]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: correctly detect when an xattr value has an invalid size
[thirdparty/kernel/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
66114cad 27#include <linux/backing-dev.h>
ac27a0ec 28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
3197ebdb 37#include <linux/ctype.h>
1330593e 38#include <linux/log2.h>
717d50e4 39#include <linux/crc16.h>
7abc52c2 40#include <linux/cleancache.h>
ac27a0ec
DK		 41#include <asm/uaccess.h>
42
bfff6873
LC		 43#include <linux/kthread.h>
44#include <linux/freezer.h>
45
3dcf5451 46#include "ext4.h"
4a092d73 47#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 48#include "ext4_jbd2.h"
ac27a0ec
DK		 49#include "xattr.h"
50#include "acl.h"
3661d286 51#include "mballoc.h"
ac27a0ec 52
9bffad1e
TT		 53#define CREATE_TRACE_POINTS
54#include <trace/events/ext4.h>
55
0b75a840
LC		 56static struct ext4_lazy_init *ext4_li_info;
57static struct mutex ext4_li_mtx;
e294a537 58static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 59
617ba13b 60static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 61 unsigned long journal_devnum);
2adf6da8 62static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 63static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 64static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
617ba13b 68static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 69static int ext4_remount(struct super_block *sb, int *flags, char *data);
70static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 71static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 72static int ext4_freeze(struct super_block *sb);
152a0836
AV		 73static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74 const char *dev_name, void *data);
2035e776
TT		 75static inline int ext2_feature_set_ok(struct super_block *sb);
76static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 77static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 78static void ext4_destroy_lazyinit_thread(void);
79static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 80static void ext4_clear_request_list(void);
c6cb7e77
EW		 81static struct inode *ext4_get_journal_inode(struct super_block *sb,
82 unsigned int journal_inum);
ac27a0ec 83
e74031fd
JK		 84/*
85 * Lock ordering
86 *
87 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
88 * i_mmap_rwsem (inode->i_mmap_rwsem)!
89 *
90 * page fault path:
91 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
92 * page lock -> i_data_sem (rw)
93 *
94 * buffered write path:
95 * sb_start_write -> i_mutex -> mmap_sem
96 * sb_start_write -> i_mutex -> transaction start -> page lock ->
97 * i_data_sem (rw)
98 *
99 * truncate:
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * i_mmap_rwsem (w) -> page lock
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * transaction start -> i_data_sem (rw)
104 *
105 * direct IO:
106 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
108 * transaction start -> i_data_sem (rw)
109 *
110 * writepages:
111 * transaction start -> page lock(s) -> i_data_sem (rw)
112 */
113
c290ea01 114#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 115static struct file_system_type ext2_fs_type = {
116 .owner = THIS_MODULE,
117 .name = "ext2",
118 .mount = ext4_mount,
119 .kill_sb = kill_block_super,
120 .fs_flags = FS_REQUIRES_DEV,
121};
7f78e035 122MODULE_ALIAS_FS("ext2");
fa7614dd 123MODULE_ALIAS("ext2");
2035e776
TT		 124#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
125#else
126#define IS_EXT2_SB(sb) (0)
127#endif
128
129
ba69f9ab
JK		 130static struct file_system_type ext3_fs_type = {
131 .owner = THIS_MODULE,
132 .name = "ext3",
152a0836 133 .mount = ext4_mount,
ba69f9ab
JK		 134 .kill_sb = kill_block_super,
135 .fs_flags = FS_REQUIRES_DEV,
136};
7f78e035 137MODULE_ALIAS_FS("ext3");
fa7614dd 138MODULE_ALIAS("ext3");
ba69f9ab 139#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 140
d25425f8
DW		 141static int ext4_verify_csum_type(struct super_block *sb,
142 struct ext4_super_block *es)
143{
e2b911c5 144 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 145 return 1;
146
147 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
148}
149
a9c47317
DW		 150static __le32 ext4_superblock_csum(struct super_block *sb,
151 struct ext4_super_block *es)
152{
153 struct ext4_sb_info *sbi = EXT4_SB(sb);
154 int offset = offsetof(struct ext4_super_block, s_checksum);
155 __u32 csum;
156
157 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
158
159 return cpu_to_le32(csum);
160}
161
c197855e
SH		 162static int ext4_superblock_csum_verify(struct super_block *sb,
163 struct ext4_super_block *es)
a9c47317 164{
9aa5d32b 165 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 166 return 1;
167
168 return es->s_checksum == ext4_superblock_csum(sb, es);
169}
170
06db49e6 171void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 172{
06db49e6
TT		 173 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
174
9aa5d32b 175 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 176 return;
177
178 es->s_checksum = ext4_superblock_csum(sb, es);
179}
180
9933fc0a
TT		 181void *ext4_kvmalloc(size_t size, gfp_t flags)
182{
183 void *ret;
184
8be04b93 185 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 186 if (!ret)
187 ret = __vmalloc(size, flags, PAGE_KERNEL);
188 return ret;
189}
190
191void *ext4_kvzalloc(size_t size, gfp_t flags)
192{
193 void *ret;
194
8be04b93 195 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 196 if (!ret)
197 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
198 return ret;
199}
200
8fadc143
AR		 201ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
202 struct ext4_group_desc *bg)
bd81d8ee 203{
3a14589c 204 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 206 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 207}
208
8fadc143
AR		 209ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
210 struct ext4_group_desc *bg)
bd81d8ee 211{
5272f837 212 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 214 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 215}
216
8fadc143
AR		 217ext4_fsblk_t ext4_inode_table(struct super_block *sb,
218 struct ext4_group_desc *bg)
bd81d8ee 219{
5272f837 220 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 222 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 223}
224
021b65bb
TT		 225__u32 ext4_free_group_clusters(struct super_block *sb,
226 struct ext4_group_desc *bg)
560671a0
AK		 227{
228 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 230 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 231}
232
233__u32 ext4_free_inodes_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
235{
236 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 238 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 239}
240
241__u32 ext4_used_dirs_count(struct super_block *sb,
242 struct ext4_group_desc *bg)
243{
244 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
245 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 246 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 247}
248
249__u32 ext4_itable_unused_count(struct super_block *sb,
250 struct ext4_group_desc *bg)
251{
252 return le16_to_cpu(bg->bg_itable_unused_lo) |
253 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 254 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 255}
256
8fadc143
AR		 257void ext4_block_bitmap_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 259{
3a14589c 260 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 263}
264
8fadc143
AR		 265void ext4_inode_bitmap_set(struct super_block *sb,
266 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 267{
5272f837 268 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 271}
272
8fadc143
AR		 273void ext4_inode_table_set(struct super_block *sb,
274 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 275{
5272f837 276 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 279}
280
021b65bb
TT		 281void ext4_free_group_clusters_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 283{
284 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
287}
288
289void ext4_free_inodes_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
291{
292 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
295}
296
297void ext4_used_dirs_set(struct super_block *sb,
298 struct ext4_group_desc *bg, __u32 count)
299{
300 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
301 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
302 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
303}
304
305void ext4_itable_unused_set(struct super_block *sb,
306 struct ext4_group_desc *bg, __u32 count)
307{
308 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
309 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
310 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
311}
312
d3d1faf6 313
1c13d5c0
TT		 314static void __save_error_info(struct super_block *sb, const char *func,
315 unsigned int line)
316{
317 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
318
319 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 320 if (bdev_read_only(sb->s_bdev))
321 return;
1c13d5c0
TT		 322 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
323 es->s_last_error_time = cpu_to_le32(get_seconds());
324 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
325 es->s_last_error_line = cpu_to_le32(line);
326 if (!es->s_first_error_time) {
327 es->s_first_error_time = es->s_last_error_time;
328 strncpy(es->s_first_error_func, func,
329 sizeof(es->s_first_error_func));
330 es->s_first_error_line = cpu_to_le32(line);
331 es->s_first_error_ino = es->s_last_error_ino;
332 es->s_first_error_block = es->s_last_error_block;
333 }
66e61a9e
TT		 334 /*
335 * Start the daily error reporting function if it hasn't been
336 * started already
337 */
338 if (!es->s_error_count)
339 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 340 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 341}
342
343static void save_error_info(struct super_block *sb, const char *func,
344 unsigned int line)
345{
346 __save_error_info(sb, func, line);
347 ext4_commit_super(sb, 1);
348}
349
bdfe0cbd
TT		 350/*
351 * The del_gendisk() function uninitializes the disk-specific data
352 * structures, including the bdi structure, without telling anyone
353 * else. Once this happens, any attempt to call mark_buffer_dirty()
354 * (for example, by ext4_commit_super), will cause a kernel OOPS.
355 * This is a kludge to prevent these oops until we can put in a proper
356 * hook in del_gendisk() to inform the VFS and file system layers.
357 */
358static int block_device_ejected(struct super_block *sb)
359{
360 struct inode *bd_inode = sb->s_bdev->bd_inode;
361 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
362
363 return bdi->dev == NULL;
364}
365
18aadd47
BJ		 366static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
367{
368 struct super_block *sb = journal->j_private;
369 struct ext4_sb_info *sbi = EXT4_SB(sb);
370 int error = is_journal_aborted(journal);
5d3ee208 371 struct ext4_journal_cb_entry *jce;
18aadd47 372
5d3ee208 373 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 374 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 375 while (!list_empty(&txn->t_private_list)) {
376 jce = list_entry(txn->t_private_list.next,
377 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 378 list_del_init(&jce->jce_list);
379 spin_unlock(&sbi->s_md_lock);
380 jce->jce_func(sb, jce, error);
381 spin_lock(&sbi->s_md_lock);
382 }
383 spin_unlock(&sbi->s_md_lock);
384}
1c13d5c0 385
ac27a0ec
DK		 386/* Deal with the reporting of failure conditions on a filesystem such as
387 * inconsistencies detected or read IO failures.
388 *
389 * On ext2, we can store the error state of the filesystem in the
617ba13b 390 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 391 * write ordering constraints on the superblock which prevent us from
392 * writing it out straight away; and given that the journal is about to
393 * be aborted, we can't rely on the current, or future, transactions to
394 * write out the superblock safely.
395 *
dab291af 396 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 397 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 398 * that error until we've noted it down and cleared it.
399 */
400
617ba13b 401static void ext4_handle_error(struct super_block *sb)
ac27a0ec 402{
ac27a0ec
DK		 403 if (sb->s_flags & MS_RDONLY)
404 return;
405
2b2d6d01 406 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 407 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 408
4ab2f15b 409 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 410 if (journal)
dab291af 411 jbd2_journal_abort(journal, -EIO);
ac27a0ec 412 }
2b2d6d01 413 if (test_opt(sb, ERRORS_RO)) {
b31e1552 414 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 415 /*
416 * Make sure updated value of ->s_mount_flags will be visible
417 * before ->s_flags update
418 */
419 smp_wmb();
ac27a0ec
DK		 420 sb->s_flags |= MS_RDONLY;
421 }
4327ba52
DJ		 422 if (test_opt(sb, ERRORS_PANIC)) {
423 if (EXT4_SB(sb)->s_journal &&
424 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
425 return;
617ba13b 426 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 427 sb->s_id);
4327ba52 428 }
ac27a0ec
DK		 429}
430
efbed4dc
TT		 431#define ext4_error_ratelimit(sb) \
432 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
433 "EXT4-fs error")
434
12062ddd 435void __ext4_error(struct super_block *sb, const char *function,
c398eda0 436 unsigned int line, const char *fmt, ...)
ac27a0ec 437{
0ff2ea7d 438 struct va_format vaf;
ac27a0ec
DK		 439 va_list args;
440
efbed4dc
TT		 441 if (ext4_error_ratelimit(sb)) {
442 va_start(args, fmt);
443 vaf.fmt = fmt;
444 vaf.va = &args;
445 printk(KERN_CRIT
446 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
447 sb->s_id, function, line, current->comm, &vaf);
448 va_end(args);
449 }
f3fc0210 450 save_error_info(sb, function, line);
617ba13b 451 ext4_handle_error(sb);
ac27a0ec
DK		 452}
453
e7c96e8e
JP		 454void __ext4_error_inode(struct inode *inode, const char *function,
455 unsigned int line, ext4_fsblk_t block,
456 const char *fmt, ...)
273df556
FM		 457{
458 va_list args;
f7c21177 459 struct va_format vaf;
1c13d5c0 460 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 461
1c13d5c0
TT		 462 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
463 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 464 if (ext4_error_ratelimit(inode->i_sb)) {
465 va_start(args, fmt);
466 vaf.fmt = fmt;
467 vaf.va = &args;
468 if (block)
469 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
470 "inode #%lu: block %llu: comm %s: %pV\n",
471 inode->i_sb->s_id, function, line, inode->i_ino,
472 block, current->comm, &vaf);
473 else
474 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
475 "inode #%lu: comm %s: %pV\n",
476 inode->i_sb->s_id, function, line, inode->i_ino,
477 current->comm, &vaf);
478 va_end(args);
479 }
1c13d5c0 480 save_error_info(inode->i_sb, function, line);
273df556
FM		 481 ext4_handle_error(inode->i_sb);
482}
483
e7c96e8e
JP		 484void __ext4_error_file(struct file *file, const char *function,
485 unsigned int line, ext4_fsblk_t block,
486 const char *fmt, ...)
273df556
FM		 487{
488 va_list args;
f7c21177 489 struct va_format vaf;
1c13d5c0 490 struct ext4_super_block *es;
496ad9aa 491 struct inode *inode = file_inode(file);
273df556
FM		 492 char pathname[80], *path;
493
1c13d5c0
TT		 494 es = EXT4_SB(inode->i_sb)->s_es;
495 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 496 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 497 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 498 if (IS_ERR(path))
499 path = "(unknown)";
500 va_start(args, fmt);
501 vaf.fmt = fmt;
502 vaf.va = &args;
503 if (block)
504 printk(KERN_CRIT
505 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
506 "block %llu: comm %s: path %s: %pV\n",
507 inode->i_sb->s_id, function, line, inode->i_ino,
508 block, current->comm, path, &vaf);
509 else
510 printk(KERN_CRIT
511 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
512 "comm %s: path %s: %pV\n",
513 inode->i_sb->s_id, function, line, inode->i_ino,
514 current->comm, path, &vaf);
515 va_end(args);
516 }
1c13d5c0 517 save_error_info(inode->i_sb, function, line);
273df556
FM		 518 ext4_handle_error(inode->i_sb);
519}
520
722887dd
TT		 521const char *ext4_decode_error(struct super_block *sb, int errno,
522 char nbuf[16])
ac27a0ec
DK		 523{
524 char *errstr = NULL;
525
526 switch (errno) {
6a797d27
DW		 527 case -EFSCORRUPTED:
528 errstr = "Corrupt filesystem";
529 break;
530 case -EFSBADCRC:
531 errstr = "Filesystem failed CRC";
532 break;
ac27a0ec
DK		 533 case -EIO:
534 errstr = "IO failure";
535 break;
536 case -ENOMEM:
537 errstr = "Out of memory";
538 break;
539 case -EROFS:
78f1ddbb
TT		 540 if (!sb || (EXT4_SB(sb)->s_journal &&
541 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 542 errstr = "Journal has aborted";
543 else
544 errstr = "Readonly filesystem";
545 break;
546 default:
547 /* If the caller passed in an extra buffer for unknown
548 * errors, textualise them now. Else we just return
549 * NULL. */
550 if (nbuf) {
551 /* Check for truncated error codes... */
552 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
553 errstr = nbuf;
554 }
555 break;
556 }
557
558 return errstr;
559}
560
617ba13b 561/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 562 * automatically and invokes the appropriate error response. */
563
c398eda0
TT		 564void __ext4_std_error(struct super_block *sb, const char *function,
565 unsigned int line, int errno)
ac27a0ec
DK		 566{
567 char nbuf[16];
568 const char *errstr;
569
570 /* Special case: if the error is EROFS, and we're not already
571 * inside a transaction, then there's really no point in logging
572 * an error. */
573 if (errno == -EROFS && journal_current_handle() == NULL &&
574 (sb->s_flags & MS_RDONLY))
575 return;
576
efbed4dc
TT		 577 if (ext4_error_ratelimit(sb)) {
578 errstr = ext4_decode_error(sb, errno, nbuf);
579 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
580 sb->s_id, function, line, errstr);
581 }
ac27a0ec 582
efbed4dc 583 save_error_info(sb, function, line);
617ba13b 584 ext4_handle_error(sb);
ac27a0ec
DK		 585}
586
587/*
617ba13b 588 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 589 * abort function may be used to deal with unrecoverable failures such
590 * as journal IO errors or ENOMEM at a critical moment in log management.
591 *
592 * We unconditionally force the filesystem into an ABORT|READONLY state,
593 * unless the error response on the fs has been set to panic in which
594 * case we take the easy way out and panic immediately.
595 */
596
c67d859e 597void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 598 unsigned int line, const char *fmt, ...)
ac27a0ec 599{
651e1c3b 600 struct va_format vaf;
ac27a0ec
DK		 601 va_list args;
602
1c13d5c0 603 save_error_info(sb, function, line);
ac27a0ec 604 va_start(args, fmt);
651e1c3b
JP		 605 vaf.fmt = fmt;
606 vaf.va = &args;
607 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
608 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 609 va_end(args);
610
1c13d5c0
TT		 611 if ((sb->s_flags & MS_RDONLY) == 0) {
612 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 613 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 614 /*
615 * Make sure updated value of ->s_mount_flags will be visible
616 * before ->s_flags update
617 */
618 smp_wmb();
619 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 620 if (EXT4_SB(sb)->s_journal)
621 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
622 save_error_info(sb, function, line);
623 }
4327ba52
DJ		 624 if (test_opt(sb, ERRORS_PANIC)) {
625 if (EXT4_SB(sb)->s_journal &&
626 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
627 return;
617ba13b 628 panic("EXT4-fs panic from previous error\n");
4327ba52 629 }
ac27a0ec
DK		 630}
631
e7c96e8e
JP		 632void __ext4_msg(struct super_block *sb,
633 const char *prefix, const char *fmt, ...)
b31e1552 634{
0ff2ea7d 635 struct va_format vaf;
b31e1552
ES		 636 va_list args;
637
efbed4dc
TT		 638 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
639 return;
640
b31e1552 641 va_start(args, fmt);
0ff2ea7d
JP		 642 vaf.fmt = fmt;
643 vaf.va = &args;
644 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 645 va_end(args);
646}
647
b03a2f7e
AD		 648#define ext4_warning_ratelimit(sb) \
649 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
650 "EXT4-fs warning")
651
12062ddd 652void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 653 unsigned int line, const char *fmt, ...)
ac27a0ec 654{
0ff2ea7d 655 struct va_format vaf;
ac27a0ec
DK		 656 va_list args;
657
b03a2f7e 658 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 659 return;
660
ac27a0ec 661 va_start(args, fmt);
0ff2ea7d
JP		 662 vaf.fmt = fmt;
663 vaf.va = &args;
664 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
665 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 666 va_end(args);
667}
668
b03a2f7e
AD		 669void __ext4_warning_inode(const struct inode *inode, const char *function,
670 unsigned int line, const char *fmt, ...)
671{
672 struct va_format vaf;
673 va_list args;
674
675 if (!ext4_warning_ratelimit(inode->i_sb))
676 return;
677
678 va_start(args, fmt);
679 vaf.fmt = fmt;
680 vaf.va = &args;
681 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
682 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
683 function, line, inode->i_ino, current->comm, &vaf);
684 va_end(args);
685}
686
e29136f8
TT		 687void __ext4_grp_locked_error(const char *function, unsigned int line,
688 struct super_block *sb, ext4_group_t grp,
689 unsigned long ino, ext4_fsblk_t block,
690 const char *fmt, ...)
5d1b1b3f
AK		 691__releases(bitlock)
692__acquires(bitlock)
693{
0ff2ea7d 694 struct va_format vaf;
5d1b1b3f
AK		 695 va_list args;
696 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
697
1c13d5c0
TT		 698 es->s_last_error_ino = cpu_to_le32(ino);
699 es->s_last_error_block = cpu_to_le64(block);
700 __save_error_info(sb, function, line);
0ff2ea7d 701
efbed4dc
TT		 702 if (ext4_error_ratelimit(sb)) {
703 va_start(args, fmt);
704 vaf.fmt = fmt;
705 vaf.va = &args;
706 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
707 sb->s_id, function, line, grp);
708 if (ino)
709 printk(KERN_CONT "inode %lu: ", ino);
710 if (block)
711 printk(KERN_CONT "block %llu:",
712 (unsigned long long) block);
713 printk(KERN_CONT "%pV\n", &vaf);
714 va_end(args);
715 }
5d1b1b3f
AK		 716
717 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 718 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 719 return;
720 }
1c13d5c0 721
5d1b1b3f
AK		 722 ext4_unlock_group(sb, grp);
723 ext4_handle_error(sb);
724 /*
725 * We only get here in the ERRORS_RO case; relocking the group
726 * may be dangerous, but nothing bad will happen since the
727 * filesystem will have already been marked read/only and the
728 * journal has been aborted. We return 1 as a hint to callers
729 * who might what to use the return value from
25985edc 730 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 731 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
732 * aggressively from the ext4 function in question, with a
733 * more appropriate error code.
734 */
735 ext4_lock_group(sb, grp);
736 return;
737}
738
617ba13b 739void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 740{
617ba13b 741 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 742
617ba13b 743 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 744 return;
745
12062ddd 746 ext4_warning(sb,
ac27a0ec
DK		 747 "updating to rev %d because of new feature flag, "
748 "running e2fsck is recommended",
617ba13b 749 EXT4_DYNAMIC_REV);
ac27a0ec 750
617ba13b
MC		 751 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
752 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
753 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 754 /* leave es->s_feature_*compat flags alone */
755 /* es->s_uuid will be set by e2fsck if empty */
756
757 /*
758 * The rest of the superblock fields should be zero, and if not it
759 * means they are likely already in use, so leave them alone. We
760 * can leave it up to e2fsck to clean up any inconsistencies there.
761 */
762}
763
764/*
765 * Open the external journal device
766 */
b31e1552 767static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 768{
769 struct block_device *bdev;
770 char b[BDEVNAME_SIZE];
771
d4d77629 772 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 773 if (IS_ERR(bdev))
774 goto fail;
775 return bdev;
776
777fail:
b31e1552 778 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 779 __bdevname(dev, b), PTR_ERR(bdev));
780 return NULL;
781}
782
783/*
784 * Release the journal device
785 */
4385bab1 786static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 787{
4385bab1 788 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 789}
790
4385bab1 791static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 792{
793 struct block_device *bdev;
ac27a0ec
DK		 794 bdev = sbi->journal_bdev;
795 if (bdev) {
4385bab1 796 ext4_blkdev_put(bdev);
ac27a0ec
DK		 797 sbi->journal_bdev = NULL;
798 }
ac27a0ec
DK		 799}
800
801static inline struct inode *orphan_list_entry(struct list_head *l)
802{
617ba13b 803 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 804}
805
617ba13b 806static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 807{
808 struct list_head *l;
809
b31e1552
ES		 810 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
811 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 812
813 printk(KERN_ERR "sb_info orphan list:\n");
814 list_for_each(l, &sbi->s_orphan) {
815 struct inode *inode = orphan_list_entry(l);
816 printk(KERN_ERR " "
817 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
818 inode->i_sb->s_id, inode->i_ino, inode,
819 inode->i_mode, inode->i_nlink,
820 NEXT_ORPHAN(inode));
821 }
822}
823
2b2d6d01 824static void ext4_put_super(struct super_block *sb)
ac27a0ec 825{
617ba13b
MC		 826 struct ext4_sb_info *sbi = EXT4_SB(sb);
827 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 828 int i, err;
ac27a0ec 829
857ac889 830 ext4_unregister_li_request(sb);
e0ccfd95
CH		 831 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
832
2e8fa54e 833 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 834 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 835
0390131b
FM		 836 if (sbi->s_journal) {
837 err = jbd2_journal_destroy(sbi->s_journal);
838 sbi->s_journal = NULL;
839 if (err < 0)
c67d859e 840 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 841 }
d4edac31 842
ebd173be 843 ext4_unregister_sysfs(sb);
d3922a77 844 ext4_es_unregister_shrinker(sbi);
9105bb14 845 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 846 ext4_release_system_zone(sb);
847 ext4_mb_release(sb);
848 ext4_ext_release(sb);
d4edac31 849
ac27a0ec 850 if (!(sb->s_flags & MS_RDONLY)) {
e2b911c5 851 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 852 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 853 }
58c5873a 854 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 855 ext4_commit_super(sb, 1);
856
ac27a0ec
DK		 857 for (i = 0; i < sbi->s_gdb_count; i++)
858 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 859 kvfree(sbi->s_group_desc);
860 kvfree(sbi->s_flex_groups);
57042651 861 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 862 percpu_counter_destroy(&sbi->s_freeinodes_counter);
863 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 864 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c8585c6f 865 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec 866#ifdef CONFIG_QUOTA
a2d4a646 867 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 868 kfree(sbi->s_qf_names[i]);
869#endif
870
871 /* Debugging code just in case the in-memory inode orphan list
872 * isn't empty. The on-disk one can be non-empty if we've
873 * detected an error and taken the fs readonly, but the
874 * in-memory list had better be clean by this point. */
875 if (!list_empty(&sbi->s_orphan))
876 dump_orphan_list(sb, sbi);
877 J_ASSERT(list_empty(&sbi->s_orphan));
878
89d96a6f 879 sync_blockdev(sb->s_bdev);
f98393a6 880 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 881 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
882 /*
883 * Invalidate the journal device's buffers. We don't want them
884 * floating about in memory - the physical journal device may
885 * hotswapped, and it breaks the `ro-after' testing code.
886 */
887 sync_blockdev(sbi->journal_bdev);
f98393a6 888 invalidate_bdev(sbi->journal_bdev);
617ba13b 889 ext4_blkdev_remove(sbi);
ac27a0ec 890 }
9c191f70
M		 891 if (sbi->s_mb_cache) {
892 ext4_xattr_destroy_cache(sbi->s_mb_cache);
893 sbi->s_mb_cache = NULL;
894 }
c5e06d10
JL		 895 if (sbi->s_mmp_tsk)
896 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 897 brelse(sbi->s_sbh);
ac27a0ec 898 sb->s_fs_info = NULL;
3197ebdb
TT		 899 /*
900 * Now that we are completely done shutting down the
901 * superblock, we need to actually destroy the kobject.
902 */
3197ebdb
TT		 903 kobject_put(&sbi->s_kobj);
904 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 905 if (sbi->s_chksum_driver)
906 crypto_free_shash(sbi->s_chksum_driver);
705895b6 907 kfree(sbi->s_blockgroup_lock);
ac27a0ec 908 kfree(sbi);
ac27a0ec
DK		 909}
910
e18b890b 911static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 912
913/*
914 * Called inside transaction, so use GFP_NOFS
915 */
617ba13b 916static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 917{
617ba13b 918 struct ext4_inode_info *ei;
ac27a0ec 919
e6b4f8da 920 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 921 if (!ei)
922 return NULL;
0b8e58a1 923
ac27a0ec 924 ei->vfs_inode.i_version = 1;
202ee5df 925 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 926 INIT_LIST_HEAD(&ei->i_prealloc_list);
927 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 928 ext4_es_init_tree(&ei->i_es_tree);
929 rwlock_init(&ei->i_es_lock);
edaa53ca 930 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 931 ei->i_es_all_nr = 0;
edaa53ca 932 ei->i_es_shk_nr = 0;
dd475925 933 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 934 ei->i_reserved_data_blocks = 0;
935 ei->i_reserved_meta_blocks = 0;
936 ei->i_allocated_meta_blocks = 0;
9d0be502 937 ei->i_da_metadata_calc_len = 0;
7e731bc9 938 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 939 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 940#ifdef CONFIG_QUOTA
941 ei->i_reserved_quota = 0;
96c7e0d9 942 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 943#endif
8aefcd55 944 ei->jinode = NULL;
2e8fa54e 945 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 946 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 947 ei->i_sync_tid = 0;
948 ei->i_datasync_tid = 0;
e27f41e1 949 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 950 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 951 return &ei->vfs_inode;
952}
953
7ff9c073
TT		 954static int ext4_drop_inode(struct inode *inode)
955{
956 int drop = generic_drop_inode(inode);
957
958 trace_ext4_drop_inode(inode, drop);
959 return drop;
960}
961
fa0d7e3d
NP		 962static void ext4_i_callback(struct rcu_head *head)
963{
964 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 965 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
966}
967
617ba13b 968static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 969{
9f7dd93d 970 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 971 ext4_msg(inode->i_sb, KERN_ERR,
972 "Inode %lu (%p): orphan list check failed!",
973 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 974 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
975 EXT4_I(inode), sizeof(struct ext4_inode_info),
976 true);
977 dump_stack();
978 }
fa0d7e3d 979 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 980}
981
51cc5068 982static void init_once(void *foo)
ac27a0ec 983{
617ba13b 984 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 985
a35afb83 986 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 987 init_rwsem(&ei->xattr_sem);
0e855ac8 988 init_rwsem(&ei->i_data_sem);
ea3d7209 989 init_rwsem(&ei->i_mmap_sem);
a35afb83 990 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 991}
992
e67bc2b3 993static int __init init_inodecache(void)
ac27a0ec 994{
617ba13b
MC		 995 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
996 sizeof(struct ext4_inode_info),
ac27a0ec 997 0, (SLAB_RECLAIM_ACCOUNT|
5d097056 998 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
20c2df83 999 init_once);
617ba13b 1000 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1001 return -ENOMEM;
1002 return 0;
1003}
1004
1005static void destroy_inodecache(void)
1006{
8c0a8537
KS		 1007 /*
1008 * Make sure all delayed rcu free inodes are flushed before we
1009 * destroy cache.
1010 */
1011 rcu_barrier();
617ba13b 1012 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1013}
1014
0930fcc1 1015void ext4_clear_inode(struct inode *inode)
ac27a0ec 1016{
0930fcc1 1017 invalidate_inode_buffers(inode);
dbd5768f 1018 clear_inode(inode);
9f754758 1019 dquot_drop(inode);
c2ea3fde 1020 ext4_discard_preallocations(inode);
51865fda 1021 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1022 if (EXT4_I(inode)->jinode) {
1023 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1024 EXT4_I(inode)->jinode);
1025 jbd2_free_inode(EXT4_I(inode)->jinode);
1026 EXT4_I(inode)->jinode = NULL;
1027 }
b7236e21 1028#ifdef CONFIG_EXT4_FS_ENCRYPTION
a7550b30 1029 fscrypt_put_encryption_info(inode, NULL);
b7236e21 1030#endif
ac27a0ec
DK		 1031}
1032
1b961ac0 1033static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1034 u64 ino, u32 generation)
ac27a0ec 1035{
ac27a0ec 1036 struct inode *inode;
ac27a0ec 1037
617ba13b 1038 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1039 return ERR_PTR(-ESTALE);
617ba13b 1040 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1041 return ERR_PTR(-ESTALE);
1042
1043 /* iget isn't really right if the inode is currently unallocated!!
1044 *
617ba13b 1045 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1046 * deleted, so we should be safe.
1047 *
1048 * Currently we don't know the generation for parent directory, so
1049 * a generation of 0 means "accept any"
1050 */
f4bb2981 1051 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1052 if (IS_ERR(inode))
1053 return ERR_CAST(inode);
1054 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1055 iput(inode);
1056 return ERR_PTR(-ESTALE);
1057 }
1b961ac0
CH		 1058
1059 return inode;
1060}
1061
1062static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1063 int fh_len, int fh_type)
1b961ac0
CH		 1064{
1065 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1066 ext4_nfs_get_inode);
1067}
1068
1069static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1070 int fh_len, int fh_type)
1b961ac0
CH		 1071{
1072 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1073 ext4_nfs_get_inode);
ac27a0ec
DK		 1074}
1075
c39a7f84
TO		 1076/*
1077 * Try to release metadata pages (indirect blocks, directories) which are
1078 * mapped via the block device. Since these pages could have journal heads
1079 * which would prevent try_to_free_buffers() from freeing them, we must use
1080 * jbd2 layer's try_to_free_buffers() function to release them.
1081 */
0b8e58a1
AD		 1082static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1083 gfp_t wait)
c39a7f84
TO		 1084{
1085 journal_t *journal = EXT4_SB(sb)->s_journal;
1086
1087 WARN_ON(PageChecked(page));
1088 if (!page_has_buffers(page))
1089 return 0;
1090 if (journal)
1091 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1092 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1093 return try_to_free_buffers(page);
1094}
1095
a7550b30
JK		 1096#ifdef CONFIG_EXT4_FS_ENCRYPTION
1097static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1098{
1099 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1100 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1101}
1102
1103static int ext4_key_prefix(struct inode *inode, u8 **key)
1104{
1105 *key = EXT4_SB(inode->i_sb)->key_prefix;
1106 return EXT4_SB(inode->i_sb)->key_prefix_size;
1107}
1108
1109static int ext4_prepare_context(struct inode *inode)
1110{
1111 return ext4_convert_inline_data(inode);
1112}
1113
1114static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1115 void *fs_data)
1116{
2f8f5e76
EB		 1117 handle_t *handle = fs_data;
1118 int res, res2, retries = 0;
1119
1120 /*
1121 * If a journal handle was specified, then the encryption context is
1122 * being set on a new inode via inheritance and is part of a larger
1123 * transaction to create the inode. Otherwise the encryption context is
1124 * being set on an existing inode in its own transaction. Only in the
1125 * latter case should the "retry on ENOSPC" logic be used.
1126 */
1127
1128 if (handle) {
1129 res = ext4_xattr_set_handle(handle, inode,
1130 EXT4_XATTR_INDEX_ENCRYPTION,
1131 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1132 ctx, len, 0);
a7550b30
JK		 1133 if (!res) {
1134 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1135 ext4_clear_inode_state(inode,
1136 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b
JK		 1137 /*
1138 * Update inode->i_flags - e.g. S_DAX may get disabled
1139 */
1140 ext4_set_inode_flags(inode);
a7550b30
JK		 1141 }
1142 return res;
1143 }
1144
2f8f5e76 1145retry:
a7550b30
JK		 1146 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1147 ext4_jbd2_credits_xattr(inode));
1148 if (IS_ERR(handle))
1149 return PTR_ERR(handle);
1150
2f8f5e76
EB		 1151 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1152 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1153 ctx, len, 0);
a7550b30
JK		 1154 if (!res) {
1155 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
a3caa24b
JK		 1156 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1157 ext4_set_inode_flags(inode);
a7550b30
JK		 1158 res = ext4_mark_inode_dirty(handle, inode);
1159 if (res)
1160 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1161 }
1162 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1163
1164 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1165 goto retry;
a7550b30
JK		 1166 if (!res)
1167 res = res2;
1168 return res;
1169}
1170
1171static int ext4_dummy_context(struct inode *inode)
1172{
1173 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1174}
1175
1176static unsigned ext4_max_namelen(struct inode *inode)
1177{
1178 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1179 EXT4_NAME_LEN;
1180}
1181
1182static struct fscrypt_operations ext4_cryptops = {
1183 .get_context = ext4_get_context,
1184 .key_prefix = ext4_key_prefix,
1185 .prepare_context = ext4_prepare_context,
1186 .set_context = ext4_set_context,
1187 .dummy_context = ext4_dummy_context,
1188 .is_encrypted = ext4_encrypted_inode,
1189 .empty_dir = ext4_empty_dir,
1190 .max_namelen = ext4_max_namelen,
1191};
1192#else
1193static struct fscrypt_operations ext4_cryptops = {
1194 .is_encrypted = ext4_encrypted_inode,
1195};
1196#endif
1197
ac27a0ec 1198#ifdef CONFIG_QUOTA
689c958c
LX		 1199static char *quotatypes[] = INITQFNAMES;
1200#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1201
617ba13b
MC		 1202static int ext4_write_dquot(struct dquot *dquot);
1203static int ext4_acquire_dquot(struct dquot *dquot);
1204static int ext4_release_dquot(struct dquot *dquot);
1205static int ext4_mark_dquot_dirty(struct dquot *dquot);
1206static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1207static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1208 struct path *path);
ca0e05e4 1209static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1210static int ext4_quota_on_mount(struct super_block *sb, int type);
1211static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1212 size_t len, loff_t off);
617ba13b 1213static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1214 const char *data, size_t len, loff_t off);
7c319d32
AK		 1215static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1216 unsigned int flags);
1217static int ext4_enable_quotas(struct super_block *sb);
8f0e8746 1218static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1219
96c7e0d9
JK		 1220static struct dquot **ext4_get_dquots(struct inode *inode)
1221{
1222 return EXT4_I(inode)->i_dquot;
1223}
1224
61e225dc 1225static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1226 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1227 .write_dquot = ext4_write_dquot,
1228 .acquire_dquot = ext4_acquire_dquot,
1229 .release_dquot = ext4_release_dquot,
1230 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1231 .write_info = ext4_write_info,
1232 .alloc_dquot = dquot_alloc,
1233 .destroy_dquot = dquot_destroy,
040cb378 1234 .get_projid = ext4_get_projid,
8f0e8746 1235 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1236};
1237
0d54b217 1238static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1239 .quota_on = ext4_quota_on,
ca0e05e4 1240 .quota_off = ext4_quota_off,
287a8095 1241 .quota_sync = dquot_quota_sync,
0a240339 1242 .get_state = dquot_get_state,
287a8095
CH		 1243 .set_info = dquot_set_dqinfo,
1244 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1245 .set_dqblk = dquot_set_dqblk,
1246 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1247};
1248#endif
1249
ee9b6d61 1250static const struct super_operations ext4_sops = {
617ba13b
MC		 1251 .alloc_inode = ext4_alloc_inode,
1252 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1253 .write_inode = ext4_write_inode,
1254 .dirty_inode = ext4_dirty_inode,
7ff9c073 1255 .drop_inode = ext4_drop_inode,
0930fcc1 1256 .evict_inode = ext4_evict_inode,
617ba13b 1257 .put_super = ext4_put_super,
617ba13b 1258 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1259 .freeze_fs = ext4_freeze,
1260 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1261 .statfs = ext4_statfs,
1262 .remount_fs = ext4_remount,
617ba13b 1263 .show_options = ext4_show_options,
ac27a0ec 1264#ifdef CONFIG_QUOTA
617ba13b
MC		 1265 .quota_read = ext4_quota_read,
1266 .quota_write = ext4_quota_write,
96c7e0d9 1267 .get_dquots = ext4_get_dquots,
ac27a0ec 1268#endif
c39a7f84 1269 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1270};
1271
39655164 1272static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1273 .fh_to_dentry = ext4_fh_to_dentry,
1274 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1275 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1276};
1277
1278enum {
1279 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1280 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1281 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1282 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1283 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1284 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1285 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1286 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1287 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1288 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1289 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1290 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1291 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1449032b 1292 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1293 Opt_lazytime, Opt_nolazytime,
1449032b 1294 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1295 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1296 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1297 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1298 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1299};
1300
a447c093 1301static const match_table_t tokens = {
ac27a0ec
DK		 1302 {Opt_bsd_df, "bsddf"},
1303 {Opt_minix_df, "minixdf"},
1304 {Opt_grpid, "grpid"},
1305 {Opt_grpid, "bsdgroups"},
1306 {Opt_nogrpid, "nogrpid"},
1307 {Opt_nogrpid, "sysvgroups"},
1308 {Opt_resgid, "resgid=%u"},
1309 {Opt_resuid, "resuid=%u"},
1310 {Opt_sb, "sb=%u"},
1311 {Opt_err_cont, "errors=continue"},
1312 {Opt_err_panic, "errors=panic"},
1313 {Opt_err_ro, "errors=remount-ro"},
1314 {Opt_nouid32, "nouid32"},
ac27a0ec 1315 {Opt_debug, "debug"},
72578c33
TT		 1316 {Opt_removed, "oldalloc"},
1317 {Opt_removed, "orlov"},
ac27a0ec
DK		 1318 {Opt_user_xattr, "user_xattr"},
1319 {Opt_nouser_xattr, "nouser_xattr"},
1320 {Opt_acl, "acl"},
1321 {Opt_noacl, "noacl"},
e3bb52ae 1322 {Opt_noload, "norecovery"},
5a916be1 1323 {Opt_noload, "noload"},
72578c33
TT		 1324 {Opt_removed, "nobh"},
1325 {Opt_removed, "bh"},
ac27a0ec 1326 {Opt_commit, "commit=%u"},
30773840
TT		 1327 {Opt_min_batch_time, "min_batch_time=%u"},
1328 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1329 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1330 {Opt_journal_path, "journal_path=%s"},
818d276c 1331 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1332 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1333 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1334 {Opt_abort, "abort"},
1335 {Opt_data_journal, "data=journal"},
1336 {Opt_data_ordered, "data=ordered"},
1337 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1338 {Opt_data_err_abort, "data_err=abort"},
1339 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1340 {Opt_offusrjquota, "usrjquota="},
1341 {Opt_usrjquota, "usrjquota=%s"},
1342 {Opt_offgrpjquota, "grpjquota="},
1343 {Opt_grpjquota, "grpjquota=%s"},
1344 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1345 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1346 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1347 {Opt_grpquota, "grpquota"},
1348 {Opt_noquota, "noquota"},
1349 {Opt_quota, "quota"},
1350 {Opt_usrquota, "usrquota"},
49da9392 1351 {Opt_prjquota, "prjquota"},
ac27a0ec 1352 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1353 {Opt_barrier, "barrier"},
1354 {Opt_nobarrier, "nobarrier"},
25ec56b5 1355 {Opt_i_version, "i_version"},
923ae0ff 1356 {Opt_dax, "dax"},
c9de560d 1357 {Opt_stripe, "stripe=%u"},
64769240 1358 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1359 {Opt_lazytime, "lazytime"},
1360 {Opt_nolazytime, "nolazytime"},
dd919b98 1361 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1362 {Opt_removed, "mblk_io_submit"},
1363 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1364 {Opt_block_validity, "block_validity"},
1365 {Opt_noblock_validity, "noblock_validity"},
240799cd 1366 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1367 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1368 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1369 {Opt_auto_da_alloc, "auto_da_alloc"},
1370 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1371 {Opt_dioread_nolock, "dioread_nolock"},
1372 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1373 {Opt_discard, "discard"},
1374 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1375 {Opt_init_itable, "init_itable=%u"},
1376 {Opt_init_itable, "init_itable"},
1377 {Opt_noinit_itable, "noinit_itable"},
df981d03 1378 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1379 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1380 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1381 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1382 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1383 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1384 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1385 {Opt_err, NULL},
ac27a0ec
DK		 1386};
1387
617ba13b 1388static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1389{
617ba13b 1390 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1391 char *options = (char *) *data;
1392
1393 if (!options || strncmp(options, "sb=", 3) != 0)
1394 return 1; /* Default location */
0b8e58a1 1395
ac27a0ec 1396 options += 3;
0b8e58a1 1397 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1398 sb_block = simple_strtoul(options, &options, 0);
1399 if (*options && *options != ',') {
4776004f 1400 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1401 (char *) *data);
1402 return 1;
1403 }
1404 if (*options == ',')
1405 options++;
1406 *data = (void *) options;
0b8e58a1 1407
ac27a0ec
DK		 1408 return sb_block;
1409}
1410
b3881f74 1411#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1412static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1413 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1414
56c50f11
DM		 1415#ifdef CONFIG_QUOTA
1416static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1417{
1418 struct ext4_sb_info *sbi = EXT4_SB(sb);
1419 char *qname;
03dafb5f 1420 int ret = -1;
56c50f11
DM		 1421
1422 if (sb_any_quota_loaded(sb) &&
1423 !sbi->s_qf_names[qtype]) {
1424 ext4_msg(sb, KERN_ERR,
1425 "Cannot change journaled "
1426 "quota options when quota turned on");
57f73c2c 1427 return -1;
56c50f11 1428 }
e2b911c5 1429 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1430 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1431 "ignored when QUOTA feature is enabled");
1432 return 1;
262b4662 1433 }
56c50f11
DM		 1434 qname = match_strdup(args);
1435 if (!qname) {
1436 ext4_msg(sb, KERN_ERR,
1437 "Not enough memory for storing quotafile name");
57f73c2c 1438 return -1;
56c50f11 1439 }
03dafb5f
CG		 1440 if (sbi->s_qf_names[qtype]) {
1441 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1442 ret = 1;
1443 else
1444 ext4_msg(sb, KERN_ERR,
1445 "%s quota file already specified",
1446 QTYPE2NAME(qtype));
1447 goto errout;
56c50f11 1448 }
03dafb5f 1449 if (strchr(qname, '/')) {
56c50f11
DM		 1450 ext4_msg(sb, KERN_ERR,
1451 "quotafile must be on filesystem root");
03dafb5f 1452 goto errout;
56c50f11 1453 }
03dafb5f 1454 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1455 set_opt(sb, QUOTA);
56c50f11 1456 return 1;
03dafb5f
CG		 1457errout:
1458 kfree(qname);
1459 return ret;
56c50f11
DM		 1460}
1461
1462static int clear_qf_name(struct super_block *sb, int qtype)
1463{
1464
1465 struct ext4_sb_info *sbi = EXT4_SB(sb);
1466
1467 if (sb_any_quota_loaded(sb) &&
1468 sbi->s_qf_names[qtype]) {
1469 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1470 " when quota turned on");
57f73c2c 1471 return -1;
56c50f11 1472 }
03dafb5f 1473 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1474 sbi->s_qf_names[qtype] = NULL;
1475 return 1;
1476}
1477#endif
1478
26092bf5
TT		 1479#define MOPT_SET 0x0001
1480#define MOPT_CLEAR 0x0002
1481#define MOPT_NOSUPPORT 0x0004
1482#define MOPT_EXPLICIT 0x0008
1483#define MOPT_CLEAR_ERR 0x0010
1484#define MOPT_GTE0 0x0020
ac27a0ec 1485#ifdef CONFIG_QUOTA
26092bf5
TT		 1486#define MOPT_Q 0
1487#define MOPT_QFMT 0x0040
1488#else
1489#define MOPT_Q MOPT_NOSUPPORT
1490#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1491#endif
26092bf5 1492#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1493#define MOPT_NO_EXT2 0x0100
1494#define MOPT_NO_EXT3 0x0200
1495#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1496#define MOPT_STRING 0x0400
26092bf5
TT		 1497
1498static const struct mount_opts {
1499 int token;
1500 int mount_opt;
1501 int flags;
1502} ext4_mount_opts[] = {
1503 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1504 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1505 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1506 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1507 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1508 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1509 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1510 MOPT_EXT4_ONLY | MOPT_SET},
1511 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1512 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1513 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1514 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1515 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1516 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1517 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1518 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1519 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1520 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1521 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1522 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1523 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1524 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1525 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1526 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1527 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1528 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1529 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1530 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1531 MOPT_NO_EXT2},
8dc0aa8c 1532 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1533 MOPT_NO_EXT2},
26092bf5
TT		 1534 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1535 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1536 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1537 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1538 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1539 {Opt_commit, 0, MOPT_GTE0},
1540 {Opt_max_batch_time, 0, MOPT_GTE0},
1541 {Opt_min_batch_time, 0, MOPT_GTE0},
1542 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1543 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1544 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1545 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1546 {Opt_resuid, 0, MOPT_GTE0},
1547 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1548 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1549 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1550 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1551 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1552 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1553 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1554 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1555 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1556 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1557#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1558 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1559 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1560#else
26092bf5
TT		 1561 {Opt_acl, 0, MOPT_NOSUPPORT},
1562 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1563#endif
26092bf5
TT		 1564 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1565 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1566 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1567 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1568 MOPT_SET | MOPT_Q},
1569 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1570 MOPT_SET | MOPT_Q},
49da9392
JK		 1571 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1572 MOPT_SET | MOPT_Q},
26092bf5 1573 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1574 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1575 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1576 {Opt_usrjquota, 0, MOPT_Q},
1577 {Opt_grpjquota, 0, MOPT_Q},
1578 {Opt_offusrjquota, 0, MOPT_Q},
1579 {Opt_offgrpjquota, 0, MOPT_Q},
1580 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1581 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1582 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1583 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1584 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1585 {Opt_err, 0, 0}
1586};
1587
1588static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1589 substring_t *args, unsigned long *journal_devnum,
1590 unsigned int *journal_ioprio, int is_remount)
1591{
1592 struct ext4_sb_info *sbi = EXT4_SB(sb);
1593 const struct mount_opts *m;
08cefc7a
EB		 1594 kuid_t uid;
1595 kgid_t gid;
26092bf5
TT		 1596 int arg = 0;
1597
57f73c2c
TT		 1598#ifdef CONFIG_QUOTA
1599 if (token == Opt_usrjquota)
1600 return set_qf_name(sb, USRQUOTA, &args[0]);
1601 else if (token == Opt_grpjquota)
1602 return set_qf_name(sb, GRPQUOTA, &args[0]);
1603 else if (token == Opt_offusrjquota)
1604 return clear_qf_name(sb, USRQUOTA);
1605 else if (token == Opt_offgrpjquota)
1606 return clear_qf_name(sb, GRPQUOTA);
1607#endif
26092bf5 1608 switch (token) {
f7048605
TT		 1609 case Opt_noacl:
1610 case Opt_nouser_xattr:
1611 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1612 break;
26092bf5
TT		 1613 case Opt_sb:
1614 return 1; /* handled by get_sb_block() */
1615 case Opt_removed:
5f3633e3 1616 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1617 return 1;
26092bf5
TT		 1618 case Opt_abort:
1619 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1620 return 1;
1621 case Opt_i_version:
1622 sb->s_flags |= MS_I_VERSION;
1623 return 1;
a26f4992
TT		 1624 case Opt_lazytime:
1625 sb->s_flags |= MS_LAZYTIME;
1626 return 1;
1627 case Opt_nolazytime:
1628 sb->s_flags &= ~MS_LAZYTIME;
1629 return 1;
26092bf5
TT		 1630 }
1631
5f3633e3
JK		 1632 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1633 if (token == m->token)
1634 break;
1635
1636 if (m->token == Opt_err) {
1637 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1638 "or missing value", opt);
1639 return -1;
1640 }
1641
8dc0aa8c
TT		 1642 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1643 ext4_msg(sb, KERN_ERR,
1644 "Mount option \"%s\" incompatible with ext2", opt);
1645 return -1;
1646 }
1647 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1648 ext4_msg(sb, KERN_ERR,
1649 "Mount option \"%s\" incompatible with ext3", opt);
1650 return -1;
1651 }
1652
ad4eec61 1653 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1654 return -1;
1655 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1656 return -1;
c93cf2d7
DM		 1657 if (m->flags & MOPT_EXPLICIT) {
1658 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1659 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1660 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1661 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1662 } else
1663 return -1;
1664 }
5f3633e3
JK		 1665 if (m->flags & MOPT_CLEAR_ERR)
1666 clear_opt(sb, ERRORS_MASK);
1667 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1668 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1669 "options when quota turned on");
1670 return -1;
1671 }
1672
1673 if (m->flags & MOPT_NOSUPPORT) {
1674 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1675 } else if (token == Opt_commit) {
1676 if (arg == 0)
1677 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1678 sbi->s_commit_interval = HZ * arg;
1679 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1680 sbi->s_max_batch_time = arg;
1681 } else if (token == Opt_min_batch_time) {
1682 sbi->s_min_batch_time = arg;
1683 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1684 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1685 ext4_msg(sb, KERN_ERR,
1686 "EXT4-fs: inode_readahead_blks must be "
1687 "0 or a power of 2 smaller than 2^31");
26092bf5 1688 return -1;
5f3633e3
JK		 1689 }
1690 sbi->s_inode_readahead_blks = arg;
1691 } else if (token == Opt_init_itable) {
1692 set_opt(sb, INIT_INODE_TABLE);
1693 if (!args->from)
1694 arg = EXT4_DEF_LI_WAIT_MULT;
1695 sbi->s_li_wait_mult = arg;
1696 } else if (token == Opt_max_dir_size_kb) {
1697 sbi->s_max_dir_size_kb = arg;
1698 } else if (token == Opt_stripe) {
1699 sbi->s_stripe = arg;
1700 } else if (token == Opt_resuid) {
1701 uid = make_kuid(current_user_ns(), arg);
1702 if (!uid_valid(uid)) {
1703 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1704 return -1;
1705 }
5f3633e3
JK		 1706 sbi->s_resuid = uid;
1707 } else if (token == Opt_resgid) {
1708 gid = make_kgid(current_user_ns(), arg);
1709 if (!gid_valid(gid)) {
1710 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1711 return -1;
1712 }
1713 sbi->s_resgid = gid;
1714 } else if (token == Opt_journal_dev) {
1715 if (is_remount) {
1716 ext4_msg(sb, KERN_ERR,
1717 "Cannot specify journal on remount");
1718 return -1;
1719 }
1720 *journal_devnum = arg;
ad4eec61
ES		 1721 } else if (token == Opt_journal_path) {
1722 char *journal_path;
1723 struct inode *journal_inode;
1724 struct path path;
1725 int error;
1726
1727 if (is_remount) {
1728 ext4_msg(sb, KERN_ERR,
1729 "Cannot specify journal on remount");
1730 return -1;
1731 }
1732 journal_path = match_strdup(&args[0]);
1733 if (!journal_path) {
1734 ext4_msg(sb, KERN_ERR, "error: could not dup "
1735 "journal device string");
1736 return -1;
1737 }
1738
1739 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1740 if (error) {
1741 ext4_msg(sb, KERN_ERR, "error: could not find "
1742 "journal device path: error %d", error);
1743 kfree(journal_path);
1744 return -1;
1745 }
1746
2b0143b5 1747 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1748 if (!S_ISBLK(journal_inode->i_mode)) {
1749 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1750 "is not a block device", journal_path);
1751 path_put(&path);
1752 kfree(journal_path);
1753 return -1;
1754 }
1755
1756 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1757 path_put(&path);
1758 kfree(journal_path);
5f3633e3
JK		 1759 } else if (token == Opt_journal_ioprio) {
1760 if (arg > 7) {
1761 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1762 " (must be 0-7)");
1763 return -1;
1764 }
1765 *journal_ioprio =
1766 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1767 } else if (token == Opt_test_dummy_encryption) {
1768#ifdef CONFIG_EXT4_FS_ENCRYPTION
1769 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1770 ext4_msg(sb, KERN_WARNING,
1771 "Test dummy encryption mode enabled");
1772#else
1773 ext4_msg(sb, KERN_WARNING,
1774 "Test dummy encryption mount option ignored");
1775#endif
5f3633e3
JK		 1776 } else if (m->flags & MOPT_DATAJ) {
1777 if (is_remount) {
1778 if (!sbi->s_journal)
1779 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1780 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1781 ext4_msg(sb, KERN_ERR,
26092bf5 1782 "Cannot change data mode on remount");
26092bf5 1783 return -1;
ac27a0ec 1784 }
26092bf5 1785 } else {
5f3633e3
JK		 1786 clear_opt(sb, DATA_FLAGS);
1787 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1788 }
5f3633e3
JK		 1789#ifdef CONFIG_QUOTA
1790 } else if (m->flags & MOPT_QFMT) {
1791 if (sb_any_quota_loaded(sb) &&
1792 sbi->s_jquota_fmt != m->mount_opt) {
1793 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1794 "quota options when quota turned on");
1795 return -1;
1796 }
e2b911c5 1797 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1798 ext4_msg(sb, KERN_INFO,
1799 "Quota format mount options ignored "
262b4662 1800 "when QUOTA feature is enabled");
c325a67c 1801 return 1;
262b4662 1802 }
5f3633e3 1803 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1804#endif
923ae0ff 1805 } else if (token == Opt_dax) {
ef83b6e8
DW		 1806#ifdef CONFIG_FS_DAX
1807 ext4_msg(sb, KERN_WARNING,
1808 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1809 sbi->s_mount_opt |= m->mount_opt;
1810#else
923ae0ff
RZ		 1811 ext4_msg(sb, KERN_INFO, "dax option not supported");
1812 return -1;
5f3633e3 1813#endif
7915a861
AN		 1814 } else if (token == Opt_data_err_abort) {
1815 sbi->s_mount_opt |= m->mount_opt;
1816 } else if (token == Opt_data_err_ignore) {
1817 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1818 } else {
1819 if (!args->from)
1820 arg = 1;
1821 if (m->flags & MOPT_CLEAR)
1822 arg = !arg;
1823 else if (unlikely(!(m->flags & MOPT_SET))) {
1824 ext4_msg(sb, KERN_WARNING,
1825 "buggy handling of option %s", opt);
1826 WARN_ON(1);
1827 return -1;
1828 }
1829 if (arg != 0)
1830 sbi->s_mount_opt |= m->mount_opt;
1831 else
1832 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1833 }
5f3633e3 1834 return 1;
26092bf5
TT		 1835}
1836
1837static int parse_options(char *options, struct super_block *sb,
1838 unsigned long *journal_devnum,
1839 unsigned int *journal_ioprio,
1840 int is_remount)
1841{
1842 struct ext4_sb_info *sbi = EXT4_SB(sb);
1843 char *p;
1844 substring_t args[MAX_OPT_ARGS];
1845 int token;
1846
1847 if (!options)
1848 return 1;
1849
1850 while ((p = strsep(&options, ",")) != NULL) {
1851 if (!*p)
1852 continue;
1853 /*
1854 * Initialize args struct so we know whether arg was
1855 * found; some options take optional arguments.
1856 */
caecd0af 1857 args[0].to = args[0].from = NULL;
26092bf5
TT		 1858 token = match_token(p, tokens, args);
1859 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1860 journal_ioprio, is_remount) < 0)
1861 return 0;
ac27a0ec
DK		 1862 }
1863#ifdef CONFIG_QUOTA
49da9392
JK		 1864 /*
1865 * We do the test below only for project quotas. 'usrquota' and
1866 * 'grpquota' mount options are allowed even without quota feature
1867 * to support legacy quotas in quota files.
1868 */
1869 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1870 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1871 "Cannot enable project quota enforcement.");
1872 return 0;
1873 }
1874 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1875 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1876 clear_opt(sb, USRQUOTA);
ac27a0ec 1877
482a7425 1878 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1879 clear_opt(sb, GRPQUOTA);
ac27a0ec 1880
56c50f11 1881 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1882 ext4_msg(sb, KERN_ERR, "old and new quota "
1883 "format mixing");
ac27a0ec
DK		 1884 return 0;
1885 }
1886
1887 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1888 ext4_msg(sb, KERN_ERR, "journaled quota format "
1889 "not specified");
ac27a0ec
DK		 1890 return 0;
1891 }
ac27a0ec
DK		 1892 }
1893#endif
261cb20c
JK		 1894 if (test_opt(sb, DIOREAD_NOLOCK)) {
1895 int blocksize =
1896 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1897
09cbfeaf 1898 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 1899 ext4_msg(sb, KERN_ERR, "can't mount with "
1900 "dioread_nolock if block size != PAGE_SIZE");
1901 return 0;
1902 }
1903 }
d4f76107
JK		 1904 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1905 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1906 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1907 "in data=ordered mode");
1908 return 0;
1909 }
ac27a0ec
DK		 1910 return 1;
1911}
1912
2adf6da8
TT		 1913static inline void ext4_show_quota_options(struct seq_file *seq,
1914 struct super_block *sb)
1915{
1916#if defined(CONFIG_QUOTA)
1917 struct ext4_sb_info *sbi = EXT4_SB(sb);
1918
1919 if (sbi->s_jquota_fmt) {
1920 char *fmtname = "";
1921
1922 switch (sbi->s_jquota_fmt) {
1923 case QFMT_VFS_OLD:
1924 fmtname = "vfsold";
1925 break;
1926 case QFMT_VFS_V0:
1927 fmtname = "vfsv0";
1928 break;
1929 case QFMT_VFS_V1:
1930 fmtname = "vfsv1";
1931 break;
1932 }
1933 seq_printf(seq, ",jqfmt=%s", fmtname);
1934 }
1935
1936 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1937 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1938
1939 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1940 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1941#endif
1942}
1943
5a916be1
TT		 1944static const char *token2str(int token)
1945{
50df9fd5 1946 const struct match_token *t;
5a916be1
TT		 1947
1948 for (t = tokens; t->token != Opt_err; t++)
1949 if (t->token == token && !strchr(t->pattern, '='))
1950 break;
1951 return t->pattern;
1952}
1953
2adf6da8
TT		 1954/*
1955 * Show an option if
1956 * - it's set to a non-default value OR
1957 * - if the per-sb default is different from the global default
1958 */
66acdcf4
TT		 1959static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1960 int nodefs)
2adf6da8 1961{
2adf6da8
TT		 1962 struct ext4_sb_info *sbi = EXT4_SB(sb);
1963 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1964 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1965 const struct mount_opts *m;
66acdcf4 1966 char sep = nodefs ? '\n' : ',';
2adf6da8 1967
66acdcf4
TT		 1968#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1969#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1970
1971 if (sbi->s_sb_block != 1)
5a916be1
TT		 1972 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1973
1974 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1975 int want_set = m->flags & MOPT_SET;
1976 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1977 (m->flags & MOPT_CLEAR_ERR))
1978 continue;
66acdcf4 1979 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1980 continue; /* skip if same as the default */
1981 if ((want_set &&
1982 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1983 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1984 continue; /* select Opt_noFoo vs Opt_Foo */
1985 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1986 }
5a916be1 1987
08cefc7a 1988 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1989 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1990 SEQ_OPTS_PRINT("resuid=%u",
1991 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1992 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1993 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1994 SEQ_OPTS_PRINT("resgid=%u",
1995 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1996 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1997 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1998 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1999 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2000 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2001 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2002 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2003 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2004 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2005 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2006 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2007 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2008 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 2009 if (sb->s_flags & MS_I_VERSION)
5a916be1 2010 SEQ_OPTS_PUTS("i_version");
66acdcf4 2011 if (nodefs || sbi->s_stripe)
5a916be1 2012 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 2013 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2014 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2015 SEQ_OPTS_PUTS("data=journal");
2016 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2017 SEQ_OPTS_PUTS("data=ordered");
2018 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2019 SEQ_OPTS_PUTS("data=writeback");
2020 }
66acdcf4
TT		 2021 if (nodefs ||
2022 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2023 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2024 sbi->s_inode_readahead_blks);
2adf6da8 2025
66acdcf4
TT		 2026 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2027 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2028 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2029 if (nodefs || sbi->s_max_dir_size_kb)
2030 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2031 if (test_opt(sb, DATA_ERR_ABORT))
2032 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 2033
2034 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2035 return 0;
2036}
2037
66acdcf4
TT		 2038static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2039{
2040 return _ext4_show_options(seq, root->d_sb, 0);
2041}
2042
ebd173be 2043int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2044{
2045 struct super_block *sb = seq->private;
2046 int rc;
2047
2048 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2049 rc = _ext4_show_options(seq, sb, 1);
2050 seq_puts(seq, "\n");
2051 return rc;
2052}
2053
617ba13b 2054static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2055 int read_only)
2056{
617ba13b 2057 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2058 int res = 0;
2059
617ba13b 2060 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2061 ext4_msg(sb, KERN_ERR, "revision level too high, "
2062 "forcing read-only mode");
ac27a0ec
DK		 2063 res = MS_RDONLY;
2064 }
2065 if (read_only)
281b5995 2066 goto done;
617ba13b 2067 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2068 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2069 "running e2fsck is recommended");
c8b459f4 2070 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2071 ext4_msg(sb, KERN_WARNING,
2072 "warning: mounting fs with errors, "
2073 "running e2fsck is recommended");
ed3ce80a 2074 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2075 le16_to_cpu(es->s_mnt_count) >=
2076 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2077 ext4_msg(sb, KERN_WARNING,
2078 "warning: maximal mount count reached, "
2079 "running e2fsck is recommended");
ac27a0ec
DK		 2080 else if (le32_to_cpu(es->s_checkinterval) &&
2081 (le32_to_cpu(es->s_lastcheck) +
2082 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 2083 ext4_msg(sb, KERN_WARNING,
2084 "warning: checktime reached, "
2085 "running e2fsck is recommended");
0b8e58a1 2086 if (!sbi->s_journal)
0390131b 2087 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2088 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2089 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2090 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 2091 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 2092 ext4_update_dynamic_rev(sb);
0390131b 2093 if (sbi->s_journal)
e2b911c5 2094 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2095
e2d67052 2096 ext4_commit_super(sb, 1);
281b5995 2097done:
ac27a0ec 2098 if (test_opt(sb, DEBUG))
a9df9a49 2099 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2100 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2101 sb->s_blocksize,
2102 sbi->s_groups_count,
617ba13b
MC		 2103 EXT4_BLOCKS_PER_GROUP(sb),
2104 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2105 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2106
7abc52c2 2107 cleancache_init_fs(sb);
ac27a0ec
DK		 2108 return res;
2109}
2110
117fff10
TT		 2111int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2112{
2113 struct ext4_sb_info *sbi = EXT4_SB(sb);
2114 struct flex_groups *new_groups;
2115 int size;
2116
2117 if (!sbi->s_log_groups_per_flex)
2118 return 0;
2119
2120 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2121 if (size <= sbi->s_flex_groups_allocated)
2122 return 0;
2123
2124 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2125 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2126 if (!new_groups) {
2127 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2128 size / (int) sizeof(struct flex_groups));
2129 return -ENOMEM;
2130 }
2131
2132 if (sbi->s_flex_groups) {
2133 memcpy(new_groups, sbi->s_flex_groups,
2134 (sbi->s_flex_groups_allocated *
2135 sizeof(struct flex_groups)));
b93b41d4 2136 kvfree(sbi->s_flex_groups);
117fff10
TT		 2137 }
2138 sbi->s_flex_groups = new_groups;
2139 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2140 return 0;
2141}
2142
772cb7c8
JS		 2143static int ext4_fill_flex_info(struct super_block *sb)
2144{
2145 struct ext4_sb_info *sbi = EXT4_SB(sb);
2146 struct ext4_group_desc *gdp = NULL;
772cb7c8 2147 ext4_group_t flex_group;
117fff10 2148 int i, err;
772cb7c8 2149
503358ae 2150 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2151 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2152 sbi->s_log_groups_per_flex = 0;
2153 return 1;
2154 }
2155
117fff10
TT		 2156 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2157 if (err)
9933fc0a 2158 goto failed;
772cb7c8 2159
772cb7c8 2160 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2161 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2162
2163 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2164 atomic_add(ext4_free_inodes_count(sb, gdp),
2165 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2166 atomic64_add(ext4_free_group_clusters(sb, gdp),
2167 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2168 atomic_add(ext4_used_dirs_count(sb, gdp),
2169 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2170 }
2171
2172 return 1;
2173failed:
2174 return 0;
2175}
2176
e2b911c5 2177static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2178 struct ext4_group_desc *gdp)
717d50e4 2179{
b47820ed 2180 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2181 __u16 crc = 0;
feb0ab32 2182 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2183 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2184
9aa5d32b 2185 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2186 /* Use new metadata_csum algorithm */
feb0ab32 2187 __u32 csum32;
b47820ed 2188 __u16 dummy_csum = 0;
feb0ab32 2189
feb0ab32
DW		 2190 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2191 sizeof(le_group));
b47820ed
DJ		 2192 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2193 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2194 sizeof(dummy_csum));
2195 offset += sizeof(dummy_csum);
2196 if (offset < sbi->s_desc_size)
2197 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2198 sbi->s_desc_size - offset);
feb0ab32
DW		 2199
2200 crc = csum32 & 0xFFFF;
2201 goto out;
717d50e4
AD		 2202 }
2203
feb0ab32 2204 /* old crc16 code */
e2b911c5 2205 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2206 return 0;
2207
feb0ab32
DW		 2208 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2209 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2210 crc = crc16(crc, (__u8 *)gdp, offset);
2211 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2212 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2213 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2214 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2215 crc = crc16(crc, (__u8 *)gdp + offset,
2216 le16_to_cpu(sbi->s_es->s_desc_size) -
2217 offset);
2218
2219out:
717d50e4
AD		 2220 return cpu_to_le16(crc);
2221}
2222
feb0ab32 2223int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2224 struct ext4_group_desc *gdp)
2225{
feb0ab32 2226 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2227 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2228 return 0;
2229
2230 return 1;
2231}
2232
feb0ab32
DW		 2233void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2234 struct ext4_group_desc *gdp)
2235{
2236 if (!ext4_has_group_desc_csum(sb))
2237 return;
e2b911c5 2238 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2239}
2240
ac27a0ec 2241/* Called at mount-time, super-block is locked */
bfff6873 2242static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2243 ext4_fsblk_t sb_block,
bfff6873 2244 ext4_group_t *first_not_zeroed)
ac27a0ec 2245{
617ba13b
MC		 2246 struct ext4_sb_info *sbi = EXT4_SB(sb);
2247 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2248 ext4_fsblk_t last_block;
bd81d8ee
LV		 2249 ext4_fsblk_t block_bitmap;
2250 ext4_fsblk_t inode_bitmap;
2251 ext4_fsblk_t inode_table;
ce421581 2252 int flexbg_flag = 0;
bfff6873 2253 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2254
e2b911c5 2255 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2256 flexbg_flag = 1;
2257
af5bc92d 2258 ext4_debug("Checking group descriptors");
ac27a0ec 2259
197cd65a
AM		 2260 for (i = 0; i < sbi->s_groups_count; i++) {
2261 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2262
ce421581 2263 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2264 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2265 else
2266 last_block = first_block +
617ba13b 2267 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2268
bfff6873
LC		 2269 if ((grp == sbi->s_groups_count) &&
2270 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2271 grp = i;
2272
8fadc143 2273 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2274 if (block_bitmap == sb_block) {
2275 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2276 "Block bitmap for group %u overlaps "
2277 "superblock", i);
2278 }
2b2d6d01 2279 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2280 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2281 "Block bitmap for group %u not in group "
b31e1552 2282 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2283 return 0;
2284 }
8fadc143 2285 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2286 if (inode_bitmap == sb_block) {
2287 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2288 "Inode bitmap for group %u overlaps "
2289 "superblock", i);
2290 }
2b2d6d01 2291 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2292 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2293 "Inode bitmap for group %u not in group "
b31e1552 2294 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2295 return 0;
2296 }
8fadc143 2297 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2298 if (inode_table == sb_block) {
2299 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2300 "Inode table for group %u overlaps "
2301 "superblock", i);
2302 }
bd81d8ee 2303 if (inode_table < first_block ||
2b2d6d01 2304 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2305 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2306 "Inode table for group %u not in group "
b31e1552 2307 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2308 return 0;
2309 }
955ce5f5 2310 ext4_lock_group(sb, i);
feb0ab32 2311 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2312 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2313 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2314 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2315 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2316 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2317 ext4_unlock_group(sb, i);
8a266467 2318 return 0;
7ee1ec4c 2319 }
717d50e4 2320 }
955ce5f5 2321 ext4_unlock_group(sb, i);
ce421581
JS		 2322 if (!flexbg_flag)
2323 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2324 }
bfff6873
LC		 2325 if (NULL != first_not_zeroed)
2326 *first_not_zeroed = grp;
ac27a0ec
DK		 2327 return 1;
2328}
2329
617ba13b 2330/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2331 * the superblock) which were deleted from all directories, but held open by
2332 * a process at the time of a crash. We walk the list and try to delete these
2333 * inodes at recovery time (only with a read-write filesystem).
2334 *
2335 * In order to keep the orphan inode chain consistent during traversal (in
2336 * case of crash during recovery), we link each inode into the superblock
2337 * orphan list_head and handle it the same way as an inode deletion during
2338 * normal operation (which journals the operations for us).
2339 *
2340 * We only do an iget() and an iput() on each inode, which is very safe if we
2341 * accidentally point at an in-use or already deleted inode. The worst that
2342 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2343 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2344 * e2fsck was run on this filesystem, and it must have already done the orphan
2345 * inode cleanup for us, so we can safely abort without any further action.
2346 */
2b2d6d01
TT		 2347static void ext4_orphan_cleanup(struct super_block *sb,
2348 struct ext4_super_block *es)
ac27a0ec
DK		 2349{
2350 unsigned int s_flags = sb->s_flags;
2c98eb5e 2351 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec
DK		 2352#ifdef CONFIG_QUOTA
2353 int i;
2354#endif
2355 if (!es->s_last_orphan) {
2356 jbd_debug(4, "no orphan inodes to clean up\n");
2357 return;
2358 }
2359
a8f48a95 2360 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2361 ext4_msg(sb, KERN_ERR, "write access "
2362 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2363 return;
2364 }
2365
d39195c3
AG		 2366 /* Check if feature set would not allow a r/w mount */
2367 if (!ext4_feature_set_ok(sb, 0)) {
2368 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2369 "unknown ROCOMPAT features");
2370 return;
2371 }
2372
617ba13b 2373 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2374 /* don't clear list on RO mount w/ errors */
2375 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2376 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2377 "clearing orphan list.\n");
c25f9bc6
ES		 2378 es->s_last_orphan = 0;
2379 }
ac27a0ec
DK		 2380 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2381 return;
2382 }
2383
2384 if (s_flags & MS_RDONLY) {
b31e1552 2385 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2386 sb->s_flags &= ~MS_RDONLY;
2387 }
2388#ifdef CONFIG_QUOTA
2389 /* Needed for iput() to work correctly and not trash data */
2390 sb->s_flags |= MS_ACTIVE;
2391 /* Turn on quotas so that they are updated correctly */
a2d4a646 2392 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2393 if (EXT4_SB(sb)->s_qf_names[i]) {
2394 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2395 if (ret < 0)
b31e1552
ES		 2396 ext4_msg(sb, KERN_ERR,
2397 "Cannot turn on journaled "
2398 "quota: error %d", ret);
ac27a0ec
DK		 2399 }
2400 }
2401#endif
2402
2403 while (es->s_last_orphan) {
2404 struct inode *inode;
2405
c65d5c6c
VN		 2406 /*
2407 * We may have encountered an error during cleanup; if
2408 * so, skip the rest.
2409 */
2410 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2411 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2412 es->s_last_orphan = 0;
2413 break;
2414 }
2415
97bd42b9
JB		 2416 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2417 if (IS_ERR(inode)) {
ac27a0ec
DK		 2418 es->s_last_orphan = 0;
2419 break;
2420 }
2421
617ba13b 2422 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2423 dquot_initialize(inode);
ac27a0ec 2424 if (inode->i_nlink) {
566370a2
PT		 2425 if (test_opt(sb, DEBUG))
2426 ext4_msg(sb, KERN_DEBUG,
2427 "%s: truncating inode %lu to %lld bytes",
2428 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2429 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2430 inode->i_ino, inode->i_size);
5955102c 2431 inode_lock(inode);
55f252c9 2432 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2433 ret = ext4_truncate(inode);
2434 if (ret)
2435 ext4_std_error(inode->i_sb, ret);
5955102c 2436 inode_unlock(inode);
ac27a0ec
DK		 2437 nr_truncates++;
2438 } else {
566370a2
PT		 2439 if (test_opt(sb, DEBUG))
2440 ext4_msg(sb, KERN_DEBUG,
2441 "%s: deleting unreferenced inode %lu",
2442 __func__, inode->i_ino);
ac27a0ec
DK		 2443 jbd_debug(2, "deleting unreferenced inode %lu\n",
2444 inode->i_ino);
2445 nr_orphans++;
2446 }
2447 iput(inode); /* The delete magic happens here! */
2448 }
2449
2b2d6d01 2450#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2451
2452 if (nr_orphans)
b31e1552
ES		 2453 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2454 PLURAL(nr_orphans));
ac27a0ec 2455 if (nr_truncates)
b31e1552
ES		 2456 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2457 PLURAL(nr_truncates));
ac27a0ec
DK		 2458#ifdef CONFIG_QUOTA
2459 /* Turn quotas off */
a2d4a646 2460 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2461 if (sb_dqopt(sb)->files[i])
287a8095 2462 dquot_quota_off(sb, i);
ac27a0ec
DK		 2463 }
2464#endif
2465 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2466}
0b8e58a1 2467
cd2291a4
ES		 2468/*
2469 * Maximal extent format file size.
2470 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2471 * extent format containers, within a sector_t, and within i_blocks
2472 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2473 * so that won't be a limiting factor.
2474 *
f17722f9
LC		 2475 * However there is other limiting factor. We do store extents in the form
2476 * of starting block and length, hence the resulting length of the extent
2477 * covering maximum file size must fit into on-disk format containers as
2478 * well. Given that length is always by 1 unit bigger than max unit (because
2479 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2480 *
cd2291a4
ES		 2481 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2482 */
f287a1a5 2483static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2484{
2485 loff_t res;
2486 loff_t upper_limit = MAX_LFS_FILESIZE;
2487
2488 /* small i_blocks in vfs inode? */
f287a1a5 2489 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2490 /*
90c699a9 2491 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2492 * i_block represent total blocks in 512 bytes
2493 * 32 == size of vfs inode i_blocks * 8
2494 */
2495 upper_limit = (1LL << 32) - 1;
2496
2497 /* total blocks in file system block size */
2498 upper_limit >>= (blkbits - 9);
2499 upper_limit <<= blkbits;
2500 }
2501
f17722f9
LC		 2502 /*
2503 * 32-bit extent-start container, ee_block. We lower the maxbytes
2504 * by one fs block, so ee_len can cover the extent of maximum file
2505 * size
2506 */
2507 res = (1LL << 32) - 1;
cd2291a4 2508 res <<= blkbits;
cd2291a4
ES		 2509
2510 /* Sanity check against vm- & vfs- imposed limits */
2511 if (res > upper_limit)
2512 res = upper_limit;
2513
2514 return res;
2515}
ac27a0ec 2516
ac27a0ec 2517/*
cd2291a4 2518 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2519 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2520 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2521 */
f287a1a5 2522static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2523{
617ba13b 2524 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2525 int meta_blocks;
2526 loff_t upper_limit;
0b8e58a1
AD		 2527 /* This is calculated to be the largest file size for a dense, block
2528 * mapped file such that the file's total number of 512-byte sectors,
2529 * including data and all indirect blocks, does not exceed (2^48 - 1).
2530 *
2531 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2532 * number of 512-byte sectors of the file.
0fc1b451
AK		 2533 */
2534
f287a1a5 2535 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2536 /*
90c699a9 2537 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2538 * the inode i_block field represents total file blocks in
2539 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2540 */
2541 upper_limit = (1LL << 32) - 1;
2542
2543 /* total blocks in file system block size */
2544 upper_limit >>= (bits - 9);
2545
2546 } else {
8180a562
AK		 2547 /*
2548 * We use 48 bit ext4_inode i_blocks
2549 * With EXT4_HUGE_FILE_FL set the i_blocks
2550 * represent total number of blocks in
2551 * file system block size
2552 */
0fc1b451
AK		 2553 upper_limit = (1LL << 48) - 1;
2554
0fc1b451
AK		 2555 }
2556
2557 /* indirect blocks */
2558 meta_blocks = 1;
2559 /* double indirect blocks */
2560 meta_blocks += 1 + (1LL << (bits-2));
2561 /* tripple indirect blocks */
2562 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2563
2564 upper_limit -= meta_blocks;
2565 upper_limit <<= bits;
ac27a0ec
DK		 2566
2567 res += 1LL << (bits-2);
2568 res += 1LL << (2*(bits-2));
2569 res += 1LL << (3*(bits-2));
2570 res <<= bits;
2571 if (res > upper_limit)
2572 res = upper_limit;
0fc1b451
AK		 2573
2574 if (res > MAX_LFS_FILESIZE)
2575 res = MAX_LFS_FILESIZE;
2576
ac27a0ec
DK		 2577 return res;
2578}
2579
617ba13b 2580static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2581 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2582{
617ba13b 2583 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2584 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2585 int has_super = 0;
2586
2587 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2588
e2b911c5 2589 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2590 return logical_sb_block + nr + 1;
ac27a0ec 2591 bg = sbi->s_desc_per_block * nr;
617ba13b 2592 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2593 has_super = 1;
0b8e58a1 2594
bd63f6b0
DW		 2595 /*
2596 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2597 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2598 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2599 * compensate.
2600 */
2601 if (sb->s_blocksize == 1024 && nr == 0 &&
2602 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2603 has_super++;
2604
617ba13b 2605 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2606}
2607
c9de560d
AT		 2608/**
2609 * ext4_get_stripe_size: Get the stripe size.
2610 * @sbi: In memory super block info
2611 *
2612 * If we have specified it via mount option, then
2613 * use the mount option value. If the value specified at mount time is
2614 * greater than the blocks per group use the super block value.
2615 * If the super block value is greater than blocks per group return 0.
2616 * Allocator needs it be less than blocks per group.
2617 *
2618 */
2619static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2620{
2621 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2622 unsigned long stripe_width =
2623 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2624 int ret;
c9de560d
AT		 2625
2626 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2627 ret = sbi->s_stripe;
2628 else if (stripe_width <= sbi->s_blocks_per_group)
2629 ret = stripe_width;
2630 else if (stride <= sbi->s_blocks_per_group)
2631 ret = stride;
2632 else
2633 ret = 0;
c9de560d 2634
3eb08658
DE		 2635 /*
2636 * If the stripe width is 1, this makes no sense and
2637 * we set it to 0 to turn off stripe handling code.
2638 */
2639 if (ret <= 1)
2640 ret = 0;
c9de560d 2641
3eb08658 2642 return ret;
c9de560d 2643}
ac27a0ec 2644
a13fb1a4
ES		 2645/*
2646 * Check whether this filesystem can be mounted based on
2647 * the features present and the RDONLY/RDWR mount requested.
2648 * Returns 1 if this filesystem can be mounted as requested,
2649 * 0 if it cannot be.
2650 */
2651static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2652{
e2b911c5 2653 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2654 ext4_msg(sb, KERN_ERR,
2655 "Couldn't mount because of "
2656 "unsupported optional features (%x)",
2657 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2658 ~EXT4_FEATURE_INCOMPAT_SUPP));
2659 return 0;
2660 }
2661
2662 if (readonly)
2663 return 1;
2664
e2b911c5 2665 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b
DW		 2666 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2667 sb->s_flags |= MS_RDONLY;
2668 return 1;
2669 }
2670
a13fb1a4 2671 /* Check that feature set is OK for a read-write mount */
e2b911c5 2672 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2673 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2674 "unsupported optional features (%x)",
2675 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2676 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2677 return 0;
2678 }
2679 /*
2680 * Large file size enabled file system can only be mounted
2681 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2682 */
e2b911c5 2683 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2684 if (sizeof(blkcnt_t) < sizeof(u64)) {
2685 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2686 "cannot be mounted RDWR without "
2687 "CONFIG_LBDAF");
2688 return 0;
2689 }
2690 }
e2b911c5 2691 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2692 ext4_msg(sb, KERN_ERR,
2693 "Can't support bigalloc feature without "
2694 "extents feature\n");
2695 return 0;
2696 }
7c319d32
AK		 2697
2698#ifndef CONFIG_QUOTA
e2b911c5 2699 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2700 ext4_msg(sb, KERN_ERR,
2701 "Filesystem with quota feature cannot be mounted RDWR "
2702 "without CONFIG_QUOTA");
2703 return 0;
2704 }
689c958c
LX		 2705 if (ext4_has_feature_project(sb) && !readonly) {
2706 ext4_msg(sb, KERN_ERR,
2707 "Filesystem with project quota feature cannot be mounted RDWR "
2708 "without CONFIG_QUOTA");
2709 return 0;
2710 }
7c319d32 2711#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2712 return 1;
2713}
2714
66e61a9e
TT		 2715/*
2716 * This function is called once a day if we have errors logged
2717 * on the file system
2718 */
2719static void print_daily_error_info(unsigned long arg)
2720{
2721 struct super_block *sb = (struct super_block *) arg;
2722 struct ext4_sb_info *sbi;
2723 struct ext4_super_block *es;
2724
2725 sbi = EXT4_SB(sb);
2726 es = sbi->s_es;
2727
2728 if (es->s_error_count)
ae0f78de
TT		 2729 /* fsck newer than v1.41.13 is needed to clean this condition. */
2730 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2731 le32_to_cpu(es->s_error_count));
2732 if (es->s_first_error_time) {
ae0f78de 2733 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2734 sb->s_id, le32_to_cpu(es->s_first_error_time),
2735 (int) sizeof(es->s_first_error_func),
2736 es->s_first_error_func,
2737 le32_to_cpu(es->s_first_error_line));
2738 if (es->s_first_error_ino)
651e1c3b 2739 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2740 le32_to_cpu(es->s_first_error_ino));
2741 if (es->s_first_error_block)
651e1c3b 2742 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2743 le64_to_cpu(es->s_first_error_block));
651e1c3b 2744 printk(KERN_CONT "\n");
66e61a9e
TT		 2745 }
2746 if (es->s_last_error_time) {
ae0f78de 2747 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2748 sb->s_id, le32_to_cpu(es->s_last_error_time),
2749 (int) sizeof(es->s_last_error_func),
2750 es->s_last_error_func,
2751 le32_to_cpu(es->s_last_error_line));
2752 if (es->s_last_error_ino)
651e1c3b 2753 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2754 le32_to_cpu(es->s_last_error_ino));
2755 if (es->s_last_error_block)
651e1c3b 2756 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2757 le64_to_cpu(es->s_last_error_block));
651e1c3b 2758 printk(KERN_CONT "\n");
66e61a9e
TT		 2759 }
2760 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2761}
2762
bfff6873
LC		 2763/* Find next suitable group and run ext4_init_inode_table */
2764static int ext4_run_li_request(struct ext4_li_request *elr)
2765{
2766 struct ext4_group_desc *gdp = NULL;
2767 ext4_group_t group, ngroups;
2768 struct super_block *sb;
2769 unsigned long timeout = 0;
2770 int ret = 0;
2771
2772 sb = elr->lr_super;
2773 ngroups = EXT4_SB(sb)->s_groups_count;
2774
2775 for (group = elr->lr_next_group; group < ngroups; group++) {
2776 gdp = ext4_get_group_desc(sb, group, NULL);
2777 if (!gdp) {
2778 ret = 1;
2779 break;
2780 }
2781
2782 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2783 break;
2784 }
2785
7f511862 2786 if (group >= ngroups)
bfff6873
LC		 2787 ret = 1;
2788
2789 if (!ret) {
2790 timeout = jiffies;
2791 ret = ext4_init_inode_table(sb, group,
2792 elr->lr_timeout ? 0 : 1);
2793 if (elr->lr_timeout == 0) {
51ce6511
LC		 2794 timeout = (jiffies - timeout) *
2795 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2796 elr->lr_timeout = timeout;
2797 }
2798 elr->lr_next_sched = jiffies + elr->lr_timeout;
2799 elr->lr_next_group = group + 1;
2800 }
bfff6873
LC		 2801 return ret;
2802}
2803
2804/*
2805 * Remove lr_request from the list_request and free the
4ed5c033 2806 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2807 */
2808static void ext4_remove_li_request(struct ext4_li_request *elr)
2809{
2810 struct ext4_sb_info *sbi;
2811
2812 if (!elr)
2813 return;
2814
2815 sbi = elr->lr_sbi;
2816
2817 list_del(&elr->lr_request);
2818 sbi->s_li_request = NULL;
2819 kfree(elr);
2820}
2821
2822static void ext4_unregister_li_request(struct super_block *sb)
2823{
1bb933fb
LC		 2824 mutex_lock(&ext4_li_mtx);
2825 if (!ext4_li_info) {
2826 mutex_unlock(&ext4_li_mtx);
bfff6873 2827 return;
1bb933fb 2828 }
bfff6873
LC		 2829
2830 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2831 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2832 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2833 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2834}
2835
8f1f7453
ES		 2836static struct task_struct *ext4_lazyinit_task;
2837
bfff6873
LC		 2838/*
2839 * This is the function where ext4lazyinit thread lives. It walks
2840 * through the request list searching for next scheduled filesystem.
2841 * When such a fs is found, run the lazy initialization request
2842 * (ext4_rn_li_request) and keep track of the time spend in this
2843 * function. Based on that time we compute next schedule time of
2844 * the request. When walking through the list is complete, compute
2845 * next waking time and put itself into sleep.
2846 */
2847static int ext4_lazyinit_thread(void *arg)
2848{
2849 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2850 struct list_head *pos, *n;
2851 struct ext4_li_request *elr;
4ed5c033 2852 unsigned long next_wakeup, cur;
bfff6873
LC		 2853
2854 BUG_ON(NULL == eli);
2855
bfff6873
LC		 2856cont_thread:
2857 while (true) {
2858 next_wakeup = MAX_JIFFY_OFFSET;
2859
2860 mutex_lock(&eli->li_list_mtx);
2861 if (list_empty(&eli->li_request_list)) {
2862 mutex_unlock(&eli->li_list_mtx);
2863 goto exit_thread;
2864 }
bfff6873 2865 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 2866 int err = 0;
2867 int progress = 0;
bfff6873
LC		 2868 elr = list_entry(pos, struct ext4_li_request,
2869 lr_request);
2870
e22834f0
DM		 2871 if (time_before(jiffies, elr->lr_next_sched)) {
2872 if (time_before(elr->lr_next_sched, next_wakeup))
2873 next_wakeup = elr->lr_next_sched;
2874 continue;
2875 }
2876 if (down_read_trylock(&elr->lr_super->s_umount)) {
2877 if (sb_start_write_trylock(elr->lr_super)) {
2878 progress = 1;
2879 /*
2880 * We hold sb->s_umount, sb can not
2881 * be removed from the list, it is
2882 * now safe to drop li_list_mtx
2883 */
2884 mutex_unlock(&eli->li_list_mtx);
2885 err = ext4_run_li_request(elr);
2886 sb_end_write(elr->lr_super);
2887 mutex_lock(&eli->li_list_mtx);
2888 n = pos->next;
b2c78cd0 2889 }
e22834f0
DM		 2890 up_read((&elr->lr_super->s_umount));
2891 }
2892 /* error, remove the lazy_init job */
2893 if (err) {
2894 ext4_remove_li_request(elr);
2895 continue;
2896 }
2897 if (!progress) {
2898 elr->lr_next_sched = jiffies +
2899 (prandom_u32()
2900 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 2901 }
bfff6873
LC		 2902 if (time_before(elr->lr_next_sched, next_wakeup))
2903 next_wakeup = elr->lr_next_sched;
2904 }
2905 mutex_unlock(&eli->li_list_mtx);
2906
a0acae0e 2907 try_to_freeze();
bfff6873 2908
4ed5c033
LC		 2909 cur = jiffies;
2910 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2911 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2912 cond_resched();
2913 continue;
2914 }
2915
4ed5c033
LC		 2916 schedule_timeout_interruptible(next_wakeup - cur);
2917
8f1f7453
ES		 2918 if (kthread_should_stop()) {
2919 ext4_clear_request_list();
2920 goto exit_thread;
2921 }
bfff6873
LC		 2922 }
2923
2924exit_thread:
2925 /*
2926 * It looks like the request list is empty, but we need
2927 * to check it under the li_list_mtx lock, to prevent any
2928 * additions into it, and of course we should lock ext4_li_mtx
2929 * to atomically free the list and ext4_li_info, because at
2930 * this point another ext4 filesystem could be registering
2931 * new one.
2932 */
2933 mutex_lock(&ext4_li_mtx);
2934 mutex_lock(&eli->li_list_mtx);
2935 if (!list_empty(&eli->li_request_list)) {
2936 mutex_unlock(&eli->li_list_mtx);
2937 mutex_unlock(&ext4_li_mtx);
2938 goto cont_thread;
2939 }
2940 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2941 kfree(ext4_li_info);
2942 ext4_li_info = NULL;
2943 mutex_unlock(&ext4_li_mtx);
2944
2945 return 0;
2946}
2947
2948static void ext4_clear_request_list(void)
2949{
2950 struct list_head *pos, *n;
2951 struct ext4_li_request *elr;
2952
2953 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2954 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2955 elr = list_entry(pos, struct ext4_li_request,
2956 lr_request);
2957 ext4_remove_li_request(elr);
2958 }
2959 mutex_unlock(&ext4_li_info->li_list_mtx);
2960}
2961
2962static int ext4_run_lazyinit_thread(void)
2963{
8f1f7453
ES		 2964 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2965 ext4_li_info, "ext4lazyinit");
2966 if (IS_ERR(ext4_lazyinit_task)) {
2967 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2968 ext4_clear_request_list();
bfff6873
LC		 2969 kfree(ext4_li_info);
2970 ext4_li_info = NULL;
92b97816 2971 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2972 "initialization thread\n",
2973 err);
2974 return err;
2975 }
2976 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2977 return 0;
2978}
2979
2980/*
2981 * Check whether it make sense to run itable init. thread or not.
2982 * If there is at least one uninitialized inode table, return
2983 * corresponding group number, else the loop goes through all
2984 * groups and return total number of groups.
2985 */
2986static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2987{
2988 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2989 struct ext4_group_desc *gdp = NULL;
2990
2991 for (group = 0; group < ngroups; group++) {
2992 gdp = ext4_get_group_desc(sb, group, NULL);
2993 if (!gdp)
2994 continue;
2995
2996 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2997 break;
2998 }
2999
3000 return group;
3001}
3002
3003static int ext4_li_info_new(void)
3004{
3005 struct ext4_lazy_init *eli = NULL;
3006
3007 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3008 if (!eli)
3009 return -ENOMEM;
3010
bfff6873
LC		 3011 INIT_LIST_HEAD(&eli->li_request_list);
3012 mutex_init(&eli->li_list_mtx);
3013
bfff6873
LC		 3014 eli->li_state |= EXT4_LAZYINIT_QUIT;
3015
3016 ext4_li_info = eli;
3017
3018 return 0;
3019}
3020
3021static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3022 ext4_group_t start)
3023{
3024 struct ext4_sb_info *sbi = EXT4_SB(sb);
3025 struct ext4_li_request *elr;
bfff6873
LC		 3026
3027 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3028 if (!elr)
3029 return NULL;
3030
3031 elr->lr_super = sb;
3032 elr->lr_sbi = sbi;
3033 elr->lr_next_group = start;
3034
3035 /*
3036 * Randomize first schedule time of the request to
3037 * spread the inode table initialization requests
3038 * better.
3039 */
dd1f723b
TT		 3040 elr->lr_next_sched = jiffies + (prandom_u32() %
3041 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3042 return elr;
3043}
3044
7f511862
TT		 3045int ext4_register_li_request(struct super_block *sb,
3046 ext4_group_t first_not_zeroed)
bfff6873
LC		 3047{
3048 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3049 struct ext4_li_request *elr = NULL;
bfff6873 3050 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3051 int ret = 0;
bfff6873 3052
7f511862 3053 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3054 if (sbi->s_li_request != NULL) {
3055 /*
3056 * Reset timeout so it can be computed again, because
3057 * s_li_wait_mult might have changed.
3058 */
3059 sbi->s_li_request->lr_timeout = 0;
7f511862 3060 goto out;
51ce6511 3061 }
bfff6873
LC		 3062
3063 if (first_not_zeroed == ngroups ||
3064 (sb->s_flags & MS_RDONLY) ||
55ff3840 3065 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3066 goto out;
bfff6873
LC		 3067
3068 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3069 if (!elr) {
3070 ret = -ENOMEM;
3071 goto out;
3072 }
bfff6873
LC		 3073
3074 if (NULL == ext4_li_info) {
3075 ret = ext4_li_info_new();
3076 if (ret)
3077 goto out;
3078 }
3079
3080 mutex_lock(&ext4_li_info->li_list_mtx);
3081 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3082 mutex_unlock(&ext4_li_info->li_list_mtx);
3083
3084 sbi->s_li_request = elr;
46e4690b
TM		 3085 /*
3086 * set elr to NULL here since it has been inserted to
3087 * the request_list and the removal and free of it is
3088 * handled by ext4_clear_request_list from now on.
3089 */
3090 elr = NULL;
bfff6873
LC		 3091
3092 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3093 ret = ext4_run_lazyinit_thread();
3094 if (ret)
3095 goto out;
3096 }
bfff6873 3097out:
beed5ecb
NK		 3098 mutex_unlock(&ext4_li_mtx);
3099 if (ret)
bfff6873 3100 kfree(elr);
bfff6873
LC		 3101 return ret;
3102}
3103
3104/*
3105 * We do not need to lock anything since this is called on
3106 * module unload.
3107 */
3108static void ext4_destroy_lazyinit_thread(void)
3109{
3110 /*
3111 * If thread exited earlier
3112 * there's nothing to be done.
3113 */
8f1f7453 3114 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3115 return;
3116
8f1f7453 3117 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3118}
3119
25ed6e8a
DW		 3120static int set_journal_csum_feature_set(struct super_block *sb)
3121{
3122 int ret = 1;
3123 int compat, incompat;
3124 struct ext4_sb_info *sbi = EXT4_SB(sb);
3125
9aa5d32b 3126 if (ext4_has_metadata_csum(sb)) {
db9ee220 3127 /* journal checksum v3 */
25ed6e8a 3128 compat = 0;
db9ee220 3129 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3130 } else {
3131 /* journal checksum v1 */
3132 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3133 incompat = 0;
3134 }
3135
feb8c6d3
DW		 3136 jbd2_journal_clear_features(sbi->s_journal,
3137 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3138 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3139 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3140 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3141 ret = jbd2_journal_set_features(sbi->s_journal,
3142 compat, 0,
3143 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3144 incompat);
3145 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3146 ret = jbd2_journal_set_features(sbi->s_journal,
3147 compat, 0,
3148 incompat);
3149 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3150 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3151 } else {
feb8c6d3
DW		 3152 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3153 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3154 }
3155
3156 return ret;
3157}
3158
952fc18e
TT		 3159/*
3160 * Note: calculating the overhead so we can be compatible with
3161 * historical BSD practice is quite difficult in the face of
3162 * clusters/bigalloc. This is because multiple metadata blocks from
3163 * different block group can end up in the same allocation cluster.
3164 * Calculating the exact overhead in the face of clustered allocation
3165 * requires either O(all block bitmaps) in memory or O(number of block
3166 * groups**2) in time. We will still calculate the superblock for
3167 * older file systems --- and if we come across with a bigalloc file
3168 * system with zero in s_overhead_clusters the estimate will be close to
3169 * correct especially for very large cluster sizes --- but for newer
3170 * file systems, it's better to calculate this figure once at mkfs
3171 * time, and store it in the superblock. If the superblock value is
3172 * present (even for non-bigalloc file systems), we will use it.
3173 */
3174static int count_overhead(struct super_block *sb, ext4_group_t grp,
3175 char *buf)
3176{
3177 struct ext4_sb_info *sbi = EXT4_SB(sb);
3178 struct ext4_group_desc *gdp;
3179 ext4_fsblk_t first_block, last_block, b;
3180 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3181 int s, j, count = 0;
3182
e2b911c5 3183 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3184 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3185 sbi->s_itb_per_group + 2);
3186
952fc18e
TT		 3187 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3188 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3189 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3190 for (i = 0; i < ngroups; i++) {
3191 gdp = ext4_get_group_desc(sb, i, NULL);
3192 b = ext4_block_bitmap(sb, gdp);
3193 if (b >= first_block && b <= last_block) {
3194 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3195 count++;
3196 }
3197 b = ext4_inode_bitmap(sb, gdp);
3198 if (b >= first_block && b <= last_block) {
3199 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3200 count++;
3201 }
3202 b = ext4_inode_table(sb, gdp);
3203 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3204 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3205 int c = EXT4_B2C(sbi, b - first_block);
3206 ext4_set_bit(c, buf);
3207 count++;
3208 }
3209 if (i != grp)
3210 continue;
3211 s = 0;
3212 if (ext4_bg_has_super(sb, grp)) {
3213 ext4_set_bit(s++, buf);
3214 count++;
3215 }
c48ae41b
TT		 3216 j = ext4_bg_num_gdb(sb, grp);
3217 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3218 ext4_error(sb, "Invalid number of block group "
3219 "descriptor blocks: %d", j);
3220 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3221 }
c48ae41b
TT		 3222 count += j;
3223 for (; j > 0; j--)
3224 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3225 }
3226 if (!count)
3227 return 0;
3228 return EXT4_CLUSTERS_PER_GROUP(sb) -
3229 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3230}
3231
3232/*
3233 * Compute the overhead and stash it in sbi->s_overhead
3234 */
3235int ext4_calculate_overhead(struct super_block *sb)
3236{
3237 struct ext4_sb_info *sbi = EXT4_SB(sb);
3238 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3239 struct inode *j_inode;
3240 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3241 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3242 ext4_fsblk_t overhead = 0;
4fdb5543 3243 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3244
952fc18e
TT		 3245 if (!buf)
3246 return -ENOMEM;
3247
3248 /*
3249 * Compute the overhead (FS structures). This is constant
3250 * for a given filesystem unless the number of block groups
3251 * changes so we cache the previous value until it does.
3252 */
3253
3254 /*
3255 * All of the blocks before first_data_block are overhead
3256 */
3257 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3258
3259 /*
3260 * Add the overhead found in each block group
3261 */
3262 for (i = 0; i < ngroups; i++) {
3263 int blks;
3264
3265 blks = count_overhead(sb, i, buf);
3266 overhead += blks;
3267 if (blks)
3268 memset(buf, 0, PAGE_SIZE);
3269 cond_resched();
3270 }
3c816ded
EW		 3271
3272 /*
3273 * Add the internal journal blocks whether the journal has been
3274 * loaded or not
3275 */
b003b524 3276 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3277 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3278 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3279 j_inode = ext4_get_journal_inode(sb, j_inum);
3280 if (j_inode) {
3281 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3282 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3283 iput(j_inode);
3284 } else {
3285 ext4_msg(sb, KERN_ERR, "can't get journal size");
3286 }
3287 }
952fc18e
TT		 3288 sbi->s_overhead = overhead;
3289 smp_wmb();
3290 free_page((unsigned long) buf);
3291 return 0;
3292}
3293
b5799018 3294static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3295{
3296 ext4_fsblk_t resv_clusters;
b5799018 3297 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3298
30fac0f7
JK		 3299 /*
3300 * There's no need to reserve anything when we aren't using extents.
3301 * The space estimates are exact, there are no unwritten extents,
3302 * hole punching doesn't need new metadata... This is needed especially
3303 * to keep ext2/3 backward compatibility.
3304 */
e2b911c5 3305 if (!ext4_has_feature_extents(sb))
b5799018 3306 return;
27dd4385
LC		 3307 /*
3308 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3309 * This should cover the situations where we can not afford to run
3310 * out of space like for example punch hole, or converting
556615dc 3311 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3312 * allocation would require 1, or 2 blocks, higher numbers are
3313 * very rare.
3314 */
b5799018
TT		 3315 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3316 sbi->s_cluster_bits);
27dd4385
LC		 3317
3318 do_div(resv_clusters, 50);
3319 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3320
b5799018 3321 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3322}
3323
2b2d6d01 3324static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3325{
d4c402d9 3326 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3327 struct buffer_head *bh;
617ba13b 3328 struct ext4_super_block *es = NULL;
5aee0f8a 3329 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3330 ext4_fsblk_t block;
3331 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3332 ext4_fsblk_t logical_sb_block;
ac27a0ec 3333 unsigned long offset = 0;
ac27a0ec
DK		 3334 unsigned long journal_devnum = 0;
3335 unsigned long def_mount_opts;
3336 struct inode *root;
0390131b 3337 const char *descr;
dcc7dae3 3338 int ret = -ENOMEM;
281b5995 3339 int blocksize, clustersize;
4ec11028
TT		 3340 unsigned int db_count;
3341 unsigned int i;
281b5995 3342 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3343 __u64 blocks_count;
07aa2ea1 3344 int err = 0;
b3881f74 3345 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3346 ext4_group_t first_not_zeroed;
ac27a0ec 3347
5aee0f8a
TT		 3348 if ((data && !orig_data) || !sbi)
3349 goto out_free_base;
705895b6
PE		 3350
3351 sbi->s_blockgroup_lock =
3352 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3353 if (!sbi->s_blockgroup_lock)
3354 goto out_free_base;
3355
ac27a0ec 3356 sb->s_fs_info = sbi;
2c0544b2 3357 sbi->s_sb = sb;
240799cd 3358 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3359 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3360 if (sb->s_bdev->bd_part)
3361 sbi->s_sectors_written_start =
3362 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3363
9f6200bb 3364 /* Cleanup superblock name */
ec3904dc 3365 strreplace(sb->s_id, '/', '!');
9f6200bb 3366
07aa2ea1 3367 /* -EINVAL is default */
dcc7dae3 3368 ret = -EINVAL;
617ba13b 3369 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3370 if (!blocksize) {
b31e1552 3371 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3372 goto out_fail;
3373 }
3374
3375 /*
617ba13b 3376 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3377 * block sizes. We need to calculate the offset from buffer start.
3378 */
617ba13b 3379 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3380 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3381 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3382 } else {
70bbb3e0 3383 logical_sb_block = sb_block;
ac27a0ec
DK		 3384 }
3385
a8ac900b 3386 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3387 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3388 goto out_fail;
3389 }
3390 /*
3391 * Note: s_es must be initialized as soon as possible because
617ba13b 3392 * some ext4 macro-instructions depend on its value
ac27a0ec 3393 */
2716b802 3394 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3395 sbi->s_es = es;
3396 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3397 if (sb->s_magic != EXT4_SUPER_MAGIC)
3398 goto cantfind_ext4;
afc32f7e 3399 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3400
feb0ab32 3401 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3402 if (ext4_has_feature_metadata_csum(sb) &&
3403 ext4_has_feature_gdt_csum(sb))
363307e6 3404 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3405 "redundant flags; please run fsck.");
3406
d25425f8
DW		 3407 /* Check for a known checksum algorithm */
3408 if (!ext4_verify_csum_type(sb, es)) {
3409 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3410 "unknown checksum algorithm.");
3411 silent = 1;
3412 goto cantfind_ext4;
3413 }
3414
0441984a 3415 /* Load the checksum driver */
e2b911c5 3416 if (ext4_has_feature_metadata_csum(sb)) {
0441984a
DW		 3417 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3418 if (IS_ERR(sbi->s_chksum_driver)) {
3419 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3420 ret = PTR_ERR(sbi->s_chksum_driver);
3421 sbi->s_chksum_driver = NULL;
3422 goto failed_mount;
3423 }
3424 }
3425
a9c47317
DW		 3426 /* Check superblock checksum */
3427 if (!ext4_superblock_csum_verify(sb, es)) {
3428 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3429 "invalid superblock checksum. Run e2fsck?");
3430 silent = 1;
6a797d27 3431 ret = -EFSBADCRC;
a9c47317
DW		 3432 goto cantfind_ext4;
3433 }
3434
3435 /* Precompute checksum seed for all metadata */
e2b911c5 3436 if (ext4_has_feature_csum_seed(sb))
8c81bd8f
DW		 3437 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3438 else if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3439 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3440 sizeof(es->s_uuid));
3441
ac27a0ec
DK		 3442 /* Set defaults before we parse the mount options */
3443 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3444 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3445 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3446 set_opt(sb, DEBUG);
87f26807 3447 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3448 set_opt(sb, GRPID);
617ba13b 3449 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3450 set_opt(sb, NO_UID32);
ea663336 3451 /* xattr user namespace & acls are now defaulted on */
ea663336 3452 set_opt(sb, XATTR_USER);
03010a33 3453#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3454 set_opt(sb, POSIX_ACL);
2e7842b8 3455#endif
98c1a759
DW		 3456 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3457 if (ext4_has_metadata_csum(sb))
3458 set_opt(sb, JOURNAL_CHECKSUM);
3459
617ba13b 3460 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3461 set_opt(sb, JOURNAL_DATA);
617ba13b 3462 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3463 set_opt(sb, ORDERED_DATA);
617ba13b 3464 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3465 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3466
3467 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3468 set_opt(sb, ERRORS_PANIC);
bb4f397a 3469 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3470 set_opt(sb, ERRORS_CONT);
bb4f397a 3471 else
fd8c37ec 3472 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3473 /* block_validity enabled by default; disable with noblock_validity */
3474 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3475 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3476 set_opt(sb, DISCARD);
ac27a0ec 3477
08cefc7a
EB		 3478 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3479 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3480 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3481 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3482 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3483
8b67f04a 3484 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3485 set_opt(sb, BARRIER);
ac27a0ec 3486
dd919b98
AK		 3487 /*
3488 * enable delayed allocation by default
3489 * Use -o nodelalloc to turn it off
3490 */
bc0b75f7 3491 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3492 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3493 set_opt(sb, DELALLOC);
dd919b98 3494
51ce6511
LC		 3495 /*
3496 * set default s_li_wait_mult for lazyinit, for the case there is
3497 * no mount option specified.
3498 */
3499 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3500
5aee0f8a
TT		 3501 if (sbi->s_es->s_mount_opts[0]) {
3502 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3503 sizeof(sbi->s_es->s_mount_opts),
3504 GFP_KERNEL);
3505 if (!s_mount_opts)
3506 goto failed_mount;
3507 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3508 &journal_ioprio, 0)) {
3509 ext4_msg(sb, KERN_WARNING,
3510 "failed to parse options in superblock: %s",
3511 s_mount_opts);
3512 }
3513 kfree(s_mount_opts);
8b67f04a 3514 }
5a916be1 3515 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3516 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3517 &journal_ioprio, 0))
ac27a0ec
DK		 3518 goto failed_mount;
3519
56889787
TT		 3520 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3521 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3522 "with data=journal disables delayed "
3523 "allocation and O_DIRECT support!\n");
3524 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3525 ext4_msg(sb, KERN_ERR, "can't mount with "
3526 "both data=journal and delalloc");
3527 goto failed_mount;
3528 }
3529 if (test_opt(sb, DIOREAD_NOLOCK)) {
3530 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3531 "both data=journal and dioread_nolock");
56889787
TT		 3532 goto failed_mount;
3533 }
923ae0ff
RZ		 3534 if (test_opt(sb, DAX)) {
3535 ext4_msg(sb, KERN_ERR, "can't mount with "
3536 "both data=journal and dax");
3537 goto failed_mount;
3538 }
56889787
TT		 3539 if (test_opt(sb, DELALLOC))
3540 clear_opt(sb, DELALLOC);
001e4a87
TH		 3541 } else {
3542 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3543 }
3544
ac27a0ec 3545 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3546 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3547
617ba13b 3548 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3549 (ext4_has_compat_features(sb) ||
3550 ext4_has_ro_compat_features(sb) ||
3551 ext4_has_incompat_features(sb)))
b31e1552
ES		 3552 ext4_msg(sb, KERN_WARNING,
3553 "feature flags set on rev 0 fs, "
3554 "running e2fsck is recommended");
469108ff 3555
ed3654eb
TT		 3556 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3557 set_opt2(sb, HURD_COMPAT);
e2b911c5 3558 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3559 ext4_msg(sb, KERN_ERR,
3560 "The Hurd can't support 64-bit file systems");
3561 goto failed_mount;
3562 }
3563 }
3564
2035e776
TT		 3565 if (IS_EXT2_SB(sb)) {
3566 if (ext2_feature_set_ok(sb))
3567 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3568 "using the ext4 subsystem");
3569 else {
3570 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3571 "to feature incompatibilities");
3572 goto failed_mount;
3573 }
3574 }
3575
3576 if (IS_EXT3_SB(sb)) {
3577 if (ext3_feature_set_ok(sb))
3578 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3579 "using the ext4 subsystem");
3580 else {
3581 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3582 "to feature incompatibilities");
3583 goto failed_mount;
3584 }
3585 }
3586
ac27a0ec
DK		 3587 /*
3588 * Check feature flags regardless of the revision level, since we
3589 * previously didn't change the revision level when setting the flags,
3590 * so there is a chance incompat flags are set on a rev 0 filesystem.
3591 */
a13fb1a4 3592 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3593 goto failed_mount;
a13fb1a4 3594
261cb20c 3595 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3596 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3597 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3598 ext4_msg(sb, KERN_ERR,
9e47a4c9
TT		 3599 "Unsupported filesystem blocksize %d (%d log_block_size)",
3600 blocksize, le32_to_cpu(es->s_log_block_size));
3601 goto failed_mount;
3602 }
3603 if (le32_to_cpu(es->s_log_block_size) >
3604 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3605 ext4_msg(sb, KERN_ERR,
3606 "Invalid log block size: %u",
3607 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3608 goto failed_mount;
3609 }
3610
5b9554dc
TT		 3611 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3612 ext4_msg(sb, KERN_ERR,
3613 "Number of reserved GDT blocks insanely large: %d",
3614 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3615 goto failed_mount;
3616 }
3617
923ae0ff 3618 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
87eefeb4
TK		 3619 err = bdev_dax_supported(sb, blocksize);
3620 if (err)
923ae0ff 3621 goto failed_mount;
923ae0ff
RZ		 3622 }
3623
e2b911c5 3624 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3625 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3626 es->s_encryption_level);
3627 goto failed_mount;
3628 }
3629
ac27a0ec 3630 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3631 /* Validate the filesystem blocksize */
3632 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3633 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3634 blocksize);
ac27a0ec
DK		 3635 goto failed_mount;
3636 }
3637
2b2d6d01 3638 brelse(bh);
70bbb3e0
AM		 3639 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3640 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3641 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3642 if (!bh) {
b31e1552
ES		 3643 ext4_msg(sb, KERN_ERR,
3644 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3645 goto failed_mount;
3646 }
2716b802 3647 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3648 sbi->s_es = es;
617ba13b 3649 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3650 ext4_msg(sb, KERN_ERR,
3651 "Magic mismatch, very weird!");
ac27a0ec
DK		 3652 goto failed_mount;
3653 }
3654 }
3655
e2b911c5 3656 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3657 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3658 has_huge_files);
3659 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3660
617ba13b
MC		 3661 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3662 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3663 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3664 } else {
3665 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3666 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3667 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3668 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3669 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3670 ext4_msg(sb, KERN_ERR,
3671 "unsupported inode size: %d",
2b2d6d01 3672 sbi->s_inode_size);
ac27a0ec
DK		 3673 goto failed_mount;
3674 }
ef7f3835
KS		 3675 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3676 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3677 }
0b8e58a1 3678
0d1ee42f 3679 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3680 if (ext4_has_feature_64bit(sb)) {
8fadc143 3681 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3682 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3683 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3684 ext4_msg(sb, KERN_ERR,
3685 "unsupported descriptor size %lu",
0d1ee42f
AR		 3686 sbi->s_desc_size);
3687 goto failed_mount;
3688 }
3689 } else
3690 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3691
ac27a0ec 3692 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3693 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 3694
617ba13b 3695 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3696 if (sbi->s_inodes_per_block == 0)
617ba13b 3697 goto cantfind_ext4;
cd6bb35b
TT		 3698 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3699 sbi->s_inodes_per_group > blocksize * 8) {
3700 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3701 sbi->s_blocks_per_group);
3702 goto failed_mount;
3703 }
ac27a0ec
DK		 3704 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3705 sbi->s_inodes_per_block;
0d1ee42f 3706 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3707 sbi->s_sbh = bh;
3708 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3709 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3710 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3711
2b2d6d01 3712 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3713 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3714 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3715 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3716 i = le32_to_cpu(es->s_flags);
3717 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3718 sbi->s_hash_unsigned = 3;
3719 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3720#ifdef __CHAR_UNSIGNED__
23301410
TT		 3721 if (!(sb->s_flags & MS_RDONLY))
3722 es->s_flags |=
3723 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3724 sbi->s_hash_unsigned = 3;
f99b2589 3725#else
23301410
TT		 3726 if (!(sb->s_flags & MS_RDONLY))
3727 es->s_flags |=
3728 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3729#endif
23301410 3730 }
f99b2589 3731 }
ac27a0ec 3732
281b5995
TT		 3733 /* Handle clustersize */
3734 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3735 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3736 if (has_bigalloc) {
3737 if (clustersize < blocksize) {
3738 ext4_msg(sb, KERN_ERR,
3739 "cluster size (%d) smaller than "
3740 "block size (%d)", clustersize, blocksize);
3741 goto failed_mount;
3742 }
9e47a4c9
TT		 3743 if (le32_to_cpu(es->s_log_cluster_size) >
3744 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3745 ext4_msg(sb, KERN_ERR,
3746 "Invalid log cluster size: %u",
3747 le32_to_cpu(es->s_log_cluster_size));
3748 goto failed_mount;
3749 }
281b5995
TT		 3750 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3751 le32_to_cpu(es->s_log_block_size);
3752 sbi->s_clusters_per_group =
3753 le32_to_cpu(es->s_clusters_per_group);
3754 if (sbi->s_clusters_per_group > blocksize * 8) {
3755 ext4_msg(sb, KERN_ERR,
3756 "#clusters per group too big: %lu",
3757 sbi->s_clusters_per_group);
3758 goto failed_mount;
3759 }
3760 if (sbi->s_blocks_per_group !=
3761 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3762 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3763 "clusters per group (%lu) inconsistent",
3764 sbi->s_blocks_per_group,
3765 sbi->s_clusters_per_group);
3766 goto failed_mount;
3767 }
3768 } else {
3769 if (clustersize != blocksize) {
3770 ext4_warning(sb, "fragment/cluster size (%d) != "
3771 "block size (%d)", clustersize,
3772 blocksize);
3773 clustersize = blocksize;
3774 }
3775 if (sbi->s_blocks_per_group > blocksize * 8) {
3776 ext4_msg(sb, KERN_ERR,
3777 "#blocks per group too big: %lu",
3778 sbi->s_blocks_per_group);
3779 goto failed_mount;
3780 }
3781 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3782 sbi->s_cluster_bits = 0;
ac27a0ec 3783 }
281b5995
TT		 3784 sbi->s_cluster_ratio = clustersize / blocksize;
3785
960fd856
TT		 3786 /* Do we have standard group size of clustersize * 8 blocks ? */
3787 if (sbi->s_blocks_per_group == clustersize << 3)
3788 set_opt2(sb, STD_GROUP_SIZE);
3789
bf43d84b
ES		 3790 /*
3791 * Test whether we have more sectors than will fit in sector_t,
3792 * and whether the max offset is addressable by the page cache.
3793 */
5a9ae68a 3794 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3795 ext4_blocks_count(es));
5a9ae68a 3796 if (err) {
b31e1552 3797 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3798 " too large to mount safely on this system");
ac27a0ec 3799 if (sizeof(sector_t) < 8)
90c699a9 3800 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3801 goto failed_mount;
3802 }
3803
617ba13b
MC		 3804 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3805 goto cantfind_ext4;
e7c95593 3806
0f2ddca6
FTN		 3807 /* check blocks count against device size */
3808 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3809 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3810 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3811 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3812 ext4_blocks_count(es), blocks_count);
3813 goto failed_mount;
3814 }
3815
0b8e58a1
AD		 3816 /*
3817 * It makes no sense for the first data block to be beyond the end
3818 * of the filesystem.
3819 */
3820 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3821 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3822 "block %u is beyond end of filesystem (%llu)",
3823 le32_to_cpu(es->s_first_data_block),
3824 ext4_blocks_count(es));
e7c95593
ES		 3825 goto failed_mount;
3826 }
bd81d8ee
LV		 3827 blocks_count = (ext4_blocks_count(es) -
3828 le32_to_cpu(es->s_first_data_block) +
3829 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3830 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3831 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3832 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3833 "(block count %llu, first data block %u, "
b31e1552 3834 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3835 ext4_blocks_count(es),
3836 le32_to_cpu(es->s_first_data_block),
3837 EXT4_BLOCKS_PER_GROUP(sb));
3838 goto failed_mount;
3839 }
bd81d8ee 3840 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3841 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3842 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3843 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3844 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3845 sbi->s_group_desc = ext4_kvmalloc(db_count *
3846 sizeof(struct buffer_head *),
3847 GFP_KERNEL);
ac27a0ec 3848 if (sbi->s_group_desc == NULL) {
b31e1552 3849 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3850 ret = -ENOMEM;
ac27a0ec
DK		 3851 goto failed_mount;
3852 }
3853
705895b6 3854 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3855
3856 for (i = 0; i < db_count; i++) {
70bbb3e0 3857 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3858 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3859 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3860 ext4_msg(sb, KERN_ERR,
3861 "can't read group descriptor %d", i);
ac27a0ec
DK		 3862 db_count = i;
3863 goto failed_mount2;
3864 }
3865 }
829fa70d 3866 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 3867 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 3868 ret = -EFSCORRUPTED;
f9ae9cf5 3869 goto failed_mount2;
ac27a0ec 3870 }
772cb7c8 3871
f9ae9cf5 3872 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3873 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3874 spin_lock_init(&sbi->s_next_gen_lock);
3875
04ecddb7
JM		 3876 setup_timer(&sbi->s_err_report, print_daily_error_info,
3877 (unsigned long) sb);
04496411 3878
a75ae78f 3879 /* Register extent status tree shrinker */
eb68d0e2 3880 if (ext4_es_register_shrinker(sbi))
ce7e010a 3881 goto failed_mount3;
ce7e010a 3882
c9de560d 3883 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3884 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3885
f9ae9cf5
TT		 3886 /*
3887 * set up enough so that it can read an inode
3888 */
f6e63f90 3889 sb->s_op = &ext4_sops;
617ba13b
MC		 3890 sb->s_export_op = &ext4_export_ops;
3891 sb->s_xattr = ext4_xattr_handlers;
a7550b30 3892 sb->s_cop = &ext4_cryptops;
ac27a0ec 3893#ifdef CONFIG_QUOTA
617ba13b 3894 sb->dq_op = &ext4_quota_operations;
e2b911c5 3895 if (ext4_has_feature_quota(sb))
1fa5efe3 3896 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3897 else
3898 sb->s_qcop = &ext4_qctl_operations;
689c958c 3899 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 3900#endif
f2fa2ffc
AK		 3901 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3902
ac27a0ec 3903 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3904 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3905
3906 sb->s_root = NULL;
3907
3908 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 3909 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 3910
e2b911c5 3911 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
c5e06d10 3912 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3913 goto failed_mount3a;
c5e06d10 3914
ac27a0ec
DK		 3915 /*
3916 * The first inode we look at is the journal inode. Don't try
3917 * root first: it may be modified in the journal!
3918 */
e2b911c5 3919 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
617ba13b 3920 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3921 goto failed_mount3a;
0390131b 3922 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
e2b911c5 3923 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 3924 ext4_msg(sb, KERN_ERR, "required journal recovery "
3925 "suppressed and not mounted read-only");
744692dc 3926 goto failed_mount_wq;
ac27a0ec 3927 } else {
1e381f60
DM		 3928 /* Nojournal mode, all journal mount options are illegal */
3929 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3930 ext4_msg(sb, KERN_ERR, "can't mount with "
3931 "journal_checksum, fs mounted w/o journal");
3932 goto failed_mount_wq;
3933 }
3934 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3935 ext4_msg(sb, KERN_ERR, "can't mount with "
3936 "journal_async_commit, fs mounted w/o journal");
3937 goto failed_mount_wq;
3938 }
3939 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3940 ext4_msg(sb, KERN_ERR, "can't mount with "
3941 "commit=%lu, fs mounted w/o journal",
3942 sbi->s_commit_interval / HZ);
3943 goto failed_mount_wq;
3944 }
3945 if (EXT4_MOUNT_DATA_FLAGS &
3946 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3947 ext4_msg(sb, KERN_ERR, "can't mount with "
3948 "data=, fs mounted w/o journal");
3949 goto failed_mount_wq;
3950 }
3951 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3952 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 3953 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3954 sbi->s_journal = NULL;
3955 needs_recovery = 0;
3956 goto no_journal;
ac27a0ec
DK		 3957 }
3958
e2b911c5 3959 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 3960 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3961 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3962 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3963 goto failed_mount_wq;
eb40a09c
JS		 3964 }
3965
25ed6e8a
DW		 3966 if (!set_journal_csum_feature_set(sb)) {
3967 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3968 "feature set");
3969 goto failed_mount_wq;
d4da6c9c 3970 }
818d276c 3971
ac27a0ec
DK		 3972 /* We have now updated the journal if required, so we can
3973 * validate the data journaling mode. */
3974 switch (test_opt(sb, DATA_FLAGS)) {
3975 case 0:
3976 /* No mode set, assume a default based on the journal
63f57933
AM		 3977 * capabilities: ORDERED_DATA if the journal can
3978 * cope, else JOURNAL_DATA
3979 */
dab291af
MC		 3980 if (jbd2_journal_check_available_features
3981 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3982 set_opt(sb, ORDERED_DATA);
ac27a0ec 3983 else
fd8c37ec 3984 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3985 break;
3986
617ba13b
MC		 3987 case EXT4_MOUNT_ORDERED_DATA:
3988 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3989 if (!jbd2_journal_check_available_features
3990 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3991 ext4_msg(sb, KERN_ERR, "Journal does not support "
3992 "requested data journaling mode");
744692dc 3993 goto failed_mount_wq;
ac27a0ec
DK		 3994 }
3995 default:
3996 break;
3997 }
b3881f74 3998 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3999
18aadd47
BJ		 4000 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4001
ce7e010a 4002no_journal:
2335d05f
AG		 4003 sbi->s_mb_cache = ext4_xattr_create_cache();
4004 if (!sbi->s_mb_cache) {
4005 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4006 goto failed_mount_wq;
9c191f70
M		 4007 }
4008
e2b911c5 4009 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 4010 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 4011 ext4_msg(sb, KERN_ERR,
4012 "Unsupported blocksize for fs encryption");
4013 goto failed_mount_wq;
4014 }
4015
e2b911c5
DW		 4016 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4017 !ext4_has_feature_encrypt(sb)) {
4018 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4019 ext4_commit_super(sb, 1);
4020 }
4021
952fc18e
TT		 4022 /*
4023 * Get the # of file system overhead blocks from the
4024 * superblock if present.
4025 */
4026 if (es->s_overhead_clusters)
4027 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4028 else {
07aa2ea1
LC		 4029 err = ext4_calculate_overhead(sb);
4030 if (err)
952fc18e
TT		 4031 goto failed_mount_wq;
4032 }
4033
fd89d5f2
TH		 4034 /*
4035 * The maximum number of concurrent works can be high and
4036 * concurrency isn't really necessary. Limit it to 1.
4037 */
2e8fa54e
JK		 4038 EXT4_SB(sb)->rsv_conversion_wq =
4039 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4040 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4041 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4042 ret = -ENOMEM;
2e8fa54e
JK		 4043 goto failed_mount4;
4044 }
4045
ac27a0ec 4046 /*
dab291af 4047 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4048 * so we can safely mount the rest of the filesystem now.
4049 */
4050
1d1fe1ee
DH		 4051 root = ext4_iget(sb, EXT4_ROOT_INO);
4052 if (IS_ERR(root)) {
b31e1552 4053 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4054 ret = PTR_ERR(root);
32a9bb57 4055 root = NULL;
ac27a0ec
DK		 4056 goto failed_mount4;
4057 }
4058 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4059 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4060 iput(root);
ac27a0ec
DK		 4061 goto failed_mount4;
4062 }
48fde701 4063 sb->s_root = d_make_root(root);
1d1fe1ee 4064 if (!sb->s_root) {
b31e1552 4065 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4066 ret = -ENOMEM;
4067 goto failed_mount4;
4068 }
ac27a0ec 4069
7e84b621
ES		 4070 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4071 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4072
4073 /* determine the minimum size of new large inodes, if present */
4074 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4075 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4076 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4077 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4078 if (sbi->s_want_extra_isize <
4079 le16_to_cpu(es->s_want_extra_isize))
4080 sbi->s_want_extra_isize =
4081 le16_to_cpu(es->s_want_extra_isize);
4082 if (sbi->s_want_extra_isize <
4083 le16_to_cpu(es->s_min_extra_isize))
4084 sbi->s_want_extra_isize =
4085 le16_to_cpu(es->s_min_extra_isize);
4086 }
4087 }
4088 /* Check if enough inode space is available */
4089 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4090 sbi->s_inode_size) {
4091 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4092 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4093 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4094 "available");
ef7f3835
KS		 4095 }
4096
b5799018 4097 ext4_set_resv_clusters(sb);
27dd4385 4098
6fd058f7
TT		 4099 err = ext4_setup_system_zone(sb);
4100 if (err) {
b31e1552 4101 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4102 "zone (%d)", err);
f9ae9cf5
TT		 4103 goto failed_mount4a;
4104 }
4105
4106 ext4_ext_init(sb);
4107 err = ext4_mb_init(sb);
4108 if (err) {
4109 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4110 err);
dcf2d804 4111 goto failed_mount5;
c2774d84
AK		 4112 }
4113
d5e03cbb
TT		 4114 block = ext4_count_free_clusters(sb);
4115 ext4_free_blocks_count_set(sbi->s_es,
4116 EXT4_C2B(sbi, block));
908c7f19
TH		 4117 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4118 GFP_KERNEL);
d5e03cbb
TT		 4119 if (!err) {
4120 unsigned long freei = ext4_count_free_inodes(sb);
4121 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4122 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4123 GFP_KERNEL);
d5e03cbb
TT		 4124 }
4125 if (!err)
4126 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4127 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4128 if (!err)
908c7f19
TH		 4129 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4130 GFP_KERNEL);
c8585c6f
DJ		 4131 if (!err)
4132 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4133
d5e03cbb
TT		 4134 if (err) {
4135 ext4_msg(sb, KERN_ERR, "insufficient memory");
4136 goto failed_mount6;
4137 }
4138
e2b911c5 4139 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4140 if (!ext4_fill_flex_info(sb)) {
4141 ext4_msg(sb, KERN_ERR,
4142 "unable to initialize "
4143 "flex_bg meta info!");
4144 goto failed_mount6;
4145 }
4146
bfff6873
LC		 4147 err = ext4_register_li_request(sb, first_not_zeroed);
4148 if (err)
dcf2d804 4149 goto failed_mount6;
bfff6873 4150
b5799018 4151 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4152 if (err)
4153 goto failed_mount7;
3197ebdb 4154
9b2ff357
JK		 4155#ifdef CONFIG_QUOTA
4156 /* Enable quota usage during mount. */
e2b911c5 4157 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
9b2ff357
JK		 4158 err = ext4_enable_quotas(sb);
4159 if (err)
4160 goto failed_mount8;
4161 }
4162#endif /* CONFIG_QUOTA */
4163
617ba13b
MC		 4164 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4165 ext4_orphan_cleanup(sb, es);
4166 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4167 if (needs_recovery) {
b31e1552 4168 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4169 ext4_mark_recovery_complete(sb, es);
4170 }
4171 if (EXT4_SB(sb)->s_journal) {
4172 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4173 descr = " journalled data mode";
4174 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4175 descr = " ordered data mode";
4176 else
4177 descr = " writeback data mode";
4178 } else
4179 descr = "out journal";
4180
79add3a3
LC		 4181 if (test_opt(sb, DISCARD)) {
4182 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4183 if (!blk_queue_discard(q))
4184 ext4_msg(sb, KERN_WARNING,
4185 "mounting with \"discard\" option, but "
4186 "the device does not support discard");
4187 }
4188
e294a537
TT		 4189 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4190 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4191 "Opts: %.*s%s%s", descr,
4192 (int) sizeof(sbi->s_es->s_mount_opts),
4193 sbi->s_es->s_mount_opts,
e294a537 4194 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4195
66e61a9e
TT		 4196 if (es->s_error_count)
4197 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4198
efbed4dc
TT		 4199 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4200 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4201 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4202 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4203
d4c402d9 4204 kfree(orig_data);
a7550b30
JK		 4205#ifdef CONFIG_EXT4_FS_ENCRYPTION
4206 memcpy(sbi->key_prefix, EXT4_KEY_DESC_PREFIX,
4207 EXT4_KEY_DESC_PREFIX_SIZE);
4208 sbi->key_prefix_size = EXT4_KEY_DESC_PREFIX_SIZE;
4209#endif
ac27a0ec
DK		 4210 return 0;
4211
617ba13b 4212cantfind_ext4:
ac27a0ec 4213 if (!silent)
b31e1552 4214 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4215 goto failed_mount;
4216
72ba7450
TT		 4217#ifdef CONFIG_QUOTA
4218failed_mount8:
ebd173be 4219 ext4_unregister_sysfs(sb);
72ba7450 4220#endif
dcf2d804
TM		 4221failed_mount7:
4222 ext4_unregister_li_request(sb);
4223failed_mount6:
f9ae9cf5 4224 ext4_mb_release(sb);
d5e03cbb 4225 if (sbi->s_flex_groups)
b93b41d4 4226 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4227 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4228 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4229 percpu_counter_destroy(&sbi->s_dirs_counter);
4230 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4231failed_mount5:
f9ae9cf5
TT		 4232 ext4_ext_release(sb);
4233 ext4_release_system_zone(sb);
4234failed_mount4a:
94bf608a 4235 dput(sb->s_root);
32a9bb57 4236 sb->s_root = NULL;
94bf608a 4237failed_mount4:
b31e1552 4238 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4239 if (EXT4_SB(sb)->rsv_conversion_wq)
4240 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4241failed_mount_wq:
82939d79
JK		 4242 if (sbi->s_mb_cache) {
4243 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4244 sbi->s_mb_cache = NULL;
4245 }
0390131b
FM		 4246 if (sbi->s_journal) {
4247 jbd2_journal_destroy(sbi->s_journal);
4248 sbi->s_journal = NULL;
4249 }
50460fe8 4250failed_mount3a:
d3922a77 4251 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4252failed_mount3:
9105bb14 4253 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4254 if (sbi->s_mmp_tsk)
4255 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4256failed_mount2:
4257 for (i = 0; i < db_count; i++)
4258 brelse(sbi->s_group_desc[i]);
b93b41d4 4259 kvfree(sbi->s_group_desc);
ac27a0ec 4260failed_mount:
0441984a
DW		 4261 if (sbi->s_chksum_driver)
4262 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4263#ifdef CONFIG_QUOTA
a2d4a646 4264 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4265 kfree(sbi->s_qf_names[i]);
4266#endif
617ba13b 4267 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4268 brelse(bh);
4269out_fail:
4270 sb->s_fs_info = NULL;
f6830165 4271 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4272out_free_base:
ac27a0ec 4273 kfree(sbi);
d4c402d9 4274 kfree(orig_data);
07aa2ea1 4275 return err ? err : ret;
ac27a0ec
DK		 4276}
4277
4278/*
4279 * Setup any per-fs journal parameters now. We'll do this both on
4280 * initial mount, once the journal has been initialised but before we've
4281 * done any recovery; and again on any subsequent remount.
4282 */
617ba13b 4283static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4284{
617ba13b 4285 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4286
30773840
TT		 4287 journal->j_commit_interval = sbi->s_commit_interval;
4288 journal->j_min_batch_time = sbi->s_min_batch_time;
4289 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4290
a931da6a 4291 write_lock(&journal->j_state_lock);
ac27a0ec 4292 if (test_opt(sb, BARRIER))
dab291af 4293 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4294 else
dab291af 4295 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4296 if (test_opt(sb, DATA_ERR_ABORT))
4297 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4298 else
4299 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4300 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4301}
4302
c6cb7e77
EW		 4303static struct inode *ext4_get_journal_inode(struct super_block *sb,
4304 unsigned int journal_inum)
ac27a0ec
DK		 4305{
4306 struct inode *journal_inode;
ac27a0ec 4307
c6cb7e77
EW		 4308 /*
4309 * Test for the existence of a valid inode on disk. Bad things
4310 * happen if we iget() an unused inode, as the subsequent iput()
4311 * will try to delete it.
4312 */
1d1fe1ee
DH		 4313 journal_inode = ext4_iget(sb, journal_inum);
4314 if (IS_ERR(journal_inode)) {
b31e1552 4315 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4316 return NULL;
4317 }
4318 if (!journal_inode->i_nlink) {
4319 make_bad_inode(journal_inode);
4320 iput(journal_inode);
b31e1552 4321 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4322 return NULL;
4323 }
4324
e5f8eab8 4325 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4326 journal_inode, journal_inode->i_size);
1d1fe1ee 4327 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4328 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4329 iput(journal_inode);
4330 return NULL;
4331 }
c6cb7e77
EW		 4332 return journal_inode;
4333}
4334
4335static journal_t *ext4_get_journal(struct super_block *sb,
4336 unsigned int journal_inum)
4337{
4338 struct inode *journal_inode;
4339 journal_t *journal;
4340
4341 BUG_ON(!ext4_has_feature_journal(sb));
4342
4343 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4344 if (!journal_inode)
4345 return NULL;
ac27a0ec 4346
dab291af 4347 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4348 if (!journal) {
b31e1552 4349 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4350 iput(journal_inode);
4351 return NULL;
4352 }
4353 journal->j_private = sb;
617ba13b 4354 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4355 return journal;
4356}
4357
617ba13b 4358static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4359 dev_t j_dev)
4360{
2b2d6d01 4361 struct buffer_head *bh;
ac27a0ec 4362 journal_t *journal;
617ba13b
MC		 4363 ext4_fsblk_t start;
4364 ext4_fsblk_t len;
ac27a0ec 4365 int hblock, blocksize;
617ba13b 4366 ext4_fsblk_t sb_block;
ac27a0ec 4367 unsigned long offset;
2b2d6d01 4368 struct ext4_super_block *es;
ac27a0ec
DK		 4369 struct block_device *bdev;
4370
e2b911c5 4371 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4372
b31e1552 4373 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4374 if (bdev == NULL)
4375 return NULL;
4376
ac27a0ec 4377 blocksize = sb->s_blocksize;
e1defc4f 4378 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4379 if (blocksize < hblock) {
b31e1552
ES		 4380 ext4_msg(sb, KERN_ERR,
4381 "blocksize too small for journal device");
ac27a0ec
DK		 4382 goto out_bdev;
4383 }
4384
617ba13b
MC		 4385 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4386 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4387 set_blocksize(bdev, blocksize);
4388 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4389 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4390 "external journal");
ac27a0ec
DK		 4391 goto out_bdev;
4392 }
4393
2716b802 4394 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4395 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4396 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4397 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4398 ext4_msg(sb, KERN_ERR, "external journal has "
4399 "bad superblock");
ac27a0ec
DK		 4400 brelse(bh);
4401 goto out_bdev;
4402 }
4403
df4763be
DW		 4404 if ((le32_to_cpu(es->s_feature_ro_compat) &
4405 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4406 es->s_checksum != ext4_superblock_csum(sb, es)) {
4407 ext4_msg(sb, KERN_ERR, "external journal has "
4408 "corrupt superblock");
4409 brelse(bh);
4410 goto out_bdev;
4411 }
4412
617ba13b 4413 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4414 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4415 brelse(bh);
4416 goto out_bdev;
4417 }
4418
bd81d8ee 4419 len = ext4_blocks_count(es);
ac27a0ec
DK		 4420 start = sb_block + 1;
4421 brelse(bh); /* we're done with the superblock */
4422
dab291af 4423 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4424 start, len, blocksize);
4425 if (!journal) {
b31e1552 4426 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4427 goto out_bdev;
4428 }
4429 journal->j_private = sb;
dfec8a14 4430 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4431 wait_on_buffer(journal->j_sb_buffer);
4432 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4433 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4434 goto out_journal;
4435 }
4436 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4437 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4438 "user (unsupported) - %d",
ac27a0ec
DK		 4439 be32_to_cpu(journal->j_superblock->s_nr_users));
4440 goto out_journal;
4441 }
617ba13b
MC		 4442 EXT4_SB(sb)->journal_bdev = bdev;
4443 ext4_init_journal_params(sb, journal);
ac27a0ec 4444 return journal;
0b8e58a1 4445
ac27a0ec 4446out_journal:
dab291af 4447 jbd2_journal_destroy(journal);
ac27a0ec 4448out_bdev:
617ba13b 4449 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4450 return NULL;
4451}
4452
617ba13b
MC		 4453static int ext4_load_journal(struct super_block *sb,
4454 struct ext4_super_block *es,
ac27a0ec
DK		 4455 unsigned long journal_devnum)
4456{
4457 journal_t *journal;
4458 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4459 dev_t journal_dev;
4460 int err = 0;
4461 int really_read_only;
4462
e2b911c5 4463 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4464
ac27a0ec
DK		 4465 if (journal_devnum &&
4466 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4467 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4468 "numbers have changed");
ac27a0ec
DK		 4469 journal_dev = new_decode_dev(journal_devnum);
4470 } else
4471 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4472
4473 really_read_only = bdev_read_only(sb->s_bdev);
4474
4475 /*
4476 * Are we loading a blank journal or performing recovery after a
4477 * crash? For recovery, we need to check in advance whether we
4478 * can get read-write access to the device.
4479 */
e2b911c5 4480 if (ext4_has_feature_journal_needs_recovery(sb)) {
ac27a0ec 4481 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4482 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4483 "required on readonly filesystem");
ac27a0ec 4484 if (really_read_only) {
b31e1552
ES		 4485 ext4_msg(sb, KERN_ERR, "write access "
4486 "unavailable, cannot proceed");
ac27a0ec
DK		 4487 return -EROFS;
4488 }
b31e1552
ES		 4489 ext4_msg(sb, KERN_INFO, "write access will "
4490 "be enabled during recovery");
ac27a0ec
DK		 4491 }
4492 }
4493
4494 if (journal_inum && journal_dev) {
b31e1552
ES		 4495 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4496 "and inode journals!");
ac27a0ec
DK		 4497 return -EINVAL;
4498 }
4499
4500 if (journal_inum) {
617ba13b 4501 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4502 return -EINVAL;
4503 } else {
617ba13b 4504 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4505 return -EINVAL;
4506 }
4507
90576c0b 4508 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4509 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4510
e2b911c5 4511 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4512 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4513 if (!err) {
4514 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4515 if (save)
4516 memcpy(save, ((char *) es) +
4517 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4518 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4519 if (save)
4520 memcpy(((char *) es) + EXT4_S_ERR_START,
4521 save, EXT4_S_ERR_LEN);
4522 kfree(save);
4523 }
ac27a0ec
DK		 4524
4525 if (err) {
b31e1552 4526 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4527 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4528 return err;
4529 }
4530
617ba13b
MC		 4531 EXT4_SB(sb)->s_journal = journal;
4532 ext4_clear_journal_err(sb, es);
ac27a0ec 4533
c41303ce 4534 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4535 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4536 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4537
4538 /* Make sure we flush the recovery flag to disk. */
e2d67052 4539 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4540 }
4541
4542 return 0;
4543}
4544
e2d67052 4545static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4546{
e2d67052 4547 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4548 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4549 int error = 0;
ac27a0ec 4550
bdfe0cbd 4551 if (!sbh || block_device_ejected(sb))
c4be0c1d 4552 return error;
71290b36
TT		 4553 /*
4554 * If the file system is mounted read-only, don't update the
4555 * superblock write time. This avoids updating the superblock
4556 * write time when we are mounting the root file system
4557 * read/only but we need to replay the journal; at that point,
4558 * for people who are east of GMT and who make their clock
4559 * tick in localtime for Windows bug-for-bug compatibility,
4560 * the clock is set in the future, and this will cause e2fsck
4561 * to complain and force a full file system check.
4562 */
4563 if (!(sb->s_flags & MS_RDONLY))
4564 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4565 if (sb->s_bdev->bd_part)
4566 es->s_kbytes_written =
4567 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4568 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4569 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4570 else
4571 es->s_kbytes_written =
4572 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4573 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4574 ext4_free_blocks_count_set(es,
57042651
TT		 4575 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4576 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4577 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4578 es->s_free_inodes_count =
4579 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4580 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4581 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4582 ext4_superblock_csum_set(sb);
1566a48a
TT		 4583 if (sync)
4584 lock_buffer(sbh);
4743f839
PKS		 4585 if (buffer_write_io_error(sbh)) {
4586 /*
4587 * Oh, dear. A previous attempt to write the
4588 * superblock failed. This could happen because the
4589 * USB device was yanked out. Or it could happen to
4590 * be a transient write error and maybe the block will
4591 * be remapped. Nothing we can do but to retry the
4592 * write and hope for the best.
4593 */
4594 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4595 "superblock detected");
4596 clear_buffer_write_io_error(sbh);
4597 set_buffer_uptodate(sbh);
4598 }
ac27a0ec 4599 mark_buffer_dirty(sbh);
914258bf 4600 if (sync) {
1566a48a 4601 unlock_buffer(sbh);
564bc402
DJ		 4602 error = __sync_dirty_buffer(sbh,
4603 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
c4be0c1d
TS		 4604 if (error)
4605 return error;
4606
4607 error = buffer_write_io_error(sbh);
4608 if (error) {
b31e1552
ES		 4609 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4610 "superblock");
914258bf
TT		 4611 clear_buffer_write_io_error(sbh);
4612 set_buffer_uptodate(sbh);
4613 }
4614 }
c4be0c1d 4615 return error;
ac27a0ec
DK		 4616}
4617
ac27a0ec
DK		 4618/*
4619 * Have we just finished recovery? If so, and if we are mounting (or
4620 * remounting) the filesystem readonly, then we will end up with a
4621 * consistent fs on disk. Record that fact.
4622 */
2b2d6d01
TT		 4623static void ext4_mark_recovery_complete(struct super_block *sb,
4624 struct ext4_super_block *es)
ac27a0ec 4625{
617ba13b 4626 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4627
e2b911c5 4628 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4629 BUG_ON(journal != NULL);
4630 return;
4631 }
dab291af 4632 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4633 if (jbd2_journal_flush(journal) < 0)
4634 goto out;
4635
e2b911c5 4636 if (ext4_has_feature_journal_needs_recovery(sb) &&
ac27a0ec 4637 sb->s_flags & MS_RDONLY) {
e2b911c5 4638 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4639 ext4_commit_super(sb, 1);
ac27a0ec 4640 }
7ffe1ea8
HK		 4641
4642out:
dab291af 4643 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4644}
4645
4646/*
4647 * If we are mounting (or read-write remounting) a filesystem whose journal
4648 * has recorded an error from a previous lifetime, move that error to the
4649 * main filesystem now.
4650 */
2b2d6d01
TT		 4651static void ext4_clear_journal_err(struct super_block *sb,
4652 struct ext4_super_block *es)
ac27a0ec
DK		 4653{
4654 journal_t *journal;
4655 int j_errno;
4656 const char *errstr;
4657
e2b911c5 4658 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4659
617ba13b 4660 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4661
4662 /*
4663 * Now check for any error status which may have been recorded in the
617ba13b 4664 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4665 */
4666
dab291af 4667 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4668 if (j_errno) {
4669 char nbuf[16];
4670
617ba13b 4671 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4672 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4673 "from previous mount: %s", errstr);
12062ddd 4674 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4675
617ba13b
MC		 4676 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4677 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4678 ext4_commit_super(sb, 1);
ac27a0ec 4679
dab291af 4680 jbd2_journal_clear_err(journal);
d796c52e 4681 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4682 }
4683}
4684
4685/*
4686 * Force the running and committing transactions to commit,
4687 * and wait on the commit.
4688 */
617ba13b 4689int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4690{
4691 journal_t *journal;
ac27a0ec
DK		 4692
4693 if (sb->s_flags & MS_RDONLY)
4694 return 0;
4695
617ba13b 4696 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4697 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4698}
4699
617ba13b 4700static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4701{
14ce0cb4 4702 int ret = 0;
9eddacf9 4703 tid_t target;
06a407f1 4704 bool needs_barrier = false;
8d5d02e6 4705 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4706
9bffad1e 4707 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4708 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4709 /*
4710 * Writeback quota in non-journalled quota case - journalled quota has
4711 * no dirty dquots
4712 */
4713 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4714 /*
4715 * Data writeback is possible w/o journal transaction, so barrier must
4716 * being sent at the end of the function. But we can skip it if
4717 * transaction_commit will do it for us.
4718 */
bda32530
TT		 4719 if (sbi->s_journal) {
4720 target = jbd2_get_latest_transaction(sbi->s_journal);
4721 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4722 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4723 needs_barrier = true;
4724
4725 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4726 if (wait)
4727 ret = jbd2_log_wait_commit(sbi->s_journal,
4728 target);
4729 }
4730 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4731 needs_barrier = true;
06a407f1
DM		 4732 if (needs_barrier) {
4733 int err;
4734 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4735 if (!ret)
4736 ret = err;
0390131b 4737 }
06a407f1
DM		 4738
4739 return ret;
4740}
4741
ac27a0ec
DK		 4742/*
4743 * LVM calls this function before a (read-only) snapshot is created. This
4744 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4745 *
4746 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4747 * state independently. It relies on upper layer to stop all data & metadata
4748 * modifications.
ac27a0ec 4749 */
c4be0c1d 4750static int ext4_freeze(struct super_block *sb)
ac27a0ec 4751{
c4be0c1d
TS		 4752 int error = 0;
4753 journal_t *journal;
ac27a0ec 4754
9ca92389
TT		 4755 if (sb->s_flags & MS_RDONLY)
4756 return 0;
ac27a0ec 4757
9ca92389 4758 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4759
bb044576
TT		 4760 if (journal) {
4761 /* Now we set up the journal barrier. */
4762 jbd2_journal_lock_updates(journal);
ac27a0ec 4763
bb044576
TT		 4764 /*
4765 * Don't clear the needs_recovery flag if we failed to
4766 * flush the journal.
4767 */
4768 error = jbd2_journal_flush(journal);
4769 if (error < 0)
4770 goto out;
c642dc9e
ES		 4771
4772 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4773 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4774 }
9ca92389 4775
9ca92389 4776 error = ext4_commit_super(sb, 1);
6b0310fb 4777out:
bb044576
TT		 4778 if (journal)
4779 /* we rely on upper layer to stop further updates */
4780 jbd2_journal_unlock_updates(journal);
6b0310fb 4781 return error;
ac27a0ec
DK		 4782}
4783
4784/*
4785 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4786 * flag here, even though the filesystem is not technically dirty yet.
4787 */
c4be0c1d 4788static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4789{
9ca92389
TT		 4790 if (sb->s_flags & MS_RDONLY)
4791 return 0;
4792
c642dc9e
ES		 4793 if (EXT4_SB(sb)->s_journal) {
4794 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4795 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4796 }
4797
9ca92389 4798 ext4_commit_super(sb, 1);
c4be0c1d 4799 return 0;
ac27a0ec
DK		 4800}
4801
673c6100
TT		 4802/*
4803 * Structure to save mount options for ext4_remount's benefit
4804 */
4805struct ext4_mount_options {
4806 unsigned long s_mount_opt;
a2595b8a 4807 unsigned long s_mount_opt2;
08cefc7a
EB		 4808 kuid_t s_resuid;
4809 kgid_t s_resgid;
673c6100
TT		 4810 unsigned long s_commit_interval;
4811 u32 s_min_batch_time, s_max_batch_time;
4812#ifdef CONFIG_QUOTA
4813 int s_jquota_fmt;
a2d4a646 4814 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4815#endif
4816};
4817
2b2d6d01 4818static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4819{
2b2d6d01 4820 struct ext4_super_block *es;
617ba13b 4821 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4822 unsigned long old_sb_flags;
617ba13b 4823 struct ext4_mount_options old_opts;
c79d967d 4824 int enable_quota = 0;
8a266467 4825 ext4_group_t g;
b3881f74 4826 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4827 int err = 0;
ac27a0ec 4828#ifdef CONFIG_QUOTA
03dafb5f 4829 int i, j;
ac27a0ec 4830#endif
d4c402d9 4831 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4832
4833 /* Store the original options */
4834 old_sb_flags = sb->s_flags;
4835 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4836 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4837 old_opts.s_resuid = sbi->s_resuid;
4838 old_opts.s_resgid = sbi->s_resgid;
4839 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4840 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4841 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4842#ifdef CONFIG_QUOTA
4843 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4844 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4845 if (sbi->s_qf_names[i]) {
4846 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4847 GFP_KERNEL);
4848 if (!old_opts.s_qf_names[i]) {
4849 for (j = 0; j < i; j++)
4850 kfree(old_opts.s_qf_names[j]);
3e36a163 4851 kfree(orig_data);
03dafb5f
CG		 4852 return -ENOMEM;
4853 }
4854 } else
4855 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4856#endif
b3881f74
TT		 4857 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4858 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4859
661aa520 4860 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4861 err = -EINVAL;
4862 goto restore_opts;
4863 }
4864
6b992ff2 4865 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4866 test_opt(sb, JOURNAL_CHECKSUM)) {
4867 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4868 "during remount not supported; ignoring");
4869 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4870 }
4871
6ae6514b
PS		 4872 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4873 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4874 ext4_msg(sb, KERN_ERR, "can't mount with "
4875 "both data=journal and delalloc");
4876 err = -EINVAL;
4877 goto restore_opts;
4878 }
4879 if (test_opt(sb, DIOREAD_NOLOCK)) {
4880 ext4_msg(sb, KERN_ERR, "can't mount with "
4881 "both data=journal and dioread_nolock");
4882 err = -EINVAL;
4883 goto restore_opts;
4884 }
923ae0ff
RZ		 4885 if (test_opt(sb, DAX)) {
4886 ext4_msg(sb, KERN_ERR, "can't mount with "
4887 "both data=journal and dax");
4888 err = -EINVAL;
4889 goto restore_opts;
4890 }
4891 }
4892
4893 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4894 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4895 "dax flag with busy inodes while remounting");
4896 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4897 }
4898
4ab2f15b 4899 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4900 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4901
4902 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4903 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4904
4905 es = sbi->s_es;
4906
b3881f74 4907 if (sbi->s_journal) {
0390131b 4908 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4909 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4910 }
ac27a0ec 4911
a2fd66d0
TT		 4912 if (*flags & MS_LAZYTIME)
4913 sb->s_flags |= MS_LAZYTIME;
4914
661aa520 4915 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4916 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4917 err = -EROFS;
4918 goto restore_opts;
4919 }
4920
4921 if (*flags & MS_RDONLY) {
38c03b34
TT		 4922 err = sync_filesystem(sb);
4923 if (err < 0)
4924 goto restore_opts;
0f0dd62f
CH		 4925 err = dquot_suspend(sb, -1);
4926 if (err < 0)
c79d967d 4927 goto restore_opts;
c79d967d 4928
ac27a0ec
DK		 4929 /*
4930 * First of all, the unconditional stuff we have to do
4931 * to disable replay of the journal when we next remount
4932 */
4933 sb->s_flags |= MS_RDONLY;
4934
4935 /*
4936 * OK, test if we are remounting a valid rw partition
4937 * readonly, and if so set the rdonly flag and then
4938 * mark the partition as valid again.
4939 */
617ba13b
MC		 4940 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4941 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4942 es->s_state = cpu_to_le16(sbi->s_mount_state);
4943
a63c9eb2 4944 if (sbi->s_journal)
0390131b 4945 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4946 } else {
a13fb1a4 4947 /* Make sure we can mount this feature set readwrite */
e2b911c5 4948 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 4949 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4950 err = -EROFS;
4951 goto restore_opts;
4952 }
8a266467
TT		 4953 /*
4954 * Make sure the group descriptor checksums
0b8e58a1 4955 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4956 */
4957 for (g = 0; g < sbi->s_groups_count; g++) {
4958 struct ext4_group_desc *gdp =
4959 ext4_get_group_desc(sb, g, NULL);
4960
feb0ab32 4961 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4962 ext4_msg(sb, KERN_ERR,
4963 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 4964 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 4965 le16_to_cpu(gdp->bg_checksum));
6a797d27 4966 err = -EFSBADCRC;
8a266467
TT		 4967 goto restore_opts;
4968 }
4969 }
4970
ead6596b
ES		 4971 /*
4972 * If we have an unprocessed orphan list hanging
4973 * around from a previously readonly bdev mount,
4974 * require a full umount/remount for now.
4975 */
4976 if (es->s_last_orphan) {
b31e1552 4977 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4978 "remount RDWR because of unprocessed "
4979 "orphan inode list. Please "
b31e1552 4980 "umount/remount instead");
ead6596b
ES		 4981 err = -EINVAL;
4982 goto restore_opts;
4983 }
4984
ac27a0ec
DK		 4985 /*
4986 * Mounting a RDONLY partition read-write, so reread
4987 * and store the current valid flag. (It may have
4988 * been changed by e2fsck since we originally mounted
4989 * the partition.)
4990 */
0390131b
FM		 4991 if (sbi->s_journal)
4992 ext4_clear_journal_err(sb, es);
ac27a0ec 4993 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4994 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4995 sb->s_flags &= ~MS_RDONLY;
e2b911c5 4996 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 4997 if (ext4_multi_mount_protect(sb,
4998 le64_to_cpu(es->s_mmp_block))) {
4999 err = -EROFS;
5000 goto restore_opts;
5001 }
c79d967d 5002 enable_quota = 1;
ac27a0ec
DK		 5003 }
5004 }
bfff6873
LC		 5005
5006 /*
5007 * Reinitialize lazy itable initialization thread based on
5008 * current settings
5009 */
5010 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5011 ext4_unregister_li_request(sb);
5012 else {
5013 ext4_group_t first_not_zeroed;
5014 first_not_zeroed = ext4_has_uninit_itable(sb);
5015 ext4_register_li_request(sb, first_not_zeroed);
5016 }
5017
6fd058f7 5018 ext4_setup_system_zone(sb);
d096ad0f 5019 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5020 ext4_commit_super(sb, 1);
0390131b 5021
ac27a0ec
DK		 5022#ifdef CONFIG_QUOTA
5023 /* Release old quota file names */
a2d4a646 5024 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5025 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5026 if (enable_quota) {
5027 if (sb_any_quota_suspended(sb))
5028 dquot_resume(sb, -1);
e2b911c5 5029 else if (ext4_has_feature_quota(sb)) {
7c319d32 5030 err = ext4_enable_quotas(sb);
07724f98 5031 if (err)
7c319d32 5032 goto restore_opts;
7c319d32
AK		 5033 }
5034 }
ac27a0ec 5035#endif
d4c402d9 5036
a26f4992 5037 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5038 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5039 kfree(orig_data);
ac27a0ec 5040 return 0;
0b8e58a1 5041
ac27a0ec
DK		 5042restore_opts:
5043 sb->s_flags = old_sb_flags;
5044 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5045 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5046 sbi->s_resuid = old_opts.s_resuid;
5047 sbi->s_resgid = old_opts.s_resgid;
5048 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5049 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5050 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5051#ifdef CONFIG_QUOTA
5052 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5053 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5054 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5055 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5056 }
5057#endif
d4c402d9 5058 kfree(orig_data);
ac27a0ec
DK		 5059 return err;
5060}
5061
689c958c
LX		 5062#ifdef CONFIG_QUOTA
5063static int ext4_statfs_project(struct super_block *sb,
5064 kprojid_t projid, struct kstatfs *buf)
5065{
5066 struct kqid qid;
5067 struct dquot *dquot;
5068 u64 limit;
5069 u64 curblock;
5070
5071 qid = make_kqid_projid(projid);
5072 dquot = dqget(sb, qid);
5073 if (IS_ERR(dquot))
5074 return PTR_ERR(dquot);
5075 spin_lock(&dq_data_lock);
5076
5077 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5078 dquot->dq_dqb.dqb_bsoftlimit :
5079 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5080 if (limit && buf->f_blocks > limit) {
5081 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5082 buf->f_blocks = limit;
5083 buf->f_bfree = buf->f_bavail =
5084 (buf->f_blocks > curblock) ?
5085 (buf->f_blocks - curblock) : 0;
5086 }
5087
5088 limit = dquot->dq_dqb.dqb_isoftlimit ?
5089 dquot->dq_dqb.dqb_isoftlimit :
5090 dquot->dq_dqb.dqb_ihardlimit;
5091 if (limit && buf->f_files > limit) {
5092 buf->f_files = limit;
5093 buf->f_ffree =
5094 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5095 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5096 }
5097
5098 spin_unlock(&dq_data_lock);
5099 dqput(dquot);
5100 return 0;
5101}
5102#endif
5103
2b2d6d01 5104static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5105{
5106 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5107 struct ext4_sb_info *sbi = EXT4_SB(sb);
5108 struct ext4_super_block *es = sbi->s_es;
27dd4385 5109 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5110 u64 fsid;
d02a9391 5111 s64 bfree;
27dd4385 5112 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5113
952fc18e
TT		 5114 if (!test_opt(sb, MINIX_DF))
5115 overhead = sbi->s_overhead;
ac27a0ec 5116
617ba13b 5117 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5118 buf->f_bsize = sb->s_blocksize;
b72f78cb 5119 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5120 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5121 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5122 /* prevent underflow in case that few free space is available */
57042651 5123 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5124 buf->f_bavail = buf->f_bfree -
5125 (ext4_r_blocks_count(es) + resv_blocks);
5126 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5127 buf->f_bavail = 0;
5128 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5129 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5130 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5131 fsid = le64_to_cpup((void *)es->s_uuid) ^
5132 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5133 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5134 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5135
689c958c
LX		 5136#ifdef CONFIG_QUOTA
5137 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5138 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5139 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5140#endif
ac27a0ec
DK		 5141 return 0;
5142}
5143
0b8e58a1
AD		 5144/* Helper function for writing quotas on sync - we need to start transaction
5145 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5146 * Process 1 Process 2
617ba13b 5147 * ext4_create() quota_sync()
a269eb18 5148 * jbd2_journal_start() write_dquot()
871a2931 5149 * dquot_initialize() down(dqio_mutex)
dab291af 5150 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5151 *
5152 */
5153
5154#ifdef CONFIG_QUOTA
5155
5156static inline struct inode *dquot_to_inode(struct dquot *dquot)
5157{
4c376dca 5158 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5159}
5160
617ba13b 5161static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5162{
5163 int ret, err;
5164 handle_t *handle;
5165 struct inode *inode;
5166
5167 inode = dquot_to_inode(dquot);
9924a92a 5168 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5169 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5170 if (IS_ERR(handle))
5171 return PTR_ERR(handle);
5172 ret = dquot_commit(dquot);
617ba13b 5173 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5174 if (!ret)
5175 ret = err;
5176 return ret;
5177}
5178
617ba13b 5179static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5180{
5181 int ret, err;
5182 handle_t *handle;
5183
9924a92a 5184 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5185 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5186 if (IS_ERR(handle))
5187 return PTR_ERR(handle);
5188 ret = dquot_acquire(dquot);
617ba13b 5189 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5190 if (!ret)
5191 ret = err;
5192 return ret;
5193}
5194
617ba13b 5195static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5196{
5197 int ret, err;
5198 handle_t *handle;
5199
9924a92a 5200 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5201 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5202 if (IS_ERR(handle)) {
5203 /* Release dquot anyway to avoid endless cycle in dqput() */
5204 dquot_release(dquot);
ac27a0ec 5205 return PTR_ERR(handle);
9c3013e9 5206 }
ac27a0ec 5207 ret = dquot_release(dquot);
617ba13b 5208 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5209 if (!ret)
5210 ret = err;
5211 return ret;
5212}
5213
617ba13b 5214static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5215{
262b4662
JK		 5216 struct super_block *sb = dquot->dq_sb;
5217 struct ext4_sb_info *sbi = EXT4_SB(sb);
5218
2c8be6b2 5219 /* Are we journaling quotas? */
e2b911c5 5220 if (ext4_has_feature_quota(sb) ||
262b4662 5221 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5222 dquot_mark_dquot_dirty(dquot);
617ba13b 5223 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5224 } else {
5225 return dquot_mark_dquot_dirty(dquot);
5226 }
5227}
5228
617ba13b 5229static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5230{
5231 int ret, err;
5232 handle_t *handle;
5233
5234 /* Data block + inode block */
2b0143b5 5235 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5236 if (IS_ERR(handle))
5237 return PTR_ERR(handle);
5238 ret = dquot_commit_info(sb, type);
617ba13b 5239 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5240 if (!ret)
5241 ret = err;
5242 return ret;
5243}
5244
5245/*
5246 * Turn on quotas during mount time - we need to find
5247 * the quota file and such...
5248 */
617ba13b 5249static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5250{
287a8095
CH		 5251 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5252 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5253}
5254
daf647d2
TT		 5255static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5256{
5257 struct ext4_inode_info *ei = EXT4_I(inode);
5258
5259 /* The first argument of lockdep_set_subclass has to be
5260 * *exactly* the same as the argument to init_rwsem() --- in
5261 * this case, in init_once() --- or lockdep gets unhappy
5262 * because the name of the lock is set using the
5263 * stringification of the argument to init_rwsem().
5264 */
5265 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5266 lockdep_set_subclass(&ei->i_data_sem, subclass);
5267}
5268
ac27a0ec
DK		 5269/*
5270 * Standard function to be called on quota_on
5271 */
617ba13b 5272static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5273 struct path *path)
ac27a0ec
DK		 5274{
5275 int err;
ac27a0ec
DK		 5276
5277 if (!test_opt(sb, QUOTA))
5278 return -EINVAL;
0623543b 5279
ac27a0ec 5280 /* Quotafile not on the same filesystem? */
d8c9584e 5281 if (path->dentry->d_sb != sb)
ac27a0ec 5282 return -EXDEV;
0623543b
JK		 5283 /* Journaling quota? */
5284 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5285 /* Quotafile not in fs root? */
f00c9e44 5286 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5287 ext4_msg(sb, KERN_WARNING,
5288 "Quota file not on filesystem root. "
5289 "Journaled quota will not work");
2b2d6d01 5290 }
0623543b
JK		 5291
5292 /*
5293 * When we journal data on quota file, we have to flush journal to see
5294 * all updates to the file when we bypass pagecache...
5295 */
0390131b 5296 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5297 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5298 /*
5299 * We don't need to lock updates but journal_flush() could
5300 * otherwise be livelocked...
5301 */
5302 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5303 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5304 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5305 if (err)
7ffe1ea8 5306 return err;
0623543b 5307 }
daf647d2
TT		 5308 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5309 err = dquot_quota_on(sb, type, format_id, path);
5310 if (err)
5311 lockdep_set_quota_inode(path->dentry->d_inode,
5312 I_DATA_SEM_NORMAL);
5313 return err;
ac27a0ec
DK		 5314}
5315
7c319d32
AK		 5316static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5317 unsigned int flags)
5318{
5319 int err;
5320 struct inode *qf_inode;
a2d4a646 5321 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5322 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5323 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5324 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5325 };
5326
e2b911c5 5327 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5328
5329 if (!qf_inums[type])
5330 return -EPERM;
5331
5332 qf_inode = ext4_iget(sb, qf_inums[type]);
5333 if (IS_ERR(qf_inode)) {
5334 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5335 return PTR_ERR(qf_inode);
5336 }
5337
bcb13850
JK		 5338 /* Don't account quota for quota files to avoid recursion */
5339 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5340 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5341 err = dquot_enable(qf_inode, type, format_id, flags);
5342 iput(qf_inode);
daf647d2
TT		 5343 if (err)
5344 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5345
5346 return err;
5347}
5348
5349/* Enable usage tracking for all quota types. */
5350static int ext4_enable_quotas(struct super_block *sb)
5351{
5352 int type, err = 0;
a2d4a646 5353 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5354 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5355 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5356 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5357 };
49da9392
JK		 5358 bool quota_mopt[EXT4_MAXQUOTAS] = {
5359 test_opt(sb, USRQUOTA),
5360 test_opt(sb, GRPQUOTA),
5361 test_opt(sb, PRJQUOTA),
5362 };
7c319d32
AK		 5363
5364 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5365 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5366 if (qf_inums[type]) {
5367 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5368 DQUOT_USAGE_ENABLED |
5369 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 5370 if (err) {
5371 ext4_warning(sb,
72ba7450
TT		 5372 "Failed to enable quota tracking "
5373 "(type=%d, err=%d). Please run "
5374 "e2fsck to fix.", type, err);
7c319d32
AK		 5375 return err;
5376 }
5377 }
5378 }
5379 return 0;
5380}
5381
ca0e05e4
DM		 5382static int ext4_quota_off(struct super_block *sb, int type)
5383{
21f97697
JK		 5384 struct inode *inode = sb_dqopt(sb)->files[type];
5385 handle_t *handle;
5386
87009d86
DM		 5387 /* Force all delayed allocation blocks to be allocated.
5388 * Caller already holds s_umount sem */
5389 if (test_opt(sb, DELALLOC))
ca0e05e4 5390 sync_filesystem(sb);
ca0e05e4 5391
0b268590
AG		 5392 if (!inode)
5393 goto out;
5394
21f97697
JK		 5395 /* Update modification times of quota files when userspace can
5396 * start looking at them */
9924a92a 5397 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5398 if (IS_ERR(handle))
5399 goto out;
eeca7ea1 5400 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5401 ext4_mark_inode_dirty(handle, inode);
5402 ext4_journal_stop(handle);
5403
5404out:
ca0e05e4
DM		 5405 return dquot_quota_off(sb, type);
5406}
5407
ac27a0ec
DK		 5408/* Read data from quotafile - avoid pagecache and such because we cannot afford
5409 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5410 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5411 * we don't have to be afraid of races */
617ba13b 5412static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5413 size_t len, loff_t off)
5414{
5415 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5416 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5417 int offset = off & (sb->s_blocksize - 1);
5418 int tocopy;
5419 size_t toread;
5420 struct buffer_head *bh;
5421 loff_t i_size = i_size_read(inode);
5422
5423 if (off > i_size)
5424 return 0;
5425 if (off+len > i_size)
5426 len = i_size-off;
5427 toread = len;
5428 while (toread > 0) {
5429 tocopy = sb->s_blocksize - offset < toread ?
5430 sb->s_blocksize - offset : toread;
1c215028
TT		 5431 bh = ext4_bread(NULL, inode, blk, 0);
5432 if (IS_ERR(bh))
5433 return PTR_ERR(bh);
ac27a0ec
DK		 5434 if (!bh) /* A hole? */
5435 memset(data, 0, tocopy);
5436 else
5437 memcpy(data, bh->b_data+offset, tocopy);
5438 brelse(bh);
5439 offset = 0;
5440 toread -= tocopy;
5441 data += tocopy;
5442 blk++;
5443 }
5444 return len;
5445}
5446
5447/* Write to quotafile (we know the transaction is already started and has
5448 * enough credits) */
617ba13b 5449static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5450 const char *data, size_t len, loff_t off)
5451{
5452 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5453 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5454 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5455 int retries = 0;
ac27a0ec
DK		 5456 struct buffer_head *bh;
5457 handle_t *handle = journal_current_handle();
5458
0390131b 5459 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5460 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5461 " cancelled because transaction is not started",
9c3013e9
JK		 5462 (unsigned long long)off, (unsigned long long)len);
5463 return -EIO;
5464 }
67eeb568
DM		 5465 /*
5466 * Since we account only one data block in transaction credits,
5467 * then it is impossible to cross a block boundary.
5468 */
5469 if (sb->s_blocksize - offset < len) {
5470 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5471 " cancelled because not block aligned",
5472 (unsigned long long)off, (unsigned long long)len);
5473 return -EIO;
5474 }
5475
c5e298ae
TT		 5476 do {
5477 bh = ext4_bread(handle, inode, blk,
5478 EXT4_GET_BLOCKS_CREATE |
5479 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5480 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5481 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5482 if (IS_ERR(bh))
5483 return PTR_ERR(bh);
67eeb568
DM		 5484 if (!bh)
5485 goto out;
5d601255 5486 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5487 err = ext4_journal_get_write_access(handle, bh);
5488 if (err) {
5489 brelse(bh);
1c215028 5490 return err;
ac27a0ec 5491 }
67eeb568
DM		 5492 lock_buffer(bh);
5493 memcpy(bh->b_data+offset, data, len);
5494 flush_dcache_page(bh->b_page);
5495 unlock_buffer(bh);
62d2b5f2 5496 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5497 brelse(bh);
ac27a0ec 5498out:
67eeb568
DM		 5499 if (inode->i_size < off + len) {
5500 i_size_write(inode, off + len);
617ba13b 5501 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5502 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5503 }
67eeb568 5504 return len;
ac27a0ec
DK		 5505}
5506
8f0e8746
TT		 5507static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5508{
5509 const struct quota_format_ops *ops;
5510
5511 if (!sb_has_quota_loaded(sb, qid->type))
5512 return -ESRCH;
5513 ops = sb_dqopt(sb)->ops[qid->type];
5514 if (!ops || !ops->get_next_id)
5515 return -ENOSYS;
5516 return dquot_get_next_id(sb, qid);
5517}
ac27a0ec
DK		 5518#endif
5519
152a0836
AV		 5520static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5521 const char *dev_name, void *data)
ac27a0ec 5522{
152a0836 5523 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5524}
5525
c290ea01 5526#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5527static inline void register_as_ext2(void)
5528{
5529 int err = register_filesystem(&ext2_fs_type);
5530 if (err)
5531 printk(KERN_WARNING
5532 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5533}
5534
5535static inline void unregister_as_ext2(void)
5536{
5537 unregister_filesystem(&ext2_fs_type);
5538}
2035e776
TT		 5539
5540static inline int ext2_feature_set_ok(struct super_block *sb)
5541{
e2b911c5 5542 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776
TT		 5543 return 0;
5544 if (sb->s_flags & MS_RDONLY)
5545 return 1;
e2b911c5 5546 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5547 return 0;
5548 return 1;
5549}
24b58424
TT		 5550#else
5551static inline void register_as_ext2(void) { }
5552static inline void unregister_as_ext2(void) { }
2035e776 5553static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5554#endif
5555
24b58424
TT		 5556static inline void register_as_ext3(void)
5557{
5558 int err = register_filesystem(&ext3_fs_type);
5559 if (err)
5560 printk(KERN_WARNING
5561 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5562}
5563
5564static inline void unregister_as_ext3(void)
5565{
5566 unregister_filesystem(&ext3_fs_type);
5567}
2035e776
TT		 5568
5569static inline int ext3_feature_set_ok(struct super_block *sb)
5570{
e2b911c5 5571 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5572 return 0;
e2b911c5 5573 if (!ext4_has_feature_journal(sb))
2035e776
TT		 5574 return 0;
5575 if (sb->s_flags & MS_RDONLY)
5576 return 1;
e2b911c5 5577 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5578 return 0;
5579 return 1;
5580}
24b58424 5581
03010a33
TT		 5582static struct file_system_type ext4_fs_type = {
5583 .owner = THIS_MODULE,
5584 .name = "ext4",
152a0836 5585 .mount = ext4_mount,
03010a33
TT		 5586 .kill_sb = kill_block_super,
5587 .fs_flags = FS_REQUIRES_DEV,
5588};
7f78e035 5589MODULE_ALIAS_FS("ext4");
03010a33 5590
e9e3bcec
ES		 5591/* Shared across all ext4 file systems */
5592wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5593
5dabfc78 5594static int __init ext4_init_fs(void)
ac27a0ec 5595{
e9e3bcec 5596 int i, err;
c9de560d 5597
e294a537 5598 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5599 ext4_li_info = NULL;
5600 mutex_init(&ext4_li_mtx);
5601
9a4c8019 5602 /* Build-time check for flags consistency */
12e9b892 5603 ext4_check_flag_values();
e9e3bcec 5604
e142d052 5605 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5606 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5607
51865fda 5608 err = ext4_init_es();
6fd058f7
TT		 5609 if (err)
5610 return err;
51865fda
ZL		 5611
5612 err = ext4_init_pageio();
5613 if (err)
b5799018 5614 goto out5;
51865fda 5615
5dabfc78 5616 err = ext4_init_system_zone();
bd2d0210 5617 if (err)
b5799018 5618 goto out4;
857ac889 5619
b5799018 5620 err = ext4_init_sysfs();
dd68314c 5621 if (err)
b5799018 5622 goto out3;
857ac889 5623
5dabfc78 5624 err = ext4_init_mballoc();
c9de560d
AT		 5625 if (err)
5626 goto out2;
ac27a0ec
DK		 5627 err = init_inodecache();
5628 if (err)
5629 goto out1;
24b58424 5630 register_as_ext3();
2035e776 5631 register_as_ext2();
03010a33 5632 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5633 if (err)
5634 goto out;
bfff6873 5635
ac27a0ec
DK		 5636 return 0;
5637out:
24b58424
TT		 5638 unregister_as_ext2();
5639 unregister_as_ext3();
ac27a0ec
DK		 5640 destroy_inodecache();
5641out1:
5dabfc78 5642 ext4_exit_mballoc();
9c191f70 5643out2:
b5799018
TT		 5644 ext4_exit_sysfs();
5645out3:
5dabfc78 5646 ext4_exit_system_zone();
b5799018 5647out4:
5dabfc78 5648 ext4_exit_pageio();
b5799018 5649out5:
51865fda
ZL		 5650 ext4_exit_es();
5651
ac27a0ec
DK		 5652 return err;
5653}
5654
5dabfc78 5655static void __exit ext4_exit_fs(void)
ac27a0ec 5656{
bfff6873 5657 ext4_destroy_lazyinit_thread();
24b58424
TT		 5658 unregister_as_ext2();
5659 unregister_as_ext3();
03010a33 5660 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5661 destroy_inodecache();
5dabfc78 5662 ext4_exit_mballoc();
b5799018 5663 ext4_exit_sysfs();
5dabfc78
TT		 5664 ext4_exit_system_zone();
5665 ext4_exit_pageio();
dd12ed14 5666 ext4_exit_es();
ac27a0ec
DK		 5667}
5668
5669MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5670MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5671MODULE_LICENSE("GPL");
5dabfc78
TT		 5672module_init(ext4_init_fs)
5673module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository