]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
x86/cpufeatures: Disentangle MSR_SPEC_CTRL enumeration from IBRS
[mirror_ubuntu-artful-kernel.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
66114cad 27#include <linux/backing-dev.h>
ac27a0ec 28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
3197ebdb 37#include <linux/ctype.h>
1330593e 38#include <linux/log2.h>
717d50e4 39#include <linux/crc16.h>
ef510424 40#include <linux/dax.h>
7abc52c2 41#include <linux/cleancache.h>
7c0f6ba6 42#include <linux/uaccess.h>
afdbd4a6 43#include <linux/user_namespace.h>
ac27a0ec 44
bfff6873
LC		 45#include <linux/kthread.h>
46#include <linux/freezer.h>
47
3dcf5451 48#include "ext4.h"
4a092d73 49#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 50#include "ext4_jbd2.h"
ac27a0ec
DK		 51#include "xattr.h"
52#include "acl.h"
3661d286 53#include "mballoc.h"
0c9ec4be 54#include "fsmap.h"
ac27a0ec 55
9bffad1e
TT		 56#define CREATE_TRACE_POINTS
57#include <trace/events/ext4.h>
58
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
e294a537 61static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 72static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 74static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 75static int ext4_freeze(struct super_block *sb);
152a0836
AV		 76static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
2035e776
TT		 78static inline int ext2_feature_set_ok(struct super_block *sb);
79static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 80static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 81static void ext4_destroy_lazyinit_thread(void);
82static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 83static void ext4_clear_request_list(void);
c6cb7e77
EW		 84static struct inode *ext4_get_journal_inode(struct super_block *sb,
85 unsigned int journal_inum);
ac27a0ec 86
e74031fd
JK		 87/*
88 * Lock ordering
89 *
90 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
91 * i_mmap_rwsem (inode->i_mmap_rwsem)!
92 *
93 * page fault path:
94 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
95 * page lock -> i_data_sem (rw)
96 *
97 * buffered write path:
98 * sb_start_write -> i_mutex -> mmap_sem
99 * sb_start_write -> i_mutex -> transaction start -> page lock ->
100 * i_data_sem (rw)
101 *
102 * truncate:
103 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
104 * i_mmap_rwsem (w) -> page lock
105 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
106 * transaction start -> i_data_sem (rw)
107 *
108 * direct IO:
109 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
110 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
111 * transaction start -> i_data_sem (rw)
112 *
113 * writepages:
114 * transaction start -> page lock(s) -> i_data_sem (rw)
115 */
116
40bd02a3
SF		 117static bool userns_mounts = false;
118module_param(userns_mounts, bool, 0644);
119MODULE_PARM_DESC(userns_mounts, "Allow mounts from unprivileged user namespaces");
120
c290ea01 121#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 122static struct file_system_type ext2_fs_type = {
123 .owner = THIS_MODULE,
124 .name = "ext2",
125 .mount = ext4_mount,
126 .kill_sb = kill_block_super,
afdbd4a6 127 .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT,
2035e776 128};
7f78e035 129MODULE_ALIAS_FS("ext2");
fa7614dd 130MODULE_ALIAS("ext2");
2035e776
TT		 131#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
132#else
133#define IS_EXT2_SB(sb) (0)
134#endif
135
136
ba69f9ab
JK		 137static struct file_system_type ext3_fs_type = {
138 .owner = THIS_MODULE,
139 .name = "ext3",
152a0836 140 .mount = ext4_mount,
ba69f9ab 141 .kill_sb = kill_block_super,
afdbd4a6 142 .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT,
ba69f9ab 143};
7f78e035 144MODULE_ALIAS_FS("ext3");
fa7614dd 145MODULE_ALIAS("ext3");
ba69f9ab 146#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 147
d25425f8
DW		 148static int ext4_verify_csum_type(struct super_block *sb,
149 struct ext4_super_block *es)
150{
e2b911c5 151 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 152 return 1;
153
154 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
155}
156
a9c47317
DW		 157static __le32 ext4_superblock_csum(struct super_block *sb,
158 struct ext4_super_block *es)
159{
160 struct ext4_sb_info *sbi = EXT4_SB(sb);
161 int offset = offsetof(struct ext4_super_block, s_checksum);
162 __u32 csum;
163
164 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
165
166 return cpu_to_le32(csum);
167}
168
c197855e
SH		 169static int ext4_superblock_csum_verify(struct super_block *sb,
170 struct ext4_super_block *es)
a9c47317 171{
9aa5d32b 172 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 173 return 1;
174
175 return es->s_checksum == ext4_superblock_csum(sb, es);
176}
177
06db49e6 178void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 179{
06db49e6
TT		 180 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
181
9aa5d32b 182 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 183 return;
184
185 es->s_checksum = ext4_superblock_csum(sb, es);
186}
187
9933fc0a
TT		 188void *ext4_kvmalloc(size_t size, gfp_t flags)
189{
190 void *ret;
191
8be04b93 192 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 193 if (!ret)
194 ret = __vmalloc(size, flags, PAGE_KERNEL);
195 return ret;
196}
197
198void *ext4_kvzalloc(size_t size, gfp_t flags)
199{
200 void *ret;
201
8be04b93 202 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 203 if (!ret)
204 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
205 return ret;
206}
207
8fadc143
AR		 208ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
209 struct ext4_group_desc *bg)
bd81d8ee 210{
3a14589c 211 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 212 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 213 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 214}
215
8fadc143
AR		 216ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
217 struct ext4_group_desc *bg)
bd81d8ee 218{
5272f837 219 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 220 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 221 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 222}
223
8fadc143
AR		 224ext4_fsblk_t ext4_inode_table(struct super_block *sb,
225 struct ext4_group_desc *bg)
bd81d8ee 226{
5272f837 227 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 228 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 229 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 230}
231
021b65bb
TT		 232__u32 ext4_free_group_clusters(struct super_block *sb,
233 struct ext4_group_desc *bg)
560671a0
AK		 234{
235 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
236 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 237 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 238}
239
240__u32 ext4_free_inodes_count(struct super_block *sb,
241 struct ext4_group_desc *bg)
242{
243 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
244 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 245 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 246}
247
248__u32 ext4_used_dirs_count(struct super_block *sb,
249 struct ext4_group_desc *bg)
250{
251 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
252 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 253 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 254}
255
256__u32 ext4_itable_unused_count(struct super_block *sb,
257 struct ext4_group_desc *bg)
258{
259 return le16_to_cpu(bg->bg_itable_unused_lo) |
260 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 261 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 262}
263
8fadc143
AR		 264void ext4_block_bitmap_set(struct super_block *sb,
265 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 266{
3a14589c 267 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 268 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
269 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 270}
271
8fadc143
AR		 272void ext4_inode_bitmap_set(struct super_block *sb,
273 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 274{
5272f837 275 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 276 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
277 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 278}
279
8fadc143
AR		 280void ext4_inode_table_set(struct super_block *sb,
281 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 282{
5272f837 283 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 284 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
285 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 286}
287
021b65bb
TT		 288void ext4_free_group_clusters_set(struct super_block *sb,
289 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 290{
291 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
292 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
293 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
294}
295
296void ext4_free_inodes_set(struct super_block *sb,
297 struct ext4_group_desc *bg, __u32 count)
298{
299 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
300 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
301 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
302}
303
304void ext4_used_dirs_set(struct super_block *sb,
305 struct ext4_group_desc *bg, __u32 count)
306{
307 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
308 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
309 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
310}
311
312void ext4_itable_unused_set(struct super_block *sb,
313 struct ext4_group_desc *bg, __u32 count)
314{
315 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
316 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
317 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
318}
319
d3d1faf6 320
1c13d5c0
TT		 321static void __save_error_info(struct super_block *sb, const char *func,
322 unsigned int line)
323{
324 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
325
326 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 327 if (bdev_read_only(sb->s_bdev))
328 return;
1c13d5c0
TT		 329 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
330 es->s_last_error_time = cpu_to_le32(get_seconds());
331 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
332 es->s_last_error_line = cpu_to_le32(line);
333 if (!es->s_first_error_time) {
334 es->s_first_error_time = es->s_last_error_time;
335 strncpy(es->s_first_error_func, func,
336 sizeof(es->s_first_error_func));
337 es->s_first_error_line = cpu_to_le32(line);
338 es->s_first_error_ino = es->s_last_error_ino;
339 es->s_first_error_block = es->s_last_error_block;
340 }
66e61a9e
TT		 341 /*
342 * Start the daily error reporting function if it hasn't been
343 * started already
344 */
345 if (!es->s_error_count)
346 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 347 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 348}
349
350static void save_error_info(struct super_block *sb, const char *func,
351 unsigned int line)
352{
353 __save_error_info(sb, func, line);
354 ext4_commit_super(sb, 1);
355}
356
bdfe0cbd
TT		 357/*
358 * The del_gendisk() function uninitializes the disk-specific data
359 * structures, including the bdi structure, without telling anyone
360 * else. Once this happens, any attempt to call mark_buffer_dirty()
361 * (for example, by ext4_commit_super), will cause a kernel OOPS.
362 * This is a kludge to prevent these oops until we can put in a proper
363 * hook in del_gendisk() to inform the VFS and file system layers.
364 */
365static int block_device_ejected(struct super_block *sb)
366{
367 struct inode *bd_inode = sb->s_bdev->bd_inode;
368 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
369
370 return bdi->dev == NULL;
371}
372
18aadd47
BJ		 373static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
374{
375 struct super_block *sb = journal->j_private;
376 struct ext4_sb_info *sbi = EXT4_SB(sb);
377 int error = is_journal_aborted(journal);
5d3ee208 378 struct ext4_journal_cb_entry *jce;
18aadd47 379
5d3ee208 380 BUG_ON(txn->t_state == T_FINISHED);
a0154344
DJ		 381
382 ext4_process_freed_data(sb, txn->t_tid);
383
18aadd47 384 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 385 while (!list_empty(&txn->t_private_list)) {
386 jce = list_entry(txn->t_private_list.next,
387 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 388 list_del_init(&jce->jce_list);
389 spin_unlock(&sbi->s_md_lock);
390 jce->jce_func(sb, jce, error);
391 spin_lock(&sbi->s_md_lock);
392 }
393 spin_unlock(&sbi->s_md_lock);
394}
1c13d5c0 395
ac27a0ec
DK		 396/* Deal with the reporting of failure conditions on a filesystem such as
397 * inconsistencies detected or read IO failures.
398 *
399 * On ext2, we can store the error state of the filesystem in the
617ba13b 400 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 401 * write ordering constraints on the superblock which prevent us from
402 * writing it out straight away; and given that the journal is about to
403 * be aborted, we can't rely on the current, or future, transactions to
404 * write out the superblock safely.
405 *
dab291af 406 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 407 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 408 * that error until we've noted it down and cleared it.
409 */
410
617ba13b 411static void ext4_handle_error(struct super_block *sb)
ac27a0ec 412{
ac27a0ec
DK		 413 if (sb->s_flags & MS_RDONLY)
414 return;
415
2b2d6d01 416 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 417 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 418
4ab2f15b 419 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 420 if (journal)
dab291af 421 jbd2_journal_abort(journal, -EIO);
ac27a0ec 422 }
2b2d6d01 423 if (test_opt(sb, ERRORS_RO)) {
b31e1552 424 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 425 /*
426 * Make sure updated value of ->s_mount_flags will be visible
427 * before ->s_flags update
428 */
429 smp_wmb();
ac27a0ec
DK		 430 sb->s_flags |= MS_RDONLY;
431 }
4327ba52
DJ		 432 if (test_opt(sb, ERRORS_PANIC)) {
433 if (EXT4_SB(sb)->s_journal &&
434 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
435 return;
617ba13b 436 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 437 sb->s_id);
4327ba52 438 }
ac27a0ec
DK		 439}
440
efbed4dc
TT		 441#define ext4_error_ratelimit(sb) \
442 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
443 "EXT4-fs error")
444
12062ddd 445void __ext4_error(struct super_block *sb, const char *function,
c398eda0 446 unsigned int line, const char *fmt, ...)
ac27a0ec 447{
0ff2ea7d 448 struct va_format vaf;
ac27a0ec
DK		 449 va_list args;
450
0db1ff22
TT		 451 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
452 return;
453
efbed4dc
TT		 454 if (ext4_error_ratelimit(sb)) {
455 va_start(args, fmt);
456 vaf.fmt = fmt;
457 vaf.va = &args;
458 printk(KERN_CRIT
459 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
460 sb->s_id, function, line, current->comm, &vaf);
461 va_end(args);
462 }
f3fc0210 463 save_error_info(sb, function, line);
617ba13b 464 ext4_handle_error(sb);
ac27a0ec
DK		 465}
466
e7c96e8e
JP		 467void __ext4_error_inode(struct inode *inode, const char *function,
468 unsigned int line, ext4_fsblk_t block,
469 const char *fmt, ...)
273df556
FM		 470{
471 va_list args;
f7c21177 472 struct va_format vaf;
1c13d5c0 473 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 474
0db1ff22
TT		 475 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
476 return;
477
1c13d5c0
TT		 478 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
479 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 480 if (ext4_error_ratelimit(inode->i_sb)) {
481 va_start(args, fmt);
482 vaf.fmt = fmt;
483 vaf.va = &args;
484 if (block)
485 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
486 "inode #%lu: block %llu: comm %s: %pV\n",
487 inode->i_sb->s_id, function, line, inode->i_ino,
488 block, current->comm, &vaf);
489 else
490 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
491 "inode #%lu: comm %s: %pV\n",
492 inode->i_sb->s_id, function, line, inode->i_ino,
493 current->comm, &vaf);
494 va_end(args);
495 }
1c13d5c0 496 save_error_info(inode->i_sb, function, line);
273df556
FM		 497 ext4_handle_error(inode->i_sb);
498}
499
e7c96e8e
JP		 500void __ext4_error_file(struct file *file, const char *function,
501 unsigned int line, ext4_fsblk_t block,
502 const char *fmt, ...)
273df556
FM		 503{
504 va_list args;
f7c21177 505 struct va_format vaf;
1c13d5c0 506 struct ext4_super_block *es;
496ad9aa 507 struct inode *inode = file_inode(file);
273df556
FM		 508 char pathname[80], *path;
509
0db1ff22
TT		 510 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
511 return;
512
1c13d5c0
TT		 513 es = EXT4_SB(inode->i_sb)->s_es;
514 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 515 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 516 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 517 if (IS_ERR(path))
518 path = "(unknown)";
519 va_start(args, fmt);
520 vaf.fmt = fmt;
521 vaf.va = &args;
522 if (block)
523 printk(KERN_CRIT
524 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
525 "block %llu: comm %s: path %s: %pV\n",
526 inode->i_sb->s_id, function, line, inode->i_ino,
527 block, current->comm, path, &vaf);
528 else
529 printk(KERN_CRIT
530 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
531 "comm %s: path %s: %pV\n",
532 inode->i_sb->s_id, function, line, inode->i_ino,
533 current->comm, path, &vaf);
534 va_end(args);
535 }
1c13d5c0 536 save_error_info(inode->i_sb, function, line);
273df556
FM		 537 ext4_handle_error(inode->i_sb);
538}
539
722887dd
TT		 540const char *ext4_decode_error(struct super_block *sb, int errno,
541 char nbuf[16])
ac27a0ec
DK		 542{
543 char *errstr = NULL;
544
545 switch (errno) {
6a797d27
DW		 546 case -EFSCORRUPTED:
547 errstr = "Corrupt filesystem";
548 break;
549 case -EFSBADCRC:
550 errstr = "Filesystem failed CRC";
551 break;
ac27a0ec
DK		 552 case -EIO:
553 errstr = "IO failure";
554 break;
555 case -ENOMEM:
556 errstr = "Out of memory";
557 break;
558 case -EROFS:
78f1ddbb
TT		 559 if (!sb || (EXT4_SB(sb)->s_journal &&
560 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 561 errstr = "Journal has aborted";
562 else
563 errstr = "Readonly filesystem";
564 break;
565 default:
566 /* If the caller passed in an extra buffer for unknown
567 * errors, textualise them now. Else we just return
568 * NULL. */
569 if (nbuf) {
570 /* Check for truncated error codes... */
571 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
572 errstr = nbuf;
573 }
574 break;
575 }
576
577 return errstr;
578}
579
617ba13b 580/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 581 * automatically and invokes the appropriate error response. */
582
c398eda0
TT		 583void __ext4_std_error(struct super_block *sb, const char *function,
584 unsigned int line, int errno)
ac27a0ec
DK		 585{
586 char nbuf[16];
587 const char *errstr;
588
0db1ff22
TT		 589 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
590 return;
591
ac27a0ec
DK		 592 /* Special case: if the error is EROFS, and we're not already
593 * inside a transaction, then there's really no point in logging
594 * an error. */
595 if (errno == -EROFS && journal_current_handle() == NULL &&
596 (sb->s_flags & MS_RDONLY))
597 return;
598
efbed4dc
TT		 599 if (ext4_error_ratelimit(sb)) {
600 errstr = ext4_decode_error(sb, errno, nbuf);
601 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
602 sb->s_id, function, line, errstr);
603 }
ac27a0ec 604
efbed4dc 605 save_error_info(sb, function, line);
617ba13b 606 ext4_handle_error(sb);
ac27a0ec
DK		 607}
608
609/*
617ba13b 610 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 611 * abort function may be used to deal with unrecoverable failures such
612 * as journal IO errors or ENOMEM at a critical moment in log management.
613 *
614 * We unconditionally force the filesystem into an ABORT|READONLY state,
615 * unless the error response on the fs has been set to panic in which
616 * case we take the easy way out and panic immediately.
617 */
618
c67d859e 619void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 620 unsigned int line, const char *fmt, ...)
ac27a0ec 621{
651e1c3b 622 struct va_format vaf;
ac27a0ec
DK		 623 va_list args;
624
0db1ff22
TT		 625 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
626 return;
627
1c13d5c0 628 save_error_info(sb, function, line);
ac27a0ec 629 va_start(args, fmt);
651e1c3b
JP		 630 vaf.fmt = fmt;
631 vaf.va = &args;
632 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
633 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 634 va_end(args);
635
1c13d5c0
TT		 636 if ((sb->s_flags & MS_RDONLY) == 0) {
637 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 638 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 639 /*
640 * Make sure updated value of ->s_mount_flags will be visible
641 * before ->s_flags update
642 */
643 smp_wmb();
644 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 645 if (EXT4_SB(sb)->s_journal)
646 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
647 save_error_info(sb, function, line);
648 }
4327ba52
DJ		 649 if (test_opt(sb, ERRORS_PANIC)) {
650 if (EXT4_SB(sb)->s_journal &&
651 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
652 return;
617ba13b 653 panic("EXT4-fs panic from previous error\n");
4327ba52 654 }
ac27a0ec
DK		 655}
656
e7c96e8e
JP		 657void __ext4_msg(struct super_block *sb,
658 const char *prefix, const char *fmt, ...)
b31e1552 659{
0ff2ea7d 660 struct va_format vaf;
b31e1552
ES		 661 va_list args;
662
efbed4dc
TT		 663 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
664 return;
665
b31e1552 666 va_start(args, fmt);
0ff2ea7d
JP		 667 vaf.fmt = fmt;
668 vaf.va = &args;
669 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 670 va_end(args);
671}
672
b03a2f7e
AD		 673#define ext4_warning_ratelimit(sb) \
674 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
675 "EXT4-fs warning")
676
12062ddd 677void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 678 unsigned int line, const char *fmt, ...)
ac27a0ec 679{
0ff2ea7d 680 struct va_format vaf;
ac27a0ec
DK		 681 va_list args;
682
b03a2f7e 683 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 684 return;
685
ac27a0ec 686 va_start(args, fmt);
0ff2ea7d
JP		 687 vaf.fmt = fmt;
688 vaf.va = &args;
689 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
690 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 691 va_end(args);
692}
693
b03a2f7e
AD		 694void __ext4_warning_inode(const struct inode *inode, const char *function,
695 unsigned int line, const char *fmt, ...)
696{
697 struct va_format vaf;
698 va_list args;
699
700 if (!ext4_warning_ratelimit(inode->i_sb))
701 return;
702
703 va_start(args, fmt);
704 vaf.fmt = fmt;
705 vaf.va = &args;
706 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
707 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
708 function, line, inode->i_ino, current->comm, &vaf);
709 va_end(args);
710}
711
e29136f8
TT		 712void __ext4_grp_locked_error(const char *function, unsigned int line,
713 struct super_block *sb, ext4_group_t grp,
714 unsigned long ino, ext4_fsblk_t block,
715 const char *fmt, ...)
5d1b1b3f
AK		 716__releases(bitlock)
717__acquires(bitlock)
718{
0ff2ea7d 719 struct va_format vaf;
5d1b1b3f
AK		 720 va_list args;
721 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
722
0db1ff22
TT		 723 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
724 return;
725
1c13d5c0
TT		 726 es->s_last_error_ino = cpu_to_le32(ino);
727 es->s_last_error_block = cpu_to_le64(block);
728 __save_error_info(sb, function, line);
0ff2ea7d 729
efbed4dc
TT		 730 if (ext4_error_ratelimit(sb)) {
731 va_start(args, fmt);
732 vaf.fmt = fmt;
733 vaf.va = &args;
734 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
735 sb->s_id, function, line, grp);
736 if (ino)
737 printk(KERN_CONT "inode %lu: ", ino);
738 if (block)
739 printk(KERN_CONT "block %llu:",
740 (unsigned long long) block);
741 printk(KERN_CONT "%pV\n", &vaf);
742 va_end(args);
743 }
5d1b1b3f
AK		 744
745 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 746 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 747 return;
748 }
1c13d5c0 749
5d1b1b3f
AK		 750 ext4_unlock_group(sb, grp);
751 ext4_handle_error(sb);
752 /*
753 * We only get here in the ERRORS_RO case; relocking the group
754 * may be dangerous, but nothing bad will happen since the
755 * filesystem will have already been marked read/only and the
756 * journal has been aborted. We return 1 as a hint to callers
757 * who might what to use the return value from
25985edc 758 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 759 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
760 * aggressively from the ext4 function in question, with a
761 * more appropriate error code.
762 */
763 ext4_lock_group(sb, grp);
764 return;
765}
766
617ba13b 767void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 768{
617ba13b 769 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 770
617ba13b 771 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 772 return;
773
12062ddd 774 ext4_warning(sb,
ac27a0ec
DK		 775 "updating to rev %d because of new feature flag, "
776 "running e2fsck is recommended",
617ba13b 777 EXT4_DYNAMIC_REV);
ac27a0ec 778
617ba13b
MC		 779 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
780 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
781 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 782 /* leave es->s_feature_*compat flags alone */
783 /* es->s_uuid will be set by e2fsck if empty */
784
785 /*
786 * The rest of the superblock fields should be zero, and if not it
787 * means they are likely already in use, so leave them alone. We
788 * can leave it up to e2fsck to clean up any inconsistencies there.
789 */
790}
791
792/*
793 * Open the external journal device
794 */
b31e1552 795static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 796{
797 struct block_device *bdev;
798 char b[BDEVNAME_SIZE];
799
d4d77629 800 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 801 if (IS_ERR(bdev))
802 goto fail;
803 return bdev;
804
805fail:
b31e1552 806 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 807 __bdevname(dev, b), PTR_ERR(bdev));
808 return NULL;
809}
810
811/*
812 * Release the journal device
813 */
4385bab1 814static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 815{
4385bab1 816 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 817}
818
4385bab1 819static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 820{
821 struct block_device *bdev;
ac27a0ec
DK		 822 bdev = sbi->journal_bdev;
823 if (bdev) {
4385bab1 824 ext4_blkdev_put(bdev);
ac27a0ec
DK		 825 sbi->journal_bdev = NULL;
826 }
ac27a0ec
DK		 827}
828
829static inline struct inode *orphan_list_entry(struct list_head *l)
830{
617ba13b 831 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 832}
833
617ba13b 834static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 835{
836 struct list_head *l;
837
b31e1552
ES		 838 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
839 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 840
841 printk(KERN_ERR "sb_info orphan list:\n");
842 list_for_each(l, &sbi->s_orphan) {
843 struct inode *inode = orphan_list_entry(l);
844 printk(KERN_ERR " "
845 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
846 inode->i_sb->s_id, inode->i_ino, inode,
847 inode->i_mode, inode->i_nlink,
848 NEXT_ORPHAN(inode));
849 }
850}
851
957153fc
JK		 852#ifdef CONFIG_QUOTA
853static int ext4_quota_off(struct super_block *sb, int type);
854
855static inline void ext4_quota_off_umount(struct super_block *sb)
856{
857 int type;
858
964edf66
JK		 859 /* Use our quota_off function to clear inode flags etc. */
860 for (type = 0; type < EXT4_MAXQUOTAS; type++)
861 ext4_quota_off(sb, type);
957153fc
JK		 862}
863#else
864static inline void ext4_quota_off_umount(struct super_block *sb)
865{
866}
867#endif
868
2b2d6d01 869static void ext4_put_super(struct super_block *sb)
ac27a0ec 870{
617ba13b
MC		 871 struct ext4_sb_info *sbi = EXT4_SB(sb);
872 struct ext4_super_block *es = sbi->s_es;
97abd7d4 873 int aborted = 0;
ef2cabf7 874 int i, err;
ac27a0ec 875
857ac889 876 ext4_unregister_li_request(sb);
957153fc 877 ext4_quota_off_umount(sb);
e0ccfd95 878
2e8fa54e 879 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 880 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 881
0390131b 882 if (sbi->s_journal) {
97abd7d4 883 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 884 err = jbd2_journal_destroy(sbi->s_journal);
885 sbi->s_journal = NULL;
97abd7d4 886 if ((err < 0) && !aborted)
c67d859e 887 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 888 }
d4edac31 889
ebd173be 890 ext4_unregister_sysfs(sb);
d3922a77 891 ext4_es_unregister_shrinker(sbi);
9105bb14 892 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 893 ext4_release_system_zone(sb);
894 ext4_mb_release(sb);
895 ext4_ext_release(sb);
d4edac31 896
97abd7d4 897 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
e2b911c5 898 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 899 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 900 }
58c5873a 901 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 902 ext4_commit_super(sb, 1);
903
ac27a0ec
DK		 904 for (i = 0; i < sbi->s_gdb_count; i++)
905 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 906 kvfree(sbi->s_group_desc);
907 kvfree(sbi->s_flex_groups);
57042651 908 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 909 percpu_counter_destroy(&sbi->s_freeinodes_counter);
910 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 911 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c8585c6f 912 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec 913#ifdef CONFIG_QUOTA
a2d4a646 914 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 915 kfree(sbi->s_qf_names[i]);
916#endif
917
918 /* Debugging code just in case the in-memory inode orphan list
919 * isn't empty. The on-disk one can be non-empty if we've
920 * detected an error and taken the fs readonly, but the
921 * in-memory list had better be clean by this point. */
922 if (!list_empty(&sbi->s_orphan))
923 dump_orphan_list(sb, sbi);
924 J_ASSERT(list_empty(&sbi->s_orphan));
925
89d96a6f 926 sync_blockdev(sb->s_bdev);
f98393a6 927 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 928 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
929 /*
930 * Invalidate the journal device's buffers. We don't want them
931 * floating about in memory - the physical journal device may
932 * hotswapped, and it breaks the `ro-after' testing code.
933 */
934 sync_blockdev(sbi->journal_bdev);
f98393a6 935 invalidate_bdev(sbi->journal_bdev);
617ba13b 936 ext4_blkdev_remove(sbi);
ac27a0ec 937 }
dec214d0
TE		 938 if (sbi->s_ea_inode_cache) {
939 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
940 sbi->s_ea_inode_cache = NULL;
941 }
47387409
TE		 942 if (sbi->s_ea_block_cache) {
943 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
944 sbi->s_ea_block_cache = NULL;
9c191f70 945 }
c5e06d10
JL		 946 if (sbi->s_mmp_tsk)
947 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 948 brelse(sbi->s_sbh);
ac27a0ec 949 sb->s_fs_info = NULL;
3197ebdb
TT		 950 /*
951 * Now that we are completely done shutting down the
952 * superblock, we need to actually destroy the kobject.
953 */
3197ebdb
TT		 954 kobject_put(&sbi->s_kobj);
955 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 956 if (sbi->s_chksum_driver)
957 crypto_free_shash(sbi->s_chksum_driver);
705895b6 958 kfree(sbi->s_blockgroup_lock);
ac27a0ec 959 kfree(sbi);
ac27a0ec
DK		 960}
961
e18b890b 962static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 963
964/*
965 * Called inside transaction, so use GFP_NOFS
966 */
617ba13b 967static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 968{
617ba13b 969 struct ext4_inode_info *ei;
ac27a0ec 970
e6b4f8da 971 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 972 if (!ei)
973 return NULL;
0b8e58a1 974
ac27a0ec 975 ei->vfs_inode.i_version = 1;
202ee5df 976 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 977 INIT_LIST_HEAD(&ei->i_prealloc_list);
978 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 979 ext4_es_init_tree(&ei->i_es_tree);
980 rwlock_init(&ei->i_es_lock);
edaa53ca 981 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 982 ei->i_es_all_nr = 0;
edaa53ca 983 ei->i_es_shk_nr = 0;
dd475925 984 ei->i_es_shrink_lblk = 0;
d2a17637 985 ei->i_reserved_data_blocks = 0;
9d0be502 986 ei->i_da_metadata_calc_len = 0;
7e731bc9 987 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 988 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 989#ifdef CONFIG_QUOTA
990 ei->i_reserved_quota = 0;
96c7e0d9 991 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 992#endif
8aefcd55 993 ei->jinode = NULL;
2e8fa54e 994 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 995 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 996 ei->i_sync_tid = 0;
997 ei->i_datasync_tid = 0;
e27f41e1 998 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 999 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 1000 return &ei->vfs_inode;
1001}
1002
7ff9c073
TT		 1003static int ext4_drop_inode(struct inode *inode)
1004{
1005 int drop = generic_drop_inode(inode);
1006
1007 trace_ext4_drop_inode(inode, drop);
1008 return drop;
1009}
1010
fa0d7e3d
NP		 1011static void ext4_i_callback(struct rcu_head *head)
1012{
1013 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 1014 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1015}
1016
617ba13b 1017static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1018{
9f7dd93d 1019 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1020 ext4_msg(inode->i_sb, KERN_ERR,
1021 "Inode %lu (%p): orphan list check failed!",
1022 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1023 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1024 EXT4_I(inode), sizeof(struct ext4_inode_info),
1025 true);
1026 dump_stack();
1027 }
fa0d7e3d 1028 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1029}
1030
51cc5068 1031static void init_once(void *foo)
ac27a0ec 1032{
617ba13b 1033 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1034
a35afb83 1035 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1036 init_rwsem(&ei->xattr_sem);
0e855ac8 1037 init_rwsem(&ei->i_data_sem);
ea3d7209 1038 init_rwsem(&ei->i_mmap_sem);
a35afb83 1039 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1040}
1041
e67bc2b3 1042static int __init init_inodecache(void)
ac27a0ec 1043{
617ba13b
MC		 1044 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1045 sizeof(struct ext4_inode_info),
ac27a0ec 1046 0, (SLAB_RECLAIM_ACCOUNT|
5d097056 1047 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
20c2df83 1048 init_once);
617ba13b 1049 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1050 return -ENOMEM;
1051 return 0;
1052}
1053
1054static void destroy_inodecache(void)
1055{
8c0a8537
KS		 1056 /*
1057 * Make sure all delayed rcu free inodes are flushed before we
1058 * destroy cache.
1059 */
1060 rcu_barrier();
617ba13b 1061 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1062}
1063
0930fcc1 1064void ext4_clear_inode(struct inode *inode)
ac27a0ec 1065{
0930fcc1 1066 invalidate_inode_buffers(inode);
dbd5768f 1067 clear_inode(inode);
9f754758 1068 dquot_drop(inode);
c2ea3fde 1069 ext4_discard_preallocations(inode);
51865fda 1070 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1071 if (EXT4_I(inode)->jinode) {
1072 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1073 EXT4_I(inode)->jinode);
1074 jbd2_free_inode(EXT4_I(inode)->jinode);
1075 EXT4_I(inode)->jinode = NULL;
1076 }
b7236e21 1077#ifdef CONFIG_EXT4_FS_ENCRYPTION
a7550b30 1078 fscrypt_put_encryption_info(inode, NULL);
b7236e21 1079#endif
ac27a0ec
DK		 1080}
1081
1b961ac0 1082static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1083 u64 ino, u32 generation)
ac27a0ec 1084{
ac27a0ec 1085 struct inode *inode;
ac27a0ec 1086
617ba13b 1087 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1088 return ERR_PTR(-ESTALE);
617ba13b 1089 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1090 return ERR_PTR(-ESTALE);
1091
1092 /* iget isn't really right if the inode is currently unallocated!!
1093 *
617ba13b 1094 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1095 * deleted, so we should be safe.
1096 *
1097 * Currently we don't know the generation for parent directory, so
1098 * a generation of 0 means "accept any"
1099 */
f4bb2981 1100 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1101 if (IS_ERR(inode))
1102 return ERR_CAST(inode);
1103 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1104 iput(inode);
1105 return ERR_PTR(-ESTALE);
1106 }
1b961ac0
CH		 1107
1108 return inode;
1109}
1110
1111static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1112 int fh_len, int fh_type)
1b961ac0
CH		 1113{
1114 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1115 ext4_nfs_get_inode);
1116}
1117
1118static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1119 int fh_len, int fh_type)
1b961ac0
CH		 1120{
1121 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1122 ext4_nfs_get_inode);
ac27a0ec
DK		 1123}
1124
c39a7f84
TO		 1125/*
1126 * Try to release metadata pages (indirect blocks, directories) which are
1127 * mapped via the block device. Since these pages could have journal heads
1128 * which would prevent try_to_free_buffers() from freeing them, we must use
1129 * jbd2 layer's try_to_free_buffers() function to release them.
1130 */
0b8e58a1
AD		 1131static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1132 gfp_t wait)
c39a7f84
TO		 1133{
1134 journal_t *journal = EXT4_SB(sb)->s_journal;
1135
1136 WARN_ON(PageChecked(page));
1137 if (!page_has_buffers(page))
1138 return 0;
1139 if (journal)
1140 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1141 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1142 return try_to_free_buffers(page);
1143}
1144
a7550b30
JK		 1145#ifdef CONFIG_EXT4_FS_ENCRYPTION
1146static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1147{
1148 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1149 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1150}
1151
a7550b30
JK		 1152static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1153 void *fs_data)
1154{
2f8f5e76 1155 handle_t *handle = fs_data;
c1a5d5f6 1156 int res, res2, credits, retries = 0;
2f8f5e76 1157
9ce0151a
EB		 1158 /*
1159 * Encrypting the root directory is not allowed because e2fsck expects
1160 * lost+found to exist and be unencrypted, and encrypting the root
1161 * directory would imply encrypting the lost+found directory as well as
1162 * the filename "lost+found" itself.
1163 */
1164 if (inode->i_ino == EXT4_ROOT_INO)
1165 return -EPERM;
2f8f5e76 1166
94840e3c
EB		 1167 res = ext4_convert_inline_data(inode);
1168 if (res)
1169 return res;
1170
2f8f5e76
EB		 1171 /*
1172 * If a journal handle was specified, then the encryption context is
1173 * being set on a new inode via inheritance and is part of a larger
1174 * transaction to create the inode. Otherwise the encryption context is
1175 * being set on an existing inode in its own transaction. Only in the
1176 * latter case should the "retry on ENOSPC" logic be used.
1177 */
a7550b30 1178
2f8f5e76
EB		 1179 if (handle) {
1180 res = ext4_xattr_set_handle(handle, inode,
1181 EXT4_XATTR_INDEX_ENCRYPTION,
1182 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1183 ctx, len, 0);
a7550b30
JK		 1184 if (!res) {
1185 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1186 ext4_clear_inode_state(inode,
1187 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b
JK		 1188 /*
1189 * Update inode->i_flags - e.g. S_DAX may get disabled
1190 */
1191 ext4_set_inode_flags(inode);
a7550b30
JK		 1192 }
1193 return res;
1194 }
1195
b8cb5a54
TE		 1196 res = dquot_initialize(inode);
1197 if (res)
1198 return res;
2f8f5e76 1199retry:
af65207c
TE		 1200 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1201 &credits);
dec214d0
TE		 1202 if (res)
1203 return res;
1204
c1a5d5f6 1205 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
a7550b30
JK		 1206 if (IS_ERR(handle))
1207 return PTR_ERR(handle);
1208
2f8f5e76
EB		 1209 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1210 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1211 ctx, len, 0);
a7550b30
JK		 1212 if (!res) {
1213 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
a3caa24b
JK		 1214 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1215 ext4_set_inode_flags(inode);
a7550b30
JK		 1216 res = ext4_mark_inode_dirty(handle, inode);
1217 if (res)
1218 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1219 }
1220 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1221
1222 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1223 goto retry;
a7550b30
JK		 1224 if (!res)
1225 res = res2;
1226 return res;
1227}
1228
c250b7dd 1229static bool ext4_dummy_context(struct inode *inode)
a7550b30
JK		 1230{
1231 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1232}
1233
1234static unsigned ext4_max_namelen(struct inode *inode)
1235{
1236 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1237 EXT4_NAME_LEN;
1238}
1239
6f69f0ed 1240static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1241 .key_prefix = "ext4:",
a7550b30 1242 .get_context = ext4_get_context,
a7550b30
JK		 1243 .set_context = ext4_set_context,
1244 .dummy_context = ext4_dummy_context,
1245 .is_encrypted = ext4_encrypted_inode,
1246 .empty_dir = ext4_empty_dir,
1247 .max_namelen = ext4_max_namelen,
1248};
1249#else
6f69f0ed 1250static const struct fscrypt_operations ext4_cryptops = {
a7550b30
JK		 1251 .is_encrypted = ext4_encrypted_inode,
1252};
1253#endif
1254
ac27a0ec 1255#ifdef CONFIG_QUOTA
d6006186 1256static const char * const quotatypes[] = INITQFNAMES;
689c958c 1257#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1258
617ba13b
MC		 1259static int ext4_write_dquot(struct dquot *dquot);
1260static int ext4_acquire_dquot(struct dquot *dquot);
1261static int ext4_release_dquot(struct dquot *dquot);
1262static int ext4_mark_dquot_dirty(struct dquot *dquot);
1263static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1264static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1265 const struct path *path);
617ba13b
MC		 1266static int ext4_quota_on_mount(struct super_block *sb, int type);
1267static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1268 size_t len, loff_t off);
617ba13b 1269static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1270 const char *data, size_t len, loff_t off);
7c319d32
AK		 1271static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1272 unsigned int flags);
1273static int ext4_enable_quotas(struct super_block *sb);
8f0e8746 1274static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1275
96c7e0d9
JK		 1276static struct dquot **ext4_get_dquots(struct inode *inode)
1277{
1278 return EXT4_I(inode)->i_dquot;
1279}
1280
61e225dc 1281static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1282 .get_reserved_space = ext4_get_reserved_space,
1283 .write_dquot = ext4_write_dquot,
1284 .acquire_dquot = ext4_acquire_dquot,
1285 .release_dquot = ext4_release_dquot,
1286 .mark_dirty = ext4_mark_dquot_dirty,
1287 .write_info = ext4_write_info,
1288 .alloc_dquot = dquot_alloc,
1289 .destroy_dquot = dquot_destroy,
1290 .get_projid = ext4_get_projid,
1291 .get_inode_usage = ext4_get_inode_usage,
1292 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1293};
1294
0d54b217 1295static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1296 .quota_on = ext4_quota_on,
ca0e05e4 1297 .quota_off = ext4_quota_off,
287a8095 1298 .quota_sync = dquot_quota_sync,
0a240339 1299 .get_state = dquot_get_state,
287a8095
CH		 1300 .set_info = dquot_set_dqinfo,
1301 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1302 .set_dqblk = dquot_set_dqblk,
1303 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1304};
1305#endif
1306
ee9b6d61 1307static const struct super_operations ext4_sops = {
617ba13b
MC		 1308 .alloc_inode = ext4_alloc_inode,
1309 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1310 .write_inode = ext4_write_inode,
1311 .dirty_inode = ext4_dirty_inode,
7ff9c073 1312 .drop_inode = ext4_drop_inode,
0930fcc1 1313 .evict_inode = ext4_evict_inode,
617ba13b 1314 .put_super = ext4_put_super,
617ba13b 1315 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1316 .freeze_fs = ext4_freeze,
1317 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1318 .statfs = ext4_statfs,
1319 .remount_fs = ext4_remount,
617ba13b 1320 .show_options = ext4_show_options,
ac27a0ec 1321#ifdef CONFIG_QUOTA
617ba13b
MC		 1322 .quota_read = ext4_quota_read,
1323 .quota_write = ext4_quota_write,
96c7e0d9 1324 .get_dquots = ext4_get_dquots,
ac27a0ec 1325#endif
c39a7f84 1326 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1327};
1328
39655164 1329static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1330 .fh_to_dentry = ext4_fh_to_dentry,
1331 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1332 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1333};
1334
1335enum {
1336 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1337 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1338 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1339 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1340 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1341 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1342 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1343 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1344 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1345 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1346 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1347 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1348 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1449032b 1349 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
670e9875 1350 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1351 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1352 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1353 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1354 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1355 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
ac27a0ec
DK		 1356};
1357
a447c093 1358static const match_table_t tokens = {
ac27a0ec
DK		 1359 {Opt_bsd_df, "bsddf"},
1360 {Opt_minix_df, "minixdf"},
1361 {Opt_grpid, "grpid"},
1362 {Opt_grpid, "bsdgroups"},
1363 {Opt_nogrpid, "nogrpid"},
1364 {Opt_nogrpid, "sysvgroups"},
1365 {Opt_resgid, "resgid=%u"},
1366 {Opt_resuid, "resuid=%u"},
1367 {Opt_sb, "sb=%u"},
1368 {Opt_err_cont, "errors=continue"},
1369 {Opt_err_panic, "errors=panic"},
1370 {Opt_err_ro, "errors=remount-ro"},
1371 {Opt_nouid32, "nouid32"},
ac27a0ec 1372 {Opt_debug, "debug"},
72578c33
TT		 1373 {Opt_removed, "oldalloc"},
1374 {Opt_removed, "orlov"},
ac27a0ec
DK		 1375 {Opt_user_xattr, "user_xattr"},
1376 {Opt_nouser_xattr, "nouser_xattr"},
1377 {Opt_acl, "acl"},
1378 {Opt_noacl, "noacl"},
e3bb52ae 1379 {Opt_noload, "norecovery"},
5a916be1 1380 {Opt_noload, "noload"},
72578c33
TT		 1381 {Opt_removed, "nobh"},
1382 {Opt_removed, "bh"},
ac27a0ec 1383 {Opt_commit, "commit=%u"},
30773840
TT		 1384 {Opt_min_batch_time, "min_batch_time=%u"},
1385 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1386 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1387 {Opt_journal_path, "journal_path=%s"},
818d276c 1388 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1389 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1390 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1391 {Opt_abort, "abort"},
1392 {Opt_data_journal, "data=journal"},
1393 {Opt_data_ordered, "data=ordered"},
1394 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1395 {Opt_data_err_abort, "data_err=abort"},
1396 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1397 {Opt_offusrjquota, "usrjquota="},
1398 {Opt_usrjquota, "usrjquota=%s"},
1399 {Opt_offgrpjquota, "grpjquota="},
1400 {Opt_grpjquota, "grpjquota=%s"},
1401 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1402 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1403 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1404 {Opt_grpquota, "grpquota"},
1405 {Opt_noquota, "noquota"},
1406 {Opt_quota, "quota"},
1407 {Opt_usrquota, "usrquota"},
49da9392 1408 {Opt_prjquota, "prjquota"},
ac27a0ec 1409 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1410 {Opt_barrier, "barrier"},
1411 {Opt_nobarrier, "nobarrier"},
25ec56b5 1412 {Opt_i_version, "i_version"},
923ae0ff 1413 {Opt_dax, "dax"},
c9de560d 1414 {Opt_stripe, "stripe=%u"},
64769240 1415 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1416 {Opt_lazytime, "lazytime"},
1417 {Opt_nolazytime, "nolazytime"},
670e9875 1418 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1419 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1420 {Opt_removed, "mblk_io_submit"},
1421 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1422 {Opt_block_validity, "block_validity"},
1423 {Opt_noblock_validity, "noblock_validity"},
240799cd 1424 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1425 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1426 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1427 {Opt_auto_da_alloc, "auto_da_alloc"},
1428 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1429 {Opt_dioread_nolock, "dioread_nolock"},
1430 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1431 {Opt_discard, "discard"},
1432 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1433 {Opt_init_itable, "init_itable=%u"},
1434 {Opt_init_itable, "init_itable"},
1435 {Opt_noinit_itable, "noinit_itable"},
df981d03 1436 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1437 {Opt_test_dummy_encryption, "test_dummy_encryption"},
cdb7ee4c
TE		 1438 {Opt_nombcache, "nombcache"},
1439 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
c7198b9c
TT		 1440 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1441 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1442 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1443 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1444 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1445 {Opt_err, NULL},
ac27a0ec
DK		 1446};
1447
617ba13b 1448static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1449{
617ba13b 1450 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1451 char *options = (char *) *data;
1452
1453 if (!options || strncmp(options, "sb=", 3) != 0)
1454 return 1; /* Default location */
0b8e58a1 1455
ac27a0ec 1456 options += 3;
0b8e58a1 1457 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1458 sb_block = simple_strtoul(options, &options, 0);
1459 if (*options && *options != ',') {
4776004f 1460 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1461 (char *) *data);
1462 return 1;
1463 }
1464 if (*options == ',')
1465 options++;
1466 *data = (void *) options;
0b8e58a1 1467
ac27a0ec
DK		 1468 return sb_block;
1469}
1470
b3881f74 1471#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1472static const char deprecated_msg[] =
1473 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1474 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1475
56c50f11
DM		 1476#ifdef CONFIG_QUOTA
1477static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1478{
1479 struct ext4_sb_info *sbi = EXT4_SB(sb);
1480 char *qname;
03dafb5f 1481 int ret = -1;
56c50f11
DM		 1482
1483 if (sb_any_quota_loaded(sb) &&
1484 !sbi->s_qf_names[qtype]) {
1485 ext4_msg(sb, KERN_ERR,
1486 "Cannot change journaled "
1487 "quota options when quota turned on");
57f73c2c 1488 return -1;
56c50f11 1489 }
e2b911c5 1490 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1491 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1492 "ignored when QUOTA feature is enabled");
1493 return 1;
262b4662 1494 }
56c50f11
DM		 1495 qname = match_strdup(args);
1496 if (!qname) {
1497 ext4_msg(sb, KERN_ERR,
1498 "Not enough memory for storing quotafile name");
57f73c2c 1499 return -1;
56c50f11 1500 }
03dafb5f
CG		 1501 if (sbi->s_qf_names[qtype]) {
1502 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1503 ret = 1;
1504 else
1505 ext4_msg(sb, KERN_ERR,
1506 "%s quota file already specified",
1507 QTYPE2NAME(qtype));
1508 goto errout;
56c50f11 1509 }
03dafb5f 1510 if (strchr(qname, '/')) {
56c50f11
DM		 1511 ext4_msg(sb, KERN_ERR,
1512 "quotafile must be on filesystem root");
03dafb5f 1513 goto errout;
56c50f11 1514 }
03dafb5f 1515 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1516 set_opt(sb, QUOTA);
56c50f11 1517 return 1;
03dafb5f
CG		 1518errout:
1519 kfree(qname);
1520 return ret;
56c50f11
DM		 1521}
1522
1523static int clear_qf_name(struct super_block *sb, int qtype)
1524{
1525
1526 struct ext4_sb_info *sbi = EXT4_SB(sb);
1527
1528 if (sb_any_quota_loaded(sb) &&
1529 sbi->s_qf_names[qtype]) {
1530 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1531 " when quota turned on");
57f73c2c 1532 return -1;
56c50f11 1533 }
03dafb5f 1534 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1535 sbi->s_qf_names[qtype] = NULL;
1536 return 1;
1537}
1538#endif
1539
26092bf5
TT		 1540#define MOPT_SET 0x0001
1541#define MOPT_CLEAR 0x0002
1542#define MOPT_NOSUPPORT 0x0004
1543#define MOPT_EXPLICIT 0x0008
1544#define MOPT_CLEAR_ERR 0x0010
1545#define MOPT_GTE0 0x0020
ac27a0ec 1546#ifdef CONFIG_QUOTA
26092bf5
TT		 1547#define MOPT_Q 0
1548#define MOPT_QFMT 0x0040
1549#else
1550#define MOPT_Q MOPT_NOSUPPORT
1551#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1552#endif
26092bf5 1553#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1554#define MOPT_NO_EXT2 0x0100
1555#define MOPT_NO_EXT3 0x0200
1556#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1557#define MOPT_STRING 0x0400
26092bf5
TT		 1558
1559static const struct mount_opts {
1560 int token;
1561 int mount_opt;
1562 int flags;
1563} ext4_mount_opts[] = {
1564 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1565 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1566 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1567 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1568 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1569 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1570 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1571 MOPT_EXT4_ONLY | MOPT_SET},
1572 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1573 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1574 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1575 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1576 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1577 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1578 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1579 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1580 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1581 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1582 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1583 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1584 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1585 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1586 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1587 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1588 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1589 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1590 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1591 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1592 MOPT_NO_EXT2},
8dc0aa8c 1593 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1594 MOPT_NO_EXT2},
26092bf5
TT		 1595 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1596 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1597 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1598 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1599 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1600 {Opt_commit, 0, MOPT_GTE0},
1601 {Opt_max_batch_time, 0, MOPT_GTE0},
1602 {Opt_min_batch_time, 0, MOPT_GTE0},
1603 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1604 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1605 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1606 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1607 {Opt_resuid, 0, MOPT_GTE0},
1608 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1609 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1610 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1611 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1612 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1613 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1614 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1615 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1616 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1617 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1618#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1619 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1620 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1621#else
26092bf5
TT		 1622 {Opt_acl, 0, MOPT_NOSUPPORT},
1623 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1624#endif
26092bf5
TT		 1625 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1626 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1627 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1628 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1629 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1630 MOPT_SET | MOPT_Q},
1631 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1632 MOPT_SET | MOPT_Q},
49da9392
JK		 1633 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1634 MOPT_SET | MOPT_Q},
26092bf5 1635 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1636 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1637 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1638 {Opt_usrjquota, 0, MOPT_Q},
1639 {Opt_grpjquota, 0, MOPT_Q},
1640 {Opt_offusrjquota, 0, MOPT_Q},
1641 {Opt_offgrpjquota, 0, MOPT_Q},
1642 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1643 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1644 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1645 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1646 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
cdb7ee4c 1647 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
26092bf5
TT		 1648 {Opt_err, 0, 0}
1649};
1650
1651static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1652 substring_t *args, unsigned long *journal_devnum,
1653 unsigned int *journal_ioprio, int is_remount)
1654{
1655 struct ext4_sb_info *sbi = EXT4_SB(sb);
1656 const struct mount_opts *m;
08cefc7a
EB		 1657 kuid_t uid;
1658 kgid_t gid;
26092bf5
TT		 1659 int arg = 0;
1660
57f73c2c
TT		 1661#ifdef CONFIG_QUOTA
1662 if (token == Opt_usrjquota)
1663 return set_qf_name(sb, USRQUOTA, &args[0]);
1664 else if (token == Opt_grpjquota)
1665 return set_qf_name(sb, GRPQUOTA, &args[0]);
1666 else if (token == Opt_offusrjquota)
1667 return clear_qf_name(sb, USRQUOTA);
1668 else if (token == Opt_offgrpjquota)
1669 return clear_qf_name(sb, GRPQUOTA);
1670#endif
26092bf5 1671 switch (token) {
f7048605
TT		 1672 case Opt_noacl:
1673 case Opt_nouser_xattr:
1674 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1675 break;
26092bf5
TT		 1676 case Opt_sb:
1677 return 1; /* handled by get_sb_block() */
1678 case Opt_removed:
5f3633e3 1679 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1680 return 1;
26092bf5
TT		 1681 case Opt_abort:
1682 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1683 return 1;
1684 case Opt_i_version:
1685 sb->s_flags |= MS_I_VERSION;
1686 return 1;
a26f4992
TT		 1687 case Opt_lazytime:
1688 sb->s_flags |= MS_LAZYTIME;
1689 return 1;
1690 case Opt_nolazytime:
1691 sb->s_flags &= ~MS_LAZYTIME;
1692 return 1;
26092bf5
TT		 1693 }
1694
5f3633e3
JK		 1695 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1696 if (token == m->token)
1697 break;
1698
1699 if (m->token == Opt_err) {
1700 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1701 "or missing value", opt);
1702 return -1;
1703 }
1704
8dc0aa8c
TT		 1705 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1706 ext4_msg(sb, KERN_ERR,
1707 "Mount option \"%s\" incompatible with ext2", opt);
1708 return -1;
1709 }
1710 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1711 ext4_msg(sb, KERN_ERR,
1712 "Mount option \"%s\" incompatible with ext3", opt);
1713 return -1;
1714 }
1715
afdbd4a6
SF		 1716 if (token == Opt_err_panic && !capable(CAP_SYS_ADMIN)) {
1717 ext4_msg(sb, KERN_ERR,
1718 "Mount option \"%s\" not allowed for unprivileged mounts",
1719 opt);
1720 return -1;
1721 }
1722
ad4eec61 1723 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1724 return -1;
1725 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1726 return -1;
c93cf2d7
DM		 1727 if (m->flags & MOPT_EXPLICIT) {
1728 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1729 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1730 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1731 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1732 } else
1733 return -1;
1734 }
5f3633e3
JK		 1735 if (m->flags & MOPT_CLEAR_ERR)
1736 clear_opt(sb, ERRORS_MASK);
1737 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1738 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1739 "options when quota turned on");
1740 return -1;
1741 }
1742
1743 if (m->flags & MOPT_NOSUPPORT) {
1744 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1745 } else if (token == Opt_commit) {
1746 if (arg == 0)
1747 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1748 sbi->s_commit_interval = HZ * arg;
670e9875
TT		 1749 } else if (token == Opt_debug_want_extra_isize) {
1750 sbi->s_want_extra_isize = arg;
5f3633e3 1751 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1752 sbi->s_max_batch_time = arg;
1753 } else if (token == Opt_min_batch_time) {
1754 sbi->s_min_batch_time = arg;
1755 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1756 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1757 ext4_msg(sb, KERN_ERR,
1758 "EXT4-fs: inode_readahead_blks must be "
1759 "0 or a power of 2 smaller than 2^31");
26092bf5 1760 return -1;
5f3633e3
JK		 1761 }
1762 sbi->s_inode_readahead_blks = arg;
1763 } else if (token == Opt_init_itable) {
1764 set_opt(sb, INIT_INODE_TABLE);
1765 if (!args->from)
1766 arg = EXT4_DEF_LI_WAIT_MULT;
1767 sbi->s_li_wait_mult = arg;
1768 } else if (token == Opt_max_dir_size_kb) {
1769 sbi->s_max_dir_size_kb = arg;
1770 } else if (token == Opt_stripe) {
1771 sbi->s_stripe = arg;
1772 } else if (token == Opt_resuid) {
afdbd4a6 1773 uid = make_kuid(sb->s_user_ns, arg);
5f3633e3
JK		 1774 if (!uid_valid(uid)) {
1775 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1776 return -1;
1777 }
5f3633e3
JK		 1778 sbi->s_resuid = uid;
1779 } else if (token == Opt_resgid) {
afdbd4a6 1780 gid = make_kgid(sb->s_user_ns, arg);
5f3633e3
JK		 1781 if (!gid_valid(gid)) {
1782 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1783 return -1;
1784 }
1785 sbi->s_resgid = gid;
1786 } else if (token == Opt_journal_dev) {
1787 if (is_remount) {
1788 ext4_msg(sb, KERN_ERR,
1789 "Cannot specify journal on remount");
1790 return -1;
1791 }
1792 *journal_devnum = arg;
ad4eec61
ES		 1793 } else if (token == Opt_journal_path) {
1794 char *journal_path;
1795 struct inode *journal_inode;
1796 struct path path;
1797 int error;
1798
1799 if (is_remount) {
1800 ext4_msg(sb, KERN_ERR,
1801 "Cannot specify journal on remount");
1802 return -1;
1803 }
1804 journal_path = match_strdup(&args[0]);
1805 if (!journal_path) {
1806 ext4_msg(sb, KERN_ERR, "error: could not dup "
1807 "journal device string");
1808 return -1;
1809 }
1810
1811 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1812 if (error) {
1813 ext4_msg(sb, KERN_ERR, "error: could not find "
1814 "journal device path: error %d", error);
1815 kfree(journal_path);
1816 return -1;
1817 }
1818
afdbd4a6
SF		 1819 /*
1820 * Refuse access for unprivileged mounts if the user does
1821 * not have rw access to the journal device via the supplied
1822 * path.
1823 */
1824 if (!capable(CAP_SYS_ADMIN) &&
1825 inode_permission(d_inode(path.dentry), MAY_READ|MAY_WRITE)) {
1826 ext4_msg(sb, KERN_ERR,
1827 "error: Insufficient access to journal path %s",
1828 journal_path);
1829 return -1;
1830 }
1831
2b0143b5 1832 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1833 if (!S_ISBLK(journal_inode->i_mode)) {
1834 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1835 "is not a block device", journal_path);
1836 path_put(&path);
1837 kfree(journal_path);
1838 return -1;
1839 }
1840
1841 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1842 path_put(&path);
1843 kfree(journal_path);
5f3633e3
JK		 1844 } else if (token == Opt_journal_ioprio) {
1845 if (arg > 7) {
1846 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1847 " (must be 0-7)");
1848 return -1;
1849 }
1850 *journal_ioprio =
1851 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1852 } else if (token == Opt_test_dummy_encryption) {
1853#ifdef CONFIG_EXT4_FS_ENCRYPTION
1854 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1855 ext4_msg(sb, KERN_WARNING,
1856 "Test dummy encryption mode enabled");
1857#else
1858 ext4_msg(sb, KERN_WARNING,
1859 "Test dummy encryption mount option ignored");
1860#endif
5f3633e3
JK		 1861 } else if (m->flags & MOPT_DATAJ) {
1862 if (is_remount) {
1863 if (!sbi->s_journal)
1864 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1865 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1866 ext4_msg(sb, KERN_ERR,
26092bf5 1867 "Cannot change data mode on remount");
26092bf5 1868 return -1;
ac27a0ec 1869 }
26092bf5 1870 } else {
5f3633e3
JK		 1871 clear_opt(sb, DATA_FLAGS);
1872 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1873 }
5f3633e3
JK		 1874#ifdef CONFIG_QUOTA
1875 } else if (m->flags & MOPT_QFMT) {
1876 if (sb_any_quota_loaded(sb) &&
1877 sbi->s_jquota_fmt != m->mount_opt) {
1878 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1879 "quota options when quota turned on");
1880 return -1;
1881 }
e2b911c5 1882 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1883 ext4_msg(sb, KERN_INFO,
1884 "Quota format mount options ignored "
262b4662 1885 "when QUOTA feature is enabled");
c325a67c 1886 return 1;
262b4662 1887 }
5f3633e3 1888 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1889#endif
923ae0ff 1890 } else if (token == Opt_dax) {
ef83b6e8
DW		 1891#ifdef CONFIG_FS_DAX
1892 ext4_msg(sb, KERN_WARNING,
1893 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1894 sbi->s_mount_opt |= m->mount_opt;
1895#else
923ae0ff
RZ		 1896 ext4_msg(sb, KERN_INFO, "dax option not supported");
1897 return -1;
5f3633e3 1898#endif
7915a861
AN		 1899 } else if (token == Opt_data_err_abort) {
1900 sbi->s_mount_opt |= m->mount_opt;
1901 } else if (token == Opt_data_err_ignore) {
1902 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1903 } else {
1904 if (!args->from)
1905 arg = 1;
1906 if (m->flags & MOPT_CLEAR)
1907 arg = !arg;
1908 else if (unlikely(!(m->flags & MOPT_SET))) {
1909 ext4_msg(sb, KERN_WARNING,
1910 "buggy handling of option %s", opt);
1911 WARN_ON(1);
1912 return -1;
1913 }
1914 if (arg != 0)
1915 sbi->s_mount_opt |= m->mount_opt;
1916 else
1917 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1918 }
5f3633e3 1919 return 1;
26092bf5
TT		 1920}
1921
1922static int parse_options(char *options, struct super_block *sb,
1923 unsigned long *journal_devnum,
1924 unsigned int *journal_ioprio,
1925 int is_remount)
1926{
1927 struct ext4_sb_info *sbi = EXT4_SB(sb);
1928 char *p;
1929 substring_t args[MAX_OPT_ARGS];
1930 int token;
1931
1932 if (!options)
1933 return 1;
1934
1935 while ((p = strsep(&options, ",")) != NULL) {
1936 if (!*p)
1937 continue;
1938 /*
1939 * Initialize args struct so we know whether arg was
1940 * found; some options take optional arguments.
1941 */
caecd0af 1942 args[0].to = args[0].from = NULL;
26092bf5
TT		 1943 token = match_token(p, tokens, args);
1944 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1945 journal_ioprio, is_remount) < 0)
1946 return 0;
ac27a0ec
DK		 1947 }
1948#ifdef CONFIG_QUOTA
49da9392
JK		 1949 /*
1950 * We do the test below only for project quotas. 'usrquota' and
1951 * 'grpquota' mount options are allowed even without quota feature
1952 * to support legacy quotas in quota files.
1953 */
1954 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1955 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1956 "Cannot enable project quota enforcement.");
1957 return 0;
1958 }
1959 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1960 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1961 clear_opt(sb, USRQUOTA);
ac27a0ec 1962
482a7425 1963 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1964 clear_opt(sb, GRPQUOTA);
ac27a0ec 1965
56c50f11 1966 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1967 ext4_msg(sb, KERN_ERR, "old and new quota "
1968 "format mixing");
ac27a0ec
DK		 1969 return 0;
1970 }
1971
1972 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1973 ext4_msg(sb, KERN_ERR, "journaled quota format "
1974 "not specified");
ac27a0ec
DK		 1975 return 0;
1976 }
ac27a0ec
DK		 1977 }
1978#endif
261cb20c
JK		 1979 if (test_opt(sb, DIOREAD_NOLOCK)) {
1980 int blocksize =
1981 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1982
09cbfeaf 1983 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 1984 ext4_msg(sb, KERN_ERR, "can't mount with "
1985 "dioread_nolock if block size != PAGE_SIZE");
1986 return 0;
1987 }
1988 }
ac27a0ec
DK		 1989 return 1;
1990}
1991
2adf6da8
TT		 1992static inline void ext4_show_quota_options(struct seq_file *seq,
1993 struct super_block *sb)
1994{
1995#if defined(CONFIG_QUOTA)
1996 struct ext4_sb_info *sbi = EXT4_SB(sb);
1997
1998 if (sbi->s_jquota_fmt) {
1999 char *fmtname = "";
2000
2001 switch (sbi->s_jquota_fmt) {
2002 case QFMT_VFS_OLD:
2003 fmtname = "vfsold";
2004 break;
2005 case QFMT_VFS_V0:
2006 fmtname = "vfsv0";
2007 break;
2008 case QFMT_VFS_V1:
2009 fmtname = "vfsv1";
2010 break;
2011 }
2012 seq_printf(seq, ",jqfmt=%s", fmtname);
2013 }
2014
2015 if (sbi->s_qf_names[USRQUOTA])
a068acf2 2016 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 2017
2018 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 2019 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 2020#endif
2021}
2022
5a916be1
TT		 2023static const char *token2str(int token)
2024{
50df9fd5 2025 const struct match_token *t;
5a916be1
TT		 2026
2027 for (t = tokens; t->token != Opt_err; t++)
2028 if (t->token == token && !strchr(t->pattern, '='))
2029 break;
2030 return t->pattern;
2031}
2032
2adf6da8
TT		 2033/*
2034 * Show an option if
2035 * - it's set to a non-default value OR
2036 * - if the per-sb default is different from the global default
2037 */
66acdcf4
TT		 2038static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2039 int nodefs)
2adf6da8 2040{
2adf6da8
TT		 2041 struct ext4_sb_info *sbi = EXT4_SB(sb);
2042 struct ext4_super_block *es = sbi->s_es;
66acdcf4 2043 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 2044 const struct mount_opts *m;
66acdcf4 2045 char sep = nodefs ? '\n' : ',';
2adf6da8 2046
66acdcf4
TT		 2047#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2048#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2049
2050 if (sbi->s_sb_block != 1)
5a916be1
TT		 2051 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2052
2053 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2054 int want_set = m->flags & MOPT_SET;
2055 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2056 (m->flags & MOPT_CLEAR_ERR))
2057 continue;
66acdcf4 2058 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2059 continue; /* skip if same as the default */
2060 if ((want_set &&
2061 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2062 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2063 continue; /* select Opt_noFoo vs Opt_Foo */
2064 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2065 }
5a916be1 2066
afdbd4a6 2067 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(sb->s_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2068 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a 2069 SEQ_OPTS_PRINT("resuid=%u",
afdbd4a6
SF		 2070 from_kuid_munged(sb->s_user_ns, sbi->s_resuid));
2071 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(sb->s_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2072 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a 2073 SEQ_OPTS_PRINT("resgid=%u",
afdbd4a6 2074 from_kgid_munged(sb->s_user_ns, sbi->s_resgid));
66acdcf4 2075 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2076 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2077 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2078 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2079 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2080 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2081 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2082 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2083 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2084 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2085 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2086 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2087 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 2088 if (sb->s_flags & MS_I_VERSION)
5a916be1 2089 SEQ_OPTS_PUTS("i_version");
66acdcf4 2090 if (nodefs || sbi->s_stripe)
5a916be1 2091 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 2092 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2093 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2094 SEQ_OPTS_PUTS("data=journal");
2095 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2096 SEQ_OPTS_PUTS("data=ordered");
2097 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2098 SEQ_OPTS_PUTS("data=writeback");
2099 }
66acdcf4
TT		 2100 if (nodefs ||
2101 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2102 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2103 sbi->s_inode_readahead_blks);
2adf6da8 2104
66acdcf4
TT		 2105 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2106 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2107 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2108 if (nodefs || sbi->s_max_dir_size_kb)
2109 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2110 if (test_opt(sb, DATA_ERR_ABORT))
2111 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 2112
2113 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2114 return 0;
2115}
2116
66acdcf4
TT		 2117static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2118{
2119 return _ext4_show_options(seq, root->d_sb, 0);
2120}
2121
ebd173be 2122int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2123{
2124 struct super_block *sb = seq->private;
2125 int rc;
2126
2127 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2128 rc = _ext4_show_options(seq, sb, 1);
2129 seq_puts(seq, "\n");
2130 return rc;
2131}
2132
617ba13b 2133static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2134 int read_only)
2135{
617ba13b 2136 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2137 int res = 0;
2138
617ba13b 2139 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2140 ext4_msg(sb, KERN_ERR, "revision level too high, "
2141 "forcing read-only mode");
ac27a0ec
DK		 2142 res = MS_RDONLY;
2143 }
2144 if (read_only)
281b5995 2145 goto done;
617ba13b 2146 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2147 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2148 "running e2fsck is recommended");
c8b459f4 2149 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2150 ext4_msg(sb, KERN_WARNING,
2151 "warning: mounting fs with errors, "
2152 "running e2fsck is recommended");
ed3ce80a 2153 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2154 le16_to_cpu(es->s_mnt_count) >=
2155 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2156 ext4_msg(sb, KERN_WARNING,
2157 "warning: maximal mount count reached, "
2158 "running e2fsck is recommended");
ac27a0ec
DK		 2159 else if (le32_to_cpu(es->s_checkinterval) &&
2160 (le32_to_cpu(es->s_lastcheck) +
2161 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 2162 ext4_msg(sb, KERN_WARNING,
2163 "warning: checktime reached, "
2164 "running e2fsck is recommended");
0b8e58a1 2165 if (!sbi->s_journal)
0390131b 2166 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2167 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2168 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2169 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 2170 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 2171 ext4_update_dynamic_rev(sb);
0390131b 2172 if (sbi->s_journal)
e2b911c5 2173 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2174
e2d67052 2175 ext4_commit_super(sb, 1);
281b5995 2176done:
ac27a0ec 2177 if (test_opt(sb, DEBUG))
a9df9a49 2178 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2179 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2180 sb->s_blocksize,
2181 sbi->s_groups_count,
617ba13b
MC		 2182 EXT4_BLOCKS_PER_GROUP(sb),
2183 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2184 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2185
7abc52c2 2186 cleancache_init_fs(sb);
ac27a0ec
DK		 2187 return res;
2188}
2189
117fff10
TT		 2190int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2191{
2192 struct ext4_sb_info *sbi = EXT4_SB(sb);
2193 struct flex_groups *new_groups;
2194 int size;
2195
2196 if (!sbi->s_log_groups_per_flex)
2197 return 0;
2198
2199 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2200 if (size <= sbi->s_flex_groups_allocated)
2201 return 0;
2202
2203 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
a7c3e901 2204 new_groups = kvzalloc(size, GFP_KERNEL);
117fff10
TT		 2205 if (!new_groups) {
2206 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2207 size / (int) sizeof(struct flex_groups));
2208 return -ENOMEM;
2209 }
2210
2211 if (sbi->s_flex_groups) {
2212 memcpy(new_groups, sbi->s_flex_groups,
2213 (sbi->s_flex_groups_allocated *
2214 sizeof(struct flex_groups)));
b93b41d4 2215 kvfree(sbi->s_flex_groups);
117fff10
TT		 2216 }
2217 sbi->s_flex_groups = new_groups;
2218 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2219 return 0;
2220}
2221
772cb7c8
JS		 2222static int ext4_fill_flex_info(struct super_block *sb)
2223{
2224 struct ext4_sb_info *sbi = EXT4_SB(sb);
2225 struct ext4_group_desc *gdp = NULL;
772cb7c8 2226 ext4_group_t flex_group;
117fff10 2227 int i, err;
772cb7c8 2228
503358ae 2229 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2230 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2231 sbi->s_log_groups_per_flex = 0;
2232 return 1;
2233 }
2234
117fff10
TT		 2235 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2236 if (err)
9933fc0a 2237 goto failed;
772cb7c8 2238
772cb7c8 2239 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2240 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2241
2242 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2243 atomic_add(ext4_free_inodes_count(sb, gdp),
2244 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2245 atomic64_add(ext4_free_group_clusters(sb, gdp),
2246 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2247 atomic_add(ext4_used_dirs_count(sb, gdp),
2248 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2249 }
2250
2251 return 1;
2252failed:
2253 return 0;
2254}
2255
e2b911c5 2256static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2257 struct ext4_group_desc *gdp)
717d50e4 2258{
b47820ed 2259 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2260 __u16 crc = 0;
feb0ab32 2261 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2262 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2263
9aa5d32b 2264 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2265 /* Use new metadata_csum algorithm */
feb0ab32 2266 __u32 csum32;
b47820ed 2267 __u16 dummy_csum = 0;
feb0ab32 2268
feb0ab32
DW		 2269 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2270 sizeof(le_group));
b47820ed
DJ		 2271 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2272 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2273 sizeof(dummy_csum));
2274 offset += sizeof(dummy_csum);
2275 if (offset < sbi->s_desc_size)
2276 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2277 sbi->s_desc_size - offset);
feb0ab32
DW		 2278
2279 crc = csum32 & 0xFFFF;
2280 goto out;
717d50e4
AD		 2281 }
2282
feb0ab32 2283 /* old crc16 code */
e2b911c5 2284 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2285 return 0;
2286
feb0ab32
DW		 2287 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2288 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2289 crc = crc16(crc, (__u8 *)gdp, offset);
2290 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2291 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2292 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2293 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2294 crc = crc16(crc, (__u8 *)gdp + offset,
2295 le16_to_cpu(sbi->s_es->s_desc_size) -
2296 offset);
2297
2298out:
717d50e4
AD		 2299 return cpu_to_le16(crc);
2300}
2301
feb0ab32 2302int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2303 struct ext4_group_desc *gdp)
2304{
feb0ab32 2305 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2306 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2307 return 0;
2308
2309 return 1;
2310}
2311
feb0ab32
DW		 2312void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2313 struct ext4_group_desc *gdp)
2314{
2315 if (!ext4_has_group_desc_csum(sb))
2316 return;
e2b911c5 2317 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2318}
2319
ac27a0ec 2320/* Called at mount-time, super-block is locked */
bfff6873 2321static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2322 ext4_fsblk_t sb_block,
bfff6873 2323 ext4_group_t *first_not_zeroed)
ac27a0ec 2324{
617ba13b
MC		 2325 struct ext4_sb_info *sbi = EXT4_SB(sb);
2326 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2327 ext4_fsblk_t last_block;
bd81d8ee
LV		 2328 ext4_fsblk_t block_bitmap;
2329 ext4_fsblk_t inode_bitmap;
2330 ext4_fsblk_t inode_table;
ce421581 2331 int flexbg_flag = 0;
bfff6873 2332 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2333
e2b911c5 2334 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2335 flexbg_flag = 1;
2336
af5bc92d 2337 ext4_debug("Checking group descriptors");
ac27a0ec 2338
197cd65a
AM		 2339 for (i = 0; i < sbi->s_groups_count; i++) {
2340 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2341
ce421581 2342 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2343 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2344 else
2345 last_block = first_block +
617ba13b 2346 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2347
bfff6873
LC		 2348 if ((grp == sbi->s_groups_count) &&
2349 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2350 grp = i;
2351
8fadc143 2352 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2353 if (block_bitmap == sb_block) {
2354 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2355 "Block bitmap for group %u overlaps "
2356 "superblock", i);
2357 }
2b2d6d01 2358 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2359 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2360 "Block bitmap for group %u not in group "
b31e1552 2361 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2362 return 0;
2363 }
8fadc143 2364 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2365 if (inode_bitmap == sb_block) {
2366 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2367 "Inode bitmap for group %u overlaps "
2368 "superblock", i);
2369 }
2b2d6d01 2370 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2371 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2372 "Inode bitmap for group %u not in group "
b31e1552 2373 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2374 return 0;
2375 }
8fadc143 2376 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2377 if (inode_table == sb_block) {
2378 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2379 "Inode table for group %u overlaps "
2380 "superblock", i);
2381 }
bd81d8ee 2382 if (inode_table < first_block ||
2b2d6d01 2383 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2384 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2385 "Inode table for group %u not in group "
b31e1552 2386 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2387 return 0;
2388 }
955ce5f5 2389 ext4_lock_group(sb, i);
feb0ab32 2390 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2391 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2392 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2393 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2394 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2395 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2396 ext4_unlock_group(sb, i);
8a266467 2397 return 0;
7ee1ec4c 2398 }
717d50e4 2399 }
955ce5f5 2400 ext4_unlock_group(sb, i);
ce421581
JS		 2401 if (!flexbg_flag)
2402 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2403 }
bfff6873
LC		 2404 if (NULL != first_not_zeroed)
2405 *first_not_zeroed = grp;
ac27a0ec
DK		 2406 return 1;
2407}
2408
617ba13b 2409/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2410 * the superblock) which were deleted from all directories, but held open by
2411 * a process at the time of a crash. We walk the list and try to delete these
2412 * inodes at recovery time (only with a read-write filesystem).
2413 *
2414 * In order to keep the orphan inode chain consistent during traversal (in
2415 * case of crash during recovery), we link each inode into the superblock
2416 * orphan list_head and handle it the same way as an inode deletion during
2417 * normal operation (which journals the operations for us).
2418 *
2419 * We only do an iget() and an iput() on each inode, which is very safe if we
2420 * accidentally point at an in-use or already deleted inode. The worst that
2421 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2422 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2423 * e2fsck was run on this filesystem, and it must have already done the orphan
2424 * inode cleanup for us, so we can safely abort without any further action.
2425 */
2b2d6d01
TT		 2426static void ext4_orphan_cleanup(struct super_block *sb,
2427 struct ext4_super_block *es)
ac27a0ec
DK		 2428{
2429 unsigned int s_flags = sb->s_flags;
2c98eb5e 2430 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec 2431#ifdef CONFIG_QUOTA
8c6cd47e 2432 int quota_update = 0;
ac27a0ec
DK		 2433 int i;
2434#endif
2435 if (!es->s_last_orphan) {
2436 jbd_debug(4, "no orphan inodes to clean up\n");
2437 return;
2438 }
2439
a8f48a95 2440 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2441 ext4_msg(sb, KERN_ERR, "write access "
2442 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2443 return;
2444 }
2445
d39195c3
AG		 2446 /* Check if feature set would not allow a r/w mount */
2447 if (!ext4_feature_set_ok(sb, 0)) {
2448 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2449 "unknown ROCOMPAT features");
2450 return;
2451 }
2452
617ba13b 2453 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2454 /* don't clear list on RO mount w/ errors */
2455 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2456 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2457 "clearing orphan list.\n");
c25f9bc6
ES		 2458 es->s_last_orphan = 0;
2459 }
ac27a0ec
DK		 2460 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2461 return;
2462 }
2463
2464 if (s_flags & MS_RDONLY) {
b31e1552 2465 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2466 sb->s_flags &= ~MS_RDONLY;
2467 }
2468#ifdef CONFIG_QUOTA
2469 /* Needed for iput() to work correctly and not trash data */
2470 sb->s_flags |= MS_ACTIVE;
8c6cd47e 2471
2472 /*
2473 * Turn on quotas which were not enabled for read-only mounts if
2474 * filesystem has quota feature, so that they are updated correctly.
2475 */
2476 if (ext4_has_feature_quota(sb) && (s_flags & MS_RDONLY)) {
2477 int ret = ext4_enable_quotas(sb);
2478
2479 if (!ret)
2480 quota_update = 1;
2481 else
2482 ext4_msg(sb, KERN_ERR,
2483 "Cannot turn on quotas: error %d", ret);
2484 }
2485
2486 /* Turn on journaled quotas used for old sytle */
a2d4a646 2487 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2488 if (EXT4_SB(sb)->s_qf_names[i]) {
2489 int ret = ext4_quota_on_mount(sb, i);
8c6cd47e 2490
2491 if (!ret)
2492 quota_update = 1;
2493 else
b31e1552
ES		 2494 ext4_msg(sb, KERN_ERR,
2495 "Cannot turn on journaled "
8c6cd47e 2496 "quota: type %d: error %d", i, ret);
ac27a0ec
DK		 2497 }
2498 }
2499#endif
2500
2501 while (es->s_last_orphan) {
2502 struct inode *inode;
2503
c65d5c6c
VN		 2504 /*
2505 * We may have encountered an error during cleanup; if
2506 * so, skip the rest.
2507 */
2508 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2509 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2510 es->s_last_orphan = 0;
2511 break;
2512 }
2513
97bd42b9
JB		 2514 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2515 if (IS_ERR(inode)) {
ac27a0ec
DK		 2516 es->s_last_orphan = 0;
2517 break;
2518 }
2519
617ba13b 2520 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2521 dquot_initialize(inode);
ac27a0ec 2522 if (inode->i_nlink) {
566370a2
PT		 2523 if (test_opt(sb, DEBUG))
2524 ext4_msg(sb, KERN_DEBUG,
2525 "%s: truncating inode %lu to %lld bytes",
2526 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2527 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2528 inode->i_ino, inode->i_size);
5955102c 2529 inode_lock(inode);
55f252c9 2530 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2531 ret = ext4_truncate(inode);
2532 if (ret)
2533 ext4_std_error(inode->i_sb, ret);
5955102c 2534 inode_unlock(inode);
ac27a0ec
DK		 2535 nr_truncates++;
2536 } else {
566370a2
PT		 2537 if (test_opt(sb, DEBUG))
2538 ext4_msg(sb, KERN_DEBUG,
2539 "%s: deleting unreferenced inode %lu",
2540 __func__, inode->i_ino);
ac27a0ec
DK		 2541 jbd_debug(2, "deleting unreferenced inode %lu\n",
2542 inode->i_ino);
2543 nr_orphans++;
2544 }
2545 iput(inode); /* The delete magic happens here! */
2546 }
2547
2b2d6d01 2548#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2549
2550 if (nr_orphans)
b31e1552
ES		 2551 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2552 PLURAL(nr_orphans));
ac27a0ec 2553 if (nr_truncates)
b31e1552
ES		 2554 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2555 PLURAL(nr_truncates));
ac27a0ec 2556#ifdef CONFIG_QUOTA
8c6cd47e 2557 /* Turn off quotas if they were enabled for orphan cleanup */
2558 if (quota_update) {
2559 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2560 if (sb_dqopt(sb)->files[i])
2561 dquot_quota_off(sb, i);
2562 }
ac27a0ec
DK		 2563 }
2564#endif
2565 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2566}
0b8e58a1 2567
cd2291a4
ES		 2568/*
2569 * Maximal extent format file size.
2570 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2571 * extent format containers, within a sector_t, and within i_blocks
2572 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2573 * so that won't be a limiting factor.
2574 *
f17722f9
LC		 2575 * However there is other limiting factor. We do store extents in the form
2576 * of starting block and length, hence the resulting length of the extent
2577 * covering maximum file size must fit into on-disk format containers as
2578 * well. Given that length is always by 1 unit bigger than max unit (because
2579 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2580 *
cd2291a4
ES		 2581 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2582 */
f287a1a5 2583static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2584{
2585 loff_t res;
2586 loff_t upper_limit = MAX_LFS_FILESIZE;
2587
2588 /* small i_blocks in vfs inode? */
f287a1a5 2589 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2590 /*
90c699a9 2591 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2592 * i_block represent total blocks in 512 bytes
2593 * 32 == size of vfs inode i_blocks * 8
2594 */
2595 upper_limit = (1LL << 32) - 1;
2596
2597 /* total blocks in file system block size */
2598 upper_limit >>= (blkbits - 9);
2599 upper_limit <<= blkbits;
2600 }
2601
f17722f9
LC		 2602 /*
2603 * 32-bit extent-start container, ee_block. We lower the maxbytes
2604 * by one fs block, so ee_len can cover the extent of maximum file
2605 * size
2606 */
2607 res = (1LL << 32) - 1;
cd2291a4 2608 res <<= blkbits;
cd2291a4
ES		 2609
2610 /* Sanity check against vm- & vfs- imposed limits */
2611 if (res > upper_limit)
2612 res = upper_limit;
2613
2614 return res;
2615}
ac27a0ec 2616
ac27a0ec 2617/*
cd2291a4 2618 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2619 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2620 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2621 */
f287a1a5 2622static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2623{
617ba13b 2624 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2625 int meta_blocks;
2626 loff_t upper_limit;
0b8e58a1
AD		 2627 /* This is calculated to be the largest file size for a dense, block
2628 * mapped file such that the file's total number of 512-byte sectors,
2629 * including data and all indirect blocks, does not exceed (2^48 - 1).
2630 *
2631 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2632 * number of 512-byte sectors of the file.
0fc1b451
AK		 2633 */
2634
f287a1a5 2635 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2636 /*
90c699a9 2637 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2638 * the inode i_block field represents total file blocks in
2639 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2640 */
2641 upper_limit = (1LL << 32) - 1;
2642
2643 /* total blocks in file system block size */
2644 upper_limit >>= (bits - 9);
2645
2646 } else {
8180a562
AK		 2647 /*
2648 * We use 48 bit ext4_inode i_blocks
2649 * With EXT4_HUGE_FILE_FL set the i_blocks
2650 * represent total number of blocks in
2651 * file system block size
2652 */
0fc1b451
AK		 2653 upper_limit = (1LL << 48) - 1;
2654
0fc1b451
AK		 2655 }
2656
2657 /* indirect blocks */
2658 meta_blocks = 1;
2659 /* double indirect blocks */
2660 meta_blocks += 1 + (1LL << (bits-2));
2661 /* tripple indirect blocks */
2662 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2663
2664 upper_limit -= meta_blocks;
2665 upper_limit <<= bits;
ac27a0ec
DK		 2666
2667 res += 1LL << (bits-2);
2668 res += 1LL << (2*(bits-2));
2669 res += 1LL << (3*(bits-2));
2670 res <<= bits;
2671 if (res > upper_limit)
2672 res = upper_limit;
0fc1b451
AK		 2673
2674 if (res > MAX_LFS_FILESIZE)
2675 res = MAX_LFS_FILESIZE;
2676
ac27a0ec
DK		 2677 return res;
2678}
2679
617ba13b 2680static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2681 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2682{
617ba13b 2683 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2684 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2685 int has_super = 0;
2686
2687 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2688
e2b911c5 2689 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2690 return logical_sb_block + nr + 1;
ac27a0ec 2691 bg = sbi->s_desc_per_block * nr;
617ba13b 2692 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2693 has_super = 1;
0b8e58a1 2694
bd63f6b0
DW		 2695 /*
2696 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2697 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2698 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2699 * compensate.
2700 */
2701 if (sb->s_blocksize == 1024 && nr == 0 &&
2702 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2703 has_super++;
2704
617ba13b 2705 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2706}
2707
c9de560d
AT		 2708/**
2709 * ext4_get_stripe_size: Get the stripe size.
2710 * @sbi: In memory super block info
2711 *
2712 * If we have specified it via mount option, then
2713 * use the mount option value. If the value specified at mount time is
2714 * greater than the blocks per group use the super block value.
2715 * If the super block value is greater than blocks per group return 0.
2716 * Allocator needs it be less than blocks per group.
2717 *
2718 */
2719static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2720{
2721 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2722 unsigned long stripe_width =
2723 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2724 int ret;
c9de560d
AT		 2725
2726 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 2727 ret = sbi->s_stripe;
5469d7c3 2728 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 2729 ret = stripe_width;
5469d7c3 2730 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 2731 ret = stride;
2732 else
2733 ret = 0;
c9de560d 2734
3eb08658
DE		 2735 /*
2736 * If the stripe width is 1, this makes no sense and
2737 * we set it to 0 to turn off stripe handling code.
2738 */
2739 if (ret <= 1)
2740 ret = 0;
c9de560d 2741
3eb08658 2742 return ret;
c9de560d 2743}
ac27a0ec 2744
a13fb1a4
ES		 2745/*
2746 * Check whether this filesystem can be mounted based on
2747 * the features present and the RDONLY/RDWR mount requested.
2748 * Returns 1 if this filesystem can be mounted as requested,
2749 * 0 if it cannot be.
2750 */
2751static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2752{
e2b911c5 2753 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2754 ext4_msg(sb, KERN_ERR,
2755 "Couldn't mount because of "
2756 "unsupported optional features (%x)",
2757 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2758 ~EXT4_FEATURE_INCOMPAT_SUPP));
2759 return 0;
2760 }
2761
2762 if (readonly)
2763 return 1;
2764
e2b911c5 2765 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b
DW		 2766 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2767 sb->s_flags |= MS_RDONLY;
2768 return 1;
2769 }
2770
a13fb1a4 2771 /* Check that feature set is OK for a read-write mount */
e2b911c5 2772 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2773 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2774 "unsupported optional features (%x)",
2775 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2776 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2777 return 0;
2778 }
2779 /*
2780 * Large file size enabled file system can only be mounted
2781 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2782 */
e2b911c5 2783 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2784 if (sizeof(blkcnt_t) < sizeof(u64)) {
2785 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2786 "cannot be mounted RDWR without "
2787 "CONFIG_LBDAF");
2788 return 0;
2789 }
2790 }
e2b911c5 2791 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2792 ext4_msg(sb, KERN_ERR,
2793 "Can't support bigalloc feature without "
2794 "extents feature\n");
2795 return 0;
2796 }
7c319d32
AK		 2797
2798#ifndef CONFIG_QUOTA
e2b911c5 2799 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2800 ext4_msg(sb, KERN_ERR,
2801 "Filesystem with quota feature cannot be mounted RDWR "
2802 "without CONFIG_QUOTA");
2803 return 0;
2804 }
689c958c
LX		 2805 if (ext4_has_feature_project(sb) && !readonly) {
2806 ext4_msg(sb, KERN_ERR,
2807 "Filesystem with project quota feature cannot be mounted RDWR "
2808 "without CONFIG_QUOTA");
2809 return 0;
2810 }
7c319d32 2811#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2812 return 1;
2813}
2814
66e61a9e
TT		 2815/*
2816 * This function is called once a day if we have errors logged
2817 * on the file system
2818 */
2819static void print_daily_error_info(unsigned long arg)
2820{
2821 struct super_block *sb = (struct super_block *) arg;
2822 struct ext4_sb_info *sbi;
2823 struct ext4_super_block *es;
2824
2825 sbi = EXT4_SB(sb);
2826 es = sbi->s_es;
2827
2828 if (es->s_error_count)
ae0f78de
TT		 2829 /* fsck newer than v1.41.13 is needed to clean this condition. */
2830 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2831 le32_to_cpu(es->s_error_count));
2832 if (es->s_first_error_time) {
ae0f78de 2833 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2834 sb->s_id, le32_to_cpu(es->s_first_error_time),
2835 (int) sizeof(es->s_first_error_func),
2836 es->s_first_error_func,
2837 le32_to_cpu(es->s_first_error_line));
2838 if (es->s_first_error_ino)
651e1c3b 2839 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2840 le32_to_cpu(es->s_first_error_ino));
2841 if (es->s_first_error_block)
651e1c3b 2842 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2843 le64_to_cpu(es->s_first_error_block));
651e1c3b 2844 printk(KERN_CONT "\n");
66e61a9e
TT		 2845 }
2846 if (es->s_last_error_time) {
ae0f78de 2847 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2848 sb->s_id, le32_to_cpu(es->s_last_error_time),
2849 (int) sizeof(es->s_last_error_func),
2850 es->s_last_error_func,
2851 le32_to_cpu(es->s_last_error_line));
2852 if (es->s_last_error_ino)
651e1c3b 2853 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2854 le32_to_cpu(es->s_last_error_ino));
2855 if (es->s_last_error_block)
651e1c3b 2856 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2857 le64_to_cpu(es->s_last_error_block));
651e1c3b 2858 printk(KERN_CONT "\n");
66e61a9e
TT		 2859 }
2860 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2861}
2862
bfff6873
LC		 2863/* Find next suitable group and run ext4_init_inode_table */
2864static int ext4_run_li_request(struct ext4_li_request *elr)
2865{
2866 struct ext4_group_desc *gdp = NULL;
2867 ext4_group_t group, ngroups;
2868 struct super_block *sb;
2869 unsigned long timeout = 0;
2870 int ret = 0;
2871
2872 sb = elr->lr_super;
2873 ngroups = EXT4_SB(sb)->s_groups_count;
2874
2875 for (group = elr->lr_next_group; group < ngroups; group++) {
2876 gdp = ext4_get_group_desc(sb, group, NULL);
2877 if (!gdp) {
2878 ret = 1;
2879 break;
2880 }
2881
2882 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2883 break;
2884 }
2885
7f511862 2886 if (group >= ngroups)
bfff6873
LC		 2887 ret = 1;
2888
2889 if (!ret) {
2890 timeout = jiffies;
2891 ret = ext4_init_inode_table(sb, group,
2892 elr->lr_timeout ? 0 : 1);
2893 if (elr->lr_timeout == 0) {
51ce6511
LC		 2894 timeout = (jiffies - timeout) *
2895 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2896 elr->lr_timeout = timeout;
2897 }
2898 elr->lr_next_sched = jiffies + elr->lr_timeout;
2899 elr->lr_next_group = group + 1;
2900 }
bfff6873
LC		 2901 return ret;
2902}
2903
2904/*
2905 * Remove lr_request from the list_request and free the
4ed5c033 2906 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2907 */
2908static void ext4_remove_li_request(struct ext4_li_request *elr)
2909{
2910 struct ext4_sb_info *sbi;
2911
2912 if (!elr)
2913 return;
2914
2915 sbi = elr->lr_sbi;
2916
2917 list_del(&elr->lr_request);
2918 sbi->s_li_request = NULL;
2919 kfree(elr);
2920}
2921
2922static void ext4_unregister_li_request(struct super_block *sb)
2923{
1bb933fb
LC		 2924 mutex_lock(&ext4_li_mtx);
2925 if (!ext4_li_info) {
2926 mutex_unlock(&ext4_li_mtx);
bfff6873 2927 return;
1bb933fb 2928 }
bfff6873
LC		 2929
2930 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2931 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2932 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2933 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2934}
2935
8f1f7453
ES		 2936static struct task_struct *ext4_lazyinit_task;
2937
bfff6873
LC		 2938/*
2939 * This is the function where ext4lazyinit thread lives. It walks
2940 * through the request list searching for next scheduled filesystem.
2941 * When such a fs is found, run the lazy initialization request
2942 * (ext4_rn_li_request) and keep track of the time spend in this
2943 * function. Based on that time we compute next schedule time of
2944 * the request. When walking through the list is complete, compute
2945 * next waking time and put itself into sleep.
2946 */
2947static int ext4_lazyinit_thread(void *arg)
2948{
2949 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2950 struct list_head *pos, *n;
2951 struct ext4_li_request *elr;
4ed5c033 2952 unsigned long next_wakeup, cur;
bfff6873
LC		 2953
2954 BUG_ON(NULL == eli);
2955
bfff6873
LC		 2956cont_thread:
2957 while (true) {
2958 next_wakeup = MAX_JIFFY_OFFSET;
2959
2960 mutex_lock(&eli->li_list_mtx);
2961 if (list_empty(&eli->li_request_list)) {
2962 mutex_unlock(&eli->li_list_mtx);
2963 goto exit_thread;
2964 }
bfff6873 2965 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 2966 int err = 0;
2967 int progress = 0;
bfff6873
LC		 2968 elr = list_entry(pos, struct ext4_li_request,
2969 lr_request);
2970
e22834f0
DM		 2971 if (time_before(jiffies, elr->lr_next_sched)) {
2972 if (time_before(elr->lr_next_sched, next_wakeup))
2973 next_wakeup = elr->lr_next_sched;
2974 continue;
2975 }
2976 if (down_read_trylock(&elr->lr_super->s_umount)) {
2977 if (sb_start_write_trylock(elr->lr_super)) {
2978 progress = 1;
2979 /*
2980 * We hold sb->s_umount, sb can not
2981 * be removed from the list, it is
2982 * now safe to drop li_list_mtx
2983 */
2984 mutex_unlock(&eli->li_list_mtx);
2985 err = ext4_run_li_request(elr);
2986 sb_end_write(elr->lr_super);
2987 mutex_lock(&eli->li_list_mtx);
2988 n = pos->next;
b2c78cd0 2989 }
e22834f0
DM		 2990 up_read((&elr->lr_super->s_umount));
2991 }
2992 /* error, remove the lazy_init job */
2993 if (err) {
2994 ext4_remove_li_request(elr);
2995 continue;
2996 }
2997 if (!progress) {
2998 elr->lr_next_sched = jiffies +
2999 (prandom_u32()
3000 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 3001 }
bfff6873
LC		 3002 if (time_before(elr->lr_next_sched, next_wakeup))
3003 next_wakeup = elr->lr_next_sched;
3004 }
3005 mutex_unlock(&eli->li_list_mtx);
3006
a0acae0e 3007 try_to_freeze();
bfff6873 3008
4ed5c033
LC		 3009 cur = jiffies;
3010 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3011 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3012 cond_resched();
3013 continue;
3014 }
3015
4ed5c033
LC		 3016 schedule_timeout_interruptible(next_wakeup - cur);
3017
8f1f7453
ES		 3018 if (kthread_should_stop()) {
3019 ext4_clear_request_list();
3020 goto exit_thread;
3021 }
bfff6873
LC		 3022 }
3023
3024exit_thread:
3025 /*
3026 * It looks like the request list is empty, but we need
3027 * to check it under the li_list_mtx lock, to prevent any
3028 * additions into it, and of course we should lock ext4_li_mtx
3029 * to atomically free the list and ext4_li_info, because at
3030 * this point another ext4 filesystem could be registering
3031 * new one.
3032 */
3033 mutex_lock(&ext4_li_mtx);
3034 mutex_lock(&eli->li_list_mtx);
3035 if (!list_empty(&eli->li_request_list)) {
3036 mutex_unlock(&eli->li_list_mtx);
3037 mutex_unlock(&ext4_li_mtx);
3038 goto cont_thread;
3039 }
3040 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3041 kfree(ext4_li_info);
3042 ext4_li_info = NULL;
3043 mutex_unlock(&ext4_li_mtx);
3044
3045 return 0;
3046}
3047
3048static void ext4_clear_request_list(void)
3049{
3050 struct list_head *pos, *n;
3051 struct ext4_li_request *elr;
3052
3053 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3054 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3055 elr = list_entry(pos, struct ext4_li_request,
3056 lr_request);
3057 ext4_remove_li_request(elr);
3058 }
3059 mutex_unlock(&ext4_li_info->li_list_mtx);
3060}
3061
3062static int ext4_run_lazyinit_thread(void)
3063{
8f1f7453
ES		 3064 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3065 ext4_li_info, "ext4lazyinit");
3066 if (IS_ERR(ext4_lazyinit_task)) {
3067 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3068 ext4_clear_request_list();
bfff6873
LC		 3069 kfree(ext4_li_info);
3070 ext4_li_info = NULL;
92b97816 3071 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3072 "initialization thread\n",
3073 err);
3074 return err;
3075 }
3076 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3077 return 0;
3078}
3079
3080/*
3081 * Check whether it make sense to run itable init. thread or not.
3082 * If there is at least one uninitialized inode table, return
3083 * corresponding group number, else the loop goes through all
3084 * groups and return total number of groups.
3085 */
3086static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3087{
3088 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3089 struct ext4_group_desc *gdp = NULL;
3090
3091 for (group = 0; group < ngroups; group++) {
3092 gdp = ext4_get_group_desc(sb, group, NULL);
3093 if (!gdp)
3094 continue;
3095
3096 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3097 break;
3098 }
3099
3100 return group;
3101}
3102
3103static int ext4_li_info_new(void)
3104{
3105 struct ext4_lazy_init *eli = NULL;
3106
3107 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3108 if (!eli)
3109 return -ENOMEM;
3110
bfff6873
LC		 3111 INIT_LIST_HEAD(&eli->li_request_list);
3112 mutex_init(&eli->li_list_mtx);
3113
bfff6873
LC		 3114 eli->li_state |= EXT4_LAZYINIT_QUIT;
3115
3116 ext4_li_info = eli;
3117
3118 return 0;
3119}
3120
3121static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3122 ext4_group_t start)
3123{
3124 struct ext4_sb_info *sbi = EXT4_SB(sb);
3125 struct ext4_li_request *elr;
bfff6873
LC		 3126
3127 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3128 if (!elr)
3129 return NULL;
3130
3131 elr->lr_super = sb;
3132 elr->lr_sbi = sbi;
3133 elr->lr_next_group = start;
3134
3135 /*
3136 * Randomize first schedule time of the request to
3137 * spread the inode table initialization requests
3138 * better.
3139 */
dd1f723b
TT		 3140 elr->lr_next_sched = jiffies + (prandom_u32() %
3141 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3142 return elr;
3143}
3144
7f511862
TT		 3145int ext4_register_li_request(struct super_block *sb,
3146 ext4_group_t first_not_zeroed)
bfff6873
LC		 3147{
3148 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3149 struct ext4_li_request *elr = NULL;
bfff6873 3150 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3151 int ret = 0;
bfff6873 3152
7f511862 3153 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3154 if (sbi->s_li_request != NULL) {
3155 /*
3156 * Reset timeout so it can be computed again, because
3157 * s_li_wait_mult might have changed.
3158 */
3159 sbi->s_li_request->lr_timeout = 0;
7f511862 3160 goto out;
51ce6511 3161 }
bfff6873
LC		 3162
3163 if (first_not_zeroed == ngroups ||
3164 (sb->s_flags & MS_RDONLY) ||
55ff3840 3165 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3166 goto out;
bfff6873
LC		 3167
3168 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3169 if (!elr) {
3170 ret = -ENOMEM;
3171 goto out;
3172 }
bfff6873
LC		 3173
3174 if (NULL == ext4_li_info) {
3175 ret = ext4_li_info_new();
3176 if (ret)
3177 goto out;
3178 }
3179
3180 mutex_lock(&ext4_li_info->li_list_mtx);
3181 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3182 mutex_unlock(&ext4_li_info->li_list_mtx);
3183
3184 sbi->s_li_request = elr;
46e4690b
TM		 3185 /*
3186 * set elr to NULL here since it has been inserted to
3187 * the request_list and the removal and free of it is
3188 * handled by ext4_clear_request_list from now on.
3189 */
3190 elr = NULL;
bfff6873
LC		 3191
3192 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3193 ret = ext4_run_lazyinit_thread();
3194 if (ret)
3195 goto out;
3196 }
bfff6873 3197out:
beed5ecb
NK		 3198 mutex_unlock(&ext4_li_mtx);
3199 if (ret)
bfff6873 3200 kfree(elr);
bfff6873
LC		 3201 return ret;
3202}
3203
3204/*
3205 * We do not need to lock anything since this is called on
3206 * module unload.
3207 */
3208static void ext4_destroy_lazyinit_thread(void)
3209{
3210 /*
3211 * If thread exited earlier
3212 * there's nothing to be done.
3213 */
8f1f7453 3214 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3215 return;
3216
8f1f7453 3217 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3218}
3219
25ed6e8a
DW		 3220static int set_journal_csum_feature_set(struct super_block *sb)
3221{
3222 int ret = 1;
3223 int compat, incompat;
3224 struct ext4_sb_info *sbi = EXT4_SB(sb);
3225
9aa5d32b 3226 if (ext4_has_metadata_csum(sb)) {
db9ee220 3227 /* journal checksum v3 */
25ed6e8a 3228 compat = 0;
db9ee220 3229 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3230 } else {
3231 /* journal checksum v1 */
3232 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3233 incompat = 0;
3234 }
3235
feb8c6d3
DW		 3236 jbd2_journal_clear_features(sbi->s_journal,
3237 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3238 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3239 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3240 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3241 ret = jbd2_journal_set_features(sbi->s_journal,
3242 compat, 0,
3243 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3244 incompat);
3245 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3246 ret = jbd2_journal_set_features(sbi->s_journal,
3247 compat, 0,
3248 incompat);
3249 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3250 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3251 } else {
feb8c6d3
DW		 3252 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3253 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3254 }
3255
3256 return ret;
3257}
3258
952fc18e
TT		 3259/*
3260 * Note: calculating the overhead so we can be compatible with
3261 * historical BSD practice is quite difficult in the face of
3262 * clusters/bigalloc. This is because multiple metadata blocks from
3263 * different block group can end up in the same allocation cluster.
3264 * Calculating the exact overhead in the face of clustered allocation
3265 * requires either O(all block bitmaps) in memory or O(number of block
3266 * groups**2) in time. We will still calculate the superblock for
3267 * older file systems --- and if we come across with a bigalloc file
3268 * system with zero in s_overhead_clusters the estimate will be close to
3269 * correct especially for very large cluster sizes --- but for newer
3270 * file systems, it's better to calculate this figure once at mkfs
3271 * time, and store it in the superblock. If the superblock value is
3272 * present (even for non-bigalloc file systems), we will use it.
3273 */
3274static int count_overhead(struct super_block *sb, ext4_group_t grp,
3275 char *buf)
3276{
3277 struct ext4_sb_info *sbi = EXT4_SB(sb);
3278 struct ext4_group_desc *gdp;
3279 ext4_fsblk_t first_block, last_block, b;
3280 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3281 int s, j, count = 0;
3282
e2b911c5 3283 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3284 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3285 sbi->s_itb_per_group + 2);
3286
952fc18e
TT		 3287 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3288 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3289 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3290 for (i = 0; i < ngroups; i++) {
3291 gdp = ext4_get_group_desc(sb, i, NULL);
3292 b = ext4_block_bitmap(sb, gdp);
3293 if (b >= first_block && b <= last_block) {
3294 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3295 count++;
3296 }
3297 b = ext4_inode_bitmap(sb, gdp);
3298 if (b >= first_block && b <= last_block) {
3299 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3300 count++;
3301 }
3302 b = ext4_inode_table(sb, gdp);
3303 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3304 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3305 int c = EXT4_B2C(sbi, b - first_block);
3306 ext4_set_bit(c, buf);
3307 count++;
3308 }
3309 if (i != grp)
3310 continue;
3311 s = 0;
3312 if (ext4_bg_has_super(sb, grp)) {
3313 ext4_set_bit(s++, buf);
3314 count++;
3315 }
c48ae41b
TT		 3316 j = ext4_bg_num_gdb(sb, grp);
3317 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3318 ext4_error(sb, "Invalid number of block group "
3319 "descriptor blocks: %d", j);
3320 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3321 }
c48ae41b
TT		 3322 count += j;
3323 for (; j > 0; j--)
3324 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3325 }
3326 if (!count)
3327 return 0;
3328 return EXT4_CLUSTERS_PER_GROUP(sb) -
3329 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3330}
3331
3332/*
3333 * Compute the overhead and stash it in sbi->s_overhead
3334 */
3335int ext4_calculate_overhead(struct super_block *sb)
3336{
3337 struct ext4_sb_info *sbi = EXT4_SB(sb);
3338 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3339 struct inode *j_inode;
3340 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3341 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3342 ext4_fsblk_t overhead = 0;
4fdb5543 3343 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3344
952fc18e
TT		 3345 if (!buf)
3346 return -ENOMEM;
3347
3348 /*
3349 * Compute the overhead (FS structures). This is constant
3350 * for a given filesystem unless the number of block groups
3351 * changes so we cache the previous value until it does.
3352 */
3353
3354 /*
3355 * All of the blocks before first_data_block are overhead
3356 */
3357 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3358
3359 /*
3360 * Add the overhead found in each block group
3361 */
3362 for (i = 0; i < ngroups; i++) {
3363 int blks;
3364
3365 blks = count_overhead(sb, i, buf);
3366 overhead += blks;
3367 if (blks)
3368 memset(buf, 0, PAGE_SIZE);
3369 cond_resched();
3370 }
3c816ded
EW		 3371
3372 /*
3373 * Add the internal journal blocks whether the journal has been
3374 * loaded or not
3375 */
b003b524 3376 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3377 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3378 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3379 j_inode = ext4_get_journal_inode(sb, j_inum);
3380 if (j_inode) {
3381 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3382 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3383 iput(j_inode);
3384 } else {
3385 ext4_msg(sb, KERN_ERR, "can't get journal size");
3386 }
3387 }
952fc18e
TT		 3388 sbi->s_overhead = overhead;
3389 smp_wmb();
3390 free_page((unsigned long) buf);
3391 return 0;
3392}
3393
b5799018 3394static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3395{
3396 ext4_fsblk_t resv_clusters;
b5799018 3397 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3398
30fac0f7
JK		 3399 /*
3400 * There's no need to reserve anything when we aren't using extents.
3401 * The space estimates are exact, there are no unwritten extents,
3402 * hole punching doesn't need new metadata... This is needed especially
3403 * to keep ext2/3 backward compatibility.
3404 */
e2b911c5 3405 if (!ext4_has_feature_extents(sb))
b5799018 3406 return;
27dd4385
LC		 3407 /*
3408 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3409 * This should cover the situations where we can not afford to run
3410 * out of space like for example punch hole, or converting
556615dc 3411 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3412 * allocation would require 1, or 2 blocks, higher numbers are
3413 * very rare.
3414 */
b5799018
TT		 3415 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3416 sbi->s_cluster_bits);
27dd4385
LC		 3417
3418 do_div(resv_clusters, 50);
3419 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3420
b5799018 3421 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3422}
3423
2b2d6d01 3424static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3425{
d4c402d9 3426 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3427 struct buffer_head *bh;
617ba13b 3428 struct ext4_super_block *es = NULL;
5aee0f8a 3429 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3430 ext4_fsblk_t block;
3431 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3432 ext4_fsblk_t logical_sb_block;
ac27a0ec 3433 unsigned long offset = 0;
ac27a0ec
DK		 3434 unsigned long journal_devnum = 0;
3435 unsigned long def_mount_opts;
3436 struct inode *root;
0390131b 3437 const char *descr;
dcc7dae3 3438 int ret = -ENOMEM;
281b5995 3439 int blocksize, clustersize;
4ec11028
TT		 3440 unsigned int db_count;
3441 unsigned int i;
281b5995 3442 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3443 __u64 blocks_count;
07aa2ea1 3444 int err = 0;
b3881f74 3445 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3446 ext4_group_t first_not_zeroed;
ac27a0ec 3447
5aee0f8a
TT		 3448 if ((data && !orig_data) || !sbi)
3449 goto out_free_base;
705895b6 3450
40bd02a3
SF		 3451 if (!userns_mounts && !capable(CAP_SYS_ADMIN)) {
3452 ret = -EPERM;
3453 goto out_free_base;
3454 }
3455
705895b6
PE		 3456 sbi->s_blockgroup_lock =
3457 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3458 if (!sbi->s_blockgroup_lock)
3459 goto out_free_base;
3460
ac27a0ec 3461 sb->s_fs_info = sbi;
2c0544b2 3462 sbi->s_sb = sb;
240799cd 3463 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3464 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3465 if (sb->s_bdev->bd_part)
3466 sbi->s_sectors_written_start =
3467 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3468
9f6200bb 3469 /* Cleanup superblock name */
ec3904dc 3470 strreplace(sb->s_id, '/', '!');
9f6200bb 3471
07aa2ea1 3472 /* -EINVAL is default */
dcc7dae3 3473 ret = -EINVAL;
617ba13b 3474 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3475 if (!blocksize) {
b31e1552 3476 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3477 goto out_fail;
3478 }
3479
3480 /*
617ba13b 3481 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3482 * block sizes. We need to calculate the offset from buffer start.
3483 */
617ba13b 3484 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3485 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3486 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3487 } else {
70bbb3e0 3488 logical_sb_block = sb_block;
ac27a0ec
DK		 3489 }
3490
a8ac900b 3491 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3492 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3493 goto out_fail;
3494 }
3495 /*
3496 * Note: s_es must be initialized as soon as possible because
617ba13b 3497 * some ext4 macro-instructions depend on its value
ac27a0ec 3498 */
2716b802 3499 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3500 sbi->s_es = es;
3501 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3502 if (sb->s_magic != EXT4_SUPER_MAGIC)
3503 goto cantfind_ext4;
afc32f7e 3504 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3505
feb0ab32 3506 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3507 if (ext4_has_feature_metadata_csum(sb) &&
3508 ext4_has_feature_gdt_csum(sb))
363307e6 3509 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3510 "redundant flags; please run fsck.");
3511
d25425f8
DW		 3512 /* Check for a known checksum algorithm */
3513 if (!ext4_verify_csum_type(sb, es)) {
3514 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3515 "unknown checksum algorithm.");
3516 silent = 1;
3517 goto cantfind_ext4;
3518 }
3519
0441984a 3520 /* Load the checksum driver */
dec214d0
TE		 3521 if (ext4_has_feature_metadata_csum(sb) ||
3522 ext4_has_feature_ea_inode(sb)) {
0441984a
DW		 3523 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3524 if (IS_ERR(sbi->s_chksum_driver)) {
3525 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3526 ret = PTR_ERR(sbi->s_chksum_driver);
3527 sbi->s_chksum_driver = NULL;
3528 goto failed_mount;
3529 }
3530 }
3531
a9c47317
DW		 3532 /* Check superblock checksum */
3533 if (!ext4_superblock_csum_verify(sb, es)) {
3534 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3535 "invalid superblock checksum. Run e2fsck?");
3536 silent = 1;
6a797d27 3537 ret = -EFSBADCRC;
a9c47317
DW		 3538 goto cantfind_ext4;
3539 }
3540
3541 /* Precompute checksum seed for all metadata */
e2b911c5 3542 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 3543 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 3544 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 3545 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3546 sizeof(es->s_uuid));
3547
ac27a0ec
DK		 3548 /* Set defaults before we parse the mount options */
3549 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3550 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3551 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3552 set_opt(sb, DEBUG);
87f26807 3553 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3554 set_opt(sb, GRPID);
617ba13b 3555 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3556 set_opt(sb, NO_UID32);
ea663336 3557 /* xattr user namespace & acls are now defaulted on */
ea663336 3558 set_opt(sb, XATTR_USER);
03010a33 3559#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3560 set_opt(sb, POSIX_ACL);
2e7842b8 3561#endif
98c1a759
DW		 3562 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3563 if (ext4_has_metadata_csum(sb))
3564 set_opt(sb, JOURNAL_CHECKSUM);
3565
617ba13b 3566 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3567 set_opt(sb, JOURNAL_DATA);
617ba13b 3568 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3569 set_opt(sb, ORDERED_DATA);
617ba13b 3570 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3571 set_opt(sb, WRITEBACK_DATA);
617ba13b 3572
afdbd4a6
SF		 3573 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) {
3574 if (!capable(CAP_SYS_ADMIN))
3575 goto failed_mount;
fd8c37ec 3576 set_opt(sb, ERRORS_PANIC);
afdbd4a6 3577 } else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) {
fd8c37ec 3578 set_opt(sb, ERRORS_CONT);
afdbd4a6 3579 } else {
fd8c37ec 3580 set_opt(sb, ERRORS_RO);
afdbd4a6 3581 }
45f1a9c3
DW		 3582 /* block_validity enabled by default; disable with noblock_validity */
3583 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3584 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3585 set_opt(sb, DISCARD);
ac27a0ec 3586
afdbd4a6
SF		 3587 sbi->s_resuid = make_kuid(sb->s_user_ns, le16_to_cpu(es->s_def_resuid));
3588 if (!uid_valid(sbi->s_resuid))
3589 sbi->s_resuid = make_kuid(sb->s_user_ns, EXT4_DEF_RESUID);
3590 sbi->s_resgid = make_kgid(sb->s_user_ns, le16_to_cpu(es->s_def_resgid));
3591 if (!gid_valid(sbi->s_resgid))
3592 sbi->s_resgid = make_kgid(sb->s_user_ns, EXT4_DEF_RESGID);
30773840
TT		 3593 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3594 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3595 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3596
8b67f04a 3597 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3598 set_opt(sb, BARRIER);
ac27a0ec 3599
dd919b98
AK		 3600 /*
3601 * enable delayed allocation by default
3602 * Use -o nodelalloc to turn it off
3603 */
bc0b75f7 3604 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3605 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3606 set_opt(sb, DELALLOC);
dd919b98 3607
51ce6511
LC		 3608 /*
3609 * set default s_li_wait_mult for lazyinit, for the case there is
3610 * no mount option specified.
3611 */
3612 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3613
5aee0f8a
TT		 3614 if (sbi->s_es->s_mount_opts[0]) {
3615 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3616 sizeof(sbi->s_es->s_mount_opts),
3617 GFP_KERNEL);
3618 if (!s_mount_opts)
3619 goto failed_mount;
3620 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3621 &journal_ioprio, 0)) {
3622 ext4_msg(sb, KERN_WARNING,
3623 "failed to parse options in superblock: %s",
3624 s_mount_opts);
3625 }
3626 kfree(s_mount_opts);
8b67f04a 3627 }
5a916be1 3628 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3629 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3630 &journal_ioprio, 0))
ac27a0ec
DK		 3631 goto failed_mount;
3632
56889787
TT		 3633 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3634 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3635 "with data=journal disables delayed "
3636 "allocation and O_DIRECT support!\n");
3637 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3638 ext4_msg(sb, KERN_ERR, "can't mount with "
3639 "both data=journal and delalloc");
3640 goto failed_mount;
3641 }
3642 if (test_opt(sb, DIOREAD_NOLOCK)) {
3643 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3644 "both data=journal and dioread_nolock");
56889787
TT		 3645 goto failed_mount;
3646 }
923ae0ff
RZ		 3647 if (test_opt(sb, DAX)) {
3648 ext4_msg(sb, KERN_ERR, "can't mount with "
3649 "both data=journal and dax");
3650 goto failed_mount;
3651 }
73b92a2a
SK		 3652 if (ext4_has_feature_encrypt(sb)) {
3653 ext4_msg(sb, KERN_WARNING,
3654 "encrypted files will use data=ordered "
3655 "instead of data journaling mode");
3656 }
56889787
TT		 3657 if (test_opt(sb, DELALLOC))
3658 clear_opt(sb, DELALLOC);
001e4a87
TH		 3659 } else {
3660 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3661 }
3662
ac27a0ec 3663 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3664 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3665
617ba13b 3666 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3667 (ext4_has_compat_features(sb) ||
3668 ext4_has_ro_compat_features(sb) ||
3669 ext4_has_incompat_features(sb)))
b31e1552
ES		 3670 ext4_msg(sb, KERN_WARNING,
3671 "feature flags set on rev 0 fs, "
3672 "running e2fsck is recommended");
469108ff 3673
ed3654eb
TT		 3674 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3675 set_opt2(sb, HURD_COMPAT);
e2b911c5 3676 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3677 ext4_msg(sb, KERN_ERR,
3678 "The Hurd can't support 64-bit file systems");
3679 goto failed_mount;
3680 }
dec214d0
TE		 3681
3682 /*
3683 * ea_inode feature uses l_i_version field which is not
3684 * available in HURD_COMPAT mode.
3685 */
3686 if (ext4_has_feature_ea_inode(sb)) {
3687 ext4_msg(sb, KERN_ERR,
3688 "ea_inode feature is not supported for Hurd");
3689 goto failed_mount;
3690 }
ed3654eb
TT		 3691 }
3692
2035e776
TT		 3693 if (IS_EXT2_SB(sb)) {
3694 if (ext2_feature_set_ok(sb))
3695 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3696 "using the ext4 subsystem");
3697 else {
3698 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3699 "to feature incompatibilities");
3700 goto failed_mount;
3701 }
3702 }
3703
3704 if (IS_EXT3_SB(sb)) {
3705 if (ext3_feature_set_ok(sb))
3706 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3707 "using the ext4 subsystem");
3708 else {
3709 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3710 "to feature incompatibilities");
3711 goto failed_mount;
3712 }
3713 }
3714
ac27a0ec
DK		 3715 /*
3716 * Check feature flags regardless of the revision level, since we
3717 * previously didn't change the revision level when setting the flags,
3718 * so there is a chance incompat flags are set on a rev 0 filesystem.
3719 */
a13fb1a4 3720 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3721 goto failed_mount;
a13fb1a4 3722
261cb20c 3723 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3724 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3725 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3726 ext4_msg(sb, KERN_ERR,
8cdf3372
TT		 3727 "Unsupported filesystem blocksize %d (%d log_block_size)",
3728 blocksize, le32_to_cpu(es->s_log_block_size));
3729 goto failed_mount;
3730 }
3731 if (le32_to_cpu(es->s_log_block_size) >
3732 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3733 ext4_msg(sb, KERN_ERR,
3734 "Invalid log block size: %u",
3735 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3736 goto failed_mount;
3737 }
3738
5b9554dc
TT		 3739 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3740 ext4_msg(sb, KERN_ERR,
3741 "Number of reserved GDT blocks insanely large: %d",
3742 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3743 goto failed_mount;
3744 }
3745
923ae0ff 3746 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
87eefeb4
TK		 3747 err = bdev_dax_supported(sb, blocksize);
3748 if (err)
923ae0ff 3749 goto failed_mount;
923ae0ff
RZ		 3750 }
3751
e2b911c5 3752 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3753 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3754 es->s_encryption_level);
3755 goto failed_mount;
3756 }
3757
ac27a0ec 3758 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3759 /* Validate the filesystem blocksize */
3760 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3761 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3762 blocksize);
ac27a0ec
DK		 3763 goto failed_mount;
3764 }
3765
2b2d6d01 3766 brelse(bh);
70bbb3e0
AM		 3767 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3768 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3769 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3770 if (!bh) {
b31e1552
ES		 3771 ext4_msg(sb, KERN_ERR,
3772 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3773 goto failed_mount;
3774 }
2716b802 3775 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3776 sbi->s_es = es;
617ba13b 3777 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3778 ext4_msg(sb, KERN_ERR,
3779 "Magic mismatch, very weird!");
ac27a0ec
DK		 3780 goto failed_mount;
3781 }
3782 }
3783
e2b911c5 3784 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3785 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3786 has_huge_files);
3787 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3788
617ba13b
MC		 3789 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3790 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3791 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3792 } else {
3793 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3794 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3795 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3796 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3797 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3798 ext4_msg(sb, KERN_ERR,
3799 "unsupported inode size: %d",
2b2d6d01 3800 sbi->s_inode_size);
ac27a0ec
DK		 3801 goto failed_mount;
3802 }
ef7f3835
KS		 3803 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3804 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3805 }
0b8e58a1 3806
0d1ee42f 3807 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3808 if (ext4_has_feature_64bit(sb)) {
8fadc143 3809 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3810 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3811 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3812 ext4_msg(sb, KERN_ERR,
3813 "unsupported descriptor size %lu",
0d1ee42f
AR		 3814 sbi->s_desc_size);
3815 goto failed_mount;
3816 }
3817 } else
3818 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3819
ac27a0ec 3820 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3821 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 3822
617ba13b 3823 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3824 if (sbi->s_inodes_per_block == 0)
617ba13b 3825 goto cantfind_ext4;
cd6bb35b
TT		 3826 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3827 sbi->s_inodes_per_group > blocksize * 8) {
3828 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3829 sbi->s_blocks_per_group);
3830 goto failed_mount;
3831 }
ac27a0ec
DK		 3832 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3833 sbi->s_inodes_per_block;
0d1ee42f 3834 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3835 sbi->s_sbh = bh;
3836 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3837 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3838 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3839
2b2d6d01 3840 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3841 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3842 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3843 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3844 i = le32_to_cpu(es->s_flags);
3845 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3846 sbi->s_hash_unsigned = 3;
3847 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3848#ifdef __CHAR_UNSIGNED__
23301410
TT		 3849 if (!(sb->s_flags & MS_RDONLY))
3850 es->s_flags |=
3851 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3852 sbi->s_hash_unsigned = 3;
f99b2589 3853#else
23301410
TT		 3854 if (!(sb->s_flags & MS_RDONLY))
3855 es->s_flags |=
3856 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3857#endif
23301410 3858 }
f99b2589 3859 }
ac27a0ec 3860
281b5995
TT		 3861 /* Handle clustersize */
3862 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3863 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3864 if (has_bigalloc) {
3865 if (clustersize < blocksize) {
3866 ext4_msg(sb, KERN_ERR,
3867 "cluster size (%d) smaller than "
3868 "block size (%d)", clustersize, blocksize);
3869 goto failed_mount;
3870 }
8cdf3372
TT		 3871 if (le32_to_cpu(es->s_log_cluster_size) >
3872 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3873 ext4_msg(sb, KERN_ERR,
3874 "Invalid log cluster size: %u",
3875 le32_to_cpu(es->s_log_cluster_size));
3876 goto failed_mount;
3877 }
281b5995
TT		 3878 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3879 le32_to_cpu(es->s_log_block_size);
3880 sbi->s_clusters_per_group =
3881 le32_to_cpu(es->s_clusters_per_group);
3882 if (sbi->s_clusters_per_group > blocksize * 8) {
3883 ext4_msg(sb, KERN_ERR,
3884 "#clusters per group too big: %lu",
3885 sbi->s_clusters_per_group);
3886 goto failed_mount;
3887 }
3888 if (sbi->s_blocks_per_group !=
3889 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3890 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3891 "clusters per group (%lu) inconsistent",
3892 sbi->s_blocks_per_group,
3893 sbi->s_clusters_per_group);
3894 goto failed_mount;
3895 }
3896 } else {
3897 if (clustersize != blocksize) {
3898 ext4_warning(sb, "fragment/cluster size (%d) != "
3899 "block size (%d)", clustersize,
3900 blocksize);
3901 clustersize = blocksize;
3902 }
3903 if (sbi->s_blocks_per_group > blocksize * 8) {
3904 ext4_msg(sb, KERN_ERR,
3905 "#blocks per group too big: %lu",
3906 sbi->s_blocks_per_group);
3907 goto failed_mount;
3908 }
3909 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3910 sbi->s_cluster_bits = 0;
ac27a0ec 3911 }
281b5995
TT		 3912 sbi->s_cluster_ratio = clustersize / blocksize;
3913
960fd856
TT		 3914 /* Do we have standard group size of clustersize * 8 blocks ? */
3915 if (sbi->s_blocks_per_group == clustersize << 3)
3916 set_opt2(sb, STD_GROUP_SIZE);
3917
bf43d84b
ES		 3918 /*
3919 * Test whether we have more sectors than will fit in sector_t,
3920 * and whether the max offset is addressable by the page cache.
3921 */
5a9ae68a 3922 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3923 ext4_blocks_count(es));
5a9ae68a 3924 if (err) {
b31e1552 3925 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3926 " too large to mount safely on this system");
ac27a0ec 3927 if (sizeof(sector_t) < 8)
90c699a9 3928 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3929 goto failed_mount;
3930 }
3931
617ba13b
MC		 3932 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3933 goto cantfind_ext4;
e7c95593 3934
0f2ddca6
FTN		 3935 /* check blocks count against device size */
3936 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3937 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3938 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3939 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3940 ext4_blocks_count(es), blocks_count);
3941 goto failed_mount;
3942 }
3943
0b8e58a1
AD		 3944 /*
3945 * It makes no sense for the first data block to be beyond the end
3946 * of the filesystem.
3947 */
3948 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3949 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3950 "block %u is beyond end of filesystem (%llu)",
3951 le32_to_cpu(es->s_first_data_block),
3952 ext4_blocks_count(es));
e7c95593
ES		 3953 goto failed_mount;
3954 }
bd81d8ee
LV		 3955 blocks_count = (ext4_blocks_count(es) -
3956 le32_to_cpu(es->s_first_data_block) +
3957 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3958 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3959 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3960 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3961 "(block count %llu, first data block %u, "
b31e1552 3962 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3963 ext4_blocks_count(es),
3964 le32_to_cpu(es->s_first_data_block),
3965 EXT4_BLOCKS_PER_GROUP(sb));
3966 goto failed_mount;
3967 }
bd81d8ee 3968 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3969 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3970 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3971 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3972 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 3973 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 3974 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 3975 ext4_msg(sb, KERN_WARNING,
3976 "first meta block group too large: %u "
3977 "(group descriptor block count %u)",
3978 le32_to_cpu(es->s_first_meta_bg), db_count);
3979 goto failed_mount;
3980 }
3981 }
a7c3e901 3982 sbi->s_group_desc = kvmalloc(db_count *
f18a5f21
TT		 3983 sizeof(struct buffer_head *),
3984 GFP_KERNEL);
ac27a0ec 3985 if (sbi->s_group_desc == NULL) {
b31e1552 3986 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3987 ret = -ENOMEM;
ac27a0ec
DK		 3988 goto failed_mount;
3989 }
3990
705895b6 3991 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 3992
85c8f176
AP		 3993 /* Pre-read the descriptors into the buffer cache */
3994 for (i = 0; i < db_count; i++) {
3995 block = descriptor_loc(sb, logical_sb_block, i);
3996 sb_breadahead(sb, block);
3997 }
3998
ac27a0ec 3999 for (i = 0; i < db_count; i++) {
70bbb3e0 4000 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 4001 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 4002 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 4003 ext4_msg(sb, KERN_ERR,
4004 "can't read group descriptor %d", i);
ac27a0ec
DK		 4005 db_count = i;
4006 goto failed_mount2;
4007 }
4008 }
829fa70d 4009 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 4010 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 4011 ret = -EFSCORRUPTED;
f9ae9cf5 4012 goto failed_mount2;
ac27a0ec 4013 }
772cb7c8 4014
f9ae9cf5 4015 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 4016 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
4017 spin_lock_init(&sbi->s_next_gen_lock);
4018
04ecddb7
JM		 4019 setup_timer(&sbi->s_err_report, print_daily_error_info,
4020 (unsigned long) sb);
04496411 4021
a75ae78f 4022 /* Register extent status tree shrinker */
eb68d0e2 4023 if (ext4_es_register_shrinker(sbi))
ce7e010a 4024 goto failed_mount3;
ce7e010a 4025
c9de560d 4026 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 4027 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 4028
f9ae9cf5
TT		 4029 /*
4030 * set up enough so that it can read an inode
4031 */
f6e63f90 4032 sb->s_op = &ext4_sops;
617ba13b
MC		 4033 sb->s_export_op = &ext4_export_ops;
4034 sb->s_xattr = ext4_xattr_handlers;
a7550b30 4035 sb->s_cop = &ext4_cryptops;
ac27a0ec 4036#ifdef CONFIG_QUOTA
617ba13b 4037 sb->dq_op = &ext4_quota_operations;
e2b911c5 4038 if (ext4_has_feature_quota(sb))
1fa5efe3 4039 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 4040 else
4041 sb->s_qcop = &ext4_qctl_operations;
689c958c 4042 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 4043#endif
85787090 4044 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 4045
ac27a0ec 4046 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4047 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 4048
4049 sb->s_root = NULL;
4050
4051 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 4052 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 4053
e2b911c5 4054 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
c5e06d10 4055 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4056 goto failed_mount3a;
c5e06d10 4057
ac27a0ec
DK		 4058 /*
4059 * The first inode we look at is the journal inode. Don't try
4060 * root first: it may be modified in the journal!
4061 */
e2b911c5 4062 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 4063 err = ext4_load_journal(sb, es, journal_devnum);
4064 if (err)
50460fe8 4065 goto failed_mount3a;
0390131b 4066 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
e2b911c5 4067 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4068 ext4_msg(sb, KERN_ERR, "required journal recovery "
4069 "suppressed and not mounted read-only");
744692dc 4070 goto failed_mount_wq;
ac27a0ec 4071 } else {
1e381f60
DM		 4072 /* Nojournal mode, all journal mount options are illegal */
4073 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4074 ext4_msg(sb, KERN_ERR, "can't mount with "
4075 "journal_checksum, fs mounted w/o journal");
4076 goto failed_mount_wq;
4077 }
4078 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4079 ext4_msg(sb, KERN_ERR, "can't mount with "
4080 "journal_async_commit, fs mounted w/o journal");
4081 goto failed_mount_wq;
4082 }
4083 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4084 ext4_msg(sb, KERN_ERR, "can't mount with "
4085 "commit=%lu, fs mounted w/o journal",
4086 sbi->s_commit_interval / HZ);
4087 goto failed_mount_wq;
4088 }
4089 if (EXT4_MOUNT_DATA_FLAGS &
4090 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4091 ext4_msg(sb, KERN_ERR, "can't mount with "
4092 "data=, fs mounted w/o journal");
4093 goto failed_mount_wq;
4094 }
4095 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4096 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4097 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4098 sbi->s_journal = NULL;
4099 needs_recovery = 0;
4100 goto no_journal;
ac27a0ec
DK		 4101 }
4102
e2b911c5 4103 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4104 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4105 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4106 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4107 goto failed_mount_wq;
eb40a09c
JS		 4108 }
4109
25ed6e8a
DW		 4110 if (!set_journal_csum_feature_set(sb)) {
4111 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4112 "feature set");
4113 goto failed_mount_wq;
d4da6c9c 4114 }
818d276c 4115
ac27a0ec
DK		 4116 /* We have now updated the journal if required, so we can
4117 * validate the data journaling mode. */
4118 switch (test_opt(sb, DATA_FLAGS)) {
4119 case 0:
4120 /* No mode set, assume a default based on the journal
63f57933
AM		 4121 * capabilities: ORDERED_DATA if the journal can
4122 * cope, else JOURNAL_DATA
4123 */
dab291af
MC		 4124 if (jbd2_journal_check_available_features
4125 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 4126 set_opt(sb, ORDERED_DATA);
ac27a0ec 4127 else
fd8c37ec 4128 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 4129 break;
4130
617ba13b
MC		 4131 case EXT4_MOUNT_ORDERED_DATA:
4132 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4133 if (!jbd2_journal_check_available_features
4134 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4135 ext4_msg(sb, KERN_ERR, "Journal does not support "
4136 "requested data journaling mode");
744692dc 4137 goto failed_mount_wq;
ac27a0ec
DK		 4138 }
4139 default:
4140 break;
4141 }
ab04df78
JK		 4142
4143 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4144 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4145 ext4_msg(sb, KERN_ERR, "can't mount with "
4146 "journal_async_commit in data=ordered mode");
4147 goto failed_mount_wq;
4148 }
4149
b3881f74 4150 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4151
18aadd47
BJ		 4152 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4153
ce7e010a 4154no_journal:
cdb7ee4c
TE		 4155 if (!test_opt(sb, NO_MBCACHE)) {
4156 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4157 if (!sbi->s_ea_block_cache) {
dec214d0 4158 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4159 "Failed to create ea_block_cache");
dec214d0
TE		 4160 goto failed_mount_wq;
4161 }
cdb7ee4c
TE		 4162
4163 if (ext4_has_feature_ea_inode(sb)) {
4164 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4165 if (!sbi->s_ea_inode_cache) {
4166 ext4_msg(sb, KERN_ERR,
4167 "Failed to create ea_inode_cache");
4168 goto failed_mount_wq;
4169 }
4170 }
9c191f70
M		 4171 }
4172
e2b911c5 4173 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 4174 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 4175 ext4_msg(sb, KERN_ERR,
4176 "Unsupported blocksize for fs encryption");
4177 goto failed_mount_wq;
4178 }
4179
e2b911c5
DW		 4180 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4181 !ext4_has_feature_encrypt(sb)) {
4182 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4183 ext4_commit_super(sb, 1);
4184 }
4185
952fc18e
TT		 4186 /*
4187 * Get the # of file system overhead blocks from the
4188 * superblock if present.
4189 */
4190 if (es->s_overhead_clusters)
4191 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4192 else {
07aa2ea1
LC		 4193 err = ext4_calculate_overhead(sb);
4194 if (err)
952fc18e
TT		 4195 goto failed_mount_wq;
4196 }
4197
fd89d5f2
TH		 4198 /*
4199 * The maximum number of concurrent works can be high and
4200 * concurrency isn't really necessary. Limit it to 1.
4201 */
2e8fa54e
JK		 4202 EXT4_SB(sb)->rsv_conversion_wq =
4203 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4204 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4205 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4206 ret = -ENOMEM;
2e8fa54e
JK		 4207 goto failed_mount4;
4208 }
4209
ac27a0ec 4210 /*
dab291af 4211 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4212 * so we can safely mount the rest of the filesystem now.
4213 */
4214
1d1fe1ee
DH		 4215 root = ext4_iget(sb, EXT4_ROOT_INO);
4216 if (IS_ERR(root)) {
b31e1552 4217 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4218 ret = PTR_ERR(root);
32a9bb57 4219 root = NULL;
ac27a0ec
DK		 4220 goto failed_mount4;
4221 }
4222 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4223 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4224 iput(root);
ac27a0ec
DK		 4225 goto failed_mount4;
4226 }
48fde701 4227 sb->s_root = d_make_root(root);
1d1fe1ee 4228 if (!sb->s_root) {
b31e1552 4229 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4230 ret = -ENOMEM;
4231 goto failed_mount4;
4232 }
ac27a0ec 4233
7e84b621
ES		 4234 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4235 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4236
4237 /* determine the minimum size of new large inodes, if present */
670e9875
TT		 4238 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4239 sbi->s_want_extra_isize == 0) {
ef7f3835
KS		 4240 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4241 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4242 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4243 if (sbi->s_want_extra_isize <
4244 le16_to_cpu(es->s_want_extra_isize))
4245 sbi->s_want_extra_isize =
4246 le16_to_cpu(es->s_want_extra_isize);
4247 if (sbi->s_want_extra_isize <
4248 le16_to_cpu(es->s_min_extra_isize))
4249 sbi->s_want_extra_isize =
4250 le16_to_cpu(es->s_min_extra_isize);
4251 }
4252 }
4253 /* Check if enough inode space is available */
4254 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4255 sbi->s_inode_size) {
4256 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4257 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4258 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4259 "available");
ef7f3835
KS		 4260 }
4261
b5799018 4262 ext4_set_resv_clusters(sb);
27dd4385 4263
6fd058f7
TT		 4264 err = ext4_setup_system_zone(sb);
4265 if (err) {
b31e1552 4266 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4267 "zone (%d)", err);
f9ae9cf5
TT		 4268 goto failed_mount4a;
4269 }
4270
4271 ext4_ext_init(sb);
4272 err = ext4_mb_init(sb);
4273 if (err) {
4274 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4275 err);
dcf2d804 4276 goto failed_mount5;
c2774d84
AK		 4277 }
4278
d5e03cbb
TT		 4279 block = ext4_count_free_clusters(sb);
4280 ext4_free_blocks_count_set(sbi->s_es,
4281 EXT4_C2B(sbi, block));
908c7f19
TH		 4282 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4283 GFP_KERNEL);
d5e03cbb
TT		 4284 if (!err) {
4285 unsigned long freei = ext4_count_free_inodes(sb);
4286 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4287 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4288 GFP_KERNEL);
d5e03cbb
TT		 4289 }
4290 if (!err)
4291 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4292 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4293 if (!err)
908c7f19
TH		 4294 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4295 GFP_KERNEL);
c8585c6f
DJ		 4296 if (!err)
4297 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4298
d5e03cbb
TT		 4299 if (err) {
4300 ext4_msg(sb, KERN_ERR, "insufficient memory");
4301 goto failed_mount6;
4302 }
4303
e2b911c5 4304 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4305 if (!ext4_fill_flex_info(sb)) {
4306 ext4_msg(sb, KERN_ERR,
4307 "unable to initialize "
4308 "flex_bg meta info!");
4309 goto failed_mount6;
4310 }
4311
bfff6873
LC		 4312 err = ext4_register_li_request(sb, first_not_zeroed);
4313 if (err)
dcf2d804 4314 goto failed_mount6;
bfff6873 4315
b5799018 4316 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4317 if (err)
4318 goto failed_mount7;
3197ebdb 4319
9b2ff357
JK		 4320#ifdef CONFIG_QUOTA
4321 /* Enable quota usage during mount. */
e2b911c5 4322 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
9b2ff357
JK		 4323 err = ext4_enable_quotas(sb);
4324 if (err)
4325 goto failed_mount8;
4326 }
4327#endif /* CONFIG_QUOTA */
4328
617ba13b
MC		 4329 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4330 ext4_orphan_cleanup(sb, es);
4331 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4332 if (needs_recovery) {
b31e1552 4333 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4334 ext4_mark_recovery_complete(sb, es);
4335 }
4336 if (EXT4_SB(sb)->s_journal) {
4337 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4338 descr = " journalled data mode";
4339 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4340 descr = " ordered data mode";
4341 else
4342 descr = " writeback data mode";
4343 } else
4344 descr = "out journal";
4345
79add3a3
LC		 4346 if (test_opt(sb, DISCARD)) {
4347 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4348 if (!blk_queue_discard(q))
4349 ext4_msg(sb, KERN_WARNING,
4350 "mounting with \"discard\" option, but "
4351 "the device does not support discard");
4352 }
4353
e294a537
TT		 4354 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4355 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4356 "Opts: %.*s%s%s", descr,
4357 (int) sizeof(sbi->s_es->s_mount_opts),
4358 sbi->s_es->s_mount_opts,
e294a537 4359 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4360
66e61a9e
TT		 4361 if (es->s_error_count)
4362 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4363
efbed4dc
TT		 4364 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4365 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4366 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4367 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4368
d4c402d9 4369 kfree(orig_data);
ac27a0ec
DK		 4370 return 0;
4371
617ba13b 4372cantfind_ext4:
ac27a0ec 4373 if (!silent)
b31e1552 4374 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4375 goto failed_mount;
4376
72ba7450
TT		 4377#ifdef CONFIG_QUOTA
4378failed_mount8:
ebd173be 4379 ext4_unregister_sysfs(sb);
72ba7450 4380#endif
dcf2d804
TM		 4381failed_mount7:
4382 ext4_unregister_li_request(sb);
4383failed_mount6:
f9ae9cf5 4384 ext4_mb_release(sb);
d5e03cbb 4385 if (sbi->s_flex_groups)
b93b41d4 4386 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4387 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4388 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4389 percpu_counter_destroy(&sbi->s_dirs_counter);
4390 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4391failed_mount5:
f9ae9cf5
TT		 4392 ext4_ext_release(sb);
4393 ext4_release_system_zone(sb);
4394failed_mount4a:
94bf608a 4395 dput(sb->s_root);
32a9bb57 4396 sb->s_root = NULL;
94bf608a 4397failed_mount4:
b31e1552 4398 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4399 if (EXT4_SB(sb)->rsv_conversion_wq)
4400 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4401failed_mount_wq:
dec214d0
TE		 4402 if (sbi->s_ea_inode_cache) {
4403 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4404 sbi->s_ea_inode_cache = NULL;
4405 }
47387409
TE		 4406 if (sbi->s_ea_block_cache) {
4407 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4408 sbi->s_ea_block_cache = NULL;
82939d79 4409 }
0390131b
FM		 4410 if (sbi->s_journal) {
4411 jbd2_journal_destroy(sbi->s_journal);
4412 sbi->s_journal = NULL;
4413 }
50460fe8 4414failed_mount3a:
d3922a77 4415 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4416failed_mount3:
9105bb14 4417 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4418 if (sbi->s_mmp_tsk)
4419 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4420failed_mount2:
4421 for (i = 0; i < db_count; i++)
4422 brelse(sbi->s_group_desc[i]);
b93b41d4 4423 kvfree(sbi->s_group_desc);
ac27a0ec 4424failed_mount:
0441984a
DW		 4425 if (sbi->s_chksum_driver)
4426 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4427#ifdef CONFIG_QUOTA
a2d4a646 4428 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4429 kfree(sbi->s_qf_names[i]);
4430#endif
617ba13b 4431 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4432 brelse(bh);
4433out_fail:
afdbd4a6 4434 /* sb->s_user_ns will be put when sb is destroyed */
ac27a0ec 4435 sb->s_fs_info = NULL;
f6830165 4436 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4437out_free_base:
ac27a0ec 4438 kfree(sbi);
d4c402d9 4439 kfree(orig_data);
07aa2ea1 4440 return err ? err : ret;
ac27a0ec
DK		 4441}
4442
4443/*
4444 * Setup any per-fs journal parameters now. We'll do this both on
4445 * initial mount, once the journal has been initialised but before we've
4446 * done any recovery; and again on any subsequent remount.
4447 */
617ba13b 4448static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4449{
617ba13b 4450 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4451
30773840
TT		 4452 journal->j_commit_interval = sbi->s_commit_interval;
4453 journal->j_min_batch_time = sbi->s_min_batch_time;
4454 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4455
a931da6a 4456 write_lock(&journal->j_state_lock);
ac27a0ec 4457 if (test_opt(sb, BARRIER))
dab291af 4458 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4459 else
dab291af 4460 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4461 if (test_opt(sb, DATA_ERR_ABORT))
4462 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4463 else
4464 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4465 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4466}
4467
c6cb7e77
EW		 4468static struct inode *ext4_get_journal_inode(struct super_block *sb,
4469 unsigned int journal_inum)
ac27a0ec
DK		 4470{
4471 struct inode *journal_inode;
ac27a0ec 4472
c6cb7e77
EW		 4473 /*
4474 * Test for the existence of a valid inode on disk. Bad things
4475 * happen if we iget() an unused inode, as the subsequent iput()
4476 * will try to delete it.
4477 */
1d1fe1ee
DH		 4478 journal_inode = ext4_iget(sb, journal_inum);
4479 if (IS_ERR(journal_inode)) {
b31e1552 4480 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4481 return NULL;
4482 }
4483 if (!journal_inode->i_nlink) {
4484 make_bad_inode(journal_inode);
4485 iput(journal_inode);
b31e1552 4486 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4487 return NULL;
4488 }
4489
e5f8eab8 4490 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4491 journal_inode, journal_inode->i_size);
1d1fe1ee 4492 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4493 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4494 iput(journal_inode);
4495 return NULL;
4496 }
c6cb7e77
EW		 4497 return journal_inode;
4498}
4499
4500static journal_t *ext4_get_journal(struct super_block *sb,
4501 unsigned int journal_inum)
4502{
4503 struct inode *journal_inode;
4504 journal_t *journal;
4505
4506 BUG_ON(!ext4_has_feature_journal(sb));
4507
4508 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4509 if (!journal_inode)
4510 return NULL;
ac27a0ec 4511
dab291af 4512 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4513 if (!journal) {
b31e1552 4514 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4515 iput(journal_inode);
4516 return NULL;
4517 }
4518 journal->j_private = sb;
617ba13b 4519 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4520 return journal;
4521}
4522
617ba13b 4523static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4524 dev_t j_dev)
4525{
2b2d6d01 4526 struct buffer_head *bh;
ac27a0ec 4527 journal_t *journal;
617ba13b
MC		 4528 ext4_fsblk_t start;
4529 ext4_fsblk_t len;
ac27a0ec 4530 int hblock, blocksize;
617ba13b 4531 ext4_fsblk_t sb_block;
ac27a0ec 4532 unsigned long offset;
2b2d6d01 4533 struct ext4_super_block *es;
ac27a0ec
DK		 4534 struct block_device *bdev;
4535
e2b911c5 4536 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4537
b31e1552 4538 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4539 if (bdev == NULL)
4540 return NULL;
4541
ac27a0ec 4542 blocksize = sb->s_blocksize;
e1defc4f 4543 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4544 if (blocksize < hblock) {
b31e1552
ES		 4545 ext4_msg(sb, KERN_ERR,
4546 "blocksize too small for journal device");
ac27a0ec
DK		 4547 goto out_bdev;
4548 }
4549
617ba13b
MC		 4550 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4551 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4552 set_blocksize(bdev, blocksize);
4553 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4554 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4555 "external journal");
ac27a0ec
DK		 4556 goto out_bdev;
4557 }
4558
2716b802 4559 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4560 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4561 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4562 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4563 ext4_msg(sb, KERN_ERR, "external journal has "
4564 "bad superblock");
ac27a0ec
DK		 4565 brelse(bh);
4566 goto out_bdev;
4567 }
4568
df4763be
DW		 4569 if ((le32_to_cpu(es->s_feature_ro_compat) &
4570 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4571 es->s_checksum != ext4_superblock_csum(sb, es)) {
4572 ext4_msg(sb, KERN_ERR, "external journal has "
4573 "corrupt superblock");
4574 brelse(bh);
4575 goto out_bdev;
4576 }
4577
617ba13b 4578 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4579 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4580 brelse(bh);
4581 goto out_bdev;
4582 }
4583
bd81d8ee 4584 len = ext4_blocks_count(es);
ac27a0ec
DK		 4585 start = sb_block + 1;
4586 brelse(bh); /* we're done with the superblock */
4587
dab291af 4588 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4589 start, len, blocksize);
4590 if (!journal) {
b31e1552 4591 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4592 goto out_bdev;
4593 }
4594 journal->j_private = sb;
dfec8a14 4595 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4596 wait_on_buffer(journal->j_sb_buffer);
4597 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4598 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4599 goto out_journal;
4600 }
4601 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4602 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4603 "user (unsupported) - %d",
ac27a0ec
DK		 4604 be32_to_cpu(journal->j_superblock->s_nr_users));
4605 goto out_journal;
4606 }
617ba13b
MC		 4607 EXT4_SB(sb)->journal_bdev = bdev;
4608 ext4_init_journal_params(sb, journal);
ac27a0ec 4609 return journal;
0b8e58a1 4610
ac27a0ec 4611out_journal:
dab291af 4612 jbd2_journal_destroy(journal);
ac27a0ec 4613out_bdev:
617ba13b 4614 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4615 return NULL;
4616}
4617
617ba13b
MC		 4618static int ext4_load_journal(struct super_block *sb,
4619 struct ext4_super_block *es,
ac27a0ec
DK		 4620 unsigned long journal_devnum)
4621{
4622 journal_t *journal;
4623 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4624 dev_t journal_dev;
4625 int err = 0;
4626 int really_read_only;
4627
e2b911c5 4628 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4629
ac27a0ec
DK		 4630 if (journal_devnum &&
4631 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4632 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4633 "numbers have changed");
ac27a0ec
DK		 4634 journal_dev = new_decode_dev(journal_devnum);
4635 } else
4636 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4637
4638 really_read_only = bdev_read_only(sb->s_bdev);
4639
4640 /*
4641 * Are we loading a blank journal or performing recovery after a
4642 * crash? For recovery, we need to check in advance whether we
4643 * can get read-write access to the device.
4644 */
e2b911c5 4645 if (ext4_has_feature_journal_needs_recovery(sb)) {
ac27a0ec 4646 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4647 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4648 "required on readonly filesystem");
ac27a0ec 4649 if (really_read_only) {
b31e1552
ES		 4650 ext4_msg(sb, KERN_ERR, "write access "
4651 "unavailable, cannot proceed");
ac27a0ec
DK		 4652 return -EROFS;
4653 }
b31e1552
ES		 4654 ext4_msg(sb, KERN_INFO, "write access will "
4655 "be enabled during recovery");
ac27a0ec
DK		 4656 }
4657 }
4658
4659 if (journal_inum && journal_dev) {
b31e1552
ES		 4660 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4661 "and inode journals!");
ac27a0ec
DK		 4662 return -EINVAL;
4663 }
4664
4665 if (journal_inum) {
617ba13b 4666 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4667 return -EINVAL;
4668 } else {
617ba13b 4669 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4670 return -EINVAL;
4671 }
4672
90576c0b 4673 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4674 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4675
e2b911c5 4676 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4677 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4678 if (!err) {
4679 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4680 if (save)
4681 memcpy(save, ((char *) es) +
4682 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4683 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4684 if (save)
4685 memcpy(((char *) es) + EXT4_S_ERR_START,
4686 save, EXT4_S_ERR_LEN);
4687 kfree(save);
4688 }
ac27a0ec
DK		 4689
4690 if (err) {
b31e1552 4691 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4692 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4693 return err;
4694 }
4695
617ba13b
MC		 4696 EXT4_SB(sb)->s_journal = journal;
4697 ext4_clear_journal_err(sb, es);
ac27a0ec 4698
c41303ce 4699 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4700 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4701 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4702
4703 /* Make sure we flush the recovery flag to disk. */
e2d67052 4704 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4705 }
4706
4707 return 0;
4708}
4709
e2d67052 4710static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4711{
e2d67052 4712 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4713 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4714 int error = 0;
ac27a0ec 4715
bdfe0cbd 4716 if (!sbh || block_device_ejected(sb))
c4be0c1d 4717 return error;
71290b36
TT		 4718 /*
4719 * If the file system is mounted read-only, don't update the
4720 * superblock write time. This avoids updating the superblock
4721 * write time when we are mounting the root file system
4722 * read/only but we need to replay the journal; at that point,
4723 * for people who are east of GMT and who make their clock
4724 * tick in localtime for Windows bug-for-bug compatibility,
4725 * the clock is set in the future, and this will cause e2fsck
4726 * to complain and force a full file system check.
4727 */
4728 if (!(sb->s_flags & MS_RDONLY))
4729 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4730 if (sb->s_bdev->bd_part)
4731 es->s_kbytes_written =
4732 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4733 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4734 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4735 else
4736 es->s_kbytes_written =
4737 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4738 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4739 ext4_free_blocks_count_set(es,
57042651
TT		 4740 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4741 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4742 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4743 es->s_free_inodes_count =
4744 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4745 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4746 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4747 ext4_superblock_csum_set(sb);
1566a48a
TT		 4748 if (sync)
4749 lock_buffer(sbh);
4743f839
PKS		 4750 if (buffer_write_io_error(sbh)) {
4751 /*
4752 * Oh, dear. A previous attempt to write the
4753 * superblock failed. This could happen because the
4754 * USB device was yanked out. Or it could happen to
4755 * be a transient write error and maybe the block will
4756 * be remapped. Nothing we can do but to retry the
4757 * write and hope for the best.
4758 */
4759 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4760 "superblock detected");
4761 clear_buffer_write_io_error(sbh);
4762 set_buffer_uptodate(sbh);
4763 }
ac27a0ec 4764 mark_buffer_dirty(sbh);
914258bf 4765 if (sync) {
1566a48a 4766 unlock_buffer(sbh);
564bc402 4767 error = __sync_dirty_buffer(sbh,
00473374 4768 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
c4be0c1d
TS		 4769 if (error)
4770 return error;
4771
4772 error = buffer_write_io_error(sbh);
4773 if (error) {
b31e1552
ES		 4774 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4775 "superblock");
914258bf
TT		 4776 clear_buffer_write_io_error(sbh);
4777 set_buffer_uptodate(sbh);
4778 }
4779 }
c4be0c1d 4780 return error;
ac27a0ec
DK		 4781}
4782
ac27a0ec
DK		 4783/*
4784 * Have we just finished recovery? If so, and if we are mounting (or
4785 * remounting) the filesystem readonly, then we will end up with a
4786 * consistent fs on disk. Record that fact.
4787 */
2b2d6d01
TT		 4788static void ext4_mark_recovery_complete(struct super_block *sb,
4789 struct ext4_super_block *es)
ac27a0ec 4790{
617ba13b 4791 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4792
e2b911c5 4793 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4794 BUG_ON(journal != NULL);
4795 return;
4796 }
dab291af 4797 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4798 if (jbd2_journal_flush(journal) < 0)
4799 goto out;
4800
e2b911c5 4801 if (ext4_has_feature_journal_needs_recovery(sb) &&
ac27a0ec 4802 sb->s_flags & MS_RDONLY) {
e2b911c5 4803 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4804 ext4_commit_super(sb, 1);
ac27a0ec 4805 }
7ffe1ea8
HK		 4806
4807out:
dab291af 4808 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4809}
4810
4811/*
4812 * If we are mounting (or read-write remounting) a filesystem whose journal
4813 * has recorded an error from a previous lifetime, move that error to the
4814 * main filesystem now.
4815 */
2b2d6d01
TT		 4816static void ext4_clear_journal_err(struct super_block *sb,
4817 struct ext4_super_block *es)
ac27a0ec
DK		 4818{
4819 journal_t *journal;
4820 int j_errno;
4821 const char *errstr;
4822
e2b911c5 4823 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4824
617ba13b 4825 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4826
4827 /*
4828 * Now check for any error status which may have been recorded in the
617ba13b 4829 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4830 */
4831
dab291af 4832 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4833 if (j_errno) {
4834 char nbuf[16];
4835
617ba13b 4836 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4837 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4838 "from previous mount: %s", errstr);
12062ddd 4839 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4840
617ba13b
MC		 4841 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4842 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4843 ext4_commit_super(sb, 1);
ac27a0ec 4844
dab291af 4845 jbd2_journal_clear_err(journal);
d796c52e 4846 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4847 }
4848}
4849
4850/*
4851 * Force the running and committing transactions to commit,
4852 * and wait on the commit.
4853 */
617ba13b 4854int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4855{
4856 journal_t *journal;
ac27a0ec
DK		 4857
4858 if (sb->s_flags & MS_RDONLY)
4859 return 0;
4860
617ba13b 4861 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4862 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4863}
4864
617ba13b 4865static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4866{
14ce0cb4 4867 int ret = 0;
9eddacf9 4868 tid_t target;
06a407f1 4869 bool needs_barrier = false;
8d5d02e6 4870 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4871
0db1ff22
TT		 4872 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4873 return 0;
4874
9bffad1e 4875 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4876 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4877 /*
4878 * Writeback quota in non-journalled quota case - journalled quota has
4879 * no dirty dquots
4880 */
4881 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4882 /*
4883 * Data writeback is possible w/o journal transaction, so barrier must
4884 * being sent at the end of the function. But we can skip it if
4885 * transaction_commit will do it for us.
4886 */
bda32530
TT		 4887 if (sbi->s_journal) {
4888 target = jbd2_get_latest_transaction(sbi->s_journal);
4889 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4890 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4891 needs_barrier = true;
4892
4893 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4894 if (wait)
4895 ret = jbd2_log_wait_commit(sbi->s_journal,
4896 target);
4897 }
4898 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4899 needs_barrier = true;
06a407f1
DM		 4900 if (needs_barrier) {
4901 int err;
4902 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4903 if (!ret)
4904 ret = err;
0390131b 4905 }
06a407f1
DM		 4906
4907 return ret;
4908}
4909
ac27a0ec
DK		 4910/*
4911 * LVM calls this function before a (read-only) snapshot is created. This
4912 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4913 *
4914 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4915 * state independently. It relies on upper layer to stop all data & metadata
4916 * modifications.
ac27a0ec 4917 */
c4be0c1d 4918static int ext4_freeze(struct super_block *sb)
ac27a0ec 4919{
c4be0c1d
TS		 4920 int error = 0;
4921 journal_t *journal;
ac27a0ec 4922
9ca92389
TT		 4923 if (sb->s_flags & MS_RDONLY)
4924 return 0;
ac27a0ec 4925
9ca92389 4926 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4927
bb044576
TT		 4928 if (journal) {
4929 /* Now we set up the journal barrier. */
4930 jbd2_journal_lock_updates(journal);
ac27a0ec 4931
bb044576
TT		 4932 /*
4933 * Don't clear the needs_recovery flag if we failed to
4934 * flush the journal.
4935 */
4936 error = jbd2_journal_flush(journal);
4937 if (error < 0)
4938 goto out;
c642dc9e
ES		 4939
4940 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4941 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4942 }
9ca92389 4943
9ca92389 4944 error = ext4_commit_super(sb, 1);
6b0310fb 4945out:
bb044576
TT		 4946 if (journal)
4947 /* we rely on upper layer to stop further updates */
4948 jbd2_journal_unlock_updates(journal);
6b0310fb 4949 return error;
ac27a0ec
DK		 4950}
4951
4952/*
4953 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4954 * flag here, even though the filesystem is not technically dirty yet.
4955 */
c4be0c1d 4956static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4957{
783d9485 4958 if ((sb->s_flags & MS_RDONLY) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 4959 return 0;
4960
c642dc9e
ES		 4961 if (EXT4_SB(sb)->s_journal) {
4962 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4963 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4964 }
4965
9ca92389 4966 ext4_commit_super(sb, 1);
c4be0c1d 4967 return 0;
ac27a0ec
DK		 4968}
4969
673c6100
TT		 4970/*
4971 * Structure to save mount options for ext4_remount's benefit
4972 */
4973struct ext4_mount_options {
4974 unsigned long s_mount_opt;
a2595b8a 4975 unsigned long s_mount_opt2;
08cefc7a
EB		 4976 kuid_t s_resuid;
4977 kgid_t s_resgid;
673c6100
TT		 4978 unsigned long s_commit_interval;
4979 u32 s_min_batch_time, s_max_batch_time;
4980#ifdef CONFIG_QUOTA
4981 int s_jquota_fmt;
a2d4a646 4982 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4983#endif
4984};
4985
2b2d6d01 4986static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4987{
2b2d6d01 4988 struct ext4_super_block *es;
617ba13b 4989 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4990 unsigned long old_sb_flags;
617ba13b 4991 struct ext4_mount_options old_opts;
c79d967d 4992 int enable_quota = 0;
8a266467 4993 ext4_group_t g;
b3881f74 4994 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4995 int err = 0;
ac27a0ec 4996#ifdef CONFIG_QUOTA
03dafb5f 4997 int i, j;
ac27a0ec 4998#endif
d4c402d9 4999 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 5000
5001 /* Store the original options */
5002 old_sb_flags = sb->s_flags;
5003 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 5004 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 5005 old_opts.s_resuid = sbi->s_resuid;
5006 old_opts.s_resgid = sbi->s_resgid;
5007 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 5008 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5009 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 5010#ifdef CONFIG_QUOTA
5011 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 5012 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 5013 if (sbi->s_qf_names[i]) {
5014 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
5015 GFP_KERNEL);
5016 if (!old_opts.s_qf_names[i]) {
5017 for (j = 0; j < i; j++)
5018 kfree(old_opts.s_qf_names[j]);
3e36a163 5019 kfree(orig_data);
03dafb5f
CG		 5020 return -ENOMEM;
5021 }
5022 } else
5023 old_opts.s_qf_names[i] = NULL;
ac27a0ec 5024#endif
b3881f74
TT		 5025 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5026 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 5027
661aa520 5028 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 5029 err = -EINVAL;
5030 goto restore_opts;
5031 }
5032
6b992ff2 5033 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 5034 test_opt(sb, JOURNAL_CHECKSUM)) {
5035 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 5036 "during remount not supported; ignoring");
5037 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 5038 }
5039
6ae6514b
PS		 5040 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5041 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5042 ext4_msg(sb, KERN_ERR, "can't mount with "
5043 "both data=journal and delalloc");
5044 err = -EINVAL;
5045 goto restore_opts;
5046 }
5047 if (test_opt(sb, DIOREAD_NOLOCK)) {
5048 ext4_msg(sb, KERN_ERR, "can't mount with "
5049 "both data=journal and dioread_nolock");
5050 err = -EINVAL;
5051 goto restore_opts;
5052 }
923ae0ff
RZ		 5053 if (test_opt(sb, DAX)) {
5054 ext4_msg(sb, KERN_ERR, "can't mount with "
5055 "both data=journal and dax");
5056 err = -EINVAL;
5057 goto restore_opts;
5058 }
ab04df78
JK		 5059 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5060 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5061 ext4_msg(sb, KERN_ERR, "can't mount with "
5062 "journal_async_commit in data=ordered mode");
5063 err = -EINVAL;
5064 goto restore_opts;
5065 }
923ae0ff
RZ		 5066 }
5067
cdb7ee4c
TE		 5068 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5069 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5070 err = -EINVAL;
5071 goto restore_opts;
5072 }
5073
923ae0ff
RZ		 5074 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5075 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5076 "dax flag with busy inodes while remounting");
5077 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 5078 }
5079
4ab2f15b 5080 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 5081 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 5082
5083 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 5084 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 5085
5086 es = sbi->s_es;
5087
b3881f74 5088 if (sbi->s_journal) {
0390131b 5089 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5090 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5091 }
ac27a0ec 5092
a2fd66d0
TT		 5093 if (*flags & MS_LAZYTIME)
5094 sb->s_flags |= MS_LAZYTIME;
5095
661aa520 5096 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 5097 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 5098 err = -EROFS;
5099 goto restore_opts;
5100 }
5101
5102 if (*flags & MS_RDONLY) {
38c03b34
TT		 5103 err = sync_filesystem(sb);
5104 if (err < 0)
5105 goto restore_opts;
0f0dd62f
CH		 5106 err = dquot_suspend(sb, -1);
5107 if (err < 0)
c79d967d 5108 goto restore_opts;
c79d967d 5109
ac27a0ec
DK		 5110 /*
5111 * First of all, the unconditional stuff we have to do
5112 * to disable replay of the journal when we next remount
5113 */
5114 sb->s_flags |= MS_RDONLY;
5115
5116 /*
5117 * OK, test if we are remounting a valid rw partition
5118 * readonly, and if so set the rdonly flag and then
5119 * mark the partition as valid again.
5120 */
617ba13b
MC		 5121 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5122 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5123 es->s_state = cpu_to_le16(sbi->s_mount_state);
5124
a63c9eb2 5125 if (sbi->s_journal)
0390131b 5126 ext4_mark_recovery_complete(sb, es);
ac27a0ec 5127 } else {
a13fb1a4 5128 /* Make sure we can mount this feature set readwrite */
e2b911c5 5129 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5130 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5131 err = -EROFS;
5132 goto restore_opts;
5133 }
8a266467
TT		 5134 /*
5135 * Make sure the group descriptor checksums
0b8e58a1 5136 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5137 */
5138 for (g = 0; g < sbi->s_groups_count; g++) {
5139 struct ext4_group_desc *gdp =
5140 ext4_get_group_desc(sb, g, NULL);
5141
feb0ab32 5142 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5143 ext4_msg(sb, KERN_ERR,
5144 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5145 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5146 le16_to_cpu(gdp->bg_checksum));
6a797d27 5147 err = -EFSBADCRC;
8a266467
TT		 5148 goto restore_opts;
5149 }
5150 }
5151
ead6596b
ES		 5152 /*
5153 * If we have an unprocessed orphan list hanging
5154 * around from a previously readonly bdev mount,
5155 * require a full umount/remount for now.
5156 */
5157 if (es->s_last_orphan) {
b31e1552 5158 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5159 "remount RDWR because of unprocessed "
5160 "orphan inode list. Please "
b31e1552 5161 "umount/remount instead");
ead6596b
ES		 5162 err = -EINVAL;
5163 goto restore_opts;
5164 }
5165
ac27a0ec
DK		 5166 /*
5167 * Mounting a RDONLY partition read-write, so reread
5168 * and store the current valid flag. (It may have
5169 * been changed by e2fsck since we originally mounted
5170 * the partition.)
5171 */
0390131b
FM		 5172 if (sbi->s_journal)
5173 ext4_clear_journal_err(sb, es);
ac27a0ec 5174 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 5175 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 5176 sb->s_flags &= ~MS_RDONLY;
e2b911c5 5177 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5178 if (ext4_multi_mount_protect(sb,
5179 le64_to_cpu(es->s_mmp_block))) {
5180 err = -EROFS;
5181 goto restore_opts;
5182 }
c79d967d 5183 enable_quota = 1;
ac27a0ec
DK		 5184 }
5185 }
bfff6873
LC		 5186
5187 /*
5188 * Reinitialize lazy itable initialization thread based on
5189 * current settings
5190 */
5191 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5192 ext4_unregister_li_request(sb);
5193 else {
5194 ext4_group_t first_not_zeroed;
5195 first_not_zeroed = ext4_has_uninit_itable(sb);
5196 ext4_register_li_request(sb, first_not_zeroed);
5197 }
5198
6fd058f7 5199 ext4_setup_system_zone(sb);
d096ad0f 5200 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5201 ext4_commit_super(sb, 1);
0390131b 5202
ac27a0ec
DK		 5203#ifdef CONFIG_QUOTA
5204 /* Release old quota file names */
a2d4a646 5205 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5206 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5207 if (enable_quota) {
5208 if (sb_any_quota_suspended(sb))
5209 dquot_resume(sb, -1);
e2b911c5 5210 else if (ext4_has_feature_quota(sb)) {
7c319d32 5211 err = ext4_enable_quotas(sb);
07724f98 5212 if (err)
7c319d32 5213 goto restore_opts;
7c319d32
AK		 5214 }
5215 }
ac27a0ec 5216#endif
d4c402d9 5217
a26f4992 5218 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5219 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5220 kfree(orig_data);
ac27a0ec 5221 return 0;
0b8e58a1 5222
ac27a0ec
DK		 5223restore_opts:
5224 sb->s_flags = old_sb_flags;
5225 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5226 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5227 sbi->s_resuid = old_opts.s_resuid;
5228 sbi->s_resgid = old_opts.s_resgid;
5229 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5230 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5231 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5232#ifdef CONFIG_QUOTA
5233 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5234 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5235 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5236 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5237 }
5238#endif
d4c402d9 5239 kfree(orig_data);
ac27a0ec
DK		 5240 return err;
5241}
5242
689c958c
LX		 5243#ifdef CONFIG_QUOTA
5244static int ext4_statfs_project(struct super_block *sb,
5245 kprojid_t projid, struct kstatfs *buf)
5246{
5247 struct kqid qid;
5248 struct dquot *dquot;
5249 u64 limit;
5250 u64 curblock;
5251
5252 qid = make_kqid_projid(projid);
5253 dquot = dqget(sb, qid);
5254 if (IS_ERR(dquot))
5255 return PTR_ERR(dquot);
5256 spin_lock(&dq_data_lock);
5257
5258 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5259 dquot->dq_dqb.dqb_bsoftlimit :
5260 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5261 if (limit && buf->f_blocks > limit) {
5262 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5263 buf->f_blocks = limit;
5264 buf->f_bfree = buf->f_bavail =
5265 (buf->f_blocks > curblock) ?
5266 (buf->f_blocks - curblock) : 0;
5267 }
5268
5269 limit = dquot->dq_dqb.dqb_isoftlimit ?
5270 dquot->dq_dqb.dqb_isoftlimit :
5271 dquot->dq_dqb.dqb_ihardlimit;
5272 if (limit && buf->f_files > limit) {
5273 buf->f_files = limit;
5274 buf->f_ffree =
5275 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5276 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5277 }
5278
5279 spin_unlock(&dq_data_lock);
5280 dqput(dquot);
5281 return 0;
5282}
5283#endif
5284
2b2d6d01 5285static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5286{
5287 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5288 struct ext4_sb_info *sbi = EXT4_SB(sb);
5289 struct ext4_super_block *es = sbi->s_es;
27dd4385 5290 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5291 u64 fsid;
d02a9391 5292 s64 bfree;
27dd4385 5293 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5294
952fc18e
TT		 5295 if (!test_opt(sb, MINIX_DF))
5296 overhead = sbi->s_overhead;
ac27a0ec 5297
617ba13b 5298 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5299 buf->f_bsize = sb->s_blocksize;
b72f78cb 5300 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5301 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5302 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5303 /* prevent underflow in case that few free space is available */
57042651 5304 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5305 buf->f_bavail = buf->f_bfree -
5306 (ext4_r_blocks_count(es) + resv_blocks);
5307 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5308 buf->f_bavail = 0;
5309 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5310 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5311 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5312 fsid = le64_to_cpup((void *)es->s_uuid) ^
5313 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5314 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5315 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5316
689c958c
LX		 5317#ifdef CONFIG_QUOTA
5318 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5319 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5320 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5321#endif
ac27a0ec
DK		 5322 return 0;
5323}
5324
0b8e58a1
AD		 5325/* Helper function for writing quotas on sync - we need to start transaction
5326 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5327 * Process 1 Process 2
617ba13b 5328 * ext4_create() quota_sync()
a269eb18 5329 * jbd2_journal_start() write_dquot()
871a2931 5330 * dquot_initialize() down(dqio_mutex)
dab291af 5331 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5332 *
5333 */
5334
5335#ifdef CONFIG_QUOTA
5336
5337static inline struct inode *dquot_to_inode(struct dquot *dquot)
5338{
4c376dca 5339 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5340}
5341
617ba13b 5342static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5343{
5344 int ret, err;
5345 handle_t *handle;
5346 struct inode *inode;
5347
5348 inode = dquot_to_inode(dquot);
9924a92a 5349 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5350 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5351 if (IS_ERR(handle))
5352 return PTR_ERR(handle);
5353 ret = dquot_commit(dquot);
617ba13b 5354 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5355 if (!ret)
5356 ret = err;
5357 return ret;
5358}
5359
617ba13b 5360static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5361{
5362 int ret, err;
5363 handle_t *handle;
5364
9924a92a 5365 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5366 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5367 if (IS_ERR(handle))
5368 return PTR_ERR(handle);
5369 ret = dquot_acquire(dquot);
617ba13b 5370 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5371 if (!ret)
5372 ret = err;
5373 return ret;
5374}
5375
617ba13b 5376static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5377{
5378 int ret, err;
5379 handle_t *handle;
5380
9924a92a 5381 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5382 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5383 if (IS_ERR(handle)) {
5384 /* Release dquot anyway to avoid endless cycle in dqput() */
5385 dquot_release(dquot);
ac27a0ec 5386 return PTR_ERR(handle);
9c3013e9 5387 }
ac27a0ec 5388 ret = dquot_release(dquot);
617ba13b 5389 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5390 if (!ret)
5391 ret = err;
5392 return ret;
5393}
5394
617ba13b 5395static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5396{
262b4662
JK		 5397 struct super_block *sb = dquot->dq_sb;
5398 struct ext4_sb_info *sbi = EXT4_SB(sb);
5399
2c8be6b2 5400 /* Are we journaling quotas? */
e2b911c5 5401 if (ext4_has_feature_quota(sb) ||
262b4662 5402 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5403 dquot_mark_dquot_dirty(dquot);
617ba13b 5404 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5405 } else {
5406 return dquot_mark_dquot_dirty(dquot);
5407 }
5408}
5409
617ba13b 5410static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5411{
5412 int ret, err;
5413 handle_t *handle;
5414
5415 /* Data block + inode block */
2b0143b5 5416 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5417 if (IS_ERR(handle))
5418 return PTR_ERR(handle);
5419 ret = dquot_commit_info(sb, type);
617ba13b 5420 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5421 if (!ret)
5422 ret = err;
5423 return ret;
5424}
5425
5426/*
5427 * Turn on quotas during mount time - we need to find
5428 * the quota file and such...
5429 */
617ba13b 5430static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5431{
287a8095
CH		 5432 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5433 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5434}
5435
daf647d2
TT		 5436static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5437{
5438 struct ext4_inode_info *ei = EXT4_I(inode);
5439
5440 /* The first argument of lockdep_set_subclass has to be
5441 * *exactly* the same as the argument to init_rwsem() --- in
5442 * this case, in init_once() --- or lockdep gets unhappy
5443 * because the name of the lock is set using the
5444 * stringification of the argument to init_rwsem().
5445 */
5446 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5447 lockdep_set_subclass(&ei->i_data_sem, subclass);
5448}
5449
ac27a0ec
DK		 5450/*
5451 * Standard function to be called on quota_on
5452 */
617ba13b 5453static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5454 const struct path *path)
ac27a0ec
DK		 5455{
5456 int err;
ac27a0ec
DK		 5457
5458 if (!test_opt(sb, QUOTA))
5459 return -EINVAL;
0623543b 5460
ac27a0ec 5461 /* Quotafile not on the same filesystem? */
d8c9584e 5462 if (path->dentry->d_sb != sb)
ac27a0ec 5463 return -EXDEV;
0623543b
JK		 5464 /* Journaling quota? */
5465 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5466 /* Quotafile not in fs root? */
f00c9e44 5467 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5468 ext4_msg(sb, KERN_WARNING,
5469 "Quota file not on filesystem root. "
5470 "Journaled quota will not work");
2b2d6d01 5471 }
0623543b
JK		 5472
5473 /*
5474 * When we journal data on quota file, we have to flush journal to see
5475 * all updates to the file when we bypass pagecache...
5476 */
0390131b 5477 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5478 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5479 /*
5480 * We don't need to lock updates but journal_flush() could
5481 * otherwise be livelocked...
5482 */
5483 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5484 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5485 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5486 if (err)
7ffe1ea8 5487 return err;
0623543b 5488 }
957153fc 5489
daf647d2
TT		 5490 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5491 err = dquot_quota_on(sb, type, format_id, path);
957153fc 5492 if (err) {
daf647d2
TT		 5493 lockdep_set_quota_inode(path->dentry->d_inode,
5494 I_DATA_SEM_NORMAL);
957153fc
JK		 5495 } else {
5496 struct inode *inode = d_inode(path->dentry);
5497 handle_t *handle;
5498
61a92987
JK		 5499 /*
5500 * Set inode flags to prevent userspace from messing with quota
5501 * files. If this fails, we return success anyway since quotas
5502 * are already enabled and this is not a hard failure.
5503 */
957153fc
JK		 5504 inode_lock(inode);
5505 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5506 if (IS_ERR(handle))
5507 goto unlock_inode;
5508 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5509 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5510 S_NOATIME | S_IMMUTABLE);
5511 ext4_mark_inode_dirty(handle, inode);
5512 ext4_journal_stop(handle);
5513 unlock_inode:
5514 inode_unlock(inode);
5515 }
daf647d2 5516 return err;
ac27a0ec
DK		 5517}
5518
7c319d32
AK		 5519static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5520 unsigned int flags)
5521{
5522 int err;
5523 struct inode *qf_inode;
a2d4a646 5524 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5525 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5526 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5527 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5528 };
5529
e2b911c5 5530 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5531
5532 if (!qf_inums[type])
5533 return -EPERM;
5534
5535 qf_inode = ext4_iget(sb, qf_inums[type]);
5536 if (IS_ERR(qf_inode)) {
5537 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5538 return PTR_ERR(qf_inode);
5539 }
5540
bcb13850
JK		 5541 /* Don't account quota for quota files to avoid recursion */
5542 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5543 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5544 err = dquot_enable(qf_inode, type, format_id, flags);
5545 iput(qf_inode);
daf647d2
TT		 5546 if (err)
5547 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5548
5549 return err;
5550}
5551
5552/* Enable usage tracking for all quota types. */
5553static int ext4_enable_quotas(struct super_block *sb)
5554{
5555 int type, err = 0;
a2d4a646 5556 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5557 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5558 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5559 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5560 };
49da9392
JK		 5561 bool quota_mopt[EXT4_MAXQUOTAS] = {
5562 test_opt(sb, USRQUOTA),
5563 test_opt(sb, GRPQUOTA),
5564 test_opt(sb, PRJQUOTA),
5565 };
7c319d32
AK		 5566
5567 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5568 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5569 if (qf_inums[type]) {
5570 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5571 DQUOT_USAGE_ENABLED |
5572 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32 5573 if (err) {
8c6cd47e 5574 for (type--; type >= 0; type--)
5575 dquot_quota_off(sb, type);
5576
7c319d32 5577 ext4_warning(sb,
72ba7450
TT		 5578 "Failed to enable quota tracking "
5579 "(type=%d, err=%d). Please run "
5580 "e2fsck to fix.", type, err);
7c319d32
AK		 5581 return err;
5582 }
5583 }
5584 }
5585 return 0;
5586}
5587
ca0e05e4
DM		 5588static int ext4_quota_off(struct super_block *sb, int type)
5589{
21f97697
JK		 5590 struct inode *inode = sb_dqopt(sb)->files[type];
5591 handle_t *handle;
957153fc 5592 int err;
21f97697 5593
87009d86
DM		 5594 /* Force all delayed allocation blocks to be allocated.
5595 * Caller already holds s_umount sem */
5596 if (test_opt(sb, DELALLOC))
ca0e05e4 5597 sync_filesystem(sb);
ca0e05e4 5598
957153fc 5599 if (!inode || !igrab(inode))
0b268590
AG		 5600 goto out;
5601
957153fc 5602 err = dquot_quota_off(sb, type);
964edf66 5603 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 5604 goto out_put;
5605
5606 inode_lock(inode);
61a92987
JK		 5607 /*
5608 * Update modification times of quota files when userspace can
5609 * start looking at them. If we fail, we return success anyway since
5610 * this is not a hard failure and quotas are already disabled.
5611 */
9924a92a 5612 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697 5613 if (IS_ERR(handle))
957153fc
JK		 5614 goto out_unlock;
5615 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5616 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 5617 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5618 ext4_mark_inode_dirty(handle, inode);
5619 ext4_journal_stop(handle);
957153fc
JK		 5620out_unlock:
5621 inode_unlock(inode);
5622out_put:
964edf66 5623 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 5624 iput(inode);
5625 return err;
21f97697 5626out:
ca0e05e4
DM		 5627 return dquot_quota_off(sb, type);
5628}
5629
ac27a0ec
DK		 5630/* Read data from quotafile - avoid pagecache and such because we cannot afford
5631 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5632 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5633 * we don't have to be afraid of races */
617ba13b 5634static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5635 size_t len, loff_t off)
5636{
5637 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5638 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5639 int offset = off & (sb->s_blocksize - 1);
5640 int tocopy;
5641 size_t toread;
5642 struct buffer_head *bh;
5643 loff_t i_size = i_size_read(inode);
5644
5645 if (off > i_size)
5646 return 0;
5647 if (off+len > i_size)
5648 len = i_size-off;
5649 toread = len;
5650 while (toread > 0) {
5651 tocopy = sb->s_blocksize - offset < toread ?
5652 sb->s_blocksize - offset : toread;
1c215028
TT		 5653 bh = ext4_bread(NULL, inode, blk, 0);
5654 if (IS_ERR(bh))
5655 return PTR_ERR(bh);
ac27a0ec
DK		 5656 if (!bh) /* A hole? */
5657 memset(data, 0, tocopy);
5658 else
5659 memcpy(data, bh->b_data+offset, tocopy);
5660 brelse(bh);
5661 offset = 0;
5662 toread -= tocopy;
5663 data += tocopy;
5664 blk++;
5665 }
5666 return len;
5667}
5668
5669/* Write to quotafile (we know the transaction is already started and has
5670 * enough credits) */
617ba13b 5671static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5672 const char *data, size_t len, loff_t off)
5673{
5674 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5675 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5676 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5677 int retries = 0;
ac27a0ec
DK		 5678 struct buffer_head *bh;
5679 handle_t *handle = journal_current_handle();
5680
0390131b 5681 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5682 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5683 " cancelled because transaction is not started",
9c3013e9
JK		 5684 (unsigned long long)off, (unsigned long long)len);
5685 return -EIO;
5686 }
67eeb568
DM		 5687 /*
5688 * Since we account only one data block in transaction credits,
5689 * then it is impossible to cross a block boundary.
5690 */
5691 if (sb->s_blocksize - offset < len) {
5692 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5693 " cancelled because not block aligned",
5694 (unsigned long long)off, (unsigned long long)len);
5695 return -EIO;
5696 }
5697
c5e298ae
TT		 5698 do {
5699 bh = ext4_bread(handle, inode, blk,
5700 EXT4_GET_BLOCKS_CREATE |
5701 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5702 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5703 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5704 if (IS_ERR(bh))
5705 return PTR_ERR(bh);
67eeb568
DM		 5706 if (!bh)
5707 goto out;
5d601255 5708 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5709 err = ext4_journal_get_write_access(handle, bh);
5710 if (err) {
5711 brelse(bh);
1c215028 5712 return err;
ac27a0ec 5713 }
67eeb568
DM		 5714 lock_buffer(bh);
5715 memcpy(bh->b_data+offset, data, len);
5716 flush_dcache_page(bh->b_page);
5717 unlock_buffer(bh);
62d2b5f2 5718 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5719 brelse(bh);
ac27a0ec 5720out:
67eeb568
DM		 5721 if (inode->i_size < off + len) {
5722 i_size_write(inode, off + len);
617ba13b 5723 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5724 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5725 }
67eeb568 5726 return len;
ac27a0ec
DK		 5727}
5728
8f0e8746
TT		 5729static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5730{
5731 const struct quota_format_ops *ops;
5732
5733 if (!sb_has_quota_loaded(sb, qid->type))
5734 return -ESRCH;
5735 ops = sb_dqopt(sb)->ops[qid->type];
5736 if (!ops || !ops->get_next_id)
5737 return -ENOSYS;
5738 return dquot_get_next_id(sb, qid);
5739}
ac27a0ec
DK		 5740#endif
5741
152a0836
AV		 5742static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5743 const char *dev_name, void *data)
ac27a0ec 5744{
152a0836 5745 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5746}
5747
c290ea01 5748#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5749static inline void register_as_ext2(void)
5750{
5751 int err = register_filesystem(&ext2_fs_type);
5752 if (err)
5753 printk(KERN_WARNING
5754 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5755}
5756
5757static inline void unregister_as_ext2(void)
5758{
5759 unregister_filesystem(&ext2_fs_type);
5760}
2035e776
TT		 5761
5762static inline int ext2_feature_set_ok(struct super_block *sb)
5763{
e2b911c5 5764 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776
TT		 5765 return 0;
5766 if (sb->s_flags & MS_RDONLY)
5767 return 1;
e2b911c5 5768 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5769 return 0;
5770 return 1;
5771}
24b58424
TT		 5772#else
5773static inline void register_as_ext2(void) { }
5774static inline void unregister_as_ext2(void) { }
2035e776 5775static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5776#endif
5777
24b58424
TT		 5778static inline void register_as_ext3(void)
5779{
5780 int err = register_filesystem(&ext3_fs_type);
5781 if (err)
5782 printk(KERN_WARNING
5783 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5784}
5785
5786static inline void unregister_as_ext3(void)
5787{
5788 unregister_filesystem(&ext3_fs_type);
5789}
2035e776
TT		 5790
5791static inline int ext3_feature_set_ok(struct super_block *sb)
5792{
e2b911c5 5793 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5794 return 0;
e2b911c5 5795 if (!ext4_has_feature_journal(sb))
2035e776
TT		 5796 return 0;
5797 if (sb->s_flags & MS_RDONLY)
5798 return 1;
e2b911c5 5799 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5800 return 0;
5801 return 1;
5802}
24b58424 5803
03010a33
TT		 5804static struct file_system_type ext4_fs_type = {
5805 .owner = THIS_MODULE,
5806 .name = "ext4",
152a0836 5807 .mount = ext4_mount,
03010a33 5808 .kill_sb = kill_block_super,
afdbd4a6 5809 .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT,
03010a33 5810};
7f78e035 5811MODULE_ALIAS_FS("ext4");
03010a33 5812
e9e3bcec
ES		 5813/* Shared across all ext4 file systems */
5814wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5815
5dabfc78 5816static int __init ext4_init_fs(void)
ac27a0ec 5817{
e9e3bcec 5818 int i, err;
c9de560d 5819
e294a537 5820 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5821 ext4_li_info = NULL;
5822 mutex_init(&ext4_li_mtx);
5823
9a4c8019 5824 /* Build-time check for flags consistency */
12e9b892 5825 ext4_check_flag_values();
e9e3bcec 5826
e142d052 5827 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5828 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5829
51865fda 5830 err = ext4_init_es();
6fd058f7
TT		 5831 if (err)
5832 return err;
51865fda
ZL		 5833
5834 err = ext4_init_pageio();
5835 if (err)
b5799018 5836 goto out5;
51865fda 5837
5dabfc78 5838 err = ext4_init_system_zone();
bd2d0210 5839 if (err)
b5799018 5840 goto out4;
857ac889 5841
b5799018 5842 err = ext4_init_sysfs();
dd68314c 5843 if (err)
b5799018 5844 goto out3;
857ac889 5845
5dabfc78 5846 err = ext4_init_mballoc();
c9de560d
AT		 5847 if (err)
5848 goto out2;
ac27a0ec
DK		 5849 err = init_inodecache();
5850 if (err)
5851 goto out1;
24b58424 5852 register_as_ext3();
2035e776 5853 register_as_ext2();
03010a33 5854 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5855 if (err)
5856 goto out;
bfff6873 5857
ac27a0ec
DK		 5858 return 0;
5859out:
24b58424
TT		 5860 unregister_as_ext2();
5861 unregister_as_ext3();
ac27a0ec
DK		 5862 destroy_inodecache();
5863out1:
5dabfc78 5864 ext4_exit_mballoc();
9c191f70 5865out2:
b5799018
TT		 5866 ext4_exit_sysfs();
5867out3:
5dabfc78 5868 ext4_exit_system_zone();
b5799018 5869out4:
5dabfc78 5870 ext4_exit_pageio();
b5799018 5871out5:
51865fda
ZL		 5872 ext4_exit_es();
5873
ac27a0ec
DK		 5874 return err;
5875}
5876
5dabfc78 5877static void __exit ext4_exit_fs(void)
ac27a0ec 5878{
bfff6873 5879 ext4_destroy_lazyinit_thread();
24b58424
TT		 5880 unregister_as_ext2();
5881 unregister_as_ext3();
03010a33 5882 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5883 destroy_inodecache();
5dabfc78 5884 ext4_exit_mballoc();
b5799018 5885 ext4_exit_sysfs();
5dabfc78
TT		 5886 ext4_exit_system_zone();
5887 ext4_exit_pageio();
dd12ed14 5888 ext4_exit_es();
ac27a0ec
DK		 5889}
5890
5891MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5892MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5893MODULE_LICENSE("GPL");
5dabfc78
TT		 5894module_init(ext4_init_fs)
5895module_exit(ext4_exit_fs)
Ubuntu Artful kernel mirror