]> 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: fix in-superblock mount options processing
[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>
ac27a0ec
DK		 41#include <asm/uaccess.h>
42
bfff6873
LC		 43#include <linux/kthread.h>
44#include <linux/freezer.h>
45
3dcf5451 46#include "ext4.h"
4a092d73 47#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 48#include "ext4_jbd2.h"
ac27a0ec
DK		 49#include "xattr.h"
50#include "acl.h"
3661d286 51#include "mballoc.h"
ac27a0ec 52
9bffad1e
TT		 53#define CREATE_TRACE_POINTS
54#include <trace/events/ext4.h>
55
0b75a840
LC		 56static struct ext4_lazy_init *ext4_li_info;
57static struct mutex ext4_li_mtx;
e294a537 58static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 59
617ba13b 60static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 61 unsigned long journal_devnum);
2adf6da8 62static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 63static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 64static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
617ba13b 68static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 69static int ext4_remount(struct super_block *sb, int *flags, char *data);
70static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 71static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 72static int ext4_freeze(struct super_block *sb);
152a0836
AV		 73static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74 const char *dev_name, void *data);
2035e776
TT		 75static inline int ext2_feature_set_ok(struct super_block *sb);
76static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 77static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 78static void ext4_destroy_lazyinit_thread(void);
79static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 80static void ext4_clear_request_list(void);
c6cb7e77
EW		 81static struct inode *ext4_get_journal_inode(struct super_block *sb,
82 unsigned int journal_inum);
ac27a0ec 83
e74031fd
JK		 84/*
85 * Lock ordering
86 *
87 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
88 * i_mmap_rwsem (inode->i_mmap_rwsem)!
89 *
90 * page fault path:
91 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
92 * page lock -> i_data_sem (rw)
93 *
94 * buffered write path:
95 * sb_start_write -> i_mutex -> mmap_sem
96 * sb_start_write -> i_mutex -> transaction start -> page lock ->
97 * i_data_sem (rw)
98 *
99 * truncate:
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * i_mmap_rwsem (w) -> page lock
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * transaction start -> i_data_sem (rw)
104 *
105 * direct IO:
106 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
108 * transaction start -> i_data_sem (rw)
109 *
110 * writepages:
111 * transaction start -> page lock(s) -> i_data_sem (rw)
112 */
113
c290ea01 114#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 115static struct file_system_type ext2_fs_type = {
116 .owner = THIS_MODULE,
117 .name = "ext2",
118 .mount = ext4_mount,
119 .kill_sb = kill_block_super,
120 .fs_flags = FS_REQUIRES_DEV,
121};
7f78e035 122MODULE_ALIAS_FS("ext2");
fa7614dd 123MODULE_ALIAS("ext2");
2035e776
TT		 124#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
125#else
126#define IS_EXT2_SB(sb) (0)
127#endif
128
129
ba69f9ab
JK		 130static struct file_system_type ext3_fs_type = {
131 .owner = THIS_MODULE,
132 .name = "ext3",
152a0836 133 .mount = ext4_mount,
ba69f9ab
JK		 134 .kill_sb = kill_block_super,
135 .fs_flags = FS_REQUIRES_DEV,
136};
7f78e035 137MODULE_ALIAS_FS("ext3");
fa7614dd 138MODULE_ALIAS("ext3");
ba69f9ab 139#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 140
d25425f8
DW		 141static int ext4_verify_csum_type(struct super_block *sb,
142 struct ext4_super_block *es)
143{
e2b911c5 144 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 145 return 1;
146
147 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
148}
149
a9c47317
DW		 150static __le32 ext4_superblock_csum(struct super_block *sb,
151 struct ext4_super_block *es)
152{
153 struct ext4_sb_info *sbi = EXT4_SB(sb);
154 int offset = offsetof(struct ext4_super_block, s_checksum);
155 __u32 csum;
156
157 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
158
159 return cpu_to_le32(csum);
160}
161
c197855e
SH		 162static int ext4_superblock_csum_verify(struct super_block *sb,
163 struct ext4_super_block *es)
a9c47317 164{
9aa5d32b 165 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 166 return 1;
167
168 return es->s_checksum == ext4_superblock_csum(sb, es);
169}
170
06db49e6 171void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 172{
06db49e6
TT		 173 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
174
9aa5d32b 175 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 176 return;
177
178 es->s_checksum = ext4_superblock_csum(sb, es);
179}
180
9933fc0a
TT		 181void *ext4_kvmalloc(size_t size, gfp_t flags)
182{
183 void *ret;
184
8be04b93 185 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 186 if (!ret)
187 ret = __vmalloc(size, flags, PAGE_KERNEL);
188 return ret;
189}
190
191void *ext4_kvzalloc(size_t size, gfp_t flags)
192{
193 void *ret;
194
8be04b93 195 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 196 if (!ret)
197 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
198 return ret;
199}
200
8fadc143
AR		 201ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
202 struct ext4_group_desc *bg)
bd81d8ee 203{
3a14589c 204 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 206 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 207}
208
8fadc143
AR		 209ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
210 struct ext4_group_desc *bg)
bd81d8ee 211{
5272f837 212 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 214 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 215}
216
8fadc143
AR		 217ext4_fsblk_t ext4_inode_table(struct super_block *sb,
218 struct ext4_group_desc *bg)
bd81d8ee 219{
5272f837 220 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 222 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 223}
224
021b65bb
TT		 225__u32 ext4_free_group_clusters(struct super_block *sb,
226 struct ext4_group_desc *bg)
560671a0
AK		 227{
228 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 230 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 231}
232
233__u32 ext4_free_inodes_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
235{
236 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 238 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 239}
240
241__u32 ext4_used_dirs_count(struct super_block *sb,
242 struct ext4_group_desc *bg)
243{
244 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
245 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 246 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 247}
248
249__u32 ext4_itable_unused_count(struct super_block *sb,
250 struct ext4_group_desc *bg)
251{
252 return le16_to_cpu(bg->bg_itable_unused_lo) |
253 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 254 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 255}
256
8fadc143
AR		 257void ext4_block_bitmap_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 259{
3a14589c 260 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 263}
264
8fadc143
AR		 265void ext4_inode_bitmap_set(struct super_block *sb,
266 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 267{
5272f837 268 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 271}
272
8fadc143
AR		 273void ext4_inode_table_set(struct super_block *sb,
274 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 275{
5272f837 276 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 279}
280
021b65bb
TT		 281void ext4_free_group_clusters_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 283{
284 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
287}
288
289void ext4_free_inodes_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
291{
292 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
295}
296
297void ext4_used_dirs_set(struct super_block *sb,
298 struct ext4_group_desc *bg, __u32 count)
299{
300 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
301 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
302 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
303}
304
305void ext4_itable_unused_set(struct super_block *sb,
306 struct ext4_group_desc *bg, __u32 count)
307{
308 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
309 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
310 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
311}
312
d3d1faf6 313
1c13d5c0
TT		 314static void __save_error_info(struct super_block *sb, const char *func,
315 unsigned int line)
316{
317 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
318
319 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 320 if (bdev_read_only(sb->s_bdev))
321 return;
1c13d5c0
TT		 322 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
323 es->s_last_error_time = cpu_to_le32(get_seconds());
324 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
325 es->s_last_error_line = cpu_to_le32(line);
326 if (!es->s_first_error_time) {
327 es->s_first_error_time = es->s_last_error_time;
328 strncpy(es->s_first_error_func, func,
329 sizeof(es->s_first_error_func));
330 es->s_first_error_line = cpu_to_le32(line);
331 es->s_first_error_ino = es->s_last_error_ino;
332 es->s_first_error_block = es->s_last_error_block;
333 }
66e61a9e
TT		 334 /*
335 * Start the daily error reporting function if it hasn't been
336 * started already
337 */
338 if (!es->s_error_count)
339 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 340 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 341}
342
343static void save_error_info(struct super_block *sb, const char *func,
344 unsigned int line)
345{
346 __save_error_info(sb, func, line);
347 ext4_commit_super(sb, 1);
348}
349
bdfe0cbd
TT		 350/*
351 * The del_gendisk() function uninitializes the disk-specific data
352 * structures, including the bdi structure, without telling anyone
353 * else. Once this happens, any attempt to call mark_buffer_dirty()
354 * (for example, by ext4_commit_super), will cause a kernel OOPS.
355 * This is a kludge to prevent these oops until we can put in a proper
356 * hook in del_gendisk() to inform the VFS and file system layers.
357 */
358static int block_device_ejected(struct super_block *sb)
359{
360 struct inode *bd_inode = sb->s_bdev->bd_inode;
361 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
362
363 return bdi->dev == NULL;
364}
365
18aadd47
BJ		 366static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
367{
368 struct super_block *sb = journal->j_private;
369 struct ext4_sb_info *sbi = EXT4_SB(sb);
370 int error = is_journal_aborted(journal);
5d3ee208 371 struct ext4_journal_cb_entry *jce;
18aadd47 372
5d3ee208 373 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 374 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 375 while (!list_empty(&txn->t_private_list)) {
376 jce = list_entry(txn->t_private_list.next,
377 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 378 list_del_init(&jce->jce_list);
379 spin_unlock(&sbi->s_md_lock);
380 jce->jce_func(sb, jce, error);
381 spin_lock(&sbi->s_md_lock);
382 }
383 spin_unlock(&sbi->s_md_lock);
384}
1c13d5c0 385
ac27a0ec
DK		 386/* Deal with the reporting of failure conditions on a filesystem such as
387 * inconsistencies detected or read IO failures.
388 *
389 * On ext2, we can store the error state of the filesystem in the
617ba13b 390 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 391 * write ordering constraints on the superblock which prevent us from
392 * writing it out straight away; and given that the journal is about to
393 * be aborted, we can't rely on the current, or future, transactions to
394 * write out the superblock safely.
395 *
dab291af 396 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 397 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 398 * that error until we've noted it down and cleared it.
399 */
400
617ba13b 401static void ext4_handle_error(struct super_block *sb)
ac27a0ec 402{
ac27a0ec
DK		 403 if (sb->s_flags & MS_RDONLY)
404 return;
405
2b2d6d01 406 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 407 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 408
4ab2f15b 409 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 410 if (journal)
dab291af 411 jbd2_journal_abort(journal, -EIO);
ac27a0ec 412 }
2b2d6d01 413 if (test_opt(sb, ERRORS_RO)) {
b31e1552 414 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 415 /*
416 * Make sure updated value of ->s_mount_flags will be visible
417 * before ->s_flags update
418 */
419 smp_wmb();
ac27a0ec
DK		 420 sb->s_flags |= MS_RDONLY;
421 }
4327ba52
DJ		 422 if (test_opt(sb, ERRORS_PANIC)) {
423 if (EXT4_SB(sb)->s_journal &&
424 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
425 return;
617ba13b 426 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 427 sb->s_id);
4327ba52 428 }
ac27a0ec
DK		 429}
430
efbed4dc
TT		 431#define ext4_error_ratelimit(sb) \
432 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
433 "EXT4-fs error")
434
12062ddd 435void __ext4_error(struct super_block *sb, const char *function,
c398eda0 436 unsigned int line, const char *fmt, ...)
ac27a0ec 437{
0ff2ea7d 438 struct va_format vaf;
ac27a0ec
DK		 439 va_list args;
440
efbed4dc
TT		 441 if (ext4_error_ratelimit(sb)) {
442 va_start(args, fmt);
443 vaf.fmt = fmt;
444 vaf.va = &args;
445 printk(KERN_CRIT
446 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
447 sb->s_id, function, line, current->comm, &vaf);
448 va_end(args);
449 }
f3fc0210 450 save_error_info(sb, function, line);
617ba13b 451 ext4_handle_error(sb);
ac27a0ec
DK		 452}
453
e7c96e8e
JP		 454void __ext4_error_inode(struct inode *inode, const char *function,
455 unsigned int line, ext4_fsblk_t block,
456 const char *fmt, ...)
273df556
FM		 457{
458 va_list args;
f7c21177 459 struct va_format vaf;
1c13d5c0 460 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 461
1c13d5c0
TT		 462 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
463 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 464 if (ext4_error_ratelimit(inode->i_sb)) {
465 va_start(args, fmt);
466 vaf.fmt = fmt;
467 vaf.va = &args;
468 if (block)
469 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
470 "inode #%lu: block %llu: comm %s: %pV\n",
471 inode->i_sb->s_id, function, line, inode->i_ino,
472 block, current->comm, &vaf);
473 else
474 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
475 "inode #%lu: comm %s: %pV\n",
476 inode->i_sb->s_id, function, line, inode->i_ino,
477 current->comm, &vaf);
478 va_end(args);
479 }
1c13d5c0 480 save_error_info(inode->i_sb, function, line);
273df556
FM		 481 ext4_handle_error(inode->i_sb);
482}
483
e7c96e8e
JP		 484void __ext4_error_file(struct file *file, const char *function,
485 unsigned int line, ext4_fsblk_t block,
486 const char *fmt, ...)
273df556
FM		 487{
488 va_list args;
f7c21177 489 struct va_format vaf;
1c13d5c0 490 struct ext4_super_block *es;
496ad9aa 491 struct inode *inode = file_inode(file);
273df556
FM		 492 char pathname[80], *path;
493
1c13d5c0
TT		 494 es = EXT4_SB(inode->i_sb)->s_es;
495 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc 496 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 497 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 498 if (IS_ERR(path))
499 path = "(unknown)";
500 va_start(args, fmt);
501 vaf.fmt = fmt;
502 vaf.va = &args;
503 if (block)
504 printk(KERN_CRIT
505 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
506 "block %llu: comm %s: path %s: %pV\n",
507 inode->i_sb->s_id, function, line, inode->i_ino,
508 block, current->comm, path, &vaf);
509 else
510 printk(KERN_CRIT
511 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
512 "comm %s: path %s: %pV\n",
513 inode->i_sb->s_id, function, line, inode->i_ino,
514 current->comm, path, &vaf);
515 va_end(args);
516 }
1c13d5c0 517 save_error_info(inode->i_sb, function, line);
273df556
FM		 518 ext4_handle_error(inode->i_sb);
519}
520
722887dd
TT		 521const char *ext4_decode_error(struct super_block *sb, int errno,
522 char nbuf[16])
ac27a0ec
DK		 523{
524 char *errstr = NULL;
525
526 switch (errno) {
6a797d27
DW		 527 case -EFSCORRUPTED:
528 errstr = "Corrupt filesystem";
529 break;
530 case -EFSBADCRC:
531 errstr = "Filesystem failed CRC";
532 break;
ac27a0ec
DK		 533 case -EIO:
534 errstr = "IO failure";
535 break;
536 case -ENOMEM:
537 errstr = "Out of memory";
538 break;
539 case -EROFS:
78f1ddbb
TT		 540 if (!sb || (EXT4_SB(sb)->s_journal &&
541 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 542 errstr = "Journal has aborted";
543 else
544 errstr = "Readonly filesystem";
545 break;
546 default:
547 /* If the caller passed in an extra buffer for unknown
548 * errors, textualise them now. Else we just return
549 * NULL. */
550 if (nbuf) {
551 /* Check for truncated error codes... */
552 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
553 errstr = nbuf;
554 }
555 break;
556 }
557
558 return errstr;
559}
560
617ba13b 561/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 562 * automatically and invokes the appropriate error response. */
563
c398eda0
TT		 564void __ext4_std_error(struct super_block *sb, const char *function,
565 unsigned int line, int errno)
ac27a0ec
DK		 566{
567 char nbuf[16];
568 const char *errstr;
569
570 /* Special case: if the error is EROFS, and we're not already
571 * inside a transaction, then there's really no point in logging
572 * an error. */
573 if (errno == -EROFS && journal_current_handle() == NULL &&
574 (sb->s_flags & MS_RDONLY))
575 return;
576
efbed4dc
TT		 577 if (ext4_error_ratelimit(sb)) {
578 errstr = ext4_decode_error(sb, errno, nbuf);
579 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
580 sb->s_id, function, line, errstr);
581 }
ac27a0ec 582
efbed4dc 583 save_error_info(sb, function, line);
617ba13b 584 ext4_handle_error(sb);
ac27a0ec
DK		 585}
586
587/*
617ba13b 588 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 589 * abort function may be used to deal with unrecoverable failures such
590 * as journal IO errors or ENOMEM at a critical moment in log management.
591 *
592 * We unconditionally force the filesystem into an ABORT|READONLY state,
593 * unless the error response on the fs has been set to panic in which
594 * case we take the easy way out and panic immediately.
595 */
596
c67d859e 597void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 598 unsigned int line, const char *fmt, ...)
ac27a0ec 599{
651e1c3b 600 struct va_format vaf;
ac27a0ec
DK		 601 va_list args;
602
1c13d5c0 603 save_error_info(sb, function, line);
ac27a0ec 604 va_start(args, fmt);
651e1c3b
JP		 605 vaf.fmt = fmt;
606 vaf.va = &args;
607 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
608 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 609 va_end(args);
610
1c13d5c0
TT		 611 if ((sb->s_flags & MS_RDONLY) == 0) {
612 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 613 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 614 /*
615 * Make sure updated value of ->s_mount_flags will be visible
616 * before ->s_flags update
617 */
618 smp_wmb();
619 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 620 if (EXT4_SB(sb)->s_journal)
621 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
622 save_error_info(sb, function, line);
623 }
4327ba52
DJ		 624 if (test_opt(sb, ERRORS_PANIC)) {
625 if (EXT4_SB(sb)->s_journal &&
626 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
627 return;
617ba13b 628 panic("EXT4-fs panic from previous error\n");
4327ba52 629 }
ac27a0ec
DK		 630}
631
e7c96e8e
JP		 632void __ext4_msg(struct super_block *sb,
633 const char *prefix, const char *fmt, ...)
b31e1552 634{
0ff2ea7d 635 struct va_format vaf;
b31e1552
ES		 636 va_list args;
637
efbed4dc
TT		 638 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
639 return;
640
b31e1552 641 va_start(args, fmt);
0ff2ea7d
JP		 642 vaf.fmt = fmt;
643 vaf.va = &args;
644 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 645 va_end(args);
646}
647
b03a2f7e
AD		 648#define ext4_warning_ratelimit(sb) \
649 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
650 "EXT4-fs warning")
651
12062ddd 652void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 653 unsigned int line, const char *fmt, ...)
ac27a0ec 654{
0ff2ea7d 655 struct va_format vaf;
ac27a0ec
DK		 656 va_list args;
657
b03a2f7e 658 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 659 return;
660
ac27a0ec 661 va_start(args, fmt);
0ff2ea7d
JP		 662 vaf.fmt = fmt;
663 vaf.va = &args;
664 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
665 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 666 va_end(args);
667}
668
b03a2f7e
AD		 669void __ext4_warning_inode(const struct inode *inode, const char *function,
670 unsigned int line, const char *fmt, ...)
671{
672 struct va_format vaf;
673 va_list args;
674
675 if (!ext4_warning_ratelimit(inode->i_sb))
676 return;
677
678 va_start(args, fmt);
679 vaf.fmt = fmt;
680 vaf.va = &args;
681 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
682 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
683 function, line, inode->i_ino, current->comm, &vaf);
684 va_end(args);
685}
686
e29136f8
TT		 687void __ext4_grp_locked_error(const char *function, unsigned int line,
688 struct super_block *sb, ext4_group_t grp,
689 unsigned long ino, ext4_fsblk_t block,
690 const char *fmt, ...)
5d1b1b3f
AK		 691__releases(bitlock)
692__acquires(bitlock)
693{
0ff2ea7d 694 struct va_format vaf;
5d1b1b3f
AK		 695 va_list args;
696 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
697
1c13d5c0
TT		 698 es->s_last_error_ino = cpu_to_le32(ino);
699 es->s_last_error_block = cpu_to_le64(block);
700 __save_error_info(sb, function, line);
0ff2ea7d 701
efbed4dc
TT		 702 if (ext4_error_ratelimit(sb)) {
703 va_start(args, fmt);
704 vaf.fmt = fmt;
705 vaf.va = &args;
706 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
707 sb->s_id, function, line, grp);
708 if (ino)
709 printk(KERN_CONT "inode %lu: ", ino);
710 if (block)
711 printk(KERN_CONT "block %llu:",
712 (unsigned long long) block);
713 printk(KERN_CONT "%pV\n", &vaf);
714 va_end(args);
715 }
5d1b1b3f
AK		 716
717 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 718 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 719 return;
720 }
1c13d5c0 721
5d1b1b3f
AK		 722 ext4_unlock_group(sb, grp);
723 ext4_handle_error(sb);
724 /*
725 * We only get here in the ERRORS_RO case; relocking the group
726 * may be dangerous, but nothing bad will happen since the
727 * filesystem will have already been marked read/only and the
728 * journal has been aborted. We return 1 as a hint to callers
729 * who might what to use the return value from
25985edc 730 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 731 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
732 * aggressively from the ext4 function in question, with a
733 * more appropriate error code.
734 */
735 ext4_lock_group(sb, grp);
736 return;
737}
738
617ba13b 739void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 740{
617ba13b 741 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 742
617ba13b 743 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 744 return;
745
12062ddd 746 ext4_warning(sb,
ac27a0ec
DK		 747 "updating to rev %d because of new feature flag, "
748 "running e2fsck is recommended",
617ba13b 749 EXT4_DYNAMIC_REV);
ac27a0ec 750
617ba13b
MC		 751 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
752 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
753 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 754 /* leave es->s_feature_*compat flags alone */
755 /* es->s_uuid will be set by e2fsck if empty */
756
757 /*
758 * The rest of the superblock fields should be zero, and if not it
759 * means they are likely already in use, so leave them alone. We
760 * can leave it up to e2fsck to clean up any inconsistencies there.
761 */
762}
763
764/*
765 * Open the external journal device
766 */
b31e1552 767static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 768{
769 struct block_device *bdev;
770 char b[BDEVNAME_SIZE];
771
d4d77629 772 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 773 if (IS_ERR(bdev))
774 goto fail;
775 return bdev;
776
777fail:
b31e1552 778 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 779 __bdevname(dev, b), PTR_ERR(bdev));
780 return NULL;
781}
782
783/*
784 * Release the journal device
785 */
4385bab1 786static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 787{
4385bab1 788 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 789}
790
4385bab1 791static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 792{
793 struct block_device *bdev;
ac27a0ec
DK		 794 bdev = sbi->journal_bdev;
795 if (bdev) {
4385bab1 796 ext4_blkdev_put(bdev);
ac27a0ec
DK		 797 sbi->journal_bdev = NULL;
798 }
ac27a0ec
DK		 799}
800
801static inline struct inode *orphan_list_entry(struct list_head *l)
802{
617ba13b 803 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 804}
805
617ba13b 806static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 807{
808 struct list_head *l;
809
b31e1552
ES		 810 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
811 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 812
813 printk(KERN_ERR "sb_info orphan list:\n");
814 list_for_each(l, &sbi->s_orphan) {
815 struct inode *inode = orphan_list_entry(l);
816 printk(KERN_ERR " "
817 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
818 inode->i_sb->s_id, inode->i_ino, inode,
819 inode->i_mode, inode->i_nlink,
820 NEXT_ORPHAN(inode));
821 }
822}
823
2b2d6d01 824static void ext4_put_super(struct super_block *sb)
ac27a0ec 825{
617ba13b
MC		 826 struct ext4_sb_info *sbi = EXT4_SB(sb);
827 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 828 int i, err;
ac27a0ec 829
857ac889 830 ext4_unregister_li_request(sb);
e0ccfd95
CH		 831 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
832
2e8fa54e 833 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 834 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 835
0390131b
FM		 836 if (sbi->s_journal) {
837 err = jbd2_journal_destroy(sbi->s_journal);
838 sbi->s_journal = NULL;
839 if (err < 0)
c67d859e 840 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 841 }
d4edac31 842
ebd173be 843 ext4_unregister_sysfs(sb);
d3922a77 844 ext4_es_unregister_shrinker(sbi);
9105bb14 845 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 846 ext4_release_system_zone(sb);
847 ext4_mb_release(sb);
848 ext4_ext_release(sb);
d4edac31 849
ac27a0ec 850 if (!(sb->s_flags & MS_RDONLY)) {
e2b911c5 851 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 852 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 853 }
58c5873a 854 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 855 ext4_commit_super(sb, 1);
856
ac27a0ec
DK		 857 for (i = 0; i < sbi->s_gdb_count; i++)
858 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 859 kvfree(sbi->s_group_desc);
860 kvfree(sbi->s_flex_groups);
57042651 861 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 862 percpu_counter_destroy(&sbi->s_freeinodes_counter);
863 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 864 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c8585c6f 865 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
ac27a0ec
DK		 866 brelse(sbi->s_sbh);
867#ifdef CONFIG_QUOTA
a2d4a646 868 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 869 kfree(sbi->s_qf_names[i]);
870#endif
871
872 /* Debugging code just in case the in-memory inode orphan list
873 * isn't empty. The on-disk one can be non-empty if we've
874 * detected an error and taken the fs readonly, but the
875 * in-memory list had better be clean by this point. */
876 if (!list_empty(&sbi->s_orphan))
877 dump_orphan_list(sb, sbi);
878 J_ASSERT(list_empty(&sbi->s_orphan));
879
89d96a6f 880 sync_blockdev(sb->s_bdev);
f98393a6 881 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 882 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
883 /*
884 * Invalidate the journal device's buffers. We don't want them
885 * floating about in memory - the physical journal device may
886 * hotswapped, and it breaks the `ro-after' testing code.
887 */
888 sync_blockdev(sbi->journal_bdev);
f98393a6 889 invalidate_bdev(sbi->journal_bdev);
617ba13b 890 ext4_blkdev_remove(sbi);
ac27a0ec 891 }
9c191f70
M		 892 if (sbi->s_mb_cache) {
893 ext4_xattr_destroy_cache(sbi->s_mb_cache);
894 sbi->s_mb_cache = NULL;
895 }
c5e06d10
JL		 896 if (sbi->s_mmp_tsk)
897 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 898 sb->s_fs_info = NULL;
3197ebdb
TT		 899 /*
900 * Now that we are completely done shutting down the
901 * superblock, we need to actually destroy the kobject.
902 */
3197ebdb
TT		 903 kobject_put(&sbi->s_kobj);
904 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 905 if (sbi->s_chksum_driver)
906 crypto_free_shash(sbi->s_chksum_driver);
705895b6 907 kfree(sbi->s_blockgroup_lock);
ac27a0ec 908 kfree(sbi);
ac27a0ec
DK		 909}
910
e18b890b 911static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 912
913/*
914 * Called inside transaction, so use GFP_NOFS
915 */
617ba13b 916static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 917{
617ba13b 918 struct ext4_inode_info *ei;
ac27a0ec 919
e6b4f8da 920 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 921 if (!ei)
922 return NULL;
0b8e58a1 923
ac27a0ec 924 ei->vfs_inode.i_version = 1;
202ee5df 925 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 926 INIT_LIST_HEAD(&ei->i_prealloc_list);
927 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 928 ext4_es_init_tree(&ei->i_es_tree);
929 rwlock_init(&ei->i_es_lock);
edaa53ca 930 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 931 ei->i_es_all_nr = 0;
edaa53ca 932 ei->i_es_shk_nr = 0;
dd475925 933 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 934 ei->i_reserved_data_blocks = 0;
935 ei->i_reserved_meta_blocks = 0;
936 ei->i_allocated_meta_blocks = 0;
9d0be502 937 ei->i_da_metadata_calc_len = 0;
7e731bc9 938 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 939 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 940#ifdef CONFIG_QUOTA
941 ei->i_reserved_quota = 0;
96c7e0d9 942 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 943#endif
8aefcd55 944 ei->jinode = NULL;
2e8fa54e 945 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 946 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 947 ei->i_sync_tid = 0;
948 ei->i_datasync_tid = 0;
e27f41e1 949 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 950 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
ac27a0ec
DK		 951 return &ei->vfs_inode;
952}
953
7ff9c073
TT		 954static int ext4_drop_inode(struct inode *inode)
955{
956 int drop = generic_drop_inode(inode);
957
958 trace_ext4_drop_inode(inode, drop);
959 return drop;
960}
961
fa0d7e3d
NP		 962static void ext4_i_callback(struct rcu_head *head)
963{
964 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 965 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
966}
967
617ba13b 968static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 969{
9f7dd93d 970 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 971 ext4_msg(inode->i_sb, KERN_ERR,
972 "Inode %lu (%p): orphan list check failed!",
973 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 974 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
975 EXT4_I(inode), sizeof(struct ext4_inode_info),
976 true);
977 dump_stack();
978 }
fa0d7e3d 979 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 980}
981
51cc5068 982static void init_once(void *foo)
ac27a0ec 983{
617ba13b 984 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 985
a35afb83 986 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 987 init_rwsem(&ei->xattr_sem);
0e855ac8 988 init_rwsem(&ei->i_data_sem);
ea3d7209 989 init_rwsem(&ei->i_mmap_sem);
a35afb83 990 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 991}
992
e67bc2b3 993static int __init init_inodecache(void)
ac27a0ec 994{
617ba13b
MC		 995 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
996 sizeof(struct ext4_inode_info),
ac27a0ec 997 0, (SLAB_RECLAIM_ACCOUNT|
5d097056 998 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
20c2df83 999 init_once);
617ba13b 1000 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1001 return -ENOMEM;
1002 return 0;
1003}
1004
1005static void destroy_inodecache(void)
1006{
8c0a8537
KS		 1007 /*
1008 * Make sure all delayed rcu free inodes are flushed before we
1009 * destroy cache.
1010 */
1011 rcu_barrier();
617ba13b 1012 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1013}
1014
0930fcc1 1015void ext4_clear_inode(struct inode *inode)
ac27a0ec 1016{
0930fcc1 1017 invalidate_inode_buffers(inode);
dbd5768f 1018 clear_inode(inode);
9f754758 1019 dquot_drop(inode);
c2ea3fde 1020 ext4_discard_preallocations(inode);
51865fda 1021 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1022 if (EXT4_I(inode)->jinode) {
1023 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1024 EXT4_I(inode)->jinode);
1025 jbd2_free_inode(EXT4_I(inode)->jinode);
1026 EXT4_I(inode)->jinode = NULL;
1027 }
b7236e21 1028#ifdef CONFIG_EXT4_FS_ENCRYPTION
a7550b30 1029 fscrypt_put_encryption_info(inode, NULL);
b7236e21 1030#endif
ac27a0ec
DK		 1031}
1032
1b961ac0 1033static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1034 u64 ino, u32 generation)
ac27a0ec 1035{
ac27a0ec 1036 struct inode *inode;
ac27a0ec 1037
617ba13b 1038 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1039 return ERR_PTR(-ESTALE);
617ba13b 1040 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1041 return ERR_PTR(-ESTALE);
1042
1043 /* iget isn't really right if the inode is currently unallocated!!
1044 *
617ba13b 1045 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1046 * deleted, so we should be safe.
1047 *
1048 * Currently we don't know the generation for parent directory, so
1049 * a generation of 0 means "accept any"
1050 */
f4bb2981 1051 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1052 if (IS_ERR(inode))
1053 return ERR_CAST(inode);
1054 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1055 iput(inode);
1056 return ERR_PTR(-ESTALE);
1057 }
1b961ac0
CH		 1058
1059 return inode;
1060}
1061
1062static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1063 int fh_len, int fh_type)
1b961ac0
CH		 1064{
1065 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1066 ext4_nfs_get_inode);
1067}
1068
1069static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1070 int fh_len, int fh_type)
1b961ac0
CH		 1071{
1072 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1073 ext4_nfs_get_inode);
ac27a0ec
DK		 1074}
1075
c39a7f84
TO		 1076/*
1077 * Try to release metadata pages (indirect blocks, directories) which are
1078 * mapped via the block device. Since these pages could have journal heads
1079 * which would prevent try_to_free_buffers() from freeing them, we must use
1080 * jbd2 layer's try_to_free_buffers() function to release them.
1081 */
0b8e58a1
AD		 1082static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1083 gfp_t wait)
c39a7f84
TO		 1084{
1085 journal_t *journal = EXT4_SB(sb)->s_journal;
1086
1087 WARN_ON(PageChecked(page));
1088 if (!page_has_buffers(page))
1089 return 0;
1090 if (journal)
1091 return jbd2_journal_try_to_free_buffers(journal, page,
d0164adc 1092 wait & ~__GFP_DIRECT_RECLAIM);
c39a7f84
TO		 1093 return try_to_free_buffers(page);
1094}
1095
a7550b30
JK		 1096#ifdef CONFIG_EXT4_FS_ENCRYPTION
1097static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1098{
1099 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1100 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1101}
1102
1103static int ext4_key_prefix(struct inode *inode, u8 **key)
1104{
1105 *key = EXT4_SB(inode->i_sb)->key_prefix;
1106 return EXT4_SB(inode->i_sb)->key_prefix_size;
1107}
1108
1109static int ext4_prepare_context(struct inode *inode)
1110{
1111 return ext4_convert_inline_data(inode);
1112}
1113
1114static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1115 void *fs_data)
1116{
1117 handle_t *handle;
1118 int res, res2;
1119
1120 /* fs_data is null when internally used. */
1121 if (fs_data) {
1122 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1123 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx,
1124 len, 0);
1125 if (!res) {
1126 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1127 ext4_clear_inode_state(inode,
1128 EXT4_STATE_MAY_INLINE_DATA);
1129 }
1130 return res;
1131 }
1132
1133 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1134 ext4_jbd2_credits_xattr(inode));
1135 if (IS_ERR(handle))
1136 return PTR_ERR(handle);
1137
1138 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1139 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx,
1140 len, 0);
1141 if (!res) {
1142 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1143 res = ext4_mark_inode_dirty(handle, inode);
1144 if (res)
1145 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1146 }
1147 res2 = ext4_journal_stop(handle);
1148 if (!res)
1149 res = res2;
1150 return res;
1151}
1152
1153static int ext4_dummy_context(struct inode *inode)
1154{
1155 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1156}
1157
1158static unsigned ext4_max_namelen(struct inode *inode)
1159{
1160 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1161 EXT4_NAME_LEN;
1162}
1163
1164static struct fscrypt_operations ext4_cryptops = {
1165 .get_context = ext4_get_context,
1166 .key_prefix = ext4_key_prefix,
1167 .prepare_context = ext4_prepare_context,
1168 .set_context = ext4_set_context,
1169 .dummy_context = ext4_dummy_context,
1170 .is_encrypted = ext4_encrypted_inode,
1171 .empty_dir = ext4_empty_dir,
1172 .max_namelen = ext4_max_namelen,
1173};
1174#else
1175static struct fscrypt_operations ext4_cryptops = {
1176 .is_encrypted = ext4_encrypted_inode,
1177};
1178#endif
1179
ac27a0ec 1180#ifdef CONFIG_QUOTA
689c958c
LX		 1181static char *quotatypes[] = INITQFNAMES;
1182#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1183
617ba13b
MC		 1184static int ext4_write_dquot(struct dquot *dquot);
1185static int ext4_acquire_dquot(struct dquot *dquot);
1186static int ext4_release_dquot(struct dquot *dquot);
1187static int ext4_mark_dquot_dirty(struct dquot *dquot);
1188static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1189static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1190 struct path *path);
ca0e05e4 1191static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1192static int ext4_quota_on_mount(struct super_block *sb, int type);
1193static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1194 size_t len, loff_t off);
617ba13b 1195static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1196 const char *data, size_t len, loff_t off);
7c319d32
AK		 1197static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1198 unsigned int flags);
1199static int ext4_enable_quotas(struct super_block *sb);
8f0e8746 1200static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
ac27a0ec 1201
96c7e0d9
JK		 1202static struct dquot **ext4_get_dquots(struct inode *inode)
1203{
1204 return EXT4_I(inode)->i_dquot;
1205}
1206
61e225dc 1207static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1208 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1209 .write_dquot = ext4_write_dquot,
1210 .acquire_dquot = ext4_acquire_dquot,
1211 .release_dquot = ext4_release_dquot,
1212 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1213 .write_info = ext4_write_info,
1214 .alloc_dquot = dquot_alloc,
1215 .destroy_dquot = dquot_destroy,
040cb378 1216 .get_projid = ext4_get_projid,
8f0e8746 1217 .get_next_id = ext4_get_next_id,
ac27a0ec
DK		 1218};
1219
0d54b217 1220static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1221 .quota_on = ext4_quota_on,
ca0e05e4 1222 .quota_off = ext4_quota_off,
287a8095 1223 .quota_sync = dquot_quota_sync,
0a240339 1224 .get_state = dquot_get_state,
287a8095
CH		 1225 .set_info = dquot_set_dqinfo,
1226 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1227 .set_dqblk = dquot_set_dqblk,
1228 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1229};
1230#endif
1231
ee9b6d61 1232static const struct super_operations ext4_sops = {
617ba13b
MC		 1233 .alloc_inode = ext4_alloc_inode,
1234 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1235 .write_inode = ext4_write_inode,
1236 .dirty_inode = ext4_dirty_inode,
7ff9c073 1237 .drop_inode = ext4_drop_inode,
0930fcc1 1238 .evict_inode = ext4_evict_inode,
617ba13b 1239 .put_super = ext4_put_super,
617ba13b 1240 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1241 .freeze_fs = ext4_freeze,
1242 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1243 .statfs = ext4_statfs,
1244 .remount_fs = ext4_remount,
617ba13b 1245 .show_options = ext4_show_options,
ac27a0ec 1246#ifdef CONFIG_QUOTA
617ba13b
MC		 1247 .quota_read = ext4_quota_read,
1248 .quota_write = ext4_quota_write,
96c7e0d9 1249 .get_dquots = ext4_get_dquots,
ac27a0ec 1250#endif
c39a7f84 1251 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1252};
1253
39655164 1254static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1255 .fh_to_dentry = ext4_fh_to_dentry,
1256 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1257 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1258};
1259
1260enum {
1261 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1262 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1263 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1264 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1265 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1266 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1267 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1268 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1269 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1270 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1271 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1272 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
49da9392 1273 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1449032b 1274 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1275 Opt_lazytime, Opt_nolazytime,
1449032b 1276 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1277 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1278 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1279 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1280 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1281};
1282
a447c093 1283static const match_table_t tokens = {
ac27a0ec
DK		 1284 {Opt_bsd_df, "bsddf"},
1285 {Opt_minix_df, "minixdf"},
1286 {Opt_grpid, "grpid"},
1287 {Opt_grpid, "bsdgroups"},
1288 {Opt_nogrpid, "nogrpid"},
1289 {Opt_nogrpid, "sysvgroups"},
1290 {Opt_resgid, "resgid=%u"},
1291 {Opt_resuid, "resuid=%u"},
1292 {Opt_sb, "sb=%u"},
1293 {Opt_err_cont, "errors=continue"},
1294 {Opt_err_panic, "errors=panic"},
1295 {Opt_err_ro, "errors=remount-ro"},
1296 {Opt_nouid32, "nouid32"},
ac27a0ec 1297 {Opt_debug, "debug"},
72578c33
TT		 1298 {Opt_removed, "oldalloc"},
1299 {Opt_removed, "orlov"},
ac27a0ec
DK		 1300 {Opt_user_xattr, "user_xattr"},
1301 {Opt_nouser_xattr, "nouser_xattr"},
1302 {Opt_acl, "acl"},
1303 {Opt_noacl, "noacl"},
e3bb52ae 1304 {Opt_noload, "norecovery"},
5a916be1 1305 {Opt_noload, "noload"},
72578c33
TT		 1306 {Opt_removed, "nobh"},
1307 {Opt_removed, "bh"},
ac27a0ec 1308 {Opt_commit, "commit=%u"},
30773840
TT		 1309 {Opt_min_batch_time, "min_batch_time=%u"},
1310 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1311 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1312 {Opt_journal_path, "journal_path=%s"},
818d276c 1313 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1314 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1315 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1316 {Opt_abort, "abort"},
1317 {Opt_data_journal, "data=journal"},
1318 {Opt_data_ordered, "data=ordered"},
1319 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1320 {Opt_data_err_abort, "data_err=abort"},
1321 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1322 {Opt_offusrjquota, "usrjquota="},
1323 {Opt_usrjquota, "usrjquota=%s"},
1324 {Opt_offgrpjquota, "grpjquota="},
1325 {Opt_grpjquota, "grpjquota=%s"},
1326 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1327 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1328 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1329 {Opt_grpquota, "grpquota"},
1330 {Opt_noquota, "noquota"},
1331 {Opt_quota, "quota"},
1332 {Opt_usrquota, "usrquota"},
49da9392 1333 {Opt_prjquota, "prjquota"},
ac27a0ec 1334 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1335 {Opt_barrier, "barrier"},
1336 {Opt_nobarrier, "nobarrier"},
25ec56b5 1337 {Opt_i_version, "i_version"},
923ae0ff 1338 {Opt_dax, "dax"},
c9de560d 1339 {Opt_stripe, "stripe=%u"},
64769240 1340 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1341 {Opt_lazytime, "lazytime"},
1342 {Opt_nolazytime, "nolazytime"},
dd919b98 1343 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1344 {Opt_removed, "mblk_io_submit"},
1345 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1346 {Opt_block_validity, "block_validity"},
1347 {Opt_noblock_validity, "noblock_validity"},
240799cd 1348 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1349 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1350 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1351 {Opt_auto_da_alloc, "auto_da_alloc"},
1352 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1353 {Opt_dioread_nolock, "dioread_nolock"},
1354 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1355 {Opt_discard, "discard"},
1356 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1357 {Opt_init_itable, "init_itable=%u"},
1358 {Opt_init_itable, "init_itable"},
1359 {Opt_noinit_itable, "noinit_itable"},
df981d03 1360 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1361 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1362 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1363 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1364 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1365 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1366 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1367 {Opt_err, NULL},
ac27a0ec
DK		 1368};
1369
617ba13b 1370static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1371{
617ba13b 1372 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1373 char *options = (char *) *data;
1374
1375 if (!options || strncmp(options, "sb=", 3) != 0)
1376 return 1; /* Default location */
0b8e58a1 1377
ac27a0ec 1378 options += 3;
0b8e58a1 1379 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1380 sb_block = simple_strtoul(options, &options, 0);
1381 if (*options && *options != ',') {
4776004f 1382 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1383 (char *) *data);
1384 return 1;
1385 }
1386 if (*options == ',')
1387 options++;
1388 *data = (void *) options;
0b8e58a1 1389
ac27a0ec
DK		 1390 return sb_block;
1391}
1392
b3881f74 1393#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1394static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1395 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1396
56c50f11
DM		 1397#ifdef CONFIG_QUOTA
1398static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1399{
1400 struct ext4_sb_info *sbi = EXT4_SB(sb);
1401 char *qname;
03dafb5f 1402 int ret = -1;
56c50f11
DM		 1403
1404 if (sb_any_quota_loaded(sb) &&
1405 !sbi->s_qf_names[qtype]) {
1406 ext4_msg(sb, KERN_ERR,
1407 "Cannot change journaled "
1408 "quota options when quota turned on");
57f73c2c 1409 return -1;
56c50f11 1410 }
e2b911c5 1411 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1412 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1413 "ignored when QUOTA feature is enabled");
1414 return 1;
262b4662 1415 }
56c50f11
DM		 1416 qname = match_strdup(args);
1417 if (!qname) {
1418 ext4_msg(sb, KERN_ERR,
1419 "Not enough memory for storing quotafile name");
57f73c2c 1420 return -1;
56c50f11 1421 }
03dafb5f
CG		 1422 if (sbi->s_qf_names[qtype]) {
1423 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1424 ret = 1;
1425 else
1426 ext4_msg(sb, KERN_ERR,
1427 "%s quota file already specified",
1428 QTYPE2NAME(qtype));
1429 goto errout;
56c50f11 1430 }
03dafb5f 1431 if (strchr(qname, '/')) {
56c50f11
DM		 1432 ext4_msg(sb, KERN_ERR,
1433 "quotafile must be on filesystem root");
03dafb5f 1434 goto errout;
56c50f11 1435 }
03dafb5f 1436 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1437 set_opt(sb, QUOTA);
56c50f11 1438 return 1;
03dafb5f
CG		 1439errout:
1440 kfree(qname);
1441 return ret;
56c50f11
DM		 1442}
1443
1444static int clear_qf_name(struct super_block *sb, int qtype)
1445{
1446
1447 struct ext4_sb_info *sbi = EXT4_SB(sb);
1448
1449 if (sb_any_quota_loaded(sb) &&
1450 sbi->s_qf_names[qtype]) {
1451 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1452 " when quota turned on");
57f73c2c 1453 return -1;
56c50f11 1454 }
03dafb5f 1455 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1456 sbi->s_qf_names[qtype] = NULL;
1457 return 1;
1458}
1459#endif
1460
26092bf5
TT		 1461#define MOPT_SET 0x0001
1462#define MOPT_CLEAR 0x0002
1463#define MOPT_NOSUPPORT 0x0004
1464#define MOPT_EXPLICIT 0x0008
1465#define MOPT_CLEAR_ERR 0x0010
1466#define MOPT_GTE0 0x0020
ac27a0ec 1467#ifdef CONFIG_QUOTA
26092bf5
TT		 1468#define MOPT_Q 0
1469#define MOPT_QFMT 0x0040
1470#else
1471#define MOPT_Q MOPT_NOSUPPORT
1472#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1473#endif
26092bf5 1474#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1475#define MOPT_NO_EXT2 0x0100
1476#define MOPT_NO_EXT3 0x0200
1477#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1478#define MOPT_STRING 0x0400
26092bf5
TT		 1479
1480static const struct mount_opts {
1481 int token;
1482 int mount_opt;
1483 int flags;
1484} ext4_mount_opts[] = {
1485 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1486 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1487 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1488 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1489 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1490 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1491 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1492 MOPT_EXT4_ONLY | MOPT_SET},
1493 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1494 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1495 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1496 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1497 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1498 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1499 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1500 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1501 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1502 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1503 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1504 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1505 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1506 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1507 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1508 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1509 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1510 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1511 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1512 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1513 MOPT_NO_EXT2},
8dc0aa8c 1514 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1515 MOPT_NO_EXT2},
26092bf5
TT		 1516 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1517 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1518 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1519 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1520 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1521 {Opt_commit, 0, MOPT_GTE0},
1522 {Opt_max_batch_time, 0, MOPT_GTE0},
1523 {Opt_min_batch_time, 0, MOPT_GTE0},
1524 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1525 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1526 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1527 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1528 {Opt_resuid, 0, MOPT_GTE0},
1529 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1530 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1531 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1532 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1533 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1534 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1535 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1536 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1537 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1538 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1539#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1540 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1541 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1542#else
26092bf5
TT		 1543 {Opt_acl, 0, MOPT_NOSUPPORT},
1544 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1545#endif
26092bf5
TT		 1546 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1547 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1548 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1549 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1550 MOPT_SET | MOPT_Q},
1551 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1552 MOPT_SET | MOPT_Q},
49da9392
JK		 1553 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1554 MOPT_SET | MOPT_Q},
26092bf5 1555 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1556 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1557 MOPT_CLEAR | MOPT_Q},
26092bf5
TT		 1558 {Opt_usrjquota, 0, MOPT_Q},
1559 {Opt_grpjquota, 0, MOPT_Q},
1560 {Opt_offusrjquota, 0, MOPT_Q},
1561 {Opt_offgrpjquota, 0, MOPT_Q},
1562 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1563 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1564 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1565 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1566 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1567 {Opt_err, 0, 0}
1568};
1569
1570static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1571 substring_t *args, unsigned long *journal_devnum,
1572 unsigned int *journal_ioprio, int is_remount)
1573{
1574 struct ext4_sb_info *sbi = EXT4_SB(sb);
1575 const struct mount_opts *m;
08cefc7a
EB		 1576 kuid_t uid;
1577 kgid_t gid;
26092bf5
TT		 1578 int arg = 0;
1579
57f73c2c
TT		 1580#ifdef CONFIG_QUOTA
1581 if (token == Opt_usrjquota)
1582 return set_qf_name(sb, USRQUOTA, &args[0]);
1583 else if (token == Opt_grpjquota)
1584 return set_qf_name(sb, GRPQUOTA, &args[0]);
1585 else if (token == Opt_offusrjquota)
1586 return clear_qf_name(sb, USRQUOTA);
1587 else if (token == Opt_offgrpjquota)
1588 return clear_qf_name(sb, GRPQUOTA);
1589#endif
26092bf5 1590 switch (token) {
f7048605
TT		 1591 case Opt_noacl:
1592 case Opt_nouser_xattr:
1593 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1594 break;
26092bf5
TT		 1595 case Opt_sb:
1596 return 1; /* handled by get_sb_block() */
1597 case Opt_removed:
5f3633e3 1598 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1599 return 1;
26092bf5
TT		 1600 case Opt_abort:
1601 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1602 return 1;
1603 case Opt_i_version:
1604 sb->s_flags |= MS_I_VERSION;
1605 return 1;
a26f4992
TT		 1606 case Opt_lazytime:
1607 sb->s_flags |= MS_LAZYTIME;
1608 return 1;
1609 case Opt_nolazytime:
1610 sb->s_flags &= ~MS_LAZYTIME;
1611 return 1;
26092bf5
TT		 1612 }
1613
5f3633e3
JK		 1614 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1615 if (token == m->token)
1616 break;
1617
1618 if (m->token == Opt_err) {
1619 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1620 "or missing value", opt);
1621 return -1;
1622 }
1623
8dc0aa8c
TT		 1624 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1625 ext4_msg(sb, KERN_ERR,
1626 "Mount option \"%s\" incompatible with ext2", opt);
1627 return -1;
1628 }
1629 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1630 ext4_msg(sb, KERN_ERR,
1631 "Mount option \"%s\" incompatible with ext3", opt);
1632 return -1;
1633 }
1634
ad4eec61 1635 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1636 return -1;
1637 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1638 return -1;
c93cf2d7
DM		 1639 if (m->flags & MOPT_EXPLICIT) {
1640 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1641 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 1642 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1643 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 1644 } else
1645 return -1;
1646 }
5f3633e3
JK		 1647 if (m->flags & MOPT_CLEAR_ERR)
1648 clear_opt(sb, ERRORS_MASK);
1649 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1650 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1651 "options when quota turned on");
1652 return -1;
1653 }
1654
1655 if (m->flags & MOPT_NOSUPPORT) {
1656 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1657 } else if (token == Opt_commit) {
1658 if (arg == 0)
1659 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1660 sbi->s_commit_interval = HZ * arg;
1661 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1662 sbi->s_max_batch_time = arg;
1663 } else if (token == Opt_min_batch_time) {
1664 sbi->s_min_batch_time = arg;
1665 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1666 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1667 ext4_msg(sb, KERN_ERR,
1668 "EXT4-fs: inode_readahead_blks must be "
1669 "0 or a power of 2 smaller than 2^31");
26092bf5 1670 return -1;
5f3633e3
JK		 1671 }
1672 sbi->s_inode_readahead_blks = arg;
1673 } else if (token == Opt_init_itable) {
1674 set_opt(sb, INIT_INODE_TABLE);
1675 if (!args->from)
1676 arg = EXT4_DEF_LI_WAIT_MULT;
1677 sbi->s_li_wait_mult = arg;
1678 } else if (token == Opt_max_dir_size_kb) {
1679 sbi->s_max_dir_size_kb = arg;
1680 } else if (token == Opt_stripe) {
1681 sbi->s_stripe = arg;
1682 } else if (token == Opt_resuid) {
1683 uid = make_kuid(current_user_ns(), arg);
1684 if (!uid_valid(uid)) {
1685 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1686 return -1;
1687 }
5f3633e3
JK		 1688 sbi->s_resuid = uid;
1689 } else if (token == Opt_resgid) {
1690 gid = make_kgid(current_user_ns(), arg);
1691 if (!gid_valid(gid)) {
1692 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1693 return -1;
1694 }
1695 sbi->s_resgid = gid;
1696 } else if (token == Opt_journal_dev) {
1697 if (is_remount) {
1698 ext4_msg(sb, KERN_ERR,
1699 "Cannot specify journal on remount");
1700 return -1;
1701 }
1702 *journal_devnum = arg;
ad4eec61
ES		 1703 } else if (token == Opt_journal_path) {
1704 char *journal_path;
1705 struct inode *journal_inode;
1706 struct path path;
1707 int error;
1708
1709 if (is_remount) {
1710 ext4_msg(sb, KERN_ERR,
1711 "Cannot specify journal on remount");
1712 return -1;
1713 }
1714 journal_path = match_strdup(&args[0]);
1715 if (!journal_path) {
1716 ext4_msg(sb, KERN_ERR, "error: could not dup "
1717 "journal device string");
1718 return -1;
1719 }
1720
1721 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1722 if (error) {
1723 ext4_msg(sb, KERN_ERR, "error: could not find "
1724 "journal device path: error %d", error);
1725 kfree(journal_path);
1726 return -1;
1727 }
1728
2b0143b5 1729 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1730 if (!S_ISBLK(journal_inode->i_mode)) {
1731 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1732 "is not a block device", journal_path);
1733 path_put(&path);
1734 kfree(journal_path);
1735 return -1;
1736 }
1737
1738 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1739 path_put(&path);
1740 kfree(journal_path);
5f3633e3
JK		 1741 } else if (token == Opt_journal_ioprio) {
1742 if (arg > 7) {
1743 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1744 " (must be 0-7)");
1745 return -1;
1746 }
1747 *journal_ioprio =
1748 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1749 } else if (token == Opt_test_dummy_encryption) {
1750#ifdef CONFIG_EXT4_FS_ENCRYPTION
1751 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1752 ext4_msg(sb, KERN_WARNING,
1753 "Test dummy encryption mode enabled");
1754#else
1755 ext4_msg(sb, KERN_WARNING,
1756 "Test dummy encryption mount option ignored");
1757#endif
5f3633e3
JK		 1758 } else if (m->flags & MOPT_DATAJ) {
1759 if (is_remount) {
1760 if (!sbi->s_journal)
1761 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1762 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1763 ext4_msg(sb, KERN_ERR,
26092bf5 1764 "Cannot change data mode on remount");
26092bf5 1765 return -1;
ac27a0ec 1766 }
26092bf5 1767 } else {
5f3633e3
JK		 1768 clear_opt(sb, DATA_FLAGS);
1769 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1770 }
5f3633e3
JK		 1771#ifdef CONFIG_QUOTA
1772 } else if (m->flags & MOPT_QFMT) {
1773 if (sb_any_quota_loaded(sb) &&
1774 sbi->s_jquota_fmt != m->mount_opt) {
1775 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1776 "quota options when quota turned on");
1777 return -1;
1778 }
e2b911c5 1779 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1780 ext4_msg(sb, KERN_INFO,
1781 "Quota format mount options ignored "
262b4662 1782 "when QUOTA feature is enabled");
c325a67c 1783 return 1;
262b4662 1784 }
5f3633e3 1785 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 1786#endif
923ae0ff 1787 } else if (token == Opt_dax) {
ef83b6e8
DW		 1788#ifdef CONFIG_FS_DAX
1789 ext4_msg(sb, KERN_WARNING,
1790 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1791 sbi->s_mount_opt |= m->mount_opt;
1792#else
923ae0ff
RZ		 1793 ext4_msg(sb, KERN_INFO, "dax option not supported");
1794 return -1;
5f3633e3 1795#endif
7915a861
AN		 1796 } else if (token == Opt_data_err_abort) {
1797 sbi->s_mount_opt |= m->mount_opt;
1798 } else if (token == Opt_data_err_ignore) {
1799 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 1800 } else {
1801 if (!args->from)
1802 arg = 1;
1803 if (m->flags & MOPT_CLEAR)
1804 arg = !arg;
1805 else if (unlikely(!(m->flags & MOPT_SET))) {
1806 ext4_msg(sb, KERN_WARNING,
1807 "buggy handling of option %s", opt);
1808 WARN_ON(1);
1809 return -1;
1810 }
1811 if (arg != 0)
1812 sbi->s_mount_opt |= m->mount_opt;
1813 else
1814 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1815 }
5f3633e3 1816 return 1;
26092bf5
TT		 1817}
1818
1819static int parse_options(char *options, struct super_block *sb,
1820 unsigned long *journal_devnum,
1821 unsigned int *journal_ioprio,
1822 int is_remount)
1823{
1824 struct ext4_sb_info *sbi = EXT4_SB(sb);
1825 char *p;
1826 substring_t args[MAX_OPT_ARGS];
1827 int token;
1828
1829 if (!options)
1830 return 1;
1831
1832 while ((p = strsep(&options, ",")) != NULL) {
1833 if (!*p)
1834 continue;
1835 /*
1836 * Initialize args struct so we know whether arg was
1837 * found; some options take optional arguments.
1838 */
caecd0af 1839 args[0].to = args[0].from = NULL;
26092bf5
TT		 1840 token = match_token(p, tokens, args);
1841 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1842 journal_ioprio, is_remount) < 0)
1843 return 0;
ac27a0ec
DK		 1844 }
1845#ifdef CONFIG_QUOTA
49da9392
JK		 1846 /*
1847 * We do the test below only for project quotas. 'usrquota' and
1848 * 'grpquota' mount options are allowed even without quota feature
1849 * to support legacy quotas in quota files.
1850 */
1851 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1852 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1853 "Cannot enable project quota enforcement.");
1854 return 0;
1855 }
1856 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1857 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1858 clear_opt(sb, USRQUOTA);
ac27a0ec 1859
482a7425 1860 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1861 clear_opt(sb, GRPQUOTA);
ac27a0ec 1862
56c50f11 1863 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1864 ext4_msg(sb, KERN_ERR, "old and new quota "
1865 "format mixing");
ac27a0ec
DK		 1866 return 0;
1867 }
1868
1869 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1870 ext4_msg(sb, KERN_ERR, "journaled quota format "
1871 "not specified");
ac27a0ec
DK		 1872 return 0;
1873 }
ac27a0ec
DK		 1874 }
1875#endif
261cb20c
JK		 1876 if (test_opt(sb, DIOREAD_NOLOCK)) {
1877 int blocksize =
1878 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1879
09cbfeaf 1880 if (blocksize < PAGE_SIZE) {
261cb20c
JK		 1881 ext4_msg(sb, KERN_ERR, "can't mount with "
1882 "dioread_nolock if block size != PAGE_SIZE");
1883 return 0;
1884 }
1885 }
d4f76107
JK		 1886 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1887 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1888 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1889 "in data=ordered mode");
1890 return 0;
1891 }
ac27a0ec
DK		 1892 return 1;
1893}
1894
2adf6da8
TT		 1895static inline void ext4_show_quota_options(struct seq_file *seq,
1896 struct super_block *sb)
1897{
1898#if defined(CONFIG_QUOTA)
1899 struct ext4_sb_info *sbi = EXT4_SB(sb);
1900
1901 if (sbi->s_jquota_fmt) {
1902 char *fmtname = "";
1903
1904 switch (sbi->s_jquota_fmt) {
1905 case QFMT_VFS_OLD:
1906 fmtname = "vfsold";
1907 break;
1908 case QFMT_VFS_V0:
1909 fmtname = "vfsv0";
1910 break;
1911 case QFMT_VFS_V1:
1912 fmtname = "vfsv1";
1913 break;
1914 }
1915 seq_printf(seq, ",jqfmt=%s", fmtname);
1916 }
1917
1918 if (sbi->s_qf_names[USRQUOTA])
a068acf2 1919 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2adf6da8
TT		 1920
1921 if (sbi->s_qf_names[GRPQUOTA])
a068acf2 1922 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1923#endif
1924}
1925
5a916be1
TT		 1926static const char *token2str(int token)
1927{
50df9fd5 1928 const struct match_token *t;
5a916be1
TT		 1929
1930 for (t = tokens; t->token != Opt_err; t++)
1931 if (t->token == token && !strchr(t->pattern, '='))
1932 break;
1933 return t->pattern;
1934}
1935
2adf6da8
TT		 1936/*
1937 * Show an option if
1938 * - it's set to a non-default value OR
1939 * - if the per-sb default is different from the global default
1940 */
66acdcf4
TT		 1941static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1942 int nodefs)
2adf6da8 1943{
2adf6da8
TT		 1944 struct ext4_sb_info *sbi = EXT4_SB(sb);
1945 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1946 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1947 const struct mount_opts *m;
66acdcf4 1948 char sep = nodefs ? '\n' : ',';
2adf6da8 1949
66acdcf4
TT		 1950#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1951#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1952
1953 if (sbi->s_sb_block != 1)
5a916be1
TT		 1954 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1955
1956 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1957 int want_set = m->flags & MOPT_SET;
1958 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1959 (m->flags & MOPT_CLEAR_ERR))
1960 continue;
66acdcf4 1961 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1962 continue; /* skip if same as the default */
1963 if ((want_set &&
1964 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1965 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1966 continue; /* select Opt_noFoo vs Opt_Foo */
1967 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1968 }
5a916be1 1969
08cefc7a 1970 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1971 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1972 SEQ_OPTS_PRINT("resuid=%u",
1973 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1974 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1975 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1976 SEQ_OPTS_PRINT("resgid=%u",
1977 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1978 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1979 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1980 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1981 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1982 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1983 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1984 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1985 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1986 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1987 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1988 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1989 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1990 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1991 if (sb->s_flags & MS_I_VERSION)
5a916be1 1992 SEQ_OPTS_PUTS("i_version");
66acdcf4 1993 if (nodefs || sbi->s_stripe)
5a916be1 1994 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1995 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1996 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1997 SEQ_OPTS_PUTS("data=journal");
1998 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1999 SEQ_OPTS_PUTS("data=ordered");
2000 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2001 SEQ_OPTS_PUTS("data=writeback");
2002 }
66acdcf4
TT		 2003 if (nodefs ||
2004 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2005 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2006 sbi->s_inode_readahead_blks);
2adf6da8 2007
66acdcf4
TT		 2008 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2009 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2010 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2011 if (nodefs || sbi->s_max_dir_size_kb)
2012 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2013 if (test_opt(sb, DATA_ERR_ABORT))
2014 SEQ_OPTS_PUTS("data_err=abort");
2adf6da8
TT		 2015
2016 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2017 return 0;
2018}
2019
66acdcf4
TT		 2020static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2021{
2022 return _ext4_show_options(seq, root->d_sb, 0);
2023}
2024
ebd173be 2025int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2026{
2027 struct super_block *sb = seq->private;
2028 int rc;
2029
2030 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2031 rc = _ext4_show_options(seq, sb, 1);
2032 seq_puts(seq, "\n");
2033 return rc;
2034}
2035
617ba13b 2036static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2037 int read_only)
2038{
617ba13b 2039 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2040 int res = 0;
2041
617ba13b 2042 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2043 ext4_msg(sb, KERN_ERR, "revision level too high, "
2044 "forcing read-only mode");
ac27a0ec
DK		 2045 res = MS_RDONLY;
2046 }
2047 if (read_only)
281b5995 2048 goto done;
617ba13b 2049 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2050 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2051 "running e2fsck is recommended");
c8b459f4 2052 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2053 ext4_msg(sb, KERN_WARNING,
2054 "warning: mounting fs with errors, "
2055 "running e2fsck is recommended");
ed3ce80a 2056 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2057 le16_to_cpu(es->s_mnt_count) >=
2058 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2059 ext4_msg(sb, KERN_WARNING,
2060 "warning: maximal mount count reached, "
2061 "running e2fsck is recommended");
ac27a0ec
DK		 2062 else if (le32_to_cpu(es->s_checkinterval) &&
2063 (le32_to_cpu(es->s_lastcheck) +
2064 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 2065 ext4_msg(sb, KERN_WARNING,
2066 "warning: checktime reached, "
2067 "running e2fsck is recommended");
0b8e58a1 2068 if (!sbi->s_journal)
0390131b 2069 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2070 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2071 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2072 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 2073 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 2074 ext4_update_dynamic_rev(sb);
0390131b 2075 if (sbi->s_journal)
e2b911c5 2076 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2077
e2d67052 2078 ext4_commit_super(sb, 1);
281b5995 2079done:
ac27a0ec 2080 if (test_opt(sb, DEBUG))
a9df9a49 2081 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2082 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2083 sb->s_blocksize,
2084 sbi->s_groups_count,
617ba13b
MC		 2085 EXT4_BLOCKS_PER_GROUP(sb),
2086 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2087 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2088
7abc52c2 2089 cleancache_init_fs(sb);
ac27a0ec
DK		 2090 return res;
2091}
2092
117fff10
TT		 2093int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2094{
2095 struct ext4_sb_info *sbi = EXT4_SB(sb);
2096 struct flex_groups *new_groups;
2097 int size;
2098
2099 if (!sbi->s_log_groups_per_flex)
2100 return 0;
2101
2102 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2103 if (size <= sbi->s_flex_groups_allocated)
2104 return 0;
2105
2106 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2107 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2108 if (!new_groups) {
2109 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2110 size / (int) sizeof(struct flex_groups));
2111 return -ENOMEM;
2112 }
2113
2114 if (sbi->s_flex_groups) {
2115 memcpy(new_groups, sbi->s_flex_groups,
2116 (sbi->s_flex_groups_allocated *
2117 sizeof(struct flex_groups)));
b93b41d4 2118 kvfree(sbi->s_flex_groups);
117fff10
TT		 2119 }
2120 sbi->s_flex_groups = new_groups;
2121 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2122 return 0;
2123}
2124
772cb7c8
JS		 2125static int ext4_fill_flex_info(struct super_block *sb)
2126{
2127 struct ext4_sb_info *sbi = EXT4_SB(sb);
2128 struct ext4_group_desc *gdp = NULL;
772cb7c8 2129 ext4_group_t flex_group;
117fff10 2130 int i, err;
772cb7c8 2131
503358ae 2132 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2133 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2134 sbi->s_log_groups_per_flex = 0;
2135 return 1;
2136 }
2137
117fff10
TT		 2138 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2139 if (err)
9933fc0a 2140 goto failed;
772cb7c8 2141
772cb7c8 2142 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2143 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2144
2145 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2146 atomic_add(ext4_free_inodes_count(sb, gdp),
2147 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2148 atomic64_add(ext4_free_group_clusters(sb, gdp),
2149 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2150 atomic_add(ext4_used_dirs_count(sb, gdp),
2151 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2152 }
2153
2154 return 1;
2155failed:
2156 return 0;
2157}
2158
e2b911c5 2159static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2160 struct ext4_group_desc *gdp)
717d50e4 2161{
b47820ed 2162 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2163 __u16 crc = 0;
feb0ab32 2164 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2165 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2166
9aa5d32b 2167 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2168 /* Use new metadata_csum algorithm */
feb0ab32 2169 __u32 csum32;
b47820ed 2170 __u16 dummy_csum = 0;
feb0ab32 2171
feb0ab32
DW		 2172 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2173 sizeof(le_group));
b47820ed
DJ		 2174 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2175 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2176 sizeof(dummy_csum));
2177 offset += sizeof(dummy_csum);
2178 if (offset < sbi->s_desc_size)
2179 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2180 sbi->s_desc_size - offset);
feb0ab32
DW		 2181
2182 crc = csum32 & 0xFFFF;
2183 goto out;
717d50e4
AD		 2184 }
2185
feb0ab32 2186 /* old crc16 code */
e2b911c5 2187 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2188 return 0;
2189
feb0ab32
DW		 2190 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2191 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2192 crc = crc16(crc, (__u8 *)gdp, offset);
2193 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2194 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2195 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2196 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2197 crc = crc16(crc, (__u8 *)gdp + offset,
2198 le16_to_cpu(sbi->s_es->s_desc_size) -
2199 offset);
2200
2201out:
717d50e4
AD		 2202 return cpu_to_le16(crc);
2203}
2204
feb0ab32 2205int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2206 struct ext4_group_desc *gdp)
2207{
feb0ab32 2208 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2209 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2210 return 0;
2211
2212 return 1;
2213}
2214
feb0ab32
DW		 2215void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2216 struct ext4_group_desc *gdp)
2217{
2218 if (!ext4_has_group_desc_csum(sb))
2219 return;
e2b911c5 2220 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2221}
2222
ac27a0ec 2223/* Called at mount-time, super-block is locked */
bfff6873 2224static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2225 ext4_fsblk_t sb_block,
bfff6873 2226 ext4_group_t *first_not_zeroed)
ac27a0ec 2227{
617ba13b
MC		 2228 struct ext4_sb_info *sbi = EXT4_SB(sb);
2229 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2230 ext4_fsblk_t last_block;
bd81d8ee
LV		 2231 ext4_fsblk_t block_bitmap;
2232 ext4_fsblk_t inode_bitmap;
2233 ext4_fsblk_t inode_table;
ce421581 2234 int flexbg_flag = 0;
bfff6873 2235 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2236
e2b911c5 2237 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2238 flexbg_flag = 1;
2239
af5bc92d 2240 ext4_debug("Checking group descriptors");
ac27a0ec 2241
197cd65a
AM		 2242 for (i = 0; i < sbi->s_groups_count; i++) {
2243 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2244
ce421581 2245 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2246 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2247 else
2248 last_block = first_block +
617ba13b 2249 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2250
bfff6873
LC		 2251 if ((grp == sbi->s_groups_count) &&
2252 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2253 grp = i;
2254
8fadc143 2255 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2256 if (block_bitmap == sb_block) {
2257 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2258 "Block bitmap for group %u overlaps "
2259 "superblock", i);
2260 }
2b2d6d01 2261 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2262 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2263 "Block bitmap for group %u not in group "
b31e1552 2264 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2265 return 0;
2266 }
8fadc143 2267 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2268 if (inode_bitmap == sb_block) {
2269 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2270 "Inode bitmap for group %u overlaps "
2271 "superblock", i);
2272 }
2b2d6d01 2273 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2274 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2275 "Inode bitmap for group %u not in group "
b31e1552 2276 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2277 return 0;
2278 }
8fadc143 2279 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2280 if (inode_table == sb_block) {
2281 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2282 "Inode table for group %u overlaps "
2283 "superblock", i);
2284 }
bd81d8ee 2285 if (inode_table < first_block ||
2b2d6d01 2286 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2287 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2288 "Inode table for group %u not in group "
b31e1552 2289 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2290 return 0;
2291 }
955ce5f5 2292 ext4_lock_group(sb, i);
feb0ab32 2293 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2294 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2295 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2296 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2297 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2298 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2299 ext4_unlock_group(sb, i);
8a266467 2300 return 0;
7ee1ec4c 2301 }
717d50e4 2302 }
955ce5f5 2303 ext4_unlock_group(sb, i);
ce421581
JS		 2304 if (!flexbg_flag)
2305 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2306 }
bfff6873
LC		 2307 if (NULL != first_not_zeroed)
2308 *first_not_zeroed = grp;
ac27a0ec
DK		 2309 return 1;
2310}
2311
617ba13b 2312/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2313 * the superblock) which were deleted from all directories, but held open by
2314 * a process at the time of a crash. We walk the list and try to delete these
2315 * inodes at recovery time (only with a read-write filesystem).
2316 *
2317 * In order to keep the orphan inode chain consistent during traversal (in
2318 * case of crash during recovery), we link each inode into the superblock
2319 * orphan list_head and handle it the same way as an inode deletion during
2320 * normal operation (which journals the operations for us).
2321 *
2322 * We only do an iget() and an iput() on each inode, which is very safe if we
2323 * accidentally point at an in-use or already deleted inode. The worst that
2324 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2325 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2326 * e2fsck was run on this filesystem, and it must have already done the orphan
2327 * inode cleanup for us, so we can safely abort without any further action.
2328 */
2b2d6d01
TT		 2329static void ext4_orphan_cleanup(struct super_block *sb,
2330 struct ext4_super_block *es)
ac27a0ec
DK		 2331{
2332 unsigned int s_flags = sb->s_flags;
2c98eb5e 2333 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec
DK		 2334#ifdef CONFIG_QUOTA
2335 int i;
2336#endif
2337 if (!es->s_last_orphan) {
2338 jbd_debug(4, "no orphan inodes to clean up\n");
2339 return;
2340 }
2341
a8f48a95 2342 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2343 ext4_msg(sb, KERN_ERR, "write access "
2344 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2345 return;
2346 }
2347
d39195c3
AG		 2348 /* Check if feature set would not allow a r/w mount */
2349 if (!ext4_feature_set_ok(sb, 0)) {
2350 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2351 "unknown ROCOMPAT features");
2352 return;
2353 }
2354
617ba13b 2355 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2356 /* don't clear list on RO mount w/ errors */
2357 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2358 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2359 "clearing orphan list.\n");
c25f9bc6
ES		 2360 es->s_last_orphan = 0;
2361 }
ac27a0ec
DK		 2362 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2363 return;
2364 }
2365
2366 if (s_flags & MS_RDONLY) {
b31e1552 2367 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2368 sb->s_flags &= ~MS_RDONLY;
2369 }
2370#ifdef CONFIG_QUOTA
2371 /* Needed for iput() to work correctly and not trash data */
2372 sb->s_flags |= MS_ACTIVE;
2373 /* Turn on quotas so that they are updated correctly */
a2d4a646 2374 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2375 if (EXT4_SB(sb)->s_qf_names[i]) {
2376 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2377 if (ret < 0)
b31e1552
ES		 2378 ext4_msg(sb, KERN_ERR,
2379 "Cannot turn on journaled "
2380 "quota: error %d", ret);
ac27a0ec
DK		 2381 }
2382 }
2383#endif
2384
2385 while (es->s_last_orphan) {
2386 struct inode *inode;
2387
c65d5c6c
VN		 2388 /*
2389 * We may have encountered an error during cleanup; if
2390 * so, skip the rest.
2391 */
2392 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2393 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2394 es->s_last_orphan = 0;
2395 break;
2396 }
2397
97bd42b9
JB		 2398 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2399 if (IS_ERR(inode)) {
ac27a0ec
DK		 2400 es->s_last_orphan = 0;
2401 break;
2402 }
2403
617ba13b 2404 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2405 dquot_initialize(inode);
ac27a0ec 2406 if (inode->i_nlink) {
566370a2
PT		 2407 if (test_opt(sb, DEBUG))
2408 ext4_msg(sb, KERN_DEBUG,
2409 "%s: truncating inode %lu to %lld bytes",
2410 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2411 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2412 inode->i_ino, inode->i_size);
5955102c 2413 inode_lock(inode);
55f252c9 2414 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 2415 ret = ext4_truncate(inode);
2416 if (ret)
2417 ext4_std_error(inode->i_sb, ret);
5955102c 2418 inode_unlock(inode);
ac27a0ec
DK		 2419 nr_truncates++;
2420 } else {
566370a2
PT		 2421 if (test_opt(sb, DEBUG))
2422 ext4_msg(sb, KERN_DEBUG,
2423 "%s: deleting unreferenced inode %lu",
2424 __func__, inode->i_ino);
ac27a0ec
DK		 2425 jbd_debug(2, "deleting unreferenced inode %lu\n",
2426 inode->i_ino);
2427 nr_orphans++;
2428 }
2429 iput(inode); /* The delete magic happens here! */
2430 }
2431
2b2d6d01 2432#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2433
2434 if (nr_orphans)
b31e1552
ES		 2435 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2436 PLURAL(nr_orphans));
ac27a0ec 2437 if (nr_truncates)
b31e1552
ES		 2438 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2439 PLURAL(nr_truncates));
ac27a0ec
DK		 2440#ifdef CONFIG_QUOTA
2441 /* Turn quotas off */
a2d4a646 2442 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2443 if (sb_dqopt(sb)->files[i])
287a8095 2444 dquot_quota_off(sb, i);
ac27a0ec
DK		 2445 }
2446#endif
2447 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2448}
0b8e58a1 2449
cd2291a4
ES		 2450/*
2451 * Maximal extent format file size.
2452 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2453 * extent format containers, within a sector_t, and within i_blocks
2454 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2455 * so that won't be a limiting factor.
2456 *
f17722f9
LC		 2457 * However there is other limiting factor. We do store extents in the form
2458 * of starting block and length, hence the resulting length of the extent
2459 * covering maximum file size must fit into on-disk format containers as
2460 * well. Given that length is always by 1 unit bigger than max unit (because
2461 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2462 *
cd2291a4
ES		 2463 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2464 */
f287a1a5 2465static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2466{
2467 loff_t res;
2468 loff_t upper_limit = MAX_LFS_FILESIZE;
2469
2470 /* small i_blocks in vfs inode? */
f287a1a5 2471 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2472 /*
90c699a9 2473 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2474 * i_block represent total blocks in 512 bytes
2475 * 32 == size of vfs inode i_blocks * 8
2476 */
2477 upper_limit = (1LL << 32) - 1;
2478
2479 /* total blocks in file system block size */
2480 upper_limit >>= (blkbits - 9);
2481 upper_limit <<= blkbits;
2482 }
2483
f17722f9
LC		 2484 /*
2485 * 32-bit extent-start container, ee_block. We lower the maxbytes
2486 * by one fs block, so ee_len can cover the extent of maximum file
2487 * size
2488 */
2489 res = (1LL << 32) - 1;
cd2291a4 2490 res <<= blkbits;
cd2291a4
ES		 2491
2492 /* Sanity check against vm- & vfs- imposed limits */
2493 if (res > upper_limit)
2494 res = upper_limit;
2495
2496 return res;
2497}
ac27a0ec 2498
ac27a0ec 2499/*
cd2291a4 2500 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2501 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2502 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2503 */
f287a1a5 2504static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2505{
617ba13b 2506 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2507 int meta_blocks;
2508 loff_t upper_limit;
0b8e58a1
AD		 2509 /* This is calculated to be the largest file size for a dense, block
2510 * mapped file such that the file's total number of 512-byte sectors,
2511 * including data and all indirect blocks, does not exceed (2^48 - 1).
2512 *
2513 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2514 * number of 512-byte sectors of the file.
0fc1b451
AK		 2515 */
2516
f287a1a5 2517 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2518 /*
90c699a9 2519 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2520 * the inode i_block field represents total file blocks in
2521 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2522 */
2523 upper_limit = (1LL << 32) - 1;
2524
2525 /* total blocks in file system block size */
2526 upper_limit >>= (bits - 9);
2527
2528 } else {
8180a562
AK		 2529 /*
2530 * We use 48 bit ext4_inode i_blocks
2531 * With EXT4_HUGE_FILE_FL set the i_blocks
2532 * represent total number of blocks in
2533 * file system block size
2534 */
0fc1b451
AK		 2535 upper_limit = (1LL << 48) - 1;
2536
0fc1b451
AK		 2537 }
2538
2539 /* indirect blocks */
2540 meta_blocks = 1;
2541 /* double indirect blocks */
2542 meta_blocks += 1 + (1LL << (bits-2));
2543 /* tripple indirect blocks */
2544 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2545
2546 upper_limit -= meta_blocks;
2547 upper_limit <<= bits;
ac27a0ec
DK		 2548
2549 res += 1LL << (bits-2);
2550 res += 1LL << (2*(bits-2));
2551 res += 1LL << (3*(bits-2));
2552 res <<= bits;
2553 if (res > upper_limit)
2554 res = upper_limit;
0fc1b451
AK		 2555
2556 if (res > MAX_LFS_FILESIZE)
2557 res = MAX_LFS_FILESIZE;
2558
ac27a0ec
DK		 2559 return res;
2560}
2561
617ba13b 2562static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2563 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2564{
617ba13b 2565 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2566 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2567 int has_super = 0;
2568
2569 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2570
e2b911c5 2571 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 2572 return logical_sb_block + nr + 1;
ac27a0ec 2573 bg = sbi->s_desc_per_block * nr;
617ba13b 2574 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2575 has_super = 1;
0b8e58a1 2576
bd63f6b0
DW		 2577 /*
2578 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2579 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2580 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2581 * compensate.
2582 */
2583 if (sb->s_blocksize == 1024 && nr == 0 &&
2584 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2585 has_super++;
2586
617ba13b 2587 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2588}
2589
c9de560d
AT		 2590/**
2591 * ext4_get_stripe_size: Get the stripe size.
2592 * @sbi: In memory super block info
2593 *
2594 * If we have specified it via mount option, then
2595 * use the mount option value. If the value specified at mount time is
2596 * greater than the blocks per group use the super block value.
2597 * If the super block value is greater than blocks per group return 0.
2598 * Allocator needs it be less than blocks per group.
2599 *
2600 */
2601static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2602{
2603 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2604 unsigned long stripe_width =
2605 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2606 int ret;
c9de560d
AT		 2607
2608 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2609 ret = sbi->s_stripe;
2610 else if (stripe_width <= sbi->s_blocks_per_group)
2611 ret = stripe_width;
2612 else if (stride <= sbi->s_blocks_per_group)
2613 ret = stride;
2614 else
2615 ret = 0;
c9de560d 2616
3eb08658
DE		 2617 /*
2618 * If the stripe width is 1, this makes no sense and
2619 * we set it to 0 to turn off stripe handling code.
2620 */
2621 if (ret <= 1)
2622 ret = 0;
c9de560d 2623
3eb08658 2624 return ret;
c9de560d 2625}
ac27a0ec 2626
a13fb1a4
ES		 2627/*
2628 * Check whether this filesystem can be mounted based on
2629 * the features present and the RDONLY/RDWR mount requested.
2630 * Returns 1 if this filesystem can be mounted as requested,
2631 * 0 if it cannot be.
2632 */
2633static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2634{
e2b911c5 2635 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 2636 ext4_msg(sb, KERN_ERR,
2637 "Couldn't mount because of "
2638 "unsupported optional features (%x)",
2639 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2640 ~EXT4_FEATURE_INCOMPAT_SUPP));
2641 return 0;
2642 }
2643
2644 if (readonly)
2645 return 1;
2646
e2b911c5 2647 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b
DW		 2648 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2649 sb->s_flags |= MS_RDONLY;
2650 return 1;
2651 }
2652
a13fb1a4 2653 /* Check that feature set is OK for a read-write mount */
e2b911c5 2654 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 2655 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2656 "unsupported optional features (%x)",
2657 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2658 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2659 return 0;
2660 }
2661 /*
2662 * Large file size enabled file system can only be mounted
2663 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2664 */
e2b911c5 2665 if (ext4_has_feature_huge_file(sb)) {
a13fb1a4
ES		 2666 if (sizeof(blkcnt_t) < sizeof(u64)) {
2667 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2668 "cannot be mounted RDWR without "
2669 "CONFIG_LBDAF");
2670 return 0;
2671 }
2672 }
e2b911c5 2673 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 2674 ext4_msg(sb, KERN_ERR,
2675 "Can't support bigalloc feature without "
2676 "extents feature\n");
2677 return 0;
2678 }
7c319d32
AK		 2679
2680#ifndef CONFIG_QUOTA
e2b911c5 2681 if (ext4_has_feature_quota(sb) && !readonly) {
7c319d32
AK		 2682 ext4_msg(sb, KERN_ERR,
2683 "Filesystem with quota feature cannot be mounted RDWR "
2684 "without CONFIG_QUOTA");
2685 return 0;
2686 }
689c958c
LX		 2687 if (ext4_has_feature_project(sb) && !readonly) {
2688 ext4_msg(sb, KERN_ERR,
2689 "Filesystem with project quota feature cannot be mounted RDWR "
2690 "without CONFIG_QUOTA");
2691 return 0;
2692 }
7c319d32 2693#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2694 return 1;
2695}
2696
66e61a9e
TT		 2697/*
2698 * This function is called once a day if we have errors logged
2699 * on the file system
2700 */
2701static void print_daily_error_info(unsigned long arg)
2702{
2703 struct super_block *sb = (struct super_block *) arg;
2704 struct ext4_sb_info *sbi;
2705 struct ext4_super_block *es;
2706
2707 sbi = EXT4_SB(sb);
2708 es = sbi->s_es;
2709
2710 if (es->s_error_count)
ae0f78de
TT		 2711 /* fsck newer than v1.41.13 is needed to clean this condition. */
2712 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2713 le32_to_cpu(es->s_error_count));
2714 if (es->s_first_error_time) {
ae0f78de 2715 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2716 sb->s_id, le32_to_cpu(es->s_first_error_time),
2717 (int) sizeof(es->s_first_error_func),
2718 es->s_first_error_func,
2719 le32_to_cpu(es->s_first_error_line));
2720 if (es->s_first_error_ino)
651e1c3b 2721 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2722 le32_to_cpu(es->s_first_error_ino));
2723 if (es->s_first_error_block)
651e1c3b 2724 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2725 le64_to_cpu(es->s_first_error_block));
651e1c3b 2726 printk(KERN_CONT "\n");
66e61a9e
TT		 2727 }
2728 if (es->s_last_error_time) {
ae0f78de 2729 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2730 sb->s_id, le32_to_cpu(es->s_last_error_time),
2731 (int) sizeof(es->s_last_error_func),
2732 es->s_last_error_func,
2733 le32_to_cpu(es->s_last_error_line));
2734 if (es->s_last_error_ino)
651e1c3b 2735 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 2736 le32_to_cpu(es->s_last_error_ino));
2737 if (es->s_last_error_block)
651e1c3b 2738 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 2739 le64_to_cpu(es->s_last_error_block));
651e1c3b 2740 printk(KERN_CONT "\n");
66e61a9e
TT		 2741 }
2742 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2743}
2744
bfff6873
LC		 2745/* Find next suitable group and run ext4_init_inode_table */
2746static int ext4_run_li_request(struct ext4_li_request *elr)
2747{
2748 struct ext4_group_desc *gdp = NULL;
2749 ext4_group_t group, ngroups;
2750 struct super_block *sb;
2751 unsigned long timeout = 0;
2752 int ret = 0;
2753
2754 sb = elr->lr_super;
2755 ngroups = EXT4_SB(sb)->s_groups_count;
2756
2757 for (group = elr->lr_next_group; group < ngroups; group++) {
2758 gdp = ext4_get_group_desc(sb, group, NULL);
2759 if (!gdp) {
2760 ret = 1;
2761 break;
2762 }
2763
2764 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2765 break;
2766 }
2767
7f511862 2768 if (group >= ngroups)
bfff6873
LC		 2769 ret = 1;
2770
2771 if (!ret) {
2772 timeout = jiffies;
2773 ret = ext4_init_inode_table(sb, group,
2774 elr->lr_timeout ? 0 : 1);
2775 if (elr->lr_timeout == 0) {
51ce6511
LC		 2776 timeout = (jiffies - timeout) *
2777 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2778 elr->lr_timeout = timeout;
2779 }
2780 elr->lr_next_sched = jiffies + elr->lr_timeout;
2781 elr->lr_next_group = group + 1;
2782 }
bfff6873
LC		 2783 return ret;
2784}
2785
2786/*
2787 * Remove lr_request from the list_request and free the
4ed5c033 2788 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2789 */
2790static void ext4_remove_li_request(struct ext4_li_request *elr)
2791{
2792 struct ext4_sb_info *sbi;
2793
2794 if (!elr)
2795 return;
2796
2797 sbi = elr->lr_sbi;
2798
2799 list_del(&elr->lr_request);
2800 sbi->s_li_request = NULL;
2801 kfree(elr);
2802}
2803
2804static void ext4_unregister_li_request(struct super_block *sb)
2805{
1bb933fb
LC		 2806 mutex_lock(&ext4_li_mtx);
2807 if (!ext4_li_info) {
2808 mutex_unlock(&ext4_li_mtx);
bfff6873 2809 return;
1bb933fb 2810 }
bfff6873
LC		 2811
2812 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2813 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2814 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2815 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2816}
2817
8f1f7453
ES		 2818static struct task_struct *ext4_lazyinit_task;
2819
bfff6873
LC		 2820/*
2821 * This is the function where ext4lazyinit thread lives. It walks
2822 * through the request list searching for next scheduled filesystem.
2823 * When such a fs is found, run the lazy initialization request
2824 * (ext4_rn_li_request) and keep track of the time spend in this
2825 * function. Based on that time we compute next schedule time of
2826 * the request. When walking through the list is complete, compute
2827 * next waking time and put itself into sleep.
2828 */
2829static int ext4_lazyinit_thread(void *arg)
2830{
2831 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2832 struct list_head *pos, *n;
2833 struct ext4_li_request *elr;
4ed5c033 2834 unsigned long next_wakeup, cur;
bfff6873
LC		 2835
2836 BUG_ON(NULL == eli);
2837
bfff6873
LC		 2838cont_thread:
2839 while (true) {
2840 next_wakeup = MAX_JIFFY_OFFSET;
2841
2842 mutex_lock(&eli->li_list_mtx);
2843 if (list_empty(&eli->li_request_list)) {
2844 mutex_unlock(&eli->li_list_mtx);
2845 goto exit_thread;
2846 }
bfff6873 2847 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 2848 int err = 0;
2849 int progress = 0;
bfff6873
LC		 2850 elr = list_entry(pos, struct ext4_li_request,
2851 lr_request);
2852
e22834f0
DM		 2853 if (time_before(jiffies, elr->lr_next_sched)) {
2854 if (time_before(elr->lr_next_sched, next_wakeup))
2855 next_wakeup = elr->lr_next_sched;
2856 continue;
2857 }
2858 if (down_read_trylock(&elr->lr_super->s_umount)) {
2859 if (sb_start_write_trylock(elr->lr_super)) {
2860 progress = 1;
2861 /*
2862 * We hold sb->s_umount, sb can not
2863 * be removed from the list, it is
2864 * now safe to drop li_list_mtx
2865 */
2866 mutex_unlock(&eli->li_list_mtx);
2867 err = ext4_run_li_request(elr);
2868 sb_end_write(elr->lr_super);
2869 mutex_lock(&eli->li_list_mtx);
2870 n = pos->next;
b2c78cd0 2871 }
e22834f0
DM		 2872 up_read((&elr->lr_super->s_umount));
2873 }
2874 /* error, remove the lazy_init job */
2875 if (err) {
2876 ext4_remove_li_request(elr);
2877 continue;
2878 }
2879 if (!progress) {
2880 elr->lr_next_sched = jiffies +
2881 (prandom_u32()
2882 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 2883 }
bfff6873
LC		 2884 if (time_before(elr->lr_next_sched, next_wakeup))
2885 next_wakeup = elr->lr_next_sched;
2886 }
2887 mutex_unlock(&eli->li_list_mtx);
2888
a0acae0e 2889 try_to_freeze();
bfff6873 2890
4ed5c033
LC		 2891 cur = jiffies;
2892 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2893 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2894 cond_resched();
2895 continue;
2896 }
2897
4ed5c033
LC		 2898 schedule_timeout_interruptible(next_wakeup - cur);
2899
8f1f7453
ES		 2900 if (kthread_should_stop()) {
2901 ext4_clear_request_list();
2902 goto exit_thread;
2903 }
bfff6873
LC		 2904 }
2905
2906exit_thread:
2907 /*
2908 * It looks like the request list is empty, but we need
2909 * to check it under the li_list_mtx lock, to prevent any
2910 * additions into it, and of course we should lock ext4_li_mtx
2911 * to atomically free the list and ext4_li_info, because at
2912 * this point another ext4 filesystem could be registering
2913 * new one.
2914 */
2915 mutex_lock(&ext4_li_mtx);
2916 mutex_lock(&eli->li_list_mtx);
2917 if (!list_empty(&eli->li_request_list)) {
2918 mutex_unlock(&eli->li_list_mtx);
2919 mutex_unlock(&ext4_li_mtx);
2920 goto cont_thread;
2921 }
2922 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2923 kfree(ext4_li_info);
2924 ext4_li_info = NULL;
2925 mutex_unlock(&ext4_li_mtx);
2926
2927 return 0;
2928}
2929
2930static void ext4_clear_request_list(void)
2931{
2932 struct list_head *pos, *n;
2933 struct ext4_li_request *elr;
2934
2935 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2936 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2937 elr = list_entry(pos, struct ext4_li_request,
2938 lr_request);
2939 ext4_remove_li_request(elr);
2940 }
2941 mutex_unlock(&ext4_li_info->li_list_mtx);
2942}
2943
2944static int ext4_run_lazyinit_thread(void)
2945{
8f1f7453
ES		 2946 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2947 ext4_li_info, "ext4lazyinit");
2948 if (IS_ERR(ext4_lazyinit_task)) {
2949 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2950 ext4_clear_request_list();
bfff6873
LC		 2951 kfree(ext4_li_info);
2952 ext4_li_info = NULL;
92b97816 2953 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2954 "initialization thread\n",
2955 err);
2956 return err;
2957 }
2958 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2959 return 0;
2960}
2961
2962/*
2963 * Check whether it make sense to run itable init. thread or not.
2964 * If there is at least one uninitialized inode table, return
2965 * corresponding group number, else the loop goes through all
2966 * groups and return total number of groups.
2967 */
2968static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2969{
2970 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2971 struct ext4_group_desc *gdp = NULL;
2972
2973 for (group = 0; group < ngroups; group++) {
2974 gdp = ext4_get_group_desc(sb, group, NULL);
2975 if (!gdp)
2976 continue;
2977
2978 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2979 break;
2980 }
2981
2982 return group;
2983}
2984
2985static int ext4_li_info_new(void)
2986{
2987 struct ext4_lazy_init *eli = NULL;
2988
2989 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2990 if (!eli)
2991 return -ENOMEM;
2992
bfff6873
LC		 2993 INIT_LIST_HEAD(&eli->li_request_list);
2994 mutex_init(&eli->li_list_mtx);
2995
bfff6873
LC		 2996 eli->li_state |= EXT4_LAZYINIT_QUIT;
2997
2998 ext4_li_info = eli;
2999
3000 return 0;
3001}
3002
3003static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3004 ext4_group_t start)
3005{
3006 struct ext4_sb_info *sbi = EXT4_SB(sb);
3007 struct ext4_li_request *elr;
bfff6873
LC		 3008
3009 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3010 if (!elr)
3011 return NULL;
3012
3013 elr->lr_super = sb;
3014 elr->lr_sbi = sbi;
3015 elr->lr_next_group = start;
3016
3017 /*
3018 * Randomize first schedule time of the request to
3019 * spread the inode table initialization requests
3020 * better.
3021 */
dd1f723b
TT		 3022 elr->lr_next_sched = jiffies + (prandom_u32() %
3023 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3024 return elr;
3025}
3026
7f511862
TT		 3027int ext4_register_li_request(struct super_block *sb,
3028 ext4_group_t first_not_zeroed)
bfff6873
LC		 3029{
3030 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3031 struct ext4_li_request *elr = NULL;
bfff6873 3032 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3033 int ret = 0;
bfff6873 3034
7f511862 3035 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3036 if (sbi->s_li_request != NULL) {
3037 /*
3038 * Reset timeout so it can be computed again, because
3039 * s_li_wait_mult might have changed.
3040 */
3041 sbi->s_li_request->lr_timeout = 0;
7f511862 3042 goto out;
51ce6511 3043 }
bfff6873
LC		 3044
3045 if (first_not_zeroed == ngroups ||
3046 (sb->s_flags & MS_RDONLY) ||
55ff3840 3047 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3048 goto out;
bfff6873
LC		 3049
3050 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3051 if (!elr) {
3052 ret = -ENOMEM;
3053 goto out;
3054 }
bfff6873
LC		 3055
3056 if (NULL == ext4_li_info) {
3057 ret = ext4_li_info_new();
3058 if (ret)
3059 goto out;
3060 }
3061
3062 mutex_lock(&ext4_li_info->li_list_mtx);
3063 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3064 mutex_unlock(&ext4_li_info->li_list_mtx);
3065
3066 sbi->s_li_request = elr;
46e4690b
TM		 3067 /*
3068 * set elr to NULL here since it has been inserted to
3069 * the request_list and the removal and free of it is
3070 * handled by ext4_clear_request_list from now on.
3071 */
3072 elr = NULL;
bfff6873
LC		 3073
3074 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3075 ret = ext4_run_lazyinit_thread();
3076 if (ret)
3077 goto out;
3078 }
bfff6873 3079out:
beed5ecb
NK		 3080 mutex_unlock(&ext4_li_mtx);
3081 if (ret)
bfff6873 3082 kfree(elr);
bfff6873
LC		 3083 return ret;
3084}
3085
3086/*
3087 * We do not need to lock anything since this is called on
3088 * module unload.
3089 */
3090static void ext4_destroy_lazyinit_thread(void)
3091{
3092 /*
3093 * If thread exited earlier
3094 * there's nothing to be done.
3095 */
8f1f7453 3096 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3097 return;
3098
8f1f7453 3099 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3100}
3101
25ed6e8a
DW		 3102static int set_journal_csum_feature_set(struct super_block *sb)
3103{
3104 int ret = 1;
3105 int compat, incompat;
3106 struct ext4_sb_info *sbi = EXT4_SB(sb);
3107
9aa5d32b 3108 if (ext4_has_metadata_csum(sb)) {
db9ee220 3109 /* journal checksum v3 */
25ed6e8a 3110 compat = 0;
db9ee220 3111 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3112 } else {
3113 /* journal checksum v1 */
3114 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3115 incompat = 0;
3116 }
3117
feb8c6d3
DW		 3118 jbd2_journal_clear_features(sbi->s_journal,
3119 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3120 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3121 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3122 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3123 ret = jbd2_journal_set_features(sbi->s_journal,
3124 compat, 0,
3125 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3126 incompat);
3127 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3128 ret = jbd2_journal_set_features(sbi->s_journal,
3129 compat, 0,
3130 incompat);
3131 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3132 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3133 } else {
feb8c6d3
DW		 3134 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3135 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3136 }
3137
3138 return ret;
3139}
3140
952fc18e
TT		 3141/*
3142 * Note: calculating the overhead so we can be compatible with
3143 * historical BSD practice is quite difficult in the face of
3144 * clusters/bigalloc. This is because multiple metadata blocks from
3145 * different block group can end up in the same allocation cluster.
3146 * Calculating the exact overhead in the face of clustered allocation
3147 * requires either O(all block bitmaps) in memory or O(number of block
3148 * groups**2) in time. We will still calculate the superblock for
3149 * older file systems --- and if we come across with a bigalloc file
3150 * system with zero in s_overhead_clusters the estimate will be close to
3151 * correct especially for very large cluster sizes --- but for newer
3152 * file systems, it's better to calculate this figure once at mkfs
3153 * time, and store it in the superblock. If the superblock value is
3154 * present (even for non-bigalloc file systems), we will use it.
3155 */
3156static int count_overhead(struct super_block *sb, ext4_group_t grp,
3157 char *buf)
3158{
3159 struct ext4_sb_info *sbi = EXT4_SB(sb);
3160 struct ext4_group_desc *gdp;
3161 ext4_fsblk_t first_block, last_block, b;
3162 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3163 int s, j, count = 0;
3164
e2b911c5 3165 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3166 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3167 sbi->s_itb_per_group + 2);
3168
952fc18e
TT		 3169 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3170 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3171 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3172 for (i = 0; i < ngroups; i++) {
3173 gdp = ext4_get_group_desc(sb, i, NULL);
3174 b = ext4_block_bitmap(sb, gdp);
3175 if (b >= first_block && b <= last_block) {
3176 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3177 count++;
3178 }
3179 b = ext4_inode_bitmap(sb, gdp);
3180 if (b >= first_block && b <= last_block) {
3181 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3182 count++;
3183 }
3184 b = ext4_inode_table(sb, gdp);
3185 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3186 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3187 int c = EXT4_B2C(sbi, b - first_block);
3188 ext4_set_bit(c, buf);
3189 count++;
3190 }
3191 if (i != grp)
3192 continue;
3193 s = 0;
3194 if (ext4_bg_has_super(sb, grp)) {
3195 ext4_set_bit(s++, buf);
3196 count++;
3197 }
3198 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3199 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3200 count++;
3201 }
3202 }
3203 if (!count)
3204 return 0;
3205 return EXT4_CLUSTERS_PER_GROUP(sb) -
3206 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3207}
3208
3209/*
3210 * Compute the overhead and stash it in sbi->s_overhead
3211 */
3212int ext4_calculate_overhead(struct super_block *sb)
3213{
3214 struct ext4_sb_info *sbi = EXT4_SB(sb);
3215 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3216 struct inode *j_inode;
3217 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3218 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3219 ext4_fsblk_t overhead = 0;
4fdb5543 3220 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3221
952fc18e
TT		 3222 if (!buf)
3223 return -ENOMEM;
3224
3225 /*
3226 * Compute the overhead (FS structures). This is constant
3227 * for a given filesystem unless the number of block groups
3228 * changes so we cache the previous value until it does.
3229 */
3230
3231 /*
3232 * All of the blocks before first_data_block are overhead
3233 */
3234 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3235
3236 /*
3237 * Add the overhead found in each block group
3238 */
3239 for (i = 0; i < ngroups; i++) {
3240 int blks;
3241
3242 blks = count_overhead(sb, i, buf);
3243 overhead += blks;
3244 if (blks)
3245 memset(buf, 0, PAGE_SIZE);
3246 cond_resched();
3247 }
3c816ded
EW		 3248
3249 /*
3250 * Add the internal journal blocks whether the journal has been
3251 * loaded or not
3252 */
b003b524 3253 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3254 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3c816ded
EW		 3255 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3256 j_inode = ext4_get_journal_inode(sb, j_inum);
3257 if (j_inode) {
3258 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3259 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3260 iput(j_inode);
3261 } else {
3262 ext4_msg(sb, KERN_ERR, "can't get journal size");
3263 }
3264 }
952fc18e
TT		 3265 sbi->s_overhead = overhead;
3266 smp_wmb();
3267 free_page((unsigned long) buf);
3268 return 0;
3269}
3270
b5799018 3271static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3272{
3273 ext4_fsblk_t resv_clusters;
b5799018 3274 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3275
30fac0f7
JK		 3276 /*
3277 * There's no need to reserve anything when we aren't using extents.
3278 * The space estimates are exact, there are no unwritten extents,
3279 * hole punching doesn't need new metadata... This is needed especially
3280 * to keep ext2/3 backward compatibility.
3281 */
e2b911c5 3282 if (!ext4_has_feature_extents(sb))
b5799018 3283 return;
27dd4385
LC		 3284 /*
3285 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3286 * This should cover the situations where we can not afford to run
3287 * out of space like for example punch hole, or converting
556615dc 3288 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3289 * allocation would require 1, or 2 blocks, higher numbers are
3290 * very rare.
3291 */
b5799018
TT		 3292 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3293 sbi->s_cluster_bits);
27dd4385
LC		 3294
3295 do_div(resv_clusters, 50);
3296 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3297
b5799018 3298 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3299}
3300
2b2d6d01 3301static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3302{
d4c402d9 3303 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3304 struct buffer_head *bh;
617ba13b 3305 struct ext4_super_block *es = NULL;
5aee0f8a 3306 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
617ba13b
MC		 3307 ext4_fsblk_t block;
3308 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3309 ext4_fsblk_t logical_sb_block;
ac27a0ec 3310 unsigned long offset = 0;
ac27a0ec
DK		 3311 unsigned long journal_devnum = 0;
3312 unsigned long def_mount_opts;
3313 struct inode *root;
0390131b 3314 const char *descr;
dcc7dae3 3315 int ret = -ENOMEM;
281b5995 3316 int blocksize, clustersize;
4ec11028
TT		 3317 unsigned int db_count;
3318 unsigned int i;
281b5995 3319 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3320 __u64 blocks_count;
07aa2ea1 3321 int err = 0;
b3881f74 3322 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3323 ext4_group_t first_not_zeroed;
ac27a0ec 3324
5aee0f8a
TT		 3325 if ((data && !orig_data) || !sbi)
3326 goto out_free_base;
705895b6
PE		 3327
3328 sbi->s_blockgroup_lock =
3329 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 3330 if (!sbi->s_blockgroup_lock)
3331 goto out_free_base;
3332
ac27a0ec 3333 sb->s_fs_info = sbi;
2c0544b2 3334 sbi->s_sb = sb;
240799cd 3335 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3336 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3337 if (sb->s_bdev->bd_part)
3338 sbi->s_sectors_written_start =
3339 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3340
9f6200bb 3341 /* Cleanup superblock name */
ec3904dc 3342 strreplace(sb->s_id, '/', '!');
9f6200bb 3343
07aa2ea1 3344 /* -EINVAL is default */
dcc7dae3 3345 ret = -EINVAL;
617ba13b 3346 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3347 if (!blocksize) {
b31e1552 3348 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3349 goto out_fail;
3350 }
3351
3352 /*
617ba13b 3353 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3354 * block sizes. We need to calculate the offset from buffer start.
3355 */
617ba13b 3356 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3357 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3358 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3359 } else {
70bbb3e0 3360 logical_sb_block = sb_block;
ac27a0ec
DK		 3361 }
3362
a8ac900b 3363 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3364 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3365 goto out_fail;
3366 }
3367 /*
3368 * Note: s_es must be initialized as soon as possible because
617ba13b 3369 * some ext4 macro-instructions depend on its value
ac27a0ec 3370 */
2716b802 3371 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3372 sbi->s_es = es;
3373 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3374 if (sb->s_magic != EXT4_SUPER_MAGIC)
3375 goto cantfind_ext4;
afc32f7e 3376 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3377
feb0ab32 3378 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 3379 if (ext4_has_feature_metadata_csum(sb) &&
3380 ext4_has_feature_gdt_csum(sb))
363307e6 3381 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3382 "redundant flags; please run fsck.");
3383
d25425f8
DW		 3384 /* Check for a known checksum algorithm */
3385 if (!ext4_verify_csum_type(sb, es)) {
3386 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3387 "unknown checksum algorithm.");
3388 silent = 1;
3389 goto cantfind_ext4;
3390 }
3391
0441984a 3392 /* Load the checksum driver */
e2b911c5 3393 if (ext4_has_feature_metadata_csum(sb)) {
0441984a
DW		 3394 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3395 if (IS_ERR(sbi->s_chksum_driver)) {
3396 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3397 ret = PTR_ERR(sbi->s_chksum_driver);
3398 sbi->s_chksum_driver = NULL;
3399 goto failed_mount;
3400 }
3401 }
3402
a9c47317
DW		 3403 /* Check superblock checksum */
3404 if (!ext4_superblock_csum_verify(sb, es)) {
3405 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3406 "invalid superblock checksum. Run e2fsck?");
3407 silent = 1;
6a797d27 3408 ret = -EFSBADCRC;
a9c47317
DW		 3409 goto cantfind_ext4;
3410 }
3411
3412 /* Precompute checksum seed for all metadata */
e2b911c5 3413 if (ext4_has_feature_csum_seed(sb))
8c81bd8f
DW		 3414 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3415 else if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3416 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3417 sizeof(es->s_uuid));
3418
ac27a0ec
DK		 3419 /* Set defaults before we parse the mount options */
3420 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3421 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3422 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3423 set_opt(sb, DEBUG);
87f26807 3424 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3425 set_opt(sb, GRPID);
617ba13b 3426 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3427 set_opt(sb, NO_UID32);
ea663336 3428 /* xattr user namespace & acls are now defaulted on */
ea663336 3429 set_opt(sb, XATTR_USER);
03010a33 3430#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3431 set_opt(sb, POSIX_ACL);
2e7842b8 3432#endif
98c1a759
DW		 3433 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3434 if (ext4_has_metadata_csum(sb))
3435 set_opt(sb, JOURNAL_CHECKSUM);
3436
617ba13b 3437 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3438 set_opt(sb, JOURNAL_DATA);
617ba13b 3439 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3440 set_opt(sb, ORDERED_DATA);
617ba13b 3441 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3442 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3443
3444 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3445 set_opt(sb, ERRORS_PANIC);
bb4f397a 3446 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3447 set_opt(sb, ERRORS_CONT);
bb4f397a 3448 else
fd8c37ec 3449 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3450 /* block_validity enabled by default; disable with noblock_validity */
3451 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3452 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3453 set_opt(sb, DISCARD);
ac27a0ec 3454
08cefc7a
EB		 3455 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3456 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3457 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3458 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3459 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3460
8b67f04a 3461 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3462 set_opt(sb, BARRIER);
ac27a0ec 3463
dd919b98
AK		 3464 /*
3465 * enable delayed allocation by default
3466 * Use -o nodelalloc to turn it off
3467 */
bc0b75f7 3468 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3469 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3470 set_opt(sb, DELALLOC);
dd919b98 3471
51ce6511
LC		 3472 /*
3473 * set default s_li_wait_mult for lazyinit, for the case there is
3474 * no mount option specified.
3475 */
3476 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3477
5aee0f8a
TT		 3478 if (sbi->s_es->s_mount_opts[0]) {
3479 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3480 sizeof(sbi->s_es->s_mount_opts),
3481 GFP_KERNEL);
3482 if (!s_mount_opts)
3483 goto failed_mount;
3484 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3485 &journal_ioprio, 0)) {
3486 ext4_msg(sb, KERN_WARNING,
3487 "failed to parse options in superblock: %s",
3488 s_mount_opts);
3489 }
3490 kfree(s_mount_opts);
8b67f04a 3491 }
5a916be1 3492 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3493 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3494 &journal_ioprio, 0))
ac27a0ec
DK		 3495 goto failed_mount;
3496
56889787
TT		 3497 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3498 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3499 "with data=journal disables delayed "
3500 "allocation and O_DIRECT support!\n");
3501 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3502 ext4_msg(sb, KERN_ERR, "can't mount with "
3503 "both data=journal and delalloc");
3504 goto failed_mount;
3505 }
3506 if (test_opt(sb, DIOREAD_NOLOCK)) {
3507 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3508 "both data=journal and dioread_nolock");
56889787
TT		 3509 goto failed_mount;
3510 }
923ae0ff
RZ		 3511 if (test_opt(sb, DAX)) {
3512 ext4_msg(sb, KERN_ERR, "can't mount with "
3513 "both data=journal and dax");
3514 goto failed_mount;
3515 }
56889787
TT		 3516 if (test_opt(sb, DELALLOC))
3517 clear_opt(sb, DELALLOC);
001e4a87
TH		 3518 } else {
3519 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 3520 }
3521
ac27a0ec 3522 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3523 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3524
617ba13b 3525 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 3526 (ext4_has_compat_features(sb) ||
3527 ext4_has_ro_compat_features(sb) ||
3528 ext4_has_incompat_features(sb)))
b31e1552
ES		 3529 ext4_msg(sb, KERN_WARNING,
3530 "feature flags set on rev 0 fs, "
3531 "running e2fsck is recommended");
469108ff 3532
ed3654eb
TT		 3533 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3534 set_opt2(sb, HURD_COMPAT);
e2b911c5 3535 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 3536 ext4_msg(sb, KERN_ERR,
3537 "The Hurd can't support 64-bit file systems");
3538 goto failed_mount;
3539 }
3540 }
3541
2035e776
TT		 3542 if (IS_EXT2_SB(sb)) {
3543 if (ext2_feature_set_ok(sb))
3544 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3545 "using the ext4 subsystem");
3546 else {
3547 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3548 "to feature incompatibilities");
3549 goto failed_mount;
3550 }
3551 }
3552
3553 if (IS_EXT3_SB(sb)) {
3554 if (ext3_feature_set_ok(sb))
3555 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3556 "using the ext4 subsystem");
3557 else {
3558 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3559 "to feature incompatibilities");
3560 goto failed_mount;
3561 }
3562 }
3563
ac27a0ec
DK		 3564 /*
3565 * Check feature flags regardless of the revision level, since we
3566 * previously didn't change the revision level when setting the flags,
3567 * so there is a chance incompat flags are set on a rev 0 filesystem.
3568 */
a13fb1a4 3569 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3570 goto failed_mount;
a13fb1a4 3571
261cb20c 3572 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3573 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3574 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552 3575 ext4_msg(sb, KERN_ERR,
9e47a4c9
TT		 3576 "Unsupported filesystem blocksize %d (%d log_block_size)",
3577 blocksize, le32_to_cpu(es->s_log_block_size));
3578 goto failed_mount;
3579 }
3580 if (le32_to_cpu(es->s_log_block_size) >
3581 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3582 ext4_msg(sb, KERN_ERR,
3583 "Invalid log block size: %u",
3584 le32_to_cpu(es->s_log_block_size));
ac27a0ec
DK		 3585 goto failed_mount;
3586 }
3587
5b9554dc
TT		 3588 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3589 ext4_msg(sb, KERN_ERR,
3590 "Number of reserved GDT blocks insanely large: %d",
3591 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3592 goto failed_mount;
3593 }
3594
923ae0ff 3595 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
87eefeb4
TK		 3596 err = bdev_dax_supported(sb, blocksize);
3597 if (err)
923ae0ff 3598 goto failed_mount;
923ae0ff
RZ		 3599 }
3600
e2b911c5 3601 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 3602 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3603 es->s_encryption_level);
3604 goto failed_mount;
3605 }
3606
ac27a0ec 3607 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3608 /* Validate the filesystem blocksize */
3609 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3610 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3611 blocksize);
ac27a0ec
DK		 3612 goto failed_mount;
3613 }
3614
2b2d6d01 3615 brelse(bh);
70bbb3e0
AM		 3616 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3617 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3618 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3619 if (!bh) {
b31e1552
ES		 3620 ext4_msg(sb, KERN_ERR,
3621 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3622 goto failed_mount;
3623 }
2716b802 3624 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3625 sbi->s_es = es;
617ba13b 3626 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3627 ext4_msg(sb, KERN_ERR,
3628 "Magic mismatch, very weird!");
ac27a0ec
DK		 3629 goto failed_mount;
3630 }
3631 }
3632
e2b911c5 3633 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 3634 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3635 has_huge_files);
3636 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3637
617ba13b
MC		 3638 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3639 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3640 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3641 } else {
3642 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3643 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3644 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3645 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3646 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3647 ext4_msg(sb, KERN_ERR,
3648 "unsupported inode size: %d",
2b2d6d01 3649 sbi->s_inode_size);
ac27a0ec
DK		 3650 goto failed_mount;
3651 }
ef7f3835
KS		 3652 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3653 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3654 }
0b8e58a1 3655
0d1ee42f 3656 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 3657 if (ext4_has_feature_64bit(sb)) {
8fadc143 3658 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3659 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3660 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3661 ext4_msg(sb, KERN_ERR,
3662 "unsupported descriptor size %lu",
0d1ee42f
AR		 3663 sbi->s_desc_size);
3664 goto failed_mount;
3665 }
3666 } else
3667 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3668
ac27a0ec 3669 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3670 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3671 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3672 goto cantfind_ext4;
0b8e58a1 3673
617ba13b 3674 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3675 if (sbi->s_inodes_per_block == 0)
617ba13b 3676 goto cantfind_ext4;
ac27a0ec
DK		 3677 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3678 sbi->s_inodes_per_block;
0d1ee42f 3679 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3680 sbi->s_sbh = bh;
3681 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3682 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3683 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3684
2b2d6d01 3685 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3686 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3687 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 3688 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 3689 i = le32_to_cpu(es->s_flags);
3690 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3691 sbi->s_hash_unsigned = 3;
3692 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3693#ifdef __CHAR_UNSIGNED__
23301410
TT		 3694 if (!(sb->s_flags & MS_RDONLY))
3695 es->s_flags |=
3696 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3697 sbi->s_hash_unsigned = 3;
f99b2589 3698#else
23301410
TT		 3699 if (!(sb->s_flags & MS_RDONLY))
3700 es->s_flags |=
3701 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3702#endif
23301410 3703 }
f99b2589 3704 }
ac27a0ec 3705
281b5995
TT		 3706 /* Handle clustersize */
3707 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
e2b911c5 3708 has_bigalloc = ext4_has_feature_bigalloc(sb);
281b5995
TT		 3709 if (has_bigalloc) {
3710 if (clustersize < blocksize) {
3711 ext4_msg(sb, KERN_ERR,
3712 "cluster size (%d) smaller than "
3713 "block size (%d)", clustersize, blocksize);
3714 goto failed_mount;
3715 }
9e47a4c9
TT		 3716 if (le32_to_cpu(es->s_log_cluster_size) >
3717 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3718 ext4_msg(sb, KERN_ERR,
3719 "Invalid log cluster size: %u",
3720 le32_to_cpu(es->s_log_cluster_size));
3721 goto failed_mount;
3722 }
281b5995
TT		 3723 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3724 le32_to_cpu(es->s_log_block_size);
3725 sbi->s_clusters_per_group =
3726 le32_to_cpu(es->s_clusters_per_group);
3727 if (sbi->s_clusters_per_group > blocksize * 8) {
3728 ext4_msg(sb, KERN_ERR,
3729 "#clusters per group too big: %lu",
3730 sbi->s_clusters_per_group);
3731 goto failed_mount;
3732 }
3733 if (sbi->s_blocks_per_group !=
3734 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3735 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3736 "clusters per group (%lu) inconsistent",
3737 sbi->s_blocks_per_group,
3738 sbi->s_clusters_per_group);
3739 goto failed_mount;
3740 }
3741 } else {
3742 if (clustersize != blocksize) {
3743 ext4_warning(sb, "fragment/cluster size (%d) != "
3744 "block size (%d)", clustersize,
3745 blocksize);
3746 clustersize = blocksize;
3747 }
3748 if (sbi->s_blocks_per_group > blocksize * 8) {
3749 ext4_msg(sb, KERN_ERR,
3750 "#blocks per group too big: %lu",
3751 sbi->s_blocks_per_group);
3752 goto failed_mount;
3753 }
3754 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3755 sbi->s_cluster_bits = 0;
ac27a0ec 3756 }
281b5995
TT		 3757 sbi->s_cluster_ratio = clustersize / blocksize;
3758
ac27a0ec 3759 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3760 ext4_msg(sb, KERN_ERR,
3761 "#inodes per group too big: %lu",
2b2d6d01 3762 sbi->s_inodes_per_group);
ac27a0ec
DK		 3763 goto failed_mount;
3764 }
3765
960fd856
TT		 3766 /* Do we have standard group size of clustersize * 8 blocks ? */
3767 if (sbi->s_blocks_per_group == clustersize << 3)
3768 set_opt2(sb, STD_GROUP_SIZE);
3769
bf43d84b
ES		 3770 /*
3771 * Test whether we have more sectors than will fit in sector_t,
3772 * and whether the max offset is addressable by the page cache.
3773 */
5a9ae68a 3774 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3775 ext4_blocks_count(es));
5a9ae68a 3776 if (err) {
b31e1552 3777 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3778 " too large to mount safely on this system");
ac27a0ec 3779 if (sizeof(sector_t) < 8)
90c699a9 3780 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3781 goto failed_mount;
3782 }
3783
617ba13b
MC		 3784 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3785 goto cantfind_ext4;
e7c95593 3786
0f2ddca6
FTN		 3787 /* check blocks count against device size */
3788 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3789 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3790 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3791 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3792 ext4_blocks_count(es), blocks_count);
3793 goto failed_mount;
3794 }
3795
0b8e58a1
AD		 3796 /*
3797 * It makes no sense for the first data block to be beyond the end
3798 * of the filesystem.
3799 */
3800 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3801 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3802 "block %u is beyond end of filesystem (%llu)",
3803 le32_to_cpu(es->s_first_data_block),
3804 ext4_blocks_count(es));
e7c95593
ES		 3805 goto failed_mount;
3806 }
bd81d8ee
LV		 3807 blocks_count = (ext4_blocks_count(es) -
3808 le32_to_cpu(es->s_first_data_block) +
3809 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3810 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3811 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3812 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3813 "(block count %llu, first data block %u, "
b31e1552 3814 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3815 ext4_blocks_count(es),
3816 le32_to_cpu(es->s_first_data_block),
3817 EXT4_BLOCKS_PER_GROUP(sb));
3818 goto failed_mount;
3819 }
bd81d8ee 3820 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3821 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3822 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3823 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3824 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3825 sbi->s_group_desc = ext4_kvmalloc(db_count *
3826 sizeof(struct buffer_head *),
3827 GFP_KERNEL);
ac27a0ec 3828 if (sbi->s_group_desc == NULL) {
b31e1552 3829 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3830 ret = -ENOMEM;
ac27a0ec
DK		 3831 goto failed_mount;
3832 }
3833
705895b6 3834 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3835
3836 for (i = 0; i < db_count; i++) {
70bbb3e0 3837 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3838 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3839 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3840 ext4_msg(sb, KERN_ERR,
3841 "can't read group descriptor %d", i);
ac27a0ec
DK		 3842 db_count = i;
3843 goto failed_mount2;
3844 }
3845 }
829fa70d 3846 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 3847 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 3848 ret = -EFSCORRUPTED;
f9ae9cf5 3849 goto failed_mount2;
ac27a0ec 3850 }
772cb7c8 3851
f9ae9cf5 3852 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3853 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3854 spin_lock_init(&sbi->s_next_gen_lock);
3855
04ecddb7
JM		 3856 setup_timer(&sbi->s_err_report, print_daily_error_info,
3857 (unsigned long) sb);
04496411 3858
a75ae78f 3859 /* Register extent status tree shrinker */
eb68d0e2 3860 if (ext4_es_register_shrinker(sbi))
ce7e010a 3861 goto failed_mount3;
ce7e010a 3862
c9de560d 3863 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3864 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3865
f9ae9cf5
TT		 3866 /*
3867 * set up enough so that it can read an inode
3868 */
f6e63f90 3869 sb->s_op = &ext4_sops;
617ba13b
MC		 3870 sb->s_export_op = &ext4_export_ops;
3871 sb->s_xattr = ext4_xattr_handlers;
a7550b30 3872 sb->s_cop = &ext4_cryptops;
ac27a0ec 3873#ifdef CONFIG_QUOTA
617ba13b 3874 sb->dq_op = &ext4_quota_operations;
e2b911c5 3875 if (ext4_has_feature_quota(sb))
1fa5efe3 3876 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3877 else
3878 sb->s_qcop = &ext4_qctl_operations;
689c958c 3879 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 3880#endif
f2fa2ffc
AK		 3881 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3882
ac27a0ec 3883 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3884 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3885
3886 sb->s_root = NULL;
3887
3888 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 3889 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 3890
e2b911c5 3891 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
c5e06d10 3892 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 3893 goto failed_mount3a;
c5e06d10 3894
ac27a0ec
DK		 3895 /*
3896 * The first inode we look at is the journal inode. Don't try
3897 * root first: it may be modified in the journal!
3898 */
e2b911c5 3899 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
617ba13b 3900 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 3901 goto failed_mount3a;
0390131b 3902 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
e2b911c5 3903 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 3904 ext4_msg(sb, KERN_ERR, "required journal recovery "
3905 "suppressed and not mounted read-only");
744692dc 3906 goto failed_mount_wq;
ac27a0ec 3907 } else {
1e381f60
DM		 3908 /* Nojournal mode, all journal mount options are illegal */
3909 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3910 ext4_msg(sb, KERN_ERR, "can't mount with "
3911 "journal_checksum, fs mounted w/o journal");
3912 goto failed_mount_wq;
3913 }
3914 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3915 ext4_msg(sb, KERN_ERR, "can't mount with "
3916 "journal_async_commit, fs mounted w/o journal");
3917 goto failed_mount_wq;
3918 }
3919 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3920 ext4_msg(sb, KERN_ERR, "can't mount with "
3921 "commit=%lu, fs mounted w/o journal",
3922 sbi->s_commit_interval / HZ);
3923 goto failed_mount_wq;
3924 }
3925 if (EXT4_MOUNT_DATA_FLAGS &
3926 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3927 ext4_msg(sb, KERN_ERR, "can't mount with "
3928 "data=, fs mounted w/o journal");
3929 goto failed_mount_wq;
3930 }
3931 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3932 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 3933 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3934 sbi->s_journal = NULL;
3935 needs_recovery = 0;
3936 goto no_journal;
ac27a0ec
DK		 3937 }
3938
e2b911c5 3939 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 3940 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3941 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3942 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3943 goto failed_mount_wq;
eb40a09c
JS		 3944 }
3945
25ed6e8a
DW		 3946 if (!set_journal_csum_feature_set(sb)) {
3947 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3948 "feature set");
3949 goto failed_mount_wq;
d4da6c9c 3950 }
818d276c 3951
ac27a0ec
DK		 3952 /* We have now updated the journal if required, so we can
3953 * validate the data journaling mode. */
3954 switch (test_opt(sb, DATA_FLAGS)) {
3955 case 0:
3956 /* No mode set, assume a default based on the journal
63f57933
AM		 3957 * capabilities: ORDERED_DATA if the journal can
3958 * cope, else JOURNAL_DATA
3959 */
dab291af
MC		 3960 if (jbd2_journal_check_available_features
3961 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3962 set_opt(sb, ORDERED_DATA);
ac27a0ec 3963 else
fd8c37ec 3964 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3965 break;
3966
617ba13b
MC		 3967 case EXT4_MOUNT_ORDERED_DATA:
3968 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3969 if (!jbd2_journal_check_available_features
3970 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3971 ext4_msg(sb, KERN_ERR, "Journal does not support "
3972 "requested data journaling mode");
744692dc 3973 goto failed_mount_wq;
ac27a0ec
DK		 3974 }
3975 default:
3976 break;
3977 }
b3881f74 3978 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3979
18aadd47
BJ		 3980 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3981
ce7e010a 3982no_journal:
2335d05f
AG		 3983 sbi->s_mb_cache = ext4_xattr_create_cache();
3984 if (!sbi->s_mb_cache) {
3985 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3986 goto failed_mount_wq;
9c191f70
M		 3987 }
3988
e2b911c5 3989 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
09cbfeaf 3990 (blocksize != PAGE_SIZE)) {
1cb767cd
TT		 3991 ext4_msg(sb, KERN_ERR,
3992 "Unsupported blocksize for fs encryption");
3993 goto failed_mount_wq;
3994 }
3995
e2b911c5
DW		 3996 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3997 !ext4_has_feature_encrypt(sb)) {
3998 ext4_set_feature_encrypt(sb);
6ddb2447
TT		 3999 ext4_commit_super(sb, 1);
4000 }
4001
952fc18e
TT		 4002 /*
4003 * Get the # of file system overhead blocks from the
4004 * superblock if present.
4005 */
4006 if (es->s_overhead_clusters)
4007 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4008 else {
07aa2ea1
LC		 4009 err = ext4_calculate_overhead(sb);
4010 if (err)
952fc18e
TT		 4011 goto failed_mount_wq;
4012 }
4013
fd89d5f2
TH		 4014 /*
4015 * The maximum number of concurrent works can be high and
4016 * concurrency isn't really necessary. Limit it to 1.
4017 */
2e8fa54e
JK		 4018 EXT4_SB(sb)->rsv_conversion_wq =
4019 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4020 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4021 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4022 ret = -ENOMEM;
2e8fa54e
JK		 4023 goto failed_mount4;
4024 }
4025
ac27a0ec 4026 /*
dab291af 4027 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4028 * so we can safely mount the rest of the filesystem now.
4029 */
4030
1d1fe1ee
DH		 4031 root = ext4_iget(sb, EXT4_ROOT_INO);
4032 if (IS_ERR(root)) {
b31e1552 4033 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4034 ret = PTR_ERR(root);
32a9bb57 4035 root = NULL;
ac27a0ec
DK		 4036 goto failed_mount4;
4037 }
4038 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4039 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4040 iput(root);
ac27a0ec
DK		 4041 goto failed_mount4;
4042 }
48fde701 4043 sb->s_root = d_make_root(root);
1d1fe1ee 4044 if (!sb->s_root) {
b31e1552 4045 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4046 ret = -ENOMEM;
4047 goto failed_mount4;
4048 }
ac27a0ec 4049
7e84b621
ES		 4050 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4051 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4052
4053 /* determine the minimum size of new large inodes, if present */
4054 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4055 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4056 EXT4_GOOD_OLD_INODE_SIZE;
e2b911c5 4057 if (ext4_has_feature_extra_isize(sb)) {
ef7f3835
KS		 4058 if (sbi->s_want_extra_isize <
4059 le16_to_cpu(es->s_want_extra_isize))
4060 sbi->s_want_extra_isize =
4061 le16_to_cpu(es->s_want_extra_isize);
4062 if (sbi->s_want_extra_isize <
4063 le16_to_cpu(es->s_min_extra_isize))
4064 sbi->s_want_extra_isize =
4065 le16_to_cpu(es->s_min_extra_isize);
4066 }
4067 }
4068 /* Check if enough inode space is available */
4069 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4070 sbi->s_inode_size) {
4071 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4072 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4073 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4074 "available");
ef7f3835
KS		 4075 }
4076
b5799018 4077 ext4_set_resv_clusters(sb);
27dd4385 4078
6fd058f7
TT		 4079 err = ext4_setup_system_zone(sb);
4080 if (err) {
b31e1552 4081 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4082 "zone (%d)", err);
f9ae9cf5
TT		 4083 goto failed_mount4a;
4084 }
4085
4086 ext4_ext_init(sb);
4087 err = ext4_mb_init(sb);
4088 if (err) {
4089 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4090 err);
dcf2d804 4091 goto failed_mount5;
c2774d84
AK		 4092 }
4093
d5e03cbb
TT		 4094 block = ext4_count_free_clusters(sb);
4095 ext4_free_blocks_count_set(sbi->s_es,
4096 EXT4_C2B(sbi, block));
908c7f19
TH		 4097 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4098 GFP_KERNEL);
d5e03cbb
TT		 4099 if (!err) {
4100 unsigned long freei = ext4_count_free_inodes(sb);
4101 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4102 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4103 GFP_KERNEL);
d5e03cbb
TT		 4104 }
4105 if (!err)
4106 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4107 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4108 if (!err)
908c7f19
TH		 4109 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4110 GFP_KERNEL);
c8585c6f
DJ		 4111 if (!err)
4112 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4113
d5e03cbb
TT		 4114 if (err) {
4115 ext4_msg(sb, KERN_ERR, "insufficient memory");
4116 goto failed_mount6;
4117 }
4118
e2b911c5 4119 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 4120 if (!ext4_fill_flex_info(sb)) {
4121 ext4_msg(sb, KERN_ERR,
4122 "unable to initialize "
4123 "flex_bg meta info!");
4124 goto failed_mount6;
4125 }
4126
bfff6873
LC		 4127 err = ext4_register_li_request(sb, first_not_zeroed);
4128 if (err)
dcf2d804 4129 goto failed_mount6;
bfff6873 4130
b5799018 4131 err = ext4_register_sysfs(sb);
dcf2d804
TM		 4132 if (err)
4133 goto failed_mount7;
3197ebdb 4134
9b2ff357
JK		 4135#ifdef CONFIG_QUOTA
4136 /* Enable quota usage during mount. */
e2b911c5 4137 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
9b2ff357
JK		 4138 err = ext4_enable_quotas(sb);
4139 if (err)
4140 goto failed_mount8;
4141 }
4142#endif /* CONFIG_QUOTA */
4143
617ba13b
MC		 4144 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4145 ext4_orphan_cleanup(sb, es);
4146 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4147 if (needs_recovery) {
b31e1552 4148 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4149 ext4_mark_recovery_complete(sb, es);
4150 }
4151 if (EXT4_SB(sb)->s_journal) {
4152 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4153 descr = " journalled data mode";
4154 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4155 descr = " ordered data mode";
4156 else
4157 descr = " writeback data mode";
4158 } else
4159 descr = "out journal";
4160
79add3a3
LC		 4161 if (test_opt(sb, DISCARD)) {
4162 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4163 if (!blk_queue_discard(q))
4164 ext4_msg(sb, KERN_WARNING,
4165 "mounting with \"discard\" option, but "
4166 "the device does not support discard");
4167 }
4168
e294a537
TT		 4169 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4170 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5aee0f8a
TT		 4171 "Opts: %.*s%s%s", descr,
4172 (int) sizeof(sbi->s_es->s_mount_opts),
4173 sbi->s_es->s_mount_opts,
e294a537 4174 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4175
66e61a9e
TT		 4176 if (es->s_error_count)
4177 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4178
efbed4dc
TT		 4179 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4180 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4181 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4182 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4183
d4c402d9 4184 kfree(orig_data);
a7550b30
JK		 4185#ifdef CONFIG_EXT4_FS_ENCRYPTION
4186 memcpy(sbi->key_prefix, EXT4_KEY_DESC_PREFIX,
4187 EXT4_KEY_DESC_PREFIX_SIZE);
4188 sbi->key_prefix_size = EXT4_KEY_DESC_PREFIX_SIZE;
4189#endif
ac27a0ec
DK		 4190 return 0;
4191
617ba13b 4192cantfind_ext4:
ac27a0ec 4193 if (!silent)
b31e1552 4194 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4195 goto failed_mount;
4196
72ba7450
TT		 4197#ifdef CONFIG_QUOTA
4198failed_mount8:
ebd173be 4199 ext4_unregister_sysfs(sb);
72ba7450 4200#endif
dcf2d804
TM		 4201failed_mount7:
4202 ext4_unregister_li_request(sb);
4203failed_mount6:
f9ae9cf5 4204 ext4_mb_release(sb);
d5e03cbb 4205 if (sbi->s_flex_groups)
b93b41d4 4206 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4207 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4208 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4209 percpu_counter_destroy(&sbi->s_dirs_counter);
4210 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4211failed_mount5:
f9ae9cf5
TT		 4212 ext4_ext_release(sb);
4213 ext4_release_system_zone(sb);
4214failed_mount4a:
94bf608a 4215 dput(sb->s_root);
32a9bb57 4216 sb->s_root = NULL;
94bf608a 4217failed_mount4:
b31e1552 4218 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4219 if (EXT4_SB(sb)->rsv_conversion_wq)
4220 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4221failed_mount_wq:
82939d79
JK		 4222 if (sbi->s_mb_cache) {
4223 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4224 sbi->s_mb_cache = NULL;
4225 }
0390131b
FM		 4226 if (sbi->s_journal) {
4227 jbd2_journal_destroy(sbi->s_journal);
4228 sbi->s_journal = NULL;
4229 }
50460fe8 4230failed_mount3a:
d3922a77 4231 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4232failed_mount3:
9105bb14 4233 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4234 if (sbi->s_mmp_tsk)
4235 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4236failed_mount2:
4237 for (i = 0; i < db_count; i++)
4238 brelse(sbi->s_group_desc[i]);
b93b41d4 4239 kvfree(sbi->s_group_desc);
ac27a0ec 4240failed_mount:
0441984a
DW		 4241 if (sbi->s_chksum_driver)
4242 crypto_free_shash(sbi->s_chksum_driver);
ac27a0ec 4243#ifdef CONFIG_QUOTA
a2d4a646 4244 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4245 kfree(sbi->s_qf_names[i]);
4246#endif
617ba13b 4247 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4248 brelse(bh);
4249out_fail:
4250 sb->s_fs_info = NULL;
f6830165 4251 kfree(sbi->s_blockgroup_lock);
5aee0f8a 4252out_free_base:
ac27a0ec 4253 kfree(sbi);
d4c402d9 4254 kfree(orig_data);
07aa2ea1 4255 return err ? err : ret;
ac27a0ec
DK		 4256}
4257
4258/*
4259 * Setup any per-fs journal parameters now. We'll do this both on
4260 * initial mount, once the journal has been initialised but before we've
4261 * done any recovery; and again on any subsequent remount.
4262 */
617ba13b 4263static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4264{
617ba13b 4265 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4266
30773840
TT		 4267 journal->j_commit_interval = sbi->s_commit_interval;
4268 journal->j_min_batch_time = sbi->s_min_batch_time;
4269 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4270
a931da6a 4271 write_lock(&journal->j_state_lock);
ac27a0ec 4272 if (test_opt(sb, BARRIER))
dab291af 4273 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4274 else
dab291af 4275 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4276 if (test_opt(sb, DATA_ERR_ABORT))
4277 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4278 else
4279 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4280 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4281}
4282
c6cb7e77
EW		 4283static struct inode *ext4_get_journal_inode(struct super_block *sb,
4284 unsigned int journal_inum)
ac27a0ec
DK		 4285{
4286 struct inode *journal_inode;
ac27a0ec 4287
c6cb7e77
EW		 4288 /*
4289 * Test for the existence of a valid inode on disk. Bad things
4290 * happen if we iget() an unused inode, as the subsequent iput()
4291 * will try to delete it.
4292 */
1d1fe1ee
DH		 4293 journal_inode = ext4_iget(sb, journal_inum);
4294 if (IS_ERR(journal_inode)) {
b31e1552 4295 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4296 return NULL;
4297 }
4298 if (!journal_inode->i_nlink) {
4299 make_bad_inode(journal_inode);
4300 iput(journal_inode);
b31e1552 4301 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4302 return NULL;
4303 }
4304
e5f8eab8 4305 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4306 journal_inode, journal_inode->i_size);
1d1fe1ee 4307 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4308 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4309 iput(journal_inode);
4310 return NULL;
4311 }
c6cb7e77
EW		 4312 return journal_inode;
4313}
4314
4315static journal_t *ext4_get_journal(struct super_block *sb,
4316 unsigned int journal_inum)
4317{
4318 struct inode *journal_inode;
4319 journal_t *journal;
4320
4321 BUG_ON(!ext4_has_feature_journal(sb));
4322
4323 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4324 if (!journal_inode)
4325 return NULL;
ac27a0ec 4326
dab291af 4327 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4328 if (!journal) {
b31e1552 4329 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4330 iput(journal_inode);
4331 return NULL;
4332 }
4333 journal->j_private = sb;
617ba13b 4334 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4335 return journal;
4336}
4337
617ba13b 4338static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4339 dev_t j_dev)
4340{
2b2d6d01 4341 struct buffer_head *bh;
ac27a0ec 4342 journal_t *journal;
617ba13b
MC		 4343 ext4_fsblk_t start;
4344 ext4_fsblk_t len;
ac27a0ec 4345 int hblock, blocksize;
617ba13b 4346 ext4_fsblk_t sb_block;
ac27a0ec 4347 unsigned long offset;
2b2d6d01 4348 struct ext4_super_block *es;
ac27a0ec
DK		 4349 struct block_device *bdev;
4350
e2b911c5 4351 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4352
b31e1552 4353 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4354 if (bdev == NULL)
4355 return NULL;
4356
ac27a0ec 4357 blocksize = sb->s_blocksize;
e1defc4f 4358 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4359 if (blocksize < hblock) {
b31e1552
ES		 4360 ext4_msg(sb, KERN_ERR,
4361 "blocksize too small for journal device");
ac27a0ec
DK		 4362 goto out_bdev;
4363 }
4364
617ba13b
MC		 4365 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4366 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4367 set_blocksize(bdev, blocksize);
4368 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4369 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4370 "external journal");
ac27a0ec
DK		 4371 goto out_bdev;
4372 }
4373
2716b802 4374 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4375 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4376 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4377 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4378 ext4_msg(sb, KERN_ERR, "external journal has "
4379 "bad superblock");
ac27a0ec
DK		 4380 brelse(bh);
4381 goto out_bdev;
4382 }
4383
df4763be
DW		 4384 if ((le32_to_cpu(es->s_feature_ro_compat) &
4385 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4386 es->s_checksum != ext4_superblock_csum(sb, es)) {
4387 ext4_msg(sb, KERN_ERR, "external journal has "
4388 "corrupt superblock");
4389 brelse(bh);
4390 goto out_bdev;
4391 }
4392
617ba13b 4393 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4394 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4395 brelse(bh);
4396 goto out_bdev;
4397 }
4398
bd81d8ee 4399 len = ext4_blocks_count(es);
ac27a0ec
DK		 4400 start = sb_block + 1;
4401 brelse(bh); /* we're done with the superblock */
4402
dab291af 4403 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4404 start, len, blocksize);
4405 if (!journal) {
b31e1552 4406 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4407 goto out_bdev;
4408 }
4409 journal->j_private = sb;
dfec8a14 4410 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4411 wait_on_buffer(journal->j_sb_buffer);
4412 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4413 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4414 goto out_journal;
4415 }
4416 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4417 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4418 "user (unsupported) - %d",
ac27a0ec
DK		 4419 be32_to_cpu(journal->j_superblock->s_nr_users));
4420 goto out_journal;
4421 }
617ba13b
MC		 4422 EXT4_SB(sb)->journal_bdev = bdev;
4423 ext4_init_journal_params(sb, journal);
ac27a0ec 4424 return journal;
0b8e58a1 4425
ac27a0ec 4426out_journal:
dab291af 4427 jbd2_journal_destroy(journal);
ac27a0ec 4428out_bdev:
617ba13b 4429 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4430 return NULL;
4431}
4432
617ba13b
MC		 4433static int ext4_load_journal(struct super_block *sb,
4434 struct ext4_super_block *es,
ac27a0ec
DK		 4435 unsigned long journal_devnum)
4436{
4437 journal_t *journal;
4438 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4439 dev_t journal_dev;
4440 int err = 0;
4441 int really_read_only;
4442
e2b911c5 4443 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4444
ac27a0ec
DK		 4445 if (journal_devnum &&
4446 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4447 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4448 "numbers have changed");
ac27a0ec
DK		 4449 journal_dev = new_decode_dev(journal_devnum);
4450 } else
4451 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4452
4453 really_read_only = bdev_read_only(sb->s_bdev);
4454
4455 /*
4456 * Are we loading a blank journal or performing recovery after a
4457 * crash? For recovery, we need to check in advance whether we
4458 * can get read-write access to the device.
4459 */
e2b911c5 4460 if (ext4_has_feature_journal_needs_recovery(sb)) {
ac27a0ec 4461 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4462 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4463 "required on readonly filesystem");
ac27a0ec 4464 if (really_read_only) {
b31e1552
ES		 4465 ext4_msg(sb, KERN_ERR, "write access "
4466 "unavailable, cannot proceed");
ac27a0ec
DK		 4467 return -EROFS;
4468 }
b31e1552
ES		 4469 ext4_msg(sb, KERN_INFO, "write access will "
4470 "be enabled during recovery");
ac27a0ec
DK		 4471 }
4472 }
4473
4474 if (journal_inum && journal_dev) {
b31e1552
ES		 4475 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4476 "and inode journals!");
ac27a0ec
DK		 4477 return -EINVAL;
4478 }
4479
4480 if (journal_inum) {
617ba13b 4481 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4482 return -EINVAL;
4483 } else {
617ba13b 4484 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4485 return -EINVAL;
4486 }
4487
90576c0b 4488 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4489 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4490
e2b911c5 4491 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 4492 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4493 if (!err) {
4494 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4495 if (save)
4496 memcpy(save, ((char *) es) +
4497 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4498 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4499 if (save)
4500 memcpy(((char *) es) + EXT4_S_ERR_START,
4501 save, EXT4_S_ERR_LEN);
4502 kfree(save);
4503 }
ac27a0ec
DK		 4504
4505 if (err) {
b31e1552 4506 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4507 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4508 return err;
4509 }
4510
617ba13b
MC		 4511 EXT4_SB(sb)->s_journal = journal;
4512 ext4_clear_journal_err(sb, es);
ac27a0ec 4513
c41303ce 4514 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4515 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4516 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4517
4518 /* Make sure we flush the recovery flag to disk. */
e2d67052 4519 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4520 }
4521
4522 return 0;
4523}
4524
e2d67052 4525static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4526{
e2d67052 4527 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4528 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4529 int error = 0;
ac27a0ec 4530
bdfe0cbd 4531 if (!sbh || block_device_ejected(sb))
c4be0c1d 4532 return error;
71290b36
TT		 4533 /*
4534 * If the file system is mounted read-only, don't update the
4535 * superblock write time. This avoids updating the superblock
4536 * write time when we are mounting the root file system
4537 * read/only but we need to replay the journal; at that point,
4538 * for people who are east of GMT and who make their clock
4539 * tick in localtime for Windows bug-for-bug compatibility,
4540 * the clock is set in the future, and this will cause e2fsck
4541 * to complain and force a full file system check.
4542 */
4543 if (!(sb->s_flags & MS_RDONLY))
4544 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4545 if (sb->s_bdev->bd_part)
4546 es->s_kbytes_written =
4547 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4548 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4549 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4550 else
4551 es->s_kbytes_written =
4552 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4553 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4554 ext4_free_blocks_count_set(es,
57042651
TT		 4555 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4556 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4557 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4558 es->s_free_inodes_count =
4559 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4560 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4561 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4562 ext4_superblock_csum_set(sb);
1566a48a
TT		 4563 if (sync)
4564 lock_buffer(sbh);
4743f839
PKS		 4565 if (buffer_write_io_error(sbh)) {
4566 /*
4567 * Oh, dear. A previous attempt to write the
4568 * superblock failed. This could happen because the
4569 * USB device was yanked out. Or it could happen to
4570 * be a transient write error and maybe the block will
4571 * be remapped. Nothing we can do but to retry the
4572 * write and hope for the best.
4573 */
4574 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4575 "superblock detected");
4576 clear_buffer_write_io_error(sbh);
4577 set_buffer_uptodate(sbh);
4578 }
ac27a0ec 4579 mark_buffer_dirty(sbh);
914258bf 4580 if (sync) {
1566a48a 4581 unlock_buffer(sbh);
564bc402
DJ		 4582 error = __sync_dirty_buffer(sbh,
4583 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
c4be0c1d
TS		 4584 if (error)
4585 return error;
4586
4587 error = buffer_write_io_error(sbh);
4588 if (error) {
b31e1552
ES		 4589 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4590 "superblock");
914258bf
TT		 4591 clear_buffer_write_io_error(sbh);
4592 set_buffer_uptodate(sbh);
4593 }
4594 }
c4be0c1d 4595 return error;
ac27a0ec
DK		 4596}
4597
ac27a0ec
DK		 4598/*
4599 * Have we just finished recovery? If so, and if we are mounting (or
4600 * remounting) the filesystem readonly, then we will end up with a
4601 * consistent fs on disk. Record that fact.
4602 */
2b2d6d01
TT		 4603static void ext4_mark_recovery_complete(struct super_block *sb,
4604 struct ext4_super_block *es)
ac27a0ec 4605{
617ba13b 4606 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4607
e2b911c5 4608 if (!ext4_has_feature_journal(sb)) {
0390131b
FM		 4609 BUG_ON(journal != NULL);
4610 return;
4611 }
dab291af 4612 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4613 if (jbd2_journal_flush(journal) < 0)
4614 goto out;
4615
e2b911c5 4616 if (ext4_has_feature_journal_needs_recovery(sb) &&
ac27a0ec 4617 sb->s_flags & MS_RDONLY) {
e2b911c5 4618 ext4_clear_feature_journal_needs_recovery(sb);
e2d67052 4619 ext4_commit_super(sb, 1);
ac27a0ec 4620 }
7ffe1ea8
HK		 4621
4622out:
dab291af 4623 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4624}
4625
4626/*
4627 * If we are mounting (or read-write remounting) a filesystem whose journal
4628 * has recorded an error from a previous lifetime, move that error to the
4629 * main filesystem now.
4630 */
2b2d6d01
TT		 4631static void ext4_clear_journal_err(struct super_block *sb,
4632 struct ext4_super_block *es)
ac27a0ec
DK		 4633{
4634 journal_t *journal;
4635 int j_errno;
4636 const char *errstr;
4637
e2b911c5 4638 BUG_ON(!ext4_has_feature_journal(sb));
0390131b 4639
617ba13b 4640 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4641
4642 /*
4643 * Now check for any error status which may have been recorded in the
617ba13b 4644 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4645 */
4646
dab291af 4647 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4648 if (j_errno) {
4649 char nbuf[16];
4650
617ba13b 4651 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4652 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4653 "from previous mount: %s", errstr);
12062ddd 4654 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4655
617ba13b
MC		 4656 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4657 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4658 ext4_commit_super(sb, 1);
ac27a0ec 4659
dab291af 4660 jbd2_journal_clear_err(journal);
d796c52e 4661 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4662 }
4663}
4664
4665/*
4666 * Force the running and committing transactions to commit,
4667 * and wait on the commit.
4668 */
617ba13b 4669int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4670{
4671 journal_t *journal;
ac27a0ec
DK		 4672
4673 if (sb->s_flags & MS_RDONLY)
4674 return 0;
4675
617ba13b 4676 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4677 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4678}
4679
617ba13b 4680static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4681{
14ce0cb4 4682 int ret = 0;
9eddacf9 4683 tid_t target;
06a407f1 4684 bool needs_barrier = false;
8d5d02e6 4685 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4686
9bffad1e 4687 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4688 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4689 /*
4690 * Writeback quota in non-journalled quota case - journalled quota has
4691 * no dirty dquots
4692 */
4693 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4694 /*
4695 * Data writeback is possible w/o journal transaction, so barrier must
4696 * being sent at the end of the function. But we can skip it if
4697 * transaction_commit will do it for us.
4698 */
bda32530
TT		 4699 if (sbi->s_journal) {
4700 target = jbd2_get_latest_transaction(sbi->s_journal);
4701 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4702 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4703 needs_barrier = true;
4704
4705 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4706 if (wait)
4707 ret = jbd2_log_wait_commit(sbi->s_journal,
4708 target);
4709 }
4710 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4711 needs_barrier = true;
06a407f1
DM		 4712 if (needs_barrier) {
4713 int err;
4714 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4715 if (!ret)
4716 ret = err;
0390131b 4717 }
06a407f1
DM		 4718
4719 return ret;
4720}
4721
ac27a0ec
DK		 4722/*
4723 * LVM calls this function before a (read-only) snapshot is created. This
4724 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4725 *
4726 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4727 * state independently. It relies on upper layer to stop all data & metadata
4728 * modifications.
ac27a0ec 4729 */
c4be0c1d 4730static int ext4_freeze(struct super_block *sb)
ac27a0ec 4731{
c4be0c1d
TS		 4732 int error = 0;
4733 journal_t *journal;
ac27a0ec 4734
9ca92389
TT		 4735 if (sb->s_flags & MS_RDONLY)
4736 return 0;
ac27a0ec 4737
9ca92389 4738 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4739
bb044576
TT		 4740 if (journal) {
4741 /* Now we set up the journal barrier. */
4742 jbd2_journal_lock_updates(journal);
ac27a0ec 4743
bb044576
TT		 4744 /*
4745 * Don't clear the needs_recovery flag if we failed to
4746 * flush the journal.
4747 */
4748 error = jbd2_journal_flush(journal);
4749 if (error < 0)
4750 goto out;
c642dc9e
ES		 4751
4752 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 4753 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 4754 }
9ca92389 4755
9ca92389 4756 error = ext4_commit_super(sb, 1);
6b0310fb 4757out:
bb044576
TT		 4758 if (journal)
4759 /* we rely on upper layer to stop further updates */
4760 jbd2_journal_unlock_updates(journal);
6b0310fb 4761 return error;
ac27a0ec
DK		 4762}
4763
4764/*
4765 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4766 * flag here, even though the filesystem is not technically dirty yet.
4767 */
c4be0c1d 4768static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4769{
9ca92389
TT		 4770 if (sb->s_flags & MS_RDONLY)
4771 return 0;
4772
c642dc9e
ES		 4773 if (EXT4_SB(sb)->s_journal) {
4774 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 4775 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 4776 }
4777
9ca92389 4778 ext4_commit_super(sb, 1);
c4be0c1d 4779 return 0;
ac27a0ec
DK		 4780}
4781
673c6100
TT		 4782/*
4783 * Structure to save mount options for ext4_remount's benefit
4784 */
4785struct ext4_mount_options {
4786 unsigned long s_mount_opt;
a2595b8a 4787 unsigned long s_mount_opt2;
08cefc7a
EB		 4788 kuid_t s_resuid;
4789 kgid_t s_resgid;
673c6100
TT		 4790 unsigned long s_commit_interval;
4791 u32 s_min_batch_time, s_max_batch_time;
4792#ifdef CONFIG_QUOTA
4793 int s_jquota_fmt;
a2d4a646 4794 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4795#endif
4796};
4797
2b2d6d01 4798static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4799{
2b2d6d01 4800 struct ext4_super_block *es;
617ba13b 4801 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4802 unsigned long old_sb_flags;
617ba13b 4803 struct ext4_mount_options old_opts;
c79d967d 4804 int enable_quota = 0;
8a266467 4805 ext4_group_t g;
b3881f74 4806 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4807 int err = 0;
ac27a0ec 4808#ifdef CONFIG_QUOTA
03dafb5f 4809 int i, j;
ac27a0ec 4810#endif
d4c402d9 4811 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4812
4813 /* Store the original options */
4814 old_sb_flags = sb->s_flags;
4815 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4816 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4817 old_opts.s_resuid = sbi->s_resuid;
4818 old_opts.s_resgid = sbi->s_resgid;
4819 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4820 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4821 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4822#ifdef CONFIG_QUOTA
4823 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4824 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4825 if (sbi->s_qf_names[i]) {
4826 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4827 GFP_KERNEL);
4828 if (!old_opts.s_qf_names[i]) {
4829 for (j = 0; j < i; j++)
4830 kfree(old_opts.s_qf_names[j]);
3e36a163 4831 kfree(orig_data);
03dafb5f
CG		 4832 return -ENOMEM;
4833 }
4834 } else
4835 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4836#endif
b3881f74
TT		 4837 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4838 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4839
661aa520 4840 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4841 err = -EINVAL;
4842 goto restore_opts;
4843 }
4844
6b992ff2 4845 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4846 test_opt(sb, JOURNAL_CHECKSUM)) {
4847 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4848 "during remount not supported; ignoring");
4849 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4850 }
4851
6ae6514b
PS		 4852 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4853 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4854 ext4_msg(sb, KERN_ERR, "can't mount with "
4855 "both data=journal and delalloc");
4856 err = -EINVAL;
4857 goto restore_opts;
4858 }
4859 if (test_opt(sb, DIOREAD_NOLOCK)) {
4860 ext4_msg(sb, KERN_ERR, "can't mount with "
4861 "both data=journal and dioread_nolock");
4862 err = -EINVAL;
4863 goto restore_opts;
4864 }
923ae0ff
RZ		 4865 if (test_opt(sb, DAX)) {
4866 ext4_msg(sb, KERN_ERR, "can't mount with "
4867 "both data=journal and dax");
4868 err = -EINVAL;
4869 goto restore_opts;
4870 }
4871 }
4872
4873 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4874 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4875 "dax flag with busy inodes while remounting");
4876 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4877 }
4878
4ab2f15b 4879 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4880 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4881
4882 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4883 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4884
4885 es = sbi->s_es;
4886
b3881f74 4887 if (sbi->s_journal) {
0390131b 4888 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4889 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4890 }
ac27a0ec 4891
a2fd66d0
TT		 4892 if (*flags & MS_LAZYTIME)
4893 sb->s_flags |= MS_LAZYTIME;
4894
661aa520 4895 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4896 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4897 err = -EROFS;
4898 goto restore_opts;
4899 }
4900
4901 if (*flags & MS_RDONLY) {
38c03b34
TT		 4902 err = sync_filesystem(sb);
4903 if (err < 0)
4904 goto restore_opts;
0f0dd62f
CH		 4905 err = dquot_suspend(sb, -1);
4906 if (err < 0)
c79d967d 4907 goto restore_opts;
c79d967d 4908
ac27a0ec
DK		 4909 /*
4910 * First of all, the unconditional stuff we have to do
4911 * to disable replay of the journal when we next remount
4912 */
4913 sb->s_flags |= MS_RDONLY;
4914
4915 /*
4916 * OK, test if we are remounting a valid rw partition
4917 * readonly, and if so set the rdonly flag and then
4918 * mark the partition as valid again.
4919 */
617ba13b
MC		 4920 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4921 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4922 es->s_state = cpu_to_le16(sbi->s_mount_state);
4923
a63c9eb2 4924 if (sbi->s_journal)
0390131b 4925 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4926 } else {
a13fb1a4 4927 /* Make sure we can mount this feature set readwrite */
e2b911c5 4928 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 4929 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4930 err = -EROFS;
4931 goto restore_opts;
4932 }
8a266467
TT		 4933 /*
4934 * Make sure the group descriptor checksums
0b8e58a1 4935 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4936 */
4937 for (g = 0; g < sbi->s_groups_count; g++) {
4938 struct ext4_group_desc *gdp =
4939 ext4_get_group_desc(sb, g, NULL);
4940
feb0ab32 4941 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4942 ext4_msg(sb, KERN_ERR,
4943 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 4944 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 4945 le16_to_cpu(gdp->bg_checksum));
6a797d27 4946 err = -EFSBADCRC;
8a266467
TT		 4947 goto restore_opts;
4948 }
4949 }
4950
ead6596b
ES		 4951 /*
4952 * If we have an unprocessed orphan list hanging
4953 * around from a previously readonly bdev mount,
4954 * require a full umount/remount for now.
4955 */
4956 if (es->s_last_orphan) {
b31e1552 4957 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4958 "remount RDWR because of unprocessed "
4959 "orphan inode list. Please "
b31e1552 4960 "umount/remount instead");
ead6596b
ES		 4961 err = -EINVAL;
4962 goto restore_opts;
4963 }
4964
ac27a0ec
DK		 4965 /*
4966 * Mounting a RDONLY partition read-write, so reread
4967 * and store the current valid flag. (It may have
4968 * been changed by e2fsck since we originally mounted
4969 * the partition.)
4970 */
0390131b
FM		 4971 if (sbi->s_journal)
4972 ext4_clear_journal_err(sb, es);
ac27a0ec 4973 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4974 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4975 sb->s_flags &= ~MS_RDONLY;
e2b911c5 4976 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 4977 if (ext4_multi_mount_protect(sb,
4978 le64_to_cpu(es->s_mmp_block))) {
4979 err = -EROFS;
4980 goto restore_opts;
4981 }
c79d967d 4982 enable_quota = 1;
ac27a0ec
DK		 4983 }
4984 }
bfff6873
LC		 4985
4986 /*
4987 * Reinitialize lazy itable initialization thread based on
4988 * current settings
4989 */
4990 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4991 ext4_unregister_li_request(sb);
4992 else {
4993 ext4_group_t first_not_zeroed;
4994 first_not_zeroed = ext4_has_uninit_itable(sb);
4995 ext4_register_li_request(sb, first_not_zeroed);
4996 }
4997
6fd058f7 4998 ext4_setup_system_zone(sb);
d096ad0f 4999 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5000 ext4_commit_super(sb, 1);
0390131b 5001
ac27a0ec
DK		 5002#ifdef CONFIG_QUOTA
5003 /* Release old quota file names */
a2d4a646 5004 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5005 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5006 if (enable_quota) {
5007 if (sb_any_quota_suspended(sb))
5008 dquot_resume(sb, -1);
e2b911c5 5009 else if (ext4_has_feature_quota(sb)) {
7c319d32 5010 err = ext4_enable_quotas(sb);
07724f98 5011 if (err)
7c319d32 5012 goto restore_opts;
7c319d32
AK		 5013 }
5014 }
ac27a0ec 5015#endif
d4c402d9 5016
a26f4992 5017 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5018 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5019 kfree(orig_data);
ac27a0ec 5020 return 0;
0b8e58a1 5021
ac27a0ec
DK		 5022restore_opts:
5023 sb->s_flags = old_sb_flags;
5024 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5025 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5026 sbi->s_resuid = old_opts.s_resuid;
5027 sbi->s_resgid = old_opts.s_resgid;
5028 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5029 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5030 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5031#ifdef CONFIG_QUOTA
5032 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5033 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5034 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5035 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5036 }
5037#endif
d4c402d9 5038 kfree(orig_data);
ac27a0ec
DK		 5039 return err;
5040}
5041
689c958c
LX		 5042#ifdef CONFIG_QUOTA
5043static int ext4_statfs_project(struct super_block *sb,
5044 kprojid_t projid, struct kstatfs *buf)
5045{
5046 struct kqid qid;
5047 struct dquot *dquot;
5048 u64 limit;
5049 u64 curblock;
5050
5051 qid = make_kqid_projid(projid);
5052 dquot = dqget(sb, qid);
5053 if (IS_ERR(dquot))
5054 return PTR_ERR(dquot);
5055 spin_lock(&dq_data_lock);
5056
5057 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5058 dquot->dq_dqb.dqb_bsoftlimit :
5059 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5060 if (limit && buf->f_blocks > limit) {
5061 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5062 buf->f_blocks = limit;
5063 buf->f_bfree = buf->f_bavail =
5064 (buf->f_blocks > curblock) ?
5065 (buf->f_blocks - curblock) : 0;
5066 }
5067
5068 limit = dquot->dq_dqb.dqb_isoftlimit ?
5069 dquot->dq_dqb.dqb_isoftlimit :
5070 dquot->dq_dqb.dqb_ihardlimit;
5071 if (limit && buf->f_files > limit) {
5072 buf->f_files = limit;
5073 buf->f_ffree =
5074 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5075 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5076 }
5077
5078 spin_unlock(&dq_data_lock);
5079 dqput(dquot);
5080 return 0;
5081}
5082#endif
5083
2b2d6d01 5084static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5085{
5086 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5087 struct ext4_sb_info *sbi = EXT4_SB(sb);
5088 struct ext4_super_block *es = sbi->s_es;
27dd4385 5089 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5090 u64 fsid;
d02a9391 5091 s64 bfree;
27dd4385 5092 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5093
952fc18e
TT		 5094 if (!test_opt(sb, MINIX_DF))
5095 overhead = sbi->s_overhead;
ac27a0ec 5096
617ba13b 5097 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5098 buf->f_bsize = sb->s_blocksize;
b72f78cb 5099 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5100 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5101 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5102 /* prevent underflow in case that few free space is available */
57042651 5103 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5104 buf->f_bavail = buf->f_bfree -
5105 (ext4_r_blocks_count(es) + resv_blocks);
5106 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5107 buf->f_bavail = 0;
5108 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5109 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5110 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5111 fsid = le64_to_cpup((void *)es->s_uuid) ^
5112 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5113 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5114 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5115
689c958c
LX		 5116#ifdef CONFIG_QUOTA
5117 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5118 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5119 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5120#endif
ac27a0ec
DK		 5121 return 0;
5122}
5123
0b8e58a1
AD		 5124/* Helper function for writing quotas on sync - we need to start transaction
5125 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5126 * Process 1 Process 2
617ba13b 5127 * ext4_create() quota_sync()
a269eb18 5128 * jbd2_journal_start() write_dquot()
871a2931 5129 * dquot_initialize() down(dqio_mutex)
dab291af 5130 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5131 *
5132 */
5133
5134#ifdef CONFIG_QUOTA
5135
5136static inline struct inode *dquot_to_inode(struct dquot *dquot)
5137{
4c376dca 5138 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5139}
5140
617ba13b 5141static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5142{
5143 int ret, err;
5144 handle_t *handle;
5145 struct inode *inode;
5146
5147 inode = dquot_to_inode(dquot);
9924a92a 5148 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5149 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5150 if (IS_ERR(handle))
5151 return PTR_ERR(handle);
5152 ret = dquot_commit(dquot);
617ba13b 5153 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5154 if (!ret)
5155 ret = err;
5156 return ret;
5157}
5158
617ba13b 5159static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5160{
5161 int ret, err;
5162 handle_t *handle;
5163
9924a92a 5164 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5165 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5166 if (IS_ERR(handle))
5167 return PTR_ERR(handle);
5168 ret = dquot_acquire(dquot);
617ba13b 5169 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5170 if (!ret)
5171 ret = err;
5172 return ret;
5173}
5174
617ba13b 5175static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5176{
5177 int ret, err;
5178 handle_t *handle;
5179
9924a92a 5180 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5181 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5182 if (IS_ERR(handle)) {
5183 /* Release dquot anyway to avoid endless cycle in dqput() */
5184 dquot_release(dquot);
ac27a0ec 5185 return PTR_ERR(handle);
9c3013e9 5186 }
ac27a0ec 5187 ret = dquot_release(dquot);
617ba13b 5188 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5189 if (!ret)
5190 ret = err;
5191 return ret;
5192}
5193
617ba13b 5194static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5195{
262b4662
JK		 5196 struct super_block *sb = dquot->dq_sb;
5197 struct ext4_sb_info *sbi = EXT4_SB(sb);
5198
2c8be6b2 5199 /* Are we journaling quotas? */
e2b911c5 5200 if (ext4_has_feature_quota(sb) ||
262b4662 5201 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5202 dquot_mark_dquot_dirty(dquot);
617ba13b 5203 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5204 } else {
5205 return dquot_mark_dquot_dirty(dquot);
5206 }
5207}
5208
617ba13b 5209static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5210{
5211 int ret, err;
5212 handle_t *handle;
5213
5214 /* Data block + inode block */
2b0143b5 5215 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5216 if (IS_ERR(handle))
5217 return PTR_ERR(handle);
5218 ret = dquot_commit_info(sb, type);
617ba13b 5219 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5220 if (!ret)
5221 ret = err;
5222 return ret;
5223}
5224
5225/*
5226 * Turn on quotas during mount time - we need to find
5227 * the quota file and such...
5228 */
617ba13b 5229static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5230{
287a8095
CH		 5231 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5232 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5233}
5234
daf647d2
TT		 5235static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5236{
5237 struct ext4_inode_info *ei = EXT4_I(inode);
5238
5239 /* The first argument of lockdep_set_subclass has to be
5240 * *exactly* the same as the argument to init_rwsem() --- in
5241 * this case, in init_once() --- or lockdep gets unhappy
5242 * because the name of the lock is set using the
5243 * stringification of the argument to init_rwsem().
5244 */
5245 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5246 lockdep_set_subclass(&ei->i_data_sem, subclass);
5247}
5248
ac27a0ec
DK		 5249/*
5250 * Standard function to be called on quota_on
5251 */
617ba13b 5252static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5253 struct path *path)
ac27a0ec
DK		 5254{
5255 int err;
ac27a0ec
DK		 5256
5257 if (!test_opt(sb, QUOTA))
5258 return -EINVAL;
0623543b 5259
ac27a0ec 5260 /* Quotafile not on the same filesystem? */
d8c9584e 5261 if (path->dentry->d_sb != sb)
ac27a0ec 5262 return -EXDEV;
0623543b
JK		 5263 /* Journaling quota? */
5264 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5265 /* Quotafile not in fs root? */
f00c9e44 5266 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5267 ext4_msg(sb, KERN_WARNING,
5268 "Quota file not on filesystem root. "
5269 "Journaled quota will not work");
2b2d6d01 5270 }
0623543b
JK		 5271
5272 /*
5273 * When we journal data on quota file, we have to flush journal to see
5274 * all updates to the file when we bypass pagecache...
5275 */
0390131b 5276 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5277 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5278 /*
5279 * We don't need to lock updates but journal_flush() could
5280 * otherwise be livelocked...
5281 */
5282 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5283 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5284 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5285 if (err)
7ffe1ea8 5286 return err;
0623543b 5287 }
daf647d2
TT		 5288 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5289 err = dquot_quota_on(sb, type, format_id, path);
5290 if (err)
5291 lockdep_set_quota_inode(path->dentry->d_inode,
5292 I_DATA_SEM_NORMAL);
5293 return err;
ac27a0ec
DK		 5294}
5295
7c319d32
AK		 5296static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5297 unsigned int flags)
5298{
5299 int err;
5300 struct inode *qf_inode;
a2d4a646 5301 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 5302 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 5303 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5304 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 5305 };
5306
e2b911c5 5307 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 5308
5309 if (!qf_inums[type])
5310 return -EPERM;
5311
5312 qf_inode = ext4_iget(sb, qf_inums[type]);
5313 if (IS_ERR(qf_inode)) {
5314 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5315 return PTR_ERR(qf_inode);
5316 }
5317
bcb13850
JK		 5318 /* Don't account quota for quota files to avoid recursion */
5319 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 5320 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7c319d32
AK		 5321 err = dquot_enable(qf_inode, type, format_id, flags);
5322 iput(qf_inode);
daf647d2
TT		 5323 if (err)
5324 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7c319d32
AK		 5325
5326 return err;
5327}
5328
5329/* Enable usage tracking for all quota types. */
5330static int ext4_enable_quotas(struct super_block *sb)
5331{
5332 int type, err = 0;
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 5337 };
49da9392
JK		 5338 bool quota_mopt[EXT4_MAXQUOTAS] = {
5339 test_opt(sb, USRQUOTA),
5340 test_opt(sb, GRPQUOTA),
5341 test_opt(sb, PRJQUOTA),
5342 };
7c319d32
AK		 5343
5344 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5345 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5346 if (qf_inums[type]) {
5347 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 5348 DQUOT_USAGE_ENABLED |
5349 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 5350 if (err) {
5351 ext4_warning(sb,
72ba7450
TT		 5352 "Failed to enable quota tracking "
5353 "(type=%d, err=%d). Please run "
5354 "e2fsck to fix.", type, err);
7c319d32
AK		 5355 return err;
5356 }
5357 }
5358 }
5359 return 0;
5360}
5361
ca0e05e4
DM		 5362static int ext4_quota_off(struct super_block *sb, int type)
5363{
21f97697
JK		 5364 struct inode *inode = sb_dqopt(sb)->files[type];
5365 handle_t *handle;
5366
87009d86
DM		 5367 /* Force all delayed allocation blocks to be allocated.
5368 * Caller already holds s_umount sem */
5369 if (test_opt(sb, DELALLOC))
ca0e05e4 5370 sync_filesystem(sb);
ca0e05e4 5371
0b268590
AG		 5372 if (!inode)
5373 goto out;
5374
21f97697
JK		 5375 /* Update modification times of quota files when userspace can
5376 * start looking at them */
9924a92a 5377 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5378 if (IS_ERR(handle))
5379 goto out;
eeca7ea1 5380 inode->i_mtime = inode->i_ctime = current_time(inode);
21f97697
JK		 5381 ext4_mark_inode_dirty(handle, inode);
5382 ext4_journal_stop(handle);
5383
5384out:
ca0e05e4
DM		 5385 return dquot_quota_off(sb, type);
5386}
5387
ac27a0ec
DK		 5388/* Read data from quotafile - avoid pagecache and such because we cannot afford
5389 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5390 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5391 * we don't have to be afraid of races */
617ba13b 5392static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5393 size_t len, loff_t off)
5394{
5395 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5396 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5397 int offset = off & (sb->s_blocksize - 1);
5398 int tocopy;
5399 size_t toread;
5400 struct buffer_head *bh;
5401 loff_t i_size = i_size_read(inode);
5402
5403 if (off > i_size)
5404 return 0;
5405 if (off+len > i_size)
5406 len = i_size-off;
5407 toread = len;
5408 while (toread > 0) {
5409 tocopy = sb->s_blocksize - offset < toread ?
5410 sb->s_blocksize - offset : toread;
1c215028
TT		 5411 bh = ext4_bread(NULL, inode, blk, 0);
5412 if (IS_ERR(bh))
5413 return PTR_ERR(bh);
ac27a0ec
DK		 5414 if (!bh) /* A hole? */
5415 memset(data, 0, tocopy);
5416 else
5417 memcpy(data, bh->b_data+offset, tocopy);
5418 brelse(bh);
5419 offset = 0;
5420 toread -= tocopy;
5421 data += tocopy;
5422 blk++;
5423 }
5424 return len;
5425}
5426
5427/* Write to quotafile (we know the transaction is already started and has
5428 * enough credits) */
617ba13b 5429static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5430 const char *data, size_t len, loff_t off)
5431{
5432 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5433 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5434 int err, offset = off & (sb->s_blocksize - 1);
c5e298ae 5435 int retries = 0;
ac27a0ec
DK		 5436 struct buffer_head *bh;
5437 handle_t *handle = journal_current_handle();
5438
0390131b 5439 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5440 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5441 " cancelled because transaction is not started",
9c3013e9
JK		 5442 (unsigned long long)off, (unsigned long long)len);
5443 return -EIO;
5444 }
67eeb568
DM		 5445 /*
5446 * Since we account only one data block in transaction credits,
5447 * then it is impossible to cross a block boundary.
5448 */
5449 if (sb->s_blocksize - offset < len) {
5450 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5451 " cancelled because not block aligned",
5452 (unsigned long long)off, (unsigned long long)len);
5453 return -EIO;
5454 }
5455
c5e298ae
TT		 5456 do {
5457 bh = ext4_bread(handle, inode, blk,
5458 EXT4_GET_BLOCKS_CREATE |
5459 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5460 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5461 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 5462 if (IS_ERR(bh))
5463 return PTR_ERR(bh);
67eeb568
DM		 5464 if (!bh)
5465 goto out;
5d601255 5466 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5467 err = ext4_journal_get_write_access(handle, bh);
5468 if (err) {
5469 brelse(bh);
1c215028 5470 return err;
ac27a0ec 5471 }
67eeb568
DM		 5472 lock_buffer(bh);
5473 memcpy(bh->b_data+offset, data, len);
5474 flush_dcache_page(bh->b_page);
5475 unlock_buffer(bh);
62d2b5f2 5476 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5477 brelse(bh);
ac27a0ec 5478out:
67eeb568
DM		 5479 if (inode->i_size < off + len) {
5480 i_size_write(inode, off + len);
617ba13b 5481 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5482 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5483 }
67eeb568 5484 return len;
ac27a0ec
DK		 5485}
5486
8f0e8746
TT		 5487static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5488{
5489 const struct quota_format_ops *ops;
5490
5491 if (!sb_has_quota_loaded(sb, qid->type))
5492 return -ESRCH;
5493 ops = sb_dqopt(sb)->ops[qid->type];
5494 if (!ops || !ops->get_next_id)
5495 return -ENOSYS;
5496 return dquot_get_next_id(sb, qid);
5497}
ac27a0ec
DK		 5498#endif
5499
152a0836
AV		 5500static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5501 const char *dev_name, void *data)
ac27a0ec 5502{
152a0836 5503 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5504}
5505
c290ea01 5506#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 5507static inline void register_as_ext2(void)
5508{
5509 int err = register_filesystem(&ext2_fs_type);
5510 if (err)
5511 printk(KERN_WARNING
5512 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5513}
5514
5515static inline void unregister_as_ext2(void)
5516{
5517 unregister_filesystem(&ext2_fs_type);
5518}
2035e776
TT		 5519
5520static inline int ext2_feature_set_ok(struct super_block *sb)
5521{
e2b911c5 5522 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776
TT		 5523 return 0;
5524 if (sb->s_flags & MS_RDONLY)
5525 return 1;
e2b911c5 5526 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 5527 return 0;
5528 return 1;
5529}
24b58424
TT		 5530#else
5531static inline void register_as_ext2(void) { }
5532static inline void unregister_as_ext2(void) { }
2035e776 5533static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5534#endif
5535
24b58424
TT		 5536static inline void register_as_ext3(void)
5537{
5538 int err = register_filesystem(&ext3_fs_type);
5539 if (err)
5540 printk(KERN_WARNING
5541 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5542}
5543
5544static inline void unregister_as_ext3(void)
5545{
5546 unregister_filesystem(&ext3_fs_type);
5547}
2035e776
TT		 5548
5549static inline int ext3_feature_set_ok(struct super_block *sb)
5550{
e2b911c5 5551 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 5552 return 0;
e2b911c5 5553 if (!ext4_has_feature_journal(sb))
2035e776
TT		 5554 return 0;
5555 if (sb->s_flags & MS_RDONLY)
5556 return 1;
e2b911c5 5557 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 5558 return 0;
5559 return 1;
5560}
24b58424 5561
03010a33
TT		 5562static struct file_system_type ext4_fs_type = {
5563 .owner = THIS_MODULE,
5564 .name = "ext4",
152a0836 5565 .mount = ext4_mount,
03010a33
TT		 5566 .kill_sb = kill_block_super,
5567 .fs_flags = FS_REQUIRES_DEV,
5568};
7f78e035 5569MODULE_ALIAS_FS("ext4");
03010a33 5570
e9e3bcec
ES		 5571/* Shared across all ext4 file systems */
5572wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 5573
5dabfc78 5574static int __init ext4_init_fs(void)
ac27a0ec 5575{
e9e3bcec 5576 int i, err;
c9de560d 5577
e294a537 5578 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8
AV		 5579 ext4_li_info = NULL;
5580 mutex_init(&ext4_li_mtx);
5581
9a4c8019 5582 /* Build-time check for flags consistency */
12e9b892 5583 ext4_check_flag_values();
e9e3bcec 5584
e142d052 5585 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 5586 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 5587
51865fda 5588 err = ext4_init_es();
6fd058f7
TT		 5589 if (err)
5590 return err;
51865fda
ZL		 5591
5592 err = ext4_init_pageio();
5593 if (err)
b5799018 5594 goto out5;
51865fda 5595
5dabfc78 5596 err = ext4_init_system_zone();
bd2d0210 5597 if (err)
b5799018 5598 goto out4;
857ac889 5599
b5799018 5600 err = ext4_init_sysfs();
dd68314c 5601 if (err)
b5799018 5602 goto out3;
857ac889 5603
5dabfc78 5604 err = ext4_init_mballoc();
c9de560d
AT		 5605 if (err)
5606 goto out2;
ac27a0ec
DK		 5607 err = init_inodecache();
5608 if (err)
5609 goto out1;
24b58424 5610 register_as_ext3();
2035e776 5611 register_as_ext2();
03010a33 5612 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5613 if (err)
5614 goto out;
bfff6873 5615
ac27a0ec
DK		 5616 return 0;
5617out:
24b58424
TT		 5618 unregister_as_ext2();
5619 unregister_as_ext3();
ac27a0ec
DK		 5620 destroy_inodecache();
5621out1:
5dabfc78 5622 ext4_exit_mballoc();
9c191f70 5623out2:
b5799018
TT		 5624 ext4_exit_sysfs();
5625out3:
5dabfc78 5626 ext4_exit_system_zone();
b5799018 5627out4:
5dabfc78 5628 ext4_exit_pageio();
b5799018 5629out5:
51865fda
ZL		 5630 ext4_exit_es();
5631
ac27a0ec
DK		 5632 return err;
5633}
5634
5dabfc78 5635static void __exit ext4_exit_fs(void)
ac27a0ec 5636{
bfff6873 5637 ext4_destroy_lazyinit_thread();
24b58424
TT		 5638 unregister_as_ext2();
5639 unregister_as_ext3();
03010a33 5640 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5641 destroy_inodecache();
5dabfc78 5642 ext4_exit_mballoc();
b5799018 5643 ext4_exit_sysfs();
5dabfc78
TT		 5644 ext4_exit_system_zone();
5645 ext4_exit_pageio();
dd12ed14 5646 ext4_exit_es();
ac27a0ec
DK		 5647}
5648
5649MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5650MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5651MODULE_LICENSE("GPL");
5dabfc78
TT		 5652module_init(ext4_init_fs)
5653module_exit(ext4_exit_fs)
Ubuntu Focal Linux kernel mirror