]> git.proxmox.com Git - mirror_ubuntu-focal-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-focal-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: do not polute the extents cache while shifting extents
[mirror_ubuntu-focal-kernel.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
66114cad 27#include <linux/backing-dev.h>
ac27a0ec 28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
3197ebdb 37#include <linux/ctype.h>
1330593e 38#include <linux/log2.h>
717d50e4 39#include <linux/crc16.h>
7abc52c2 40#include <linux/cleancache.h>
7c0f6ba6 41#include <linux/uaccess.h>
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
efbed4dc
TT		 441 if (ext4_error_ratelimit(sb)) {
442 va_start(args, fmt);
443 vaf.fmt = fmt;
444 vaf.va = &args;
445 printk(KERN_CRIT
446 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
447 sb->s_id, function, line, current->comm, &vaf);
448 va_end(args);
449 }
f3fc0210 450 save_error_info(sb, function, line);
617ba13b 451 ext4_handle_error(sb);
ac27a0ec
DK		 452}
453
e7c96e8e
JP		 454void __ext4_error_inode(struct inode *inode, const char *function,
455 unsigned int line, ext4_fsblk_t block,
456 const char *fmt, ...)
273df556
FM		 457{
458 va_list args;
f7c21177 459 struct va_format vaf;
1c13d5c0 460 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 461
1c13d5c0
TT		 462 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
463 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 464 if (ext4_error_ratelimit(inode->i_sb)) {
465 va_start(args, fmt);
466 vaf.fmt = fmt;
467 vaf.va = &args;
468 if (block)
469 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
470 "inode #%lu: block %llu: comm %s: %pV\n",
471 inode->i_sb->s_id, function, line, inode->i_ino,
472 block, current->comm, &vaf);
473 else
474 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
475 "inode #%lu: comm %s: %pV\n",
476 inode->i_sb->s_id, function, line, inode->i_ino,
477 current->comm, &vaf);
478 va_end(args);
479 }
1c13d5c0 480 save_error_info(inode->i_sb, function, line);
273df556
FM		 481 ext4_handle_error(inode->i_sb);
482}
483
e7c96e8e
JP		 484void __ext4_error_file(struct file *file, const char *function,
485 unsigned int line, ext4_fsblk_t block,
486 const char *fmt, ...)
273df556
FM		 487{
488 va_list args;
f7c21177 489 struct va_format vaf;
1c13d5c0 490 struct ext4_super_block *es;
496ad9aa 491 struct inode *inode = file_inode(file);
273df556
FM		 492 char pathname[80], *path;
493
1c13d5c0
TT		 494 es = EXT4_SB(inode->i_sb)->s_es;
495 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 496 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 497 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 498 if (IS_ERR(path))
499 path = "(unknown)";
500 va_start(args, fmt);
501 vaf.fmt = fmt;
502 vaf.va = &args;
503 if (block)
504 printk(KERN_CRIT
505 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
506 "block %llu: comm %s: path %s: %pV\n",
507 inode->i_sb->s_id, function, line, inode->i_ino,
508 block, current->comm, path, &vaf);
509 else
510 printk(KERN_CRIT
511 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
512 "comm %s: path %s: %pV\n",
513 inode->i_sb->s_id, function, line, inode->i_ino,
514 current->comm, path, &vaf);
515 va_end(args);
516 }
1c13d5c0 517 save_error_info(inode->i_sb, function, line);
273df556
FM		 518 ext4_handle_error(inode->i_sb);
519}
520
722887dd
TT		 521const char *ext4_decode_error(struct super_block *sb, int errno,
522 char nbuf[16])
ac27a0ec
DK		 523{
524 char *errstr = NULL;
525
526 switch (errno) {
6a797d27
DW		 527 case -EFSCORRUPTED:
528 errstr = "Corrupt filesystem";
529 break;
530 case -EFSBADCRC:
531 errstr = "Filesystem failed CRC";
532 break;
ac27a0ec
DK		 533 case -EIO:
534 errstr = "IO failure";
535 break;
536 case -ENOMEM:
537 errstr = "Out of memory";
538 break;
539 case -EROFS:
78f1ddbb
TT		 540 if (!sb || (EXT4_SB(sb)->s_journal &&
541 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 542 errstr = "Journal has aborted";
543 else
544 errstr = "Readonly filesystem";
545 break;
546 default:
547 /* If the caller passed in an extra buffer for unknown
548 * errors, textualise them now. Else we just return
549 * NULL. */
550 if (nbuf) {
551 /* Check for truncated error codes... */
552 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
553 errstr = nbuf;
554 }
555 break;
556 }
557
558 return errstr;
559}
560
617ba13b 561/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 562 * automatically and invokes the appropriate error response. */
563
c398eda0
TT		 564void __ext4_std_error(struct super_block *sb, const char *function,
565 unsigned int line, int errno)
ac27a0ec
DK		 566{
567 char nbuf[16];
568 const char *errstr;
569
570 /* Special case: if the error is EROFS, and we're not already
571 * inside a transaction, then there's really no point in logging
572 * an error. */
573 if (errno == -EROFS && journal_current_handle() == NULL &&
574 (sb->s_flags & MS_RDONLY))
575 return;
576
efbed4dc
TT		 577 if (ext4_error_ratelimit(sb)) {
578 errstr = ext4_decode_error(sb, errno, nbuf);
579 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
580 sb->s_id, function, line, errstr);
581 }
ac27a0ec 582
efbed4dc 583 save_error_info(sb, function, line);
617ba13b 584 ext4_handle_error(sb);
ac27a0ec
DK		 585}
586
587/*
617ba13b 588 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 589 * abort function may be used to deal with unrecoverable failures such
590 * as journal IO errors or ENOMEM at a critical moment in log management.
591 *
592 * We unconditionally force the filesystem into an ABORT|READONLY state,
593 * unless the error response on the fs has been set to panic in which
594 * case we take the easy way out and panic immediately.
595 */
596
c67d859e 597void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 598 unsigned int line, const char *fmt, ...)
ac27a0ec 599{
651e1c3b 600 struct va_format vaf;
ac27a0ec
DK		 601 va_list args;
602
1c13d5c0 603 save_error_info(sb, function, line);
ac27a0ec 604 va_start(args, fmt);
651e1c3b
JP		 605 vaf.fmt = fmt;
606 vaf.va = &args;
607 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
608 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 609 va_end(args);
610
1c13d5c0
TT		 611 if ((sb->s_flags & MS_RDONLY) == 0) {
612 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 613 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 614 /*
615 * Make sure updated value of ->s_mount_flags will be visible
616 * before ->s_flags update
617 */
618 smp_wmb();
619 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 620 if (EXT4_SB(sb)->s_journal)
621 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
622 save_error_info(sb, function, line);
623 }
4327ba52
DJ		 624 if (test_opt(sb, ERRORS_PANIC)) {
625 if (EXT4_SB(sb)->s_journal &&
626 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
627 return;
617ba13b 628 panic("EXT4-fs panic from previous error\n");
4327ba52 629 }
ac27a0ec
DK		 630}
631
e7c96e8e
JP		 632void __ext4_msg(struct super_block *sb,
633 const char *prefix, const char *fmt, ...)
b31e1552 634{
0ff2ea7d 635 struct va_format vaf;
b31e1552
ES		 636 va_list args;
637
efbed4dc
TT		 638 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
639 return;
640
b31e1552 641 va_start(args, fmt);
0ff2ea7d
JP		 642 vaf.fmt = fmt;
643 vaf.va = &args;
644 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 645 va_end(args);
646}
647
b03a2f7e
AD		 648#define ext4_warning_ratelimit(sb) \
649 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
650 "EXT4-fs warning")
651
12062ddd 652void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 653 unsigned int line, const char *fmt, ...)
ac27a0ec 654{
0ff2ea7d 655 struct va_format vaf;
ac27a0ec
DK		 656 va_list args;
657
b03a2f7e 658 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 659 return;
660
ac27a0ec 661 va_start(args, fmt);
0ff2ea7d
JP		 662 vaf.fmt = fmt;
663 vaf.va = &args;
664 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
665 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 666 va_end(args);
667}
668
b03a2f7e
AD		 669void __ext4_warning_inode(const struct inode *inode, const char *function,
670 unsigned int line, const char *fmt, ...)
671{
672 struct va_format vaf;
673 va_list args;
674
675 if (!ext4_warning_ratelimit(inode->i_sb))
676 return;
677
678 va_start(args, fmt);
679 vaf.fmt = fmt;
680 vaf.va = &args;
681 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
682 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
683 function, line, inode->i_ino, current->comm, &vaf);
684 va_end(args);
685}
686
e29136f8
TT		 687void __ext4_grp_locked_error(const char *function, unsigned int line,
688 struct super_block *sb, ext4_group_t grp,
689 unsigned long ino, ext4_fsblk_t block,
690 const char *fmt, ...)
5d1b1b3f
AK		 691__releases(bitlock)
692__acquires(bitlock)
693{
0ff2ea7d 694 struct va_format vaf;
5d1b1b3f
AK		 695 va_list args;
696 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
697
1c13d5c0
TT		 698 es->s_last_error_ino = cpu_to_le32(ino);
699 es->s_last_error_block = cpu_to_le64(block);
700 __save_error_info(sb, function, line);
0ff2ea7d 701
efbed4dc
TT		 702 if (ext4_error_ratelimit(sb)) {
703 va_start(args, fmt);
704 vaf.fmt = fmt;
705 vaf.va = &args;
706 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
707 sb->s_id, function, line, grp);
708 if (ino)
709 printk(KERN_CONT "inode %lu: ", ino);
710 if (block)
711 printk(KERN_CONT "block %llu:",
712 (unsigned long long) block);
713 printk(KERN_CONT "%pV\n", &vaf);
714 va_end(args);
715 }
5d1b1b3f
AK		 716
717 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 718 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 719 return;
720 }
1c13d5c0 721
5d1b1b3f
AK		 722 ext4_unlock_group(sb, grp);
723 ext4_handle_error(sb);
724 /*
725 * We only get here in the ERRORS_RO case; relocking the group
726 * may be dangerous, but nothing bad will happen since the
727 * filesystem will have already been marked read/only and the
728 * journal has been aborted. We return 1 as a hint to callers
729 * who might what to use the return value from
25985edc 730 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 731 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
732 * aggressively from the ext4 function in question, with a
733 * more appropriate error code.
734 */
735 ext4_lock_group(sb, grp);
736 return;
737}
738
617ba13b 739void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 740{
617ba13b 741 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 742
617ba13b 743 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 744 return;
745
12062ddd 746 ext4_warning(sb,
ac27a0ec
DK		 747 "updating to rev %d because of new feature flag, "
748 "running e2fsck is recommended",
617ba13b 749 EXT4_DYNAMIC_REV);
ac27a0ec 750
617ba13b
MC		 751 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
752 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
753 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 754 /* leave es->s_feature_*compat flags alone */
755 /* es->s_uuid will be set by e2fsck if empty */
756
757 /*
758 * The rest of the superblock fields should be zero, and if not it
759 * means they are likely already in use, so leave them alone. We
760 * can leave it up to e2fsck to clean up any inconsistencies there.
761 */
762}
763
764/*
765 * Open the external journal device
766 */
b31e1552 767static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 768{
769 struct block_device *bdev;
770 char b[BDEVNAME_SIZE];
771
d4d77629 772 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 773 if (IS_ERR(bdev))
774 goto fail;
775 return bdev;
776
777fail:
b31e1552 778 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 779 __bdevname(dev, b), PTR_ERR(bdev));
780 return NULL;
781}
782
783/*
784 * Release the journal device
785 */
4385bab1 786static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 787{
4385bab1 788 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 789}
790
4385bab1 791static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 792{
793 struct block_device *bdev;
ac27a0ec
DK		 794 bdev = sbi->journal_bdev;
795 if (bdev) {
4385bab1 796 ext4_blkdev_put(bdev);
ac27a0ec
DK		 797 sbi->journal_bdev = NULL;
798 }
ac27a0ec
DK		 799}
800
801static inline struct inode *orphan_list_entry(struct list_head *l)
802{
617ba13b 803 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 804}
805
617ba13b 806static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 807{
808 struct list_head *l;
809
b31e1552
ES		 810 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
811 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 812
813 printk(KERN_ERR "sb_info orphan list:\n");
814 list_for_each(l, &sbi->s_orphan) {
815 struct inode *inode = orphan_list_entry(l);
816 printk(KERN_ERR " "
817 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
818 inode->i_sb->s_id, inode->i_ino, inode,
819 inode->i_mode, inode->i_nlink,
820 NEXT_ORPHAN(inode));
821 }
822}
823
2b2d6d01 824static void ext4_put_super(struct super_block *sb)
ac27a0ec 825{
617ba13b
MC		 826 struct ext4_sb_info *sbi = EXT4_SB(sb);
827 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 828 int i, err;
ac27a0ec 829
857ac889 830 ext4_unregister_li_request(sb);
e0ccfd95
CH		 831 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
832
2e8fa54e 833 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 834 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 835
0390131b
FM		 836 if (sbi->s_journal) {
837 err = jbd2_journal_destroy(sbi->s_journal);
838 sbi->s_journal = NULL;
839 if (err < 0)
c67d859e 840 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 841 }
d4edac31 842
ebd173be 843 ext4_unregister_sysfs(sb);
d3922a77 844 ext4_es_unregister_shrinker(sbi);
9105bb14 845 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 846 ext4_release_system_zone(sb);
847 ext4_mb_release(sb);
848 ext4_ext_release(sb);
d4edac31 849
ac27a0ec 850 if (!(sb->s_flags & MS_RDONLY)) {
e2b911c5 851 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 852 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 853 }
58c5873a 854 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 855 ext4_commit_super(sb, 1);
856
ac27a0ec
DK		 857 for (i = 0; i < sbi->s_gdb_count; i++)
858 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 859 kvfree(sbi->s_group_desc);
860 kvfree(sbi->s_flex_groups);
57042651 861 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 862 percpu_counter_destroy(&sbi->s_freeinodes_counter);
863 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 864 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c8585c6f 865 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec 866#ifdef CONFIG_QUOTA
a2d4a646 867 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 868 kfree(sbi->s_qf_names[i]);
869#endif
870
871 /* Debugging code just in case the in-memory inode orphan list
872 * isn't empty. The on-disk one can be non-empty if we've
873 * detected an error and taken the fs readonly, but the
874 * in-memory list had better be clean by this point. */
875 if (!list_empty(&sbi->s_orphan))
876 dump_orphan_list(sb, sbi);
877 J_ASSERT(list_empty(&sbi->s_orphan));
878
89d96a6f 879 sync_blockdev(sb->s_bdev);
f98393a6 880 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 881 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
882 /*
883 * Invalidate the journal device's buffers. We don't want them
884 * floating about in memory - the physical journal device may
885 * hotswapped, and it breaks the `ro-after' testing code.
886 */
887 sync_blockdev(sbi->journal_bdev);
f98393a6 888 invalidate_bdev(sbi->journal_bdev);
617ba13b 889 ext4_blkdev_remove(sbi);
ac27a0ec 890 }
9c191f70
M		 891 if (sbi->s_mb_cache) {
892 ext4_xattr_destroy_cache(sbi->s_mb_cache);
893 sbi->s_mb_cache = NULL;
894 }
c5e06d10
JL		 895 if (sbi->s_mmp_tsk)
896 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 897 brelse(sbi->s_sbh);
ac27a0ec 898 sb->s_fs_info = NULL;
3197ebdb
TT		 899 /*
900 * Now that we are completely done shutting down the
901 * superblock, we need to actually destroy the kobject.
902 */
3197ebdb
TT		 903 kobject_put(&sbi->s_kobj);
904 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 905 if (sbi->s_chksum_driver)
906 crypto_free_shash(sbi->s_chksum_driver);
705895b6 907 kfree(sbi->s_blockgroup_lock);
ac27a0ec 908 kfree(sbi);
ac27a0ec
DK		 909}
910
e18b890b 911static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 912
913/*
914 * Called inside transaction, so use GFP_NOFS
915 */
617ba13b 916static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 917{
617ba13b 918 struct ext4_inode_info *ei;
ac27a0ec 919
e6b4f8da 920 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 921 if (!ei)
922 return NULL;
0b8e58a1 923
ac27a0ec 924 ei->vfs_inode.i_version = 1;
202ee5df 925 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 926 INIT_LIST_HEAD(&ei->i_prealloc_list);
927 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 928 ext4_es_init_tree(&ei->i_es_tree);
929 rwlock_init(&ei->i_es_lock);
edaa53ca 930 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 931 ei->i_es_all_nr = 0;
edaa53ca 932 ei->i_es_shk_nr = 0;
dd475925 933 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 934 ei->i_reserved_data_blocks = 0;
935 ei->i_reserved_meta_blocks = 0;
936 ei->i_allocated_meta_blocks = 0;
9d0be502 937 ei->i_da_metadata_calc_len = 0;
7e731bc9 938 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 939 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 940#ifdef CONFIG_QUOTA
941 ei->i_reserved_quota = 0;
96c7e0d9 942 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 943#endif
8aefcd55 944 ei->jinode = NULL;
2e8fa54e 945 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 946 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 947 ei->i_sync_tid = 0;
948 ei->i_datasync_tid = 0;
e27f41e1 949 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 950 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 951 return &ei->vfs_inode;
952}
953
7ff9c073
TT		 954static int ext4_drop_inode(struct inode *inode)
955{
956 int drop = generic_drop_inode(inode);
957
958 trace_ext4_drop_inode(inode, drop);
959 return drop;
960}
961
fa0d7e3d
NP		 962static void ext4_i_callback(struct rcu_head *head)
963{
964 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 965 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
966}
967
617ba13b 968static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 969{
9f7dd93d 970 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 971 ext4_msg(inode->i_sb, KERN_ERR,
972 "Inode %lu (%p): orphan list check failed!",
973 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 974 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
975 EXT4_I(inode), sizeof(struct ext4_inode_info),
976 true);
977 dump_stack();
978 }
fa0d7e3d 979 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 980}
981
51cc5068 982static void init_once(void *foo)
ac27a0ec 983{
617ba13b 984 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 985
a35afb83 986 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 987 init_rwsem(&ei->xattr_sem);
0e855ac8 988 init_rwsem(&ei->i_data_sem);
ea3d7209 989 init_rwsem(&ei->i_mmap_sem);
a35afb83 990 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 991}
992
e67bc2b3 993static int __init init_inodecache(void)
ac27a0ec 994{
617ba13b
MC		 995 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
996 sizeof(struct ext4_inode_info),
ac27a0ec 997 0, (SLAB_RECLAIM_ACCOUNT|
5d097056 998 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
20c2df83 999 init_once);
617ba13b 1000 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1001 return -ENOMEM;
1002 return 0;
1003}
1004
1005static void destroy_inodecache(void)
1006{
8c0a8537
KS		 1007 /*
1008 * Make sure all delayed rcu free inodes are flushed before we
1009 * destroy cache.
1010 */
1011 rcu_barrier();
617ba13b 1012 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1013}
1014
0930fcc1 1015void ext4_clear_inode(struct inode *inode)
ac27a0ec 1016{
0930fcc1 1017 invalidate_inode_buffers(inode);
dbd5768f 1018 clear_inode(inode);
9f754758 1019 dquot_drop(inode);
c2ea3fde 1020 ext4_discard_preallocations(inode);
51865fda 1021 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1022 if (EXT4_I(inode)->jinode) {
1023 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1024 EXT4_I(inode)->jinode);
1025 jbd2_free_inode(EXT4_I(inode)->jinode);
1026 EXT4_I(inode)->jinode = NULL;
1027 }
b7236e21 1028#ifdef CONFIG_EXT4_FS_ENCRYPTION
a7550b30 1029 fscrypt_put_encryption_info(inode, NULL);
b7236e21 1030#endif
ac27a0ec
DK		 1031}
1032
1b961ac0 1033static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1034 u64 ino, u32 generation)
ac27a0ec 1035{
ac27a0ec 1036 struct inode *inode;
ac27a0ec 1037
617ba13b 1038 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1039 return ERR_PTR(-ESTALE);
617ba13b 1040 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1041 return ERR_PTR(-ESTALE);
1042
1043 /* iget isn't really right if the inode is currently unallocated!!
1044 *
617ba13b 1045 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1046 * deleted, so we should be safe.
1047 *
1048 * Currently we don't know the generation for parent directory, so
1049 * a generation of 0 means "accept any"
1050 */
f4bb2981 1051 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1052 if (IS_ERR(inode))
1053 return ERR_CAST(inode);
1054 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1055 iput(inode);
1056 return ERR_PTR(-ESTALE);
1057 }
1b961ac0
CH		 1058
1059 return inode;
1060}
1061
1062static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1063 int fh_len, int fh_type)
1b961ac0
CH		 1064{
1065 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1066 ext4_nfs_get_inode);
1067}
1068
1069static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1070 int fh_len, int fh_type)
1b961ac0
CH		 1071{
1072 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1073 ext4_nfs_get_inode);
ac27a0ec
DK		 1074}
1075
c39a7f84
TO		 1076/*
1077 * Try to release metadata pages (indirect blocks, directories) which are
1078 * mapped via the block device. Since these pages could have journal heads
1079 * which would prevent try_to_free_buffers() from freeing them, we must use
1080 * jbd2 layer's try_to_free_buffers() function to release them.
1081 */
0b8e58a1
AD		 1082static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1083 gfp_t wait)
c39a7f84
TO		 1084{
1085 journal_t *journal = EXT4_SB(sb)->s_journal;
1086
1087 WARN_ON(PageChecked(page));
1088 if (!page_has_buffers(page))
1089 return 0;
1090 if (journal)
1091 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1092 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1093 return try_to_free_buffers(page);
1094}
1095
a7550b30
JK		 1096#ifdef CONFIG_EXT4_FS_ENCRYPTION
1097static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1098{
1099 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1100 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1101}
1102
a7550b30
JK		 1103static int ext4_prepare_context(struct inode *inode)
1104{
1105 return ext4_convert_inline_data(inode);
1106}
1107
1108static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1109 void *fs_data)
1110{
2f8f5e76
EB		 1111 handle_t *handle = fs_data;
1112 int res, res2, retries = 0;
1113
1114 /*
1115 * If a journal handle was specified, then the encryption context is
1116 * being set on a new inode via inheritance and is part of a larger
1117 * transaction to create the inode. Otherwise the encryption context is
1118 * being set on an existing inode in its own transaction. Only in the
1119 * latter case should the "retry on ENOSPC" logic be used.
1120 */
a7550b30 1121
2f8f5e76
EB		 1122 if (handle) {
1123 res = ext4_xattr_set_handle(handle, inode,
1124 EXT4_XATTR_INDEX_ENCRYPTION,
1125 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1126 ctx, len, 0);
a7550b30
JK		 1127 if (!res) {
1128 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1129 ext4_clear_inode_state(inode,
1130 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b
JK		 1131 /*
1132 * Update inode->i_flags - e.g. S_DAX may get disabled
1133 */
1134 ext4_set_inode_flags(inode);
a7550b30
JK		 1135 }
1136 return res;
1137 }
1138
2f8f5e76 1139retry:
a7550b30
JK		 1140 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1141 ext4_jbd2_credits_xattr(inode));
1142 if (IS_ERR(handle))
1143 return PTR_ERR(handle);
1144
2f8f5e76
EB		 1145 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1146 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1147 ctx, len, 0);
a7550b30
JK		 1148 if (!res) {
1149 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
a3caa24b
JK		 1150 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1151 ext4_set_inode_flags(inode);
a7550b30
JK		 1152 res = ext4_mark_inode_dirty(handle, inode);
1153 if (res)
1154 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1155 }
1156 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1157
1158 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1159 goto retry;
a7550b30
JK		 1160 if (!res)
1161 res = res2;
1162 return res;
1163}
1164
1165static int ext4_dummy_context(struct inode *inode)
1166{
1167 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1168}
1169
1170static unsigned ext4_max_namelen(struct inode *inode)
1171{
1172 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1173 EXT4_NAME_LEN;
1174}
1175
1176static struct fscrypt_operations ext4_cryptops = {
a5d431ef 1177 .key_prefix = "ext4:",
a7550b30 1178 .get_context = ext4_get_context,
a7550b30
JK		 1179 .prepare_context = ext4_prepare_context,
1180 .set_context = ext4_set_context,
1181 .dummy_context = ext4_dummy_context,
1182 .is_encrypted = ext4_encrypted_inode,
1183 .empty_dir = ext4_empty_dir,
1184 .max_namelen = ext4_max_namelen,
1185};
1186#else
1187static struct fscrypt_operations ext4_cryptops = {
1188 .is_encrypted = ext4_encrypted_inode,
1189};
1190#endif
1191
ac27a0ec 1192#ifdef CONFIG_QUOTA
689c958c
LX		 1193static char *quotatypes[] = INITQFNAMES;
1194#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1195
617ba13b
MC		 1196static int ext4_write_dquot(struct dquot *dquot);
1197static int ext4_acquire_dquot(struct dquot *dquot);
1198static int ext4_release_dquot(struct dquot *dquot);
1199static int ext4_mark_dquot_dirty(struct dquot *dquot);
1200static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1201static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1202 const struct path *path);
ca0e05e4 1203static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1204static int ext4_quota_on_mount(struct super_block *sb, int type);
1205static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1206 size_t len, loff_t off);
617ba13b 1207static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1208 const char *data, size_t len, loff_t off);
7c319d32
AK		 1209static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1210 unsigned int flags);
1211static int ext4_enable_quotas(struct super_block *sb);
8f0e8746 1212static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1213
96c7e0d9
JK		 1214static struct dquot **ext4_get_dquots(struct inode *inode)
1215{
1216 return EXT4_I(inode)->i_dquot;
1217}
1218
61e225dc 1219static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1220 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1221 .write_dquot = ext4_write_dquot,
1222 .acquire_dquot = ext4_acquire_dquot,
1223 .release_dquot = ext4_release_dquot,
1224 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1225 .write_info = ext4_write_info,
1226 .alloc_dquot = dquot_alloc,
1227 .destroy_dquot = dquot_destroy,
040cb378 1228 .get_projid = ext4_get_projid,
8f0e8746 1229 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1230};
1231
0d54b217 1232static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1233 .quota_on = ext4_quota_on,
ca0e05e4 1234 .quota_off = ext4_quota_off,
287a8095 1235 .quota_sync = dquot_quota_sync,
0a240339 1236 .get_state = dquot_get_state,
287a8095
CH		 1237 .set_info = dquot_set_dqinfo,
1238 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1239 .set_dqblk = dquot_set_dqblk,
1240 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1241};
1242#endif
1243
ee9b6d61 1244static const struct super_operations ext4_sops = {
617ba13b
MC		 1245 .alloc_inode = ext4_alloc_inode,
1246 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1247 .write_inode = ext4_write_inode,
1248 .dirty_inode = ext4_dirty_inode,
7ff9c073 1249 .drop_inode = ext4_drop_inode,
0930fcc1 1250 .evict_inode = ext4_evict_inode,
617ba13b 1251 .put_super = ext4_put_super,
617ba13b 1252 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1253 .freeze_fs = ext4_freeze,
1254 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1255 .statfs = ext4_statfs,
1256 .remount_fs = ext4_remount,
617ba13b 1257 .show_options = ext4_show_options,
ac27a0ec 1258#ifdef CONFIG_QUOTA
617ba13b
MC		 1259 .quota_read = ext4_quota_read,
1260 .quota_write = ext4_quota_write,
96c7e0d9 1261 .get_dquots = ext4_get_dquots,
ac27a0ec 1262#endif
c39a7f84 1263 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1264};
1265
39655164 1266static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1267 .fh_to_dentry = ext4_fh_to_dentry,
1268 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1269 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1270};
1271
1272enum {
1273 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1274 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1275 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1276 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1277 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1278 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1279 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1280 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1281 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1282 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1283 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1284 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1285 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1449032b 1286 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1287 Opt_lazytime, Opt_nolazytime,
1449032b 1288 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1289 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1290 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1291 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1292 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1293};
1294
a447c093 1295static const match_table_t tokens = {
ac27a0ec
DK		 1296 {Opt_bsd_df, "bsddf"},
1297 {Opt_minix_df, "minixdf"},
1298 {Opt_grpid, "grpid"},
1299 {Opt_grpid, "bsdgroups"},
1300 {Opt_nogrpid, "nogrpid"},
1301 {Opt_nogrpid, "sysvgroups"},
1302 {Opt_resgid, "resgid=%u"},
1303 {Opt_resuid, "resuid=%u"},
1304 {Opt_sb, "sb=%u"},
1305 {Opt_err_cont, "errors=continue"},
1306 {Opt_err_panic, "errors=panic"},
1307 {Opt_err_ro, "errors=remount-ro"},
1308 {Opt_nouid32, "nouid32"},
ac27a0ec 1309 {Opt_debug, "debug"},
72578c33
TT		 1310 {Opt_removed, "oldalloc"},
1311 {Opt_removed, "orlov"},
ac27a0ec
DK		 1312 {Opt_user_xattr, "user_xattr"},
1313 {Opt_nouser_xattr, "nouser_xattr"},
1314 {Opt_acl, "acl"},
1315 {Opt_noacl, "noacl"},
e3bb52ae 1316 {Opt_noload, "norecovery"},
5a916be1 1317 {Opt_noload, "noload"},
72578c33
TT		 1318 {Opt_removed, "nobh"},
1319 {Opt_removed, "bh"},
ac27a0ec 1320 {Opt_commit, "commit=%u"},
30773840
TT		 1321 {Opt_min_batch_time, "min_batch_time=%u"},
1322 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1323 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1324 {Opt_journal_path, "journal_path=%s"},
818d276c 1325 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1326 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1327 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1328 {Opt_abort, "abort"},
1329 {Opt_data_journal, "data=journal"},
1330 {Opt_data_ordered, "data=ordered"},
1331 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1332 {Opt_data_err_abort, "data_err=abort"},
1333 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1334 {Opt_offusrjquota, "usrjquota="},
1335 {Opt_usrjquota, "usrjquota=%s"},
1336 {Opt_offgrpjquota, "grpjquota="},
1337 {Opt_grpjquota, "grpjquota=%s"},
1338 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1339 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1340 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1341 {Opt_grpquota, "grpquota"},
1342 {Opt_noquota, "noquota"},
1343 {Opt_quota, "quota"},
1344 {Opt_usrquota, "usrquota"},
49da9392 1345 {Opt_prjquota, "prjquota"},
ac27a0ec 1346 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1347 {Opt_barrier, "barrier"},
1348 {Opt_nobarrier, "nobarrier"},
25ec56b5 1349 {Opt_i_version, "i_version"},
923ae0ff 1350 {Opt_dax, "dax"},
c9de560d 1351 {Opt_stripe, "stripe=%u"},
64769240 1352 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1353 {Opt_lazytime, "lazytime"},
1354 {Opt_nolazytime, "nolazytime"},
dd919b98 1355 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1356 {Opt_removed, "mblk_io_submit"},
1357 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1358 {Opt_block_validity, "block_validity"},
1359 {Opt_noblock_validity, "noblock_validity"},
240799cd 1360 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1361 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1362 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1363 {Opt_auto_da_alloc, "auto_da_alloc"},
1364 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1365 {Opt_dioread_nolock, "dioread_nolock"},
1366 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1367 {Opt_discard, "discard"},
1368 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1369 {Opt_init_itable, "init_itable=%u"},
1370 {Opt_init_itable, "init_itable"},
1371 {Opt_noinit_itable, "noinit_itable"},
df981d03 1372 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1373 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1374 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1375 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1376 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1377 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1378 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1379 {Opt_err, NULL},
ac27a0ec
DK		 1380};
1381
617ba13b 1382static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1383{
617ba13b 1384 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1385 char *options = (char *) *data;
1386
1387 if (!options || strncmp(options, "sb=", 3) != 0)
1388 return 1; /* Default location */
0b8e58a1 1389
ac27a0ec 1390 options += 3;
0b8e58a1 1391 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1392 sb_block = simple_strtoul(options, &options, 0);
1393 if (*options && *options != ',') {
4776004f 1394 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1395 (char *) *data);
1396 return 1;
1397 }
1398 if (*options == ',')
1399 options++;
1400 *data = (void *) options;
0b8e58a1 1401
ac27a0ec
DK		 1402 return sb_block;
1403}
1404
b3881f74 1405#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1406static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1407 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1408
56c50f11
DM		 1409#ifdef CONFIG_QUOTA
1410static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1411{
1412 struct ext4_sb_info *sbi = EXT4_SB(sb);
1413 char *qname;
03dafb5f 1414 int ret = -1;
56c50f11
DM		 1415
1416 if (sb_any_quota_loaded(sb) &&
1417 !sbi->s_qf_names[qtype]) {
1418 ext4_msg(sb, KERN_ERR,
1419 "Cannot change journaled "
1420 "quota options when quota turned on");
57f73c2c 1421 return -1;
56c50f11 1422 }
e2b911c5 1423 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1424 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1425 "ignored when QUOTA feature is enabled");
1426 return 1;
262b4662 1427 }
56c50f11
DM		 1428 qname = match_strdup(args);
1429 if (!qname) {
1430 ext4_msg(sb, KERN_ERR,
1431 "Not enough memory for storing quotafile name");
57f73c2c 1432 return -1;
56c50f11 1433 }
03dafb5f
CG		 1434 if (sbi->s_qf_names[qtype]) {
1435 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1436 ret = 1;
1437 else
1438 ext4_msg(sb, KERN_ERR,
1439 "%s quota file already specified",
1440 QTYPE2NAME(qtype));
1441 goto errout;
56c50f11 1442 }
03dafb5f 1443 if (strchr(qname, '/')) {
56c50f11
DM		 1444 ext4_msg(sb, KERN_ERR,
1445 "quotafile must be on filesystem root");
03dafb5f 1446 goto errout;
56c50f11 1447 }
03dafb5f 1448 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1449 set_opt(sb, QUOTA);
56c50f11 1450 return 1;
03dafb5f
CG		 1451errout:
1452 kfree(qname);
1453 return ret;
56c50f11
DM		 1454}
1455
1456static int clear_qf_name(struct super_block *sb, int qtype)
1457{
1458
1459 struct ext4_sb_info *sbi = EXT4_SB(sb);
1460
1461 if (sb_any_quota_loaded(sb) &&
1462 sbi->s_qf_names[qtype]) {
1463 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1464 " when quota turned on");
57f73c2c 1465 return -1;
56c50f11 1466 }
03dafb5f 1467 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1468 sbi->s_qf_names[qtype] = NULL;
1469 return 1;
1470}
1471#endif
1472
26092bf5
TT		 1473#define MOPT_SET 0x0001
1474#define MOPT_CLEAR 0x0002
1475#define MOPT_NOSUPPORT 0x0004
1476#define MOPT_EXPLICIT 0x0008
1477#define MOPT_CLEAR_ERR 0x0010
1478#define MOPT_GTE0 0x0020
ac27a0ec 1479#ifdef CONFIG_QUOTA
26092bf5
TT		 1480#define MOPT_Q 0
1481#define MOPT_QFMT 0x0040
1482#else
1483#define MOPT_Q MOPT_NOSUPPORT
1484#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1485#endif
26092bf5 1486#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1487#define MOPT_NO_EXT2 0x0100
1488#define MOPT_NO_EXT3 0x0200
1489#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1490#define MOPT_STRING 0x0400
26092bf5
TT		 1491
1492static const struct mount_opts {
1493 int token;
1494 int mount_opt;
1495 int flags;
1496} ext4_mount_opts[] = {
1497 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1498 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1499 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1500 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1501 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1502 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1503 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1504 MOPT_EXT4_ONLY | MOPT_SET},
1505 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1506 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1507 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1508 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1509 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1510 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1511 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1512 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1513 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1514 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1515 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1516 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1517 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1518 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1519 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1520 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1521 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1522 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1523 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1524 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1525 MOPT_NO_EXT2},
8dc0aa8c 1526 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1527 MOPT_NO_EXT2},
26092bf5
TT		 1528 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1529 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1530 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1531 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1532 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1533 {Opt_commit, 0, MOPT_GTE0},
1534 {Opt_max_batch_time, 0, MOPT_GTE0},
1535 {Opt_min_batch_time, 0, MOPT_GTE0},
1536 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1537 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1538 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1539 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1540 {Opt_resuid, 0, MOPT_GTE0},
1541 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1542 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1543 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1544 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1545 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1546 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1547 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1548 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1549 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1550 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1551#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1552 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1553 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1554#else
26092bf5
TT		 1555 {Opt_acl, 0, MOPT_NOSUPPORT},
1556 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1557#endif
26092bf5
TT		 1558 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1559 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1560 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1561 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1562 MOPT_SET | MOPT_Q},
1563 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1564 MOPT_SET | MOPT_Q},
49da9392
JK		 1565 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1566 MOPT_SET | MOPT_Q},
26092bf5 1567 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1568 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1569 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1570 {Opt_usrjquota, 0, MOPT_Q},
1571 {Opt_grpjquota, 0, MOPT_Q},
1572 {Opt_offusrjquota, 0, MOPT_Q},
1573 {Opt_offgrpjquota, 0, MOPT_Q},
1574 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1575 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1576 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1577 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1578 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1579 {Opt_err, 0, 0}
1580};
1581
1582static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1583 substring_t *args, unsigned long *journal_devnum,
1584 unsigned int *journal_ioprio, int is_remount)
1585{
1586 struct ext4_sb_info *sbi = EXT4_SB(sb);
1587 const struct mount_opts *m;
08cefc7a
EB		 1588 kuid_t uid;
1589 kgid_t gid;
26092bf5
TT		 1590 int arg = 0;
1591
57f73c2c
TT		 1592#ifdef CONFIG_QUOTA
1593 if (token == Opt_usrjquota)
1594 return set_qf_name(sb, USRQUOTA, &args[0]);
1595 else if (token == Opt_grpjquota)
1596 return set_qf_name(sb, GRPQUOTA, &args[0]);
1597 else if (token == Opt_offusrjquota)
1598 return clear_qf_name(sb, USRQUOTA);
1599 else if (token == Opt_offgrpjquota)
1600 return clear_qf_name(sb, GRPQUOTA);
1601#endif
26092bf5 1602 switch (token) {
f7048605
TT		 1603 case Opt_noacl:
1604 case Opt_nouser_xattr:
1605 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1606 break;
26092bf5
TT		 1607 case Opt_sb:
1608 return 1; /* handled by get_sb_block() */
1609 case Opt_removed:
5f3633e3 1610 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1611 return 1;
26092bf5
TT		 1612 case Opt_abort:
1613 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1614 return 1;
1615 case Opt_i_version:
1616 sb->s_flags |= MS_I_VERSION;
1617 return 1;
a26f4992
TT		 1618 case Opt_lazytime:
1619 sb->s_flags |= MS_LAZYTIME;
1620 return 1;
1621 case Opt_nolazytime:
1622 sb->s_flags &= ~MS_LAZYTIME;
1623 return 1;
26092bf5
TT		 1624 }
1625
5f3633e3
JK		 1626 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1627 if (token == m->token)
1628 break;
1629
1630 if (m->token == Opt_err) {
1631 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1632 "or missing value", opt);
1633 return -1;
1634 }
1635
8dc0aa8c
TT		 1636 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1637 ext4_msg(sb, KERN_ERR,
1638 "Mount option \"%s\" incompatible with ext2", opt);
1639 return -1;
1640 }
1641 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1642 ext4_msg(sb, KERN_ERR,
1643 "Mount option \"%s\" incompatible with ext3", opt);
1644 return -1;
1645 }
1646
ad4eec61 1647 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1648 return -1;
1649 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1650 return -1;
c93cf2d7
DM		 1651 if (m->flags & MOPT_EXPLICIT) {
1652 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1653 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1654 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1655 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1656 } else
1657 return -1;
1658 }
5f3633e3
JK		 1659 if (m->flags & MOPT_CLEAR_ERR)
1660 clear_opt(sb, ERRORS_MASK);
1661 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1662 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1663 "options when quota turned on");
1664 return -1;
1665 }
1666
1667 if (m->flags & MOPT_NOSUPPORT) {
1668 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1669 } else if (token == Opt_commit) {
1670 if (arg == 0)
1671 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1672 sbi->s_commit_interval = HZ * arg;
1673 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1674 sbi->s_max_batch_time = arg;
1675 } else if (token == Opt_min_batch_time) {
1676 sbi->s_min_batch_time = arg;
1677 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1678 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1679 ext4_msg(sb, KERN_ERR,
1680 "EXT4-fs: inode_readahead_blks must be "
1681 "0 or a power of 2 smaller than 2^31");
26092bf5 1682 return -1;
5f3633e3
JK		 1683 }
1684 sbi->s_inode_readahead_blks = arg;
1685 } else if (token == Opt_init_itable) {
1686 set_opt(sb, INIT_INODE_TABLE);
1687 if (!args->from)
1688 arg = EXT4_DEF_LI_WAIT_MULT;
1689 sbi->s_li_wait_mult = arg;
1690 } else if (token == Opt_max_dir_size_kb) {
1691 sbi->s_max_dir_size_kb = arg;
1692 } else if (token == Opt_stripe) {
1693 sbi->s_stripe = arg;
1694 } else if (token == Opt_resuid) {
1695 uid = make_kuid(current_user_ns(), arg);
1696 if (!uid_valid(uid)) {
1697 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1698 return -1;
1699 }
5f3633e3
JK		 1700 sbi->s_resuid = uid;
1701 } else if (token == Opt_resgid) {
1702 gid = make_kgid(current_user_ns(), arg);
1703 if (!gid_valid(gid)) {
1704 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1705 return -1;
1706 }
1707 sbi->s_resgid = gid;
1708 } else if (token == Opt_journal_dev) {
1709 if (is_remount) {
1710 ext4_msg(sb, KERN_ERR,
1711 "Cannot specify journal on remount");
1712 return -1;
1713 }
1714 *journal_devnum = arg;
ad4eec61
ES		 1715 } else if (token == Opt_journal_path) {
1716 char *journal_path;
1717 struct inode *journal_inode;
1718 struct path path;
1719 int error;
1720
1721 if (is_remount) {
1722 ext4_msg(sb, KERN_ERR,
1723 "Cannot specify journal on remount");
1724 return -1;
1725 }
1726 journal_path = match_strdup(&args[0]);
1727 if (!journal_path) {
1728 ext4_msg(sb, KERN_ERR, "error: could not dup "
1729 "journal device string");
1730 return -1;
1731 }
1732
1733 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1734 if (error) {
1735 ext4_msg(sb, KERN_ERR, "error: could not find "
1736 "journal device path: error %d", error);
1737 kfree(journal_path);
1738 return -1;
1739 }
1740
2b0143b5 1741 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1742 if (!S_ISBLK(journal_inode->i_mode)) {
1743 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1744 "is not a block device", journal_path);
1745 path_put(&path);
1746 kfree(journal_path);
1747 return -1;
1748 }
1749
1750 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1751 path_put(&path);
1752 kfree(journal_path);
5f3633e3
JK		 1753 } else if (token == Opt_journal_ioprio) {
1754 if (arg > 7) {
1755 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1756 " (must be 0-7)");
1757 return -1;
1758 }
1759 *journal_ioprio =
1760 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1761 } else if (token == Opt_test_dummy_encryption) {
1762#ifdef CONFIG_EXT4_FS_ENCRYPTION
1763 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1764 ext4_msg(sb, KERN_WARNING,
1765 "Test dummy encryption mode enabled");
1766#else
1767 ext4_msg(sb, KERN_WARNING,
1768 "Test dummy encryption mount option ignored");
1769#endif
5f3633e3
JK		 1770 } else if (m->flags & MOPT_DATAJ) {
1771 if (is_remount) {
1772 if (!sbi->s_journal)
1773 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1774 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1775 ext4_msg(sb, KERN_ERR,
26092bf5 1776 "Cannot change data mode on remount");
26092bf5 1777 return -1;
ac27a0ec 1778 }
26092bf5 1779 } else {
5f3633e3
JK		 1780 clear_opt(sb, DATA_FLAGS);
1781 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1782 }
5f3633e3
JK		 1783#ifdef CONFIG_QUOTA
1784 } else if (m->flags & MOPT_QFMT) {
1785 if (sb_any_quota_loaded(sb) &&
1786 sbi->s_jquota_fmt != m->mount_opt) {
1787 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1788 "quota options when quota turned on");
1789 return -1;
1790 }
e2b911c5 1791 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1792 ext4_msg(sb, KERN_INFO,
1793 "Quota format mount options ignored "
262b4662 1794 "when QUOTA feature is enabled");
c325a67c 1795 return 1;
262b4662 1796 }
5f3633e3 1797 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1798#endif
923ae0ff 1799 } else if (token == Opt_dax) {
ef83b6e8
DW		 1800#ifdef CONFIG_FS_DAX
1801 ext4_msg(sb, KERN_WARNING,
1802 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1803 sbi->s_mount_opt |= m->mount_opt;
1804#else
923ae0ff
RZ		 1805 ext4_msg(sb, KERN_INFO, "dax option not supported");
1806 return -1;
5f3633e3 1807#endif
7915a861
AN		 1808 } else if (token == Opt_data_err_abort) {
1809 sbi->s_mount_opt |= m->mount_opt;
1810 } else if (token == Opt_data_err_ignore) {
1811 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1812 } else {
1813 if (!args->from)
1814 arg = 1;
1815 if (m->flags & MOPT_CLEAR)
1816 arg = !arg;
1817 else if (unlikely(!(m->flags & MOPT_SET))) {
1818 ext4_msg(sb, KERN_WARNING,
1819 "buggy handling of option %s", opt);
1820 WARN_ON(1);
1821 return -1;
1822 }
1823 if (arg != 0)
1824 sbi->s_mount_opt |= m->mount_opt;
1825 else
1826 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1827 }
5f3633e3 1828 return 1;
26092bf5
TT		 1829}
1830
1831static int parse_options(char *options, struct super_block *sb,
1832 unsigned long *journal_devnum,
1833 unsigned int *journal_ioprio,
1834 int is_remount)
1835{
1836 struct ext4_sb_info *sbi = EXT4_SB(sb);
1837 char *p;
1838 substring_t args[MAX_OPT_ARGS];
1839 int token;
1840
1841 if (!options)
1842 return 1;
1843
1844 while ((p = strsep(&options, ",")) != NULL) {
1845 if (!*p)
1846 continue;
1847 /*
1848 * Initialize args struct so we know whether arg was
1849 * found; some options take optional arguments.
1850 */
caecd0af 1851 args[0].to = args[0].from = NULL;
26092bf5
TT		 1852 token = match_token(p, tokens, args);
1853 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1854 journal_ioprio, is_remount) < 0)
1855 return 0;
ac27a0ec
DK		 1856 }
1857#ifdef CONFIG_QUOTA
49da9392
JK		 1858 /*
1859 * We do the test below only for project quotas. 'usrquota' and
1860 * 'grpquota' mount options are allowed even without quota feature
1861 * to support legacy quotas in quota files.
1862 */
1863 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1864 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1865 "Cannot enable project quota enforcement.");
1866 return 0;
1867 }
1868 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1869 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1870 clear_opt(sb, USRQUOTA);
ac27a0ec 1871
482a7425 1872 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1873 clear_opt(sb, GRPQUOTA);
ac27a0ec 1874
56c50f11 1875 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1876 ext4_msg(sb, KERN_ERR, "old and new quota "
1877 "format mixing");
ac27a0ec
DK		 1878 return 0;
1879 }
1880
1881 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1882 ext4_msg(sb, KERN_ERR, "journaled quota format "
1883 "not specified");
ac27a0ec
DK		 1884 return 0;
1885 }
ac27a0ec
DK		 1886 }
1887#endif
261cb20c
JK		 1888 if (test_opt(sb, DIOREAD_NOLOCK)) {
1889 int blocksize =
1890 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1891
09cbfeaf 1892 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 1893 ext4_msg(sb, KERN_ERR, "can't mount with "
1894 "dioread_nolock if block size != PAGE_SIZE");
1895 return 0;
1896 }
1897 }
ac27a0ec
DK		 1898 return 1;
1899}
1900
2adf6da8
TT		 1901static inline void ext4_show_quota_options(struct seq_file *seq,
1902 struct super_block *sb)
1903{
1904#if defined(CONFIG_QUOTA)
1905 struct ext4_sb_info *sbi = EXT4_SB(sb);
1906
1907 if (sbi->s_jquota_fmt) {
1908 char *fmtname = "";
1909
1910 switch (sbi->s_jquota_fmt) {
1911 case QFMT_VFS_OLD:
1912 fmtname = "vfsold";
1913 break;
1914 case QFMT_VFS_V0:
1915 fmtname = "vfsv0";
1916 break;
1917 case QFMT_VFS_V1:
1918 fmtname = "vfsv1";
1919 break;
1920 }
1921 seq_printf(seq, ",jqfmt=%s", fmtname);
1922 }
1923
1924 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1925 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1926
1927 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1928 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1929#endif
1930}
1931
5a916be1
TT		 1932static const char *token2str(int token)
1933{
50df9fd5 1934 const struct match_token *t;
5a916be1
TT		 1935
1936 for (t = tokens; t->token != Opt_err; t++)
1937 if (t->token == token && !strchr(t->pattern, '='))
1938 break;
1939 return t->pattern;
1940}
1941
2adf6da8
TT		 1942/*
1943 * Show an option if
1944 * - it's set to a non-default value OR
1945 * - if the per-sb default is different from the global default
1946 */
66acdcf4
TT		 1947static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1948 int nodefs)
2adf6da8 1949{
2adf6da8
TT		 1950 struct ext4_sb_info *sbi = EXT4_SB(sb);
1951 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1952 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1953 const struct mount_opts *m;
66acdcf4 1954 char sep = nodefs ? '\n' : ',';
2adf6da8 1955
66acdcf4
TT		 1956#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1957#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1958
1959 if (sbi->s_sb_block != 1)
5a916be1
TT		 1960 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1961
1962 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1963 int want_set = m->flags & MOPT_SET;
1964 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1965 (m->flags & MOPT_CLEAR_ERR))
1966 continue;
66acdcf4 1967 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1968 continue; /* skip if same as the default */
1969 if ((want_set &&
1970 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1971 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1972 continue; /* select Opt_noFoo vs Opt_Foo */
1973 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1974 }
5a916be1 1975
08cefc7a 1976 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1977 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1978 SEQ_OPTS_PRINT("resuid=%u",
1979 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1980 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1981 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1982 SEQ_OPTS_PRINT("resgid=%u",
1983 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1984 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1985 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1986 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1987 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1988 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1989 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1990 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1991 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1992 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1993 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1994 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1995 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1996 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1997 if (sb->s_flags & MS_I_VERSION)
5a916be1 1998 SEQ_OPTS_PUTS("i_version");
66acdcf4 1999 if (nodefs || sbi->s_stripe)
5a916be1 2000 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 2001 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2002 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2003 SEQ_OPTS_PUTS("data=journal");
2004 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2005 SEQ_OPTS_PUTS("data=ordered");
2006 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2007 SEQ_OPTS_PUTS("data=writeback");
2008 }
66acdcf4
TT		 2009 if (nodefs ||
2010 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2011 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2012 sbi->s_inode_readahead_blks);
2adf6da8 2013
66acdcf4
TT		 2014 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2015 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2016 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2017 if (nodefs || sbi->s_max_dir_size_kb)
2018 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2019 if (test_opt(sb, DATA_ERR_ABORT))
2020 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 2021
2022 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2023 return 0;
2024}
2025
66acdcf4
TT		 2026static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2027{
2028 return _ext4_show_options(seq, root->d_sb, 0);
2029}
2030
ebd173be 2031int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2032{
2033 struct super_block *sb = seq->private;
2034 int rc;
2035
2036 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2037 rc = _ext4_show_options(seq, sb, 1);
2038 seq_puts(seq, "\n");
2039 return rc;
2040}
2041
617ba13b 2042static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2043 int read_only)
2044{
617ba13b 2045 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2046 int res = 0;
2047
617ba13b 2048 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2049 ext4_msg(sb, KERN_ERR, "revision level too high, "
2050 "forcing read-only mode");
ac27a0ec
DK		 2051 res = MS_RDONLY;
2052 }
2053 if (read_only)
281b5995 2054 goto done;
617ba13b 2055 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2056 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2057 "running e2fsck is recommended");
c8b459f4 2058 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2059 ext4_msg(sb, KERN_WARNING,
2060 "warning: mounting fs with errors, "
2061 "running e2fsck is recommended");
ed3ce80a 2062 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2063 le16_to_cpu(es->s_mnt_count) >=
2064 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2065 ext4_msg(sb, KERN_WARNING,
2066 "warning: maximal mount count reached, "
2067 "running e2fsck is recommended");
ac27a0ec
DK		 2068 else if (le32_to_cpu(es->s_checkinterval) &&
2069 (le32_to_cpu(es->s_lastcheck) +
2070 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 2071 ext4_msg(sb, KERN_WARNING,
2072 "warning: checktime reached, "
2073 "running e2fsck is recommended");
0b8e58a1 2074 if (!sbi->s_journal)
0390131b 2075 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2076 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2077 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2078 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 2079 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 2080 ext4_update_dynamic_rev(sb);
0390131b 2081 if (sbi->s_journal)
e2b911c5 2082 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2083
e2d67052 2084 ext4_commit_super(sb, 1);
281b5995 2085done:
ac27a0ec 2086 if (test_opt(sb, DEBUG))
a9df9a49 2087 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2088 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2089 sb->s_blocksize,
2090 sbi->s_groups_count,
617ba13b
MC		 2091 EXT4_BLOCKS_PER_GROUP(sb),
2092 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2093 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2094
7abc52c2 2095 cleancache_init_fs(sb);
ac27a0ec
DK		 2096 return res;
2097}
2098
117fff10
TT		 2099int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2100{
2101 struct ext4_sb_info *sbi = EXT4_SB(sb);
2102 struct flex_groups *new_groups;
2103 int size;
2104
2105 if (!sbi->s_log_groups_per_flex)
2106 return 0;
2107
2108 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2109 if (size <= sbi->s_flex_groups_allocated)
2110 return 0;
2111
2112 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2113 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2114 if (!new_groups) {
2115 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2116 size / (int) sizeof(struct flex_groups));
2117 return -ENOMEM;
2118 }
2119
2120 if (sbi->s_flex_groups) {
2121 memcpy(new_groups, sbi->s_flex_groups,
2122 (sbi->s_flex_groups_allocated *
2123 sizeof(struct flex_groups)));
b93b41d4 2124 kvfree(sbi->s_flex_groups);
117fff10
TT		 2125 }
2126 sbi->s_flex_groups = new_groups;
2127 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2128 return 0;
2129}
2130
772cb7c8
JS		 2131static int ext4_fill_flex_info(struct super_block *sb)
2132{
2133 struct ext4_sb_info *sbi = EXT4_SB(sb);
2134 struct ext4_group_desc *gdp = NULL;
772cb7c8 2135 ext4_group_t flex_group;
117fff10 2136 int i, err;
772cb7c8 2137
503358ae 2138 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2139 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2140 sbi->s_log_groups_per_flex = 0;
2141 return 1;
2142 }
2143
117fff10
TT		 2144 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2145 if (err)
9933fc0a 2146 goto failed;
772cb7c8 2147
772cb7c8 2148 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2149 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2150
2151 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2152 atomic_add(ext4_free_inodes_count(sb, gdp),
2153 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2154 atomic64_add(ext4_free_group_clusters(sb, gdp),
2155 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2156 atomic_add(ext4_used_dirs_count(sb, gdp),
2157 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2158 }
2159
2160 return 1;
2161failed:
2162 return 0;
2163}
2164
e2b911c5 2165static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2166 struct ext4_group_desc *gdp)
717d50e4 2167{
b47820ed 2168 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2169 __u16 crc = 0;
feb0ab32 2170 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2171 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2172
9aa5d32b 2173 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2174 /* Use new metadata_csum algorithm */
feb0ab32 2175 __u32 csum32;
b47820ed 2176 __u16 dummy_csum = 0;
feb0ab32 2177
feb0ab32
DW		 2178 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2179 sizeof(le_group));
b47820ed
DJ		 2180 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2181 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2182 sizeof(dummy_csum));
2183 offset += sizeof(dummy_csum);
2184 if (offset < sbi->s_desc_size)
2185 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2186 sbi->s_desc_size - offset);
feb0ab32
DW		 2187
2188 crc = csum32 & 0xFFFF;
2189 goto out;
717d50e4
AD		 2190 }
2191
feb0ab32 2192 /* old crc16 code */
e2b911c5 2193 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2194 return 0;
2195
feb0ab32
DW		 2196 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2197 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2198 crc = crc16(crc, (__u8 *)gdp, offset);
2199 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2200 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2201 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2202 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2203 crc = crc16(crc, (__u8 *)gdp + offset,
2204 le16_to_cpu(sbi->s_es->s_desc_size) -
2205 offset);
2206
2207out:
717d50e4
AD		 2208 return cpu_to_le16(crc);
2209}
2210
feb0ab32 2211int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2212 struct ext4_group_desc *gdp)
2213{
feb0ab32 2214 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2215 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2216 return 0;
2217
2218 return 1;
2219}
2220
feb0ab32
DW		 2221void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2222 struct ext4_group_desc *gdp)
2223{
2224 if (!ext4_has_group_desc_csum(sb))
2225 return;
e2b911c5 2226 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2227}
2228
ac27a0ec 2229/* Called at mount-time, super-block is locked */
bfff6873 2230static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2231 ext4_fsblk_t sb_block,
bfff6873 2232 ext4_group_t *first_not_zeroed)
ac27a0ec 2233{
617ba13b
MC		 2234 struct ext4_sb_info *sbi = EXT4_SB(sb);
2235 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2236 ext4_fsblk_t last_block;
bd81d8ee
LV		 2237 ext4_fsblk_t block_bitmap;
2238 ext4_fsblk_t inode_bitmap;
2239 ext4_fsblk_t inode_table;
ce421581 2240 int flexbg_flag = 0;
bfff6873 2241 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2242
e2b911c5 2243 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2244 flexbg_flag = 1;
2245
af5bc92d 2246 ext4_debug("Checking group descriptors");
ac27a0ec 2247
197cd65a
AM		 2248 for (i = 0; i < sbi->s_groups_count; i++) {
2249 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2250
ce421581 2251 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2252 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2253 else
2254 last_block = first_block +
617ba13b 2255 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2256
bfff6873
LC		 2257 if ((grp == sbi->s_groups_count) &&
2258 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2259 grp = i;
2260
8fadc143 2261 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2262 if (block_bitmap == sb_block) {
2263 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2264 "Block bitmap for group %u overlaps "
2265 "superblock", i);
2266 }
2b2d6d01 2267 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2268 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2269 "Block bitmap for group %u not in group "
b31e1552 2270 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2271 return 0;
2272 }
8fadc143 2273 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2274 if (inode_bitmap == sb_block) {
2275 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2276 "Inode bitmap for group %u overlaps "
2277 "superblock", i);
2278 }
2b2d6d01 2279 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2280 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2281 "Inode bitmap for group %u not in group "
b31e1552 2282 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2283 return 0;
2284 }
8fadc143 2285 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2286 if (inode_table == sb_block) {
2287 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2288 "Inode table for group %u overlaps "
2289 "superblock", i);
2290 }
bd81d8ee 2291 if (inode_table < first_block ||
2b2d6d01 2292 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2293 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2294 "Inode table for group %u not in group "
b31e1552 2295 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2296 return 0;
2297 }
955ce5f5 2298 ext4_lock_group(sb, i);
feb0ab32 2299 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2300 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2301 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2302 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2303 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2304 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2305 ext4_unlock_group(sb, i);
8a266467 2306 return 0;
7ee1ec4c 2307 }
717d50e4 2308 }
955ce5f5 2309 ext4_unlock_group(sb, i);
ce421581
JS		 2310 if (!flexbg_flag)
2311 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2312 }
bfff6873
LC		 2313 if (NULL != first_not_zeroed)
2314 *first_not_zeroed = grp;
ac27a0ec
DK		 2315 return 1;
2316}
2317
617ba13b 2318/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2319 * the superblock) which were deleted from all directories, but held open by
2320 * a process at the time of a crash. We walk the list and try to delete these
2321 * inodes at recovery time (only with a read-write filesystem).
2322 *
2323 * In order to keep the orphan inode chain consistent during traversal (in
2324 * case of crash during recovery), we link each inode into the superblock
2325 * orphan list_head and handle it the same way as an inode deletion during
2326 * normal operation (which journals the operations for us).
2327 *
2328 * We only do an iget() and an iput() on each inode, which is very safe if we
2329 * accidentally point at an in-use or already deleted inode. The worst that
2330 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2331 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2332 * e2fsck was run on this filesystem, and it must have already done the orphan
2333 * inode cleanup for us, so we can safely abort without any further action.
2334 */
2b2d6d01
TT		 2335static void ext4_orphan_cleanup(struct super_block *sb,
2336 struct ext4_super_block *es)
ac27a0ec
DK		 2337{
2338 unsigned int s_flags = sb->s_flags;
2c98eb5e 2339 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec
DK		 2340#ifdef CONFIG_QUOTA
2341 int i;
2342#endif
2343 if (!es->s_last_orphan) {
2344 jbd_debug(4, "no orphan inodes to clean up\n");
2345 return;
2346 }
2347
a8f48a95 2348 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2349 ext4_msg(sb, KERN_ERR, "write access "
2350 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2351 return;
2352 }
2353
d39195c3
AG		 2354 /* Check if feature set would not allow a r/w mount */
2355 if (!ext4_feature_set_ok(sb, 0)) {
2356 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2357 "unknown ROCOMPAT features");
2358 return;
2359 }
2360
617ba13b 2361 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2362 /* don't clear list on RO mount w/ errors */
2363 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2364 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2365 "clearing orphan list.\n");
c25f9bc6
ES		 2366 es->s_last_orphan = 0;
2367 }
ac27a0ec
DK		 2368 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2369 return;
2370 }
2371
2372 if (s_flags & MS_RDONLY) {
b31e1552 2373 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2374 sb->s_flags &= ~MS_RDONLY;
2375 }
2376#ifdef CONFIG_QUOTA
2377 /* Needed for iput() to work correctly and not trash data */
2378 sb->s_flags |= MS_ACTIVE;
2379 /* Turn on quotas so that they are updated correctly */
a2d4a646 2380 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2381 if (EXT4_SB(sb)->s_qf_names[i]) {
2382 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2383 if (ret < 0)
b31e1552
ES		 2384 ext4_msg(sb, KERN_ERR,
2385 "Cannot turn on journaled "
2386 "quota: error %d", ret);
ac27a0ec
DK		 2387 }
2388 }
2389#endif
2390
2391 while (es->s_last_orphan) {
2392 struct inode *inode;
2393
c65d5c6c
VN		 2394 /*
2395 * We may have encountered an error during cleanup; if
2396 * so, skip the rest.
2397 */
2398 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2399 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2400 es->s_last_orphan = 0;
2401 break;
2402 }
2403
97bd42b9
JB		 2404 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2405 if (IS_ERR(inode)) {
ac27a0ec
DK		 2406 es->s_last_orphan = 0;
2407 break;
2408 }
2409
617ba13b 2410 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2411 dquot_initialize(inode);
ac27a0ec 2412 if (inode->i_nlink) {
566370a2
PT		 2413 if (test_opt(sb, DEBUG))
2414 ext4_msg(sb, KERN_DEBUG,
2415 "%s: truncating inode %lu to %lld bytes",
2416 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2417 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2418 inode->i_ino, inode->i_size);
5955102c 2419 inode_lock(inode);
55f252c9 2420 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2421 ret = ext4_truncate(inode);
2422 if (ret)
2423 ext4_std_error(inode->i_sb, ret);
5955102c 2424 inode_unlock(inode);
ac27a0ec
DK		 2425 nr_truncates++;
2426 } else {
566370a2
PT		 2427 if (test_opt(sb, DEBUG))
2428 ext4_msg(sb, KERN_DEBUG,
2429 "%s: deleting unreferenced inode %lu",
2430 __func__, inode->i_ino);
ac27a0ec
DK		 2431 jbd_debug(2, "deleting unreferenced inode %lu\n",
2432 inode->i_ino);
2433 nr_orphans++;
2434 }
2435 iput(inode); /* The delete magic happens here! */
2436 }
2437
2b2d6d01 2438#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2439
2440 if (nr_orphans)
b31e1552
ES		 2441 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2442 PLURAL(nr_orphans));
ac27a0ec 2443 if (nr_truncates)
b31e1552
ES		 2444 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2445 PLURAL(nr_truncates));
ac27a0ec
DK		 2446#ifdef CONFIG_QUOTA
2447 /* Turn quotas off */
a2d4a646 2448 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2449 if (sb_dqopt(sb)->files[i])
287a8095 2450 dquot_quota_off(sb, i);
ac27a0ec
DK		 2451 }
2452#endif
2453 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2454}
0b8e58a1 2455
cd2291a4
ES		 2456/*
2457 * Maximal extent format file size.
2458 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2459 * extent format containers, within a sector_t, and within i_blocks
2460 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2461 * so that won't be a limiting factor.
2462 *
f17722f9
LC		 2463 * However there is other limiting factor. We do store extents in the form
2464 * of starting block and length, hence the resulting length of the extent
2465 * covering maximum file size must fit into on-disk format containers as
2466 * well. Given that length is always by 1 unit bigger than max unit (because
2467 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2468 *
cd2291a4
ES		 2469 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2470 */
f287a1a5 2471static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2472{
2473 loff_t res;
2474 loff_t upper_limit = MAX_LFS_FILESIZE;
2475
2476 /* small i_blocks in vfs inode? */
f287a1a5 2477 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2478 /*
90c699a9 2479 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2480 * i_block represent total blocks in 512 bytes
2481 * 32 == size of vfs inode i_blocks * 8
2482 */
2483 upper_limit = (1LL << 32) - 1;
2484
2485 /* total blocks in file system block size */
2486 upper_limit >>= (blkbits - 9);
2487 upper_limit <<= blkbits;
2488 }
2489
f17722f9
LC		 2490 /*
2491 * 32-bit extent-start container, ee_block. We lower the maxbytes
2492 * by one fs block, so ee_len can cover the extent of maximum file
2493 * size
2494 */
2495 res = (1LL << 32) - 1;
cd2291a4 2496 res <<= blkbits;
cd2291a4
ES		 2497
2498 /* Sanity check against vm- & vfs- imposed limits */
2499 if (res > upper_limit)
2500 res = upper_limit;
2501
2502 return res;
2503}
ac27a0ec 2504
ac27a0ec 2505/*
cd2291a4 2506 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2507 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2508 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2509 */
f287a1a5 2510static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2511{
617ba13b 2512 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2513 int meta_blocks;
2514 loff_t upper_limit;
0b8e58a1
AD		 2515 /* This is calculated to be the largest file size for a dense, block
2516 * mapped file such that the file's total number of 512-byte sectors,
2517 * including data and all indirect blocks, does not exceed (2^48 - 1).
2518 *
2519 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2520 * number of 512-byte sectors of the file.
0fc1b451
AK		 2521 */
2522
f287a1a5 2523 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2524 /*
90c699a9 2525 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2526 * the inode i_block field represents total file blocks in
2527 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2528 */
2529 upper_limit = (1LL << 32) - 1;
2530
2531 /* total blocks in file system block size */
2532 upper_limit >>= (bits - 9);
2533
2534 } else {
8180a562
AK		 2535 /*
2536 * We use 48 bit ext4_inode i_blocks
2537 * With EXT4_HUGE_FILE_FL set the i_blocks
2538 * represent total number of blocks in
2539 * file system block size
2540 */
0fc1b451
AK		 2541 upper_limit = (1LL << 48) - 1;
2542
0fc1b451
AK		 2543 }
2544
2545 /* indirect blocks */
2546 meta_blocks = 1;
2547 /* double indirect blocks */
2548 meta_blocks += 1 + (1LL << (bits-2));
2549 /* tripple indirect blocks */
2550 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2551
2552 upper_limit -= meta_blocks;
2553 upper_limit <<= bits;
ac27a0ec
DK		 2554
2555 res += 1LL << (bits-2);
2556 res += 1LL << (2*(bits-2));
2557 res += 1LL << (3*(bits-2));
2558 res <<= bits;
2559 if (res > upper_limit)
2560 res = upper_limit;
0fc1b451
AK		 2561
2562 if (res > MAX_LFS_FILESIZE)
2563 res = MAX_LFS_FILESIZE;
2564
ac27a0ec
DK		 2565 return res;
2566}
2567
617ba13b 2568static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2569 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2570{
617ba13b 2571 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2572 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2573 int has_super = 0;
2574
2575 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2576
e2b911c5 2577 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2578 return logical_sb_block + nr + 1;
ac27a0ec 2579 bg = sbi->s_desc_per_block * nr;
617ba13b 2580 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2581 has_super = 1;
0b8e58a1 2582
bd63f6b0
DW		 2583 /*
2584 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2585 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2586 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2587 * compensate.
2588 */
2589 if (sb->s_blocksize == 1024 && nr == 0 &&
2590 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2591 has_super++;
2592
617ba13b 2593 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2594}
2595
c9de560d
AT		 2596/**
2597 * ext4_get_stripe_size: Get the stripe size.
2598 * @sbi: In memory super block info
2599 *
2600 * If we have specified it via mount option, then
2601 * use the mount option value. If the value specified at mount time is
2602 * greater than the blocks per group use the super block value.
2603 * If the super block value is greater than blocks per group return 0.
2604 * Allocator needs it be less than blocks per group.
2605 *
2606 */
2607static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2608{
2609 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2610 unsigned long stripe_width =
2611 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2612 int ret;
c9de560d
AT		 2613
2614 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2615 ret = sbi->s_stripe;
2616 else if (stripe_width <= sbi->s_blocks_per_group)
2617 ret = stripe_width;
2618 else if (stride <= sbi->s_blocks_per_group)
2619 ret = stride;
2620 else
2621 ret = 0;
c9de560d 2622
3eb08658
DE		 2623 /*
2624 * If the stripe width is 1, this makes no sense and
2625 * we set it to 0 to turn off stripe handling code.
2626 */
2627 if (ret <= 1)
2628 ret = 0;
c9de560d 2629
3eb08658 2630 return ret;
c9de560d 2631}
ac27a0ec 2632
a13fb1a4
ES		 2633/*
2634 * Check whether this filesystem can be mounted based on
2635 * the features present and the RDONLY/RDWR mount requested.
2636 * Returns 1 if this filesystem can be mounted as requested,
2637 * 0 if it cannot be.
2638 */
2639static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2640{
e2b911c5 2641 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2642 ext4_msg(sb, KERN_ERR,
2643 "Couldn't mount because of "
2644 "unsupported optional features (%x)",
2645 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2646 ~EXT4_FEATURE_INCOMPAT_SUPP));
2647 return 0;
2648 }
2649
2650 if (readonly)
2651 return 1;
2652
e2b911c5 2653 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b
DW		 2654 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2655 sb->s_flags |= MS_RDONLY;
2656 return 1;
2657 }
2658
a13fb1a4 2659 /* Check that feature set is OK for a read-write mount */
e2b911c5 2660 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2661 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2662 "unsupported optional features (%x)",
2663 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2664 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2665 return 0;
2666 }
2667 /*
2668 * Large file size enabled file system can only be mounted
2669 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2670 */
e2b911c5 2671 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2672 if (sizeof(blkcnt_t) < sizeof(u64)) {
2673 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2674 "cannot be mounted RDWR without "
2675 "CONFIG_LBDAF");
2676 return 0;
2677 }
2678 }
e2b911c5 2679 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2680 ext4_msg(sb, KERN_ERR,
2681 "Can't support bigalloc feature without "
2682 "extents feature\n");
2683 return 0;
2684 }
7c319d32
AK		 2685
2686#ifndef CONFIG_QUOTA
e2b911c5 2687 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2688 ext4_msg(sb, KERN_ERR,
2689 "Filesystem with quota feature cannot be mounted RDWR "
2690 "without CONFIG_QUOTA");
2691 return 0;
2692 }
689c958c
LX		 2693 if (ext4_has_feature_project(sb) && !readonly) {
2694 ext4_msg(sb, KERN_ERR,
2695 "Filesystem with project quota feature cannot be mounted RDWR "
2696 "without CONFIG_QUOTA");
2697 return 0;
2698 }
7c319d32 2699#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2700 return 1;
2701}
2702
66e61a9e
TT		 2703/*
2704 * This function is called once a day if we have errors logged
2705 * on the file system
2706 */
2707static void print_daily_error_info(unsigned long arg)
2708{
2709 struct super_block *sb = (struct super_block *) arg;
2710 struct ext4_sb_info *sbi;
2711 struct ext4_super_block *es;
2712
2713 sbi = EXT4_SB(sb);
2714 es = sbi->s_es;
2715
2716 if (es->s_error_count)
ae0f78de
TT		 2717 /* fsck newer than v1.41.13 is needed to clean this condition. */
2718 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2719 le32_to_cpu(es->s_error_count));
2720 if (es->s_first_error_time) {
ae0f78de 2721 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2722 sb->s_id, le32_to_cpu(es->s_first_error_time),
2723 (int) sizeof(es->s_first_error_func),
2724 es->s_first_error_func,
2725 le32_to_cpu(es->s_first_error_line));
2726 if (es->s_first_error_ino)
651e1c3b 2727 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2728 le32_to_cpu(es->s_first_error_ino));
2729 if (es->s_first_error_block)
651e1c3b 2730 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2731 le64_to_cpu(es->s_first_error_block));
651e1c3b 2732 printk(KERN_CONT "\n");
66e61a9e
TT		 2733 }
2734 if (es->s_last_error_time) {
ae0f78de 2735 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2736 sb->s_id, le32_to_cpu(es->s_last_error_time),
2737 (int) sizeof(es->s_last_error_func),
2738 es->s_last_error_func,
2739 le32_to_cpu(es->s_last_error_line));
2740 if (es->s_last_error_ino)
651e1c3b 2741 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2742 le32_to_cpu(es->s_last_error_ino));
2743 if (es->s_last_error_block)
651e1c3b 2744 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2745 le64_to_cpu(es->s_last_error_block));
651e1c3b 2746 printk(KERN_CONT "\n");
66e61a9e
TT		 2747 }
2748 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2749}
2750
bfff6873
LC		 2751/* Find next suitable group and run ext4_init_inode_table */
2752static int ext4_run_li_request(struct ext4_li_request *elr)
2753{
2754 struct ext4_group_desc *gdp = NULL;
2755 ext4_group_t group, ngroups;
2756 struct super_block *sb;
2757 unsigned long timeout = 0;
2758 int ret = 0;
2759
2760 sb = elr->lr_super;
2761 ngroups = EXT4_SB(sb)->s_groups_count;
2762
2763 for (group = elr->lr_next_group; group < ngroups; group++) {
2764 gdp = ext4_get_group_desc(sb, group, NULL);
2765 if (!gdp) {
2766 ret = 1;
2767 break;
2768 }
2769
2770 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2771 break;
2772 }
2773
7f511862 2774 if (group >= ngroups)
bfff6873
LC		 2775 ret = 1;
2776
2777 if (!ret) {
2778 timeout = jiffies;
2779 ret = ext4_init_inode_table(sb, group,
2780 elr->lr_timeout ? 0 : 1);
2781 if (elr->lr_timeout == 0) {
51ce6511
LC		 2782 timeout = (jiffies - timeout) *
2783 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2784 elr->lr_timeout = timeout;
2785 }
2786 elr->lr_next_sched = jiffies + elr->lr_timeout;
2787 elr->lr_next_group = group + 1;
2788 }
bfff6873
LC		 2789 return ret;
2790}
2791
2792/*
2793 * Remove lr_request from the list_request and free the
4ed5c033 2794 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2795 */
2796static void ext4_remove_li_request(struct ext4_li_request *elr)
2797{
2798 struct ext4_sb_info *sbi;
2799
2800 if (!elr)
2801 return;
2802
2803 sbi = elr->lr_sbi;
2804
2805 list_del(&elr->lr_request);
2806 sbi->s_li_request = NULL;
2807 kfree(elr);
2808}
2809
2810static void ext4_unregister_li_request(struct super_block *sb)
2811{
1bb933fb
LC		 2812 mutex_lock(&ext4_li_mtx);
2813 if (!ext4_li_info) {
2814 mutex_unlock(&ext4_li_mtx);
bfff6873 2815 return;
1bb933fb 2816 }
bfff6873
LC		 2817
2818 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2819 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2820 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2821 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2822}
2823
8f1f7453
ES		 2824static struct task_struct *ext4_lazyinit_task;
2825
bfff6873
LC		 2826/*
2827 * This is the function where ext4lazyinit thread lives. It walks
2828 * through the request list searching for next scheduled filesystem.
2829 * When such a fs is found, run the lazy initialization request
2830 * (ext4_rn_li_request) and keep track of the time spend in this
2831 * function. Based on that time we compute next schedule time of
2832 * the request. When walking through the list is complete, compute
2833 * next waking time and put itself into sleep.
2834 */
2835static int ext4_lazyinit_thread(void *arg)
2836{
2837 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2838 struct list_head *pos, *n;
2839 struct ext4_li_request *elr;
4ed5c033 2840 unsigned long next_wakeup, cur;
bfff6873
LC		 2841
2842 BUG_ON(NULL == eli);
2843
bfff6873
LC		 2844cont_thread:
2845 while (true) {
2846 next_wakeup = MAX_JIFFY_OFFSET;
2847
2848 mutex_lock(&eli->li_list_mtx);
2849 if (list_empty(&eli->li_request_list)) {
2850 mutex_unlock(&eli->li_list_mtx);
2851 goto exit_thread;
2852 }
bfff6873 2853 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 2854 int err = 0;
2855 int progress = 0;
bfff6873
LC		 2856 elr = list_entry(pos, struct ext4_li_request,
2857 lr_request);
2858
e22834f0
DM		 2859 if (time_before(jiffies, elr->lr_next_sched)) {
2860 if (time_before(elr->lr_next_sched, next_wakeup))
2861 next_wakeup = elr->lr_next_sched;
2862 continue;
2863 }
2864 if (down_read_trylock(&elr->lr_super->s_umount)) {
2865 if (sb_start_write_trylock(elr->lr_super)) {
2866 progress = 1;
2867 /*
2868 * We hold sb->s_umount, sb can not
2869 * be removed from the list, it is
2870 * now safe to drop li_list_mtx
2871 */
2872 mutex_unlock(&eli->li_list_mtx);
2873 err = ext4_run_li_request(elr);
2874 sb_end_write(elr->lr_super);
2875 mutex_lock(&eli->li_list_mtx);
2876 n = pos->next;
b2c78cd0 2877 }
e22834f0
DM		 2878 up_read((&elr->lr_super->s_umount));
2879 }
2880 /* error, remove the lazy_init job */
2881 if (err) {
2882 ext4_remove_li_request(elr);
2883 continue;
2884 }
2885 if (!progress) {
2886 elr->lr_next_sched = jiffies +
2887 (prandom_u32()
2888 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 2889 }
bfff6873
LC		 2890 if (time_before(elr->lr_next_sched, next_wakeup))
2891 next_wakeup = elr->lr_next_sched;
2892 }
2893 mutex_unlock(&eli->li_list_mtx);
2894
a0acae0e 2895 try_to_freeze();
bfff6873 2896
4ed5c033
LC		 2897 cur = jiffies;
2898 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2899 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2900 cond_resched();
2901 continue;
2902 }
2903
4ed5c033
LC		 2904 schedule_timeout_interruptible(next_wakeup - cur);
2905
8f1f7453
ES		 2906 if (kthread_should_stop()) {
2907 ext4_clear_request_list();
2908 goto exit_thread;
2909 }
bfff6873
LC		 2910 }
2911
2912exit_thread:
2913 /*
2914 * It looks like the request list is empty, but we need
2915 * to check it under the li_list_mtx lock, to prevent any
2916 * additions into it, and of course we should lock ext4_li_mtx
2917 * to atomically free the list and ext4_li_info, because at
2918 * this point another ext4 filesystem could be registering
2919 * new one.
2920 */
2921 mutex_lock(&ext4_li_mtx);
2922 mutex_lock(&eli->li_list_mtx);
2923 if (!list_empty(&eli->li_request_list)) {
2924 mutex_unlock(&eli->li_list_mtx);
2925 mutex_unlock(&ext4_li_mtx);
2926 goto cont_thread;
2927 }
2928 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2929 kfree(ext4_li_info);
2930 ext4_li_info = NULL;
2931 mutex_unlock(&ext4_li_mtx);
2932
2933 return 0;
2934}
2935
2936static void ext4_clear_request_list(void)
2937{
2938 struct list_head *pos, *n;
2939 struct ext4_li_request *elr;
2940
2941 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2942 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2943 elr = list_entry(pos, struct ext4_li_request,
2944 lr_request);
2945 ext4_remove_li_request(elr);
2946 }
2947 mutex_unlock(&ext4_li_info->li_list_mtx);
2948}
2949
2950static int ext4_run_lazyinit_thread(void)
2951{
8f1f7453
ES		 2952 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2953 ext4_li_info, "ext4lazyinit");
2954 if (IS_ERR(ext4_lazyinit_task)) {
2955 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2956 ext4_clear_request_list();
bfff6873
LC		 2957 kfree(ext4_li_info);
2958 ext4_li_info = NULL;
92b97816 2959 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2960 "initialization thread\n",
2961 err);
2962 return err;
2963 }
2964 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2965 return 0;
2966}
2967
2968/*
2969 * Check whether it make sense to run itable init. thread or not.
2970 * If there is at least one uninitialized inode table, return
2971 * corresponding group number, else the loop goes through all
2972 * groups and return total number of groups.
2973 */
2974static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2975{
2976 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2977 struct ext4_group_desc *gdp = NULL;
2978
2979 for (group = 0; group < ngroups; group++) {
2980 gdp = ext4_get_group_desc(sb, group, NULL);
2981 if (!gdp)
2982 continue;
2983
2984 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2985 break;
2986 }
2987
2988 return group;
2989}
2990
2991static int ext4_li_info_new(void)
2992{
2993 struct ext4_lazy_init *eli = NULL;
2994
2995 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2996 if (!eli)
2997 return -ENOMEM;
2998
bfff6873
LC		 2999 INIT_LIST_HEAD(&eli->li_request_list);
3000 mutex_init(&eli->li_list_mtx);
3001
bfff6873
LC		 3002 eli->li_state |= EXT4_LAZYINIT_QUIT;
3003
3004 ext4_li_info = eli;
3005
3006 return 0;
3007}
3008
3009static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3010 ext4_group_t start)
3011{
3012 struct ext4_sb_info *sbi = EXT4_SB(sb);
3013 struct ext4_li_request *elr;
bfff6873
LC		 3014
3015 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3016 if (!elr)
3017 return NULL;
3018
3019 elr->lr_super = sb;
3020 elr->lr_sbi = sbi;
3021 elr->lr_next_group = start;
3022
3023 /*
3024 * Randomize first schedule time of the request to
3025 * spread the inode table initialization requests
3026 * better.
3027 */
dd1f723b
TT		 3028 elr->lr_next_sched = jiffies + (prandom_u32() %
3029 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3030 return elr;
3031}
3032
7f511862
TT		 3033int ext4_register_li_request(struct super_block *sb,
3034 ext4_group_t first_not_zeroed)
bfff6873
LC		 3035{
3036 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3037 struct ext4_li_request *elr = NULL;
bfff6873 3038 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3039 int ret = 0;
bfff6873 3040
7f511862 3041 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3042 if (sbi->s_li_request != NULL) {
3043 /*
3044 * Reset timeout so it can be computed again, because
3045 * s_li_wait_mult might have changed.
3046 */
3047 sbi->s_li_request->lr_timeout = 0;
7f511862 3048 goto out;
51ce6511 3049 }
bfff6873
LC		 3050
3051 if (first_not_zeroed == ngroups ||
3052 (sb->s_flags & MS_RDONLY) ||
55ff3840 3053 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3054 goto out;
bfff6873
LC		 3055
3056 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3057 if (!elr) {
3058 ret = -ENOMEM;
3059 goto out;
3060 }
bfff6873
LC		 3061
3062 if (NULL == ext4_li_info) {
3063 ret = ext4_li_info_new();
3064 if (ret)
3065 goto out;
3066 }
3067
3068 mutex_lock(&ext4_li_info->li_list_mtx);
3069 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3070 mutex_unlock(&ext4_li_info->li_list_mtx);
3071
3072 sbi->s_li_request = elr;
46e4690b
TM		 3073 /*
3074 * set elr to NULL here since it has been inserted to
3075 * the request_list and the removal and free of it is
3076 * handled by ext4_clear_request_list from now on.
3077 */
3078 elr = NULL;
bfff6873
LC		 3079
3080 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3081 ret = ext4_run_lazyinit_thread();
3082 if (ret)
3083 goto out;
3084 }
bfff6873 3085out:
beed5ecb
NK		 3086 mutex_unlock(&ext4_li_mtx);
3087 if (ret)
bfff6873 3088 kfree(elr);
bfff6873
LC		 3089 return ret;
3090}
3091
3092/*
3093 * We do not need to lock anything since this is called on
3094 * module unload.
3095 */
3096static void ext4_destroy_lazyinit_thread(void)
3097{
3098 /*
3099 * If thread exited earlier
3100 * there's nothing to be done.
3101 */
8f1f7453 3102 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3103 return;
3104
8f1f7453 3105 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3106}
3107
25ed6e8a
DW		 3108static int set_journal_csum_feature_set(struct super_block *sb)
3109{
3110 int ret = 1;
3111 int compat, incompat;
3112 struct ext4_sb_info *sbi = EXT4_SB(sb);
3113
9aa5d32b 3114 if (ext4_has_metadata_csum(sb)) {
db9ee220 3115 /* journal checksum v3 */
25ed6e8a 3116 compat = 0;
db9ee220 3117 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3118 } else {
3119 /* journal checksum v1 */
3120 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3121 incompat = 0;
3122 }
3123
feb8c6d3
DW		 3124 jbd2_journal_clear_features(sbi->s_journal,
3125 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3126 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3127 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3128 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3129 ret = jbd2_journal_set_features(sbi->s_journal,
3130 compat, 0,
3131 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3132 incompat);
3133 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3134 ret = jbd2_journal_set_features(sbi->s_journal,
3135 compat, 0,
3136 incompat);
3137 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3138 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3139 } else {
feb8c6d3
DW		 3140 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3141 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3142 }
3143
3144 return ret;
3145}
3146
952fc18e
TT		 3147/*
3148 * Note: calculating the overhead so we can be compatible with
3149 * historical BSD practice is quite difficult in the face of
3150 * clusters/bigalloc. This is because multiple metadata blocks from
3151 * different block group can end up in the same allocation cluster.
3152 * Calculating the exact overhead in the face of clustered allocation
3153 * requires either O(all block bitmaps) in memory or O(number of block
3154 * groups**2) in time. We will still calculate the superblock for
3155 * older file systems --- and if we come across with a bigalloc file
3156 * system with zero in s_overhead_clusters the estimate will be close to
3157 * correct especially for very large cluster sizes --- but for newer
3158 * file systems, it's better to calculate this figure once at mkfs
3159 * time, and store it in the superblock. If the superblock value is
3160 * present (even for non-bigalloc file systems), we will use it.
3161 */
3162static int count_overhead(struct super_block *sb, ext4_group_t grp,
3163 char *buf)
3164{
3165 struct ext4_sb_info *sbi = EXT4_SB(sb);
3166 struct ext4_group_desc *gdp;
3167 ext4_fsblk_t first_block, last_block, b;
3168 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3169 int s, j, count = 0;
3170
e2b911c5 3171 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3172 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3173 sbi->s_itb_per_group + 2);
3174
952fc18e
TT		 3175 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3176 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3177 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3178 for (i = 0; i < ngroups; i++) {
3179 gdp = ext4_get_group_desc(sb, i, NULL);
3180 b = ext4_block_bitmap(sb, gdp);
3181 if (b >= first_block && b <= last_block) {
3182 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3183 count++;
3184 }
3185 b = ext4_inode_bitmap(sb, gdp);
3186 if (b >= first_block && b <= last_block) {
3187 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3188 count++;
3189 }
3190 b = ext4_inode_table(sb, gdp);
3191 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3192 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3193 int c = EXT4_B2C(sbi, b - first_block);
3194 ext4_set_bit(c, buf);
3195 count++;
3196 }
3197 if (i != grp)
3198 continue;
3199 s = 0;
3200 if (ext4_bg_has_super(sb, grp)) {
3201 ext4_set_bit(s++, buf);
3202 count++;
3203 }
c48ae41b
TT		 3204 j = ext4_bg_num_gdb(sb, grp);
3205 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3206 ext4_error(sb, "Invalid number of block group "
3207 "descriptor blocks: %d", j);
3208 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3209 }
c48ae41b
TT		 3210 count += j;
3211 for (; j > 0; j--)
3212 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3213 }
3214 if (!count)
3215 return 0;
3216 return EXT4_CLUSTERS_PER_GROUP(sb) -
3217 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3218}
3219
3220/*
3221 * Compute the overhead and stash it in sbi->s_overhead
3222 */
3223int ext4_calculate_overhead(struct super_block *sb)
3224{
3225 struct ext4_sb_info *sbi = EXT4_SB(sb);
3226 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3227 struct inode *j_inode;
3228 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3229 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3230 ext4_fsblk_t overhead = 0;
4fdb5543 3231 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3232
952fc18e
TT		 3233 if (!buf)
3234 return -ENOMEM;
3235
3236 /*
3237 * Compute the overhead (FS structures). This is constant
3238 * for a given filesystem unless the number of block groups
3239 * changes so we cache the previous value until it does.
3240 */
3241
3242 /*
3243 * All of the blocks before first_data_block are overhead
3244 */
3245 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3246
3247 /*
3248 * Add the overhead found in each block group
3249 */
3250 for (i = 0; i < ngroups; i++) {
3251 int blks;
3252
3253 blks = count_overhead(sb, i, buf);
3254 overhead += blks;
3255 if (blks)
3256 memset(buf, 0, PAGE_SIZE);
3257 cond_resched();
3258 }
3c816ded
EW		 3259
3260 /*
3261 * Add the internal journal blocks whether the journal has been
3262 * loaded or not
3263 */
b003b524 3264 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3265 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3266 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3267 j_inode = ext4_get_journal_inode(sb, j_inum);
3268 if (j_inode) {
3269 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3270 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3271 iput(j_inode);
3272 } else {
3273 ext4_msg(sb, KERN_ERR, "can't get journal size");
3274 }
3275 }
952fc18e
TT		 3276 sbi->s_overhead = overhead;
3277 smp_wmb();
3278 free_page((unsigned long) buf);
3279 return 0;
3280}
3281
b5799018 3282static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3283{
3284 ext4_fsblk_t resv_clusters;
b5799018 3285 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3286
30fac0f7
JK		 3287 /*
3288 * There's no need to reserve anything when we aren't using extents.
3289 * The space estimates are exact, there are no unwritten extents,
3290 * hole punching doesn't need new metadata... This is needed especially
3291 * to keep ext2/3 backward compatibility.
3292 */
e2b911c5 3293 if (!ext4_has_feature_extents(sb))
b5799018 3294 return;
27dd4385
LC		 3295 /*
3296 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3297 * This should cover the situations where we can not afford to run
3298 * out of space like for example punch hole, or converting
556615dc 3299 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3300 * allocation would require 1, or 2 blocks, higher numbers are
3301 * very rare.
3302 */
b5799018
TT		 3303 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3304 sbi->s_cluster_bits);
27dd4385
LC		 3305
3306 do_div(resv_clusters, 50);
3307 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3308
b5799018 3309 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3310}
3311
2b2d6d01 3312static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3313{
d4c402d9 3314 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3315 struct buffer_head *bh;
617ba13b 3316 struct ext4_super_block *es = NULL;
5aee0f8a 3317 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3318 ext4_fsblk_t block;
3319 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3320 ext4_fsblk_t logical_sb_block;
ac27a0ec 3321 unsigned long offset = 0;
ac27a0ec
DK		 3322 unsigned long journal_devnum = 0;
3323 unsigned long def_mount_opts;
3324 struct inode *root;
0390131b 3325 const char *descr;
dcc7dae3 3326 int ret = -ENOMEM;
281b5995 3327 int blocksize, clustersize;
4ec11028
TT		 3328 unsigned int db_count;
3329 unsigned int i;
281b5995 3330 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3331 __u64 blocks_count;
07aa2ea1 3332 int err = 0;
b3881f74 3333 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3334 ext4_group_t first_not_zeroed;
ac27a0ec 3335
5aee0f8a
TT		 3336 if ((data && !orig_data) || !sbi)
3337 goto out_free_base;
705895b6
PE		 3338
3339 sbi->s_blockgroup_lock =
3340 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3341 if (!sbi->s_blockgroup_lock)
3342 goto out_free_base;
3343
ac27a0ec 3344 sb->s_fs_info = sbi;
2c0544b2 3345 sbi->s_sb = sb;
240799cd 3346 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3347 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3348 if (sb->s_bdev->bd_part)
3349 sbi->s_sectors_written_start =
3350 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3351
9f6200bb 3352 /* Cleanup superblock name */
ec3904dc 3353 strreplace(sb->s_id, '/', '!');
9f6200bb 3354
07aa2ea1 3355 /* -EINVAL is default */
dcc7dae3 3356 ret = -EINVAL;
617ba13b 3357 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3358 if (!blocksize) {
b31e1552 3359 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3360 goto out_fail;
3361 }
3362
3363 /*
617ba13b 3364 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3365 * block sizes. We need to calculate the offset from buffer start.
3366 */
617ba13b 3367 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3368 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3369 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3370 } else {
70bbb3e0 3371 logical_sb_block = sb_block;
ac27a0ec
DK		 3372 }
3373
a8ac900b 3374 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3375 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3376 goto out_fail;
3377 }
3378 /*
3379 * Note: s_es must be initialized as soon as possible because
617ba13b 3380 * some ext4 macro-instructions depend on its value
ac27a0ec 3381 */
2716b802 3382 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3383 sbi->s_es = es;
3384 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3385 if (sb->s_magic != EXT4_SUPER_MAGIC)
3386 goto cantfind_ext4;
afc32f7e 3387 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3388
feb0ab32 3389 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3390 if (ext4_has_feature_metadata_csum(sb) &&
3391 ext4_has_feature_gdt_csum(sb))
363307e6 3392 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3393 "redundant flags; please run fsck.");
3394
d25425f8
DW		 3395 /* Check for a known checksum algorithm */
3396 if (!ext4_verify_csum_type(sb, es)) {
3397 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3398 "unknown checksum algorithm.");
3399 silent = 1;
3400 goto cantfind_ext4;
3401 }
3402
0441984a 3403 /* Load the checksum driver */
e2b911c5 3404 if (ext4_has_feature_metadata_csum(sb)) {
0441984a
DW		 3405 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3406 if (IS_ERR(sbi->s_chksum_driver)) {
3407 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3408 ret = PTR_ERR(sbi->s_chksum_driver);
3409 sbi->s_chksum_driver = NULL;
3410 goto failed_mount;
3411 }
3412 }
3413
a9c47317
DW		 3414 /* Check superblock checksum */
3415 if (!ext4_superblock_csum_verify(sb, es)) {
3416 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3417 "invalid superblock checksum. Run e2fsck?");
3418 silent = 1;
6a797d27 3419 ret = -EFSBADCRC;
a9c47317
DW		 3420 goto cantfind_ext4;
3421 }
3422
3423 /* Precompute checksum seed for all metadata */
e2b911c5 3424 if (ext4_has_feature_csum_seed(sb))
8c81bd8f
DW		 3425 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3426 else if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3427 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3428 sizeof(es->s_uuid));
3429
ac27a0ec
DK		 3430 /* Set defaults before we parse the mount options */
3431 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3432 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3433 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3434 set_opt(sb, DEBUG);
87f26807 3435 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3436 set_opt(sb, GRPID);
617ba13b 3437 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3438 set_opt(sb, NO_UID32);
ea663336 3439 /* xattr user namespace & acls are now defaulted on */
ea663336 3440 set_opt(sb, XATTR_USER);
03010a33 3441#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3442 set_opt(sb, POSIX_ACL);
2e7842b8 3443#endif
98c1a759
DW		 3444 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3445 if (ext4_has_metadata_csum(sb))
3446 set_opt(sb, JOURNAL_CHECKSUM);
3447
617ba13b 3448 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3449 set_opt(sb, JOURNAL_DATA);
617ba13b 3450 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3451 set_opt(sb, ORDERED_DATA);
617ba13b 3452 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3453 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3454
3455 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3456 set_opt(sb, ERRORS_PANIC);
bb4f397a 3457 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3458 set_opt(sb, ERRORS_CONT);
bb4f397a 3459 else
fd8c37ec 3460 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3461 /* block_validity enabled by default; disable with noblock_validity */
3462 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3463 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3464 set_opt(sb, DISCARD);
ac27a0ec 3465
08cefc7a
EB		 3466 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3467 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3468 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3469 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3470 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3471
8b67f04a 3472 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3473 set_opt(sb, BARRIER);
ac27a0ec 3474
dd919b98
AK		 3475 /*
3476 * enable delayed allocation by default
3477 * Use -o nodelalloc to turn it off
3478 */
bc0b75f7 3479 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3480 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3481 set_opt(sb, DELALLOC);
dd919b98 3482
51ce6511
LC		 3483 /*
3484 * set default s_li_wait_mult for lazyinit, for the case there is
3485 * no mount option specified.
3486 */
3487 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3488
5aee0f8a
TT		 3489 if (sbi->s_es->s_mount_opts[0]) {
3490 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3491 sizeof(sbi->s_es->s_mount_opts),
3492 GFP_KERNEL);
3493 if (!s_mount_opts)
3494 goto failed_mount;
3495 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3496 &journal_ioprio, 0)) {
3497 ext4_msg(sb, KERN_WARNING,
3498 "failed to parse options in superblock: %s",
3499 s_mount_opts);
3500 }
3501 kfree(s_mount_opts);
8b67f04a 3502 }
5a916be1 3503 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3504 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3505 &journal_ioprio, 0))
ac27a0ec
DK		 3506 goto failed_mount;
3507
56889787
TT		 3508 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3509 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3510 "with data=journal disables delayed "
3511 "allocation and O_DIRECT support!\n");
3512 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3513 ext4_msg(sb, KERN_ERR, "can't mount with "
3514 "both data=journal and delalloc");
3515 goto failed_mount;
3516 }
3517 if (test_opt(sb, DIOREAD_NOLOCK)) {
3518 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3519 "both data=journal and dioread_nolock");
56889787
TT		 3520 goto failed_mount;
3521 }
923ae0ff
RZ		 3522 if (test_opt(sb, DAX)) {
3523 ext4_msg(sb, KERN_ERR, "can't mount with "
3524 "both data=journal and dax");
3525 goto failed_mount;
3526 }
73b92a2a
SK		 3527 if (ext4_has_feature_encrypt(sb)) {
3528 ext4_msg(sb, KERN_WARNING,
3529 "encrypted files will use data=ordered "
3530 "instead of data journaling mode");
3531 }
56889787
TT		 3532 if (test_opt(sb, DELALLOC))
3533 clear_opt(sb, DELALLOC);
001e4a87
TH		 3534 } else {
3535 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3536 }
3537
ac27a0ec 3538 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3539 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3540
617ba13b 3541 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3542 (ext4_has_compat_features(sb) ||
3543 ext4_has_ro_compat_features(sb) ||
3544 ext4_has_incompat_features(sb)))
b31e1552
ES		 3545 ext4_msg(sb, KERN_WARNING,
3546 "feature flags set on rev 0 fs, "
3547 "running e2fsck is recommended");
469108ff 3548
ed3654eb
TT		 3549 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3550 set_opt2(sb, HURD_COMPAT);
e2b911c5 3551 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3552 ext4_msg(sb, KERN_ERR,
3553 "The Hurd can't support 64-bit file systems");
3554 goto failed_mount;
3555 }
3556 }
3557
2035e776
TT		 3558 if (IS_EXT2_SB(sb)) {
3559 if (ext2_feature_set_ok(sb))
3560 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3561 "using the ext4 subsystem");
3562 else {
3563 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3564 "to feature incompatibilities");
3565 goto failed_mount;
3566 }
3567 }
3568
3569 if (IS_EXT3_SB(sb)) {
3570 if (ext3_feature_set_ok(sb))
3571 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3572 "using the ext4 subsystem");
3573 else {
3574 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3575 "to feature incompatibilities");
3576 goto failed_mount;
3577 }
3578 }
3579
ac27a0ec
DK		 3580 /*
3581 * Check feature flags regardless of the revision level, since we
3582 * previously didn't change the revision level when setting the flags,
3583 * so there is a chance incompat flags are set on a rev 0 filesystem.
3584 */
a13fb1a4 3585 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3586 goto failed_mount;
a13fb1a4 3587
261cb20c 3588 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3589 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3590 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3591 ext4_msg(sb, KERN_ERR,
8cdf3372
TT		 3592 "Unsupported filesystem blocksize %d (%d log_block_size)",
3593 blocksize, le32_to_cpu(es->s_log_block_size));
3594 goto failed_mount;
3595 }
3596 if (le32_to_cpu(es->s_log_block_size) >
3597 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3598 ext4_msg(sb, KERN_ERR,
3599 "Invalid log block size: %u",
3600 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3601 goto failed_mount;
3602 }
3603
5b9554dc
TT		 3604 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3605 ext4_msg(sb, KERN_ERR,
3606 "Number of reserved GDT blocks insanely large: %d",
3607 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3608 goto failed_mount;
3609 }
3610
923ae0ff 3611 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
87eefeb4
TK		 3612 err = bdev_dax_supported(sb, blocksize);
3613 if (err)
923ae0ff 3614 goto failed_mount;
923ae0ff
RZ		 3615 }
3616
e2b911c5 3617 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3618 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3619 es->s_encryption_level);
3620 goto failed_mount;
3621 }
3622
ac27a0ec 3623 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3624 /* Validate the filesystem blocksize */
3625 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3626 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3627 blocksize);
ac27a0ec
DK		 3628 goto failed_mount;
3629 }
3630
2b2d6d01 3631 brelse(bh);
70bbb3e0
AM		 3632 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3633 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3634 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3635 if (!bh) {
b31e1552
ES		 3636 ext4_msg(sb, KERN_ERR,
3637 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3638 goto failed_mount;
3639 }
2716b802 3640 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3641 sbi->s_es = es;
617ba13b 3642 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3643 ext4_msg(sb, KERN_ERR,
3644 "Magic mismatch, very weird!");
ac27a0ec
DK		 3645 goto failed_mount;
3646 }
3647 }
3648
e2b911c5 3649 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3650 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3651 has_huge_files);
3652 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3653
617ba13b
MC		 3654 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3655 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3656 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3657 } else {
3658 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3659 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3660 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3661 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3662 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3663 ext4_msg(sb, KERN_ERR,
3664 "unsupported inode size: %d",
2b2d6d01 3665 sbi->s_inode_size);
ac27a0ec
DK		 3666 goto failed_mount;
3667 }
ef7f3835
KS		 3668 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3669 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3670 }
0b8e58a1 3671
0d1ee42f 3672 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3673 if (ext4_has_feature_64bit(sb)) {
8fadc143 3674 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3675 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3676 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3677 ext4_msg(sb, KERN_ERR,
3678 "unsupported descriptor size %lu",
0d1ee42f
AR		 3679 sbi->s_desc_size);
3680 goto failed_mount;
3681 }
3682 } else
3683 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3684
ac27a0ec 3685 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3686 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 3687
617ba13b 3688 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3689 if (sbi->s_inodes_per_block == 0)
617ba13b 3690 goto cantfind_ext4;
cd6bb35b
TT		 3691 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3692 sbi->s_inodes_per_group > blocksize * 8) {
3693 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3694 sbi->s_blocks_per_group);
3695 goto failed_mount;
3696 }
ac27a0ec
DK		 3697 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3698 sbi->s_inodes_per_block;
0d1ee42f 3699 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3700 sbi->s_sbh = bh;
3701 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3702 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3703 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3704
2b2d6d01 3705 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3706 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3707 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3708 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3709 i = le32_to_cpu(es->s_flags);
3710 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3711 sbi->s_hash_unsigned = 3;
3712 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3713#ifdef __CHAR_UNSIGNED__
23301410
TT		 3714 if (!(sb->s_flags & MS_RDONLY))
3715 es->s_flags |=
3716 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3717 sbi->s_hash_unsigned = 3;
f99b2589 3718#else
23301410
TT		 3719 if (!(sb->s_flags & MS_RDONLY))
3720 es->s_flags |=
3721 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3722#endif
23301410 3723 }
f99b2589 3724 }
ac27a0ec 3725
281b5995
TT		 3726 /* Handle clustersize */
3727 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3728 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3729 if (has_bigalloc) {
3730 if (clustersize < blocksize) {
3731 ext4_msg(sb, KERN_ERR,
3732 "cluster size (%d) smaller than "
3733 "block size (%d)", clustersize, blocksize);
3734 goto failed_mount;
3735 }
8cdf3372
TT		 3736 if (le32_to_cpu(es->s_log_cluster_size) >
3737 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3738 ext4_msg(sb, KERN_ERR,
3739 "Invalid log cluster size: %u",
3740 le32_to_cpu(es->s_log_cluster_size));
3741 goto failed_mount;
3742 }
281b5995
TT		 3743 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3744 le32_to_cpu(es->s_log_block_size);
3745 sbi->s_clusters_per_group =
3746 le32_to_cpu(es->s_clusters_per_group);
3747 if (sbi->s_clusters_per_group > blocksize * 8) {
3748 ext4_msg(sb, KERN_ERR,
3749 "#clusters per group too big: %lu",
3750 sbi->s_clusters_per_group);
3751 goto failed_mount;
3752 }
3753 if (sbi->s_blocks_per_group !=
3754 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3755 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3756 "clusters per group (%lu) inconsistent",
3757 sbi->s_blocks_per_group,
3758 sbi->s_clusters_per_group);
3759 goto failed_mount;
3760 }
3761 } else {
3762 if (clustersize != blocksize) {
3763 ext4_warning(sb, "fragment/cluster size (%d) != "
3764 "block size (%d)", clustersize,
3765 blocksize);
3766 clustersize = blocksize;
3767 }
3768 if (sbi->s_blocks_per_group > blocksize * 8) {
3769 ext4_msg(sb, KERN_ERR,
3770 "#blocks per group too big: %lu",
3771 sbi->s_blocks_per_group);
3772 goto failed_mount;
3773 }
3774 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3775 sbi->s_cluster_bits = 0;
ac27a0ec 3776 }
281b5995
TT		 3777 sbi->s_cluster_ratio = clustersize / blocksize;
3778
960fd856
TT		 3779 /* Do we have standard group size of clustersize * 8 blocks ? */
3780 if (sbi->s_blocks_per_group == clustersize << 3)
3781 set_opt2(sb, STD_GROUP_SIZE);
3782
bf43d84b
ES		 3783 /*
3784 * Test whether we have more sectors than will fit in sector_t,
3785 * and whether the max offset is addressable by the page cache.
3786 */
5a9ae68a 3787 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3788 ext4_blocks_count(es));
5a9ae68a 3789 if (err) {
b31e1552 3790 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3791 " too large to mount safely on this system");
ac27a0ec 3792 if (sizeof(sector_t) < 8)
90c699a9 3793 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3794 goto failed_mount;
3795 }
3796
617ba13b
MC		 3797 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3798 goto cantfind_ext4;
e7c95593 3799
0f2ddca6
FTN		 3800 /* check blocks count against device size */
3801 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3802 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3803 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3804 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3805 ext4_blocks_count(es), blocks_count);
3806 goto failed_mount;
3807 }
3808
0b8e58a1
AD		 3809 /*
3810 * It makes no sense for the first data block to be beyond the end
3811 * of the filesystem.
3812 */
3813 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3814 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3815 "block %u is beyond end of filesystem (%llu)",
3816 le32_to_cpu(es->s_first_data_block),
3817 ext4_blocks_count(es));
e7c95593
ES		 3818 goto failed_mount;
3819 }
bd81d8ee
LV		 3820 blocks_count = (ext4_blocks_count(es) -
3821 le32_to_cpu(es->s_first_data_block) +
3822 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3823 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3824 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3825 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3826 "(block count %llu, first data block %u, "
b31e1552 3827 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3828 ext4_blocks_count(es),
3829 le32_to_cpu(es->s_first_data_block),
3830 EXT4_BLOCKS_PER_GROUP(sb));
3831 goto failed_mount;
3832 }
bd81d8ee 3833 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3834 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3835 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3836 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3837 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd
EG		 3838 if (ext4_has_feature_meta_bg(sb)) {
3839 if (le32_to_cpu(es->s_first_meta_bg) >= db_count) {
3840 ext4_msg(sb, KERN_WARNING,
3841 "first meta block group too large: %u "
3842 "(group descriptor block count %u)",
3843 le32_to_cpu(es->s_first_meta_bg), db_count);
3844 goto failed_mount;
3845 }
3846 }
f18a5f21
TT		 3847 sbi->s_group_desc = ext4_kvmalloc(db_count *
3848 sizeof(struct buffer_head *),
3849 GFP_KERNEL);
ac27a0ec 3850 if (sbi->s_group_desc == NULL) {
b31e1552 3851 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3852 ret = -ENOMEM;
ac27a0ec
DK		 3853 goto failed_mount;
3854 }
3855
705895b6 3856 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3857
3858 for (i = 0; i < db_count; i++) {
70bbb3e0 3859 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3860 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3861 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3862 ext4_msg(sb, KERN_ERR,
3863 "can't read group descriptor %d", i);
ac27a0ec
DK		 3864 db_count = i;
3865 goto failed_mount2;
3866 }
3867 }
829fa70d 3868 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 3869 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 3870 ret = -EFSCORRUPTED;
f9ae9cf5 3871 goto failed_mount2;
ac27a0ec 3872 }
772cb7c8 3873
f9ae9cf5 3874 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3875 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3876 spin_lock_init(&sbi->s_next_gen_lock);
3877
04ecddb7
JM		 3878 setup_timer(&sbi->s_err_report, print_daily_error_info,
3879 (unsigned long) sb);
04496411 3880
a75ae78f 3881 /* Register extent status tree shrinker */
eb68d0e2 3882 if (ext4_es_register_shrinker(sbi))
ce7e010a 3883 goto failed_mount3;
ce7e010a 3884
c9de560d 3885 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3886 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3887
f9ae9cf5
TT		 3888 /*
3889 * set up enough so that it can read an inode
3890 */
f6e63f90 3891 sb->s_op = &ext4_sops;
617ba13b
MC		 3892 sb->s_export_op = &ext4_export_ops;
3893 sb->s_xattr = ext4_xattr_handlers;
a7550b30 3894 sb->s_cop = &ext4_cryptops;
ac27a0ec 3895#ifdef CONFIG_QUOTA
617ba13b 3896 sb->dq_op = &ext4_quota_operations;
e2b911c5 3897 if (ext4_has_feature_quota(sb))
1fa5efe3 3898 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3899 else
3900 sb->s_qcop = &ext4_qctl_operations;
689c958c 3901 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 3902#endif
f2fa2ffc
AK		 3903 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3904
ac27a0ec 3905 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3906 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3907
3908 sb->s_root = NULL;
3909
3910 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 3911 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 3912
e2b911c5 3913 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
c5e06d10 3914 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3915 goto failed_mount3a;
c5e06d10 3916
ac27a0ec
DK		 3917 /*
3918 * The first inode we look at is the journal inode. Don't try
3919 * root first: it may be modified in the journal!
3920 */
e2b911c5 3921 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
617ba13b 3922 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3923 goto failed_mount3a;
0390131b 3924 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
e2b911c5 3925 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 3926 ext4_msg(sb, KERN_ERR, "required journal recovery "
3927 "suppressed and not mounted read-only");
744692dc 3928 goto failed_mount_wq;
ac27a0ec 3929 } else {
1e381f60
DM		 3930 /* Nojournal mode, all journal mount options are illegal */
3931 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3932 ext4_msg(sb, KERN_ERR, "can't mount with "
3933 "journal_checksum, fs mounted w/o journal");
3934 goto failed_mount_wq;
3935 }
3936 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3937 ext4_msg(sb, KERN_ERR, "can't mount with "
3938 "journal_async_commit, fs mounted w/o journal");
3939 goto failed_mount_wq;
3940 }
3941 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3942 ext4_msg(sb, KERN_ERR, "can't mount with "
3943 "commit=%lu, fs mounted w/o journal",
3944 sbi->s_commit_interval / HZ);
3945 goto failed_mount_wq;
3946 }
3947 if (EXT4_MOUNT_DATA_FLAGS &
3948 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3949 ext4_msg(sb, KERN_ERR, "can't mount with "
3950 "data=, fs mounted w/o journal");
3951 goto failed_mount_wq;
3952 }
3953 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3954 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 3955 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3956 sbi->s_journal = NULL;
3957 needs_recovery = 0;
3958 goto no_journal;
ac27a0ec
DK		 3959 }
3960
e2b911c5 3961 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 3962 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3963 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3964 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3965 goto failed_mount_wq;
eb40a09c
JS		 3966 }
3967
25ed6e8a
DW		 3968 if (!set_journal_csum_feature_set(sb)) {
3969 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3970 "feature set");
3971 goto failed_mount_wq;
d4da6c9c 3972 }
818d276c 3973
ac27a0ec
DK		 3974 /* We have now updated the journal if required, so we can
3975 * validate the data journaling mode. */
3976 switch (test_opt(sb, DATA_FLAGS)) {
3977 case 0:
3978 /* No mode set, assume a default based on the journal
63f57933
AM		 3979 * capabilities: ORDERED_DATA if the journal can
3980 * cope, else JOURNAL_DATA
3981 */
dab291af
MC		 3982 if (jbd2_journal_check_available_features
3983 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3984 set_opt(sb, ORDERED_DATA);
ac27a0ec 3985 else
fd8c37ec 3986 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3987 break;
3988
617ba13b
MC		 3989 case EXT4_MOUNT_ORDERED_DATA:
3990 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3991 if (!jbd2_journal_check_available_features
3992 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3993 ext4_msg(sb, KERN_ERR, "Journal does not support "
3994 "requested data journaling mode");
744692dc 3995 goto failed_mount_wq;
ac27a0ec
DK		 3996 }
3997 default:
3998 break;
3999 }
ab04df78
JK		 4000
4001 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4002 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4003 ext4_msg(sb, KERN_ERR, "can't mount with "
4004 "journal_async_commit in data=ordered mode");
4005 goto failed_mount_wq;
4006 }
4007
b3881f74 4008 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4009
18aadd47
BJ		 4010 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4011
ce7e010a 4012no_journal:
2335d05f
AG		 4013 sbi->s_mb_cache = ext4_xattr_create_cache();
4014 if (!sbi->s_mb_cache) {
4015 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4016 goto failed_mount_wq;
9c191f70
M		 4017 }
4018
e2b911c5 4019 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 4020 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 4021 ext4_msg(sb, KERN_ERR,
4022 "Unsupported blocksize for fs encryption");
4023 goto failed_mount_wq;
4024 }
4025
e2b911c5
DW		 4026 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4027 !ext4_has_feature_encrypt(sb)) {
4028 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 4029 ext4_commit_super(sb, 1);
4030 }
4031
952fc18e
TT		 4032 /*
4033 * Get the # of file system overhead blocks from the
4034 * superblock if present.
4035 */
4036 if (es->s_overhead_clusters)
4037 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4038 else {
07aa2ea1
LC		 4039 err = ext4_calculate_overhead(sb);
4040 if (err)
952fc18e
TT		 4041 goto failed_mount_wq;
4042 }
4043
fd89d5f2
TH		 4044 /*
4045 * The maximum number of concurrent works can be high and
4046 * concurrency isn't really necessary. Limit it to 1.
4047 */
2e8fa54e
JK		 4048 EXT4_SB(sb)->rsv_conversion_wq =
4049 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4050 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4051 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4052 ret = -ENOMEM;
2e8fa54e
JK		 4053 goto failed_mount4;
4054 }
4055
ac27a0ec 4056 /*
dab291af 4057 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4058 * so we can safely mount the rest of the filesystem now.
4059 */
4060
1d1fe1ee
DH		 4061 root = ext4_iget(sb, EXT4_ROOT_INO);
4062 if (IS_ERR(root)) {
b31e1552 4063 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4064 ret = PTR_ERR(root);
32a9bb57 4065 root = NULL;
ac27a0ec
DK		 4066 goto failed_mount4;
4067 }
4068 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4069 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4070 iput(root);
ac27a0ec
DK		 4071 goto failed_mount4;
4072 }
48fde701 4073 sb->s_root = d_make_root(root);
1d1fe1ee 4074 if (!sb->s_root) {
b31e1552 4075 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4076 ret = -ENOMEM;
4077 goto failed_mount4;
4078 }
ac27a0ec 4079
7e84b621
ES		 4080 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4081 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4082
4083 /* determine the minimum size of new large inodes, if present */
4084 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4085 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4086 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4087 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4088 if (sbi->s_want_extra_isize <
4089 le16_to_cpu(es->s_want_extra_isize))
4090 sbi->s_want_extra_isize =
4091 le16_to_cpu(es->s_want_extra_isize);
4092 if (sbi->s_want_extra_isize <
4093 le16_to_cpu(es->s_min_extra_isize))
4094 sbi->s_want_extra_isize =
4095 le16_to_cpu(es->s_min_extra_isize);
4096 }
4097 }
4098 /* Check if enough inode space is available */
4099 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4100 sbi->s_inode_size) {
4101 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4102 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4103 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4104 "available");
ef7f3835
KS		 4105 }
4106
b5799018 4107 ext4_set_resv_clusters(sb);
27dd4385 4108
6fd058f7
TT		 4109 err = ext4_setup_system_zone(sb);
4110 if (err) {
b31e1552 4111 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4112 "zone (%d)", err);
f9ae9cf5
TT		 4113 goto failed_mount4a;
4114 }
4115
4116 ext4_ext_init(sb);
4117 err = ext4_mb_init(sb);
4118 if (err) {
4119 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4120 err);
dcf2d804 4121 goto failed_mount5;
c2774d84
AK		 4122 }
4123
d5e03cbb
TT		 4124 block = ext4_count_free_clusters(sb);
4125 ext4_free_blocks_count_set(sbi->s_es,
4126 EXT4_C2B(sbi, block));
908c7f19
TH		 4127 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4128 GFP_KERNEL);
d5e03cbb
TT		 4129 if (!err) {
4130 unsigned long freei = ext4_count_free_inodes(sb);
4131 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4132 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4133 GFP_KERNEL);
d5e03cbb
TT		 4134 }
4135 if (!err)
4136 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4137 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4138 if (!err)
908c7f19
TH		 4139 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4140 GFP_KERNEL);
c8585c6f
DJ		 4141 if (!err)
4142 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4143
d5e03cbb
TT		 4144 if (err) {
4145 ext4_msg(sb, KERN_ERR, "insufficient memory");
4146 goto failed_mount6;
4147 }
4148
e2b911c5 4149 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4150 if (!ext4_fill_flex_info(sb)) {
4151 ext4_msg(sb, KERN_ERR,
4152 "unable to initialize "
4153 "flex_bg meta info!");
4154 goto failed_mount6;
4155 }
4156
bfff6873
LC		 4157 err = ext4_register_li_request(sb, first_not_zeroed);
4158 if (err)
dcf2d804 4159 goto failed_mount6;
bfff6873 4160
b5799018 4161 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4162 if (err)
4163 goto failed_mount7;
3197ebdb 4164
9b2ff357
JK		 4165#ifdef CONFIG_QUOTA
4166 /* Enable quota usage during mount. */
e2b911c5 4167 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
9b2ff357
JK		 4168 err = ext4_enable_quotas(sb);
4169 if (err)
4170 goto failed_mount8;
4171 }
4172#endif /* CONFIG_QUOTA */
4173
617ba13b
MC		 4174 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4175 ext4_orphan_cleanup(sb, es);
4176 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4177 if (needs_recovery) {
b31e1552 4178 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4179 ext4_mark_recovery_complete(sb, es);
4180 }
4181 if (EXT4_SB(sb)->s_journal) {
4182 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4183 descr = " journalled data mode";
4184 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4185 descr = " ordered data mode";
4186 else
4187 descr = " writeback data mode";
4188 } else
4189 descr = "out journal";
4190
79add3a3
LC		 4191 if (test_opt(sb, DISCARD)) {
4192 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4193 if (!blk_queue_discard(q))
4194 ext4_msg(sb, KERN_WARNING,
4195 "mounting with \"discard\" option, but "
4196 "the device does not support discard");
4197 }
4198
e294a537
TT		 4199 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4200 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4201 "Opts: %.*s%s%s", descr,
4202 (int) sizeof(sbi->s_es->s_mount_opts),
4203 sbi->s_es->s_mount_opts,
e294a537 4204 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4205
66e61a9e
TT		 4206 if (es->s_error_count)
4207 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4208
efbed4dc
TT		 4209 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4210 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4211 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4212 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4213
d4c402d9 4214 kfree(orig_data);
ac27a0ec
DK		 4215 return 0;
4216
617ba13b 4217cantfind_ext4:
ac27a0ec 4218 if (!silent)
b31e1552 4219 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4220 goto failed_mount;
4221
72ba7450
TT		 4222#ifdef CONFIG_QUOTA
4223failed_mount8:
ebd173be 4224 ext4_unregister_sysfs(sb);
72ba7450 4225#endif
dcf2d804
TM		 4226failed_mount7:
4227 ext4_unregister_li_request(sb);
4228failed_mount6:
f9ae9cf5 4229 ext4_mb_release(sb);
d5e03cbb 4230 if (sbi->s_flex_groups)
b93b41d4 4231 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4232 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4233 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4234 percpu_counter_destroy(&sbi->s_dirs_counter);
4235 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4236failed_mount5:
f9ae9cf5
TT		 4237 ext4_ext_release(sb);
4238 ext4_release_system_zone(sb);
4239failed_mount4a:
94bf608a 4240 dput(sb->s_root);
32a9bb57 4241 sb->s_root = NULL;
94bf608a 4242failed_mount4:
b31e1552 4243 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4244 if (EXT4_SB(sb)->rsv_conversion_wq)
4245 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4246failed_mount_wq:
82939d79
JK		 4247 if (sbi->s_mb_cache) {
4248 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4249 sbi->s_mb_cache = NULL;
4250 }
0390131b
FM		 4251 if (sbi->s_journal) {
4252 jbd2_journal_destroy(sbi->s_journal);
4253 sbi->s_journal = NULL;
4254 }
50460fe8 4255failed_mount3a:
d3922a77 4256 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4257failed_mount3:
9105bb14 4258 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4259 if (sbi->s_mmp_tsk)
4260 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4261failed_mount2:
4262 for (i = 0; i < db_count; i++)
4263 brelse(sbi->s_group_desc[i]);
b93b41d4 4264 kvfree(sbi->s_group_desc);
ac27a0ec 4265failed_mount:
0441984a
DW		 4266 if (sbi->s_chksum_driver)
4267 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4268#ifdef CONFIG_QUOTA
a2d4a646 4269 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4270 kfree(sbi->s_qf_names[i]);
4271#endif
617ba13b 4272 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4273 brelse(bh);
4274out_fail:
4275 sb->s_fs_info = NULL;
f6830165 4276 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4277out_free_base:
ac27a0ec 4278 kfree(sbi);
d4c402d9 4279 kfree(orig_data);
07aa2ea1 4280 return err ? err : ret;
ac27a0ec
DK		 4281}
4282
4283/*
4284 * Setup any per-fs journal parameters now. We'll do this both on
4285 * initial mount, once the journal has been initialised but before we've
4286 * done any recovery; and again on any subsequent remount.
4287 */
617ba13b 4288static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4289{
617ba13b 4290 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4291
30773840
TT		 4292 journal->j_commit_interval = sbi->s_commit_interval;
4293 journal->j_min_batch_time = sbi->s_min_batch_time;
4294 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4295
a931da6a 4296 write_lock(&journal->j_state_lock);
ac27a0ec 4297 if (test_opt(sb, BARRIER))
dab291af 4298 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4299 else
dab291af 4300 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4301 if (test_opt(sb, DATA_ERR_ABORT))
4302 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4303 else
4304 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4305 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4306}
4307
c6cb7e77
EW		 4308static struct inode *ext4_get_journal_inode(struct super_block *sb,
4309 unsigned int journal_inum)
ac27a0ec
DK		 4310{
4311 struct inode *journal_inode;
ac27a0ec 4312
c6cb7e77
EW		 4313 /*
4314 * Test for the existence of a valid inode on disk. Bad things
4315 * happen if we iget() an unused inode, as the subsequent iput()
4316 * will try to delete it.
4317 */
1d1fe1ee
DH		 4318 journal_inode = ext4_iget(sb, journal_inum);
4319 if (IS_ERR(journal_inode)) {
b31e1552 4320 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4321 return NULL;
4322 }
4323 if (!journal_inode->i_nlink) {
4324 make_bad_inode(journal_inode);
4325 iput(journal_inode);
b31e1552 4326 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4327 return NULL;
4328 }
4329
e5f8eab8 4330 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4331 journal_inode, journal_inode->i_size);
1d1fe1ee 4332 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4333 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4334 iput(journal_inode);
4335 return NULL;
4336 }
c6cb7e77
EW		 4337 return journal_inode;
4338}
4339
4340static journal_t *ext4_get_journal(struct super_block *sb,
4341 unsigned int journal_inum)
4342{
4343 struct inode *journal_inode;
4344 journal_t *journal;
4345
4346 BUG_ON(!ext4_has_feature_journal(sb));
4347
4348 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4349 if (!journal_inode)
4350 return NULL;
ac27a0ec 4351
dab291af 4352 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4353 if (!journal) {
b31e1552 4354 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4355 iput(journal_inode);
4356 return NULL;
4357 }
4358 journal->j_private = sb;
617ba13b 4359 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4360 return journal;
4361}
4362
617ba13b 4363static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4364 dev_t j_dev)
4365{
2b2d6d01 4366 struct buffer_head *bh;
ac27a0ec 4367 journal_t *journal;
617ba13b
MC		 4368 ext4_fsblk_t start;
4369 ext4_fsblk_t len;
ac27a0ec 4370 int hblock, blocksize;
617ba13b 4371 ext4_fsblk_t sb_block;
ac27a0ec 4372 unsigned long offset;
2b2d6d01 4373 struct ext4_super_block *es;
ac27a0ec
DK		 4374 struct block_device *bdev;
4375
e2b911c5 4376 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4377
b31e1552 4378 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4379 if (bdev == NULL)
4380 return NULL;
4381
ac27a0ec 4382 blocksize = sb->s_blocksize;
e1defc4f 4383 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4384 if (blocksize < hblock) {
b31e1552
ES		 4385 ext4_msg(sb, KERN_ERR,
4386 "blocksize too small for journal device");
ac27a0ec
DK		 4387 goto out_bdev;
4388 }
4389
617ba13b
MC		 4390 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4391 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4392 set_blocksize(bdev, blocksize);
4393 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4394 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4395 "external journal");
ac27a0ec
DK		 4396 goto out_bdev;
4397 }
4398
2716b802 4399 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4400 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4401 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4402 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4403 ext4_msg(sb, KERN_ERR, "external journal has "
4404 "bad superblock");
ac27a0ec
DK		 4405 brelse(bh);
4406 goto out_bdev;
4407 }
4408
df4763be
DW		 4409 if ((le32_to_cpu(es->s_feature_ro_compat) &
4410 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4411 es->s_checksum != ext4_superblock_csum(sb, es)) {
4412 ext4_msg(sb, KERN_ERR, "external journal has "
4413 "corrupt superblock");
4414 brelse(bh);
4415 goto out_bdev;
4416 }
4417
617ba13b 4418 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4419 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4420 brelse(bh);
4421 goto out_bdev;
4422 }
4423
bd81d8ee 4424 len = ext4_blocks_count(es);
ac27a0ec
DK		 4425 start = sb_block + 1;
4426 brelse(bh); /* we're done with the superblock */
4427
dab291af 4428 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4429 start, len, blocksize);
4430 if (!journal) {
b31e1552 4431 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4432 goto out_bdev;
4433 }
4434 journal->j_private = sb;
dfec8a14 4435 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4436 wait_on_buffer(journal->j_sb_buffer);
4437 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4438 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4439 goto out_journal;
4440 }
4441 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4442 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4443 "user (unsupported) - %d",
ac27a0ec
DK		 4444 be32_to_cpu(journal->j_superblock->s_nr_users));
4445 goto out_journal;
4446 }
617ba13b
MC		 4447 EXT4_SB(sb)->journal_bdev = bdev;
4448 ext4_init_journal_params(sb, journal);
ac27a0ec 4449 return journal;
0b8e58a1 4450
ac27a0ec 4451out_journal:
dab291af 4452 jbd2_journal_destroy(journal);
ac27a0ec 4453out_bdev:
617ba13b 4454 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4455 return NULL;
4456}
4457
617ba13b
MC		 4458static int ext4_load_journal(struct super_block *sb,
4459 struct ext4_super_block *es,
ac27a0ec
DK		 4460 unsigned long journal_devnum)
4461{
4462 journal_t *journal;
4463 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4464 dev_t journal_dev;
4465 int err = 0;
4466 int really_read_only;
4467
e2b911c5 4468 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4469
ac27a0ec
DK		 4470 if (journal_devnum &&
4471 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4472 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4473 "numbers have changed");
ac27a0ec
DK		 4474 journal_dev = new_decode_dev(journal_devnum);
4475 } else
4476 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4477
4478 really_read_only = bdev_read_only(sb->s_bdev);
4479
4480 /*
4481 * Are we loading a blank journal or performing recovery after a
4482 * crash? For recovery, we need to check in advance whether we
4483 * can get read-write access to the device.
4484 */
e2b911c5 4485 if (ext4_has_feature_journal_needs_recovery(sb)) {
ac27a0ec 4486 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4487 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4488 "required on readonly filesystem");
ac27a0ec 4489 if (really_read_only) {
b31e1552
ES		 4490 ext4_msg(sb, KERN_ERR, "write access "
4491 "unavailable, cannot proceed");
ac27a0ec
DK		 4492 return -EROFS;
4493 }
b31e1552
ES		 4494 ext4_msg(sb, KERN_INFO, "write access will "
4495 "be enabled during recovery");
ac27a0ec
DK		 4496 }
4497 }
4498
4499 if (journal_inum && journal_dev) {
b31e1552
ES		 4500 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4501 "and inode journals!");
ac27a0ec
DK		 4502 return -EINVAL;
4503 }
4504
4505 if (journal_inum) {
617ba13b 4506 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4507 return -EINVAL;
4508 } else {
617ba13b 4509 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4510 return -EINVAL;
4511 }
4512
90576c0b 4513 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4514 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4515
e2b911c5 4516 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4517 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4518 if (!err) {
4519 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4520 if (save)
4521 memcpy(save, ((char *) es) +
4522 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4523 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4524 if (save)
4525 memcpy(((char *) es) + EXT4_S_ERR_START,
4526 save, EXT4_S_ERR_LEN);
4527 kfree(save);
4528 }
ac27a0ec
DK		 4529
4530 if (err) {
b31e1552 4531 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4532 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4533 return err;
4534 }
4535
617ba13b
MC		 4536 EXT4_SB(sb)->s_journal = journal;
4537 ext4_clear_journal_err(sb, es);
ac27a0ec 4538
c41303ce 4539 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4540 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4541 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4542
4543 /* Make sure we flush the recovery flag to disk. */
e2d67052 4544 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4545 }
4546
4547 return 0;
4548}
4549
e2d67052 4550static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4551{
e2d67052 4552 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4553 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4554 int error = 0;
ac27a0ec 4555
bdfe0cbd 4556 if (!sbh || block_device_ejected(sb))
c4be0c1d 4557 return error;
71290b36
TT		 4558 /*
4559 * If the file system is mounted read-only, don't update the
4560 * superblock write time. This avoids updating the superblock
4561 * write time when we are mounting the root file system
4562 * read/only but we need to replay the journal; at that point,
4563 * for people who are east of GMT and who make their clock
4564 * tick in localtime for Windows bug-for-bug compatibility,
4565 * the clock is set in the future, and this will cause e2fsck
4566 * to complain and force a full file system check.
4567 */
4568 if (!(sb->s_flags & MS_RDONLY))
4569 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4570 if (sb->s_bdev->bd_part)
4571 es->s_kbytes_written =
4572 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4573 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4574 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4575 else
4576 es->s_kbytes_written =
4577 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4578 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4579 ext4_free_blocks_count_set(es,
57042651
TT		 4580 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4581 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4582 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4583 es->s_free_inodes_count =
4584 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4585 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4586 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4587 ext4_superblock_csum_set(sb);
1566a48a
TT		 4588 if (sync)
4589 lock_buffer(sbh);
4743f839
PKS		 4590 if (buffer_write_io_error(sbh)) {
4591 /*
4592 * Oh, dear. A previous attempt to write the
4593 * superblock failed. This could happen because the
4594 * USB device was yanked out. Or it could happen to
4595 * be a transient write error and maybe the block will
4596 * be remapped. Nothing we can do but to retry the
4597 * write and hope for the best.
4598 */
4599 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4600 "superblock detected");
4601 clear_buffer_write_io_error(sbh);
4602 set_buffer_uptodate(sbh);
4603 }
ac27a0ec 4604 mark_buffer_dirty(sbh);
914258bf 4605 if (sync) {
1566a48a 4606 unlock_buffer(sbh);
564bc402 4607 error = __sync_dirty_buffer(sbh,
70fd7614 4608 test_opt(sb, BARRIER) ? REQ_FUA : REQ_SYNC);
c4be0c1d
TS		 4609 if (error)
4610 return error;
4611
4612 error = buffer_write_io_error(sbh);
4613 if (error) {
b31e1552
ES		 4614 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4615 "superblock");
914258bf
TT		 4616 clear_buffer_write_io_error(sbh);
4617 set_buffer_uptodate(sbh);
4618 }
4619 }
c4be0c1d 4620 return error;
ac27a0ec
DK		 4621}
4622
ac27a0ec
DK		 4623/*
4624 * Have we just finished recovery? If so, and if we are mounting (or
4625 * remounting) the filesystem readonly, then we will end up with a
4626 * consistent fs on disk. Record that fact.
4627 */
2b2d6d01
TT		 4628static void ext4_mark_recovery_complete(struct super_block *sb,
4629 struct ext4_super_block *es)
ac27a0ec 4630{
617ba13b 4631 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4632
e2b911c5 4633 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4634 BUG_ON(journal != NULL);
4635 return;
4636 }
dab291af 4637 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4638 if (jbd2_journal_flush(journal) < 0)
4639 goto out;
4640
e2b911c5 4641 if (ext4_has_feature_journal_needs_recovery(sb) &&
ac27a0ec 4642 sb->s_flags & MS_RDONLY) {
e2b911c5 4643 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4644 ext4_commit_super(sb, 1);
ac27a0ec 4645 }
7ffe1ea8
HK		 4646
4647out:
dab291af 4648 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4649}
4650
4651/*
4652 * If we are mounting (or read-write remounting) a filesystem whose journal
4653 * has recorded an error from a previous lifetime, move that error to the
4654 * main filesystem now.
4655 */
2b2d6d01
TT		 4656static void ext4_clear_journal_err(struct super_block *sb,
4657 struct ext4_super_block *es)
ac27a0ec
DK		 4658{
4659 journal_t *journal;
4660 int j_errno;
4661 const char *errstr;
4662
e2b911c5 4663 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4664
617ba13b 4665 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4666
4667 /*
4668 * Now check for any error status which may have been recorded in the
617ba13b 4669 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4670 */
4671
dab291af 4672 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4673 if (j_errno) {
4674 char nbuf[16];
4675
617ba13b 4676 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4677 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4678 "from previous mount: %s", errstr);
12062ddd 4679 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4680
617ba13b
MC		 4681 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4682 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4683 ext4_commit_super(sb, 1);
ac27a0ec 4684
dab291af 4685 jbd2_journal_clear_err(journal);
d796c52e 4686 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4687 }
4688}
4689
4690/*
4691 * Force the running and committing transactions to commit,
4692 * and wait on the commit.
4693 */
617ba13b 4694int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4695{
4696 journal_t *journal;
ac27a0ec
DK		 4697
4698 if (sb->s_flags & MS_RDONLY)
4699 return 0;
4700
617ba13b 4701 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4702 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4703}
4704
617ba13b 4705static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4706{
14ce0cb4 4707 int ret = 0;
9eddacf9 4708 tid_t target;
06a407f1 4709 bool needs_barrier = false;
8d5d02e6 4710 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4711
9bffad1e 4712 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4713 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4714 /*
4715 * Writeback quota in non-journalled quota case - journalled quota has
4716 * no dirty dquots
4717 */
4718 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4719 /*
4720 * Data writeback is possible w/o journal transaction, so barrier must
4721 * being sent at the end of the function. But we can skip it if
4722 * transaction_commit will do it for us.
4723 */
bda32530
TT		 4724 if (sbi->s_journal) {
4725 target = jbd2_get_latest_transaction(sbi->s_journal);
4726 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4727 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4728 needs_barrier = true;
4729
4730 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4731 if (wait)
4732 ret = jbd2_log_wait_commit(sbi->s_journal,
4733 target);
4734 }
4735 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4736 needs_barrier = true;
06a407f1
DM		 4737 if (needs_barrier) {
4738 int err;
4739 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4740 if (!ret)
4741 ret = err;
0390131b 4742 }
06a407f1
DM		 4743
4744 return ret;
4745}
4746
ac27a0ec
DK		 4747/*
4748 * LVM calls this function before a (read-only) snapshot is created. This
4749 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4750 *
4751 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4752 * state independently. It relies on upper layer to stop all data & metadata
4753 * modifications.
ac27a0ec 4754 */
c4be0c1d 4755static int ext4_freeze(struct super_block *sb)
ac27a0ec 4756{
c4be0c1d
TS		 4757 int error = 0;
4758 journal_t *journal;
ac27a0ec 4759
9ca92389
TT		 4760 if (sb->s_flags & MS_RDONLY)
4761 return 0;
ac27a0ec 4762
9ca92389 4763 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4764
bb044576
TT		 4765 if (journal) {
4766 /* Now we set up the journal barrier. */
4767 jbd2_journal_lock_updates(journal);
ac27a0ec 4768
bb044576
TT		 4769 /*
4770 * Don't clear the needs_recovery flag if we failed to
4771 * flush the journal.
4772 */
4773 error = jbd2_journal_flush(journal);
4774 if (error < 0)
4775 goto out;
c642dc9e
ES		 4776
4777 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4778 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4779 }
9ca92389 4780
9ca92389 4781 error = ext4_commit_super(sb, 1);
6b0310fb 4782out:
bb044576
TT		 4783 if (journal)
4784 /* we rely on upper layer to stop further updates */
4785 jbd2_journal_unlock_updates(journal);
6b0310fb 4786 return error;
ac27a0ec
DK		 4787}
4788
4789/*
4790 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4791 * flag here, even though the filesystem is not technically dirty yet.
4792 */
c4be0c1d 4793static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4794{
9ca92389
TT		 4795 if (sb->s_flags & MS_RDONLY)
4796 return 0;
4797
c642dc9e
ES		 4798 if (EXT4_SB(sb)->s_journal) {
4799 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4800 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4801 }
4802
9ca92389 4803 ext4_commit_super(sb, 1);
c4be0c1d 4804 return 0;
ac27a0ec
DK		 4805}
4806
673c6100
TT		 4807/*
4808 * Structure to save mount options for ext4_remount's benefit
4809 */
4810struct ext4_mount_options {
4811 unsigned long s_mount_opt;
a2595b8a 4812 unsigned long s_mount_opt2;
08cefc7a
EB		 4813 kuid_t s_resuid;
4814 kgid_t s_resgid;
673c6100
TT		 4815 unsigned long s_commit_interval;
4816 u32 s_min_batch_time, s_max_batch_time;
4817#ifdef CONFIG_QUOTA
4818 int s_jquota_fmt;
a2d4a646 4819 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4820#endif
4821};
4822
2b2d6d01 4823static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4824{
2b2d6d01 4825 struct ext4_super_block *es;
617ba13b 4826 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4827 unsigned long old_sb_flags;
617ba13b 4828 struct ext4_mount_options old_opts;
c79d967d 4829 int enable_quota = 0;
8a266467 4830 ext4_group_t g;
b3881f74 4831 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4832 int err = 0;
ac27a0ec 4833#ifdef CONFIG_QUOTA
03dafb5f 4834 int i, j;
ac27a0ec 4835#endif
d4c402d9 4836 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4837
4838 /* Store the original options */
4839 old_sb_flags = sb->s_flags;
4840 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4841 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4842 old_opts.s_resuid = sbi->s_resuid;
4843 old_opts.s_resgid = sbi->s_resgid;
4844 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4845 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4846 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4847#ifdef CONFIG_QUOTA
4848 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4849 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4850 if (sbi->s_qf_names[i]) {
4851 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4852 GFP_KERNEL);
4853 if (!old_opts.s_qf_names[i]) {
4854 for (j = 0; j < i; j++)
4855 kfree(old_opts.s_qf_names[j]);
3e36a163 4856 kfree(orig_data);
03dafb5f
CG		 4857 return -ENOMEM;
4858 }
4859 } else
4860 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4861#endif
b3881f74
TT		 4862 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4863 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4864
661aa520 4865 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4866 err = -EINVAL;
4867 goto restore_opts;
4868 }
4869
6b992ff2 4870 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4871 test_opt(sb, JOURNAL_CHECKSUM)) {
4872 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4873 "during remount not supported; ignoring");
4874 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4875 }
4876
6ae6514b
PS		 4877 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4878 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4879 ext4_msg(sb, KERN_ERR, "can't mount with "
4880 "both data=journal and delalloc");
4881 err = -EINVAL;
4882 goto restore_opts;
4883 }
4884 if (test_opt(sb, DIOREAD_NOLOCK)) {
4885 ext4_msg(sb, KERN_ERR, "can't mount with "
4886 "both data=journal and dioread_nolock");
4887 err = -EINVAL;
4888 goto restore_opts;
4889 }
923ae0ff
RZ		 4890 if (test_opt(sb, DAX)) {
4891 ext4_msg(sb, KERN_ERR, "can't mount with "
4892 "both data=journal and dax");
4893 err = -EINVAL;
4894 goto restore_opts;
4895 }
ab04df78
JK		 4896 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
4897 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4898 ext4_msg(sb, KERN_ERR, "can't mount with "
4899 "journal_async_commit in data=ordered mode");
4900 err = -EINVAL;
4901 goto restore_opts;
4902 }
923ae0ff
RZ		 4903 }
4904
4905 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4906 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4907 "dax flag with busy inodes while remounting");
4908 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4909 }
4910
4ab2f15b 4911 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4912 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4913
4914 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4915 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4916
4917 es = sbi->s_es;
4918
b3881f74 4919 if (sbi->s_journal) {
0390131b 4920 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4921 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4922 }
ac27a0ec 4923
a2fd66d0
TT		 4924 if (*flags & MS_LAZYTIME)
4925 sb->s_flags |= MS_LAZYTIME;
4926
661aa520 4927 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4928 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4929 err = -EROFS;
4930 goto restore_opts;
4931 }
4932
4933 if (*flags & MS_RDONLY) {
38c03b34
TT		 4934 err = sync_filesystem(sb);
4935 if (err < 0)
4936 goto restore_opts;
0f0dd62f
CH		 4937 err = dquot_suspend(sb, -1);
4938 if (err < 0)
c79d967d 4939 goto restore_opts;
c79d967d 4940
ac27a0ec
DK		 4941 /*
4942 * First of all, the unconditional stuff we have to do
4943 * to disable replay of the journal when we next remount
4944 */
4945 sb->s_flags |= MS_RDONLY;
4946
4947 /*
4948 * OK, test if we are remounting a valid rw partition
4949 * readonly, and if so set the rdonly flag and then
4950 * mark the partition as valid again.
4951 */
617ba13b
MC		 4952 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4953 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4954 es->s_state = cpu_to_le16(sbi->s_mount_state);
4955
a63c9eb2 4956 if (sbi->s_journal)
0390131b 4957 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4958 } else {
a13fb1a4 4959 /* Make sure we can mount this feature set readwrite */
e2b911c5 4960 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 4961 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4962 err = -EROFS;
4963 goto restore_opts;
4964 }
8a266467
TT		 4965 /*
4966 * Make sure the group descriptor checksums
0b8e58a1 4967 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4968 */
4969 for (g = 0; g < sbi->s_groups_count; g++) {
4970 struct ext4_group_desc *gdp =
4971 ext4_get_group_desc(sb, g, NULL);
4972
feb0ab32 4973 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4974 ext4_msg(sb, KERN_ERR,
4975 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 4976 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 4977 le16_to_cpu(gdp->bg_checksum));
6a797d27 4978 err = -EFSBADCRC;
8a266467
TT		 4979 goto restore_opts;
4980 }
4981 }
4982
ead6596b
ES		 4983 /*
4984 * If we have an unprocessed orphan list hanging
4985 * around from a previously readonly bdev mount,
4986 * require a full umount/remount for now.
4987 */
4988 if (es->s_last_orphan) {
b31e1552 4989 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4990 "remount RDWR because of unprocessed "
4991 "orphan inode list. Please "
b31e1552 4992 "umount/remount instead");
ead6596b
ES		 4993 err = -EINVAL;
4994 goto restore_opts;
4995 }
4996
ac27a0ec
DK		 4997 /*
4998 * Mounting a RDONLY partition read-write, so reread
4999 * and store the current valid flag. (It may have
5000 * been changed by e2fsck since we originally mounted
5001 * the partition.)
5002 */
0390131b
FM		 5003 if (sbi->s_journal)
5004 ext4_clear_journal_err(sb, es);
ac27a0ec 5005 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 5006 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 5007 sb->s_flags &= ~MS_RDONLY;
e2b911c5 5008 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 5009 if (ext4_multi_mount_protect(sb,
5010 le64_to_cpu(es->s_mmp_block))) {
5011 err = -EROFS;
5012 goto restore_opts;
5013 }
c79d967d 5014 enable_quota = 1;
ac27a0ec
DK		 5015 }
5016 }
bfff6873
LC		 5017
5018 /*
5019 * Reinitialize lazy itable initialization thread based on
5020 * current settings
5021 */
5022 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5023 ext4_unregister_li_request(sb);
5024 else {
5025 ext4_group_t first_not_zeroed;
5026 first_not_zeroed = ext4_has_uninit_itable(sb);
5027 ext4_register_li_request(sb, first_not_zeroed);
5028 }
5029
6fd058f7 5030 ext4_setup_system_zone(sb);
d096ad0f 5031 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5032 ext4_commit_super(sb, 1);
0390131b 5033
ac27a0ec
DK		 5034#ifdef CONFIG_QUOTA
5035 /* Release old quota file names */
a2d4a646 5036 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5037 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5038 if (enable_quota) {
5039 if (sb_any_quota_suspended(sb))
5040 dquot_resume(sb, -1);
e2b911c5 5041 else if (ext4_has_feature_quota(sb)) {
7c319d32 5042 err = ext4_enable_quotas(sb);
07724f98 5043 if (err)
7c319d32 5044 goto restore_opts;
7c319d32
AK		 5045 }
5046 }
ac27a0ec 5047#endif
d4c402d9 5048
a26f4992 5049 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5050 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5051 kfree(orig_data);
ac27a0ec 5052 return 0;
0b8e58a1 5053
ac27a0ec
DK		 5054restore_opts:
5055 sb->s_flags = old_sb_flags;
5056 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5057 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5058 sbi->s_resuid = old_opts.s_resuid;
5059 sbi->s_resgid = old_opts.s_resgid;
5060 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5061 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5062 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5063#ifdef CONFIG_QUOTA
5064 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5065 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5066 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5067 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5068 }
5069#endif
d4c402d9 5070 kfree(orig_data);
ac27a0ec
DK		 5071 return err;
5072}
5073
689c958c
LX		 5074#ifdef CONFIG_QUOTA
5075static int ext4_statfs_project(struct super_block *sb,
5076 kprojid_t projid, struct kstatfs *buf)
5077{
5078 struct kqid qid;
5079 struct dquot *dquot;
5080 u64 limit;
5081 u64 curblock;
5082
5083 qid = make_kqid_projid(projid);
5084 dquot = dqget(sb, qid);
5085 if (IS_ERR(dquot))
5086 return PTR_ERR(dquot);
5087 spin_lock(&dq_data_lock);
5088
5089 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5090 dquot->dq_dqb.dqb_bsoftlimit :
5091 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5092 if (limit && buf->f_blocks > limit) {
5093 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5094 buf->f_blocks = limit;
5095 buf->f_bfree = buf->f_bavail =
5096 (buf->f_blocks > curblock) ?
5097 (buf->f_blocks - curblock) : 0;
5098 }
5099
5100 limit = dquot->dq_dqb.dqb_isoftlimit ?
5101 dquot->dq_dqb.dqb_isoftlimit :
5102 dquot->dq_dqb.dqb_ihardlimit;
5103 if (limit && buf->f_files > limit) {
5104 buf->f_files = limit;
5105 buf->f_ffree =
5106 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5107 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5108 }
5109
5110 spin_unlock(&dq_data_lock);
5111 dqput(dquot);
5112 return 0;
5113}
5114#endif
5115
2b2d6d01 5116static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5117{
5118 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5119 struct ext4_sb_info *sbi = EXT4_SB(sb);
5120 struct ext4_super_block *es = sbi->s_es;
27dd4385 5121 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5122 u64 fsid;
d02a9391 5123 s64 bfree;
27dd4385 5124 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5125
952fc18e
TT		 5126 if (!test_opt(sb, MINIX_DF))
5127 overhead = sbi->s_overhead;
ac27a0ec 5128
617ba13b 5129 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5130 buf->f_bsize = sb->s_blocksize;
b72f78cb 5131 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5132 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5133 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5134 /* prevent underflow in case that few free space is available */
57042651 5135 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5136 buf->f_bavail = buf->f_bfree -
5137 (ext4_r_blocks_count(es) + resv_blocks);
5138 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5139 buf->f_bavail = 0;
5140 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5141 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5142 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5143 fsid = le64_to_cpup((void *)es->s_uuid) ^
5144 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5145 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5146 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5147
689c958c
LX		 5148#ifdef CONFIG_QUOTA
5149 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5150 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5151 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5152#endif
ac27a0ec
DK		 5153 return 0;
5154}
5155
0b8e58a1
AD		 5156/* Helper function for writing quotas on sync - we need to start transaction
5157 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5158 * Process 1 Process 2
617ba13b 5159 * ext4_create() quota_sync()
a269eb18 5160 * jbd2_journal_start() write_dquot()
871a2931 5161 * dquot_initialize() down(dqio_mutex)
dab291af 5162 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5163 *
5164 */
5165
5166#ifdef CONFIG_QUOTA
5167
5168static inline struct inode *dquot_to_inode(struct dquot *dquot)
5169{
4c376dca 5170 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5171}
5172
617ba13b 5173static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5174{
5175 int ret, err;
5176 handle_t *handle;
5177 struct inode *inode;
5178
5179 inode = dquot_to_inode(dquot);
9924a92a 5180 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5181 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5182 if (IS_ERR(handle))
5183 return PTR_ERR(handle);
5184 ret = dquot_commit(dquot);
617ba13b 5185 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5186 if (!ret)
5187 ret = err;
5188 return ret;
5189}
5190
617ba13b 5191static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5192{
5193 int ret, err;
5194 handle_t *handle;
5195
9924a92a 5196 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5197 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5198 if (IS_ERR(handle))
5199 return PTR_ERR(handle);
5200 ret = dquot_acquire(dquot);
617ba13b 5201 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5202 if (!ret)
5203 ret = err;
5204 return ret;
5205}
5206
617ba13b 5207static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5208{
5209 int ret, err;
5210 handle_t *handle;
5211
9924a92a 5212 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5213 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5214 if (IS_ERR(handle)) {
5215 /* Release dquot anyway to avoid endless cycle in dqput() */
5216 dquot_release(dquot);
ac27a0ec 5217 return PTR_ERR(handle);
9c3013e9 5218 }
ac27a0ec 5219 ret = dquot_release(dquot);
617ba13b 5220 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5221 if (!ret)
5222 ret = err;
5223 return ret;
5224}
5225
617ba13b 5226static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5227{
262b4662
JK		 5228 struct super_block *sb = dquot->dq_sb;
5229 struct ext4_sb_info *sbi = EXT4_SB(sb);
5230
2c8be6b2 5231 /* Are we journaling quotas? */
e2b911c5 5232 if (ext4_has_feature_quota(sb) ||
262b4662 5233 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5234 dquot_mark_dquot_dirty(dquot);
617ba13b 5235 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5236 } else {
5237 return dquot_mark_dquot_dirty(dquot);
5238 }
5239}
5240
617ba13b 5241static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5242{
5243 int ret, err;
5244 handle_t *handle;
5245
5246 /* Data block + inode block */
2b0143b5 5247 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5248 if (IS_ERR(handle))
5249 return PTR_ERR(handle);
5250 ret = dquot_commit_info(sb, type);
617ba13b 5251 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5252 if (!ret)
5253 ret = err;
5254 return ret;
5255}
5256
5257/*
5258 * Turn on quotas during mount time - we need to find
5259 * the quota file and such...
5260 */
617ba13b 5261static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5262{
287a8095
CH		 5263 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5264 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5265}
5266
daf647d2
TT		 5267static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5268{
5269 struct ext4_inode_info *ei = EXT4_I(inode);
5270
5271 /* The first argument of lockdep_set_subclass has to be
5272 * *exactly* the same as the argument to init_rwsem() --- in
5273 * this case, in init_once() --- or lockdep gets unhappy
5274 * because the name of the lock is set using the
5275 * stringification of the argument to init_rwsem().
5276 */
5277 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5278 lockdep_set_subclass(&ei->i_data_sem, subclass);
5279}
5280
ac27a0ec
DK		 5281/*
5282 * Standard function to be called on quota_on
5283 */
617ba13b 5284static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 5285 const struct path *path)
ac27a0ec
DK		 5286{
5287 int err;
ac27a0ec
DK		 5288
5289 if (!test_opt(sb, QUOTA))
5290 return -EINVAL;
0623543b 5291
ac27a0ec 5292 /* Quotafile not on the same filesystem? */
d8c9584e 5293 if (path->dentry->d_sb != sb)
ac27a0ec 5294 return -EXDEV;
0623543b
JK		 5295 /* Journaling quota? */
5296 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5297 /* Quotafile not in fs root? */
f00c9e44 5298 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5299 ext4_msg(sb, KERN_WARNING,
5300 "Quota file not on filesystem root. "
5301 "Journaled quota will not work");
2b2d6d01 5302 }
0623543b
JK		 5303
5304 /*
5305 * When we journal data on quota file, we have to flush journal to see
5306 * all updates to the file when we bypass pagecache...
5307 */
0390131b 5308 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5309 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5310 /*
5311 * We don't need to lock updates but journal_flush() could
5312 * otherwise be livelocked...
5313 */
5314 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5315 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5316 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5317 if (err)
7ffe1ea8 5318 return err;
0623543b 5319 }
daf647d2
TT		 5320 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5321 err = dquot_quota_on(sb, type, format_id, path);
5322 if (err)
5323 lockdep_set_quota_inode(path->dentry->d_inode,
5324 I_DATA_SEM_NORMAL);
5325 return err;
ac27a0ec
DK		 5326}
5327
7c319d32
AK		 5328static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5329 unsigned int flags)
5330{
5331 int err;
5332 struct inode *qf_inode;
a2d4a646 5333 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5334 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5335 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5336 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5337 };
5338
e2b911c5 5339 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5340
5341 if (!qf_inums[type])
5342 return -EPERM;
5343
5344 qf_inode = ext4_iget(sb, qf_inums[type]);
5345 if (IS_ERR(qf_inode)) {
5346 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5347 return PTR_ERR(qf_inode);
5348 }
5349
bcb13850
JK		 5350 /* Don't account quota for quota files to avoid recursion */
5351 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5352 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5353 err = dquot_enable(qf_inode, type, format_id, flags);
5354 iput(qf_inode);
daf647d2
TT		 5355 if (err)
5356 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5357
5358 return err;
5359}
5360
5361/* Enable usage tracking for all quota types. */
5362static int ext4_enable_quotas(struct super_block *sb)
5363{
5364 int type, err = 0;
a2d4a646 5365 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5366 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5367 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5368 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 5369 };
49da9392
JK		 5370 bool quota_mopt[EXT4_MAXQUOTAS] = {
5371 test_opt(sb, USRQUOTA),
5372 test_opt(sb, GRPQUOTA),
5373 test_opt(sb, PRJQUOTA),
5374 };
7c319d32
AK		 5375
5376 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5377 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5378 if (qf_inums[type]) {
5379 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5380 DQUOT_USAGE_ENABLED |
5381 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 5382 if (err) {
5383 ext4_warning(sb,
72ba7450
TT		 5384 "Failed to enable quota tracking "
5385 "(type=%d, err=%d). Please run "
5386 "e2fsck to fix.", type, err);
7c319d32
AK		 5387 return err;
5388 }
5389 }
5390 }
5391 return 0;
5392}
5393
ca0e05e4
DM		 5394static int ext4_quota_off(struct super_block *sb, int type)
5395{
21f97697
JK		 5396 struct inode *inode = sb_dqopt(sb)->files[type];
5397 handle_t *handle;
5398
87009d86
DM		 5399 /* Force all delayed allocation blocks to be allocated.
5400 * Caller already holds s_umount sem */
5401 if (test_opt(sb, DELALLOC))
ca0e05e4 5402 sync_filesystem(sb);
ca0e05e4 5403
0b268590
AG		 5404 if (!inode)
5405 goto out;
5406
21f97697
JK		 5407 /* Update modification times of quota files when userspace can
5408 * start looking at them */
9924a92a 5409 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5410 if (IS_ERR(handle))
5411 goto out;
eeca7ea1 5412 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5413 ext4_mark_inode_dirty(handle, inode);
5414 ext4_journal_stop(handle);
5415
5416out:
ca0e05e4
DM		 5417 return dquot_quota_off(sb, type);
5418}
5419
ac27a0ec
DK		 5420/* Read data from quotafile - avoid pagecache and such because we cannot afford
5421 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5422 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5423 * we don't have to be afraid of races */
617ba13b 5424static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5425 size_t len, loff_t off)
5426{
5427 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5428 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5429 int offset = off & (sb->s_blocksize - 1);
5430 int tocopy;
5431 size_t toread;
5432 struct buffer_head *bh;
5433 loff_t i_size = i_size_read(inode);
5434
5435 if (off > i_size)
5436 return 0;
5437 if (off+len > i_size)
5438 len = i_size-off;
5439 toread = len;
5440 while (toread > 0) {
5441 tocopy = sb->s_blocksize - offset < toread ?
5442 sb->s_blocksize - offset : toread;
1c215028
TT		 5443 bh = ext4_bread(NULL, inode, blk, 0);
5444 if (IS_ERR(bh))
5445 return PTR_ERR(bh);
ac27a0ec
DK		 5446 if (!bh) /* A hole? */
5447 memset(data, 0, tocopy);
5448 else
5449 memcpy(data, bh->b_data+offset, tocopy);
5450 brelse(bh);
5451 offset = 0;
5452 toread -= tocopy;
5453 data += tocopy;
5454 blk++;
5455 }
5456 return len;
5457}
5458
5459/* Write to quotafile (we know the transaction is already started and has
5460 * enough credits) */
617ba13b 5461static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5462 const char *data, size_t len, loff_t off)
5463{
5464 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5465 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5466 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5467 int retries = 0;
ac27a0ec
DK		 5468 struct buffer_head *bh;
5469 handle_t *handle = journal_current_handle();
5470
0390131b 5471 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5472 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5473 " cancelled because transaction is not started",
9c3013e9
JK		 5474 (unsigned long long)off, (unsigned long long)len);
5475 return -EIO;
5476 }
67eeb568
DM		 5477 /*
5478 * Since we account only one data block in transaction credits,
5479 * then it is impossible to cross a block boundary.
5480 */
5481 if (sb->s_blocksize - offset < len) {
5482 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5483 " cancelled because not block aligned",
5484 (unsigned long long)off, (unsigned long long)len);
5485 return -EIO;
5486 }
5487
c5e298ae
TT		 5488 do {
5489 bh = ext4_bread(handle, inode, blk,
5490 EXT4_GET_BLOCKS_CREATE |
5491 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5492 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5493 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5494 if (IS_ERR(bh))
5495 return PTR_ERR(bh);
67eeb568
DM		 5496 if (!bh)
5497 goto out;
5d601255 5498 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5499 err = ext4_journal_get_write_access(handle, bh);
5500 if (err) {
5501 brelse(bh);
1c215028 5502 return err;
ac27a0ec 5503 }
67eeb568
DM		 5504 lock_buffer(bh);
5505 memcpy(bh->b_data+offset, data, len);
5506 flush_dcache_page(bh->b_page);
5507 unlock_buffer(bh);
62d2b5f2 5508 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5509 brelse(bh);
ac27a0ec 5510out:
67eeb568
DM		 5511 if (inode->i_size < off + len) {
5512 i_size_write(inode, off + len);
617ba13b 5513 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5514 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5515 }
67eeb568 5516 return len;
ac27a0ec
DK		 5517}
5518
8f0e8746
TT		 5519static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5520{
5521 const struct quota_format_ops *ops;
5522
5523 if (!sb_has_quota_loaded(sb, qid->type))
5524 return -ESRCH;
5525 ops = sb_dqopt(sb)->ops[qid->type];
5526 if (!ops || !ops->get_next_id)
5527 return -ENOSYS;
5528 return dquot_get_next_id(sb, qid);
5529}
ac27a0ec
DK		 5530#endif
5531
152a0836
AV		 5532static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5533 const char *dev_name, void *data)
ac27a0ec 5534{
152a0836 5535 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5536}
5537
c290ea01 5538#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5539static inline void register_as_ext2(void)
5540{
5541 int err = register_filesystem(&ext2_fs_type);
5542 if (err)
5543 printk(KERN_WARNING
5544 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5545}
5546
5547static inline void unregister_as_ext2(void)
5548{
5549 unregister_filesystem(&ext2_fs_type);
5550}
2035e776
TT		 5551
5552static inline int ext2_feature_set_ok(struct super_block *sb)
5553{
e2b911c5 5554 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776
TT		 5555 return 0;
5556 if (sb->s_flags & MS_RDONLY)
5557 return 1;
e2b911c5 5558 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5559 return 0;
5560 return 1;
5561}
24b58424
TT		 5562#else
5563static inline void register_as_ext2(void) { }
5564static inline void unregister_as_ext2(void) { }
2035e776 5565static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5566#endif
5567
24b58424
TT		 5568static inline void register_as_ext3(void)
5569{
5570 int err = register_filesystem(&ext3_fs_type);
5571 if (err)
5572 printk(KERN_WARNING
5573 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5574}
5575
5576static inline void unregister_as_ext3(void)
5577{
5578 unregister_filesystem(&ext3_fs_type);
5579}
2035e776
TT		 5580
5581static inline int ext3_feature_set_ok(struct super_block *sb)
5582{
e2b911c5 5583 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5584 return 0;
e2b911c5 5585 if (!ext4_has_feature_journal(sb))
2035e776
TT		 5586 return 0;
5587 if (sb->s_flags & MS_RDONLY)
5588 return 1;
e2b911c5 5589 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5590 return 0;
5591 return 1;
5592}
24b58424 5593
03010a33
TT		 5594static struct file_system_type ext4_fs_type = {
5595 .owner = THIS_MODULE,
5596 .name = "ext4",
152a0836 5597 .mount = ext4_mount,
03010a33
TT		 5598 .kill_sb = kill_block_super,
5599 .fs_flags = FS_REQUIRES_DEV,
5600};
7f78e035 5601MODULE_ALIAS_FS("ext4");
03010a33 5602
e9e3bcec
ES		 5603/* Shared across all ext4 file systems */
5604wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5605
5dabfc78 5606static int __init ext4_init_fs(void)
ac27a0ec 5607{
e9e3bcec 5608 int i, err;
c9de560d 5609
e294a537 5610 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5611 ext4_li_info = NULL;
5612 mutex_init(&ext4_li_mtx);
5613
9a4c8019 5614 /* Build-time check for flags consistency */
12e9b892 5615 ext4_check_flag_values();
e9e3bcec 5616
e142d052 5617 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5618 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5619
51865fda 5620 err = ext4_init_es();
6fd058f7
TT		 5621 if (err)
5622 return err;
51865fda
ZL		 5623
5624 err = ext4_init_pageio();
5625 if (err)
b5799018 5626 goto out5;
51865fda 5627
5dabfc78 5628 err = ext4_init_system_zone();
bd2d0210 5629 if (err)
b5799018 5630 goto out4;
857ac889 5631
b5799018 5632 err = ext4_init_sysfs();
dd68314c 5633 if (err)
b5799018 5634 goto out3;
857ac889 5635
5dabfc78 5636 err = ext4_init_mballoc();
c9de560d
AT		 5637 if (err)
5638 goto out2;
ac27a0ec
DK		 5639 err = init_inodecache();
5640 if (err)
5641 goto out1;
24b58424 5642 register_as_ext3();
2035e776 5643 register_as_ext2();
03010a33 5644 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5645 if (err)
5646 goto out;
bfff6873 5647
ac27a0ec
DK		 5648 return 0;
5649out:
24b58424
TT		 5650 unregister_as_ext2();
5651 unregister_as_ext3();
ac27a0ec
DK		 5652 destroy_inodecache();
5653out1:
5dabfc78 5654 ext4_exit_mballoc();
9c191f70 5655out2:
b5799018
TT		 5656 ext4_exit_sysfs();
5657out3:
5dabfc78 5658 ext4_exit_system_zone();
b5799018 5659out4:
5dabfc78 5660 ext4_exit_pageio();
b5799018 5661out5:
51865fda
ZL		 5662 ext4_exit_es();
5663
ac27a0ec
DK		 5664 return err;
5665}
5666
5dabfc78 5667static void __exit ext4_exit_fs(void)
ac27a0ec 5668{
bfff6873 5669 ext4_destroy_lazyinit_thread();
24b58424
TT		 5670 unregister_as_ext2();
5671 unregister_as_ext3();
03010a33 5672 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5673 destroy_inodecache();
5dabfc78 5674 ext4_exit_mballoc();
b5799018 5675 ext4_exit_sysfs();
5dabfc78
TT		 5676 ext4_exit_system_zone();
5677 ext4_exit_pageio();
dd12ed14 5678 ext4_exit_es();
ac27a0ec
DK		 5679}
5680
5681MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5682MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5683MODULE_LICENSE("GPL");
5dabfc78
TT		 5684module_init(ext4_init_fs)
5685module_exit(ext4_exit_fs)
Ubuntu Focal Linux kernel mirror