]> git.ipfire.org Git - thirdparty/linux.git/blame - fs/ext4/super.c
projects / thirdparty / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge remote-tracking branches 'asoc/topic/samsung', 'asoc/topic/sgtl5000', 'asoc...
[thirdparty/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>
7c0f6ba6 41#include <linux/uaccess.h>
ac27a0ec 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
0db1ff22
TT		 441 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
442 return;
443
efbed4dc
TT		 444 if (ext4_error_ratelimit(sb)) {
445 va_start(args, fmt);
446 vaf.fmt = fmt;
447 vaf.va = &args;
448 printk(KERN_CRIT
449 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
450 sb->s_id, function, line, current->comm, &vaf);
451 va_end(args);
452 }
f3fc0210 453 save_error_info(sb, function, line);
617ba13b 454 ext4_handle_error(sb);
ac27a0ec
DK		 455}
456
e7c96e8e
JP		 457void __ext4_error_inode(struct inode *inode, const char *function,
458 unsigned int line, ext4_fsblk_t block,
459 const char *fmt, ...)
273df556
FM		 460{
461 va_list args;
f7c21177 462 struct va_format vaf;
1c13d5c0 463 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 464
0db1ff22
TT		 465 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
466 return;
467
1c13d5c0
TT		 468 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
469 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 470 if (ext4_error_ratelimit(inode->i_sb)) {
471 va_start(args, fmt);
472 vaf.fmt = fmt;
473 vaf.va = &args;
474 if (block)
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
476 "inode #%lu: block %llu: comm %s: %pV\n",
477 inode->i_sb->s_id, function, line, inode->i_ino,
478 block, current->comm, &vaf);
479 else
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
481 "inode #%lu: comm %s: %pV\n",
482 inode->i_sb->s_id, function, line, inode->i_ino,
483 current->comm, &vaf);
484 va_end(args);
485 }
1c13d5c0 486 save_error_info(inode->i_sb, function, line);
273df556
FM		 487 ext4_handle_error(inode->i_sb);
488}
489
e7c96e8e
JP		 490void __ext4_error_file(struct file *file, const char *function,
491 unsigned int line, ext4_fsblk_t block,
492 const char *fmt, ...)
273df556
FM		 493{
494 va_list args;
f7c21177 495 struct va_format vaf;
1c13d5c0 496 struct ext4_super_block *es;
496ad9aa 497 struct inode *inode = file_inode(file);
273df556
FM		 498 char pathname[80], *path;
499
0db1ff22
TT		 500 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
501 return;
502
1c13d5c0
TT		 503 es = EXT4_SB(inode->i_sb)->s_es;
504 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 505 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 506 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 507 if (IS_ERR(path))
508 path = "(unknown)";
509 va_start(args, fmt);
510 vaf.fmt = fmt;
511 vaf.va = &args;
512 if (block)
513 printk(KERN_CRIT
514 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
515 "block %llu: comm %s: path %s: %pV\n",
516 inode->i_sb->s_id, function, line, inode->i_ino,
517 block, current->comm, path, &vaf);
518 else
519 printk(KERN_CRIT
520 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
521 "comm %s: path %s: %pV\n",
522 inode->i_sb->s_id, function, line, inode->i_ino,
523 current->comm, path, &vaf);
524 va_end(args);
525 }
1c13d5c0 526 save_error_info(inode->i_sb, function, line);
273df556
FM		 527 ext4_handle_error(inode->i_sb);
528}
529
722887dd
TT		 530const char *ext4_decode_error(struct super_block *sb, int errno,
531 char nbuf[16])
ac27a0ec
DK		 532{
533 char *errstr = NULL;
534
535 switch (errno) {
6a797d27
DW		 536 case -EFSCORRUPTED:
537 errstr = "Corrupt filesystem";
538 break;
539 case -EFSBADCRC:
540 errstr = "Filesystem failed CRC";
541 break;
ac27a0ec
DK		 542 case -EIO:
543 errstr = "IO failure";
544 break;
545 case -ENOMEM:
546 errstr = "Out of memory";
547 break;
548 case -EROFS:
78f1ddbb
TT		 549 if (!sb || (EXT4_SB(sb)->s_journal &&
550 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 551 errstr = "Journal has aborted";
552 else
553 errstr = "Readonly filesystem";
554 break;
555 default:
556 /* If the caller passed in an extra buffer for unknown
557 * errors, textualise them now. Else we just return
558 * NULL. */
559 if (nbuf) {
560 /* Check for truncated error codes... */
561 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
562 errstr = nbuf;
563 }
564 break;
565 }
566
567 return errstr;
568}
569
617ba13b 570/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 571 * automatically and invokes the appropriate error response. */
572
c398eda0
TT		 573void __ext4_std_error(struct super_block *sb, const char *function,
574 unsigned int line, int errno)
ac27a0ec
DK		 575{
576 char nbuf[16];
577 const char *errstr;
578
0db1ff22
TT		 579 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
580 return;
581
ac27a0ec
DK		 582 /* Special case: if the error is EROFS, and we're not already
583 * inside a transaction, then there's really no point in logging
584 * an error. */
585 if (errno == -EROFS && journal_current_handle() == NULL &&
586 (sb->s_flags & MS_RDONLY))
587 return;
588
efbed4dc
TT		 589 if (ext4_error_ratelimit(sb)) {
590 errstr = ext4_decode_error(sb, errno, nbuf);
591 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
592 sb->s_id, function, line, errstr);
593 }
ac27a0ec 594
efbed4dc 595 save_error_info(sb, function, line);
617ba13b 596 ext4_handle_error(sb);
ac27a0ec
DK		 597}
598
599/*
617ba13b 600 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 601 * abort function may be used to deal with unrecoverable failures such
602 * as journal IO errors or ENOMEM at a critical moment in log management.
603 *
604 * We unconditionally force the filesystem into an ABORT|READONLY state,
605 * unless the error response on the fs has been set to panic in which
606 * case we take the easy way out and panic immediately.
607 */
608
c67d859e 609void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 610 unsigned int line, const char *fmt, ...)
ac27a0ec 611{
651e1c3b 612 struct va_format vaf;
ac27a0ec
DK		 613 va_list args;
614
0db1ff22
TT		 615 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
616 return;
617
1c13d5c0 618 save_error_info(sb, function, line);
ac27a0ec 619 va_start(args, fmt);
651e1c3b
JP		 620 vaf.fmt = fmt;
621 vaf.va = &args;
622 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
623 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 624 va_end(args);
625
1c13d5c0
TT		 626 if ((sb->s_flags & MS_RDONLY) == 0) {
627 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 628 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 629 /*
630 * Make sure updated value of ->s_mount_flags will be visible
631 * before ->s_flags update
632 */
633 smp_wmb();
634 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 635 if (EXT4_SB(sb)->s_journal)
636 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
637 save_error_info(sb, function, line);
638 }
4327ba52
DJ		 639 if (test_opt(sb, ERRORS_PANIC)) {
640 if (EXT4_SB(sb)->s_journal &&
641 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
642 return;
617ba13b 643 panic("EXT4-fs panic from previous error\n");
4327ba52 644 }
ac27a0ec
DK		 645}
646
e7c96e8e
JP		 647void __ext4_msg(struct super_block *sb,
648 const char *prefix, const char *fmt, ...)
b31e1552 649{
0ff2ea7d 650 struct va_format vaf;
b31e1552
ES		 651 va_list args;
652
efbed4dc
TT		 653 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
654 return;
655
b31e1552 656 va_start(args, fmt);
0ff2ea7d
JP		 657 vaf.fmt = fmt;
658 vaf.va = &args;
659 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 660 va_end(args);
661}
662
b03a2f7e
AD		 663#define ext4_warning_ratelimit(sb) \
664 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
665 "EXT4-fs warning")
666
12062ddd 667void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 668 unsigned int line, const char *fmt, ...)
ac27a0ec 669{
0ff2ea7d 670 struct va_format vaf;
ac27a0ec
DK		 671 va_list args;
672
b03a2f7e 673 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 674 return;
675
ac27a0ec 676 va_start(args, fmt);
0ff2ea7d
JP		 677 vaf.fmt = fmt;
678 vaf.va = &args;
679 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
680 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 681 va_end(args);
682}
683
b03a2f7e
AD		 684void __ext4_warning_inode(const struct inode *inode, const char *function,
685 unsigned int line, const char *fmt, ...)
686{
687 struct va_format vaf;
688 va_list args;
689
690 if (!ext4_warning_ratelimit(inode->i_sb))
691 return;
692
693 va_start(args, fmt);
694 vaf.fmt = fmt;
695 vaf.va = &args;
696 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
697 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
698 function, line, inode->i_ino, current->comm, &vaf);
699 va_end(args);
700}
701
e29136f8
TT		 702void __ext4_grp_locked_error(const char *function, unsigned int line,
703 struct super_block *sb, ext4_group_t grp,
704 unsigned long ino, ext4_fsblk_t block,
705 const char *fmt, ...)
5d1b1b3f
AK		 706__releases(bitlock)
707__acquires(bitlock)
708{
0ff2ea7d 709 struct va_format vaf;
5d1b1b3f
AK		 710 va_list args;
711 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
712
0db1ff22
TT		 713 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
714 return;
715
1c13d5c0
TT		 716 es->s_last_error_ino = cpu_to_le32(ino);
717 es->s_last_error_block = cpu_to_le64(block);
718 __save_error_info(sb, function, line);
0ff2ea7d 719
efbed4dc
TT		 720 if (ext4_error_ratelimit(sb)) {
721 va_start(args, fmt);
722 vaf.fmt = fmt;
723 vaf.va = &args;
724 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
725 sb->s_id, function, line, grp);
726 if (ino)
727 printk(KERN_CONT "inode %lu: ", ino);
728 if (block)
729 printk(KERN_CONT "block %llu:",
730 (unsigned long long) block);
731 printk(KERN_CONT "%pV\n", &vaf);
732 va_end(args);
733 }
5d1b1b3f
AK		 734
735 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 736 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 737 return;
738 }
1c13d5c0 739
5d1b1b3f
AK		 740 ext4_unlock_group(sb, grp);
741 ext4_handle_error(sb);
742 /*
743 * We only get here in the ERRORS_RO case; relocking the group
744 * may be dangerous, but nothing bad will happen since the
745 * filesystem will have already been marked read/only and the
746 * journal has been aborted. We return 1 as a hint to callers
747 * who might what to use the return value from
25985edc 748 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 749 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
750 * aggressively from the ext4 function in question, with a
751 * more appropriate error code.
752 */
753 ext4_lock_group(sb, grp);
754 return;
755}
756
617ba13b 757void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 758{
617ba13b 759 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 760
617ba13b 761 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 762 return;
763
12062ddd 764 ext4_warning(sb,
ac27a0ec
DK		 765 "updating to rev %d because of new feature flag, "
766 "running e2fsck is recommended",
617ba13b 767 EXT4_DYNAMIC_REV);
ac27a0ec 768
617ba13b
MC		 769 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
770 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
771 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 772 /* leave es->s_feature_*compat flags alone */
773 /* es->s_uuid will be set by e2fsck if empty */
774
775 /*
776 * The rest of the superblock fields should be zero, and if not it
777 * means they are likely already in use, so leave them alone. We
778 * can leave it up to e2fsck to clean up any inconsistencies there.
779 */
780}
781
782/*
783 * Open the external journal device
784 */
b31e1552 785static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 786{
787 struct block_device *bdev;
788 char b[BDEVNAME_SIZE];
789
d4d77629 790 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 791 if (IS_ERR(bdev))
792 goto fail;
793 return bdev;
794
795fail:
b31e1552 796 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 797 __bdevname(dev, b), PTR_ERR(bdev));
798 return NULL;
799}
800
801/*
802 * Release the journal device
803 */
4385bab1 804static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 805{
4385bab1 806 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 807}
808
4385bab1 809static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 810{
811 struct block_device *bdev;
ac27a0ec
DK		 812 bdev = sbi->journal_bdev;
813 if (bdev) {
4385bab1 814 ext4_blkdev_put(bdev);
ac27a0ec
DK		 815 sbi->journal_bdev = NULL;
816 }
ac27a0ec
DK		 817}
818
819static inline struct inode *orphan_list_entry(struct list_head *l)
820{
617ba13b 821 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 822}
823
617ba13b 824static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 825{
826 struct list_head *l;
827
b31e1552
ES		 828 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
829 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 830
831 printk(KERN_ERR "sb_info orphan list:\n");
832 list_for_each(l, &sbi->s_orphan) {
833 struct inode *inode = orphan_list_entry(l);
834 printk(KERN_ERR " "
835 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
836 inode->i_sb->s_id, inode->i_ino, inode,
837 inode->i_mode, inode->i_nlink,
838 NEXT_ORPHAN(inode));
839 }
840}
841
2b2d6d01 842static void ext4_put_super(struct super_block *sb)
ac27a0ec 843{
617ba13b
MC		 844 struct ext4_sb_info *sbi = EXT4_SB(sb);
845 struct ext4_super_block *es = sbi->s_es;
97abd7d4 846 int aborted = 0;
ef2cabf7 847 int i, err;
ac27a0ec 848
857ac889 849 ext4_unregister_li_request(sb);
e0ccfd95
CH		 850 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
851
2e8fa54e 852 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 853 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 854
0390131b 855 if (sbi->s_journal) {
97abd7d4 856 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 857 err = jbd2_journal_destroy(sbi->s_journal);
858 sbi->s_journal = NULL;
97abd7d4 859 if ((err < 0) && !aborted)
c67d859e 860 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 861 }
d4edac31 862
ebd173be 863 ext4_unregister_sysfs(sb);
d3922a77 864 ext4_es_unregister_shrinker(sbi);
9105bb14 865 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 866 ext4_release_system_zone(sb);
867 ext4_mb_release(sb);
868 ext4_ext_release(sb);
d4edac31 869
97abd7d4 870 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
e2b911c5 871 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 872 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 873 }
58c5873a 874 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 875 ext4_commit_super(sb, 1);
876
ac27a0ec
DK		 877 for (i = 0; i < sbi->s_gdb_count; i++)
878 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 879 kvfree(sbi->s_group_desc);
880 kvfree(sbi->s_flex_groups);
57042651 881 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 882 percpu_counter_destroy(&sbi->s_freeinodes_counter);
883 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 884 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c8585c6f 885 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec 886#ifdef CONFIG_QUOTA
a2d4a646 887 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 888 kfree(sbi->s_qf_names[i]);
889#endif
890
891 /* Debugging code just in case the in-memory inode orphan list
892 * isn't empty. The on-disk one can be non-empty if we've
893 * detected an error and taken the fs readonly, but the
894 * in-memory list had better be clean by this point. */
895 if (!list_empty(&sbi->s_orphan))
896 dump_orphan_list(sb, sbi);
897 J_ASSERT(list_empty(&sbi->s_orphan));
898
89d96a6f 899 sync_blockdev(sb->s_bdev);
f98393a6 900 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 901 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
902 /*
903 * Invalidate the journal device's buffers. We don't want them
904 * floating about in memory - the physical journal device may
905 * hotswapped, and it breaks the `ro-after' testing code.
906 */
907 sync_blockdev(sbi->journal_bdev);
f98393a6 908 invalidate_bdev(sbi->journal_bdev);
617ba13b 909 ext4_blkdev_remove(sbi);
ac27a0ec 910 }
9c191f70
M		 911 if (sbi->s_mb_cache) {
912 ext4_xattr_destroy_cache(sbi->s_mb_cache);
913 sbi->s_mb_cache = NULL;
914 }
c5e06d10
JL		 915 if (sbi->s_mmp_tsk)
916 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 917 brelse(sbi->s_sbh);
ac27a0ec 918 sb->s_fs_info = NULL;
3197ebdb
TT		 919 /*
920 * Now that we are completely done shutting down the
921 * superblock, we need to actually destroy the kobject.
922 */
3197ebdb
TT		 923 kobject_put(&sbi->s_kobj);
924 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 925 if (sbi->s_chksum_driver)
926 crypto_free_shash(sbi->s_chksum_driver);
705895b6 927 kfree(sbi->s_blockgroup_lock);
ac27a0ec 928 kfree(sbi);
ac27a0ec
DK		 929}
930
e18b890b 931static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 932
933/*
934 * Called inside transaction, so use GFP_NOFS
935 */
617ba13b 936static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 937{
617ba13b 938 struct ext4_inode_info *ei;
ac27a0ec 939
e6b4f8da 940 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 941 if (!ei)
942 return NULL;
0b8e58a1 943
ac27a0ec 944 ei->vfs_inode.i_version = 1;
202ee5df 945 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 946 INIT_LIST_HEAD(&ei->i_prealloc_list);
947 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 948 ext4_es_init_tree(&ei->i_es_tree);
949 rwlock_init(&ei->i_es_lock);
edaa53ca 950 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 951 ei->i_es_all_nr = 0;
edaa53ca 952 ei->i_es_shk_nr = 0;
dd475925 953 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 954 ei->i_reserved_data_blocks = 0;
955 ei->i_reserved_meta_blocks = 0;
956 ei->i_allocated_meta_blocks = 0;
9d0be502 957 ei->i_da_metadata_calc_len = 0;
7e731bc9 958 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 959 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 960#ifdef CONFIG_QUOTA
961 ei->i_reserved_quota = 0;
96c7e0d9 962 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 963#endif
8aefcd55 964 ei->jinode = NULL;
2e8fa54e 965 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 966 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 967 ei->i_sync_tid = 0;
968 ei->i_datasync_tid = 0;
e27f41e1 969 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 970 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 971 return &ei->vfs_inode;
972}
973
7ff9c073
TT		 974static int ext4_drop_inode(struct inode *inode)
975{
976 int drop = generic_drop_inode(inode);
977
978 trace_ext4_drop_inode(inode, drop);
979 return drop;
980}
981
fa0d7e3d
NP		 982static void ext4_i_callback(struct rcu_head *head)
983{
984 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 985 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
986}
987
617ba13b 988static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 989{
9f7dd93d 990 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 991 ext4_msg(inode->i_sb, KERN_ERR,
992 "Inode %lu (%p): orphan list check failed!",
993 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 994 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
995 EXT4_I(inode), sizeof(struct ext4_inode_info),
996 true);
997 dump_stack();
998 }
fa0d7e3d 999 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 1000}
1001
51cc5068 1002static void init_once(void *foo)
ac27a0ec 1003{
617ba13b 1004 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1005
a35afb83 1006 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1007 init_rwsem(&ei->xattr_sem);
0e855ac8 1008 init_rwsem(&ei->i_data_sem);
ea3d7209 1009 init_rwsem(&ei->i_mmap_sem);
a35afb83 1010 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1011}
1012
e67bc2b3 1013static int __init init_inodecache(void)
ac27a0ec 1014{
617ba13b
MC		 1015 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1016 sizeof(struct ext4_inode_info),
ac27a0ec 1017 0, (SLAB_RECLAIM_ACCOUNT|
5d097056 1018 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
20c2df83 1019 init_once);
617ba13b 1020 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1021 return -ENOMEM;
1022 return 0;
1023}
1024
1025static void destroy_inodecache(void)
1026{
8c0a8537
KS		 1027 /*
1028 * Make sure all delayed rcu free inodes are flushed before we
1029 * destroy cache.
1030 */
1031 rcu_barrier();
617ba13b 1032 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1033}
1034
0930fcc1 1035void ext4_clear_inode(struct inode *inode)
ac27a0ec 1036{
0930fcc1 1037 invalidate_inode_buffers(inode);
dbd5768f 1038 clear_inode(inode);
9f754758 1039 dquot_drop(inode);
c2ea3fde 1040 ext4_discard_preallocations(inode);
51865fda 1041 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1042 if (EXT4_I(inode)->jinode) {
1043 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1044 EXT4_I(inode)->jinode);
1045 jbd2_free_inode(EXT4_I(inode)->jinode);
1046 EXT4_I(inode)->jinode = NULL;
1047 }
b7236e21 1048#ifdef CONFIG_EXT4_FS_ENCRYPTION
a7550b30 1049 fscrypt_put_encryption_info(inode, NULL);
b7236e21 1050#endif
ac27a0ec
DK		 1051}
1052
1b961ac0 1053static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1054 u64 ino, u32 generation)
ac27a0ec 1055{
ac27a0ec 1056 struct inode *inode;
ac27a0ec 1057
617ba13b 1058 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1059 return ERR_PTR(-ESTALE);
617ba13b 1060 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1061 return ERR_PTR(-ESTALE);
1062
1063 /* iget isn't really right if the inode is currently unallocated!!
1064 *
617ba13b 1065 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1066 * deleted, so we should be safe.
1067 *
1068 * Currently we don't know the generation for parent directory, so
1069 * a generation of 0 means "accept any"
1070 */
f4bb2981 1071 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1072 if (IS_ERR(inode))
1073 return ERR_CAST(inode);
1074 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1075 iput(inode);
1076 return ERR_PTR(-ESTALE);
1077 }
1b961ac0
CH		 1078
1079 return inode;
1080}
1081
1082static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1083 int fh_len, int fh_type)
1b961ac0
CH		 1084{
1085 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1086 ext4_nfs_get_inode);
1087}
1088
1089static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1090 int fh_len, int fh_type)
1b961ac0
CH		 1091{
1092 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1093 ext4_nfs_get_inode);
ac27a0ec
DK		 1094}
1095
c39a7f84
TO		 1096/*
1097 * Try to release metadata pages (indirect blocks, directories) which are
1098 * mapped via the block device. Since these pages could have journal heads
1099 * which would prevent try_to_free_buffers() from freeing them, we must use
1100 * jbd2 layer's try_to_free_buffers() function to release them.
1101 */
0b8e58a1
AD		 1102static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1103 gfp_t wait)
c39a7f84
TO		 1104{
1105 journal_t *journal = EXT4_SB(sb)->s_journal;
1106
1107 WARN_ON(PageChecked(page));
1108 if (!page_has_buffers(page))
1109 return 0;
1110 if (journal)
1111 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1112 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1113 return try_to_free_buffers(page);
1114}
1115
a7550b30
JK		 1116#ifdef CONFIG_EXT4_FS_ENCRYPTION
1117static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1118{
1119 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1120 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1121}
1122
a7550b30
JK		 1123static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1124 void *fs_data)
1125{
2f8f5e76
EB		 1126 handle_t *handle = fs_data;
1127 int res, res2, retries = 0;
1128
94840e3c
EB		 1129 res = ext4_convert_inline_data(inode);
1130 if (res)
1131 return res;
1132
2f8f5e76
EB		 1133 /*
1134 * If a journal handle was specified, then the encryption context is
1135 * being set on a new inode via inheritance and is part of a larger
1136 * transaction to create the inode. Otherwise the encryption context is
1137 * being set on an existing inode in its own transaction. Only in the
1138 * latter case should the "retry on ENOSPC" logic be used.
1139 */
a7550b30 1140
2f8f5e76
EB		 1141 if (handle) {
1142 res = ext4_xattr_set_handle(handle, inode,
1143 EXT4_XATTR_INDEX_ENCRYPTION,
1144 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1145 ctx, len, 0);
a7550b30
JK		 1146 if (!res) {
1147 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1148 ext4_clear_inode_state(inode,
1149 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b
JK		 1150 /*
1151 * Update inode->i_flags - e.g. S_DAX may get disabled
1152 */
1153 ext4_set_inode_flags(inode);
a7550b30
JK		 1154 }
1155 return res;
1156 }
1157
2f8f5e76 1158retry:
a7550b30
JK		 1159 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1160 ext4_jbd2_credits_xattr(inode));
1161 if (IS_ERR(handle))
1162 return PTR_ERR(handle);
1163
2f8f5e76
EB		 1164 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1165 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1166 ctx, len, 0);
a7550b30
JK		 1167 if (!res) {
1168 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
a3caa24b
JK		 1169 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1170 ext4_set_inode_flags(inode);
a7550b30
JK		 1171 res = ext4_mark_inode_dirty(handle, inode);
1172 if (res)
1173 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1174 }
1175 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1176
1177 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1178 goto retry;
a7550b30
JK		 1179 if (!res)
1180 res = res2;
1181 return res;
1182}
1183
1184static int ext4_dummy_context(struct inode *inode)
1185{
1186 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1187}
1188
1189static unsigned ext4_max_namelen(struct inode *inode)
1190{
1191 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1192 EXT4_NAME_LEN;
1193}
1194
6f69f0ed 1195static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1196 .key_prefix = "ext4:",
a7550b30 1197 .get_context = ext4_get_context,
a7550b30
JK		 1198 .set_context = ext4_set_context,
1199 .dummy_context = ext4_dummy_context,
1200 .is_encrypted = ext4_encrypted_inode,
1201 .empty_dir = ext4_empty_dir,
1202 .max_namelen = ext4_max_namelen,
1203};
1204#else
6f69f0ed 1205static const struct fscrypt_operations ext4_cryptops = {
a7550b30
JK		 1206 .is_encrypted = ext4_encrypted_inode,
1207};
1208#endif
1209
ac27a0ec 1210#ifdef CONFIG_QUOTA
689c958c
LX		 1211static char *quotatypes[] = INITQFNAMES;
1212#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1213
617ba13b
MC		 1214static int ext4_write_dquot(struct dquot *dquot);
1215static int ext4_acquire_dquot(struct dquot *dquot);
1216static int ext4_release_dquot(struct dquot *dquot);
1217static int ext4_mark_dquot_dirty(struct dquot *dquot);
1218static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1219static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1220 const struct path *path);
ca0e05e4 1221static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1222static int ext4_quota_on_mount(struct super_block *sb, int type);
1223static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1224 size_t len, loff_t off);
617ba13b 1225static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1226 const char *data, size_t len, loff_t off);
7c319d32
AK		 1227static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1228 unsigned int flags);
1229static int ext4_enable_quotas(struct super_block *sb);
8f0e8746 1230static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1231
96c7e0d9
JK		 1232static struct dquot **ext4_get_dquots(struct inode *inode)
1233{
1234 return EXT4_I(inode)->i_dquot;
1235}
1236
61e225dc 1237static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1238 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1239 .write_dquot = ext4_write_dquot,
1240 .acquire_dquot = ext4_acquire_dquot,
1241 .release_dquot = ext4_release_dquot,
1242 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1243 .write_info = ext4_write_info,
1244 .alloc_dquot = dquot_alloc,
1245 .destroy_dquot = dquot_destroy,
040cb378 1246 .get_projid = ext4_get_projid,
8f0e8746 1247 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1248};
1249
0d54b217 1250static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1251 .quota_on = ext4_quota_on,
ca0e05e4 1252 .quota_off = ext4_quota_off,
287a8095 1253 .quota_sync = dquot_quota_sync,
0a240339 1254 .get_state = dquot_get_state,
287a8095
CH		 1255 .set_info = dquot_set_dqinfo,
1256 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1257 .set_dqblk = dquot_set_dqblk,
1258 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1259};
1260#endif
1261
ee9b6d61 1262static const struct super_operations ext4_sops = {
617ba13b
MC		 1263 .alloc_inode = ext4_alloc_inode,
1264 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1265 .write_inode = ext4_write_inode,
1266 .dirty_inode = ext4_dirty_inode,
7ff9c073 1267 .drop_inode = ext4_drop_inode,
0930fcc1 1268 .evict_inode = ext4_evict_inode,
617ba13b 1269 .put_super = ext4_put_super,
617ba13b 1270 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1271 .freeze_fs = ext4_freeze,
1272 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1273 .statfs = ext4_statfs,
1274 .remount_fs = ext4_remount,
617ba13b 1275 .show_options = ext4_show_options,
ac27a0ec 1276#ifdef CONFIG_QUOTA
617ba13b
MC		 1277 .quota_read = ext4_quota_read,
1278 .quota_write = ext4_quota_write,
96c7e0d9 1279 .get_dquots = ext4_get_dquots,
ac27a0ec 1280#endif
c39a7f84 1281 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1282};
1283
39655164 1284static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1285 .fh_to_dentry = ext4_fh_to_dentry,
1286 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1287 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1288};
1289
1290enum {
1291 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1292 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1293 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1294 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1295 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1296 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1297 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1298 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1299 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1300 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1301 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1302 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1303 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1449032b 1304 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
670e9875 1305 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1306 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1307 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1308 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1309 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1310 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1311};
1312
a447c093 1313static const match_table_t tokens = {
ac27a0ec
DK		 1314 {Opt_bsd_df, "bsddf"},
1315 {Opt_minix_df, "minixdf"},
1316 {Opt_grpid, "grpid"},
1317 {Opt_grpid, "bsdgroups"},
1318 {Opt_nogrpid, "nogrpid"},
1319 {Opt_nogrpid, "sysvgroups"},
1320 {Opt_resgid, "resgid=%u"},
1321 {Opt_resuid, "resuid=%u"},
1322 {Opt_sb, "sb=%u"},
1323 {Opt_err_cont, "errors=continue"},
1324 {Opt_err_panic, "errors=panic"},
1325 {Opt_err_ro, "errors=remount-ro"},
1326 {Opt_nouid32, "nouid32"},
ac27a0ec 1327 {Opt_debug, "debug"},
72578c33
TT		 1328 {Opt_removed, "oldalloc"},
1329 {Opt_removed, "orlov"},
ac27a0ec
DK		 1330 {Opt_user_xattr, "user_xattr"},
1331 {Opt_nouser_xattr, "nouser_xattr"},
1332 {Opt_acl, "acl"},
1333 {Opt_noacl, "noacl"},
e3bb52ae 1334 {Opt_noload, "norecovery"},
5a916be1 1335 {Opt_noload, "noload"},
72578c33
TT		 1336 {Opt_removed, "nobh"},
1337 {Opt_removed, "bh"},
ac27a0ec 1338 {Opt_commit, "commit=%u"},
30773840
TT		 1339 {Opt_min_batch_time, "min_batch_time=%u"},
1340 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1341 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1342 {Opt_journal_path, "journal_path=%s"},
818d276c 1343 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1344 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1345 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1346 {Opt_abort, "abort"},
1347 {Opt_data_journal, "data=journal"},
1348 {Opt_data_ordered, "data=ordered"},
1349 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1350 {Opt_data_err_abort, "data_err=abort"},
1351 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1352 {Opt_offusrjquota, "usrjquota="},
1353 {Opt_usrjquota, "usrjquota=%s"},
1354 {Opt_offgrpjquota, "grpjquota="},
1355 {Opt_grpjquota, "grpjquota=%s"},
1356 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1357 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1358 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1359 {Opt_grpquota, "grpquota"},
1360 {Opt_noquota, "noquota"},
1361 {Opt_quota, "quota"},
1362 {Opt_usrquota, "usrquota"},
49da9392 1363 {Opt_prjquota, "prjquota"},
ac27a0ec 1364 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1365 {Opt_barrier, "barrier"},
1366 {Opt_nobarrier, "nobarrier"},
25ec56b5 1367 {Opt_i_version, "i_version"},
923ae0ff 1368 {Opt_dax, "dax"},
c9de560d 1369 {Opt_stripe, "stripe=%u"},
64769240 1370 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1371 {Opt_lazytime, "lazytime"},
1372 {Opt_nolazytime, "nolazytime"},
670e9875 1373 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1374 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1375 {Opt_removed, "mblk_io_submit"},
1376 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1377 {Opt_block_validity, "block_validity"},
1378 {Opt_noblock_validity, "noblock_validity"},
240799cd 1379 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1380 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1381 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1382 {Opt_auto_da_alloc, "auto_da_alloc"},
1383 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1384 {Opt_dioread_nolock, "dioread_nolock"},
1385 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1386 {Opt_discard, "discard"},
1387 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1388 {Opt_init_itable, "init_itable=%u"},
1389 {Opt_init_itable, "init_itable"},
1390 {Opt_noinit_itable, "noinit_itable"},
df981d03 1391 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1392 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1393 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1394 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1395 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1396 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1397 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1398 {Opt_err, NULL},
ac27a0ec
DK		 1399};
1400
617ba13b 1401static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1402{
617ba13b 1403 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1404 char *options = (char *) *data;
1405
1406 if (!options || strncmp(options, "sb=", 3) != 0)
1407 return 1; /* Default location */
0b8e58a1 1408
ac27a0ec 1409 options += 3;
0b8e58a1 1410 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1411 sb_block = simple_strtoul(options, &options, 0);
1412 if (*options && *options != ',') {
4776004f 1413 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1414 (char *) *data);
1415 return 1;
1416 }
1417 if (*options == ',')
1418 options++;
1419 *data = (void *) options;
0b8e58a1 1420
ac27a0ec
DK		 1421 return sb_block;
1422}
1423
b3881f74 1424#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1425static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1426 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1427
56c50f11
DM		 1428#ifdef CONFIG_QUOTA
1429static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1430{
1431 struct ext4_sb_info *sbi = EXT4_SB(sb);
1432 char *qname;
03dafb5f 1433 int ret = -1;
56c50f11
DM		 1434
1435 if (sb_any_quota_loaded(sb) &&
1436 !sbi->s_qf_names[qtype]) {
1437 ext4_msg(sb, KERN_ERR,
1438 "Cannot change journaled "
1439 "quota options when quota turned on");
57f73c2c 1440 return -1;
56c50f11 1441 }
e2b911c5 1442 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1443 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1444 "ignored when QUOTA feature is enabled");
1445 return 1;
262b4662 1446 }
56c50f11
DM		 1447 qname = match_strdup(args);
1448 if (!qname) {
1449 ext4_msg(sb, KERN_ERR,
1450 "Not enough memory for storing quotafile name");
57f73c2c 1451 return -1;
56c50f11 1452 }
03dafb5f
CG		 1453 if (sbi->s_qf_names[qtype]) {
1454 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1455 ret = 1;
1456 else
1457 ext4_msg(sb, KERN_ERR,
1458 "%s quota file already specified",
1459 QTYPE2NAME(qtype));
1460 goto errout;
56c50f11 1461 }
03dafb5f 1462 if (strchr(qname, '/')) {
56c50f11
DM		 1463 ext4_msg(sb, KERN_ERR,
1464 "quotafile must be on filesystem root");
03dafb5f 1465 goto errout;
56c50f11 1466 }
03dafb5f 1467 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1468 set_opt(sb, QUOTA);
56c50f11 1469 return 1;
03dafb5f
CG		 1470errout:
1471 kfree(qname);
1472 return ret;
56c50f11
DM		 1473}
1474
1475static int clear_qf_name(struct super_block *sb, int qtype)
1476{
1477
1478 struct ext4_sb_info *sbi = EXT4_SB(sb);
1479
1480 if (sb_any_quota_loaded(sb) &&
1481 sbi->s_qf_names[qtype]) {
1482 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1483 " when quota turned on");
57f73c2c 1484 return -1;
56c50f11 1485 }
03dafb5f 1486 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1487 sbi->s_qf_names[qtype] = NULL;
1488 return 1;
1489}
1490#endif
1491
26092bf5
TT		 1492#define MOPT_SET 0x0001
1493#define MOPT_CLEAR 0x0002
1494#define MOPT_NOSUPPORT 0x0004
1495#define MOPT_EXPLICIT 0x0008
1496#define MOPT_CLEAR_ERR 0x0010
1497#define MOPT_GTE0 0x0020
ac27a0ec 1498#ifdef CONFIG_QUOTA
26092bf5
TT		 1499#define MOPT_Q 0
1500#define MOPT_QFMT 0x0040
1501#else
1502#define MOPT_Q MOPT_NOSUPPORT
1503#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1504#endif
26092bf5 1505#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1506#define MOPT_NO_EXT2 0x0100
1507#define MOPT_NO_EXT3 0x0200
1508#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1509#define MOPT_STRING 0x0400
26092bf5
TT		 1510
1511static const struct mount_opts {
1512 int token;
1513 int mount_opt;
1514 int flags;
1515} ext4_mount_opts[] = {
1516 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1517 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1518 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1519 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1520 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1521 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1522 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1523 MOPT_EXT4_ONLY | MOPT_SET},
1524 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1525 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1526 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1527 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1528 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1529 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1530 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1531 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1532 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1533 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1534 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1535 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1536 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1537 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1538 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1539 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1540 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1541 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1542 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1543 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1544 MOPT_NO_EXT2},
8dc0aa8c 1545 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1546 MOPT_NO_EXT2},
26092bf5
TT		 1547 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1548 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1549 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1550 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1551 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1552 {Opt_commit, 0, MOPT_GTE0},
1553 {Opt_max_batch_time, 0, MOPT_GTE0},
1554 {Opt_min_batch_time, 0, MOPT_GTE0},
1555 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1556 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1557 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1558 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1559 {Opt_resuid, 0, MOPT_GTE0},
1560 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1561 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1562 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1563 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1564 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1565 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1566 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1567 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1568 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1569 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1570#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1571 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1572 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1573#else
26092bf5
TT		 1574 {Opt_acl, 0, MOPT_NOSUPPORT},
1575 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1576#endif
26092bf5
TT		 1577 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1578 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1579 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1580 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1581 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1582 MOPT_SET | MOPT_Q},
1583 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1584 MOPT_SET | MOPT_Q},
49da9392
JK		 1585 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1586 MOPT_SET | MOPT_Q},
26092bf5 1587 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1588 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1589 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1590 {Opt_usrjquota, 0, MOPT_Q},
1591 {Opt_grpjquota, 0, MOPT_Q},
1592 {Opt_offusrjquota, 0, MOPT_Q},
1593 {Opt_offgrpjquota, 0, MOPT_Q},
1594 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1595 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1596 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1597 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1598 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1599 {Opt_err, 0, 0}
1600};
1601
1602static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1603 substring_t *args, unsigned long *journal_devnum,
1604 unsigned int *journal_ioprio, int is_remount)
1605{
1606 struct ext4_sb_info *sbi = EXT4_SB(sb);
1607 const struct mount_opts *m;
08cefc7a
EB		 1608 kuid_t uid;
1609 kgid_t gid;
26092bf5
TT		 1610 int arg = 0;
1611
57f73c2c
TT		 1612#ifdef CONFIG_QUOTA
1613 if (token == Opt_usrjquota)
1614 return set_qf_name(sb, USRQUOTA, &args[0]);
1615 else if (token == Opt_grpjquota)
1616 return set_qf_name(sb, GRPQUOTA, &args[0]);
1617 else if (token == Opt_offusrjquota)
1618 return clear_qf_name(sb, USRQUOTA);
1619 else if (token == Opt_offgrpjquota)
1620 return clear_qf_name(sb, GRPQUOTA);
1621#endif
26092bf5 1622 switch (token) {
f7048605
TT		 1623 case Opt_noacl:
1624 case Opt_nouser_xattr:
1625 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1626 break;
26092bf5
TT		 1627 case Opt_sb:
1628 return 1; /* handled by get_sb_block() */
1629 case Opt_removed:
5f3633e3 1630 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1631 return 1;
26092bf5
TT		 1632 case Opt_abort:
1633 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1634 return 1;
1635 case Opt_i_version:
1636 sb->s_flags |= MS_I_VERSION;
1637 return 1;
a26f4992
TT		 1638 case Opt_lazytime:
1639 sb->s_flags |= MS_LAZYTIME;
1640 return 1;
1641 case Opt_nolazytime:
1642 sb->s_flags &= ~MS_LAZYTIME;
1643 return 1;
26092bf5
TT		 1644 }
1645
5f3633e3
JK		 1646 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1647 if (token == m->token)
1648 break;
1649
1650 if (m->token == Opt_err) {
1651 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1652 "or missing value", opt);
1653 return -1;
1654 }
1655
8dc0aa8c
TT		 1656 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1657 ext4_msg(sb, KERN_ERR,
1658 "Mount option \"%s\" incompatible with ext2", opt);
1659 return -1;
1660 }
1661 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1662 ext4_msg(sb, KERN_ERR,
1663 "Mount option \"%s\" incompatible with ext3", opt);
1664 return -1;
1665 }
1666
ad4eec61 1667 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1668 return -1;
1669 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1670 return -1;
c93cf2d7
DM		 1671 if (m->flags & MOPT_EXPLICIT) {
1672 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1673 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1674 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1675 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1676 } else
1677 return -1;
1678 }
5f3633e3
JK		 1679 if (m->flags & MOPT_CLEAR_ERR)
1680 clear_opt(sb, ERRORS_MASK);
1681 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1682 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1683 "options when quota turned on");
1684 return -1;
1685 }
1686
1687 if (m->flags & MOPT_NOSUPPORT) {
1688 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1689 } else if (token == Opt_commit) {
1690 if (arg == 0)
1691 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1692 sbi->s_commit_interval = HZ * arg;
670e9875
TT		 1693 } else if (token == Opt_debug_want_extra_isize) {
1694 sbi->s_want_extra_isize = arg;
5f3633e3 1695 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1696 sbi->s_max_batch_time = arg;
1697 } else if (token == Opt_min_batch_time) {
1698 sbi->s_min_batch_time = arg;
1699 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1700 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1701 ext4_msg(sb, KERN_ERR,
1702 "EXT4-fs: inode_readahead_blks must be "
1703 "0 or a power of 2 smaller than 2^31");
26092bf5 1704 return -1;
5f3633e3
JK		 1705 }
1706 sbi->s_inode_readahead_blks = arg;
1707 } else if (token == Opt_init_itable) {
1708 set_opt(sb, INIT_INODE_TABLE);
1709 if (!args->from)
1710 arg = EXT4_DEF_LI_WAIT_MULT;
1711 sbi->s_li_wait_mult = arg;
1712 } else if (token == Opt_max_dir_size_kb) {
1713 sbi->s_max_dir_size_kb = arg;
1714 } else if (token == Opt_stripe) {
1715 sbi->s_stripe = arg;
1716 } else if (token == Opt_resuid) {
1717 uid = make_kuid(current_user_ns(), arg);
1718 if (!uid_valid(uid)) {
1719 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1720 return -1;
1721 }
5f3633e3
JK		 1722 sbi->s_resuid = uid;
1723 } else if (token == Opt_resgid) {
1724 gid = make_kgid(current_user_ns(), arg);
1725 if (!gid_valid(gid)) {
1726 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1727 return -1;
1728 }
1729 sbi->s_resgid = gid;
1730 } else if (token == Opt_journal_dev) {
1731 if (is_remount) {
1732 ext4_msg(sb, KERN_ERR,
1733 "Cannot specify journal on remount");
1734 return -1;
1735 }
1736 *journal_devnum = arg;
ad4eec61
ES		 1737 } else if (token == Opt_journal_path) {
1738 char *journal_path;
1739 struct inode *journal_inode;
1740 struct path path;
1741 int error;
1742
1743 if (is_remount) {
1744 ext4_msg(sb, KERN_ERR,
1745 "Cannot specify journal on remount");
1746 return -1;
1747 }
1748 journal_path = match_strdup(&args[0]);
1749 if (!journal_path) {
1750 ext4_msg(sb, KERN_ERR, "error: could not dup "
1751 "journal device string");
1752 return -1;
1753 }
1754
1755 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1756 if (error) {
1757 ext4_msg(sb, KERN_ERR, "error: could not find "
1758 "journal device path: error %d", error);
1759 kfree(journal_path);
1760 return -1;
1761 }
1762
2b0143b5 1763 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1764 if (!S_ISBLK(journal_inode->i_mode)) {
1765 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1766 "is not a block device", journal_path);
1767 path_put(&path);
1768 kfree(journal_path);
1769 return -1;
1770 }
1771
1772 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1773 path_put(&path);
1774 kfree(journal_path);
5f3633e3
JK		 1775 } else if (token == Opt_journal_ioprio) {
1776 if (arg > 7) {
1777 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1778 " (must be 0-7)");
1779 return -1;
1780 }
1781 *journal_ioprio =
1782 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1783 } else if (token == Opt_test_dummy_encryption) {
1784#ifdef CONFIG_EXT4_FS_ENCRYPTION
1785 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1786 ext4_msg(sb, KERN_WARNING,
1787 "Test dummy encryption mode enabled");
1788#else
1789 ext4_msg(sb, KERN_WARNING,
1790 "Test dummy encryption mount option ignored");
1791#endif
5f3633e3
JK		 1792 } else if (m->flags & MOPT_DATAJ) {
1793 if (is_remount) {
1794 if (!sbi->s_journal)
1795 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1796 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1797 ext4_msg(sb, KERN_ERR,
26092bf5 1798 "Cannot change data mode on remount");
26092bf5 1799 return -1;
ac27a0ec 1800 }
26092bf5 1801 } else {
5f3633e3
JK		 1802 clear_opt(sb, DATA_FLAGS);
1803 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1804 }
5f3633e3
JK		 1805#ifdef CONFIG_QUOTA
1806 } else if (m->flags & MOPT_QFMT) {
1807 if (sb_any_quota_loaded(sb) &&
1808 sbi->s_jquota_fmt != m->mount_opt) {
1809 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1810 "quota options when quota turned on");
1811 return -1;
1812 }
e2b911c5 1813 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1814 ext4_msg(sb, KERN_INFO,
1815 "Quota format mount options ignored "
262b4662 1816 "when QUOTA feature is enabled");
c325a67c 1817 return 1;
262b4662 1818 }
5f3633e3 1819 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1820#endif
923ae0ff 1821 } else if (token == Opt_dax) {
ef83b6e8
DW		 1822#ifdef CONFIG_FS_DAX
1823 ext4_msg(sb, KERN_WARNING,
1824 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1825 sbi->s_mount_opt |= m->mount_opt;
1826#else
923ae0ff
RZ		 1827 ext4_msg(sb, KERN_INFO, "dax option not supported");
1828 return -1;
5f3633e3 1829#endif
7915a861
AN		 1830 } else if (token == Opt_data_err_abort) {
1831 sbi->s_mount_opt |= m->mount_opt;
1832 } else if (token == Opt_data_err_ignore) {
1833 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1834 } else {
1835 if (!args->from)
1836 arg = 1;
1837 if (m->flags & MOPT_CLEAR)
1838 arg = !arg;
1839 else if (unlikely(!(m->flags & MOPT_SET))) {
1840 ext4_msg(sb, KERN_WARNING,
1841 "buggy handling of option %s", opt);
1842 WARN_ON(1);
1843 return -1;
1844 }
1845 if (arg != 0)
1846 sbi->s_mount_opt |= m->mount_opt;
1847 else
1848 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1849 }
5f3633e3 1850 return 1;
26092bf5
TT		 1851}
1852
1853static int parse_options(char *options, struct super_block *sb,
1854 unsigned long *journal_devnum,
1855 unsigned int *journal_ioprio,
1856 int is_remount)
1857{
1858 struct ext4_sb_info *sbi = EXT4_SB(sb);
1859 char *p;
1860 substring_t args[MAX_OPT_ARGS];
1861 int token;
1862
1863 if (!options)
1864 return 1;
1865
1866 while ((p = strsep(&options, ",")) != NULL) {
1867 if (!*p)
1868 continue;
1869 /*
1870 * Initialize args struct so we know whether arg was
1871 * found; some options take optional arguments.
1872 */
caecd0af 1873 args[0].to = args[0].from = NULL;
26092bf5
TT		 1874 token = match_token(p, tokens, args);
1875 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1876 journal_ioprio, is_remount) < 0)
1877 return 0;
ac27a0ec
DK		 1878 }
1879#ifdef CONFIG_QUOTA
49da9392
JK		 1880 /*
1881 * We do the test below only for project quotas. 'usrquota' and
1882 * 'grpquota' mount options are allowed even without quota feature
1883 * to support legacy quotas in quota files.
1884 */
1885 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1886 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1887 "Cannot enable project quota enforcement.");
1888 return 0;
1889 }
1890 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1891 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1892 clear_opt(sb, USRQUOTA);
ac27a0ec 1893
482a7425 1894 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1895 clear_opt(sb, GRPQUOTA);
ac27a0ec 1896
56c50f11 1897 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1898 ext4_msg(sb, KERN_ERR, "old and new quota "
1899 "format mixing");
ac27a0ec
DK		 1900 return 0;
1901 }
1902
1903 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1904 ext4_msg(sb, KERN_ERR, "journaled quota format "
1905 "not specified");
ac27a0ec
DK		 1906 return 0;
1907 }
ac27a0ec
DK		 1908 }
1909#endif
261cb20c
JK		 1910 if (test_opt(sb, DIOREAD_NOLOCK)) {
1911 int blocksize =
1912 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1913
09cbfeaf 1914 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 1915 ext4_msg(sb, KERN_ERR, "can't mount with "
1916 "dioread_nolock if block size != PAGE_SIZE");
1917 return 0;
1918 }
1919 }
ac27a0ec
DK		 1920 return 1;
1921}
1922
2adf6da8
TT		 1923static inline void ext4_show_quota_options(struct seq_file *seq,
1924 struct super_block *sb)
1925{
1926#if defined(CONFIG_QUOTA)
1927 struct ext4_sb_info *sbi = EXT4_SB(sb);
1928
1929 if (sbi->s_jquota_fmt) {
1930 char *fmtname = "";
1931
1932 switch (sbi->s_jquota_fmt) {
1933 case QFMT_VFS_OLD:
1934 fmtname = "vfsold";
1935 break;
1936 case QFMT_VFS_V0:
1937 fmtname = "vfsv0";
1938 break;
1939 case QFMT_VFS_V1:
1940 fmtname = "vfsv1";
1941 break;
1942 }
1943 seq_printf(seq, ",jqfmt=%s", fmtname);
1944 }
1945
1946 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1947 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1948
1949 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1950 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1951#endif
1952}
1953
5a916be1
TT		 1954static const char *token2str(int token)
1955{
50df9fd5 1956 const struct match_token *t;
5a916be1
TT		 1957
1958 for (t = tokens; t->token != Opt_err; t++)
1959 if (t->token == token && !strchr(t->pattern, '='))
1960 break;
1961 return t->pattern;
1962}
1963
2adf6da8
TT		 1964/*
1965 * Show an option if
1966 * - it's set to a non-default value OR
1967 * - if the per-sb default is different from the global default
1968 */
66acdcf4
TT		 1969static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1970 int nodefs)
2adf6da8 1971{
2adf6da8
TT		 1972 struct ext4_sb_info *sbi = EXT4_SB(sb);
1973 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1974 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1975 const struct mount_opts *m;
66acdcf4 1976 char sep = nodefs ? '\n' : ',';
2adf6da8 1977
66acdcf4
TT		 1978#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1979#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1980
1981 if (sbi->s_sb_block != 1)
5a916be1
TT		 1982 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1983
1984 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1985 int want_set = m->flags & MOPT_SET;
1986 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1987 (m->flags & MOPT_CLEAR_ERR))
1988 continue;
66acdcf4 1989 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1990 continue; /* skip if same as the default */
1991 if ((want_set &&
1992 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1993 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1994 continue; /* select Opt_noFoo vs Opt_Foo */
1995 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1996 }
5a916be1 1997
08cefc7a 1998 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1999 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2000 SEQ_OPTS_PRINT("resuid=%u",
2001 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2002 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2003 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2004 SEQ_OPTS_PRINT("resgid=%u",
2005 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2006 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2007 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2008 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2009 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2010 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2011 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2012 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2013 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2014 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2015 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2016 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2017 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2018 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 2019 if (sb->s_flags & MS_I_VERSION)
5a916be1 2020 SEQ_OPTS_PUTS("i_version");
66acdcf4 2021 if (nodefs || sbi->s_stripe)
5a916be1 2022 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 2023 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2024 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2025 SEQ_OPTS_PUTS("data=journal");
2026 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2027 SEQ_OPTS_PUTS("data=ordered");
2028 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2029 SEQ_OPTS_PUTS("data=writeback");
2030 }
66acdcf4
TT		 2031 if (nodefs ||
2032 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2033 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2034 sbi->s_inode_readahead_blks);
2adf6da8 2035
66acdcf4
TT		 2036 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2037 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2038 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2039 if (nodefs || sbi->s_max_dir_size_kb)
2040 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2041 if (test_opt(sb, DATA_ERR_ABORT))
2042 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 2043
2044 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2045 return 0;
2046}
2047
66acdcf4
TT		 2048static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2049{
2050 return _ext4_show_options(seq, root->d_sb, 0);
2051}
2052
ebd173be 2053int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2054{
2055 struct super_block *sb = seq->private;
2056 int rc;
2057
2058 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2059 rc = _ext4_show_options(seq, sb, 1);
2060 seq_puts(seq, "\n");
2061 return rc;
2062}
2063
617ba13b 2064static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2065 int read_only)
2066{
617ba13b 2067 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2068 int res = 0;
2069
617ba13b 2070 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2071 ext4_msg(sb, KERN_ERR, "revision level too high, "
2072 "forcing read-only mode");
ac27a0ec
DK		 2073 res = MS_RDONLY;
2074 }
2075 if (read_only)
281b5995 2076 goto done;
617ba13b 2077 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2078 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2079 "running e2fsck is recommended");
c8b459f4 2080 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2081 ext4_msg(sb, KERN_WARNING,
2082 "warning: mounting fs with errors, "
2083 "running e2fsck is recommended");
ed3ce80a 2084 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2085 le16_to_cpu(es->s_mnt_count) >=
2086 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2087 ext4_msg(sb, KERN_WARNING,
2088 "warning: maximal mount count reached, "
2089 "running e2fsck is recommended");
ac27a0ec
DK		 2090 else if (le32_to_cpu(es->s_checkinterval) &&
2091 (le32_to_cpu(es->s_lastcheck) +
2092 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 2093 ext4_msg(sb, KERN_WARNING,
2094 "warning: checktime reached, "
2095 "running e2fsck is recommended");
0b8e58a1 2096 if (!sbi->s_journal)
0390131b 2097 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2098 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2099 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2100 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 2101 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 2102 ext4_update_dynamic_rev(sb);
0390131b 2103 if (sbi->s_journal)
e2b911c5 2104 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2105
e2d67052 2106 ext4_commit_super(sb, 1);
281b5995 2107done:
ac27a0ec 2108 if (test_opt(sb, DEBUG))
a9df9a49 2109 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2110 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2111 sb->s_blocksize,
2112 sbi->s_groups_count,
617ba13b
MC		 2113 EXT4_BLOCKS_PER_GROUP(sb),
2114 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2115 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2116
7abc52c2 2117 cleancache_init_fs(sb);
ac27a0ec
DK		 2118 return res;
2119}
2120
117fff10
TT		 2121int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2122{
2123 struct ext4_sb_info *sbi = EXT4_SB(sb);
2124 struct flex_groups *new_groups;
2125 int size;
2126
2127 if (!sbi->s_log_groups_per_flex)
2128 return 0;
2129
2130 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2131 if (size <= sbi->s_flex_groups_allocated)
2132 return 0;
2133
2134 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2135 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2136 if (!new_groups) {
2137 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2138 size / (int) sizeof(struct flex_groups));
2139 return -ENOMEM;
2140 }
2141
2142 if (sbi->s_flex_groups) {
2143 memcpy(new_groups, sbi->s_flex_groups,
2144 (sbi->s_flex_groups_allocated *
2145 sizeof(struct flex_groups)));
b93b41d4 2146 kvfree(sbi->s_flex_groups);
117fff10
TT		 2147 }
2148 sbi->s_flex_groups = new_groups;
2149 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2150 return 0;
2151}
2152
772cb7c8
JS		 2153static int ext4_fill_flex_info(struct super_block *sb)
2154{
2155 struct ext4_sb_info *sbi = EXT4_SB(sb);
2156 struct ext4_group_desc *gdp = NULL;
772cb7c8 2157 ext4_group_t flex_group;
117fff10 2158 int i, err;
772cb7c8 2159
503358ae 2160 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2161 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2162 sbi->s_log_groups_per_flex = 0;
2163 return 1;
2164 }
2165
117fff10
TT		 2166 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2167 if (err)
9933fc0a 2168 goto failed;
772cb7c8 2169
772cb7c8 2170 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2171 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2172
2173 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2174 atomic_add(ext4_free_inodes_count(sb, gdp),
2175 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2176 atomic64_add(ext4_free_group_clusters(sb, gdp),
2177 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2178 atomic_add(ext4_used_dirs_count(sb, gdp),
2179 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2180 }
2181
2182 return 1;
2183failed:
2184 return 0;
2185}
2186
e2b911c5 2187static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2188 struct ext4_group_desc *gdp)
717d50e4 2189{
b47820ed 2190 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2191 __u16 crc = 0;
feb0ab32 2192 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2193 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2194
9aa5d32b 2195 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2196 /* Use new metadata_csum algorithm */
feb0ab32 2197 __u32 csum32;
b47820ed 2198 __u16 dummy_csum = 0;
feb0ab32 2199
feb0ab32
DW		 2200 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2201 sizeof(le_group));
b47820ed
DJ		 2202 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2203 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2204 sizeof(dummy_csum));
2205 offset += sizeof(dummy_csum);
2206 if (offset < sbi->s_desc_size)
2207 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2208 sbi->s_desc_size - offset);
feb0ab32
DW		 2209
2210 crc = csum32 & 0xFFFF;
2211 goto out;
717d50e4
AD		 2212 }
2213
feb0ab32 2214 /* old crc16 code */
e2b911c5 2215 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2216 return 0;
2217
feb0ab32
DW		 2218 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2219 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2220 crc = crc16(crc, (__u8 *)gdp, offset);
2221 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2222 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2223 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2224 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2225 crc = crc16(crc, (__u8 *)gdp + offset,
2226 le16_to_cpu(sbi->s_es->s_desc_size) -
2227 offset);
2228
2229out:
717d50e4
AD		 2230 return cpu_to_le16(crc);
2231}
2232
feb0ab32 2233int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2234 struct ext4_group_desc *gdp)
2235{
feb0ab32 2236 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2237 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2238 return 0;
2239
2240 return 1;
2241}
2242
feb0ab32
DW		 2243void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2244 struct ext4_group_desc *gdp)
2245{
2246 if (!ext4_has_group_desc_csum(sb))
2247 return;
e2b911c5 2248 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2249}
2250
ac27a0ec 2251/* Called at mount-time, super-block is locked */
bfff6873 2252static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2253 ext4_fsblk_t sb_block,
bfff6873 2254 ext4_group_t *first_not_zeroed)
ac27a0ec 2255{
617ba13b
MC		 2256 struct ext4_sb_info *sbi = EXT4_SB(sb);
2257 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2258 ext4_fsblk_t last_block;
bd81d8ee
LV		 2259 ext4_fsblk_t block_bitmap;
2260 ext4_fsblk_t inode_bitmap;
2261 ext4_fsblk_t inode_table;
ce421581 2262 int flexbg_flag = 0;
bfff6873 2263 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2264
e2b911c5 2265 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2266 flexbg_flag = 1;
2267
af5bc92d 2268 ext4_debug("Checking group descriptors");
ac27a0ec 2269
197cd65a
AM		 2270 for (i = 0; i < sbi->s_groups_count; i++) {
2271 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2272
ce421581 2273 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2274 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2275 else
2276 last_block = first_block +
617ba13b 2277 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2278
bfff6873
LC		 2279 if ((grp == sbi->s_groups_count) &&
2280 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2281 grp = i;
2282
8fadc143 2283 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2284 if (block_bitmap == sb_block) {
2285 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2286 "Block bitmap for group %u overlaps "
2287 "superblock", i);
2288 }
2b2d6d01 2289 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2290 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2291 "Block bitmap for group %u not in group "
b31e1552 2292 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2293 return 0;
2294 }
8fadc143 2295 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2296 if (inode_bitmap == sb_block) {
2297 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2298 "Inode bitmap for group %u overlaps "
2299 "superblock", i);
2300 }
2b2d6d01 2301 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2302 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2303 "Inode bitmap for group %u not in group "
b31e1552 2304 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2305 return 0;
2306 }
8fadc143 2307 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2308 if (inode_table == sb_block) {
2309 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2310 "Inode table for group %u overlaps "
2311 "superblock", i);
2312 }
bd81d8ee 2313 if (inode_table < first_block ||
2b2d6d01 2314 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2315 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2316 "Inode table for group %u not in group "
b31e1552 2317 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2318 return 0;
2319 }
955ce5f5 2320 ext4_lock_group(sb, i);
feb0ab32 2321 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2322 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2323 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2324 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2325 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2326 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2327 ext4_unlock_group(sb, i);
8a266467 2328 return 0;
7ee1ec4c 2329 }
717d50e4 2330 }
955ce5f5 2331 ext4_unlock_group(sb, i);
ce421581
JS		 2332 if (!flexbg_flag)
2333 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2334 }
bfff6873
LC		 2335 if (NULL != first_not_zeroed)
2336 *first_not_zeroed = grp;
ac27a0ec
DK		 2337 return 1;
2338}
2339
617ba13b 2340/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2341 * the superblock) which were deleted from all directories, but held open by
2342 * a process at the time of a crash. We walk the list and try to delete these
2343 * inodes at recovery time (only with a read-write filesystem).
2344 *
2345 * In order to keep the orphan inode chain consistent during traversal (in
2346 * case of crash during recovery), we link each inode into the superblock
2347 * orphan list_head and handle it the same way as an inode deletion during
2348 * normal operation (which journals the operations for us).
2349 *
2350 * We only do an iget() and an iput() on each inode, which is very safe if we
2351 * accidentally point at an in-use or already deleted inode. The worst that
2352 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2353 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2354 * e2fsck was run on this filesystem, and it must have already done the orphan
2355 * inode cleanup for us, so we can safely abort without any further action.
2356 */
2b2d6d01
TT		 2357static void ext4_orphan_cleanup(struct super_block *sb,
2358 struct ext4_super_block *es)
ac27a0ec
DK		 2359{
2360 unsigned int s_flags = sb->s_flags;
2c98eb5e 2361 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec
DK		 2362#ifdef CONFIG_QUOTA
2363 int i;
2364#endif
2365 if (!es->s_last_orphan) {
2366 jbd_debug(4, "no orphan inodes to clean up\n");
2367 return;
2368 }
2369
a8f48a95 2370 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2371 ext4_msg(sb, KERN_ERR, "write access "
2372 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2373 return;
2374 }
2375
d39195c3
AG		 2376 /* Check if feature set would not allow a r/w mount */
2377 if (!ext4_feature_set_ok(sb, 0)) {
2378 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2379 "unknown ROCOMPAT features");
2380 return;
2381 }
2382
617ba13b 2383 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2384 /* don't clear list on RO mount w/ errors */
2385 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2386 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2387 "clearing orphan list.\n");
c25f9bc6
ES		 2388 es->s_last_orphan = 0;
2389 }
ac27a0ec
DK		 2390 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2391 return;
2392 }
2393
2394 if (s_flags & MS_RDONLY) {
b31e1552 2395 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2396 sb->s_flags &= ~MS_RDONLY;
2397 }
2398#ifdef CONFIG_QUOTA
2399 /* Needed for iput() to work correctly and not trash data */
2400 sb->s_flags |= MS_ACTIVE;
2401 /* Turn on quotas so that they are updated correctly */
a2d4a646 2402 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2403 if (EXT4_SB(sb)->s_qf_names[i]) {
2404 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2405 if (ret < 0)
b31e1552
ES		 2406 ext4_msg(sb, KERN_ERR,
2407 "Cannot turn on journaled "
2408 "quota: error %d", ret);
ac27a0ec
DK		 2409 }
2410 }
2411#endif
2412
2413 while (es->s_last_orphan) {
2414 struct inode *inode;
2415
c65d5c6c
VN		 2416 /*
2417 * We may have encountered an error during cleanup; if
2418 * so, skip the rest.
2419 */
2420 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2421 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2422 es->s_last_orphan = 0;
2423 break;
2424 }
2425
97bd42b9
JB		 2426 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2427 if (IS_ERR(inode)) {
ac27a0ec
DK		 2428 es->s_last_orphan = 0;
2429 break;
2430 }
2431
617ba13b 2432 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2433 dquot_initialize(inode);
ac27a0ec 2434 if (inode->i_nlink) {
566370a2
PT		 2435 if (test_opt(sb, DEBUG))
2436 ext4_msg(sb, KERN_DEBUG,
2437 "%s: truncating inode %lu to %lld bytes",
2438 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2439 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2440 inode->i_ino, inode->i_size);
5955102c 2441 inode_lock(inode);
55f252c9 2442 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2443 ret = ext4_truncate(inode);
2444 if (ret)
2445 ext4_std_error(inode->i_sb, ret);
5955102c 2446 inode_unlock(inode);
ac27a0ec
DK		 2447 nr_truncates++;
2448 } else {
566370a2
PT		 2449 if (test_opt(sb, DEBUG))
2450 ext4_msg(sb, KERN_DEBUG,
2451 "%s: deleting unreferenced inode %lu",
2452 __func__, inode->i_ino);
ac27a0ec
DK		 2453 jbd_debug(2, "deleting unreferenced inode %lu\n",
2454 inode->i_ino);
2455 nr_orphans++;
2456 }
2457 iput(inode); /* The delete magic happens here! */
2458 }
2459
2b2d6d01 2460#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2461
2462 if (nr_orphans)
b31e1552
ES		 2463 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2464 PLURAL(nr_orphans));
ac27a0ec 2465 if (nr_truncates)
b31e1552
ES		 2466 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2467 PLURAL(nr_truncates));
ac27a0ec
DK		 2468#ifdef CONFIG_QUOTA
2469 /* Turn quotas off */
a2d4a646 2470 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2471 if (sb_dqopt(sb)->files[i])
287a8095 2472 dquot_quota_off(sb, i);
ac27a0ec
DK		 2473 }
2474#endif
2475 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2476}
0b8e58a1 2477
cd2291a4
ES		 2478/*
2479 * Maximal extent format file size.
2480 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2481 * extent format containers, within a sector_t, and within i_blocks
2482 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2483 * so that won't be a limiting factor.
2484 *
f17722f9
LC		 2485 * However there is other limiting factor. We do store extents in the form
2486 * of starting block and length, hence the resulting length of the extent
2487 * covering maximum file size must fit into on-disk format containers as
2488 * well. Given that length is always by 1 unit bigger than max unit (because
2489 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2490 *
cd2291a4
ES		 2491 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2492 */
f287a1a5 2493static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2494{
2495 loff_t res;
2496 loff_t upper_limit = MAX_LFS_FILESIZE;
2497
2498 /* small i_blocks in vfs inode? */
f287a1a5 2499 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2500 /*
90c699a9 2501 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2502 * i_block represent total blocks in 512 bytes
2503 * 32 == size of vfs inode i_blocks * 8
2504 */
2505 upper_limit = (1LL << 32) - 1;
2506
2507 /* total blocks in file system block size */
2508 upper_limit >>= (blkbits - 9);
2509 upper_limit <<= blkbits;
2510 }
2511
f17722f9
LC		 2512 /*
2513 * 32-bit extent-start container, ee_block. We lower the maxbytes
2514 * by one fs block, so ee_len can cover the extent of maximum file
2515 * size
2516 */
2517 res = (1LL << 32) - 1;
cd2291a4 2518 res <<= blkbits;
cd2291a4
ES		 2519
2520 /* Sanity check against vm- & vfs- imposed limits */
2521 if (res > upper_limit)
2522 res = upper_limit;
2523
2524 return res;
2525}
ac27a0ec 2526
ac27a0ec 2527/*
cd2291a4 2528 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2529 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2530 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2531 */
f287a1a5 2532static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2533{
617ba13b 2534 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2535 int meta_blocks;
2536 loff_t upper_limit;
0b8e58a1
AD		 2537 /* This is calculated to be the largest file size for a dense, block
2538 * mapped file such that the file's total number of 512-byte sectors,
2539 * including data and all indirect blocks, does not exceed (2^48 - 1).
2540 *
2541 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2542 * number of 512-byte sectors of the file.
0fc1b451
AK		 2543 */
2544
f287a1a5 2545 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2546 /*
90c699a9 2547 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2548 * the inode i_block field represents total file blocks in
2549 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2550 */
2551 upper_limit = (1LL << 32) - 1;
2552
2553 /* total blocks in file system block size */
2554 upper_limit >>= (bits - 9);
2555
2556 } else {
8180a562
AK		 2557 /*
2558 * We use 48 bit ext4_inode i_blocks
2559 * With EXT4_HUGE_FILE_FL set the i_blocks
2560 * represent total number of blocks in
2561 * file system block size
2562 */
0fc1b451
AK		 2563 upper_limit = (1LL << 48) - 1;
2564
0fc1b451
AK		 2565 }
2566
2567 /* indirect blocks */
2568 meta_blocks = 1;
2569 /* double indirect blocks */
2570 meta_blocks += 1 + (1LL << (bits-2));
2571 /* tripple indirect blocks */
2572 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2573
2574 upper_limit -= meta_blocks;
2575 upper_limit <<= bits;
ac27a0ec
DK		 2576
2577 res += 1LL << (bits-2);
2578 res += 1LL << (2*(bits-2));
2579 res += 1LL << (3*(bits-2));
2580 res <<= bits;
2581 if (res > upper_limit)
2582 res = upper_limit;
0fc1b451
AK		 2583
2584 if (res > MAX_LFS_FILESIZE)
2585 res = MAX_LFS_FILESIZE;
2586
ac27a0ec
DK		 2587 return res;
2588}
2589
617ba13b 2590static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2591 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2592{
617ba13b 2593 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2594 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2595 int has_super = 0;
2596
2597 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2598
e2b911c5 2599 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2600 return logical_sb_block + nr + 1;
ac27a0ec 2601 bg = sbi->s_desc_per_block * nr;
617ba13b 2602 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2603 has_super = 1;
0b8e58a1 2604
bd63f6b0
DW		 2605 /*
2606 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2607 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2608 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2609 * compensate.
2610 */
2611 if (sb->s_blocksize == 1024 && nr == 0 &&
2612 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2613 has_super++;
2614
617ba13b 2615 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2616}
2617
c9de560d
AT		 2618/**
2619 * ext4_get_stripe_size: Get the stripe size.
2620 * @sbi: In memory super block info
2621 *
2622 * If we have specified it via mount option, then
2623 * use the mount option value. If the value specified at mount time is
2624 * greater than the blocks per group use the super block value.
2625 * If the super block value is greater than blocks per group return 0.
2626 * Allocator needs it be less than blocks per group.
2627 *
2628 */
2629static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2630{
2631 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2632 unsigned long stripe_width =
2633 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2634 int ret;
c9de560d
AT		 2635
2636 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 2637 ret = sbi->s_stripe;
5469d7c3 2638 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 2639 ret = stripe_width;
5469d7c3 2640 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 2641 ret = stride;
2642 else
2643 ret = 0;
c9de560d 2644
3eb08658
DE		 2645 /*
2646 * If the stripe width is 1, this makes no sense and
2647 * we set it to 0 to turn off stripe handling code.
2648 */
2649 if (ret <= 1)
2650 ret = 0;
c9de560d 2651
3eb08658 2652 return ret;
c9de560d 2653}
ac27a0ec 2654
a13fb1a4
ES		 2655/*
2656 * Check whether this filesystem can be mounted based on
2657 * the features present and the RDONLY/RDWR mount requested.
2658 * Returns 1 if this filesystem can be mounted as requested,
2659 * 0 if it cannot be.
2660 */
2661static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2662{
e2b911c5 2663 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2664 ext4_msg(sb, KERN_ERR,
2665 "Couldn't mount because of "
2666 "unsupported optional features (%x)",
2667 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2668 ~EXT4_FEATURE_INCOMPAT_SUPP));
2669 return 0;
2670 }
2671
2672 if (readonly)
2673 return 1;
2674
e2b911c5 2675 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b
DW		 2676 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2677 sb->s_flags |= MS_RDONLY;
2678 return 1;
2679 }
2680
a13fb1a4 2681 /* Check that feature set is OK for a read-write mount */
e2b911c5 2682 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2683 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2684 "unsupported optional features (%x)",
2685 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2686 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2687 return 0;
2688 }
2689 /*
2690 * Large file size enabled file system can only be mounted
2691 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2692 */
e2b911c5 2693 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2694 if (sizeof(blkcnt_t) < sizeof(u64)) {
2695 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2696 "cannot be mounted RDWR without "
2697 "CONFIG_LBDAF");
2698 return 0;
2699 }
2700 }
e2b911c5 2701 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2702 ext4_msg(sb, KERN_ERR,
2703 "Can't support bigalloc feature without "
2704 "extents feature\n");
2705 return 0;
2706 }
7c319d32
AK		 2707
2708#ifndef CONFIG_QUOTA
e2b911c5 2709 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2710 ext4_msg(sb, KERN_ERR,
2711 "Filesystem with quota feature cannot be mounted RDWR "
2712 "without CONFIG_QUOTA");
2713 return 0;
2714 }
689c958c
LX		 2715 if (ext4_has_feature_project(sb) && !readonly) {
2716 ext4_msg(sb, KERN_ERR,
2717 "Filesystem with project quota feature cannot be mounted RDWR "
2718 "without CONFIG_QUOTA");
2719 return 0;
2720 }
7c319d32 2721#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2722 return 1;
2723}
2724
66e61a9e
TT		 2725/*
2726 * This function is called once a day if we have errors logged
2727 * on the file system
2728 */
2729static void print_daily_error_info(unsigned long arg)
2730{
2731 struct super_block *sb = (struct super_block *) arg;
2732 struct ext4_sb_info *sbi;
2733 struct ext4_super_block *es;
2734
2735 sbi = EXT4_SB(sb);
2736 es = sbi->s_es;
2737
2738 if (es->s_error_count)
ae0f78de
TT		 2739 /* fsck newer than v1.41.13 is needed to clean this condition. */
2740 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2741 le32_to_cpu(es->s_error_count));
2742 if (es->s_first_error_time) {
ae0f78de 2743 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2744 sb->s_id, le32_to_cpu(es->s_first_error_time),
2745 (int) sizeof(es->s_first_error_func),
2746 es->s_first_error_func,
2747 le32_to_cpu(es->s_first_error_line));
2748 if (es->s_first_error_ino)
651e1c3b 2749 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2750 le32_to_cpu(es->s_first_error_ino));
2751 if (es->s_first_error_block)
651e1c3b 2752 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2753 le64_to_cpu(es->s_first_error_block));
651e1c3b 2754 printk(KERN_CONT "\n");
66e61a9e
TT		 2755 }
2756 if (es->s_last_error_time) {
ae0f78de 2757 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2758 sb->s_id, le32_to_cpu(es->s_last_error_time),
2759 (int) sizeof(es->s_last_error_func),
2760 es->s_last_error_func,
2761 le32_to_cpu(es->s_last_error_line));
2762 if (es->s_last_error_ino)
651e1c3b 2763 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2764 le32_to_cpu(es->s_last_error_ino));
2765 if (es->s_last_error_block)
651e1c3b 2766 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2767 le64_to_cpu(es->s_last_error_block));
651e1c3b 2768 printk(KERN_CONT "\n");
66e61a9e
TT		 2769 }
2770 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2771}
2772
bfff6873
LC		 2773/* Find next suitable group and run ext4_init_inode_table */
2774static int ext4_run_li_request(struct ext4_li_request *elr)
2775{
2776 struct ext4_group_desc *gdp = NULL;
2777 ext4_group_t group, ngroups;
2778 struct super_block *sb;
2779 unsigned long timeout = 0;
2780 int ret = 0;
2781
2782 sb = elr->lr_super;
2783 ngroups = EXT4_SB(sb)->s_groups_count;
2784
2785 for (group = elr->lr_next_group; group < ngroups; group++) {
2786 gdp = ext4_get_group_desc(sb, group, NULL);
2787 if (!gdp) {
2788 ret = 1;
2789 break;
2790 }
2791
2792 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2793 break;
2794 }
2795
7f511862 2796 if (group >= ngroups)
bfff6873
LC		 2797 ret = 1;
2798
2799 if (!ret) {
2800 timeout = jiffies;
2801 ret = ext4_init_inode_table(sb, group,
2802 elr->lr_timeout ? 0 : 1);
2803 if (elr->lr_timeout == 0) {
51ce6511
LC		 2804 timeout = (jiffies - timeout) *
2805 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2806 elr->lr_timeout = timeout;
2807 }
2808 elr->lr_next_sched = jiffies + elr->lr_timeout;
2809 elr->lr_next_group = group + 1;
2810 }
bfff6873
LC		 2811 return ret;
2812}
2813
2814/*
2815 * Remove lr_request from the list_request and free the
4ed5c033 2816 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2817 */
2818static void ext4_remove_li_request(struct ext4_li_request *elr)
2819{
2820 struct ext4_sb_info *sbi;
2821
2822 if (!elr)
2823 return;
2824
2825 sbi = elr->lr_sbi;
2826
2827 list_del(&elr->lr_request);
2828 sbi->s_li_request = NULL;
2829 kfree(elr);
2830}
2831
2832static void ext4_unregister_li_request(struct super_block *sb)
2833{
1bb933fb
LC		 2834 mutex_lock(&ext4_li_mtx);
2835 if (!ext4_li_info) {
2836 mutex_unlock(&ext4_li_mtx);
bfff6873 2837 return;
1bb933fb 2838 }
bfff6873
LC		 2839
2840 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2841 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2842 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2843 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2844}
2845
8f1f7453
ES		 2846static struct task_struct *ext4_lazyinit_task;
2847
bfff6873
LC		 2848/*
2849 * This is the function where ext4lazyinit thread lives. It walks
2850 * through the request list searching for next scheduled filesystem.
2851 * When such a fs is found, run the lazy initialization request
2852 * (ext4_rn_li_request) and keep track of the time spend in this
2853 * function. Based on that time we compute next schedule time of
2854 * the request. When walking through the list is complete, compute
2855 * next waking time and put itself into sleep.
2856 */
2857static int ext4_lazyinit_thread(void *arg)
2858{
2859 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2860 struct list_head *pos, *n;
2861 struct ext4_li_request *elr;
4ed5c033 2862 unsigned long next_wakeup, cur;
bfff6873
LC		 2863
2864 BUG_ON(NULL == eli);
2865
bfff6873
LC		 2866cont_thread:
2867 while (true) {
2868 next_wakeup = MAX_JIFFY_OFFSET;
2869
2870 mutex_lock(&eli->li_list_mtx);
2871 if (list_empty(&eli->li_request_list)) {
2872 mutex_unlock(&eli->li_list_mtx);
2873 goto exit_thread;
2874 }
bfff6873 2875 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 2876 int err = 0;
2877 int progress = 0;
bfff6873
LC		 2878 elr = list_entry(pos, struct ext4_li_request,
2879 lr_request);
2880
e22834f0
DM		 2881 if (time_before(jiffies, elr->lr_next_sched)) {
2882 if (time_before(elr->lr_next_sched, next_wakeup))
2883 next_wakeup = elr->lr_next_sched;
2884 continue;
2885 }
2886 if (down_read_trylock(&elr->lr_super->s_umount)) {
2887 if (sb_start_write_trylock(elr->lr_super)) {
2888 progress = 1;
2889 /*
2890 * We hold sb->s_umount, sb can not
2891 * be removed from the list, it is
2892 * now safe to drop li_list_mtx
2893 */
2894 mutex_unlock(&eli->li_list_mtx);
2895 err = ext4_run_li_request(elr);
2896 sb_end_write(elr->lr_super);
2897 mutex_lock(&eli->li_list_mtx);
2898 n = pos->next;
b2c78cd0 2899 }
e22834f0
DM		 2900 up_read((&elr->lr_super->s_umount));
2901 }
2902 /* error, remove the lazy_init job */
2903 if (err) {
2904 ext4_remove_li_request(elr);
2905 continue;
2906 }
2907 if (!progress) {
2908 elr->lr_next_sched = jiffies +
2909 (prandom_u32()
2910 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 2911 }
bfff6873
LC		 2912 if (time_before(elr->lr_next_sched, next_wakeup))
2913 next_wakeup = elr->lr_next_sched;
2914 }
2915 mutex_unlock(&eli->li_list_mtx);
2916
a0acae0e 2917 try_to_freeze();
bfff6873 2918
4ed5c033
LC		 2919 cur = jiffies;
2920 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2921 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2922 cond_resched();
2923 continue;
2924 }
2925
4ed5c033
LC		 2926 schedule_timeout_interruptible(next_wakeup - cur);
2927
8f1f7453
ES		 2928 if (kthread_should_stop()) {
2929 ext4_clear_request_list();
2930 goto exit_thread;
2931 }
bfff6873
LC		 2932 }
2933
2934exit_thread:
2935 /*
2936 * It looks like the request list is empty, but we need
2937 * to check it under the li_list_mtx lock, to prevent any
2938 * additions into it, and of course we should lock ext4_li_mtx
2939 * to atomically free the list and ext4_li_info, because at
2940 * this point another ext4 filesystem could be registering
2941 * new one.
2942 */
2943 mutex_lock(&ext4_li_mtx);
2944 mutex_lock(&eli->li_list_mtx);
2945 if (!list_empty(&eli->li_request_list)) {
2946 mutex_unlock(&eli->li_list_mtx);
2947 mutex_unlock(&ext4_li_mtx);
2948 goto cont_thread;
2949 }
2950 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2951 kfree(ext4_li_info);
2952 ext4_li_info = NULL;
2953 mutex_unlock(&ext4_li_mtx);
2954
2955 return 0;
2956}
2957
2958static void ext4_clear_request_list(void)
2959{
2960 struct list_head *pos, *n;
2961 struct ext4_li_request *elr;
2962
2963 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2964 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2965 elr = list_entry(pos, struct ext4_li_request,
2966 lr_request);
2967 ext4_remove_li_request(elr);
2968 }
2969 mutex_unlock(&ext4_li_info->li_list_mtx);
2970}
2971
2972static int ext4_run_lazyinit_thread(void)
2973{
8f1f7453
ES		 2974 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2975 ext4_li_info, "ext4lazyinit");
2976 if (IS_ERR(ext4_lazyinit_task)) {
2977 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2978 ext4_clear_request_list();
bfff6873
LC		 2979 kfree(ext4_li_info);
2980 ext4_li_info = NULL;
92b97816 2981 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2982 "initialization thread\n",
2983 err);
2984 return err;
2985 }
2986 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2987 return 0;
2988}
2989
2990/*
2991 * Check whether it make sense to run itable init. thread or not.
2992 * If there is at least one uninitialized inode table, return
2993 * corresponding group number, else the loop goes through all
2994 * groups and return total number of groups.
2995 */
2996static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2997{
2998 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2999 struct ext4_group_desc *gdp = NULL;
3000
3001 for (group = 0; group < ngroups; group++) {
3002 gdp = ext4_get_group_desc(sb, group, NULL);
3003 if (!gdp)
3004 continue;
3005
3006 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3007 break;
3008 }
3009
3010 return group;
3011}
3012
3013static int ext4_li_info_new(void)
3014{
3015 struct ext4_lazy_init *eli = NULL;
3016
3017 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3018 if (!eli)
3019 return -ENOMEM;
3020
bfff6873
LC		 3021 INIT_LIST_HEAD(&eli->li_request_list);
3022 mutex_init(&eli->li_list_mtx);
3023
bfff6873
LC		 3024 eli->li_state |= EXT4_LAZYINIT_QUIT;
3025
3026 ext4_li_info = eli;
3027
3028 return 0;
3029}
3030
3031static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3032 ext4_group_t start)
3033{
3034 struct ext4_sb_info *sbi = EXT4_SB(sb);
3035 struct ext4_li_request *elr;
bfff6873
LC		 3036
3037 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3038 if (!elr)
3039 return NULL;
3040
3041 elr->lr_super = sb;
3042 elr->lr_sbi = sbi;
3043 elr->lr_next_group = start;
3044
3045 /*
3046 * Randomize first schedule time of the request to
3047 * spread the inode table initialization requests
3048 * better.
3049 */
dd1f723b
TT		 3050 elr->lr_next_sched = jiffies + (prandom_u32() %
3051 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3052 return elr;
3053}
3054
7f511862
TT		 3055int ext4_register_li_request(struct super_block *sb,
3056 ext4_group_t first_not_zeroed)
bfff6873
LC		 3057{
3058 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3059 struct ext4_li_request *elr = NULL;
bfff6873 3060 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3061 int ret = 0;
bfff6873 3062
7f511862 3063 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3064 if (sbi->s_li_request != NULL) {
3065 /*
3066 * Reset timeout so it can be computed again, because
3067 * s_li_wait_mult might have changed.
3068 */
3069 sbi->s_li_request->lr_timeout = 0;
7f511862 3070 goto out;
51ce6511 3071 }
bfff6873
LC		 3072
3073 if (first_not_zeroed == ngroups ||
3074 (sb->s_flags & MS_RDONLY) ||
55ff3840 3075 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3076 goto out;
bfff6873
LC		 3077
3078 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3079 if (!elr) {
3080 ret = -ENOMEM;
3081 goto out;
3082 }
bfff6873
LC		 3083
3084 if (NULL == ext4_li_info) {
3085 ret = ext4_li_info_new();
3086 if (ret)
3087 goto out;
3088 }
3089
3090 mutex_lock(&ext4_li_info->li_list_mtx);
3091 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3092 mutex_unlock(&ext4_li_info->li_list_mtx);
3093
3094 sbi->s_li_request = elr;
46e4690b
TM		 3095 /*
3096 * set elr to NULL here since it has been inserted to
3097 * the request_list and the removal and free of it is
3098 * handled by ext4_clear_request_list from now on.
3099 */
3100 elr = NULL;
bfff6873
LC		 3101
3102 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3103 ret = ext4_run_lazyinit_thread();
3104 if (ret)
3105 goto out;
3106 }
bfff6873 3107out:
beed5ecb
NK		 3108 mutex_unlock(&ext4_li_mtx);
3109 if (ret)
bfff6873 3110 kfree(elr);
bfff6873
LC		 3111 return ret;
3112}
3113
3114/*
3115 * We do not need to lock anything since this is called on
3116 * module unload.
3117 */
3118static void ext4_destroy_lazyinit_thread(void)
3119{
3120 /*
3121 * If thread exited earlier
3122 * there's nothing to be done.
3123 */
8f1f7453 3124 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3125 return;
3126
8f1f7453 3127 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3128}
3129
25ed6e8a
DW		 3130static int set_journal_csum_feature_set(struct super_block *sb)
3131{
3132 int ret = 1;
3133 int compat, incompat;
3134 struct ext4_sb_info *sbi = EXT4_SB(sb);
3135
9aa5d32b 3136 if (ext4_has_metadata_csum(sb)) {
db9ee220 3137 /* journal checksum v3 */
25ed6e8a 3138 compat = 0;
db9ee220 3139 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3140 } else {
3141 /* journal checksum v1 */
3142 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3143 incompat = 0;
3144 }
3145
feb8c6d3
DW		 3146 jbd2_journal_clear_features(sbi->s_journal,
3147 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3148 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3149 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3150 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3151 ret = jbd2_journal_set_features(sbi->s_journal,
3152 compat, 0,
3153 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3154 incompat);
3155 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3156 ret = jbd2_journal_set_features(sbi->s_journal,
3157 compat, 0,
3158 incompat);
3159 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3160 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3161 } else {
feb8c6d3
DW		 3162 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3163 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3164 }
3165
3166 return ret;
3167}
3168
952fc18e
TT		 3169/*
3170 * Note: calculating the overhead so we can be compatible with
3171 * historical BSD practice is quite difficult in the face of
3172 * clusters/bigalloc. This is because multiple metadata blocks from
3173 * different block group can end up in the same allocation cluster.
3174 * Calculating the exact overhead in the face of clustered allocation
3175 * requires either O(all block bitmaps) in memory or O(number of block
3176 * groups**2) in time. We will still calculate the superblock for
3177 * older file systems --- and if we come across with a bigalloc file
3178 * system with zero in s_overhead_clusters the estimate will be close to
3179 * correct especially for very large cluster sizes --- but for newer
3180 * file systems, it's better to calculate this figure once at mkfs
3181 * time, and store it in the superblock. If the superblock value is
3182 * present (even for non-bigalloc file systems), we will use it.
3183 */
3184static int count_overhead(struct super_block *sb, ext4_group_t grp,
3185 char *buf)
3186{
3187 struct ext4_sb_info *sbi = EXT4_SB(sb);
3188 struct ext4_group_desc *gdp;
3189 ext4_fsblk_t first_block, last_block, b;
3190 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3191 int s, j, count = 0;
3192
e2b911c5 3193 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3194 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3195 sbi->s_itb_per_group + 2);
3196
952fc18e
TT		 3197 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3198 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3199 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3200 for (i = 0; i < ngroups; i++) {
3201 gdp = ext4_get_group_desc(sb, i, NULL);
3202 b = ext4_block_bitmap(sb, gdp);
3203 if (b >= first_block && b <= last_block) {
3204 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3205 count++;
3206 }
3207 b = ext4_inode_bitmap(sb, gdp);
3208 if (b >= first_block && b <= last_block) {
3209 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3210 count++;
3211 }
3212 b = ext4_inode_table(sb, gdp);
3213 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3214 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3215 int c = EXT4_B2C(sbi, b - first_block);
3216 ext4_set_bit(c, buf);
3217 count++;
3218 }
3219 if (i != grp)
3220 continue;
3221 s = 0;
3222 if (ext4_bg_has_super(sb, grp)) {
3223 ext4_set_bit(s++, buf);
3224 count++;
3225 }
c48ae41b
TT		 3226 j = ext4_bg_num_gdb(sb, grp);
3227 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3228 ext4_error(sb, "Invalid number of block group "
3229 "descriptor blocks: %d", j);
3230 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3231 }
c48ae41b
TT		 3232 count += j;
3233 for (; j > 0; j--)
3234 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3235 }
3236 if (!count)
3237 return 0;
3238 return EXT4_CLUSTERS_PER_GROUP(sb) -
3239 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3240}
3241
3242/*
3243 * Compute the overhead and stash it in sbi->s_overhead
3244 */
3245int ext4_calculate_overhead(struct super_block *sb)
3246{
3247 struct ext4_sb_info *sbi = EXT4_SB(sb);
3248 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3249 struct inode *j_inode;
3250 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3251 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3252 ext4_fsblk_t overhead = 0;
4fdb5543 3253 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3254
952fc18e
TT		 3255 if (!buf)
3256 return -ENOMEM;
3257
3258 /*
3259 * Compute the overhead (FS structures). This is constant
3260 * for a given filesystem unless the number of block groups
3261 * changes so we cache the previous value until it does.
3262 */
3263
3264 /*
3265 * All of the blocks before first_data_block are overhead
3266 */
3267 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3268
3269 /*
3270 * Add the overhead found in each block group
3271 */
3272 for (i = 0; i < ngroups; i++) {
3273 int blks;
3274
3275 blks = count_overhead(sb, i, buf);
3276 overhead += blks;
3277 if (blks)
3278 memset(buf, 0, PAGE_SIZE);
3279 cond_resched();
3280 }
3c816ded
EW		 3281
3282 /*
3283 * Add the internal journal blocks whether the journal has been
3284 * loaded or not
3285 */
b003b524 3286 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3287 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3288 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3289 j_inode = ext4_get_journal_inode(sb, j_inum);
3290 if (j_inode) {
3291 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3292 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3293 iput(j_inode);
3294 } else {
3295 ext4_msg(sb, KERN_ERR, "can't get journal size");
3296 }
3297 }
952fc18e
TT		 3298 sbi->s_overhead = overhead;
3299 smp_wmb();
3300 free_page((unsigned long) buf);
3301 return 0;
3302}
3303
b5799018 3304static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3305{
3306 ext4_fsblk_t resv_clusters;
b5799018 3307 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3308
30fac0f7
JK		 3309 /*
3310 * There's no need to reserve anything when we aren't using extents.
3311 * The space estimates are exact, there are no unwritten extents,
3312 * hole punching doesn't need new metadata... This is needed especially
3313 * to keep ext2/3 backward compatibility.
3314 */
e2b911c5 3315 if (!ext4_has_feature_extents(sb))
b5799018 3316 return;
27dd4385
LC		 3317 /*
3318 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3319 * This should cover the situations where we can not afford to run
3320 * out of space like for example punch hole, or converting
556615dc 3321 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3322 * allocation would require 1, or 2 blocks, higher numbers are
3323 * very rare.
3324 */
b5799018
TT		 3325 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3326 sbi->s_cluster_bits);
27dd4385
LC		 3327
3328 do_div(resv_clusters, 50);
3329 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3330
b5799018 3331 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3332}
3333
2b2d6d01 3334static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3335{
d4c402d9 3336 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3337 struct buffer_head *bh;
617ba13b 3338 struct ext4_super_block *es = NULL;
5aee0f8a 3339 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3340 ext4_fsblk_t block;
3341 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3342 ext4_fsblk_t logical_sb_block;
ac27a0ec 3343 unsigned long offset = 0;
ac27a0ec
DK		 3344 unsigned long journal_devnum = 0;
3345 unsigned long def_mount_opts;
3346 struct inode *root;
0390131b 3347 const char *descr;
dcc7dae3 3348 int ret = -ENOMEM;
281b5995 3349 int blocksize, clustersize;
4ec11028
TT		 3350 unsigned int db_count;
3351 unsigned int i;
281b5995 3352 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3353 __u64 blocks_count;
07aa2ea1 3354 int err = 0;
b3881f74 3355 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3356 ext4_group_t first_not_zeroed;
ac27a0ec 3357
5aee0f8a
TT		 3358 if ((data && !orig_data) || !sbi)
3359 goto out_free_base;
705895b6
PE		 3360
3361 sbi->s_blockgroup_lock =
3362 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3363 if (!sbi->s_blockgroup_lock)
3364 goto out_free_base;
3365
ac27a0ec 3366 sb->s_fs_info = sbi;
2c0544b2 3367 sbi->s_sb = sb;
240799cd 3368 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3369 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3370 if (sb->s_bdev->bd_part)
3371 sbi->s_sectors_written_start =
3372 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3373
9f6200bb 3374 /* Cleanup superblock name */
ec3904dc 3375 strreplace(sb->s_id, '/', '!');
9f6200bb 3376
07aa2ea1 3377 /* -EINVAL is default */
dcc7dae3 3378 ret = -EINVAL;
617ba13b 3379 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3380 if (!blocksize) {
b31e1552 3381 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3382 goto out_fail;
3383 }
3384
3385 /*
617ba13b 3386 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3387 * block sizes. We need to calculate the offset from buffer start.
3388 */
617ba13b 3389 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3390 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3391 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3392 } else {
70bbb3e0 3393 logical_sb_block = sb_block;
ac27a0ec
DK		 3394 }
3395
a8ac900b 3396 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3397 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3398 goto out_fail;
3399 }
3400 /*
3401 * Note: s_es must be initialized as soon as possible because
617ba13b 3402 * some ext4 macro-instructions depend on its value
ac27a0ec 3403 */
2716b802 3404 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3405 sbi->s_es = es;
3406 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3407 if (sb->s_magic != EXT4_SUPER_MAGIC)
3408 goto cantfind_ext4;
afc32f7e 3409 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3410
feb0ab32 3411 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3412 if (ext4_has_feature_metadata_csum(sb) &&
3413 ext4_has_feature_gdt_csum(sb))
363307e6 3414 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3415 "redundant flags; please run fsck.");
3416
d25425f8
DW		 3417 /* Check for a known checksum algorithm */
3418 if (!ext4_verify_csum_type(sb, es)) {
3419 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3420 "unknown checksum algorithm.");
3421 silent = 1;
3422 goto cantfind_ext4;
3423 }
3424
0441984a 3425 /* Load the checksum driver */
e2b911c5 3426 if (ext4_has_feature_metadata_csum(sb)) {
0441984a
DW		 3427 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3428 if (IS_ERR(sbi->s_chksum_driver)) {
3429 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3430 ret = PTR_ERR(sbi->s_chksum_driver);
3431 sbi->s_chksum_driver = NULL;
3432 goto failed_mount;
3433 }
3434 }
3435
a9c47317
DW		 3436 /* Check superblock checksum */
3437 if (!ext4_superblock_csum_verify(sb, es)) {
3438 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3439 "invalid superblock checksum. Run e2fsck?");
3440 silent = 1;
6a797d27 3441 ret = -EFSBADCRC;
a9c47317
DW		 3442 goto cantfind_ext4;
3443 }
3444
3445 /* Precompute checksum seed for all metadata */
e2b911c5 3446 if (ext4_has_feature_csum_seed(sb))
8c81bd8f
DW		 3447 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3448 else if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3449 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3450 sizeof(es->s_uuid));
3451
ac27a0ec
DK		 3452 /* Set defaults before we parse the mount options */
3453 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3454 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3455 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3456 set_opt(sb, DEBUG);
87f26807 3457 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3458 set_opt(sb, GRPID);
617ba13b 3459 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3460 set_opt(sb, NO_UID32);
ea663336 3461 /* xattr user namespace & acls are now defaulted on */
ea663336 3462 set_opt(sb, XATTR_USER);
03010a33 3463#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3464 set_opt(sb, POSIX_ACL);
2e7842b8 3465#endif
98c1a759
DW		 3466 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3467 if (ext4_has_metadata_csum(sb))
3468 set_opt(sb, JOURNAL_CHECKSUM);
3469
617ba13b 3470 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3471 set_opt(sb, JOURNAL_DATA);
617ba13b 3472 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3473 set_opt(sb, ORDERED_DATA);
617ba13b 3474 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3475 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3476
3477 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3478 set_opt(sb, ERRORS_PANIC);
bb4f397a 3479 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3480 set_opt(sb, ERRORS_CONT);
bb4f397a 3481 else
fd8c37ec 3482 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3483 /* block_validity enabled by default; disable with noblock_validity */
3484 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3485 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3486 set_opt(sb, DISCARD);
ac27a0ec 3487
08cefc7a
EB		 3488 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3489 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3490 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3491 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3492 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3493
8b67f04a 3494 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3495 set_opt(sb, BARRIER);
ac27a0ec 3496
dd919b98
AK		 3497 /*
3498 * enable delayed allocation by default
3499 * Use -o nodelalloc to turn it off
3500 */
bc0b75f7 3501 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3502 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3503 set_opt(sb, DELALLOC);
dd919b98 3504
51ce6511
LC		 3505 /*
3506 * set default s_li_wait_mult for lazyinit, for the case there is
3507 * no mount option specified.
3508 */
3509 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3510
5aee0f8a
TT		 3511 if (sbi->s_es->s_mount_opts[0]) {
3512 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3513 sizeof(sbi->s_es->s_mount_opts),
3514 GFP_KERNEL);
3515 if (!s_mount_opts)
3516 goto failed_mount;
3517 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3518 &journal_ioprio, 0)) {
3519 ext4_msg(sb, KERN_WARNING,
3520 "failed to parse options in superblock: %s",
3521 s_mount_opts);
3522 }
3523 kfree(s_mount_opts);
8b67f04a 3524 }
5a916be1 3525 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3526 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3527 &journal_ioprio, 0))
ac27a0ec
DK		 3528 goto failed_mount;
3529
56889787
TT		 3530 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3531 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3532 "with data=journal disables delayed "
3533 "allocation and O_DIRECT support!\n");
3534 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3535 ext4_msg(sb, KERN_ERR, "can't mount with "
3536 "both data=journal and delalloc");
3537 goto failed_mount;
3538 }
3539 if (test_opt(sb, DIOREAD_NOLOCK)) {
3540 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3541 "both data=journal and dioread_nolock");
56889787
TT		 3542 goto failed_mount;
3543 }
923ae0ff
RZ		 3544 if (test_opt(sb, DAX)) {
3545 ext4_msg(sb, KERN_ERR, "can't mount with "
3546 "both data=journal and dax");
3547 goto failed_mount;
3548 }
73b92a2a
SK		 3549 if (ext4_has_feature_encrypt(sb)) {
3550 ext4_msg(sb, KERN_WARNING,
3551 "encrypted files will use data=ordered "
3552 "instead of data journaling mode");
3553 }
56889787
TT		 3554 if (test_opt(sb, DELALLOC))
3555 clear_opt(sb, DELALLOC);
001e4a87
TH		 3556 } else {
3557 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3558 }
3559
ac27a0ec 3560 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3561 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3562
617ba13b 3563 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3564 (ext4_has_compat_features(sb) ||
3565 ext4_has_ro_compat_features(sb) ||
3566 ext4_has_incompat_features(sb)))
b31e1552
ES		 3567 ext4_msg(sb, KERN_WARNING,
3568 "feature flags set on rev 0 fs, "
3569 "running e2fsck is recommended");
469108ff 3570
ed3654eb
TT		 3571 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3572 set_opt2(sb, HURD_COMPAT);
e2b911c5 3573 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3574 ext4_msg(sb, KERN_ERR,
3575 "The Hurd can't support 64-bit file systems");
3576 goto failed_mount;
3577 }
3578 }
3579
2035e776
TT		 3580 if (IS_EXT2_SB(sb)) {
3581 if (ext2_feature_set_ok(sb))
3582 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3583 "using the ext4 subsystem");
3584 else {
3585 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3586 "to feature incompatibilities");
3587 goto failed_mount;
3588 }
3589 }
3590
3591 if (IS_EXT3_SB(sb)) {
3592 if (ext3_feature_set_ok(sb))
3593 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3594 "using the ext4 subsystem");
3595 else {
3596 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3597 "to feature incompatibilities");
3598 goto failed_mount;
3599 }
3600 }
3601
ac27a0ec
DK		 3602 /*
3603 * Check feature flags regardless of the revision level, since we
3604 * previously didn't change the revision level when setting the flags,
3605 * so there is a chance incompat flags are set on a rev 0 filesystem.
3606 */
a13fb1a4 3607 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3608 goto failed_mount;
a13fb1a4 3609
261cb20c 3610 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3611 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3612 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3613 ext4_msg(sb, KERN_ERR,
8cdf3372
TT		 3614 "Unsupported filesystem blocksize %d (%d log_block_size)",
3615 blocksize, le32_to_cpu(es->s_log_block_size));
3616 goto failed_mount;
3617 }
3618 if (le32_to_cpu(es->s_log_block_size) >
3619 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3620 ext4_msg(sb, KERN_ERR,
3621 "Invalid log block size: %u",
3622 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3623 goto failed_mount;
3624 }
3625
5b9554dc
TT		 3626 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3627 ext4_msg(sb, KERN_ERR,
3628 "Number of reserved GDT blocks insanely large: %d",
3629 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3630 goto failed_mount;
3631 }
3632
923ae0ff 3633 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
87eefeb4
TK		 3634 err = bdev_dax_supported(sb, blocksize);
3635 if (err)
923ae0ff 3636 goto failed_mount;
923ae0ff
RZ		 3637 }
3638
e2b911c5 3639 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3640 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3641 es->s_encryption_level);
3642 goto failed_mount;
3643 }
3644
ac27a0ec 3645 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3646 /* Validate the filesystem blocksize */
3647 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3648 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3649 blocksize);
ac27a0ec
DK		 3650 goto failed_mount;
3651 }
3652
2b2d6d01 3653 brelse(bh);
70bbb3e0
AM		 3654 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3655 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3656 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3657 if (!bh) {
b31e1552
ES		 3658 ext4_msg(sb, KERN_ERR,
3659 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3660 goto failed_mount;
3661 }
2716b802 3662 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3663 sbi->s_es = es;
617ba13b 3664 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3665 ext4_msg(sb, KERN_ERR,
3666 "Magic mismatch, very weird!");
ac27a0ec
DK		 3667 goto failed_mount;
3668 }
3669 }
3670
e2b911c5 3671 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3672 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3673 has_huge_files);
3674 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3675
617ba13b
MC		 3676 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3677 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3678 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3679 } else {
3680 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3681 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3682 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3683 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3684 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3685 ext4_msg(sb, KERN_ERR,
3686 "unsupported inode size: %d",
2b2d6d01 3687 sbi->s_inode_size);
ac27a0ec
DK		 3688 goto failed_mount;
3689 }
ef7f3835
KS		 3690 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3691 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3692 }
0b8e58a1 3693
0d1ee42f 3694 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3695 if (ext4_has_feature_64bit(sb)) {
8fadc143 3696 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3697 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3698 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3699 ext4_msg(sb, KERN_ERR,
3700 "unsupported descriptor size %lu",
0d1ee42f
AR		 3701 sbi->s_desc_size);
3702 goto failed_mount;
3703 }
3704 } else
3705 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3706
ac27a0ec 3707 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3708 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 3709
617ba13b 3710 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3711 if (sbi->s_inodes_per_block == 0)
617ba13b 3712 goto cantfind_ext4;
cd6bb35b
TT		 3713 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3714 sbi->s_inodes_per_group > blocksize * 8) {
3715 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3716 sbi->s_blocks_per_group);
3717 goto failed_mount;
3718 }
ac27a0ec
DK		 3719 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3720 sbi->s_inodes_per_block;
0d1ee42f 3721 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3722 sbi->s_sbh = bh;
3723 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3724 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3725 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3726
2b2d6d01 3727 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3728 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3729 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3730 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3731 i = le32_to_cpu(es->s_flags);
3732 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3733 sbi->s_hash_unsigned = 3;
3734 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3735#ifdef __CHAR_UNSIGNED__
23301410
TT		 3736 if (!(sb->s_flags & MS_RDONLY))
3737 es->s_flags |=
3738 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3739 sbi->s_hash_unsigned = 3;
f99b2589 3740#else
23301410
TT		 3741 if (!(sb->s_flags & MS_RDONLY))
3742 es->s_flags |=
3743 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3744#endif
23301410 3745 }
f99b2589 3746 }
ac27a0ec 3747
281b5995
TT		 3748 /* Handle clustersize */
3749 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3750 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3751 if (has_bigalloc) {
3752 if (clustersize < blocksize) {
3753 ext4_msg(sb, KERN_ERR,
3754 "cluster size (%d) smaller than "
3755 "block size (%d)", clustersize, blocksize);
3756 goto failed_mount;
3757 }
8cdf3372
TT		 3758 if (le32_to_cpu(es->s_log_cluster_size) >
3759 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3760 ext4_msg(sb, KERN_ERR,
3761 "Invalid log cluster size: %u",
3762 le32_to_cpu(es->s_log_cluster_size));
3763 goto failed_mount;
3764 }
281b5995
TT		 3765 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3766 le32_to_cpu(es->s_log_block_size);
3767 sbi->s_clusters_per_group =
3768 le32_to_cpu(es->s_clusters_per_group);
3769 if (sbi->s_clusters_per_group > blocksize * 8) {
3770 ext4_msg(sb, KERN_ERR,
3771 "#clusters per group too big: %lu",
3772 sbi->s_clusters_per_group);
3773 goto failed_mount;
3774 }
3775 if (sbi->s_blocks_per_group !=
3776 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3777 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3778 "clusters per group (%lu) inconsistent",
3779 sbi->s_blocks_per_group,
3780 sbi->s_clusters_per_group);
3781 goto failed_mount;
3782 }
3783 } else {
3784 if (clustersize != blocksize) {
3785 ext4_warning(sb, "fragment/cluster size (%d) != "
3786 "block size (%d)", clustersize,
3787 blocksize);
3788 clustersize = blocksize;
3789 }
3790 if (sbi->s_blocks_per_group > blocksize * 8) {
3791 ext4_msg(sb, KERN_ERR,
3792 "#blocks per group too big: %lu",
3793 sbi->s_blocks_per_group);
3794 goto failed_mount;
3795 }
3796 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3797 sbi->s_cluster_bits = 0;
ac27a0ec 3798 }
281b5995
TT		 3799 sbi->s_cluster_ratio = clustersize / blocksize;
3800
960fd856
TT		 3801 /* Do we have standard group size of clustersize * 8 blocks ? */
3802 if (sbi->s_blocks_per_group == clustersize << 3)
3803 set_opt2(sb, STD_GROUP_SIZE);
3804
bf43d84b
ES		 3805 /*
3806 * Test whether we have more sectors than will fit in sector_t,
3807 * and whether the max offset is addressable by the page cache.
3808 */
5a9ae68a 3809 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3810 ext4_blocks_count(es));
5a9ae68a 3811 if (err) {
b31e1552 3812 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3813 " too large to mount safely on this system");
ac27a0ec 3814 if (sizeof(sector_t) < 8)
90c699a9 3815 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3816 goto failed_mount;
3817 }
3818
617ba13b
MC		 3819 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3820 goto cantfind_ext4;
e7c95593 3821
0f2ddca6
FTN		 3822 /* check blocks count against device size */
3823 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3824 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3825 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3826 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3827 ext4_blocks_count(es), blocks_count);
3828 goto failed_mount;
3829 }
3830
0b8e58a1
AD		 3831 /*
3832 * It makes no sense for the first data block to be beyond the end
3833 * of the filesystem.
3834 */
3835 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3836 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3837 "block %u is beyond end of filesystem (%llu)",
3838 le32_to_cpu(es->s_first_data_block),
3839 ext4_blocks_count(es));
e7c95593
ES		 3840 goto failed_mount;
3841 }
bd81d8ee
LV		 3842 blocks_count = (ext4_blocks_count(es) -
3843 le32_to_cpu(es->s_first_data_block) +
3844 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3845 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3846 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3847 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3848 "(block count %llu, first data block %u, "
b31e1552 3849 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3850 ext4_blocks_count(es),
3851 le32_to_cpu(es->s_first_data_block),
3852 EXT4_BLOCKS_PER_GROUP(sb));
3853 goto failed_mount;
3854 }
bd81d8ee 3855 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3856 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3857 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3858 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3859 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 3860 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 3861 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 3862 ext4_msg(sb, KERN_WARNING,
3863 "first meta block group too large: %u "
3864 "(group descriptor block count %u)",
3865 le32_to_cpu(es->s_first_meta_bg), db_count);
3866 goto failed_mount;
3867 }
3868 }
f18a5f21
TT		 3869 sbi->s_group_desc = ext4_kvmalloc(db_count *
3870 sizeof(struct buffer_head *),
3871 GFP_KERNEL);
ac27a0ec 3872 if (sbi->s_group_desc == NULL) {
b31e1552 3873 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3874 ret = -ENOMEM;
ac27a0ec
DK		 3875 goto failed_mount;
3876 }
3877
705895b6 3878 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3879
3880 for (i = 0; i < db_count; i++) {
70bbb3e0 3881 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3882 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3883 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3884 ext4_msg(sb, KERN_ERR,
3885 "can't read group descriptor %d", i);
ac27a0ec
DK		 3886 db_count = i;
3887 goto failed_mount2;
3888 }
3889 }
829fa70d 3890 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 3891 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 3892 ret = -EFSCORRUPTED;
f9ae9cf5 3893 goto failed_mount2;
ac27a0ec 3894 }
772cb7c8 3895
f9ae9cf5 3896 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3897 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3898 spin_lock_init(&sbi->s_next_gen_lock);
3899
04ecddb7
JM		 3900 setup_timer(&sbi->s_err_report, print_daily_error_info,
3901 (unsigned long) sb);
04496411 3902
a75ae78f 3903 /* Register extent status tree shrinker */
eb68d0e2 3904 if (ext4_es_register_shrinker(sbi))
ce7e010a 3905 goto failed_mount3;
ce7e010a 3906
c9de560d 3907 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3908 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3909
f9ae9cf5
TT		 3910 /*
3911 * set up enough so that it can read an inode
3912 */
f6e63f90 3913 sb->s_op = &ext4_sops;
617ba13b
MC		 3914 sb->s_export_op = &ext4_export_ops;
3915 sb->s_xattr = ext4_xattr_handlers;
a7550b30 3916 sb->s_cop = &ext4_cryptops;
ac27a0ec 3917#ifdef CONFIG_QUOTA
617ba13b 3918 sb->dq_op = &ext4_quota_operations;
e2b911c5 3919 if (ext4_has_feature_quota(sb))
1fa5efe3 3920 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3921 else
3922 sb->s_qcop = &ext4_qctl_operations;
689c958c 3923 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 3924#endif
f2fa2ffc
AK		 3925 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3926
ac27a0ec 3927 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3928 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3929
3930 sb->s_root = NULL;
3931
3932 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 3933 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 3934
e2b911c5 3935 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
c5e06d10 3936 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3937 goto failed_mount3a;
c5e06d10 3938
ac27a0ec
DK		 3939 /*
3940 * The first inode we look at is the journal inode. Don't try
3941 * root first: it may be modified in the journal!
3942 */
e2b911c5 3943 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 3944 err = ext4_load_journal(sb, es, journal_devnum);
3945 if (err)
50460fe8 3946 goto failed_mount3a;
0390131b 3947 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
e2b911c5 3948 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 3949 ext4_msg(sb, KERN_ERR, "required journal recovery "
3950 "suppressed and not mounted read-only");
744692dc 3951 goto failed_mount_wq;
ac27a0ec 3952 } else {
1e381f60
DM		 3953 /* Nojournal mode, all journal mount options are illegal */
3954 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3955 ext4_msg(sb, KERN_ERR, "can't mount with "
3956 "journal_checksum, fs mounted w/o journal");
3957 goto failed_mount_wq;
3958 }
3959 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3960 ext4_msg(sb, KERN_ERR, "can't mount with "
3961 "journal_async_commit, fs mounted w/o journal");
3962 goto failed_mount_wq;
3963 }
3964 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3965 ext4_msg(sb, KERN_ERR, "can't mount with "
3966 "commit=%lu, fs mounted w/o journal",
3967 sbi->s_commit_interval / HZ);
3968 goto failed_mount_wq;
3969 }
3970 if (EXT4_MOUNT_DATA_FLAGS &
3971 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3972 ext4_msg(sb, KERN_ERR, "can't mount with "
3973 "data=, fs mounted w/o journal");
3974 goto failed_mount_wq;
3975 }
3976 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3977 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 3978 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3979 sbi->s_journal = NULL;
3980 needs_recovery = 0;
3981 goto no_journal;
ac27a0ec
DK		 3982 }
3983
e2b911c5 3984 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 3985 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3986 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3987 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3988 goto failed_mount_wq;
eb40a09c
JS		 3989 }
3990
25ed6e8a
DW		 3991 if (!set_journal_csum_feature_set(sb)) {
3992 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3993 "feature set");
3994 goto failed_mount_wq;
d4da6c9c 3995 }
818d276c 3996
ac27a0ec
DK		 3997 /* We have now updated the journal if required, so we can
3998 * validate the data journaling mode. */
3999 switch (test_opt(sb, DATA_FLAGS)) {
4000 case 0:
4001 /* No mode set, assume a default based on the journal
63f57933
AM		 4002 * capabilities: ORDERED_DATA if the journal can
4003 * cope, else JOURNAL_DATA
4004 */
dab291af
MC		 4005 if (jbd2_journal_check_available_features
4006 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 4007 set_opt(sb, ORDERED_DATA);
ac27a0ec 4008 else
fd8c37ec 4009 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 4010 break;
4011
617ba13b
MC		 4012 case EXT4_MOUNT_ORDERED_DATA:
4013 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4014 if (!jbd2_journal_check_available_features
4015 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4016 ext4_msg(sb, KERN_ERR, "Journal does not support "
4017 "requested data journaling mode");
744692dc 4018 goto failed_mount_wq;
ac27a0ec
DK		 4019 }
4020 default:
4021 break;
4022 }
ab04df78
JK		 4023
4024 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4025 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4026 ext4_msg(sb, KERN_ERR, "can't mount with "
4027 "journal_async_commit in data=ordered mode");
4028 goto failed_mount_wq;
4029 }
4030
b3881f74 4031 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4032
18aadd47
BJ		 4033 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4034
ce7e010a 4035no_journal:
2335d05f
AG		 4036 sbi->s_mb_cache = ext4_xattr_create_cache();
4037 if (!sbi->s_mb_cache) {
4038 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4039 goto failed_mount_wq;
9c191f70
M		 4040 }
4041
e2b911c5 4042 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 4043 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 4044 ext4_msg(sb, KERN_ERR,
4045 "Unsupported blocksize for fs encryption");
4046 goto failed_mount_wq;
4047 }
4048
e2b911c5
DW		 4049 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4050 !ext4_has_feature_encrypt(sb)) {
4051 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4052 ext4_commit_super(sb, 1);
4053 }
4054
952fc18e
TT		 4055 /*
4056 * Get the # of file system overhead blocks from the
4057 * superblock if present.
4058 */
4059 if (es->s_overhead_clusters)
4060 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4061 else {
07aa2ea1
LC		 4062 err = ext4_calculate_overhead(sb);
4063 if (err)
952fc18e
TT		 4064 goto failed_mount_wq;
4065 }
4066
fd89d5f2
TH		 4067 /*
4068 * The maximum number of concurrent works can be high and
4069 * concurrency isn't really necessary. Limit it to 1.
4070 */
2e8fa54e
JK		 4071 EXT4_SB(sb)->rsv_conversion_wq =
4072 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4073 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4074 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4075 ret = -ENOMEM;
2e8fa54e
JK		 4076 goto failed_mount4;
4077 }
4078
ac27a0ec 4079 /*
dab291af 4080 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4081 * so we can safely mount the rest of the filesystem now.
4082 */
4083
1d1fe1ee
DH		 4084 root = ext4_iget(sb, EXT4_ROOT_INO);
4085 if (IS_ERR(root)) {
b31e1552 4086 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4087 ret = PTR_ERR(root);
32a9bb57 4088 root = NULL;
ac27a0ec
DK		 4089 goto failed_mount4;
4090 }
4091 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4092 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4093 iput(root);
ac27a0ec
DK		 4094 goto failed_mount4;
4095 }
48fde701 4096 sb->s_root = d_make_root(root);
1d1fe1ee 4097 if (!sb->s_root) {
b31e1552 4098 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4099 ret = -ENOMEM;
4100 goto failed_mount4;
4101 }
ac27a0ec 4102
7e84b621
ES		 4103 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4104 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4105
4106 /* determine the minimum size of new large inodes, if present */
670e9875
TT		 4107 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4108 sbi->s_want_extra_isize == 0) {
ef7f3835
KS		 4109 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4110 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4111 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4112 if (sbi->s_want_extra_isize <
4113 le16_to_cpu(es->s_want_extra_isize))
4114 sbi->s_want_extra_isize =
4115 le16_to_cpu(es->s_want_extra_isize);
4116 if (sbi->s_want_extra_isize <
4117 le16_to_cpu(es->s_min_extra_isize))
4118 sbi->s_want_extra_isize =
4119 le16_to_cpu(es->s_min_extra_isize);
4120 }
4121 }
4122 /* Check if enough inode space is available */
4123 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4124 sbi->s_inode_size) {
4125 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4126 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4127 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4128 "available");
ef7f3835
KS		 4129 }
4130
b5799018 4131 ext4_set_resv_clusters(sb);
27dd4385 4132
6fd058f7
TT		 4133 err = ext4_setup_system_zone(sb);
4134 if (err) {
b31e1552 4135 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4136 "zone (%d)", err);
f9ae9cf5
TT		 4137 goto failed_mount4a;
4138 }
4139
4140 ext4_ext_init(sb);
4141 err = ext4_mb_init(sb);
4142 if (err) {
4143 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4144 err);
dcf2d804 4145 goto failed_mount5;
c2774d84
AK		 4146 }
4147
d5e03cbb
TT		 4148 block = ext4_count_free_clusters(sb);
4149 ext4_free_blocks_count_set(sbi->s_es,
4150 EXT4_C2B(sbi, block));
908c7f19
TH		 4151 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4152 GFP_KERNEL);
d5e03cbb
TT		 4153 if (!err) {
4154 unsigned long freei = ext4_count_free_inodes(sb);
4155 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4156 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4157 GFP_KERNEL);
d5e03cbb
TT		 4158 }
4159 if (!err)
4160 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4161 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4162 if (!err)
908c7f19
TH		 4163 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4164 GFP_KERNEL);
c8585c6f
DJ		 4165 if (!err)
4166 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4167
d5e03cbb
TT		 4168 if (err) {
4169 ext4_msg(sb, KERN_ERR, "insufficient memory");
4170 goto failed_mount6;
4171 }
4172
e2b911c5 4173 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4174 if (!ext4_fill_flex_info(sb)) {
4175 ext4_msg(sb, KERN_ERR,
4176 "unable to initialize "
4177 "flex_bg meta info!");
4178 goto failed_mount6;
4179 }
4180
bfff6873
LC		 4181 err = ext4_register_li_request(sb, first_not_zeroed);
4182 if (err)
dcf2d804 4183 goto failed_mount6;
bfff6873 4184
b5799018 4185 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4186 if (err)
4187 goto failed_mount7;
3197ebdb 4188
9b2ff357
JK		 4189#ifdef CONFIG_QUOTA
4190 /* Enable quota usage during mount. */
e2b911c5 4191 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
9b2ff357
JK		 4192 err = ext4_enable_quotas(sb);
4193 if (err)
4194 goto failed_mount8;
4195 }
4196#endif /* CONFIG_QUOTA */
4197
617ba13b
MC		 4198 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4199 ext4_orphan_cleanup(sb, es);
4200 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4201 if (needs_recovery) {
b31e1552 4202 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4203 ext4_mark_recovery_complete(sb, es);
4204 }
4205 if (EXT4_SB(sb)->s_journal) {
4206 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4207 descr = " journalled data mode";
4208 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4209 descr = " ordered data mode";
4210 else
4211 descr = " writeback data mode";
4212 } else
4213 descr = "out journal";
4214
79add3a3
LC		 4215 if (test_opt(sb, DISCARD)) {
4216 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4217 if (!blk_queue_discard(q))
4218 ext4_msg(sb, KERN_WARNING,
4219 "mounting with \"discard\" option, but "
4220 "the device does not support discard");
4221 }
4222
e294a537
TT		 4223 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4224 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4225 "Opts: %.*s%s%s", descr,
4226 (int) sizeof(sbi->s_es->s_mount_opts),
4227 sbi->s_es->s_mount_opts,
e294a537 4228 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4229
66e61a9e
TT		 4230 if (es->s_error_count)
4231 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4232
efbed4dc
TT		 4233 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4234 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4235 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4236 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4237
d4c402d9 4238 kfree(orig_data);
ac27a0ec
DK		 4239 return 0;
4240
617ba13b 4241cantfind_ext4:
ac27a0ec 4242 if (!silent)
b31e1552 4243 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4244 goto failed_mount;
4245
72ba7450
TT		 4246#ifdef CONFIG_QUOTA
4247failed_mount8:
ebd173be 4248 ext4_unregister_sysfs(sb);
72ba7450 4249#endif
dcf2d804
TM		 4250failed_mount7:
4251 ext4_unregister_li_request(sb);
4252failed_mount6:
f9ae9cf5 4253 ext4_mb_release(sb);
d5e03cbb 4254 if (sbi->s_flex_groups)
b93b41d4 4255 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4256 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4257 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4258 percpu_counter_destroy(&sbi->s_dirs_counter);
4259 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4260failed_mount5:
f9ae9cf5
TT		 4261 ext4_ext_release(sb);
4262 ext4_release_system_zone(sb);
4263failed_mount4a:
94bf608a 4264 dput(sb->s_root);
32a9bb57 4265 sb->s_root = NULL;
94bf608a 4266failed_mount4:
b31e1552 4267 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4268 if (EXT4_SB(sb)->rsv_conversion_wq)
4269 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4270failed_mount_wq:
82939d79
JK		 4271 if (sbi->s_mb_cache) {
4272 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4273 sbi->s_mb_cache = NULL;
4274 }
0390131b
FM		 4275 if (sbi->s_journal) {
4276 jbd2_journal_destroy(sbi->s_journal);
4277 sbi->s_journal = NULL;
4278 }
50460fe8 4279failed_mount3a:
d3922a77 4280 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4281failed_mount3:
9105bb14 4282 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4283 if (sbi->s_mmp_tsk)
4284 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4285failed_mount2:
4286 for (i = 0; i < db_count; i++)
4287 brelse(sbi->s_group_desc[i]);
b93b41d4 4288 kvfree(sbi->s_group_desc);
ac27a0ec 4289failed_mount:
0441984a
DW		 4290 if (sbi->s_chksum_driver)
4291 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4292#ifdef CONFIG_QUOTA
a2d4a646 4293 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4294 kfree(sbi->s_qf_names[i]);
4295#endif
617ba13b 4296 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4297 brelse(bh);
4298out_fail:
4299 sb->s_fs_info = NULL;
f6830165 4300 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4301out_free_base:
ac27a0ec 4302 kfree(sbi);
d4c402d9 4303 kfree(orig_data);
07aa2ea1 4304 return err ? err : ret;
ac27a0ec
DK		 4305}
4306
4307/*
4308 * Setup any per-fs journal parameters now. We'll do this both on
4309 * initial mount, once the journal has been initialised but before we've
4310 * done any recovery; and again on any subsequent remount.
4311 */
617ba13b 4312static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4313{
617ba13b 4314 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4315
30773840
TT		 4316 journal->j_commit_interval = sbi->s_commit_interval;
4317 journal->j_min_batch_time = sbi->s_min_batch_time;
4318 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4319
a931da6a 4320 write_lock(&journal->j_state_lock);
ac27a0ec 4321 if (test_opt(sb, BARRIER))
dab291af 4322 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4323 else
dab291af 4324 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4325 if (test_opt(sb, DATA_ERR_ABORT))
4326 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4327 else
4328 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4329 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4330}
4331
c6cb7e77
EW		 4332static struct inode *ext4_get_journal_inode(struct super_block *sb,
4333 unsigned int journal_inum)
ac27a0ec
DK		 4334{
4335 struct inode *journal_inode;
ac27a0ec 4336
c6cb7e77
EW		 4337 /*
4338 * Test for the existence of a valid inode on disk. Bad things
4339 * happen if we iget() an unused inode, as the subsequent iput()
4340 * will try to delete it.
4341 */
1d1fe1ee
DH		 4342 journal_inode = ext4_iget(sb, journal_inum);
4343 if (IS_ERR(journal_inode)) {
b31e1552 4344 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4345 return NULL;
4346 }
4347 if (!journal_inode->i_nlink) {
4348 make_bad_inode(journal_inode);
4349 iput(journal_inode);
b31e1552 4350 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4351 return NULL;
4352 }
4353
e5f8eab8 4354 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4355 journal_inode, journal_inode->i_size);
1d1fe1ee 4356 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4357 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4358 iput(journal_inode);
4359 return NULL;
4360 }
c6cb7e77
EW		 4361 return journal_inode;
4362}
4363
4364static journal_t *ext4_get_journal(struct super_block *sb,
4365 unsigned int journal_inum)
4366{
4367 struct inode *journal_inode;
4368 journal_t *journal;
4369
4370 BUG_ON(!ext4_has_feature_journal(sb));
4371
4372 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4373 if (!journal_inode)
4374 return NULL;
ac27a0ec 4375
dab291af 4376 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4377 if (!journal) {
b31e1552 4378 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4379 iput(journal_inode);
4380 return NULL;
4381 }
4382 journal->j_private = sb;
617ba13b 4383 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4384 return journal;
4385}
4386
617ba13b 4387static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4388 dev_t j_dev)
4389{
2b2d6d01 4390 struct buffer_head *bh;
ac27a0ec 4391 journal_t *journal;
617ba13b
MC		 4392 ext4_fsblk_t start;
4393 ext4_fsblk_t len;
ac27a0ec 4394 int hblock, blocksize;
617ba13b 4395 ext4_fsblk_t sb_block;
ac27a0ec 4396 unsigned long offset;
2b2d6d01 4397 struct ext4_super_block *es;
ac27a0ec
DK		 4398 struct block_device *bdev;
4399
e2b911c5 4400 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4401
b31e1552 4402 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4403 if (bdev == NULL)
4404 return NULL;
4405
ac27a0ec 4406 blocksize = sb->s_blocksize;
e1defc4f 4407 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4408 if (blocksize < hblock) {
b31e1552
ES		 4409 ext4_msg(sb, KERN_ERR,
4410 "blocksize too small for journal device");
ac27a0ec
DK		 4411 goto out_bdev;
4412 }
4413
617ba13b
MC		 4414 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4415 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4416 set_blocksize(bdev, blocksize);
4417 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4418 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4419 "external journal");
ac27a0ec
DK		 4420 goto out_bdev;
4421 }
4422
2716b802 4423 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4424 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4425 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4426 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4427 ext4_msg(sb, KERN_ERR, "external journal has "
4428 "bad superblock");
ac27a0ec
DK		 4429 brelse(bh);
4430 goto out_bdev;
4431 }
4432
df4763be
DW		 4433 if ((le32_to_cpu(es->s_feature_ro_compat) &
4434 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4435 es->s_checksum != ext4_superblock_csum(sb, es)) {
4436 ext4_msg(sb, KERN_ERR, "external journal has "
4437 "corrupt superblock");
4438 brelse(bh);
4439 goto out_bdev;
4440 }
4441
617ba13b 4442 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4443 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4444 brelse(bh);
4445 goto out_bdev;
4446 }
4447
bd81d8ee 4448 len = ext4_blocks_count(es);
ac27a0ec
DK		 4449 start = sb_block + 1;
4450 brelse(bh); /* we're done with the superblock */
4451
dab291af 4452 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4453 start, len, blocksize);
4454 if (!journal) {
b31e1552 4455 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4456 goto out_bdev;
4457 }
4458 journal->j_private = sb;
dfec8a14 4459 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4460 wait_on_buffer(journal->j_sb_buffer);
4461 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4462 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4463 goto out_journal;
4464 }
4465 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4466 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4467 "user (unsupported) - %d",
ac27a0ec
DK		 4468 be32_to_cpu(journal->j_superblock->s_nr_users));
4469 goto out_journal;
4470 }
617ba13b
MC		 4471 EXT4_SB(sb)->journal_bdev = bdev;
4472 ext4_init_journal_params(sb, journal);
ac27a0ec 4473 return journal;
0b8e58a1 4474
ac27a0ec 4475out_journal:
dab291af 4476 jbd2_journal_destroy(journal);
ac27a0ec 4477out_bdev:
617ba13b 4478 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4479 return NULL;
4480}
4481
617ba13b
MC		 4482static int ext4_load_journal(struct super_block *sb,
4483 struct ext4_super_block *es,
ac27a0ec
DK		 4484 unsigned long journal_devnum)
4485{
4486 journal_t *journal;
4487 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4488 dev_t journal_dev;
4489 int err = 0;
4490 int really_read_only;
4491
e2b911c5 4492 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4493
ac27a0ec
DK		 4494 if (journal_devnum &&
4495 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4496 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4497 "numbers have changed");
ac27a0ec
DK		 4498 journal_dev = new_decode_dev(journal_devnum);
4499 } else
4500 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4501
4502 really_read_only = bdev_read_only(sb->s_bdev);
4503
4504 /*
4505 * Are we loading a blank journal or performing recovery after a
4506 * crash? For recovery, we need to check in advance whether we
4507 * can get read-write access to the device.
4508 */
e2b911c5 4509 if (ext4_has_feature_journal_needs_recovery(sb)) {
ac27a0ec 4510 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4511 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4512 "required on readonly filesystem");
ac27a0ec 4513 if (really_read_only) {
b31e1552
ES		 4514 ext4_msg(sb, KERN_ERR, "write access "
4515 "unavailable, cannot proceed");
ac27a0ec
DK		 4516 return -EROFS;
4517 }
b31e1552
ES		 4518 ext4_msg(sb, KERN_INFO, "write access will "
4519 "be enabled during recovery");
ac27a0ec
DK		 4520 }
4521 }
4522
4523 if (journal_inum && journal_dev) {
b31e1552
ES		 4524 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4525 "and inode journals!");
ac27a0ec
DK		 4526 return -EINVAL;
4527 }
4528
4529 if (journal_inum) {
617ba13b 4530 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4531 return -EINVAL;
4532 } else {
617ba13b 4533 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4534 return -EINVAL;
4535 }
4536
90576c0b 4537 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4538 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4539
e2b911c5 4540 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4541 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4542 if (!err) {
4543 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4544 if (save)
4545 memcpy(save, ((char *) es) +
4546 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4547 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4548 if (save)
4549 memcpy(((char *) es) + EXT4_S_ERR_START,
4550 save, EXT4_S_ERR_LEN);
4551 kfree(save);
4552 }
ac27a0ec
DK		 4553
4554 if (err) {
b31e1552 4555 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4556 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4557 return err;
4558 }
4559
617ba13b
MC		 4560 EXT4_SB(sb)->s_journal = journal;
4561 ext4_clear_journal_err(sb, es);
ac27a0ec 4562
c41303ce 4563 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4564 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4565 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4566
4567 /* Make sure we flush the recovery flag to disk. */
e2d67052 4568 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4569 }
4570
4571 return 0;
4572}
4573
e2d67052 4574static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4575{
e2d67052 4576 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4577 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4578 int error = 0;
ac27a0ec 4579
bdfe0cbd 4580 if (!sbh || block_device_ejected(sb))
c4be0c1d 4581 return error;
71290b36
TT		 4582 /*
4583 * If the file system is mounted read-only, don't update the
4584 * superblock write time. This avoids updating the superblock
4585 * write time when we are mounting the root file system
4586 * read/only but we need to replay the journal; at that point,
4587 * for people who are east of GMT and who make their clock
4588 * tick in localtime for Windows bug-for-bug compatibility,
4589 * the clock is set in the future, and this will cause e2fsck
4590 * to complain and force a full file system check.
4591 */
4592 if (!(sb->s_flags & MS_RDONLY))
4593 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4594 if (sb->s_bdev->bd_part)
4595 es->s_kbytes_written =
4596 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4597 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4598 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4599 else
4600 es->s_kbytes_written =
4601 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4602 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4603 ext4_free_blocks_count_set(es,
57042651
TT		 4604 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4605 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4606 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4607 es->s_free_inodes_count =
4608 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4609 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4610 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4611 ext4_superblock_csum_set(sb);
1566a48a
TT		 4612 if (sync)
4613 lock_buffer(sbh);
4743f839
PKS		 4614 if (buffer_write_io_error(sbh)) {
4615 /*
4616 * Oh, dear. A previous attempt to write the
4617 * superblock failed. This could happen because the
4618 * USB device was yanked out. Or it could happen to
4619 * be a transient write error and maybe the block will
4620 * be remapped. Nothing we can do but to retry the
4621 * write and hope for the best.
4622 */
4623 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4624 "superblock detected");
4625 clear_buffer_write_io_error(sbh);
4626 set_buffer_uptodate(sbh);
4627 }
ac27a0ec 4628 mark_buffer_dirty(sbh);
914258bf 4629 if (sync) {
1566a48a 4630 unlock_buffer(sbh);
564bc402 4631 error = __sync_dirty_buffer(sbh,
70fd7614 4632 test_opt(sb, BARRIER) ? REQ_FUA : REQ_SYNC);
c4be0c1d
TS		 4633 if (error)
4634 return error;
4635
4636 error = buffer_write_io_error(sbh);
4637 if (error) {
b31e1552
ES		 4638 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4639 "superblock");
914258bf
TT		 4640 clear_buffer_write_io_error(sbh);
4641 set_buffer_uptodate(sbh);
4642 }
4643 }
c4be0c1d 4644 return error;
ac27a0ec
DK		 4645}
4646
ac27a0ec
DK		 4647/*
4648 * Have we just finished recovery? If so, and if we are mounting (or
4649 * remounting) the filesystem readonly, then we will end up with a
4650 * consistent fs on disk. Record that fact.
4651 */
2b2d6d01
TT		 4652static void ext4_mark_recovery_complete(struct super_block *sb,
4653 struct ext4_super_block *es)
ac27a0ec 4654{
617ba13b 4655 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4656
e2b911c5 4657 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4658 BUG_ON(journal != NULL);
4659 return;
4660 }
dab291af 4661 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4662 if (jbd2_journal_flush(journal) < 0)
4663 goto out;
4664
e2b911c5 4665 if (ext4_has_feature_journal_needs_recovery(sb) &&
ac27a0ec 4666 sb->s_flags & MS_RDONLY) {
e2b911c5 4667 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4668 ext4_commit_super(sb, 1);
ac27a0ec 4669 }
7ffe1ea8
HK		 4670
4671out:
dab291af 4672 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4673}
4674
4675/*
4676 * If we are mounting (or read-write remounting) a filesystem whose journal
4677 * has recorded an error from a previous lifetime, move that error to the
4678 * main filesystem now.
4679 */
2b2d6d01
TT		 4680static void ext4_clear_journal_err(struct super_block *sb,
4681 struct ext4_super_block *es)
ac27a0ec
DK		 4682{
4683 journal_t *journal;
4684 int j_errno;
4685 const char *errstr;
4686
e2b911c5 4687 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4688
617ba13b 4689 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4690
4691 /*
4692 * Now check for any error status which may have been recorded in the
617ba13b 4693 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4694 */
4695
dab291af 4696 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4697 if (j_errno) {
4698 char nbuf[16];
4699
617ba13b 4700 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4701 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4702 "from previous mount: %s", errstr);
12062ddd 4703 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4704
617ba13b
MC		 4705 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4706 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4707 ext4_commit_super(sb, 1);
ac27a0ec 4708
dab291af 4709 jbd2_journal_clear_err(journal);
d796c52e 4710 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4711 }
4712}
4713
4714/*
4715 * Force the running and committing transactions to commit,
4716 * and wait on the commit.
4717 */
617ba13b 4718int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4719{
4720 journal_t *journal;
ac27a0ec
DK		 4721
4722 if (sb->s_flags & MS_RDONLY)
4723 return 0;
4724
617ba13b 4725 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4726 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4727}
4728
617ba13b 4729static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4730{
14ce0cb4 4731 int ret = 0;
9eddacf9 4732 tid_t target;
06a407f1 4733 bool needs_barrier = false;
8d5d02e6 4734 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4735
0db1ff22
TT		 4736 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4737 return 0;
4738
9bffad1e 4739 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4740 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4741 /*
4742 * Writeback quota in non-journalled quota case - journalled quota has
4743 * no dirty dquots
4744 */
4745 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4746 /*
4747 * Data writeback is possible w/o journal transaction, so barrier must
4748 * being sent at the end of the function. But we can skip it if
4749 * transaction_commit will do it for us.
4750 */
bda32530
TT		 4751 if (sbi->s_journal) {
4752 target = jbd2_get_latest_transaction(sbi->s_journal);
4753 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4754 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4755 needs_barrier = true;
4756
4757 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4758 if (wait)
4759 ret = jbd2_log_wait_commit(sbi->s_journal,
4760 target);
4761 }
4762 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4763 needs_barrier = true;
06a407f1
DM		 4764 if (needs_barrier) {
4765 int err;
4766 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4767 if (!ret)
4768 ret = err;
0390131b 4769 }
06a407f1
DM		 4770
4771 return ret;
4772}
4773
ac27a0ec
DK		 4774/*
4775 * LVM calls this function before a (read-only) snapshot is created. This
4776 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4777 *
4778 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4779 * state independently. It relies on upper layer to stop all data & metadata
4780 * modifications.
ac27a0ec 4781 */
c4be0c1d 4782static int ext4_freeze(struct super_block *sb)
ac27a0ec 4783{
c4be0c1d
TS		 4784 int error = 0;
4785 journal_t *journal;
ac27a0ec 4786
9ca92389
TT		 4787 if (sb->s_flags & MS_RDONLY)
4788 return 0;
ac27a0ec 4789
9ca92389 4790 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4791
bb044576
TT		 4792 if (journal) {
4793 /* Now we set up the journal barrier. */
4794 jbd2_journal_lock_updates(journal);
ac27a0ec 4795
bb044576
TT		 4796 /*
4797 * Don't clear the needs_recovery flag if we failed to
4798 * flush the journal.
4799 */
4800 error = jbd2_journal_flush(journal);
4801 if (error < 0)
4802 goto out;
c642dc9e
ES		 4803
4804 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4805 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4806 }
9ca92389 4807
9ca92389 4808 error = ext4_commit_super(sb, 1);
6b0310fb 4809out:
bb044576
TT		 4810 if (journal)
4811 /* we rely on upper layer to stop further updates */
4812 jbd2_journal_unlock_updates(journal);
6b0310fb 4813 return error;
ac27a0ec
DK		 4814}
4815
4816/*
4817 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4818 * flag here, even though the filesystem is not technically dirty yet.
4819 */
c4be0c1d 4820static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4821{
783d9485 4822 if ((sb->s_flags & MS_RDONLY) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 4823 return 0;
4824
c642dc9e
ES		 4825 if (EXT4_SB(sb)->s_journal) {
4826 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4827 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4828 }
4829
9ca92389 4830 ext4_commit_super(sb, 1);
c4be0c1d 4831 return 0;
ac27a0ec
DK		 4832}
4833
673c6100
TT		 4834/*
4835 * Structure to save mount options for ext4_remount's benefit
4836 */
4837struct ext4_mount_options {
4838 unsigned long s_mount_opt;
a2595b8a 4839 unsigned long s_mount_opt2;
08cefc7a
EB		 4840 kuid_t s_resuid;
4841 kgid_t s_resgid;
673c6100
TT		 4842 unsigned long s_commit_interval;
4843 u32 s_min_batch_time, s_max_batch_time;
4844#ifdef CONFIG_QUOTA
4845 int s_jquota_fmt;
a2d4a646 4846 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4847#endif
4848};
4849
2b2d6d01 4850static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4851{
2b2d6d01 4852 struct ext4_super_block *es;
617ba13b 4853 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4854 unsigned long old_sb_flags;
617ba13b 4855 struct ext4_mount_options old_opts;
c79d967d 4856 int enable_quota = 0;
8a266467 4857 ext4_group_t g;
b3881f74 4858 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4859 int err = 0;
ac27a0ec 4860#ifdef CONFIG_QUOTA
03dafb5f 4861 int i, j;
ac27a0ec 4862#endif
d4c402d9 4863 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4864
4865 /* Store the original options */
4866 old_sb_flags = sb->s_flags;
4867 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4868 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4869 old_opts.s_resuid = sbi->s_resuid;
4870 old_opts.s_resgid = sbi->s_resgid;
4871 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4872 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4873 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4874#ifdef CONFIG_QUOTA
4875 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4876 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4877 if (sbi->s_qf_names[i]) {
4878 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4879 GFP_KERNEL);
4880 if (!old_opts.s_qf_names[i]) {
4881 for (j = 0; j < i; j++)
4882 kfree(old_opts.s_qf_names[j]);
3e36a163 4883 kfree(orig_data);
03dafb5f
CG		 4884 return -ENOMEM;
4885 }
4886 } else
4887 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4888#endif
b3881f74
TT		 4889 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4890 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4891
661aa520 4892 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4893 err = -EINVAL;
4894 goto restore_opts;
4895 }
4896
6b992ff2 4897 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4898 test_opt(sb, JOURNAL_CHECKSUM)) {
4899 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4900 "during remount not supported; ignoring");
4901 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4902 }
4903
6ae6514b
PS		 4904 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4905 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4906 ext4_msg(sb, KERN_ERR, "can't mount with "
4907 "both data=journal and delalloc");
4908 err = -EINVAL;
4909 goto restore_opts;
4910 }
4911 if (test_opt(sb, DIOREAD_NOLOCK)) {
4912 ext4_msg(sb, KERN_ERR, "can't mount with "
4913 "both data=journal and dioread_nolock");
4914 err = -EINVAL;
4915 goto restore_opts;
4916 }
923ae0ff
RZ		 4917 if (test_opt(sb, DAX)) {
4918 ext4_msg(sb, KERN_ERR, "can't mount with "
4919 "both data=journal and dax");
4920 err = -EINVAL;
4921 goto restore_opts;
4922 }
ab04df78
JK		 4923 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
4924 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4925 ext4_msg(sb, KERN_ERR, "can't mount with "
4926 "journal_async_commit in data=ordered mode");
4927 err = -EINVAL;
4928 goto restore_opts;
4929 }
923ae0ff
RZ		 4930 }
4931
4932 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4933 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4934 "dax flag with busy inodes while remounting");
4935 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4936 }
4937
4ab2f15b 4938 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4939 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4940
4941 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4942 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4943
4944 es = sbi->s_es;
4945
b3881f74 4946 if (sbi->s_journal) {
0390131b 4947 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4948 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4949 }
ac27a0ec 4950
a2fd66d0
TT		 4951 if (*flags & MS_LAZYTIME)
4952 sb->s_flags |= MS_LAZYTIME;
4953
661aa520 4954 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4955 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4956 err = -EROFS;
4957 goto restore_opts;
4958 }
4959
4960 if (*flags & MS_RDONLY) {
38c03b34
TT		 4961 err = sync_filesystem(sb);
4962 if (err < 0)
4963 goto restore_opts;
0f0dd62f
CH		 4964 err = dquot_suspend(sb, -1);
4965 if (err < 0)
c79d967d 4966 goto restore_opts;
c79d967d 4967
ac27a0ec
DK		 4968 /*
4969 * First of all, the unconditional stuff we have to do
4970 * to disable replay of the journal when we next remount
4971 */
4972 sb->s_flags |= MS_RDONLY;
4973
4974 /*
4975 * OK, test if we are remounting a valid rw partition
4976 * readonly, and if so set the rdonly flag and then
4977 * mark the partition as valid again.
4978 */
617ba13b
MC		 4979 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4980 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4981 es->s_state = cpu_to_le16(sbi->s_mount_state);
4982
a63c9eb2 4983 if (sbi->s_journal)
0390131b 4984 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4985 } else {
a13fb1a4 4986 /* Make sure we can mount this feature set readwrite */
e2b911c5 4987 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 4988 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4989 err = -EROFS;
4990 goto restore_opts;
4991 }
8a266467
TT		 4992 /*
4993 * Make sure the group descriptor checksums
0b8e58a1 4994 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4995 */
4996 for (g = 0; g < sbi->s_groups_count; g++) {
4997 struct ext4_group_desc *gdp =
4998 ext4_get_group_desc(sb, g, NULL);
4999
feb0ab32 5000 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5001 ext4_msg(sb, KERN_ERR,
5002 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5003 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5004 le16_to_cpu(gdp->bg_checksum));
6a797d27 5005 err = -EFSBADCRC;
8a266467
TT		 5006 goto restore_opts;
5007 }
5008 }
5009
ead6596b
ES		 5010 /*
5011 * If we have an unprocessed orphan list hanging
5012 * around from a previously readonly bdev mount,
5013 * require a full umount/remount for now.
5014 */
5015 if (es->s_last_orphan) {
b31e1552 5016 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5017 "remount RDWR because of unprocessed "
5018 "orphan inode list. Please "
b31e1552 5019 "umount/remount instead");
ead6596b
ES		 5020 err = -EINVAL;
5021 goto restore_opts;
5022 }
5023
ac27a0ec
DK		 5024 /*
5025 * Mounting a RDONLY partition read-write, so reread
5026 * and store the current valid flag. (It may have
5027 * been changed by e2fsck since we originally mounted
5028 * the partition.)
5029 */
0390131b
FM		 5030 if (sbi->s_journal)
5031 ext4_clear_journal_err(sb, es);
ac27a0ec 5032 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 5033 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 5034 sb->s_flags &= ~MS_RDONLY;
e2b911c5 5035 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5036 if (ext4_multi_mount_protect(sb,
5037 le64_to_cpu(es->s_mmp_block))) {
5038 err = -EROFS;
5039 goto restore_opts;
5040 }
c79d967d 5041 enable_quota = 1;
ac27a0ec
DK		 5042 }
5043 }
bfff6873
LC		 5044
5045 /*
5046 * Reinitialize lazy itable initialization thread based on
5047 * current settings
5048 */
5049 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5050 ext4_unregister_li_request(sb);
5051 else {
5052 ext4_group_t first_not_zeroed;
5053 first_not_zeroed = ext4_has_uninit_itable(sb);
5054 ext4_register_li_request(sb, first_not_zeroed);
5055 }
5056
6fd058f7 5057 ext4_setup_system_zone(sb);
d096ad0f 5058 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5059 ext4_commit_super(sb, 1);
0390131b 5060
ac27a0ec
DK		 5061#ifdef CONFIG_QUOTA
5062 /* Release old quota file names */
a2d4a646 5063 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5064 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5065 if (enable_quota) {
5066 if (sb_any_quota_suspended(sb))
5067 dquot_resume(sb, -1);
e2b911c5 5068 else if (ext4_has_feature_quota(sb)) {
7c319d32 5069 err = ext4_enable_quotas(sb);
07724f98 5070 if (err)
7c319d32 5071 goto restore_opts;
7c319d32
AK		 5072 }
5073 }
ac27a0ec 5074#endif
d4c402d9 5075
a26f4992 5076 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5077 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5078 kfree(orig_data);
ac27a0ec 5079 return 0;
0b8e58a1 5080
ac27a0ec
DK		 5081restore_opts:
5082 sb->s_flags = old_sb_flags;
5083 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5084 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5085 sbi->s_resuid = old_opts.s_resuid;
5086 sbi->s_resgid = old_opts.s_resgid;
5087 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5088 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5089 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5090#ifdef CONFIG_QUOTA
5091 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5092 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5093 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5094 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5095 }
5096#endif
d4c402d9 5097 kfree(orig_data);
ac27a0ec
DK		 5098 return err;
5099}
5100
689c958c
LX		 5101#ifdef CONFIG_QUOTA
5102static int ext4_statfs_project(struct super_block *sb,
5103 kprojid_t projid, struct kstatfs *buf)
5104{
5105 struct kqid qid;
5106 struct dquot *dquot;
5107 u64 limit;
5108 u64 curblock;
5109
5110 qid = make_kqid_projid(projid);
5111 dquot = dqget(sb, qid);
5112 if (IS_ERR(dquot))
5113 return PTR_ERR(dquot);
5114 spin_lock(&dq_data_lock);
5115
5116 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5117 dquot->dq_dqb.dqb_bsoftlimit :
5118 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5119 if (limit && buf->f_blocks > limit) {
5120 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5121 buf->f_blocks = limit;
5122 buf->f_bfree = buf->f_bavail =
5123 (buf->f_blocks > curblock) ?
5124 (buf->f_blocks - curblock) : 0;
5125 }
5126
5127 limit = dquot->dq_dqb.dqb_isoftlimit ?
5128 dquot->dq_dqb.dqb_isoftlimit :
5129 dquot->dq_dqb.dqb_ihardlimit;
5130 if (limit && buf->f_files > limit) {
5131 buf->f_files = limit;
5132 buf->f_ffree =
5133 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5134 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5135 }
5136
5137 spin_unlock(&dq_data_lock);
5138 dqput(dquot);
5139 return 0;
5140}
5141#endif
5142
2b2d6d01 5143static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5144{
5145 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5146 struct ext4_sb_info *sbi = EXT4_SB(sb);
5147 struct ext4_super_block *es = sbi->s_es;
27dd4385 5148 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5149 u64 fsid;
d02a9391 5150 s64 bfree;
27dd4385 5151 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5152
952fc18e
TT		 5153 if (!test_opt(sb, MINIX_DF))
5154 overhead = sbi->s_overhead;
ac27a0ec 5155
617ba13b 5156 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5157 buf->f_bsize = sb->s_blocksize;
b72f78cb 5158 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5159 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5160 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5161 /* prevent underflow in case that few free space is available */
57042651 5162 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5163 buf->f_bavail = buf->f_bfree -
5164 (ext4_r_blocks_count(es) + resv_blocks);
5165 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5166 buf->f_bavail = 0;
5167 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5168 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5169 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5170 fsid = le64_to_cpup((void *)es->s_uuid) ^
5171 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5172 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5173 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5174
689c958c
LX		 5175#ifdef CONFIG_QUOTA
5176 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5177 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5178 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5179#endif
ac27a0ec
DK		 5180 return 0;
5181}
5182
0b8e58a1
AD		 5183/* Helper function for writing quotas on sync - we need to start transaction
5184 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5185 * Process 1 Process 2
617ba13b 5186 * ext4_create() quota_sync()
a269eb18 5187 * jbd2_journal_start() write_dquot()
871a2931 5188 * dquot_initialize() down(dqio_mutex)
dab291af 5189 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5190 *
5191 */
5192
5193#ifdef CONFIG_QUOTA
5194
5195static inline struct inode *dquot_to_inode(struct dquot *dquot)
5196{
4c376dca 5197 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5198}
5199
617ba13b 5200static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5201{
5202 int ret, err;
5203 handle_t *handle;
5204 struct inode *inode;
5205
5206 inode = dquot_to_inode(dquot);
9924a92a 5207 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5208 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5209 if (IS_ERR(handle))
5210 return PTR_ERR(handle);
5211 ret = dquot_commit(dquot);
617ba13b 5212 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5213 if (!ret)
5214 ret = err;
5215 return ret;
5216}
5217
617ba13b 5218static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5219{
5220 int ret, err;
5221 handle_t *handle;
5222
9924a92a 5223 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5224 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5225 if (IS_ERR(handle))
5226 return PTR_ERR(handle);
5227 ret = dquot_acquire(dquot);
617ba13b 5228 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5229 if (!ret)
5230 ret = err;
5231 return ret;
5232}
5233
617ba13b 5234static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5235{
5236 int ret, err;
5237 handle_t *handle;
5238
9924a92a 5239 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5240 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5241 if (IS_ERR(handle)) {
5242 /* Release dquot anyway to avoid endless cycle in dqput() */
5243 dquot_release(dquot);
ac27a0ec 5244 return PTR_ERR(handle);
9c3013e9 5245 }
ac27a0ec 5246 ret = dquot_release(dquot);
617ba13b 5247 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5248 if (!ret)
5249 ret = err;
5250 return ret;
5251}
5252
617ba13b 5253static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5254{
262b4662
JK		 5255 struct super_block *sb = dquot->dq_sb;
5256 struct ext4_sb_info *sbi = EXT4_SB(sb);
5257
2c8be6b2 5258 /* Are we journaling quotas? */
e2b911c5 5259 if (ext4_has_feature_quota(sb) ||
262b4662 5260 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5261 dquot_mark_dquot_dirty(dquot);
617ba13b 5262 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5263 } else {
5264 return dquot_mark_dquot_dirty(dquot);
5265 }
5266}
5267
617ba13b 5268static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5269{
5270 int ret, err;
5271 handle_t *handle;
5272
5273 /* Data block + inode block */
2b0143b5 5274 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5275 if (IS_ERR(handle))
5276 return PTR_ERR(handle);
5277 ret = dquot_commit_info(sb, type);
617ba13b 5278 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5279 if (!ret)
5280 ret = err;
5281 return ret;
5282}
5283
5284/*
5285 * Turn on quotas during mount time - we need to find
5286 * the quota file and such...
5287 */
617ba13b 5288static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5289{
287a8095
CH		 5290 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5291 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5292}
5293
daf647d2
TT		 5294static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5295{
5296 struct ext4_inode_info *ei = EXT4_I(inode);
5297
5298 /* The first argument of lockdep_set_subclass has to be
5299 * *exactly* the same as the argument to init_rwsem() --- in
5300 * this case, in init_once() --- or lockdep gets unhappy
5301 * because the name of the lock is set using the
5302 * stringification of the argument to init_rwsem().
5303 */
5304 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5305 lockdep_set_subclass(&ei->i_data_sem, subclass);
5306}
5307
ac27a0ec
DK		 5308/*
5309 * Standard function to be called on quota_on
5310 */
617ba13b 5311static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5312 const struct path *path)
ac27a0ec
DK		 5313{
5314 int err;
ac27a0ec
DK		 5315
5316 if (!test_opt(sb, QUOTA))
5317 return -EINVAL;
0623543b 5318
ac27a0ec 5319 /* Quotafile not on the same filesystem? */
d8c9584e 5320 if (path->dentry->d_sb != sb)
ac27a0ec 5321 return -EXDEV;
0623543b
JK		 5322 /* Journaling quota? */
5323 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5324 /* Quotafile not in fs root? */
f00c9e44 5325 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5326 ext4_msg(sb, KERN_WARNING,
5327 "Quota file not on filesystem root. "
5328 "Journaled quota will not work");
2b2d6d01 5329 }
0623543b
JK		 5330
5331 /*
5332 * When we journal data on quota file, we have to flush journal to see
5333 * all updates to the file when we bypass pagecache...
5334 */
0390131b 5335 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5336 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5337 /*
5338 * We don't need to lock updates but journal_flush() could
5339 * otherwise be livelocked...
5340 */
5341 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5342 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5343 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5344 if (err)
7ffe1ea8 5345 return err;
0623543b 5346 }
daf647d2
TT		 5347 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5348 err = dquot_quota_on(sb, type, format_id, path);
5349 if (err)
5350 lockdep_set_quota_inode(path->dentry->d_inode,
5351 I_DATA_SEM_NORMAL);
5352 return err;
ac27a0ec
DK		 5353}
5354
7c319d32
AK		 5355static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5356 unsigned int flags)
5357{
5358 int err;
5359 struct inode *qf_inode;
a2d4a646 5360 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5361 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5362 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5363 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5364 };
5365
e2b911c5 5366 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5367
5368 if (!qf_inums[type])
5369 return -EPERM;
5370
5371 qf_inode = ext4_iget(sb, qf_inums[type]);
5372 if (IS_ERR(qf_inode)) {
5373 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5374 return PTR_ERR(qf_inode);
5375 }
5376
bcb13850
JK		 5377 /* Don't account quota for quota files to avoid recursion */
5378 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5379 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5380 err = dquot_enable(qf_inode, type, format_id, flags);
5381 iput(qf_inode);
daf647d2
TT		 5382 if (err)
5383 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5384
5385 return err;
5386}
5387
5388/* Enable usage tracking for all quota types. */
5389static int ext4_enable_quotas(struct super_block *sb)
5390{
5391 int type, err = 0;
a2d4a646 5392 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5393 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5394 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5395 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5396 };
49da9392
JK		 5397 bool quota_mopt[EXT4_MAXQUOTAS] = {
5398 test_opt(sb, USRQUOTA),
5399 test_opt(sb, GRPQUOTA),
5400 test_opt(sb, PRJQUOTA),
5401 };
7c319d32
AK		 5402
5403 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5404 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5405 if (qf_inums[type]) {
5406 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5407 DQUOT_USAGE_ENABLED |
5408 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 5409 if (err) {
5410 ext4_warning(sb,
72ba7450
TT		 5411 "Failed to enable quota tracking "
5412 "(type=%d, err=%d). Please run "
5413 "e2fsck to fix.", type, err);
7c319d32
AK		 5414 return err;
5415 }
5416 }
5417 }
5418 return 0;
5419}
5420
ca0e05e4
DM		 5421static int ext4_quota_off(struct super_block *sb, int type)
5422{
21f97697
JK		 5423 struct inode *inode = sb_dqopt(sb)->files[type];
5424 handle_t *handle;
5425
87009d86
DM		 5426 /* Force all delayed allocation blocks to be allocated.
5427 * Caller already holds s_umount sem */
5428 if (test_opt(sb, DELALLOC))
ca0e05e4 5429 sync_filesystem(sb);
ca0e05e4 5430
0b268590
AG		 5431 if (!inode)
5432 goto out;
5433
21f97697
JK		 5434 /* Update modification times of quota files when userspace can
5435 * start looking at them */
9924a92a 5436 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5437 if (IS_ERR(handle))
5438 goto out;
eeca7ea1 5439 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5440 ext4_mark_inode_dirty(handle, inode);
5441 ext4_journal_stop(handle);
5442
5443out:
ca0e05e4
DM		 5444 return dquot_quota_off(sb, type);
5445}
5446
ac27a0ec
DK		 5447/* Read data from quotafile - avoid pagecache and such because we cannot afford
5448 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5449 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5450 * we don't have to be afraid of races */
617ba13b 5451static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5452 size_t len, loff_t off)
5453{
5454 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5455 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5456 int offset = off & (sb->s_blocksize - 1);
5457 int tocopy;
5458 size_t toread;
5459 struct buffer_head *bh;
5460 loff_t i_size = i_size_read(inode);
5461
5462 if (off > i_size)
5463 return 0;
5464 if (off+len > i_size)
5465 len = i_size-off;
5466 toread = len;
5467 while (toread > 0) {
5468 tocopy = sb->s_blocksize - offset < toread ?
5469 sb->s_blocksize - offset : toread;
1c215028
TT		 5470 bh = ext4_bread(NULL, inode, blk, 0);
5471 if (IS_ERR(bh))
5472 return PTR_ERR(bh);
ac27a0ec
DK		 5473 if (!bh) /* A hole? */
5474 memset(data, 0, tocopy);
5475 else
5476 memcpy(data, bh->b_data+offset, tocopy);
5477 brelse(bh);
5478 offset = 0;
5479 toread -= tocopy;
5480 data += tocopy;
5481 blk++;
5482 }
5483 return len;
5484}
5485
5486/* Write to quotafile (we know the transaction is already started and has
5487 * enough credits) */
617ba13b 5488static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5489 const char *data, size_t len, loff_t off)
5490{
5491 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5492 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5493 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5494 int retries = 0;
ac27a0ec
DK		 5495 struct buffer_head *bh;
5496 handle_t *handle = journal_current_handle();
5497
0390131b 5498 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5499 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5500 " cancelled because transaction is not started",
9c3013e9
JK		 5501 (unsigned long long)off, (unsigned long long)len);
5502 return -EIO;
5503 }
67eeb568
DM		 5504 /*
5505 * Since we account only one data block in transaction credits,
5506 * then it is impossible to cross a block boundary.
5507 */
5508 if (sb->s_blocksize - offset < len) {
5509 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5510 " cancelled because not block aligned",
5511 (unsigned long long)off, (unsigned long long)len);
5512 return -EIO;
5513 }
5514
c5e298ae
TT		 5515 do {
5516 bh = ext4_bread(handle, inode, blk,
5517 EXT4_GET_BLOCKS_CREATE |
5518 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5519 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5520 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5521 if (IS_ERR(bh))
5522 return PTR_ERR(bh);
67eeb568
DM		 5523 if (!bh)
5524 goto out;
5d601255 5525 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5526 err = ext4_journal_get_write_access(handle, bh);
5527 if (err) {
5528 brelse(bh);
1c215028 5529 return err;
ac27a0ec 5530 }
67eeb568
DM		 5531 lock_buffer(bh);
5532 memcpy(bh->b_data+offset, data, len);
5533 flush_dcache_page(bh->b_page);
5534 unlock_buffer(bh);
62d2b5f2 5535 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5536 brelse(bh);
ac27a0ec 5537out:
67eeb568
DM		 5538 if (inode->i_size < off + len) {
5539 i_size_write(inode, off + len);
617ba13b 5540 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5541 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5542 }
67eeb568 5543 return len;
ac27a0ec
DK		 5544}
5545
8f0e8746
TT		 5546static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5547{
5548 const struct quota_format_ops *ops;
5549
5550 if (!sb_has_quota_loaded(sb, qid->type))
5551 return -ESRCH;
5552 ops = sb_dqopt(sb)->ops[qid->type];
5553 if (!ops || !ops->get_next_id)
5554 return -ENOSYS;
5555 return dquot_get_next_id(sb, qid);
5556}
ac27a0ec
DK		 5557#endif
5558
152a0836
AV		 5559static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5560 const char *dev_name, void *data)
ac27a0ec 5561{
152a0836 5562 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5563}
5564
c290ea01 5565#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5566static inline void register_as_ext2(void)
5567{
5568 int err = register_filesystem(&ext2_fs_type);
5569 if (err)
5570 printk(KERN_WARNING
5571 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5572}
5573
5574static inline void unregister_as_ext2(void)
5575{
5576 unregister_filesystem(&ext2_fs_type);
5577}
2035e776
TT		 5578
5579static inline int ext2_feature_set_ok(struct super_block *sb)
5580{
e2b911c5 5581 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776
TT		 5582 return 0;
5583 if (sb->s_flags & MS_RDONLY)
5584 return 1;
e2b911c5 5585 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5586 return 0;
5587 return 1;
5588}
24b58424
TT		 5589#else
5590static inline void register_as_ext2(void) { }
5591static inline void unregister_as_ext2(void) { }
2035e776 5592static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5593#endif
5594
24b58424
TT		 5595static inline void register_as_ext3(void)
5596{
5597 int err = register_filesystem(&ext3_fs_type);
5598 if (err)
5599 printk(KERN_WARNING
5600 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5601}
5602
5603static inline void unregister_as_ext3(void)
5604{
5605 unregister_filesystem(&ext3_fs_type);
5606}
2035e776
TT		 5607
5608static inline int ext3_feature_set_ok(struct super_block *sb)
5609{
e2b911c5 5610 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5611 return 0;
e2b911c5 5612 if (!ext4_has_feature_journal(sb))
2035e776
TT		 5613 return 0;
5614 if (sb->s_flags & MS_RDONLY)
5615 return 1;
e2b911c5 5616 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5617 return 0;
5618 return 1;
5619}
24b58424 5620
03010a33
TT		 5621static struct file_system_type ext4_fs_type = {
5622 .owner = THIS_MODULE,
5623 .name = "ext4",
152a0836 5624 .mount = ext4_mount,
03010a33
TT		 5625 .kill_sb = kill_block_super,
5626 .fs_flags = FS_REQUIRES_DEV,
5627};
7f78e035 5628MODULE_ALIAS_FS("ext4");
03010a33 5629
e9e3bcec
ES		 5630/* Shared across all ext4 file systems */
5631wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5632
5dabfc78 5633static int __init ext4_init_fs(void)
ac27a0ec 5634{
e9e3bcec 5635 int i, err;
c9de560d 5636
e294a537 5637 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5638 ext4_li_info = NULL;
5639 mutex_init(&ext4_li_mtx);
5640
9a4c8019 5641 /* Build-time check for flags consistency */
12e9b892 5642 ext4_check_flag_values();
e9e3bcec 5643
e142d052 5644 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5645 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5646
51865fda 5647 err = ext4_init_es();
6fd058f7
TT		 5648 if (err)
5649 return err;
51865fda
ZL		 5650
5651 err = ext4_init_pageio();
5652 if (err)
b5799018 5653 goto out5;
51865fda 5654
5dabfc78 5655 err = ext4_init_system_zone();
bd2d0210 5656 if (err)
b5799018 5657 goto out4;
857ac889 5658
b5799018 5659 err = ext4_init_sysfs();
dd68314c 5660 if (err)
b5799018 5661 goto out3;
857ac889 5662
5dabfc78 5663 err = ext4_init_mballoc();
c9de560d
AT		 5664 if (err)
5665 goto out2;
ac27a0ec
DK		 5666 err = init_inodecache();
5667 if (err)
5668 goto out1;
24b58424 5669 register_as_ext3();
2035e776 5670 register_as_ext2();
03010a33 5671 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5672 if (err)
5673 goto out;
bfff6873 5674
ac27a0ec
DK		 5675 return 0;
5676out:
24b58424
TT		 5677 unregister_as_ext2();
5678 unregister_as_ext3();
ac27a0ec
DK		 5679 destroy_inodecache();
5680out1:
5dabfc78 5681 ext4_exit_mballoc();
9c191f70 5682out2:
b5799018
TT		 5683 ext4_exit_sysfs();
5684out3:
5dabfc78 5685 ext4_exit_system_zone();
b5799018 5686out4:
5dabfc78 5687 ext4_exit_pageio();
b5799018 5688out5:
51865fda
ZL		 5689 ext4_exit_es();
5690
ac27a0ec
DK		 5691 return err;
5692}
5693
5dabfc78 5694static void __exit ext4_exit_fs(void)
ac27a0ec 5695{
bfff6873 5696 ext4_destroy_lazyinit_thread();
24b58424
TT		 5697 unregister_as_ext2();
5698 unregister_as_ext3();
03010a33 5699 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5700 destroy_inodecache();
5dabfc78 5701 ext4_exit_mballoc();
b5799018 5702 ext4_exit_sysfs();
5dabfc78
TT		 5703 ext4_exit_system_zone();
5704 ext4_exit_pageio();
dd12ed14 5705 ext4_exit_es();
ac27a0ec
DK		 5706}
5707
5708MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5709MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5710MODULE_LICENSE("GPL");
5dabfc78
TT		 5711module_init(ext4_init_fs)
5712module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository