer_head *bh;
ac27a0ec 4416 journal_t *journal;
617ba13b
MC		 4417 ext4_fsblk_t start;
4418 ext4_fsblk_t len;
ac27a0ec 4419 int hblock, blocksize;
617ba13b 4420 ext4_fsblk_t sb_block;
ac27a0ec 4421 unsigned long offset;
2b2d6d01 4422 struct ext4_super_block *es;
ac27a0ec
DK		 4423 struct block_device *bdev;
4424
e2b911c5 4425 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4426
b31e1552 4427 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4428 if (bdev == NULL)
4429 return NULL;
4430
ac27a0ec 4431 blocksize = sb->s_blocksize;
e1defc4f 4432 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4433 if (blocksize < hblock) {
b31e1552
ES		 4434 ext4_msg(sb, KERN_ERR,
4435 "blocksize too small for journal device");
ac27a0ec
DK		 4436 goto out_bdev;
4437 }
4438
617ba13b
MC		 4439 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4440 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4441 set_blocksize(bdev, blocksize);
4442 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4443 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4444 "external journal");
ac27a0ec
DK		 4445 goto out_bdev;
4446 }
4447
2716b802 4448 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4449 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4450 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4451 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4452 ext4_msg(sb, KERN_ERR, "external journal has "
4453 "bad superblock");
ac27a0ec
DK		 4454 brelse(bh);
4455 goto out_bdev;
4456 }
4457
df4763be
DW		 4458 if ((le32_to_cpu(es->s_feature_ro_compat) &
4459 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4460 es->s_checksum != ext4_superblock_csum(sb, es)) {
4461 ext4_msg(sb, KERN_ERR, "external journal has "
4462 "corrupt superblock");
4463 brelse(bh);
4464 goto out_bdev;
4465 }
4466
617ba13b 4467 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4468 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4469 brelse(bh);
4470 goto out_bdev;
4471 }
4472
bd81d8ee 4473 len = ext4_blocks_count(es);
ac27a0ec
DK		 4474 start = sb_block + 1;
4475 brelse(bh); /* we're done with the superblock */
4476
dab291af 4477 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4478 start, len, blocksize);
4479 if (!journal) {
b31e1552 4480 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4481 goto out_bdev;
4482 }
4483 journal->j_private = sb;
dfec8a14 4484 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4485 wait_on_buffer(journal->j_sb_buffer);
4486 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4487 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4488 goto out_journal;
4489 }
4490 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4491 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4492 "user (unsupported) - %d",
ac27a0ec
DK		 4493 be32_to_cpu(journal->j_superblock->s_nr_users));
4494 goto out_journal;
4495 }
617ba13b
MC		 4496 EXT4_SB(sb)->journal_bdev = bdev;
4497 ext4_init_journal_params(sb, journal);
ac27a0ec 4498 return journal;
0b8e58a1 4499
ac27a0ec 4500out_journal:
dab291af 4501 jbd2_journal_destroy(journal);
ac27a0ec 4502out_bdev:
617ba13b 4503 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4504 return NULL;
4505}
4506
617ba13b
MC		 4507static int ext4_load_journal(struct super_block *sb,
4508 struct ext4_super_block *es,
ac27a0ec
DK		 4509 unsigned long journal_devnum)
4510{
4511 journal_t *journal;
4512 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4513 dev_t journal_dev;
4514 int err = 0;
4515 int really_read_only;
4516
e2b911c5 4517 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4518
ac27a0ec
DK		 4519 if (journal_devnum &&
4520 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4521 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4522 "numbers have changed");
ac27a0ec
DK		 4523 journal_dev = new_decode_dev(journal_devnum);
4524 } else
4525 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4526
4527 really_read_only = bdev_read_only(sb->s_bdev);
4528
4529 /*
4530 * Are we loading a blank journal or performing recovery after a
4531 * crash? For recovery, we need to check in advance whether we
4532 * can get read-write access to the device.
4533 */
e2b911c5 4534 if (ext4_has_feature_journal_needs_recovery(sb)) {
ac27a0ec 4535 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4536 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4537 "required on readonly filesystem");
ac27a0ec 4538 if (really_read_only) {
b31e1552
ES		 4539 ext4_msg(sb, KERN_ERR, "write access "
4540 "unavailable, cannot proceed");
ac27a0ec
DK		 4541 return -EROFS;
4542 }
b31e1552
ES		 4543 ext4_msg(sb, KERN_INFO, "write access will "
4544 "be enabled during recovery");
ac27a0ec
DK		 4545 }
4546 }
4547
4548 if (journal_inum && journal_dev) {
b31e1552
ES		 4549 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4550 "and inode journals!");
ac27a0ec
DK		 4551 return -EINVAL;
4552 }
4553
4554 if (journal_inum) {
617ba13b 4555 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4556 return -EINVAL;
4557 } else {
617ba13b 4558 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4559 return -EINVAL;
4560 }
4561
90576c0b 4562 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4563 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4564
e2b911c5 4565 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4566 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4567 if (!err) {
4568 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4569 if (save)
4570 memcpy(save, ((char *) es) +
4571 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4572 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4573 if (save)
4574 memcpy(((char *) es) + EXT4_S_ERR_START,
4575 save, EXT4_S_ERR_LEN);
4576 kfree(save);
4577 }
ac27a0ec
DK		 4578
4579 if (err) {
b31e1552 4580 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4581 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4582 return err;
4583 }
4584
617ba13b
MC		 4585 EXT4_SB(sb)->s_journal = journal;
4586 ext4_clear_journal_err(sb, es);
ac27a0ec 4587
c41303ce 4588 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4589 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4590 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4591
4592 /* Make sure we flush the recovery flag to disk. */
e2d67052 4593 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4594 }
4595
4596 return 0;
4597}
4598
e2d67052 4599static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4600{
e2d67052 4601 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4602 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4603 int error = 0;
ac27a0ec 4604
bdfe0cbd 4605 if (!sbh || block_device_ejected(sb))
c4be0c1d 4606 return error;
71290b36
TT		 4607 /*
4608 * If the file system is mounted read-only, don't update the
4609 * superblock write time. This avoids updating the superblock
4610 * write time when we are mounting the root file system
4611 * read/only but we need to replay the journal; at that point,
4612 * for people who are east of GMT and who make their clock
4613 * tick in localtime for Windows bug-for-bug compatibility,
4614 * the clock is set in the future, and this will cause e2fsck
4615 * to complain and force a full file system check.
4616 */
4617 if (!(sb->s_flags & MS_RDONLY))
4618 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4619 if (sb->s_bdev->bd_part)
4620 es->s_kbytes_written =
4621 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4622 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4623 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4624 else
4625 es->s_kbytes_written =
4626 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4627 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4628 ext4_free_blocks_count_set(es,
57042651
TT		 4629 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4630 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4631 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4632 es->s_free_inodes_count =
4633 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4634 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4635 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4636 ext4_superblock_csum_set(sb);
1566a48a
TT		 4637 if (sync)
4638 lock_buffer(sbh);
4743f839
PKS		 4639 if (buffer_write_io_error(sbh)) {
4640 /*
4641 * Oh, dear. A previous attempt to write the
4642 * superblock failed. This could happen because the
4643 * USB device was yanked out. Or it could happen to
4644 * be a transient write error and maybe the block will
4645 * be remapped. Nothing we can do but to retry the
4646 * write and hope for the best.
4647 */
4648 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4649 "superblock detected");
4650 clear_buffer_write_io_error(sbh);
4651 set_buffer_uptodate(sbh);
4652 }
ac27a0ec 4653 mark_buffer_dirty(sbh);
914258bf 4654 if (sync) {
1566a48a 4655 unlock_buffer(sbh);
564bc402 4656 error = __sync_dirty_buffer(sbh,
70fd7614 4657 test_opt(sb, BARRIER) ? REQ_FUA : REQ_SYNC);
c4be0c1d
TS		 4658 if (error)
4659 return error;
4660
4661 error = buffer_write_io_error(sbh);
4662 if (error) {
b31e1552
ES		 4663 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4664 "superblock");
914258bf
TT		 4665 clear_buffer_write_io_error(sbh);
4666 set_buffer_uptodate(sbh);
4667 }
4668 }
c4be0c1d 4669 return error;
ac27a0ec
DK		 4670}
4671
ac27a0ec
DK		 4672/*
4673 * Have we just finished recovery? If so, and if we are mounting (or
4674 * remounting) the filesystem readonly, then we will end up with a
4675 * consistent fs on disk. Record that fact.
4676 */
2b2d6d01
TT		 4677static void ext4_mark_recovery_complete(struct super_block *sb,
4678 struct ext4_super_block *es)
ac27a0ec 4679{
617ba13b 4680 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4681
e2b911c5 4682 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4683 BUG_ON(journal != NULL);
4684 return;
4685 }
dab291af 4686 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4687 if (jbd2_journal_flush(journal) < 0)
4688 goto out;
4689
e2b911c5 4690 if (ext4_has_feature_journal_needs_recovery(sb) &&
ac27a0ec 4691 sb->s_flags & MS_RDONLY) {
e2b911c5 4692 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4693 ext4_commit_super(sb, 1);
ac27a0ec 4694 }
7ffe1ea8
HK		 4695
4696out:
dab291af 4697 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4698}
4699
4700/*
4701 * If we are mounting (or read-write remounting) a filesystem whose journal
4702 * has recorded an error from a previous lifetime, move that error to the
4703 * main filesystem now.
4704 */
2b2d6d01
TT		 4705static void ext4_clear_journal_err(struct super_block *sb,
4706 struct ext4_super_block *es)
ac27a0ec
DK		 4707{
4708 journal_t *journal;
4709 int j_errno;
4710 const char *errstr;
4711
e2b911c5 4712 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4713
617ba13b 4714 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4715
4716 /*
4717 * Now check for any error status which may have been recorded in the
617ba13b 4718 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4719 */
4720
dab291af 4721 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4722 if (j_errno) {
4723 char nbuf[16];
4724
617ba13b 4725 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4726 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4727 "from previous mount: %s", errstr);
12062ddd 4728 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4729
617ba13b
MC		 4730 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4731 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4732 ext4_commit_super(sb, 1);
ac27a0ec 4733
dab291af 4734 jbd2_journal_clear_err(journal);
d796c52e 4735 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4736 }
4737}
4738
4739/*
4740 * Force the running and committing transactions to commit,
4741 * and wait on the commit.
4742 */
617ba13b 4743int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4744{
4745 journal_t *journal;
ac27a0ec
DK		 4746
4747 if (sb->s_flags & MS_RDONLY)
4748 return 0;
4749
617ba13b 4750 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4751 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4752}
4753
617ba13b 4754static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4755{
14ce0cb4 4756 int ret = 0;
9eddacf9 4757 tid_t target;
06a407f1 4758 bool needs_barrier = false;
8d5d02e6 4759 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4760
9bffad1e 4761 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4762 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4763 /*
4764 * Writeback quota in non-journalled quota case - journalled quota has
4765 * no dirty dquots
4766 */
4767 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4768 /*
4769 * Data writeback is possible w/o journal transaction, so barrier must
4770 * being sent at the end of the function. But we can skip it if
4771 * transaction_commit will do it for us.
4772 */
bda32530
TT		 4773 if (sbi->s_journal) {
4774 target = jbd2_get_latest_transaction(sbi->s_journal);
4775 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4776 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4777 needs_barrier = true;
4778
4779 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4780 if (wait)
4781 ret = jbd2_log_wait_commit(sbi->s_journal,
4782 target);
4783 }
4784 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4785 needs_barrier = true;
06a407f1
DM		 4786 if (needs_barrier) {
4787 int err;
4788 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4789 if (!ret)
4790 ret = err;
0390131b 4791 }
06a407f1
DM		 4792
4793 return ret;
4794}
4795
ac27a0ec
DK		 4796/*
4797 * LVM calls this function before a (read-only) snapshot is created. This
4798 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4799 *
4800 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4801 * state independently. It relies on upper layer to stop all data & metadata
4802 * modifications.
ac27a0ec 4803 */
c4be0c1d 4804static int ext4_freeze(struct super_block *sb)
ac27a0ec 4805{
c4be0c1d
TS		 4806 int error = 0;
4807 journal_t *journal;
ac27a0ec 4808
9ca92389
TT		 4809 if (sb->s_flags & MS_RDONLY)
4810 return 0;
ac27a0ec 4811
9ca92389 4812 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4813
bb044576
TT		 4814 if (journal) {
4815 /* Now we set up the journal barrier. */
4816 jbd2_journal_lock_updates(journal);
ac27a0ec 4817
bb044576
TT		 4818 /*
4819 * Don't clear the needs_recovery flag if we failed to
4820 * flush the journal.
4821 */
4822 error = jbd2_journal_flush(journal);
4823 if (error < 0)
4824 goto out;
c642dc9e
ES		 4825
4826 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4827 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4828 }
9ca92389 4829
9ca92389 4830 error = ext4_commit_super(sb, 1);
6b0310fb 4831out:
bb044576
TT		 4832 if (journal)
4833 /* we rely on upper layer to stop further updates */
4834 jbd2_journal_unlock_updates(journal);
6b0310fb 4835 return error;
ac27a0ec
DK		 4836}
4837
4838/*
4839 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4840 * flag here, even though the filesystem is not technically dirty yet.
4841 */
c4be0c1d 4842static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4843{
9ca92389
TT		 4844 if (sb->s_flags & MS_RDONLY)
4845 return 0;
4846
c642dc9e
ES		 4847 if (EXT4_SB(sb)->s_journal) {
4848 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4849 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4850 }
4851
9ca92389 4852 ext4_commit_super(sb, 1);
c4be0c1d 4853 return 0;
ac27a0ec
DK		 4854}
4855
673c6100
TT		 4856/*
4857 * Structure to save mount options for ext4_remount's benefit
4858 */
4859struct ext4_mount_options {
4860 unsigned long s_mount_opt;
a2595b8a 4861 unsigned long s_mount_opt2;
08cefc7a
EB		 4862 kuid_t s_resuid;
4863 kgid_t s_resgid;
673c6100
TT		 4864 unsigned long s_commit_interval;
4865 u32 s_min_batch_time, s_max_batch_time;
4866#ifdef CONFIG_QUOTA
4867 int s_jquota_fmt;
a2d4a646 4868 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4869#endif
4870};
4871
2b2d6d01 4872static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4873{
2b2d6d01 4874 struct ext4_super_block *es;
617ba13b 4875 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4876 unsigned long old_sb_flags;
617ba13b 4877 struct ext4_mount_options old_opts;
c79d967d 4878 int enable_quota = 0;
8a266467 4879 ext4_group_t g;
b3881f74 4880 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4881 int err = 0;
ac27a0ec 4882#ifdef CONFIG_QUOTA
03dafb5f 4883 int i, j;
ac27a0ec 4884#endif
d4c402d9 4885 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4886
4887 /* Store the original options */
4888 old_sb_flags = sb->s_flags;
4889 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4890 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4891 old_opts.s_resuid = sbi->s_resuid;
4892 old_opts.s_resgid = sbi->s_resgid;
4893 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4894 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4895 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4896#ifdef CONFIG_QUOTA
4897 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4898 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4899 if (sbi->s_qf_names[i]) {
4900 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4901 GFP_KERNEL);
4902 if (!old_opts.s_qf_names[i]) {
4903 for (j = 0; j < i; j++)
4904 kfree(old_opts.s_qf_names[j]);
3e36a163 4905 kfree(orig_data);
03dafb5f
CG		 4906 return -ENOMEM;
4907 }
4908 } else
4909 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4910#endif
b3881f74
TT		 4911 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4912 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4913
661aa520 4914 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4915 err = -EINVAL;
4916 goto restore_opts;
4917 }
4918
6b992ff2 4919 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4920 test_opt(sb, JOURNAL_CHECKSUM)) {
4921 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4922 "during remount not supported; ignoring");
4923 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4924 }
4925
6ae6514b
PS		 4926 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4927 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4928 ext4_msg(sb, KERN_ERR, "can't mount with "
4929 "both data=journal and delalloc");
4930 err = -EINVAL;
4931 goto restore_opts;
4932 }
4933 if (test_opt(sb, DIOREAD_NOLOCK)) {
4934 ext4_msg(sb, KERN_ERR, "can't mount with "
4935 "both data=journal and dioread_nolock");
4936 err = -EINVAL;
4937 goto restore_opts;
4938 }
923ae0ff
RZ		 4939 if (test_opt(sb, DAX)) {
4940 ext4_msg(sb, KERN_ERR, "can't mount with "
4941 "both data=journal and dax");
4942 err = -EINVAL;
4943 goto restore_opts;
4944 }
ab04df78
JK		 4945 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
4946 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4947 ext4_msg(sb, KERN_ERR, "can't mount with "
4948 "journal_async_commit in data=ordered mode");
4949 err = -EINVAL;
4950 goto restore_opts;
4951 }
923ae0ff
RZ		 4952 }
4953
4954 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4955 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4956 "dax flag with busy inodes while remounting");
4957 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4958 }
4959
4ab2f15b 4960 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4961 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4962
4963 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4964 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4965
4966 es = sbi->s_es;
4967
b3881f74 4968 if (sbi->s_journal) {
0390131b 4969 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4970 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4971 }
ac27a0ec 4972
a2fd66d0
TT		 4973 if (*flags & MS_LAZYTIME)
4974 sb->s_flags |= MS_LAZYTIME;
4975
661aa520 4976 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4977 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4978 err = -EROFS;
4979 goto restore_opts;
4980 }
4981
4982 if (*flags & MS_RDONLY) {
38c03b34
TT		 4983 err = sync_filesystem(sb);
4984 if (err < 0)
4985 goto restore_opts;
0f0dd62f
CH		 4986 err = dquot_suspend(sb, -1);
4987 if (err < 0)
c79d967d 4988 goto restore_opts;
c79d967d 4989
ac27a0ec
DK		 4990 /*
4991 * First of all, the unconditional stuff we have to do
4992 * to disable replay of the journal when we next remount
4993 */
4994 sb->s_flags |= MS_RDONLY;
4995
4996 /*
4997 * OK, test if we are remounting a valid rw partition
4998 * readonly, and if so set the rdonly flag and then
4999 * mark the partition as valid again.
5000 */
617ba13b
MC		 5001 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5002 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5003 es->s_state = cpu_to_le16(sbi->s_mount_state);
5004
a63c9eb2 5005 if (sbi->s_journal)
0390131b 5006 ext4_mark_recovery_complete(sb, es);
ac27a0ec 5007 } else {
a13fb1a4 5008 /* Make sure we can mount this feature set readwrite */
e2b911c5 5009 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5010 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5011 err = -EROFS;
5012 goto restore_opts;
5013 }
8a266467
TT		 5014 /*
5015 * Make sure the group descriptor checksums
0b8e58a1 5016 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5017 */
5018 for (g = 0; g < sbi->s_groups_count; g++) {
5019 struct ext4_group_desc *gdp =
5020 ext4_get_group_desc(sb, g, NULL);
5021
feb0ab32 5022 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5023 ext4_msg(sb, KERN_ERR,
5024 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5025 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5026 le16_to_cpu(gdp->bg_checksum));
6a797d27 5027 err = -EFSBADCRC;
8a266467
TT		 5028 goto restore_opts;
5029 }
5030 }
5031
ead6596b
ES		 5032 /*
5033 * If we have an unprocessed orphan list hanging
5034 * around from a previously readonly bdev mount,
5035 * require a full umount/remount for now.
5036 */
5037 if (es->s_last_orphan) {
b31e1552 5038 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5039 "remount RDWR because of unprocessed "
5040 "orphan inode list. Please "
b31e1552 5041 "umount/remount instead");
ead6596b
ES		 5042 err = -EINVAL;
5043 goto restore_opts;
5044 }
5045
ac27a0ec
DK		 5046 /*
5047 * Mounting a RDONLY partition read-write, so reread
5048 * and store the current valid flag. (It may have
5049 * been changed by e2fsck since we originally mounted
5050 * the partition.)
5051 */
0390131b
FM		 5052 if (sbi->s_journal)
5053 ext4_clear_journal_err(sb, es);
ac27a0ec 5054 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 5055 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 5056 sb->s_flags &= ~MS_RDONLY;
e2b911c5 5057 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5058 if (ext4_multi_mount_protect(sb,
5059 le64_to_cpu(es->s_mmp_block))) {
5060 err = -EROFS;
5061 goto restore_opts;
5062 }
c79d967d 5063 enable_quota = 1;
ac27a0ec
DK		 5064 }
5065 }
bfff6873
LC		 5066
5067 /*
5068 * Reinitialize lazy itable initialization thread based on
5069 * current settings
5070 */
5071 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5072 ext4_unregister_li_request(sb);
5073 else {
5074 ext4_group_t first_not_zeroed;
5075 first_not_zeroed = ext4_has_uninit_itable(sb);
5076 ext4_register_li_request(sb, first_not_zeroed);
5077 }
5078
6fd058f7 5079 ext4_setup_system_zone(sb);
d096ad0f 5080 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5081 ext4_commit_super(sb, 1);
0390131b 5082
ac27a0ec
DK		 5083#ifdef CONFIG_QUOTA
5084 /* Release old quota file names */
a2d4a646 5085 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5086 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5087 if (enable_quota) {
5088 if (sb_any_quota_suspended(sb))
5089 dquot_resume(sb, -1);
e2b911c5 5090 else if (ext4_has_feature_quota(sb)) {
7c319d32 5091 err = ext4_enable_quotas(sb);
07724f98 5092 if (err)
7c319d32 5093 goto restore_opts;
7c319d32
AK		 5094 }
5095 }
ac27a0ec 5096#endif
d4c402d9 5097
a26f4992 5098 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5099 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5100 kfree(orig_data);
ac27a0ec 5101 return 0;
0b8e58a1 5102
ac27a0ec
DK		 5103restore_opts:
5104 sb->s_flags = old_sb_flags;
5105 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5106 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5107 sbi->s_resuid = old_opts.s_resuid;
5108 sbi->s_resgid = old_opts.s_resgid;
5109 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5110 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5111 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5112#ifdef CONFIG_QUOTA
5113 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5114 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5115 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5116 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5117 }
5118#endif
d4c402d9 5119 kfree(orig_data);
ac27a0ec
DK		 5120 return err;
5121}
5122
689c958c
LX		 5123#ifdef CONFIG_QUOTA
5124static int ext4_statfs_project(struct super_block *sb,
5125 kprojid_t projid, struct kstatfs *buf)
5126{
5127 struct kqid qid;
5128 struct dquot *dquot;
5129 u64 limit;
5130 u64 curblock;
5131
5132 qid = make_kqid_projid(projid);
5133 dquot = dqget(sb, qid);
5134 if (IS_ERR(dquot))
5135 return PTR_ERR(dquot);
5136 spin_lock(&dq_data_lock);
5137
5138 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5139 dquot->dq_dqb.dqb_bsoftlimit :
5140 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5141 if (limit && buf->f_blocks > limit) {
5142 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5143 buf->f_blocks = limit;
5144 buf->f_bfree = buf->f_bavail =
5145 (buf->f_blocks > curblock) ?
5146 (buf->f_blocks - curblock) : 0;
5147 }
5148
5149 limit = dquot->dq_dqb.dqb_isoftlimit ?
5150 dquot->dq_dqb.dqb_isoftlimit :
5151 dquot->dq_dqb.dqb_ihardlimit;
5152 if (limit && buf->f_files > limit) {
5153 buf->f_files = limit;
5154 buf->f_ffree =
5155 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5156 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5157 }
5158
5159 spin_unlock(&dq_data_lock);
5160 dqput(dquot);
5161 return 0;
5162}
5163#endif
5164
2b2d6d01 5165static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5166{
5167 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5168 struct ext4_sb_info *sbi = EXT4_SB(sb);
5169 struct ext4_super_block *es = sbi->s_es;
27dd4385 5170 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5171 u64 fsid;
d02a9391 5172 s64 bfree;
27dd4385 5173 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5174
952fc18e
TT		 5175 if (!test_opt(sb, MINIX_DF))
5176 overhead = sbi->s_overhead;
ac27a0ec 5177
617ba13b 5178 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5179 buf->f_bsize = sb->s_blocksize;
b72f78cb 5180 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5181 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5182 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5183 /* prevent underflow in case that few free space is available */
57042651 5184 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5185 buf->f_bavail = buf->f_bfree -
5186 (ext4_r_blocks_count(es) + resv_blocks);
5187 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5188 buf->f_bavail = 0;
5189 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5190 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5191 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5192 fsid = le64_to_cpup((void *)es->s_uuid) ^
5193 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5194 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5195 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5196
689c958c
LX		 5197#ifdef CONFIG_QUOTA
5198 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5199 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5200 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5201#endif
ac27a0ec
DK		 5202 return 0;
5203}
5204
0b8e58a1
AD		 5205/* Helper function for writing quotas on sync - we need to start transaction
5206 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5207 * Process 1 Process 2
617ba13b 5208 * ext4_create() quota_sync()
a269eb18 5209 * jbd2_journal_start() write_dquot()
871a2931 5210 * dquot_initialize() down(dqio_mutex)
dab291af 5211 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5212 *
5213 */
5214
5215#ifdef CONFIG_QUOTA
5216
5217static inline struct inode *dquot_to_inode(struct dquot *dquot)
5218{
4c376dca 5219 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5220}
5221
617ba13b 5222static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5223{
5224 int ret, err;
5225 handle_t *handle;
5226 struct inode *inode;
5227
5228 inode = dquot_to_inode(dquot);
9924a92a 5229 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5230 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5231 if (IS_ERR(handle))
5232 return PTR_ERR(handle);
5233 ret = dquot_commit(dquot);
617ba13b 5234 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5235 if (!ret)
5236 ret = err;
5237 return ret;
5238}
5239
617ba13b 5240static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5241{
5242 int ret, err;
5243 handle_t *handle;
5244
9924a92a 5245 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5246 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5247 if (IS_ERR(handle))
5248 return PTR_ERR(handle);
5249 ret = dquot_acquire(dquot);
617ba13b 5250 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5251 if (!ret)
5252 ret = err;
5253 return ret;
5254}
5255
617ba13b 5256static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5257{
5258 int ret, err;
5259 handle_t *handle;
5260
9924a92a 5261 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5262 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5263 if (IS_ERR(handle)) {
5264 /* Release dquot anyway to avoid endless cycle in dqput() */
5265 dquot_release(dquot);
ac27a0ec 5266 return PTR_ERR(handle);
9c3013e9 5267 }
ac27a0ec 5268 ret = dquot_release(dquot);
617ba13b 5269 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5270 if (!ret)
5271 ret = err;
5272 return ret;
5273}
5274
617ba13b 5275static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5276{
262b4662
JK		 5277 struct super_block *sb = dquot->dq_sb;
5278 struct ext4_sb_info *sbi = EXT4_SB(sb);
5279
2c8be6b2 5280 /* Are we journaling quotas? */
e2b911c5 5281 if (ext4_has_feature_quota(sb) ||
262b4662 5282 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5283 dquot_mark_dquot_dirty(dquot);
617ba13b 5284 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5285 } else {
5286 return dquot_mark_dquot_dirty(dquot);
5287 }
5288}
5289
617ba13b 5290static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5291{
5292 int ret, err;
5293 handle_t *handle;
5294
5295 /* Data block + inode block */
2b0143b5 5296 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5297 if (IS_ERR(handle))
5298 return PTR_ERR(handle);
5299 ret = dquot_commit_info(sb, type);
617ba13b 5300 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5301 if (!ret)
5302 ret = err;
5303 return ret;
5304}
5305
5306/*
5307 * Turn on quotas during mount time - we need to find
5308 * the quota file and such...
5309 */
617ba13b 5310static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5311{
287a8095
CH		 5312 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5313 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5314}
5315
daf647d2
TT		 5316static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5317{
5318 struct ext4_inode_info *ei = EXT4_I(inode);
5319
5320 /* The first argument of lockdep_set_subclass has to be
5321 * *exactly* the same as the argument to init_rwsem() --- in
5322 * this case, in init_once() --- or lockdep gets unhappy
5323 * because the name of the lock is set using the
5324 * stringification of the argument to init_rwsem().
5325 */
5326 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5327 lockdep_set_subclass(&ei->i_data_sem, subclass);
5328}
5329
ac27a0ec
DK		 5330/*
5331 * Standard function to be called on quota_on
5332 */
617ba13b 5333static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5334 const struct path *path)
ac27a0ec
DK		 5335{
5336 int err;
ac27a0ec
DK		 5337
5338 if (!test_opt(sb, QUOTA))
5339 return -EINVAL;
0623543b 5340
ac27a0ec 5341 /* Quotafile not on the same filesystem? */
d8c9584e 5342 if (path->dentry->d_sb != sb)
ac27a0ec 5343 return -EXDEV;
0623543b
JK		 5344 /* Journaling quota? */
5345 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5346 /* Quotafile not in fs root? */
f00c9e44 5347 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5348 ext4_msg(sb, KERN_WARNING,
5349 "Quota file not on filesystem root. "
5350 "Journaled quota will not work");
2b2d6d01 5351 }
0623543b
JK		 5352
5353 /*
5354 * When we journal data on quota file, we have to flush journal to see
5355 * all updates to the file when we bypass pagecache...
5356 */
0390131b 5357 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5358 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5359 /*
5360 * We don't need to lock updates but journal_flush() could
5361 * otherwise be livelocked...
5362 */
5363 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5364 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5365 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5366 if (err)
7ffe1ea8 5367 return err;
0623543b 5368 }
daf647d2
TT		 5369 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5370 err = dquot_quota_on(sb, type, format_id, path);
5371 if (err)
5372 lockdep_set_quota_inode(path->dentry->d_inode,
5373 I_DATA_SEM_NORMAL);
5374 return err;
ac27a0ec
DK		 5375}
5376
7c319d32
AK		 5377static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5378 unsigned int flags)
5379{
5380 int err;
5381 struct inode *qf_inode;
a2d4a646 5382 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5383 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5384 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5385 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5386 };
5387
e2b911c5 5388 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5389
5390 if (!qf_inums[type])
5391 return -EPERM;
5392
5393 qf_inode = ext4_iget(sb, qf_inums[type]);
5394 if (IS_ERR(qf_inode)) {
5395 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5396 return PTR_ERR(qf_inode);
5397 }
5398
bcb13850
JK		 5399 /* Don't account quota for quota files to avoid recursion */
5400 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5401 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5402 err = dquot_enable(qf_inode, type, format_id, flags);
5403 iput(qf_inode);
daf647d2
TT		 5404 if (err)
5405 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5406
5407 return err;
5408}
5409
5410/* Enable usage tracking for all quota types. */
5411static int ext4_enable_quotas(struct super_block *sb)
5412{
5413 int type, err = 0;
a2d4a646 5414 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5415 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5416 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5417 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5418 };
49da9392
JK		 5419 bool quota_mopt[EXT4_MAXQUOTAS] = {
5420 test_opt(sb, USRQUOTA),
5421 test_opt(sb, GRPQUOTA),
5422 test_opt(sb, PRJQUOTA),
5423 };
7c319d32
AK		 5424
5425 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5426 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5427 if (qf_inums[type]) {
5428 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5429 DQUOT_USAGE_ENABLED |
5430 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 5431 if (err) {
5432 ext4_warning(sb,
72ba7450
TT		 5433 "Failed to enable quota tracking "
5434 "(type=%d, err=%d). Please run "
5435 "e2fsck to fix.", type, err);
7c319d32
AK		 5436 return err;
5437 }
5438 }
5439 }
5440 return 0;
5441}
5442
ca0e05e4
DM		 5443static int ext4_quota_off(struct super_block *sb, int type)
5444{
21f97697
JK		 5445 struct inode *inode = sb_dqopt(sb)->files[type];
5446 handle_t *handle;
5447
87009d86
DM		 5448 /* Force all delayed allocation blocks to be allocated.
5449 * Caller already holds s_umount sem */
5450 if (test_opt(sb, DELALLOC))
ca0e05e4 5451 sync_filesystem(sb);
ca0e05e4 5452
0b268590
AG		 5453 if (!inode)
5454 goto out;
5455
21f97697
JK		 5456 /* Update modification times of quota files when userspace can
5457 * start looking at them */
9924a92a 5458 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5459 if (IS_ERR(handle))
5460 goto out;
eeca7ea1 5461 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5462 ext4_mark_inode_dirty(handle, inode);
5463 ext4_journal_stop(handle);
5464
5465out:
ca0e05e4
DM		 5466 return dquot_quota_off(sb, type);
5467}
5468
ac27a0ec
DK		 5469/* Read data from quotafile - avoid pagecache and such because we cannot afford
5470 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5471 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5472 * we don't have to be afraid of races */
617ba13b 5473static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5474 size_t len, loff_t off)
5475{
5476 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5477 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5478 int offset = off & (sb->s_blocksize - 1);
5479 int tocopy;
5480 size_t toread;
5481 struct buffer_head *bh;
5482 loff_t i_size = i_size_read(inode);
5483
5484 if (off > i_size)
5485 return 0;
5486 if (off+len > i_size)
5487 len = i_size-off;
5488 toread = len;
5489 while (toread > 0) {
5490 tocopy = sb->s_blocksize - offset < toread ?
5491 sb->s_blocksize - offset : toread;
1c215028
TT		 5492 bh = ext4_bread(NULL, inode, blk, 0);
5493 if (IS_ERR(bh))
5494 return PTR_ERR(bh);
ac27a0ec
DK		 5495 if (!bh) /* A hole? */
5496 memset(data, 0, tocopy);
5497 else
5498 memcpy(data, bh->b_data+offset, tocopy);
5499 brelse(bh);
5500 offset = 0;
5501 toread -= tocopy;
5502 data += tocopy;
5503 blk++;
5504 }
5505 return len;
5506}
5507
5508/* Write to quotafile (we know the transaction is already started and has
5509 * enough credits) */
617ba13b 5510static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5511 const char *data, size_t len, loff_t off)
5512{
5513 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5514 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5515 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5516 int retries = 0;
ac27a0ec
DK		 5517 struct buffer_head *bh;
5518 handle_t *handle = journal_current_handle();
5519
0390131b 5520 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5521 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5522 " cancelled because transaction is not started",
9c3013e9
JK		 5523 (unsigned long long)off, (unsigned long long)len);
5524 return -EIO;
5525 }
67eeb568
DM		 5526 /*
5527 * Since we account only one data block in transaction credits,
5528 * then it is impossible to cross a block boundary.
5529 */
5530 if (sb->s_blocksize - offset < len) {
5531 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5532 " cancelled because not block aligned",
5533 (unsigned long long)off, (unsigned long long)len);
5534 return -EIO;
5535 }
5536
c5e298ae
TT		 5537 do {
5538 bh = ext4_bread(handle, inode, blk,
5539 EXT4_GET_BLOCKS_CREATE |
5540 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5541 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5542 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5543 if (IS_ERR(bh))
5544 return PTR_ERR(bh);
67eeb568
DM		 5545 if (!bh)
5546 goto out;
5d601255 5547 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5548 err = ext4_journal_get_write_access(handle, bh);
5549 if (err) {
5550 brelse(bh);
1c215028 5551 return err;
ac27a0ec 5552 }
67eeb568
DM		 5553 lock_buffer(bh);
5554 memcpy(bh->b_data+offset, data, len);
5555 flush_dcache_page(bh->b_page);
5556 unlock_buffer(bh);
62d2b5f2 5557 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5558 brelse(bh);
ac27a0ec 5559out:
67eeb568
DM		 5560 if (inode->i_size < off + len) {
5561 i_size_write(inode, off + len);
617ba13b 5562 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5563 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5564 }
67eeb568 5565 return len;
ac27a0ec
DK		 5566}
5567
8f0e8746
TT		 5568static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5569{
5570 const struct quota_format_ops *ops;
5571
5572 if (!sb_has_quota_loaded(sb, qid->type))
5573 return -ESRCH;
5574 ops = sb_dqopt(sb)->ops[qid->type];
5575 if (!ops || !ops->get_next_id)
5576 return -ENOSYS;
5577 return dquot_get_next_id(sb, qid);
5578}
ac27a0ec
DK		 5579#endif
5580
152a0836
AV		 5581static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5582 const char *dev_name, void *data)
ac27a0ec 5583{
152a0836 5584 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5585}
5586
c290ea01 5587#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5588static inline void register_as_ext2(void)
5589{
5590 int err = register_filesystem(&ext2_fs_type);
5591 if (err)
5592 printk(KERN_WARNING
5593 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5594}
5595
5596static inline void unregister_as_ext2(void)
5597{
5598 unregister_filesystem(&ext2_fs_type);
5599}
2035e776
TT		 5600
5601static inline int ext2_feature_set_ok(struct super_block *sb)
5602{
e2b911c5 5603 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776
TT		 5604 return 0;
5605 if (sb->s_flags & MS_RDONLY)
5606 return 1;
e2b911c5 5607 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5608 return 0;
5609 return 1;
5610}
24b58424
TT		 5611#else
5612static inline void register_as_ext2(void) { }
5613static inline void unregister_as_ext2(void) { }
2035e776 5614static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5615#endif
5616
24b58424
TT		 5617static inline void register_as_ext3(void)
5618{
5619 int err = register_filesystem(&ext3_fs_type);
5620 if (err)
5621 printk(KERN_WARNING
5622 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5623}
5624
5625static inline void unregister_as_ext3(void)
5626{
5627 unregister_filesystem(&ext3_fs_type);
5628}
2035e776
TT		 5629
5630static inline int ext3_feature_set_ok(struct super_block *sb)
5631{
e2b911c5 5632 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5633 return 0;
e2b911c5 5634 if (!ext4_has_feature_journal(sb))
2035e776
TT		 5635 return 0;
5636 if (sb->s_flags & MS_RDONLY)
5637 return 1;
e2b911c5 5638 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5639 return 0;
5640 return 1;
5641}
24b58424 5642
03010a33
TT		 5643static struct file_system_type ext4_fs_type = {
5644 .owner = THIS_MODULE,
5645 .name = "ext4",
152a0836 5646 .mount = ext4_mount,
03010a33 5647 .kill_sb = kill_block_super,
acc94812 5648 .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT,
03010a33 5649};
7f78e035 5650MODULE_ALIAS_FS("ext4");
03010a33 5651
e9e3bcec
ES		 5652/* Shared across all ext4 file systems */
5653wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5654
5dabfc78 5655static int __init ext4_init_fs(void)
ac27a0ec 5656{
e9e3bcec 5657 int i, err;
c9de560d 5658
e294a537 5659 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5660 ext4_li_info = NULL;
5661 mutex_init(&ext4_li_mtx);
5662
9a4c8019 5663 /* Build-time check for flags consistency */
12e9b892 5664 ext4_check_flag_values();
e9e3bcec 5665
e142d052 5666 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5667 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5668
51865fda 5669 err = ext4_init_es();
6fd058f7
TT		 5670 if (err)
5671 return err;
51865fda
ZL		 5672
5673 err = ext4_init_pageio();
5674 if (err)
b5799018 5675 goto out5;
51865fda 5676
5dabfc78 5677 err = ext4_init_system_zone();
bd2d0210 5678 if (err)
b5799018 5679 goto out4;
857ac889 5680
b5799018 5681 err = ext4_init_sysfs();
dd68314c 5682 if (err)
b5799018 5683 goto out3;
857ac889 5684
5dabfc78 5685 err = ext4_init_mballoc();
c9de560d
AT		 5686 if (err)
5687 goto out2;
ac27a0ec
DK		 5688 err = init_inodecache();
5689 if (err)
5690 goto out1;
24b58424 5691 register_as_ext3();
2035e776 5692 register_as_ext2();
03010a33 5693 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5694 if (err)
5695 goto out;
bfff6873 5696
ac27a0ec
DK		 5697 return 0;
5698out:
24b58424
TT		 5699 unregister_as_ext2();
5700 unregister_as_ext3();
ac27a0ec
DK		 5701 destroy_inodecache();
5702out1:
5dabfc78 5703 ext4_exit_mballoc();
9c191f70 5704out2:
b5799018
TT		 5705 ext4_exit_sysfs();
5706out3:
5dabfc78 5707 ext4_exit_system_zone();
b5799018 5708out4:
5dabfc78 5709 ext4_exit_pageio();
b5799018 5710out5:
51865fda
ZL		 5711 ext4_exit_es();
5712
ac27a0ec
DK		 5713 return err;
5714}
5715
5dabfc78 5716static void __exit ext4_exit_fs(void)
ac27a0ec 5717{
bfff6873 5718 ext4_destroy_lazyinit_thread();
24b58424
TT		 5719 unregister_as_ext2();
5720 unregister_as_ext3();
03010a33 5721 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5722 destroy_inodecache();
5dabfc78 5723 ext4_exit_mballoc();
b5799018 5724 ext4_exit_sysfs();
5dabfc78
TT		 5725 ext4_exit_system_zone();
5726 ext4_exit_pageio();
dd12ed14 5727 ext4_exit_es();
ac27a0ec
DK		 5728}
5729
5730MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5731MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5732MODULE_LICENSE("GPL");
5dabfc78
TT		 5733module_init(ext4_init_fs)
5734module_exit(ext4_exit_fs)
ubuntu-zesty-kernel mirror