projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: fix possible memory leak in ext4_remount()
[linux-block.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>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
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>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
4a092d73 48#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 72static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 74static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 75static int ext4_freeze(struct super_block *sb);
152a0836
AV		 76static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
2035e776
TT		 78static inline int ext2_feature_set_ok(struct super_block *sb);
79static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 80static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 81static void ext4_destroy_lazyinit_thread(void);
82static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 83static void ext4_clear_request_list(void);
ac27a0ec 84
2035e776
TT		 85#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
86static struct file_system_type ext2_fs_type = {
87 .owner = THIS_MODULE,
88 .name = "ext2",
89 .mount = ext4_mount,
90 .kill_sb = kill_block_super,
91 .fs_flags = FS_REQUIRES_DEV,
92};
93#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
94#else
95#define IS_EXT2_SB(sb) (0)
96#endif
97
98
ba69f9ab
JK		 99#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
100static struct file_system_type ext3_fs_type = {
101 .owner = THIS_MODULE,
102 .name = "ext3",
152a0836 103 .mount = ext4_mount,
ba69f9ab
JK		 104 .kill_sb = kill_block_super,
105 .fs_flags = FS_REQUIRES_DEV,
106};
107#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
108#else
109#define IS_EXT3_SB(sb) (0)
110#endif
bd81d8ee 111
d25425f8
DW		 112static int ext4_verify_csum_type(struct super_block *sb,
113 struct ext4_super_block *es)
114{
115 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
116 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
117 return 1;
118
119 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
120}
121
a9c47317
DW		 122static __le32 ext4_superblock_csum(struct super_block *sb,
123 struct ext4_super_block *es)
124{
125 struct ext4_sb_info *sbi = EXT4_SB(sb);
126 int offset = offsetof(struct ext4_super_block, s_checksum);
127 __u32 csum;
128
129 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
130
131 return cpu_to_le32(csum);
132}
133
134int ext4_superblock_csum_verify(struct super_block *sb,
135 struct ext4_super_block *es)
136{
137 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
138 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
139 return 1;
140
141 return es->s_checksum == ext4_superblock_csum(sb, es);
142}
143
06db49e6 144void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 145{
06db49e6
TT		 146 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
147
a9c47317
DW		 148 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
149 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
150 return;
151
152 es->s_checksum = ext4_superblock_csum(sb, es);
153}
154
9933fc0a
TT		 155void *ext4_kvmalloc(size_t size, gfp_t flags)
156{
157 void *ret;
158
159 ret = kmalloc(size, flags);
160 if (!ret)
161 ret = __vmalloc(size, flags, PAGE_KERNEL);
162 return ret;
163}
164
165void *ext4_kvzalloc(size_t size, gfp_t flags)
166{
167 void *ret;
168
db9481c0 169 ret = kzalloc(size, flags);
9933fc0a
TT		 170 if (!ret)
171 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
172 return ret;
173}
174
175void ext4_kvfree(void *ptr)
176{
177 if (is_vmalloc_addr(ptr))
178 vfree(ptr);
179 else
180 kfree(ptr);
181
182}
183
8fadc143
AR		 184ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
185 struct ext4_group_desc *bg)
bd81d8ee 186{
3a14589c 187 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 188 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 189 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 190}
191
8fadc143
AR		 192ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
193 struct ext4_group_desc *bg)
bd81d8ee 194{
5272f837 195 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 196 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 197 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 198}
199
8fadc143
AR		 200ext4_fsblk_t ext4_inode_table(struct super_block *sb,
201 struct ext4_group_desc *bg)
bd81d8ee 202{
5272f837 203 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 204 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 205 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 206}
207
021b65bb
TT		 208__u32 ext4_free_group_clusters(struct super_block *sb,
209 struct ext4_group_desc *bg)
560671a0
AK		 210{
211 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
212 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 213 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 214}
215
216__u32 ext4_free_inodes_count(struct super_block *sb,
217 struct ext4_group_desc *bg)
218{
219 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
220 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 221 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 222}
223
224__u32 ext4_used_dirs_count(struct super_block *sb,
225 struct ext4_group_desc *bg)
226{
227 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
228 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 229 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 230}
231
232__u32 ext4_itable_unused_count(struct super_block *sb,
233 struct ext4_group_desc *bg)
234{
235 return le16_to_cpu(bg->bg_itable_unused_lo) |
236 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 237 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 238}
239
8fadc143
AR		 240void ext4_block_bitmap_set(struct super_block *sb,
241 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 242{
3a14589c 243 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 244 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
245 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 246}
247
8fadc143
AR		 248void ext4_inode_bitmap_set(struct super_block *sb,
249 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 250{
5272f837 251 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 252 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
253 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 254}
255
8fadc143
AR		 256void ext4_inode_table_set(struct super_block *sb,
257 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 258{
5272f837 259 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 260 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
261 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 262}
263
021b65bb
TT		 264void ext4_free_group_clusters_set(struct super_block *sb,
265 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 266{
267 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
268 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
269 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
270}
271
272void ext4_free_inodes_set(struct super_block *sb,
273 struct ext4_group_desc *bg, __u32 count)
274{
275 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
276 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
277 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
278}
279
280void ext4_used_dirs_set(struct super_block *sb,
281 struct ext4_group_desc *bg, __u32 count)
282{
283 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
284 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
285 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
286}
287
288void ext4_itable_unused_set(struct super_block *sb,
289 struct ext4_group_desc *bg, __u32 count)
290{
291 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
292 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
293 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
294}
295
d3d1faf6 296
1c13d5c0
TT		 297static void __save_error_info(struct super_block *sb, const char *func,
298 unsigned int line)
299{
300 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
301
302 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
303 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
304 es->s_last_error_time = cpu_to_le32(get_seconds());
305 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
306 es->s_last_error_line = cpu_to_le32(line);
307 if (!es->s_first_error_time) {
308 es->s_first_error_time = es->s_last_error_time;
309 strncpy(es->s_first_error_func, func,
310 sizeof(es->s_first_error_func));
311 es->s_first_error_line = cpu_to_le32(line);
312 es->s_first_error_ino = es->s_last_error_ino;
313 es->s_first_error_block = es->s_last_error_block;
314 }
66e61a9e
TT		 315 /*
316 * Start the daily error reporting function if it hasn't been
317 * started already
318 */
319 if (!es->s_error_count)
320 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 321 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 322}
323
324static void save_error_info(struct super_block *sb, const char *func,
325 unsigned int line)
326{
327 __save_error_info(sb, func, line);
328 ext4_commit_super(sb, 1);
329}
330
7c2e7087
TT		 331/*
332 * The del_gendisk() function uninitializes the disk-specific data
333 * structures, including the bdi structure, without telling anyone
334 * else. Once this happens, any attempt to call mark_buffer_dirty()
335 * (for example, by ext4_commit_super), will cause a kernel OOPS.
336 * This is a kludge to prevent these oops until we can put in a proper
337 * hook in del_gendisk() to inform the VFS and file system layers.
338 */
339static int block_device_ejected(struct super_block *sb)
340{
341 struct inode *bd_inode = sb->s_bdev->bd_inode;
342 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
343
344 return bdi->dev == NULL;
345}
346
18aadd47
BJ		 347static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
348{
349 struct super_block *sb = journal->j_private;
350 struct ext4_sb_info *sbi = EXT4_SB(sb);
351 int error = is_journal_aborted(journal);
352 struct ext4_journal_cb_entry *jce, *tmp;
353
354 spin_lock(&sbi->s_md_lock);
355 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
356 list_del_init(&jce->jce_list);
357 spin_unlock(&sbi->s_md_lock);
358 jce->jce_func(sb, jce, error);
359 spin_lock(&sbi->s_md_lock);
360 }
361 spin_unlock(&sbi->s_md_lock);
362}
1c13d5c0 363
ac27a0ec
DK		 364/* Deal with the reporting of failure conditions on a filesystem such as
365 * inconsistencies detected or read IO failures.
366 *
367 * On ext2, we can store the error state of the filesystem in the
617ba13b 368 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 369 * write ordering constraints on the superblock which prevent us from
370 * writing it out straight away; and given that the journal is about to
371 * be aborted, we can't rely on the current, or future, transactions to
372 * write out the superblock safely.
373 *
dab291af 374 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 375 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 376 * that error until we've noted it down and cleared it.
377 */
378
617ba13b 379static void ext4_handle_error(struct super_block *sb)
ac27a0ec 380{
ac27a0ec
DK		 381 if (sb->s_flags & MS_RDONLY)
382 return;
383
2b2d6d01 384 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 385 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 386
4ab2f15b 387 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 388 if (journal)
dab291af 389 jbd2_journal_abort(journal, -EIO);
ac27a0ec 390 }
2b2d6d01 391 if (test_opt(sb, ERRORS_RO)) {
b31e1552 392 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 393 sb->s_flags |= MS_RDONLY;
394 }
ac27a0ec 395 if (test_opt(sb, ERRORS_PANIC))
617ba13b 396 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 397 sb->s_id);
398}
399
12062ddd 400void __ext4_error(struct super_block *sb, const char *function,
c398eda0 401 unsigned int line, const char *fmt, ...)
ac27a0ec 402{
0ff2ea7d 403 struct va_format vaf;
ac27a0ec
DK		 404 va_list args;
405
406 va_start(args, fmt);
0ff2ea7d
JP		 407 vaf.fmt = fmt;
408 vaf.va = &args;
409 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
410 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec 411 va_end(args);
f3fc0210 412 save_error_info(sb, function, line);
ac27a0ec 413
617ba13b 414 ext4_handle_error(sb);
ac27a0ec
DK		 415}
416
c398eda0
TT		 417void ext4_error_inode(struct inode *inode, const char *function,
418 unsigned int line, ext4_fsblk_t block,
273df556
FM		 419 const char *fmt, ...)
420{
421 va_list args;
f7c21177 422 struct va_format vaf;
1c13d5c0 423 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 424
1c13d5c0
TT		 425 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
426 es->s_last_error_block = cpu_to_le64(block);
427 save_error_info(inode->i_sb, function, line);
273df556 428 va_start(args, fmt);
f7c21177
TT		 429 vaf.fmt = fmt;
430 vaf.va = &args;
c398eda0 431 if (block)
d9ee81da
JP		 432 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
433 "inode #%lu: block %llu: comm %s: %pV\n",
434 inode->i_sb->s_id, function, line, inode->i_ino,
435 block, current->comm, &vaf);
436 else
437 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
438 "inode #%lu: comm %s: %pV\n",
439 inode->i_sb->s_id, function, line, inode->i_ino,
440 current->comm, &vaf);
273df556
FM		 441 va_end(args);
442
443 ext4_handle_error(inode->i_sb);
444}
445
c398eda0 446void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 447 unsigned int line, ext4_fsblk_t block,
448 const char *fmt, ...)
273df556
FM		 449{
450 va_list args;
f7c21177 451 struct va_format vaf;
1c13d5c0 452 struct ext4_super_block *es;
273df556
FM		 453 struct inode *inode = file->f_dentry->d_inode;
454 char pathname[80], *path;
455
1c13d5c0
TT		 456 es = EXT4_SB(inode->i_sb)->s_es;
457 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
458 save_error_info(inode->i_sb, function, line);
273df556 459 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 460 if (IS_ERR(path))
273df556 461 path = "(unknown)";
f7c21177
TT		 462 va_start(args, fmt);
463 vaf.fmt = fmt;
464 vaf.va = &args;
d9ee81da
JP		 465 if (block)
466 printk(KERN_CRIT
467 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
468 "block %llu: comm %s: path %s: %pV\n",
469 inode->i_sb->s_id, function, line, inode->i_ino,
470 block, current->comm, path, &vaf);
471 else
472 printk(KERN_CRIT
473 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
474 "comm %s: path %s: %pV\n",
475 inode->i_sb->s_id, function, line, inode->i_ino,
476 current->comm, path, &vaf);
273df556
FM		 477 va_end(args);
478
479 ext4_handle_error(inode->i_sb);
480}
481
722887dd
TT		 482const char *ext4_decode_error(struct super_block *sb, int errno,
483 char nbuf[16])
ac27a0ec
DK		 484{
485 char *errstr = NULL;
486
487 switch (errno) {
488 case -EIO:
489 errstr = "IO failure";
490 break;
491 case -ENOMEM:
492 errstr = "Out of memory";
493 break;
494 case -EROFS:
78f1ddbb
TT		 495 if (!sb || (EXT4_SB(sb)->s_journal &&
496 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 497 errstr = "Journal has aborted";
498 else
499 errstr = "Readonly filesystem";
500 break;
501 default:
502 /* If the caller passed in an extra buffer for unknown
503 * errors, textualise them now. Else we just return
504 * NULL. */
505 if (nbuf) {
506 /* Check for truncated error codes... */
507 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
508 errstr = nbuf;
509 }
510 break;
511 }
512
513 return errstr;
514}
515
617ba13b 516/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 517 * automatically and invokes the appropriate error response. */
518
c398eda0
TT		 519void __ext4_std_error(struct super_block *sb, const char *function,
520 unsigned int line, int errno)
ac27a0ec
DK		 521{
522 char nbuf[16];
523 const char *errstr;
524
525 /* Special case: if the error is EROFS, and we're not already
526 * inside a transaction, then there's really no point in logging
527 * an error. */
528 if (errno == -EROFS && journal_current_handle() == NULL &&
529 (sb->s_flags & MS_RDONLY))
530 return;
531
617ba13b 532 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 533 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
534 sb->s_id, function, line, errstr);
1c13d5c0 535 save_error_info(sb, function, line);
ac27a0ec 536
617ba13b 537 ext4_handle_error(sb);
ac27a0ec
DK		 538}
539
540/*
617ba13b 541 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 542 * abort function may be used to deal with unrecoverable failures such
543 * as journal IO errors or ENOMEM at a critical moment in log management.
544 *
545 * We unconditionally force the filesystem into an ABORT|READONLY state,
546 * unless the error response on the fs has been set to panic in which
547 * case we take the easy way out and panic immediately.
548 */
549
c67d859e 550void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 551 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 552{
553 va_list args;
554
1c13d5c0 555 save_error_info(sb, function, line);
ac27a0ec 556 va_start(args, fmt);
c398eda0
TT		 557 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
558 function, line);
ac27a0ec
DK		 559 vprintk(fmt, args);
560 printk("\n");
561 va_end(args);
562
1c13d5c0
TT		 563 if ((sb->s_flags & MS_RDONLY) == 0) {
564 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
565 sb->s_flags |= MS_RDONLY;
566 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
567 if (EXT4_SB(sb)->s_journal)
568 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
569 save_error_info(sb, function, line);
570 }
ac27a0ec 571 if (test_opt(sb, ERRORS_PANIC))
617ba13b 572 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 573}
574
0ff2ea7d 575void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 576{
0ff2ea7d 577 struct va_format vaf;
b31e1552
ES		 578 va_list args;
579
580 va_start(args, fmt);
0ff2ea7d
JP		 581 vaf.fmt = fmt;
582 vaf.va = &args;
583 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 584 va_end(args);
585}
586
12062ddd 587void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 588 unsigned int line, const char *fmt, ...)
ac27a0ec 589{
0ff2ea7d 590 struct va_format vaf;
ac27a0ec
DK		 591 va_list args;
592
593 va_start(args, fmt);
0ff2ea7d
JP		 594 vaf.fmt = fmt;
595 vaf.va = &args;
596 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
597 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 598 va_end(args);
599}
600
e29136f8
TT		 601void __ext4_grp_locked_error(const char *function, unsigned int line,
602 struct super_block *sb, ext4_group_t grp,
603 unsigned long ino, ext4_fsblk_t block,
604 const char *fmt, ...)
5d1b1b3f
AK		 605__releases(bitlock)
606__acquires(bitlock)
607{
0ff2ea7d 608 struct va_format vaf;
5d1b1b3f
AK		 609 va_list args;
610 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
611
1c13d5c0
TT		 612 es->s_last_error_ino = cpu_to_le32(ino);
613 es->s_last_error_block = cpu_to_le64(block);
614 __save_error_info(sb, function, line);
0ff2ea7d 615
5d1b1b3f 616 va_start(args, fmt);
0ff2ea7d
JP		 617
618 vaf.fmt = fmt;
619 vaf.va = &args;
21149d61 620 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 621 sb->s_id, function, line, grp);
622 if (ino)
0ff2ea7d 623 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 624 if (block)
0ff2ea7d
JP		 625 printk(KERN_CONT "block %llu:", (unsigned long long) block);
626 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 627 va_end(args);
628
629 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 630 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 631 return;
632 }
1c13d5c0 633
5d1b1b3f
AK		 634 ext4_unlock_group(sb, grp);
635 ext4_handle_error(sb);
636 /*
637 * We only get here in the ERRORS_RO case; relocking the group
638 * may be dangerous, but nothing bad will happen since the
639 * filesystem will have already been marked read/only and the
640 * journal has been aborted. We return 1 as a hint to callers
641 * who might what to use the return value from
25985edc 642 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 643 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
644 * aggressively from the ext4 function in question, with a
645 * more appropriate error code.
646 */
647 ext4_lock_group(sb, grp);
648 return;
649}
650
617ba13b 651void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 652{
617ba13b 653 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 654
617ba13b 655 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 656 return;
657
12062ddd 658 ext4_warning(sb,
ac27a0ec
DK		 659 "updating to rev %d because of new feature flag, "
660 "running e2fsck is recommended",
617ba13b 661 EXT4_DYNAMIC_REV);
ac27a0ec 662
617ba13b
MC		 663 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
664 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
665 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 666 /* leave es->s_feature_*compat flags alone */
667 /* es->s_uuid will be set by e2fsck if empty */
668
669 /*
670 * The rest of the superblock fields should be zero, and if not it
671 * means they are likely already in use, so leave them alone. We
672 * can leave it up to e2fsck to clean up any inconsistencies there.
673 */
674}
675
676/*
677 * Open the external journal device
678 */
b31e1552 679static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 680{
681 struct block_device *bdev;
682 char b[BDEVNAME_SIZE];
683
d4d77629 684 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 685 if (IS_ERR(bdev))
686 goto fail;
687 return bdev;
688
689fail:
b31e1552 690 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 691 __bdevname(dev, b), PTR_ERR(bdev));
692 return NULL;
693}
694
695/*
696 * Release the journal device
697 */
617ba13b 698static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 699{
e525fd89 700 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 701}
702
617ba13b 703static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 704{
705 struct block_device *bdev;
706 int ret = -ENODEV;
707
708 bdev = sbi->journal_bdev;
709 if (bdev) {
617ba13b 710 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 711 sbi->journal_bdev = NULL;
712 }
713 return ret;
714}
715
716static inline struct inode *orphan_list_entry(struct list_head *l)
717{
617ba13b 718 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 719}
720
617ba13b 721static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 722{
723 struct list_head *l;
724
b31e1552
ES		 725 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
726 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 727
728 printk(KERN_ERR "sb_info orphan list:\n");
729 list_for_each(l, &sbi->s_orphan) {
730 struct inode *inode = orphan_list_entry(l);
731 printk(KERN_ERR " "
732 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
733 inode->i_sb->s_id, inode->i_ino, inode,
734 inode->i_mode, inode->i_nlink,
735 NEXT_ORPHAN(inode));
736 }
737}
738
2b2d6d01 739static void ext4_put_super(struct super_block *sb)
ac27a0ec 740{
617ba13b
MC		 741 struct ext4_sb_info *sbi = EXT4_SB(sb);
742 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 743 int i, err;
ac27a0ec 744
857ac889 745 ext4_unregister_li_request(sb);
e0ccfd95
CH		 746 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
747
4c0425ff
MC		 748 flush_workqueue(sbi->dio_unwritten_wq);
749 destroy_workqueue(sbi->dio_unwritten_wq);
750
0390131b
FM		 751 if (sbi->s_journal) {
752 err = jbd2_journal_destroy(sbi->s_journal);
753 sbi->s_journal = NULL;
754 if (err < 0)
c67d859e 755 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 756 }
d4edac31 757
74cd15cd 758 ext4_es_unregister_shrinker(sb);
a1c6c569 759 del_timer(&sbi->s_err_report);
d4edac31
JB		 760 ext4_release_system_zone(sb);
761 ext4_mb_release(sb);
762 ext4_ext_release(sb);
763 ext4_xattr_put_super(sb);
764
ac27a0ec 765 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 766 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 767 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 768 }
58c5873a 769 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 770 ext4_commit_super(sb, 1);
771
240799cd 772 if (sbi->s_proc) {
66acdcf4 773 remove_proc_entry("options", sbi->s_proc);
9f6200bb 774 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 775 }
3197ebdb 776 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 777
778 for (i = 0; i < sbi->s_gdb_count; i++)
779 brelse(sbi->s_group_desc[i]);
f18a5f21 780 ext4_kvfree(sbi->s_group_desc);
9933fc0a 781 ext4_kvfree(sbi->s_flex_groups);
57042651 782 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 783 percpu_counter_destroy(&sbi->s_freeinodes_counter);
784 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 785 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 786 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
ac27a0ec
DK		 787 brelse(sbi->s_sbh);
788#ifdef CONFIG_QUOTA
789 for (i = 0; i < MAXQUOTAS; i++)
790 kfree(sbi->s_qf_names[i]);
791#endif
792
793 /* Debugging code just in case the in-memory inode orphan list
794 * isn't empty. The on-disk one can be non-empty if we've
795 * detected an error and taken the fs readonly, but the
796 * in-memory list had better be clean by this point. */
797 if (!list_empty(&sbi->s_orphan))
798 dump_orphan_list(sb, sbi);
799 J_ASSERT(list_empty(&sbi->s_orphan));
800
f98393a6 801 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 802 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
803 /*
804 * Invalidate the journal device's buffers. We don't want them
805 * floating about in memory - the physical journal device may
806 * hotswapped, and it breaks the `ro-after' testing code.
807 */
808 sync_blockdev(sbi->journal_bdev);
f98393a6 809 invalidate_bdev(sbi->journal_bdev);
617ba13b 810 ext4_blkdev_remove(sbi);
ac27a0ec 811 }
c5e06d10
JL		 812 if (sbi->s_mmp_tsk)
813 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 814 sb->s_fs_info = NULL;
3197ebdb
TT		 815 /*
816 * Now that we are completely done shutting down the
817 * superblock, we need to actually destroy the kobject.
818 */
3197ebdb
TT		 819 kobject_put(&sbi->s_kobj);
820 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 821 if (sbi->s_chksum_driver)
822 crypto_free_shash(sbi->s_chksum_driver);
705895b6 823 kfree(sbi->s_blockgroup_lock);
ac27a0ec 824 kfree(sbi);
ac27a0ec
DK		 825}
826
e18b890b 827static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 828
829/*
830 * Called inside transaction, so use GFP_NOFS
831 */
617ba13b 832static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 833{
617ba13b 834 struct ext4_inode_info *ei;
ac27a0ec 835
e6b4f8da 836 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 837 if (!ei)
838 return NULL;
0b8e58a1 839
ac27a0ec 840 ei->vfs_inode.i_version = 1;
c9de560d
AT		 841 INIT_LIST_HEAD(&ei->i_prealloc_list);
842 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 843 ext4_es_init_tree(&ei->i_es_tree);
844 rwlock_init(&ei->i_es_lock);
74cd15cd
ZL		 845 INIT_LIST_HEAD(&ei->i_es_lru);
846 ei->i_es_lru_nr = 0;
d2a17637
MC		 847 ei->i_reserved_data_blocks = 0;
848 ei->i_reserved_meta_blocks = 0;
849 ei->i_allocated_meta_blocks = 0;
9d0be502 850 ei->i_da_metadata_calc_len = 0;
7e731bc9 851 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 852 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 853#ifdef CONFIG_QUOTA
854 ei->i_reserved_quota = 0;
855#endif
8aefcd55 856 ei->jinode = NULL;
c7064ef1 857 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 858 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 859 ei->i_sync_tid = 0;
860 ei->i_datasync_tid = 0;
f7ad6d2e 861 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 862 atomic_set(&ei->i_unwritten, 0);
84c17543 863 INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
0b8e58a1 864
ac27a0ec
DK		 865 return &ei->vfs_inode;
866}
867
7ff9c073
TT		 868static int ext4_drop_inode(struct inode *inode)
869{
870 int drop = generic_drop_inode(inode);
871
872 trace_ext4_drop_inode(inode, drop);
873 return drop;
874}
875
fa0d7e3d
NP		 876static void ext4_i_callback(struct rcu_head *head)
877{
878 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 879 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
880}
881
617ba13b 882static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 883{
9f7dd93d 884 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 885 ext4_msg(inode->i_sb, KERN_ERR,
886 "Inode %lu (%p): orphan list check failed!",
887 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 888 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
889 EXT4_I(inode), sizeof(struct ext4_inode_info),
890 true);
891 dump_stack();
892 }
fa0d7e3d 893 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 894}
895
51cc5068 896static void init_once(void *foo)
ac27a0ec 897{
617ba13b 898 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 899
a35afb83 900 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 901 init_rwsem(&ei->xattr_sem);
0e855ac8 902 init_rwsem(&ei->i_data_sem);
a35afb83 903 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 904}
905
906static int init_inodecache(void)
907{
617ba13b
MC		 908 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
909 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 910 0, (SLAB_RECLAIM_ACCOUNT|
911 SLAB_MEM_SPREAD),
20c2df83 912 init_once);
617ba13b 913 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 914 return -ENOMEM;
915 return 0;
916}
917
918static void destroy_inodecache(void)
919{
8c0a8537
KS		 920 /*
921 * Make sure all delayed rcu free inodes are flushed before we
922 * destroy cache.
923 */
924 rcu_barrier();
617ba13b 925 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 926}
927
0930fcc1 928void ext4_clear_inode(struct inode *inode)
ac27a0ec 929{
0930fcc1 930 invalidate_inode_buffers(inode);
dbd5768f 931 clear_inode(inode);
9f754758 932 dquot_drop(inode);
c2ea3fde 933 ext4_discard_preallocations(inode);
51865fda 934 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
74cd15cd 935 ext4_es_lru_del(inode);
8aefcd55
TT		 936 if (EXT4_I(inode)->jinode) {
937 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
938 EXT4_I(inode)->jinode);
939 jbd2_free_inode(EXT4_I(inode)->jinode);
940 EXT4_I(inode)->jinode = NULL;
941 }
ac27a0ec
DK		 942}
943
1b961ac0 944static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 945 u64 ino, u32 generation)
ac27a0ec 946{
ac27a0ec 947 struct inode *inode;
ac27a0ec 948
617ba13b 949 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 950 return ERR_PTR(-ESTALE);
617ba13b 951 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 952 return ERR_PTR(-ESTALE);
953
954 /* iget isn't really right if the inode is currently unallocated!!
955 *
617ba13b 956 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 957 * deleted, so we should be safe.
958 *
959 * Currently we don't know the generation for parent directory, so
960 * a generation of 0 means "accept any"
961 */
1d1fe1ee
DH		 962 inode = ext4_iget(sb, ino);
963 if (IS_ERR(inode))
964 return ERR_CAST(inode);
965 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 966 iput(inode);
967 return ERR_PTR(-ESTALE);
968 }
1b961ac0
CH		 969
970 return inode;
971}
972
973static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 974 int fh_len, int fh_type)
1b961ac0
CH		 975{
976 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
977 ext4_nfs_get_inode);
978}
979
980static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 981 int fh_len, int fh_type)
1b961ac0
CH		 982{
983 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
984 ext4_nfs_get_inode);
ac27a0ec
DK		 985}
986
c39a7f84
TO		 987/*
988 * Try to release metadata pages (indirect blocks, directories) which are
989 * mapped via the block device. Since these pages could have journal heads
990 * which would prevent try_to_free_buffers() from freeing them, we must use
991 * jbd2 layer's try_to_free_buffers() function to release them.
992 */
0b8e58a1
AD		 993static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
994 gfp_t wait)
c39a7f84
TO		 995{
996 journal_t *journal = EXT4_SB(sb)->s_journal;
997
998 WARN_ON(PageChecked(page));
999 if (!page_has_buffers(page))
1000 return 0;
1001 if (journal)
1002 return jbd2_journal_try_to_free_buffers(journal, page,
1003 wait & ~__GFP_WAIT);
1004 return try_to_free_buffers(page);
1005}
1006
ac27a0ec 1007#ifdef CONFIG_QUOTA
af5bc92d 1008#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1009#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1010
617ba13b
MC		 1011static int ext4_write_dquot(struct dquot *dquot);
1012static int ext4_acquire_dquot(struct dquot *dquot);
1013static int ext4_release_dquot(struct dquot *dquot);
1014static int ext4_mark_dquot_dirty(struct dquot *dquot);
1015static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1016static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1017 struct path *path);
7c319d32
AK		 1018static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1019 int format_id);
ca0e05e4 1020static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1021static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1022static int ext4_quota_on_mount(struct super_block *sb, int type);
1023static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1024 size_t len, loff_t off);
617ba13b 1025static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1026 const char *data, size_t len, loff_t off);
7c319d32
AK		 1027static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1028 unsigned int flags);
1029static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1030
61e225dc 1031static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1032 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1033 .write_dquot = ext4_write_dquot,
1034 .acquire_dquot = ext4_acquire_dquot,
1035 .release_dquot = ext4_release_dquot,
1036 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1037 .write_info = ext4_write_info,
1038 .alloc_dquot = dquot_alloc,
1039 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1040};
1041
0d54b217 1042static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1043 .quota_on = ext4_quota_on,
ca0e05e4 1044 .quota_off = ext4_quota_off,
287a8095
CH		 1045 .quota_sync = dquot_quota_sync,
1046 .get_info = dquot_get_dqinfo,
1047 .set_info = dquot_set_dqinfo,
1048 .get_dqblk = dquot_get_dqblk,
1049 .set_dqblk = dquot_set_dqblk
ac27a0ec 1050};
7c319d32
AK		 1051
1052static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1053 .quota_on_meta = ext4_quota_on_sysfile,
1054 .quota_off = ext4_quota_off_sysfile,
1055 .quota_sync = dquot_quota_sync,
1056 .get_info = dquot_get_dqinfo,
1057 .set_info = dquot_set_dqinfo,
1058 .get_dqblk = dquot_get_dqblk,
1059 .set_dqblk = dquot_set_dqblk
1060};
ac27a0ec
DK		 1061#endif
1062
ee9b6d61 1063static const struct super_operations ext4_sops = {
617ba13b
MC		 1064 .alloc_inode = ext4_alloc_inode,
1065 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1066 .write_inode = ext4_write_inode,
1067 .dirty_inode = ext4_dirty_inode,
7ff9c073 1068 .drop_inode = ext4_drop_inode,
0930fcc1 1069 .evict_inode = ext4_evict_inode,
617ba13b 1070 .put_super = ext4_put_super,
617ba13b 1071 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1072 .freeze_fs = ext4_freeze,
1073 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1074 .statfs = ext4_statfs,
1075 .remount_fs = ext4_remount,
617ba13b 1076 .show_options = ext4_show_options,
ac27a0ec 1077#ifdef CONFIG_QUOTA
617ba13b
MC		 1078 .quota_read = ext4_quota_read,
1079 .quota_write = ext4_quota_write,
ac27a0ec 1080#endif
c39a7f84 1081 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1082};
1083
9ca92389
TT		 1084static const struct super_operations ext4_nojournal_sops = {
1085 .alloc_inode = ext4_alloc_inode,
1086 .destroy_inode = ext4_destroy_inode,
1087 .write_inode = ext4_write_inode,
1088 .dirty_inode = ext4_dirty_inode,
7ff9c073 1089 .drop_inode = ext4_drop_inode,
0930fcc1 1090 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1091 .put_super = ext4_put_super,
1092 .statfs = ext4_statfs,
1093 .remount_fs = ext4_remount,
9ca92389
TT		 1094 .show_options = ext4_show_options,
1095#ifdef CONFIG_QUOTA
1096 .quota_read = ext4_quota_read,
1097 .quota_write = ext4_quota_write,
1098#endif
1099 .bdev_try_to_free_page = bdev_try_to_free_page,
1100};
1101
39655164 1102static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1103 .fh_to_dentry = ext4_fh_to_dentry,
1104 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1105 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1106};
1107
1108enum {
1109 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1110 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1111 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1112 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1113 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1114 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1115 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1116 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1117 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1118 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1119 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1120 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1121 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1122 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1123 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1124 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1125 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1126 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1127 Opt_max_dir_size_kb,
ac27a0ec
DK		 1128};
1129
a447c093 1130static const match_table_t tokens = {
ac27a0ec
DK		 1131 {Opt_bsd_df, "bsddf"},
1132 {Opt_minix_df, "minixdf"},
1133 {Opt_grpid, "grpid"},
1134 {Opt_grpid, "bsdgroups"},
1135 {Opt_nogrpid, "nogrpid"},
1136 {Opt_nogrpid, "sysvgroups"},
1137 {Opt_resgid, "resgid=%u"},
1138 {Opt_resuid, "resuid=%u"},
1139 {Opt_sb, "sb=%u"},
1140 {Opt_err_cont, "errors=continue"},
1141 {Opt_err_panic, "errors=panic"},
1142 {Opt_err_ro, "errors=remount-ro"},
1143 {Opt_nouid32, "nouid32"},
ac27a0ec 1144 {Opt_debug, "debug"},
72578c33
TT		 1145 {Opt_removed, "oldalloc"},
1146 {Opt_removed, "orlov"},
ac27a0ec
DK		 1147 {Opt_user_xattr, "user_xattr"},
1148 {Opt_nouser_xattr, "nouser_xattr"},
1149 {Opt_acl, "acl"},
1150 {Opt_noacl, "noacl"},
e3bb52ae 1151 {Opt_noload, "norecovery"},
5a916be1 1152 {Opt_noload, "noload"},
72578c33
TT		 1153 {Opt_removed, "nobh"},
1154 {Opt_removed, "bh"},
ac27a0ec 1155 {Opt_commit, "commit=%u"},
30773840
TT		 1156 {Opt_min_batch_time, "min_batch_time=%u"},
1157 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1158 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1159 {Opt_journal_checksum, "journal_checksum"},
1160 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1161 {Opt_abort, "abort"},
1162 {Opt_data_journal, "data=journal"},
1163 {Opt_data_ordered, "data=ordered"},
1164 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1165 {Opt_data_err_abort, "data_err=abort"},
1166 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1167 {Opt_offusrjquota, "usrjquota="},
1168 {Opt_usrjquota, "usrjquota=%s"},
1169 {Opt_offgrpjquota, "grpjquota="},
1170 {Opt_grpjquota, "grpjquota=%s"},
1171 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1172 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1173 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1174 {Opt_grpquota, "grpquota"},
1175 {Opt_noquota, "noquota"},
1176 {Opt_quota, "quota"},
1177 {Opt_usrquota, "usrquota"},
1178 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1179 {Opt_barrier, "barrier"},
1180 {Opt_nobarrier, "nobarrier"},
25ec56b5 1181 {Opt_i_version, "i_version"},
c9de560d 1182 {Opt_stripe, "stripe=%u"},
64769240 1183 {Opt_delalloc, "delalloc"},
dd919b98 1184 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1185 {Opt_removed, "mblk_io_submit"},
1186 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1187 {Opt_block_validity, "block_validity"},
1188 {Opt_noblock_validity, "noblock_validity"},
240799cd 1189 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1190 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1191 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1192 {Opt_auto_da_alloc, "auto_da_alloc"},
1193 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1194 {Opt_dioread_nolock, "dioread_nolock"},
1195 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1196 {Opt_discard, "discard"},
1197 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1198 {Opt_init_itable, "init_itable=%u"},
1199 {Opt_init_itable, "init_itable"},
1200 {Opt_noinit_itable, "noinit_itable"},
df981d03 1201 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1202 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1203 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1204 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1205 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1206 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1207 {Opt_err, NULL},
ac27a0ec
DK		 1208};
1209
617ba13b 1210static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1211{
617ba13b 1212 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1213 char *options = (char *) *data;
1214
1215 if (!options || strncmp(options, "sb=", 3) != 0)
1216 return 1; /* Default location */
0b8e58a1 1217
ac27a0ec 1218 options += 3;
0b8e58a1 1219 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1220 sb_block = simple_strtoul(options, &options, 0);
1221 if (*options && *options != ',') {
4776004f 1222 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1223 (char *) *data);
1224 return 1;
1225 }
1226 if (*options == ',')
1227 options++;
1228 *data = (void *) options;
0b8e58a1 1229
ac27a0ec
DK		 1230 return sb_block;
1231}
1232
b3881f74 1233#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1234static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1235 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1236
56c50f11
DM		 1237#ifdef CONFIG_QUOTA
1238static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1239{
1240 struct ext4_sb_info *sbi = EXT4_SB(sb);
1241 char *qname;
03dafb5f 1242 int ret = -1;
56c50f11
DM		 1243
1244 if (sb_any_quota_loaded(sb) &&
1245 !sbi->s_qf_names[qtype]) {
1246 ext4_msg(sb, KERN_ERR,
1247 "Cannot change journaled "
1248 "quota options when quota turned on");
57f73c2c 1249 return -1;
56c50f11
DM		 1250 }
1251 qname = match_strdup(args);
1252 if (!qname) {
1253 ext4_msg(sb, KERN_ERR,
1254 "Not enough memory for storing quotafile name");
57f73c2c 1255 return -1;
56c50f11 1256 }
03dafb5f
CG		 1257 if (sbi->s_qf_names[qtype]) {
1258 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1259 ret = 1;
1260 else
1261 ext4_msg(sb, KERN_ERR,
1262 "%s quota file already specified",
1263 QTYPE2NAME(qtype));
1264 goto errout;
56c50f11 1265 }
03dafb5f 1266 if (strchr(qname, '/')) {
56c50f11
DM		 1267 ext4_msg(sb, KERN_ERR,
1268 "quotafile must be on filesystem root");
03dafb5f 1269 goto errout;
56c50f11 1270 }
03dafb5f 1271 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1272 set_opt(sb, QUOTA);
56c50f11 1273 return 1;
03dafb5f
CG		 1274errout:
1275 kfree(qname);
1276 return ret;
56c50f11
DM		 1277}
1278
1279static int clear_qf_name(struct super_block *sb, int qtype)
1280{
1281
1282 struct ext4_sb_info *sbi = EXT4_SB(sb);
1283
1284 if (sb_any_quota_loaded(sb) &&
1285 sbi->s_qf_names[qtype]) {
1286 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1287 " when quota turned on");
57f73c2c 1288 return -1;
56c50f11 1289 }
03dafb5f 1290 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1291 sbi->s_qf_names[qtype] = NULL;
1292 return 1;
1293}
1294#endif
1295
26092bf5
TT		 1296#define MOPT_SET 0x0001
1297#define MOPT_CLEAR 0x0002
1298#define MOPT_NOSUPPORT 0x0004
1299#define MOPT_EXPLICIT 0x0008
1300#define MOPT_CLEAR_ERR 0x0010
1301#define MOPT_GTE0 0x0020
ac27a0ec 1302#ifdef CONFIG_QUOTA
26092bf5
TT		 1303#define MOPT_Q 0
1304#define MOPT_QFMT 0x0040
1305#else
1306#define MOPT_Q MOPT_NOSUPPORT
1307#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1308#endif
26092bf5 1309#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1310#define MOPT_NO_EXT2 0x0100
1311#define MOPT_NO_EXT3 0x0200
1312#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
26092bf5
TT		 1313
1314static const struct mount_opts {
1315 int token;
1316 int mount_opt;
1317 int flags;
1318} ext4_mount_opts[] = {
1319 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1320 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1321 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1322 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1323 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1324 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1325 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1326 MOPT_EXT4_ONLY | MOPT_SET},
1327 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1328 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1329 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1330 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1331 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1332 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1333 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1334 MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1335 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1336 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1337 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1338 EXT4_MOUNT_JOURNAL_CHECKSUM),
1339 MOPT_EXT4_ONLY | MOPT_SET},
1340 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1341 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1342 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1343 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1344 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1345 MOPT_NO_EXT2 | MOPT_SET},
1346 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1347 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1348 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1349 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1350 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1351 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1352 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1353 {Opt_commit, 0, MOPT_GTE0},
1354 {Opt_max_batch_time, 0, MOPT_GTE0},
1355 {Opt_min_batch_time, 0, MOPT_GTE0},
1356 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1357 {Opt_init_itable, 0, MOPT_GTE0},
1358 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1359 {Opt_resuid, 0, MOPT_GTE0},
1360 {Opt_resgid, 0, MOPT_GTE0},
1361 {Opt_journal_dev, 0, MOPT_GTE0},
1362 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1363 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1364 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1365 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1366 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1367 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1368 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1369#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1370 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1371 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1372#else
26092bf5
TT		 1373 {Opt_acl, 0, MOPT_NOSUPPORT},
1374 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1375#endif
26092bf5
TT		 1376 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1377 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1378 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1379 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1380 MOPT_SET | MOPT_Q},
1381 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1382 MOPT_SET | MOPT_Q},
1383 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1384 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1385 {Opt_usrjquota, 0, MOPT_Q},
1386 {Opt_grpjquota, 0, MOPT_Q},
1387 {Opt_offusrjquota, 0, MOPT_Q},
1388 {Opt_offgrpjquota, 0, MOPT_Q},
1389 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1390 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1391 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1392 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1393 {Opt_err, 0, 0}
1394};
1395
1396static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1397 substring_t *args, unsigned long *journal_devnum,
1398 unsigned int *journal_ioprio, int is_remount)
1399{
1400 struct ext4_sb_info *sbi = EXT4_SB(sb);
1401 const struct mount_opts *m;
08cefc7a
EB		 1402 kuid_t uid;
1403 kgid_t gid;
26092bf5
TT		 1404 int arg = 0;
1405
57f73c2c
TT		 1406#ifdef CONFIG_QUOTA
1407 if (token == Opt_usrjquota)
1408 return set_qf_name(sb, USRQUOTA, &args[0]);
1409 else if (token == Opt_grpjquota)
1410 return set_qf_name(sb, GRPQUOTA, &args[0]);
1411 else if (token == Opt_offusrjquota)
1412 return clear_qf_name(sb, USRQUOTA);
1413 else if (token == Opt_offgrpjquota)
1414 return clear_qf_name(sb, GRPQUOTA);
1415#endif
26092bf5 1416 switch (token) {
f7048605
TT		 1417 case Opt_noacl:
1418 case Opt_nouser_xattr:
1419 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1420 break;
26092bf5
TT		 1421 case Opt_sb:
1422 return 1; /* handled by get_sb_block() */
1423 case Opt_removed:
5f3633e3 1424 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1425 return 1;
26092bf5
TT		 1426 case Opt_abort:
1427 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1428 return 1;
1429 case Opt_i_version:
1430 sb->s_flags |= MS_I_VERSION;
1431 return 1;
26092bf5
TT		 1432 }
1433
5f3633e3
JK		 1434 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1435 if (token == m->token)
1436 break;
1437
1438 if (m->token == Opt_err) {
1439 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1440 "or missing value", opt);
1441 return -1;
1442 }
1443
8dc0aa8c
TT		 1444 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1445 ext4_msg(sb, KERN_ERR,
1446 "Mount option \"%s\" incompatible with ext2", opt);
1447 return -1;
1448 }
1449 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1450 ext4_msg(sb, KERN_ERR,
1451 "Mount option \"%s\" incompatible with ext3", opt);
1452 return -1;
1453 }
1454
5f3633e3
JK		 1455 if (args->from && match_int(args, &arg))
1456 return -1;
1457 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1458 return -1;
1459 if (m->flags & MOPT_EXPLICIT)
1460 set_opt2(sb, EXPLICIT_DELALLOC);
1461 if (m->flags & MOPT_CLEAR_ERR)
1462 clear_opt(sb, ERRORS_MASK);
1463 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1464 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1465 "options when quota turned on");
1466 return -1;
1467 }
1468
1469 if (m->flags & MOPT_NOSUPPORT) {
1470 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1471 } else if (token == Opt_commit) {
1472 if (arg == 0)
1473 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1474 sbi->s_commit_interval = HZ * arg;
1475 } else if (token == Opt_max_batch_time) {
1476 if (arg == 0)
1477 arg = EXT4_DEF_MAX_BATCH_TIME;
1478 sbi->s_max_batch_time = arg;
1479 } else if (token == Opt_min_batch_time) {
1480 sbi->s_min_batch_time = arg;
1481 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1482 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1483 ext4_msg(sb, KERN_ERR,
1484 "EXT4-fs: inode_readahead_blks must be "
1485 "0 or a power of 2 smaller than 2^31");
26092bf5 1486 return -1;
5f3633e3
JK		 1487 }
1488 sbi->s_inode_readahead_blks = arg;
1489 } else if (token == Opt_init_itable) {
1490 set_opt(sb, INIT_INODE_TABLE);
1491 if (!args->from)
1492 arg = EXT4_DEF_LI_WAIT_MULT;
1493 sbi->s_li_wait_mult = arg;
1494 } else if (token == Opt_max_dir_size_kb) {
1495 sbi->s_max_dir_size_kb = arg;
1496 } else if (token == Opt_stripe) {
1497 sbi->s_stripe = arg;
1498 } else if (token == Opt_resuid) {
1499 uid = make_kuid(current_user_ns(), arg);
1500 if (!uid_valid(uid)) {
1501 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1502 return -1;
1503 }
5f3633e3
JK		 1504 sbi->s_resuid = uid;
1505 } else if (token == Opt_resgid) {
1506 gid = make_kgid(current_user_ns(), arg);
1507 if (!gid_valid(gid)) {
1508 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1509 return -1;
1510 }
1511 sbi->s_resgid = gid;
1512 } else if (token == Opt_journal_dev) {
1513 if (is_remount) {
1514 ext4_msg(sb, KERN_ERR,
1515 "Cannot specify journal on remount");
1516 return -1;
1517 }
1518 *journal_devnum = arg;
1519 } else if (token == Opt_journal_ioprio) {
1520 if (arg > 7) {
1521 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1522 " (must be 0-7)");
1523 return -1;
1524 }
1525 *journal_ioprio =
1526 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1527 } else if (m->flags & MOPT_DATAJ) {
1528 if (is_remount) {
1529 if (!sbi->s_journal)
1530 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1531 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1532 ext4_msg(sb, KERN_ERR,
26092bf5 1533 "Cannot change data mode on remount");
26092bf5 1534 return -1;
ac27a0ec 1535 }
26092bf5 1536 } else {
5f3633e3
JK		 1537 clear_opt(sb, DATA_FLAGS);
1538 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1539 }
5f3633e3
JK		 1540#ifdef CONFIG_QUOTA
1541 } else if (m->flags & MOPT_QFMT) {
1542 if (sb_any_quota_loaded(sb) &&
1543 sbi->s_jquota_fmt != m->mount_opt) {
1544 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1545 "quota options when quota turned on");
1546 return -1;
1547 }
1548 sbi->s_jquota_fmt = m->mount_opt;
1549#endif
1550 } else {
1551 if (!args->from)
1552 arg = 1;
1553 if (m->flags & MOPT_CLEAR)
1554 arg = !arg;
1555 else if (unlikely(!(m->flags & MOPT_SET))) {
1556 ext4_msg(sb, KERN_WARNING,
1557 "buggy handling of option %s", opt);
1558 WARN_ON(1);
1559 return -1;
1560 }
1561 if (arg != 0)
1562 sbi->s_mount_opt |= m->mount_opt;
1563 else
1564 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1565 }
5f3633e3 1566 return 1;
26092bf5
TT		 1567}
1568
1569static int parse_options(char *options, struct super_block *sb,
1570 unsigned long *journal_devnum,
1571 unsigned int *journal_ioprio,
1572 int is_remount)
1573{
1574 struct ext4_sb_info *sbi = EXT4_SB(sb);
1575 char *p;
1576 substring_t args[MAX_OPT_ARGS];
1577 int token;
1578
1579 if (!options)
1580 return 1;
1581
1582 while ((p = strsep(&options, ",")) != NULL) {
1583 if (!*p)
1584 continue;
1585 /*
1586 * Initialize args struct so we know whether arg was
1587 * found; some options take optional arguments.
1588 */
caecd0af 1589 args[0].to = args[0].from = NULL;
26092bf5
TT		 1590 token = match_token(p, tokens, args);
1591 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1592 journal_ioprio, is_remount) < 0)
1593 return 0;
ac27a0ec
DK		 1594 }
1595#ifdef CONFIG_QUOTA
1596 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1597 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1598 clear_opt(sb, USRQUOTA);
ac27a0ec 1599
482a7425 1600 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1601 clear_opt(sb, GRPQUOTA);
ac27a0ec 1602
56c50f11 1603 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1604 ext4_msg(sb, KERN_ERR, "old and new quota "
1605 "format mixing");
ac27a0ec
DK		 1606 return 0;
1607 }
1608
1609 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1610 ext4_msg(sb, KERN_ERR, "journaled quota format "
1611 "not specified");
ac27a0ec
DK		 1612 return 0;
1613 }
1614 } else {
1615 if (sbi->s_jquota_fmt) {
b31e1552 1616 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1617 "specified with no journaling "
b31e1552 1618 "enabled");
ac27a0ec
DK		 1619 return 0;
1620 }
1621 }
1622#endif
261cb20c
JK		 1623 if (test_opt(sb, DIOREAD_NOLOCK)) {
1624 int blocksize =
1625 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1626
1627 if (blocksize < PAGE_CACHE_SIZE) {
1628 ext4_msg(sb, KERN_ERR, "can't mount with "
1629 "dioread_nolock if block size != PAGE_SIZE");
1630 return 0;
1631 }
1632 }
ac27a0ec
DK		 1633 return 1;
1634}
1635
2adf6da8
TT		 1636static inline void ext4_show_quota_options(struct seq_file *seq,
1637 struct super_block *sb)
1638{
1639#if defined(CONFIG_QUOTA)
1640 struct ext4_sb_info *sbi = EXT4_SB(sb);
1641
1642 if (sbi->s_jquota_fmt) {
1643 char *fmtname = "";
1644
1645 switch (sbi->s_jquota_fmt) {
1646 case QFMT_VFS_OLD:
1647 fmtname = "vfsold";
1648 break;
1649 case QFMT_VFS_V0:
1650 fmtname = "vfsv0";
1651 break;
1652 case QFMT_VFS_V1:
1653 fmtname = "vfsv1";
1654 break;
1655 }
1656 seq_printf(seq, ",jqfmt=%s", fmtname);
1657 }
1658
1659 if (sbi->s_qf_names[USRQUOTA])
1660 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1661
1662 if (sbi->s_qf_names[GRPQUOTA])
1663 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1664
1665 if (test_opt(sb, USRQUOTA))
1666 seq_puts(seq, ",usrquota");
1667
1668 if (test_opt(sb, GRPQUOTA))
1669 seq_puts(seq, ",grpquota");
1670#endif
1671}
1672
5a916be1
TT		 1673static const char *token2str(int token)
1674{
50df9fd5 1675 const struct match_token *t;
5a916be1
TT		 1676
1677 for (t = tokens; t->token != Opt_err; t++)
1678 if (t->token == token && !strchr(t->pattern, '='))
1679 break;
1680 return t->pattern;
1681}
1682
2adf6da8
TT		 1683/*
1684 * Show an option if
1685 * - it's set to a non-default value OR
1686 * - if the per-sb default is different from the global default
1687 */
66acdcf4
TT		 1688static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1689 int nodefs)
2adf6da8 1690{
2adf6da8
TT		 1691 struct ext4_sb_info *sbi = EXT4_SB(sb);
1692 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1693 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1694 const struct mount_opts *m;
66acdcf4 1695 char sep = nodefs ? '\n' : ',';
2adf6da8 1696
66acdcf4
TT		 1697#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1698#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1699
1700 if (sbi->s_sb_block != 1)
5a916be1
TT		 1701 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1702
1703 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1704 int want_set = m->flags & MOPT_SET;
1705 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1706 (m->flags & MOPT_CLEAR_ERR))
1707 continue;
66acdcf4 1708 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1709 continue; /* skip if same as the default */
1710 if ((want_set &&
1711 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1712 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1713 continue; /* select Opt_noFoo vs Opt_Foo */
1714 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1715 }
5a916be1 1716
08cefc7a 1717 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1718 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1719 SEQ_OPTS_PRINT("resuid=%u",
1720 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1721 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1722 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1723 SEQ_OPTS_PRINT("resgid=%u",
1724 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1725 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1726 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1727 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1728 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1729 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1730 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1731 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1732 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1733 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1734 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1735 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1736 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1737 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1738 if (sb->s_flags & MS_I_VERSION)
5a916be1 1739 SEQ_OPTS_PUTS("i_version");
66acdcf4 1740 if (nodefs || sbi->s_stripe)
5a916be1 1741 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1742 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1743 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1744 SEQ_OPTS_PUTS("data=journal");
1745 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1746 SEQ_OPTS_PUTS("data=ordered");
1747 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1748 SEQ_OPTS_PUTS("data=writeback");
1749 }
66acdcf4
TT		 1750 if (nodefs ||
1751 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1752 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1753 sbi->s_inode_readahead_blks);
2adf6da8 1754
66acdcf4
TT		 1755 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1756 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1757 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1758 if (nodefs || sbi->s_max_dir_size_kb)
1759 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1760
1761 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1762 return 0;
1763}
1764
66acdcf4
TT		 1765static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1766{
1767 return _ext4_show_options(seq, root->d_sb, 0);
1768}
1769
1770static int options_seq_show(struct seq_file *seq, void *offset)
1771{
1772 struct super_block *sb = seq->private;
1773 int rc;
1774
1775 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1776 rc = _ext4_show_options(seq, sb, 1);
1777 seq_puts(seq, "\n");
1778 return rc;
1779}
1780
1781static int options_open_fs(struct inode *inode, struct file *file)
1782{
1783 return single_open(file, options_seq_show, PDE(inode)->data);
1784}
1785
1786static const struct file_operations ext4_seq_options_fops = {
1787 .owner = THIS_MODULE,
1788 .open = options_open_fs,
1789 .read = seq_read,
1790 .llseek = seq_lseek,
1791 .release = single_release,
1792};
1793
617ba13b 1794static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1795 int read_only)
1796{
617ba13b 1797 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1798 int res = 0;
1799
617ba13b 1800 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1801 ext4_msg(sb, KERN_ERR, "revision level too high, "
1802 "forcing read-only mode");
ac27a0ec
DK		 1803 res = MS_RDONLY;
1804 }
1805 if (read_only)
281b5995 1806 goto done;
617ba13b 1807 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1808 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1809 "running e2fsck is recommended");
617ba13b 1810 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1811 ext4_msg(sb, KERN_WARNING,
1812 "warning: mounting fs with errors, "
1813 "running e2fsck is recommended");
ed3ce80a 1814 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1815 le16_to_cpu(es->s_mnt_count) >=
1816 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1817 ext4_msg(sb, KERN_WARNING,
1818 "warning: maximal mount count reached, "
1819 "running e2fsck is recommended");
ac27a0ec
DK		 1820 else if (le32_to_cpu(es->s_checkinterval) &&
1821 (le32_to_cpu(es->s_lastcheck) +
1822 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1823 ext4_msg(sb, KERN_WARNING,
1824 "warning: checktime reached, "
1825 "running e2fsck is recommended");
0b8e58a1 1826 if (!sbi->s_journal)
0390131b 1827 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1828 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1829 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1830 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1831 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1832 ext4_update_dynamic_rev(sb);
0390131b
FM		 1833 if (sbi->s_journal)
1834 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1835
e2d67052 1836 ext4_commit_super(sb, 1);
281b5995 1837done:
ac27a0ec 1838 if (test_opt(sb, DEBUG))
a9df9a49 1839 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1840 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1841 sb->s_blocksize,
1842 sbi->s_groups_count,
617ba13b
MC		 1843 EXT4_BLOCKS_PER_GROUP(sb),
1844 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1845 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1846
7abc52c2 1847 cleancache_init_fs(sb);
ac27a0ec
DK		 1848 return res;
1849}
1850
117fff10
TT		 1851int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1852{
1853 struct ext4_sb_info *sbi = EXT4_SB(sb);
1854 struct flex_groups *new_groups;
1855 int size;
1856
1857 if (!sbi->s_log_groups_per_flex)
1858 return 0;
1859
1860 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1861 if (size <= sbi->s_flex_groups_allocated)
1862 return 0;
1863
1864 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1865 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1866 if (!new_groups) {
1867 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1868 size / (int) sizeof(struct flex_groups));
1869 return -ENOMEM;
1870 }
1871
1872 if (sbi->s_flex_groups) {
1873 memcpy(new_groups, sbi->s_flex_groups,
1874 (sbi->s_flex_groups_allocated *
1875 sizeof(struct flex_groups)));
1876 ext4_kvfree(sbi->s_flex_groups);
1877 }
1878 sbi->s_flex_groups = new_groups;
1879 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1880 return 0;
1881}
1882
772cb7c8
JS		 1883static int ext4_fill_flex_info(struct super_block *sb)
1884{
1885 struct ext4_sb_info *sbi = EXT4_SB(sb);
1886 struct ext4_group_desc *gdp = NULL;
772cb7c8 1887 ext4_group_t flex_group;
d50f2ab6 1888 unsigned int groups_per_flex = 0;
117fff10 1889 int i, err;
772cb7c8 1890
503358ae 1891 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1892 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1893 sbi->s_log_groups_per_flex = 0;
1894 return 1;
1895 }
76495ec1 1896 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
772cb7c8 1897
117fff10
TT		 1898 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1899 if (err)
9933fc0a 1900 goto failed;
772cb7c8 1901
772cb7c8 1902 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1903 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1904
1905 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1906 atomic_add(ext4_free_inodes_count(sb, gdp),
1907 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1908 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1909 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1910 atomic_add(ext4_used_dirs_count(sb, gdp),
1911 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1912 }
1913
1914 return 1;
1915failed:
1916 return 0;
1917}
1918
feb0ab32
DW		 1919static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1920 struct ext4_group_desc *gdp)
717d50e4 1921{
feb0ab32 1922 int offset;
717d50e4 1923 __u16 crc = 0;
feb0ab32 1924 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1925
feb0ab32
DW		 1926 if ((sbi->s_es->s_feature_ro_compat &
1927 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1928 /* Use new metadata_csum algorithm */
1929 __u16 old_csum;
1930 __u32 csum32;
1931
1932 old_csum = gdp->bg_checksum;
1933 gdp->bg_checksum = 0;
1934 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1935 sizeof(le_group));
1936 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1937 sbi->s_desc_size);
1938 gdp->bg_checksum = old_csum;
1939
1940 crc = csum32 & 0xFFFF;
1941 goto out;
717d50e4
AD		 1942 }
1943
feb0ab32
DW		 1944 /* old crc16 code */
1945 offset = offsetof(struct ext4_group_desc, bg_checksum);
1946
1947 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1948 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1949 crc = crc16(crc, (__u8 *)gdp, offset);
1950 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1951 /* for checksum of struct ext4_group_desc do the rest...*/
1952 if ((sbi->s_es->s_feature_incompat &
1953 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1954 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1955 crc = crc16(crc, (__u8 *)gdp + offset,
1956 le16_to_cpu(sbi->s_es->s_desc_size) -
1957 offset);
1958
1959out:
717d50e4
AD		 1960 return cpu_to_le16(crc);
1961}
1962
feb0ab32 1963int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 1964 struct ext4_group_desc *gdp)
1965{
feb0ab32
DW		 1966 if (ext4_has_group_desc_csum(sb) &&
1967 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1968 block_group, gdp)))
717d50e4
AD		 1969 return 0;
1970
1971 return 1;
1972}
1973
feb0ab32
DW		 1974void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1975 struct ext4_group_desc *gdp)
1976{
1977 if (!ext4_has_group_desc_csum(sb))
1978 return;
1979 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
1980}
1981
ac27a0ec 1982/* Called at mount-time, super-block is locked */
bfff6873
LC		 1983static int ext4_check_descriptors(struct super_block *sb,
1984 ext4_group_t *first_not_zeroed)
ac27a0ec 1985{
617ba13b
MC		 1986 struct ext4_sb_info *sbi = EXT4_SB(sb);
1987 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1988 ext4_fsblk_t last_block;
bd81d8ee
LV		 1989 ext4_fsblk_t block_bitmap;
1990 ext4_fsblk_t inode_bitmap;
1991 ext4_fsblk_t inode_table;
ce421581 1992 int flexbg_flag = 0;
bfff6873 1993 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 1994
ce421581
JS		 1995 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1996 flexbg_flag = 1;
1997
af5bc92d 1998 ext4_debug("Checking group descriptors");
ac27a0ec 1999
197cd65a
AM		 2000 for (i = 0; i < sbi->s_groups_count; i++) {
2001 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2002
ce421581 2003 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2004 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2005 else
2006 last_block = first_block +
617ba13b 2007 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2008
bfff6873
LC		 2009 if ((grp == sbi->s_groups_count) &&
2010 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2011 grp = i;
2012
8fadc143 2013 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2014 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2015 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2016 "Block bitmap for group %u not in group "
b31e1552 2017 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2018 return 0;
2019 }
8fadc143 2020 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2021 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2022 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2023 "Inode bitmap for group %u not in group "
b31e1552 2024 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2025 return 0;
2026 }
8fadc143 2027 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2028 if (inode_table < first_block ||
2b2d6d01 2029 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2030 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2031 "Inode table for group %u not in group "
b31e1552 2032 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2033 return 0;
2034 }
955ce5f5 2035 ext4_lock_group(sb, i);
feb0ab32 2036 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2037 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2038 "Checksum for group %u failed (%u!=%u)",
2039 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2040 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2041 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2042 ext4_unlock_group(sb, i);
8a266467 2043 return 0;
7ee1ec4c 2044 }
717d50e4 2045 }
955ce5f5 2046 ext4_unlock_group(sb, i);
ce421581
JS		 2047 if (!flexbg_flag)
2048 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2049 }
bfff6873
LC		 2050 if (NULL != first_not_zeroed)
2051 *first_not_zeroed = grp;
ac27a0ec 2052
5dee5437
TT		 2053 ext4_free_blocks_count_set(sbi->s_es,
2054 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2055 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2056 return 1;
2057}
2058
617ba13b 2059/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2060 * the superblock) which were deleted from all directories, but held open by
2061 * a process at the time of a crash. We walk the list and try to delete these
2062 * inodes at recovery time (only with a read-write filesystem).
2063 *
2064 * In order to keep the orphan inode chain consistent during traversal (in
2065 * case of crash during recovery), we link each inode into the superblock
2066 * orphan list_head and handle it the same way as an inode deletion during
2067 * normal operation (which journals the operations for us).
2068 *
2069 * We only do an iget() and an iput() on each inode, which is very safe if we
2070 * accidentally point at an in-use or already deleted inode. The worst that
2071 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2072 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2073 * e2fsck was run on this filesystem, and it must have already done the orphan
2074 * inode cleanup for us, so we can safely abort without any further action.
2075 */
2b2d6d01
TT		 2076static void ext4_orphan_cleanup(struct super_block *sb,
2077 struct ext4_super_block *es)
ac27a0ec
DK		 2078{
2079 unsigned int s_flags = sb->s_flags;
2080 int nr_orphans = 0, nr_truncates = 0;
2081#ifdef CONFIG_QUOTA
2082 int i;
2083#endif
2084 if (!es->s_last_orphan) {
2085 jbd_debug(4, "no orphan inodes to clean up\n");
2086 return;
2087 }
2088
a8f48a95 2089 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2090 ext4_msg(sb, KERN_ERR, "write access "
2091 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2092 return;
2093 }
2094
d39195c3
AG		 2095 /* Check if feature set would not allow a r/w mount */
2096 if (!ext4_feature_set_ok(sb, 0)) {
2097 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2098 "unknown ROCOMPAT features");
2099 return;
2100 }
2101
617ba13b 2102 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2103 /* don't clear list on RO mount w/ errors */
2104 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2105 jbd_debug(1, "Errors on filesystem, "
2106 "clearing orphan list.\n");
c25f9bc6
ES		 2107 es->s_last_orphan = 0;
2108 }
ac27a0ec
DK		 2109 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2110 return;
2111 }
2112
2113 if (s_flags & MS_RDONLY) {
b31e1552 2114 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2115 sb->s_flags &= ~MS_RDONLY;
2116 }
2117#ifdef CONFIG_QUOTA
2118 /* Needed for iput() to work correctly and not trash data */
2119 sb->s_flags |= MS_ACTIVE;
2120 /* Turn on quotas so that they are updated correctly */
2121 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2122 if (EXT4_SB(sb)->s_qf_names[i]) {
2123 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2124 if (ret < 0)
b31e1552
ES		 2125 ext4_msg(sb, KERN_ERR,
2126 "Cannot turn on journaled "
2127 "quota: error %d", ret);
ac27a0ec
DK		 2128 }
2129 }
2130#endif
2131
2132 while (es->s_last_orphan) {
2133 struct inode *inode;
2134
97bd42b9
JB		 2135 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2136 if (IS_ERR(inode)) {
ac27a0ec
DK		 2137 es->s_last_orphan = 0;
2138 break;
2139 }
2140
617ba13b 2141 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2142 dquot_initialize(inode);
ac27a0ec 2143 if (inode->i_nlink) {
b31e1552
ES		 2144 ext4_msg(sb, KERN_DEBUG,
2145 "%s: truncating inode %lu to %lld bytes",
46e665e9 2146 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2147 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2148 inode->i_ino, inode->i_size);
721e3eba 2149 mutex_lock(&inode->i_mutex);
617ba13b 2150 ext4_truncate(inode);
721e3eba 2151 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2152 nr_truncates++;
2153 } else {
b31e1552
ES		 2154 ext4_msg(sb, KERN_DEBUG,
2155 "%s: deleting unreferenced inode %lu",
46e665e9 2156 __func__, inode->i_ino);
ac27a0ec
DK		 2157 jbd_debug(2, "deleting unreferenced inode %lu\n",
2158 inode->i_ino);
2159 nr_orphans++;
2160 }
2161 iput(inode); /* The delete magic happens here! */
2162 }
2163
2b2d6d01 2164#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2165
2166 if (nr_orphans)
b31e1552
ES		 2167 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2168 PLURAL(nr_orphans));
ac27a0ec 2169 if (nr_truncates)
b31e1552
ES		 2170 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2171 PLURAL(nr_truncates));
ac27a0ec
DK		 2172#ifdef CONFIG_QUOTA
2173 /* Turn quotas off */
2174 for (i = 0; i < MAXQUOTAS; i++) {
2175 if (sb_dqopt(sb)->files[i])
287a8095 2176 dquot_quota_off(sb, i);
ac27a0ec
DK		 2177 }
2178#endif
2179 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2180}
0b8e58a1 2181
cd2291a4
ES		 2182/*
2183 * Maximal extent format file size.
2184 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2185 * extent format containers, within a sector_t, and within i_blocks
2186 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2187 * so that won't be a limiting factor.
2188 *
f17722f9
LC		 2189 * However there is other limiting factor. We do store extents in the form
2190 * of starting block and length, hence the resulting length of the extent
2191 * covering maximum file size must fit into on-disk format containers as
2192 * well. Given that length is always by 1 unit bigger than max unit (because
2193 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2194 *
cd2291a4
ES		 2195 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2196 */
f287a1a5 2197static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2198{
2199 loff_t res;
2200 loff_t upper_limit = MAX_LFS_FILESIZE;
2201
2202 /* small i_blocks in vfs inode? */
f287a1a5 2203 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2204 /*
90c699a9 2205 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2206 * i_block represent total blocks in 512 bytes
2207 * 32 == size of vfs inode i_blocks * 8
2208 */
2209 upper_limit = (1LL << 32) - 1;
2210
2211 /* total blocks in file system block size */
2212 upper_limit >>= (blkbits - 9);
2213 upper_limit <<= blkbits;
2214 }
2215
f17722f9
LC		 2216 /*
2217 * 32-bit extent-start container, ee_block. We lower the maxbytes
2218 * by one fs block, so ee_len can cover the extent of maximum file
2219 * size
2220 */
2221 res = (1LL << 32) - 1;
cd2291a4 2222 res <<= blkbits;
cd2291a4
ES		 2223
2224 /* Sanity check against vm- & vfs- imposed limits */
2225 if (res > upper_limit)
2226 res = upper_limit;
2227
2228 return res;
2229}
ac27a0ec 2230
ac27a0ec 2231/*
cd2291a4 2232 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2233 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2234 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2235 */
f287a1a5 2236static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2237{
617ba13b 2238 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2239 int meta_blocks;
2240 loff_t upper_limit;
0b8e58a1
AD		 2241 /* This is calculated to be the largest file size for a dense, block
2242 * mapped file such that the file's total number of 512-byte sectors,
2243 * including data and all indirect blocks, does not exceed (2^48 - 1).
2244 *
2245 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2246 * number of 512-byte sectors of the file.
0fc1b451
AK		 2247 */
2248
f287a1a5 2249 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2250 /*
90c699a9 2251 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2252 * the inode i_block field represents total file blocks in
2253 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2254 */
2255 upper_limit = (1LL << 32) - 1;
2256
2257 /* total blocks in file system block size */
2258 upper_limit >>= (bits - 9);
2259
2260 } else {
8180a562
AK		 2261 /*
2262 * We use 48 bit ext4_inode i_blocks
2263 * With EXT4_HUGE_FILE_FL set the i_blocks
2264 * represent total number of blocks in
2265 * file system block size
2266 */
0fc1b451
AK		 2267 upper_limit = (1LL << 48) - 1;
2268
0fc1b451
AK		 2269 }
2270
2271 /* indirect blocks */
2272 meta_blocks = 1;
2273 /* double indirect blocks */
2274 meta_blocks += 1 + (1LL << (bits-2));
2275 /* tripple indirect blocks */
2276 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2277
2278 upper_limit -= meta_blocks;
2279 upper_limit <<= bits;
ac27a0ec
DK		 2280
2281 res += 1LL << (bits-2);
2282 res += 1LL << (2*(bits-2));
2283 res += 1LL << (3*(bits-2));
2284 res <<= bits;
2285 if (res > upper_limit)
2286 res = upper_limit;
0fc1b451
AK		 2287
2288 if (res > MAX_LFS_FILESIZE)
2289 res = MAX_LFS_FILESIZE;
2290
ac27a0ec
DK		 2291 return res;
2292}
2293
617ba13b 2294static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2295 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2296{
617ba13b 2297 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2298 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2299 int has_super = 0;
2300
2301 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2302
617ba13b 2303 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2304 nr < first_meta_bg)
70bbb3e0 2305 return logical_sb_block + nr + 1;
ac27a0ec 2306 bg = sbi->s_desc_per_block * nr;
617ba13b 2307 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2308 has_super = 1;
0b8e58a1 2309
617ba13b 2310 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2311}
2312
c9de560d
AT		 2313/**
2314 * ext4_get_stripe_size: Get the stripe size.
2315 * @sbi: In memory super block info
2316 *
2317 * If we have specified it via mount option, then
2318 * use the mount option value. If the value specified at mount time is
2319 * greater than the blocks per group use the super block value.
2320 * If the super block value is greater than blocks per group return 0.
2321 * Allocator needs it be less than blocks per group.
2322 *
2323 */
2324static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2325{
2326 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2327 unsigned long stripe_width =
2328 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2329 int ret;
c9de560d
AT		 2330
2331 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2332 ret = sbi->s_stripe;
2333 else if (stripe_width <= sbi->s_blocks_per_group)
2334 ret = stripe_width;
2335 else if (stride <= sbi->s_blocks_per_group)
2336 ret = stride;
2337 else
2338 ret = 0;
c9de560d 2339
3eb08658
DE		 2340 /*
2341 * If the stripe width is 1, this makes no sense and
2342 * we set it to 0 to turn off stripe handling code.
2343 */
2344 if (ret <= 1)
2345 ret = 0;
c9de560d 2346
3eb08658 2347 return ret;
c9de560d 2348}
ac27a0ec 2349
3197ebdb
TT		 2350/* sysfs supprt */
2351
2352struct ext4_attr {
2353 struct attribute attr;
2354 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2355 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2356 const char *, size_t);
2357 int offset;
2358};
2359
2360static int parse_strtoul(const char *buf,
2361 unsigned long max, unsigned long *value)
2362{
2363 char *endp;
2364
e7d2860b
AGR		 2365 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2366 endp = skip_spaces(endp);
3197ebdb
TT		 2367 if (*endp || *value > max)
2368 return -EINVAL;
2369
2370 return 0;
2371}
2372
2373static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2374 struct ext4_sb_info *sbi,
2375 char *buf)
2376{
2377 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2378 (s64) EXT4_C2B(sbi,
2379 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2380}
2381
2382static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2383 struct ext4_sb_info *sbi, char *buf)
2384{
2385 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2386
f613dfcb
TT		 2387 if (!sb->s_bdev->bd_part)
2388 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2389 return snprintf(buf, PAGE_SIZE, "%lu\n",
2390 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2391 sbi->s_sectors_written_start) >> 1);
2392}
2393
2394static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2395 struct ext4_sb_info *sbi, char *buf)
2396{
2397 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2398
f613dfcb
TT		 2399 if (!sb->s_bdev->bd_part)
2400 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2401 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2402 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2403 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2404 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2405}
2406
2407static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2408 struct ext4_sb_info *sbi,
2409 const char *buf, size_t count)
2410{
2411 unsigned long t;
2412
2413 if (parse_strtoul(buf, 0x40000000, &t))
2414 return -EINVAL;
2415
5dbd571d 2416 if (t && !is_power_of_2(t))
3197ebdb
TT		 2417 return -EINVAL;
2418
2419 sbi->s_inode_readahead_blks = t;
2420 return count;
2421}
2422
2423static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2424 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2425{
2426 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2427
2428 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2429}
2430
2431static ssize_t sbi_ui_store(struct ext4_attr *a,
2432 struct ext4_sb_info *sbi,
2433 const char *buf, size_t count)
2434{
2435 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2436 unsigned long t;
2437
2438 if (parse_strtoul(buf, 0xffffffff, &t))
2439 return -EINVAL;
2440 *ui = t;
2441 return count;
2442}
2443
2c0544b2
TT		 2444static ssize_t trigger_test_error(struct ext4_attr *a,
2445 struct ext4_sb_info *sbi,
2446 const char *buf, size_t count)
2447{
2448 int len = count;
2449
2450 if (!capable(CAP_SYS_ADMIN))
2451 return -EPERM;
2452
2453 if (len && buf[len-1] == '\n')
2454 len--;
2455
2456 if (len)
2457 ext4_error(sbi->s_sb, "%.*s", len, buf);
2458 return count;
2459}
2460
3197ebdb
TT		 2461#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2462static struct ext4_attr ext4_attr_##_name = { \
2463 .attr = {.name = __stringify(_name), .mode = _mode }, \
2464 .show = _show, \
2465 .store = _store, \
2466 .offset = offsetof(struct ext4_sb_info, _elname), \
2467}
2468#define EXT4_ATTR(name, mode, show, store) \
2469static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2470
857ac889 2471#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2472#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2473#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2474#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2475 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2476#define ATTR_LIST(name) &ext4_attr_##name.attr
2477
2478EXT4_RO_ATTR(delayed_allocation_blocks);
2479EXT4_RO_ATTR(session_write_kbytes);
2480EXT4_RO_ATTR(lifetime_write_kbytes);
2481EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2482 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2483EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2484EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2485EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2486EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2487EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2488EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2489EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2490EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
67a5da56 2491EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2492EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2493
2494static struct attribute *ext4_attrs[] = {
2495 ATTR_LIST(delayed_allocation_blocks),
2496 ATTR_LIST(session_write_kbytes),
2497 ATTR_LIST(lifetime_write_kbytes),
2498 ATTR_LIST(inode_readahead_blks),
11013911 2499 ATTR_LIST(inode_goal),
3197ebdb
TT		 2500 ATTR_LIST(mb_stats),
2501 ATTR_LIST(mb_max_to_scan),
2502 ATTR_LIST(mb_min_to_scan),
2503 ATTR_LIST(mb_order2_req),
2504 ATTR_LIST(mb_stream_req),
2505 ATTR_LIST(mb_group_prealloc),
55138e0b 2506 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2507 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2508 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2509 NULL,
2510};
2511
857ac889
LC		 2512/* Features this copy of ext4 supports */
2513EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2514EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2515EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2516
2517static struct attribute *ext4_feat_attrs[] = {
2518 ATTR_LIST(lazy_itable_init),
27ee40df 2519 ATTR_LIST(batched_discard),
5e7bbef1 2520 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2521 NULL,
2522};
2523
3197ebdb
TT		 2524static ssize_t ext4_attr_show(struct kobject *kobj,
2525 struct attribute *attr, char *buf)
2526{
2527 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2528 s_kobj);
2529 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2530
2531 return a->show ? a->show(a, sbi, buf) : 0;
2532}
2533
2534static ssize_t ext4_attr_store(struct kobject *kobj,
2535 struct attribute *attr,
2536 const char *buf, size_t len)
2537{
2538 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2539 s_kobj);
2540 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2541
2542 return a->store ? a->store(a, sbi, buf, len) : 0;
2543}
2544
2545static void ext4_sb_release(struct kobject *kobj)
2546{
2547 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2548 s_kobj);
2549 complete(&sbi->s_kobj_unregister);
2550}
2551
52cf25d0 2552static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2553 .show = ext4_attr_show,
2554 .store = ext4_attr_store,
2555};
2556
2557static struct kobj_type ext4_ktype = {
2558 .default_attrs = ext4_attrs,
2559 .sysfs_ops = &ext4_attr_ops,
2560 .release = ext4_sb_release,
2561};
2562
857ac889
LC		 2563static void ext4_feat_release(struct kobject *kobj)
2564{
2565 complete(&ext4_feat->f_kobj_unregister);
2566}
2567
2568static struct kobj_type ext4_feat_ktype = {
2569 .default_attrs = ext4_feat_attrs,
2570 .sysfs_ops = &ext4_attr_ops,
2571 .release = ext4_feat_release,
2572};
2573
a13fb1a4
ES		 2574/*
2575 * Check whether this filesystem can be mounted based on
2576 * the features present and the RDONLY/RDWR mount requested.
2577 * Returns 1 if this filesystem can be mounted as requested,
2578 * 0 if it cannot be.
2579 */
2580static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2581{
2582 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2583 ext4_msg(sb, KERN_ERR,
2584 "Couldn't mount because of "
2585 "unsupported optional features (%x)",
2586 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2587 ~EXT4_FEATURE_INCOMPAT_SUPP));
2588 return 0;
2589 }
2590
2591 if (readonly)
2592 return 1;
2593
2594 /* Check that feature set is OK for a read-write mount */
2595 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2596 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2597 "unsupported optional features (%x)",
2598 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2599 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2600 return 0;
2601 }
2602 /*
2603 * Large file size enabled file system can only be mounted
2604 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2605 */
2606 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2607 if (sizeof(blkcnt_t) < sizeof(u64)) {
2608 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2609 "cannot be mounted RDWR without "
2610 "CONFIG_LBDAF");
2611 return 0;
2612 }
2613 }
bab08ab9
TT		 2614 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2615 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2616 ext4_msg(sb, KERN_ERR,
2617 "Can't support bigalloc feature without "
2618 "extents feature\n");
2619 return 0;
2620 }
7c319d32
AK		 2621
2622#ifndef CONFIG_QUOTA
2623 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2624 !readonly) {
2625 ext4_msg(sb, KERN_ERR,
2626 "Filesystem with quota feature cannot be mounted RDWR "
2627 "without CONFIG_QUOTA");
2628 return 0;
2629 }
2630#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2631 return 1;
2632}
2633
66e61a9e
TT		 2634/*
2635 * This function is called once a day if we have errors logged
2636 * on the file system
2637 */
2638static void print_daily_error_info(unsigned long arg)
2639{
2640 struct super_block *sb = (struct super_block *) arg;
2641 struct ext4_sb_info *sbi;
2642 struct ext4_super_block *es;
2643
2644 sbi = EXT4_SB(sb);
2645 es = sbi->s_es;
2646
2647 if (es->s_error_count)
2648 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2649 le32_to_cpu(es->s_error_count));
2650 if (es->s_first_error_time) {
2651 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2652 sb->s_id, le32_to_cpu(es->s_first_error_time),
2653 (int) sizeof(es->s_first_error_func),
2654 es->s_first_error_func,
2655 le32_to_cpu(es->s_first_error_line));
2656 if (es->s_first_error_ino)
2657 printk(": inode %u",
2658 le32_to_cpu(es->s_first_error_ino));
2659 if (es->s_first_error_block)
2660 printk(": block %llu", (unsigned long long)
2661 le64_to_cpu(es->s_first_error_block));
2662 printk("\n");
2663 }
2664 if (es->s_last_error_time) {
2665 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2666 sb->s_id, le32_to_cpu(es->s_last_error_time),
2667 (int) sizeof(es->s_last_error_func),
2668 es->s_last_error_func,
2669 le32_to_cpu(es->s_last_error_line));
2670 if (es->s_last_error_ino)
2671 printk(": inode %u",
2672 le32_to_cpu(es->s_last_error_ino));
2673 if (es->s_last_error_block)
2674 printk(": block %llu", (unsigned long long)
2675 le64_to_cpu(es->s_last_error_block));
2676 printk("\n");
2677 }
2678 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2679}
2680
bfff6873
LC		 2681/* Find next suitable group and run ext4_init_inode_table */
2682static int ext4_run_li_request(struct ext4_li_request *elr)
2683{
2684 struct ext4_group_desc *gdp = NULL;
2685 ext4_group_t group, ngroups;
2686 struct super_block *sb;
2687 unsigned long timeout = 0;
2688 int ret = 0;
2689
2690 sb = elr->lr_super;
2691 ngroups = EXT4_SB(sb)->s_groups_count;
2692
8e8ad8a5 2693 sb_start_write(sb);
bfff6873
LC		 2694 for (group = elr->lr_next_group; group < ngroups; group++) {
2695 gdp = ext4_get_group_desc(sb, group, NULL);
2696 if (!gdp) {
2697 ret = 1;
2698 break;
2699 }
2700
2701 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2702 break;
2703 }
2704
7f511862 2705 if (group >= ngroups)
bfff6873
LC		 2706 ret = 1;
2707
2708 if (!ret) {
2709 timeout = jiffies;
2710 ret = ext4_init_inode_table(sb, group,
2711 elr->lr_timeout ? 0 : 1);
2712 if (elr->lr_timeout == 0) {
51ce6511
LC		 2713 timeout = (jiffies - timeout) *
2714 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2715 elr->lr_timeout = timeout;
2716 }
2717 elr->lr_next_sched = jiffies + elr->lr_timeout;
2718 elr->lr_next_group = group + 1;
2719 }
8e8ad8a5 2720 sb_end_write(sb);
bfff6873
LC		 2721
2722 return ret;
2723}
2724
2725/*
2726 * Remove lr_request from the list_request and free the
4ed5c033 2727 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2728 */
2729static void ext4_remove_li_request(struct ext4_li_request *elr)
2730{
2731 struct ext4_sb_info *sbi;
2732
2733 if (!elr)
2734 return;
2735
2736 sbi = elr->lr_sbi;
2737
2738 list_del(&elr->lr_request);
2739 sbi->s_li_request = NULL;
2740 kfree(elr);
2741}
2742
2743static void ext4_unregister_li_request(struct super_block *sb)
2744{
1bb933fb
LC		 2745 mutex_lock(&ext4_li_mtx);
2746 if (!ext4_li_info) {
2747 mutex_unlock(&ext4_li_mtx);
bfff6873 2748 return;
1bb933fb 2749 }
bfff6873
LC		 2750
2751 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2752 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2753 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2754 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2755}
2756
8f1f7453
ES		 2757static struct task_struct *ext4_lazyinit_task;
2758
bfff6873
LC		 2759/*
2760 * This is the function where ext4lazyinit thread lives. It walks
2761 * through the request list searching for next scheduled filesystem.
2762 * When such a fs is found, run the lazy initialization request
2763 * (ext4_rn_li_request) and keep track of the time spend in this
2764 * function. Based on that time we compute next schedule time of
2765 * the request. When walking through the list is complete, compute
2766 * next waking time and put itself into sleep.
2767 */
2768static int ext4_lazyinit_thread(void *arg)
2769{
2770 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2771 struct list_head *pos, *n;
2772 struct ext4_li_request *elr;
4ed5c033 2773 unsigned long next_wakeup, cur;
bfff6873
LC		 2774
2775 BUG_ON(NULL == eli);
2776
bfff6873
LC		 2777cont_thread:
2778 while (true) {
2779 next_wakeup = MAX_JIFFY_OFFSET;
2780
2781 mutex_lock(&eli->li_list_mtx);
2782 if (list_empty(&eli->li_request_list)) {
2783 mutex_unlock(&eli->li_list_mtx);
2784 goto exit_thread;
2785 }
2786
2787 list_for_each_safe(pos, n, &eli->li_request_list) {
2788 elr = list_entry(pos, struct ext4_li_request,
2789 lr_request);
2790
b2c78cd0
TT		 2791 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2792 if (ext4_run_li_request(elr) != 0) {
2793 /* error, remove the lazy_init job */
2794 ext4_remove_li_request(elr);
2795 continue;
2796 }
bfff6873
LC		 2797 }
2798
2799 if (time_before(elr->lr_next_sched, next_wakeup))
2800 next_wakeup = elr->lr_next_sched;
2801 }
2802 mutex_unlock(&eli->li_list_mtx);
2803
a0acae0e 2804 try_to_freeze();
bfff6873 2805
4ed5c033
LC		 2806 cur = jiffies;
2807 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2808 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2809 cond_resched();
2810 continue;
2811 }
2812
4ed5c033
LC		 2813 schedule_timeout_interruptible(next_wakeup - cur);
2814
8f1f7453
ES		 2815 if (kthread_should_stop()) {
2816 ext4_clear_request_list();
2817 goto exit_thread;
2818 }
bfff6873
LC		 2819 }
2820
2821exit_thread:
2822 /*
2823 * It looks like the request list is empty, but we need
2824 * to check it under the li_list_mtx lock, to prevent any
2825 * additions into it, and of course we should lock ext4_li_mtx
2826 * to atomically free the list and ext4_li_info, because at
2827 * this point another ext4 filesystem could be registering
2828 * new one.
2829 */
2830 mutex_lock(&ext4_li_mtx);
2831 mutex_lock(&eli->li_list_mtx);
2832 if (!list_empty(&eli->li_request_list)) {
2833 mutex_unlock(&eli->li_list_mtx);
2834 mutex_unlock(&ext4_li_mtx);
2835 goto cont_thread;
2836 }
2837 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2838 kfree(ext4_li_info);
2839 ext4_li_info = NULL;
2840 mutex_unlock(&ext4_li_mtx);
2841
2842 return 0;
2843}
2844
2845static void ext4_clear_request_list(void)
2846{
2847 struct list_head *pos, *n;
2848 struct ext4_li_request *elr;
2849
2850 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2851 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2852 elr = list_entry(pos, struct ext4_li_request,
2853 lr_request);
2854 ext4_remove_li_request(elr);
2855 }
2856 mutex_unlock(&ext4_li_info->li_list_mtx);
2857}
2858
2859static int ext4_run_lazyinit_thread(void)
2860{
8f1f7453
ES		 2861 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2862 ext4_li_info, "ext4lazyinit");
2863 if (IS_ERR(ext4_lazyinit_task)) {
2864 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2865 ext4_clear_request_list();
bfff6873
LC		 2866 kfree(ext4_li_info);
2867 ext4_li_info = NULL;
92b97816 2868 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2869 "initialization thread\n",
2870 err);
2871 return err;
2872 }
2873 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2874 return 0;
2875}
2876
2877/*
2878 * Check whether it make sense to run itable init. thread or not.
2879 * If there is at least one uninitialized inode table, return
2880 * corresponding group number, else the loop goes through all
2881 * groups and return total number of groups.
2882 */
2883static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2884{
2885 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2886 struct ext4_group_desc *gdp = NULL;
2887
2888 for (group = 0; group < ngroups; group++) {
2889 gdp = ext4_get_group_desc(sb, group, NULL);
2890 if (!gdp)
2891 continue;
2892
2893 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2894 break;
2895 }
2896
2897 return group;
2898}
2899
2900static int ext4_li_info_new(void)
2901{
2902 struct ext4_lazy_init *eli = NULL;
2903
2904 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2905 if (!eli)
2906 return -ENOMEM;
2907
bfff6873
LC		 2908 INIT_LIST_HEAD(&eli->li_request_list);
2909 mutex_init(&eli->li_list_mtx);
2910
bfff6873
LC		 2911 eli->li_state |= EXT4_LAZYINIT_QUIT;
2912
2913 ext4_li_info = eli;
2914
2915 return 0;
2916}
2917
2918static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2919 ext4_group_t start)
2920{
2921 struct ext4_sb_info *sbi = EXT4_SB(sb);
2922 struct ext4_li_request *elr;
2923 unsigned long rnd;
2924
2925 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2926 if (!elr)
2927 return NULL;
2928
2929 elr->lr_super = sb;
2930 elr->lr_sbi = sbi;
2931 elr->lr_next_group = start;
2932
2933 /*
2934 * Randomize first schedule time of the request to
2935 * spread the inode table initialization requests
2936 * better.
2937 */
2938 get_random_bytes(&rnd, sizeof(rnd));
2939 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2940 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2941
2942 return elr;
2943}
2944
7f511862
TT		 2945int ext4_register_li_request(struct super_block *sb,
2946 ext4_group_t first_not_zeroed)
bfff6873
LC		 2947{
2948 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 2949 struct ext4_li_request *elr = NULL;
bfff6873 2950 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2951 int ret = 0;
bfff6873 2952
7f511862 2953 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 2954 if (sbi->s_li_request != NULL) {
2955 /*
2956 * Reset timeout so it can be computed again, because
2957 * s_li_wait_mult might have changed.
2958 */
2959 sbi->s_li_request->lr_timeout = 0;
7f511862 2960 goto out;
51ce6511 2961 }
bfff6873
LC		 2962
2963 if (first_not_zeroed == ngroups ||
2964 (sb->s_flags & MS_RDONLY) ||
55ff3840 2965 !test_opt(sb, INIT_INODE_TABLE))
7f511862 2966 goto out;
bfff6873
LC		 2967
2968 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 2969 if (!elr) {
2970 ret = -ENOMEM;
2971 goto out;
2972 }
bfff6873
LC		 2973
2974 if (NULL == ext4_li_info) {
2975 ret = ext4_li_info_new();
2976 if (ret)
2977 goto out;
2978 }
2979
2980 mutex_lock(&ext4_li_info->li_list_mtx);
2981 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2982 mutex_unlock(&ext4_li_info->li_list_mtx);
2983
2984 sbi->s_li_request = elr;
46e4690b
TM		 2985 /*
2986 * set elr to NULL here since it has been inserted to
2987 * the request_list and the removal and free of it is
2988 * handled by ext4_clear_request_list from now on.
2989 */
2990 elr = NULL;
bfff6873
LC		 2991
2992 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2993 ret = ext4_run_lazyinit_thread();
2994 if (ret)
2995 goto out;
2996 }
bfff6873 2997out:
beed5ecb
NK		 2998 mutex_unlock(&ext4_li_mtx);
2999 if (ret)
bfff6873 3000 kfree(elr);
bfff6873
LC		 3001 return ret;
3002}
3003
3004/*
3005 * We do not need to lock anything since this is called on
3006 * module unload.
3007 */
3008static void ext4_destroy_lazyinit_thread(void)
3009{
3010 /*
3011 * If thread exited earlier
3012 * there's nothing to be done.
3013 */
8f1f7453 3014 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3015 return;
3016
8f1f7453 3017 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3018}
3019
25ed6e8a
DW		 3020static int set_journal_csum_feature_set(struct super_block *sb)
3021{
3022 int ret = 1;
3023 int compat, incompat;
3024 struct ext4_sb_info *sbi = EXT4_SB(sb);
3025
3026 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3027 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3028 /* journal checksum v2 */
3029 compat = 0;
3030 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3031 } else {
3032 /* journal checksum v1 */
3033 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3034 incompat = 0;
3035 }
3036
3037 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3038 ret = jbd2_journal_set_features(sbi->s_journal,
3039 compat, 0,
3040 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3041 incompat);
3042 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3043 ret = jbd2_journal_set_features(sbi->s_journal,
3044 compat, 0,
3045 incompat);
3046 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3047 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3048 } else {
3049 jbd2_journal_clear_features(sbi->s_journal,
3050 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3051 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3052 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3053 }
3054
3055 return ret;
3056}
3057
952fc18e
TT		 3058/*
3059 * Note: calculating the overhead so we can be compatible with
3060 * historical BSD practice is quite difficult in the face of
3061 * clusters/bigalloc. This is because multiple metadata blocks from
3062 * different block group can end up in the same allocation cluster.
3063 * Calculating the exact overhead in the face of clustered allocation
3064 * requires either O(all block bitmaps) in memory or O(number of block
3065 * groups**2) in time. We will still calculate the superblock for
3066 * older file systems --- and if we come across with a bigalloc file
3067 * system with zero in s_overhead_clusters the estimate will be close to
3068 * correct especially for very large cluster sizes --- but for newer
3069 * file systems, it's better to calculate this figure once at mkfs
3070 * time, and store it in the superblock. If the superblock value is
3071 * present (even for non-bigalloc file systems), we will use it.
3072 */
3073static int count_overhead(struct super_block *sb, ext4_group_t grp,
3074 char *buf)
3075{
3076 struct ext4_sb_info *sbi = EXT4_SB(sb);
3077 struct ext4_group_desc *gdp;
3078 ext4_fsblk_t first_block, last_block, b;
3079 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3080 int s, j, count = 0;
3081
0548bbb8
TT		 3082 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3083 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3084 sbi->s_itb_per_group + 2);
3085
952fc18e
TT		 3086 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3087 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3088 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3089 for (i = 0; i < ngroups; i++) {
3090 gdp = ext4_get_group_desc(sb, i, NULL);
3091 b = ext4_block_bitmap(sb, gdp);
3092 if (b >= first_block && b <= last_block) {
3093 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3094 count++;
3095 }
3096 b = ext4_inode_bitmap(sb, gdp);
3097 if (b >= first_block && b <= last_block) {
3098 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3099 count++;
3100 }
3101 b = ext4_inode_table(sb, gdp);
3102 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3103 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3104 int c = EXT4_B2C(sbi, b - first_block);
3105 ext4_set_bit(c, buf);
3106 count++;
3107 }
3108 if (i != grp)
3109 continue;
3110 s = 0;
3111 if (ext4_bg_has_super(sb, grp)) {
3112 ext4_set_bit(s++, buf);
3113 count++;
3114 }
3115 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3116 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3117 count++;
3118 }
3119 }
3120 if (!count)
3121 return 0;
3122 return EXT4_CLUSTERS_PER_GROUP(sb) -
3123 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3124}
3125
3126/*
3127 * Compute the overhead and stash it in sbi->s_overhead
3128 */
3129int ext4_calculate_overhead(struct super_block *sb)
3130{
3131 struct ext4_sb_info *sbi = EXT4_SB(sb);
3132 struct ext4_super_block *es = sbi->s_es;
3133 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3134 ext4_fsblk_t overhead = 0;
3135 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3136
952fc18e
TT		 3137 if (!buf)
3138 return -ENOMEM;
3139
3140 /*
3141 * Compute the overhead (FS structures). This is constant
3142 * for a given filesystem unless the number of block groups
3143 * changes so we cache the previous value until it does.
3144 */
3145
3146 /*
3147 * All of the blocks before first_data_block are overhead
3148 */
3149 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3150
3151 /*
3152 * Add the overhead found in each block group
3153 */
3154 for (i = 0; i < ngroups; i++) {
3155 int blks;
3156
3157 blks = count_overhead(sb, i, buf);
3158 overhead += blks;
3159 if (blks)
3160 memset(buf, 0, PAGE_SIZE);
3161 cond_resched();
3162 }
0875a2b4
ES		 3163 /* Add the journal blocks as well */
3164 if (sbi->s_journal)
3165 overhead += EXT4_B2C(sbi, sbi->s_journal->j_maxlen);
3166
952fc18e
TT		 3167 sbi->s_overhead = overhead;
3168 smp_wmb();
3169 free_page((unsigned long) buf);
3170 return 0;
3171}
3172
2b2d6d01 3173static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3174{
d4c402d9 3175 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3176 struct buffer_head *bh;
617ba13b
MC		 3177 struct ext4_super_block *es = NULL;
3178 struct ext4_sb_info *sbi;
3179 ext4_fsblk_t block;
3180 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3181 ext4_fsblk_t logical_sb_block;
ac27a0ec 3182 unsigned long offset = 0;
ac27a0ec
DK		 3183 unsigned long journal_devnum = 0;
3184 unsigned long def_mount_opts;
3185 struct inode *root;
9f6200bb 3186 char *cp;
0390131b 3187 const char *descr;
dcc7dae3 3188 int ret = -ENOMEM;
281b5995 3189 int blocksize, clustersize;
4ec11028
TT		 3190 unsigned int db_count;
3191 unsigned int i;
281b5995 3192 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3193 __u64 blocks_count;
07aa2ea1 3194 int err = 0;
b3881f74 3195 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3196 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3197
3198 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3199 if (!sbi)
dcc7dae3 3200 goto out_free_orig;
705895b6
PE		 3201
3202 sbi->s_blockgroup_lock =
3203 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3204 if (!sbi->s_blockgroup_lock) {
3205 kfree(sbi);
dcc7dae3 3206 goto out_free_orig;
705895b6 3207 }
ac27a0ec 3208 sb->s_fs_info = sbi;
2c0544b2 3209 sbi->s_sb = sb;
240799cd 3210 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3211 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3212 if (sb->s_bdev->bd_part)
3213 sbi->s_sectors_written_start =
3214 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3215
9f6200bb
TT		 3216 /* Cleanup superblock name */
3217 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3218 *cp = '!';
3219
07aa2ea1 3220 /* -EINVAL is default */
dcc7dae3 3221 ret = -EINVAL;
617ba13b 3222 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3223 if (!blocksize) {
b31e1552 3224 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3225 goto out_fail;
3226 }
3227
3228 /*
617ba13b 3229 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3230 * block sizes. We need to calculate the offset from buffer start.
3231 */
617ba13b 3232 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3233 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3234 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3235 } else {
70bbb3e0 3236 logical_sb_block = sb_block;
ac27a0ec
DK		 3237 }
3238
70bbb3e0 3239 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3240 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3241 goto out_fail;
3242 }
3243 /*
3244 * Note: s_es must be initialized as soon as possible because
617ba13b 3245 * some ext4 macro-instructions depend on its value
ac27a0ec 3246 */
2716b802 3247 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3248 sbi->s_es = es;
3249 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3250 if (sb->s_magic != EXT4_SUPER_MAGIC)
3251 goto cantfind_ext4;
afc32f7e 3252 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3253
feb0ab32
DW		 3254 /* Warn if metadata_csum and gdt_csum are both set. */
3255 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3256 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3257 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3258 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3259 "redundant flags; please run fsck.");
3260
d25425f8
DW		 3261 /* Check for a known checksum algorithm */
3262 if (!ext4_verify_csum_type(sb, es)) {
3263 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3264 "unknown checksum algorithm.");
3265 silent = 1;
3266 goto cantfind_ext4;
3267 }
3268
0441984a
DW		 3269 /* Load the checksum driver */
3270 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3271 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3272 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3273 if (IS_ERR(sbi->s_chksum_driver)) {
3274 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3275 ret = PTR_ERR(sbi->s_chksum_driver);
3276 sbi->s_chksum_driver = NULL;
3277 goto failed_mount;
3278 }
3279 }
3280
a9c47317
DW		 3281 /* Check superblock checksum */
3282 if (!ext4_superblock_csum_verify(sb, es)) {
3283 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3284 "invalid superblock checksum. Run e2fsck?");
3285 silent = 1;
3286 goto cantfind_ext4;
3287 }
3288
3289 /* Precompute checksum seed for all metadata */
3290 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3291 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3292 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3293 sizeof(es->s_uuid));
3294
ac27a0ec
DK		 3295 /* Set defaults before we parse the mount options */
3296 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3297 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3298 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3299 set_opt(sb, DEBUG);
87f26807 3300 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3301 set_opt(sb, GRPID);
617ba13b 3302 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3303 set_opt(sb, NO_UID32);
ea663336 3304 /* xattr user namespace & acls are now defaulted on */
ea663336 3305 set_opt(sb, XATTR_USER);
03010a33 3306#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3307 set_opt(sb, POSIX_ACL);
2e7842b8 3308#endif
617ba13b 3309 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3310 set_opt(sb, JOURNAL_DATA);
617ba13b 3311 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3312 set_opt(sb, ORDERED_DATA);
617ba13b 3313 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3314 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3315
3316 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3317 set_opt(sb, ERRORS_PANIC);
bb4f397a 3318 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3319 set_opt(sb, ERRORS_CONT);
bb4f397a 3320 else
fd8c37ec 3321 set_opt(sb, ERRORS_RO);
8b67f04a 3322 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3323 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3324 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3325 set_opt(sb, DISCARD);
ac27a0ec 3326
08cefc7a
EB		 3327 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3328 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3329 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3330 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3331 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3332
8b67f04a 3333 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3334 set_opt(sb, BARRIER);
ac27a0ec 3335
dd919b98
AK		 3336 /*
3337 * enable delayed allocation by default
3338 * Use -o nodelalloc to turn it off
3339 */
bc0b75f7 3340 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3341 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3342 set_opt(sb, DELALLOC);
dd919b98 3343
51ce6511
LC		 3344 /*
3345 * set default s_li_wait_mult for lazyinit, for the case there is
3346 * no mount option specified.
3347 */
3348 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3349
8b67f04a 3350 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3351 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3352 ext4_msg(sb, KERN_WARNING,
3353 "failed to parse options in superblock: %s",
3354 sbi->s_es->s_mount_opts);
3355 }
5a916be1 3356 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3357 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3358 &journal_ioprio, 0))
ac27a0ec
DK		 3359 goto failed_mount;
3360
56889787
TT		 3361 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3362 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3363 "with data=journal disables delayed "
3364 "allocation and O_DIRECT support!\n");
3365 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3366 ext4_msg(sb, KERN_ERR, "can't mount with "
3367 "both data=journal and delalloc");
3368 goto failed_mount;
3369 }
3370 if (test_opt(sb, DIOREAD_NOLOCK)) {
3371 ext4_msg(sb, KERN_ERR, "can't mount with "
3372 "both data=journal and delalloc");
3373 goto failed_mount;
3374 }
3375 if (test_opt(sb, DELALLOC))
3376 clear_opt(sb, DELALLOC);
3377 }
3378
ac27a0ec 3379 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3380 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3381
617ba13b
MC		 3382 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3383 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3384 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3385 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3386 ext4_msg(sb, KERN_WARNING,
3387 "feature flags set on rev 0 fs, "
3388 "running e2fsck is recommended");
469108ff 3389
2035e776
TT		 3390 if (IS_EXT2_SB(sb)) {
3391 if (ext2_feature_set_ok(sb))
3392 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3393 "using the ext4 subsystem");
3394 else {
3395 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3396 "to feature incompatibilities");
3397 goto failed_mount;
3398 }
3399 }
3400
3401 if (IS_EXT3_SB(sb)) {
3402 if (ext3_feature_set_ok(sb))
3403 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3404 "using the ext4 subsystem");
3405 else {
3406 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3407 "to feature incompatibilities");
3408 goto failed_mount;
3409 }
3410 }
3411
ac27a0ec
DK		 3412 /*
3413 * Check feature flags regardless of the revision level, since we
3414 * previously didn't change the revision level when setting the flags,
3415 * so there is a chance incompat flags are set on a rev 0 filesystem.
3416 */
a13fb1a4 3417 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3418 goto failed_mount;
a13fb1a4 3419
261cb20c 3420 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3421 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3422 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3423 ext4_msg(sb, KERN_ERR,
3424 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3425 goto failed_mount;
3426 }
3427
ac27a0ec 3428 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3429 /* Validate the filesystem blocksize */
3430 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3431 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3432 blocksize);
ac27a0ec
DK		 3433 goto failed_mount;
3434 }
3435
2b2d6d01 3436 brelse(bh);
70bbb3e0
AM		 3437 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3438 offset = do_div(logical_sb_block, blocksize);
3439 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3440 if (!bh) {
b31e1552
ES		 3441 ext4_msg(sb, KERN_ERR,
3442 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3443 goto failed_mount;
3444 }
2716b802 3445 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3446 sbi->s_es = es;
617ba13b 3447 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3448 ext4_msg(sb, KERN_ERR,
3449 "Magic mismatch, very weird!");
ac27a0ec
DK		 3450 goto failed_mount;
3451 }
3452 }
3453
a13fb1a4
ES		 3454 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3455 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3456 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3457 has_huge_files);
3458 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3459
617ba13b
MC		 3460 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3461 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3462 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3463 } else {
3464 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3465 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3466 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3467 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3468 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3469 ext4_msg(sb, KERN_ERR,
3470 "unsupported inode size: %d",
2b2d6d01 3471 sbi->s_inode_size);
ac27a0ec
DK		 3472 goto failed_mount;
3473 }
ef7f3835
KS		 3474 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3475 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3476 }
0b8e58a1 3477
0d1ee42f
AR		 3478 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3479 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3480 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3481 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3482 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3483 ext4_msg(sb, KERN_ERR,
3484 "unsupported descriptor size %lu",
0d1ee42f
AR		 3485 sbi->s_desc_size);
3486 goto failed_mount;
3487 }
3488 } else
3489 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3490
ac27a0ec 3491 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3492 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3493 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3494 goto cantfind_ext4;
0b8e58a1 3495
617ba13b 3496 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3497 if (sbi->s_inodes_per_block == 0)
617ba13b 3498 goto cantfind_ext4;
ac27a0ec
DK		 3499 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3500 sbi->s_inodes_per_block;
0d1ee42f 3501 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3502 sbi->s_sbh = bh;
3503 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3504 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3505 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3506
2b2d6d01 3507 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3508 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3509 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3510 i = le32_to_cpu(es->s_flags);
3511 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3512 sbi->s_hash_unsigned = 3;
3513 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3514#ifdef __CHAR_UNSIGNED__
3515 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3516 sbi->s_hash_unsigned = 3;
3517#else
3518 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3519#endif
f99b2589 3520 }
ac27a0ec 3521
281b5995
TT		 3522 /* Handle clustersize */
3523 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3524 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3525 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3526 if (has_bigalloc) {
3527 if (clustersize < blocksize) {
3528 ext4_msg(sb, KERN_ERR,
3529 "cluster size (%d) smaller than "
3530 "block size (%d)", clustersize, blocksize);
3531 goto failed_mount;
3532 }
3533 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3534 le32_to_cpu(es->s_log_block_size);
3535 sbi->s_clusters_per_group =
3536 le32_to_cpu(es->s_clusters_per_group);
3537 if (sbi->s_clusters_per_group > blocksize * 8) {
3538 ext4_msg(sb, KERN_ERR,
3539 "#clusters per group too big: %lu",
3540 sbi->s_clusters_per_group);
3541 goto failed_mount;
3542 }
3543 if (sbi->s_blocks_per_group !=
3544 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3545 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3546 "clusters per group (%lu) inconsistent",
3547 sbi->s_blocks_per_group,
3548 sbi->s_clusters_per_group);
3549 goto failed_mount;
3550 }
3551 } else {
3552 if (clustersize != blocksize) {
3553 ext4_warning(sb, "fragment/cluster size (%d) != "
3554 "block size (%d)", clustersize,
3555 blocksize);
3556 clustersize = blocksize;
3557 }
3558 if (sbi->s_blocks_per_group > blocksize * 8) {
3559 ext4_msg(sb, KERN_ERR,
3560 "#blocks per group too big: %lu",
3561 sbi->s_blocks_per_group);
3562 goto failed_mount;
3563 }
3564 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3565 sbi->s_cluster_bits = 0;
ac27a0ec 3566 }
281b5995
TT		 3567 sbi->s_cluster_ratio = clustersize / blocksize;
3568
ac27a0ec 3569 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3570 ext4_msg(sb, KERN_ERR,
3571 "#inodes per group too big: %lu",
2b2d6d01 3572 sbi->s_inodes_per_group);
ac27a0ec
DK		 3573 goto failed_mount;
3574 }
3575
bf43d84b
ES		 3576 /*
3577 * Test whether we have more sectors than will fit in sector_t,
3578 * and whether the max offset is addressable by the page cache.
3579 */
5a9ae68a 3580 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3581 ext4_blocks_count(es));
5a9ae68a 3582 if (err) {
b31e1552 3583 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3584 " too large to mount safely on this system");
ac27a0ec 3585 if (sizeof(sector_t) < 8)
90c699a9 3586 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3587 goto failed_mount;
3588 }
3589
617ba13b
MC		 3590 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3591 goto cantfind_ext4;
e7c95593 3592
0f2ddca6
FTN		 3593 /* check blocks count against device size */
3594 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3595 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3596 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3597 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3598 ext4_blocks_count(es), blocks_count);
3599 goto failed_mount;
3600 }
3601
0b8e58a1
AD		 3602 /*
3603 * It makes no sense for the first data block to be beyond the end
3604 * of the filesystem.
3605 */
3606 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3607 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3608 "block %u is beyond end of filesystem (%llu)",
3609 le32_to_cpu(es->s_first_data_block),
3610 ext4_blocks_count(es));
e7c95593
ES		 3611 goto failed_mount;
3612 }
bd81d8ee
LV		 3613 blocks_count = (ext4_blocks_count(es) -
3614 le32_to_cpu(es->s_first_data_block) +
3615 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3616 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3617 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3618 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3619 "(block count %llu, first data block %u, "
b31e1552 3620 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3621 ext4_blocks_count(es),
3622 le32_to_cpu(es->s_first_data_block),
3623 EXT4_BLOCKS_PER_GROUP(sb));
3624 goto failed_mount;
3625 }
bd81d8ee 3626 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3627 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3628 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3629 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3630 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3631 sbi->s_group_desc = ext4_kvmalloc(db_count *
3632 sizeof(struct buffer_head *),
3633 GFP_KERNEL);
ac27a0ec 3634 if (sbi->s_group_desc == NULL) {
b31e1552 3635 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3636 ret = -ENOMEM;
ac27a0ec
DK		 3637 goto failed_mount;
3638 }
3639
9f6200bb
TT		 3640 if (ext4_proc_root)
3641 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3642
66acdcf4
TT		 3643 if (sbi->s_proc)
3644 proc_create_data("options", S_IRUGO, sbi->s_proc,
3645 &ext4_seq_options_fops, sb);
3646
705895b6 3647 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3648
3649 for (i = 0; i < db_count; i++) {
70bbb3e0 3650 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3651 sbi->s_group_desc[i] = sb_bread(sb, block);
3652 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3653 ext4_msg(sb, KERN_ERR,
3654 "can't read group descriptor %d", i);
ac27a0ec
DK		 3655 db_count = i;
3656 goto failed_mount2;
3657 }
3658 }
bfff6873 3659 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3660 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3661 goto failed_mount2;
3662 }
772cb7c8
JS		 3663 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3664 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3665 ext4_msg(sb, KERN_ERR,
3666 "unable to initialize "
3667 "flex_bg meta info!");
772cb7c8
JS		 3668 goto failed_mount2;
3669 }
3670
ac27a0ec
DK		 3671 sbi->s_gdb_count = db_count;
3672 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3673 spin_lock_init(&sbi->s_next_gen_lock);
3674
04496411
TM		 3675 init_timer(&sbi->s_err_report);
3676 sbi->s_err_report.function = print_daily_error_info;
3677 sbi->s_err_report.data = (unsigned long) sb;
3678
57042651 3679 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3680 ext4_count_free_clusters(sb));
ce7e010a
TT		 3681 if (!err) {
3682 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3683 ext4_count_free_inodes(sb));
3684 }
3685 if (!err) {
3686 err = percpu_counter_init(&sbi->s_dirs_counter,
3687 ext4_count_dirs(sb));
3688 }
3689 if (!err) {
57042651 3690 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a 3691 }
1ac6466f
TT		 3692 if (!err) {
3693 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3694 }
ce7e010a
TT		 3695 if (err) {
3696 ext4_msg(sb, KERN_ERR, "insufficient memory");
3697 goto failed_mount3;
3698 }
3699
c9de560d 3700 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3701 sbi->s_max_writeback_mb_bump = 128;
67a5da56 3702 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3703
74cd15cd
ZL		 3704 /* Register extent status tree shrinker */
3705 ext4_es_register_shrinker(sb);
3706
ac27a0ec
DK		 3707 /*
3708 * set up enough so that it can read an inode
3709 */
9ca92389
TT		 3710 if (!test_opt(sb, NOLOAD) &&
3711 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3712 sb->s_op = &ext4_sops;
3713 else
3714 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3715 sb->s_export_op = &ext4_export_ops;
3716 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3717#ifdef CONFIG_QUOTA
617ba13b
MC		 3718 sb->s_qcop = &ext4_qctl_operations;
3719 sb->dq_op = &ext4_quota_operations;
7c319d32
AK		 3720
3721 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3722 /* Use qctl operations for hidden quota files. */
3723 sb->s_qcop = &ext4_qctl_sysfile_operations;
3724 }
ac27a0ec 3725#endif
f2fa2ffc
AK		 3726 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3727
ac27a0ec 3728 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3729 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3730
3731 sb->s_root = NULL;
3732
3733 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3734 EXT4_HAS_INCOMPAT_FEATURE(sb,
3735 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3736
c5e06d10
JL		 3737 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3738 !(sb->s_flags & MS_RDONLY))
3739 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3740 goto failed_mount3;
3741
ac27a0ec
DK		 3742 /*
3743 * The first inode we look at is the journal inode. Don't try
3744 * root first: it may be modified in the journal!
3745 */
3746 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3747 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3748 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3749 goto failed_mount3;
0390131b
FM		 3750 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3751 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3752 ext4_msg(sb, KERN_ERR, "required journal recovery "
3753 "suppressed and not mounted read-only");
744692dc 3754 goto failed_mount_wq;
ac27a0ec 3755 } else {
fd8c37ec 3756 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3757 sbi->s_journal = NULL;
3758 needs_recovery = 0;
3759 goto no_journal;
ac27a0ec
DK		 3760 }
3761
f32aaf2d 3762 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3763 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3764 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3765 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3766 goto failed_mount_wq;
eb40a09c
JS		 3767 }
3768
25ed6e8a
DW		 3769 if (!set_journal_csum_feature_set(sb)) {
3770 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3771 "feature set");
3772 goto failed_mount_wq;
d4da6c9c 3773 }
818d276c 3774
ac27a0ec
DK		 3775 /* We have now updated the journal if required, so we can
3776 * validate the data journaling mode. */
3777 switch (test_opt(sb, DATA_FLAGS)) {
3778 case 0:
3779 /* No mode set, assume a default based on the journal
63f57933
AM		 3780 * capabilities: ORDERED_DATA if the journal can
3781 * cope, else JOURNAL_DATA
3782 */
dab291af
MC		 3783 if (jbd2_journal_check_available_features
3784 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3785 set_opt(sb, ORDERED_DATA);
ac27a0ec 3786 else
fd8c37ec 3787 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3788 break;
3789
617ba13b
MC		 3790 case EXT4_MOUNT_ORDERED_DATA:
3791 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3792 if (!jbd2_journal_check_available_features
3793 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3794 ext4_msg(sb, KERN_ERR, "Journal does not support "
3795 "requested data journaling mode");
744692dc 3796 goto failed_mount_wq;
ac27a0ec
DK		 3797 }
3798 default:
3799 break;
3800 }
b3881f74 3801 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3802
18aadd47
BJ		 3803 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3804
ce7e010a
TT		 3805 /*
3806 * The journal may have updated the bg summary counts, so we
3807 * need to update the global counters.
3808 */
57042651 3809 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3810 ext4_count_free_clusters(sb));
ce7e010a
TT		 3811 percpu_counter_set(&sbi->s_freeinodes_counter,
3812 ext4_count_free_inodes(sb));
3813 percpu_counter_set(&sbi->s_dirs_counter,
3814 ext4_count_dirs(sb));
57042651 3815 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3816
ce7e010a 3817no_journal:
952fc18e
TT		 3818 /*
3819 * Get the # of file system overhead blocks from the
3820 * superblock if present.
3821 */
3822 if (es->s_overhead_clusters)
3823 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3824 else {
07aa2ea1
LC		 3825 err = ext4_calculate_overhead(sb);
3826 if (err)
952fc18e
TT		 3827 goto failed_mount_wq;
3828 }
3829
fd89d5f2
TH		 3830 /*
3831 * The maximum number of concurrent works can be high and
3832 * concurrency isn't really necessary. Limit it to 1.
3833 */
3834 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3835 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3836 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3837 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
07aa2ea1 3838 ret = -ENOMEM;
4c0425ff
MC		 3839 goto failed_mount_wq;
3840 }
3841
ac27a0ec 3842 /*
dab291af 3843 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3844 * so we can safely mount the rest of the filesystem now.
3845 */
3846
1d1fe1ee
DH		 3847 root = ext4_iget(sb, EXT4_ROOT_INO);
3848 if (IS_ERR(root)) {
b31e1552 3849 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3850 ret = PTR_ERR(root);
32a9bb57 3851 root = NULL;
ac27a0ec
DK		 3852 goto failed_mount4;
3853 }
3854 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3855 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3856 iput(root);
ac27a0ec
DK		 3857 goto failed_mount4;
3858 }
48fde701 3859 sb->s_root = d_make_root(root);
1d1fe1ee 3860 if (!sb->s_root) {
b31e1552 3861 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3862 ret = -ENOMEM;
3863 goto failed_mount4;
3864 }
ac27a0ec 3865
7e84b621
ES		 3866 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3867 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3868
3869 /* determine the minimum size of new large inodes, if present */
3870 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3871 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3872 EXT4_GOOD_OLD_INODE_SIZE;
3873 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3874 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3875 if (sbi->s_want_extra_isize <
3876 le16_to_cpu(es->s_want_extra_isize))
3877 sbi->s_want_extra_isize =
3878 le16_to_cpu(es->s_want_extra_isize);
3879 if (sbi->s_want_extra_isize <
3880 le16_to_cpu(es->s_min_extra_isize))
3881 sbi->s_want_extra_isize =
3882 le16_to_cpu(es->s_min_extra_isize);
3883 }
3884 }
3885 /* Check if enough inode space is available */
3886 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3887 sbi->s_inode_size) {
3888 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3889 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3890 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3891 "available");
ef7f3835
KS		 3892 }
3893
6fd058f7
TT		 3894 err = ext4_setup_system_zone(sb);
3895 if (err) {
b31e1552 3896 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3897 "zone (%d)", err);
94bf608a 3898 goto failed_mount4a;
6fd058f7
TT		 3899 }
3900
c2774d84 3901 ext4_ext_init(sb);
9d99012f 3902 err = ext4_mb_init(sb);
c2774d84 3903 if (err) {
421f91d2 3904 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3905 err);
dcf2d804 3906 goto failed_mount5;
c2774d84
AK		 3907 }
3908
bfff6873
LC		 3909 err = ext4_register_li_request(sb, first_not_zeroed);
3910 if (err)
dcf2d804 3911 goto failed_mount6;
bfff6873 3912
3197ebdb
TT		 3913 sbi->s_kobj.kset = ext4_kset;
3914 init_completion(&sbi->s_kobj_unregister);
3915 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3916 "%s", sb->s_id);
dcf2d804
TM		 3917 if (err)
3918 goto failed_mount7;
3197ebdb 3919
617ba13b
MC		 3920 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3921 ext4_orphan_cleanup(sb, es);
3922 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3923 if (needs_recovery) {
b31e1552 3924 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3925 ext4_mark_recovery_complete(sb, es);
3926 }
3927 if (EXT4_SB(sb)->s_journal) {
3928 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3929 descr = " journalled data mode";
3930 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3931 descr = " ordered data mode";
3932 else
3933 descr = " writeback data mode";
3934 } else
3935 descr = "out journal";
3936
7c319d32
AK		 3937#ifdef CONFIG_QUOTA
3938 /* Enable quota usage during mount. */
3939 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3940 !(sb->s_flags & MS_RDONLY)) {
07aa2ea1
LC		 3941 err = ext4_enable_quotas(sb);
3942 if (err)
72ba7450 3943 goto failed_mount8;
7c319d32
AK		 3944 }
3945#endif /* CONFIG_QUOTA */
3946
79add3a3
LC		 3947 if (test_opt(sb, DISCARD)) {
3948 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3949 if (!blk_queue_discard(q))
3950 ext4_msg(sb, KERN_WARNING,
3951 "mounting with \"discard\" option, but "
3952 "the device does not support discard");
3953 }
3954
d4c402d9 3955 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3956 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3957 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3958
66e61a9e
TT		 3959 if (es->s_error_count)
3960 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3961
d4c402d9 3962 kfree(orig_data);
ac27a0ec
DK		 3963 return 0;
3964
617ba13b 3965cantfind_ext4:
ac27a0ec 3966 if (!silent)
b31e1552 3967 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3968 goto failed_mount;
3969
72ba7450
TT		 3970#ifdef CONFIG_QUOTA
3971failed_mount8:
3972 kobject_del(&sbi->s_kobj);
3973#endif
dcf2d804
TM		 3974failed_mount7:
3975 ext4_unregister_li_request(sb);
3976failed_mount6:
dcf2d804 3977 ext4_mb_release(sb);
94bf608a
AV		 3978failed_mount5:
3979 ext4_ext_release(sb);
dcf2d804 3980 ext4_release_system_zone(sb);
94bf608a
AV		 3981failed_mount4a:
3982 dput(sb->s_root);
32a9bb57 3983 sb->s_root = NULL;
94bf608a 3984failed_mount4:
b31e1552 3985 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3986 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3987failed_mount_wq:
0390131b
FM		 3988 if (sbi->s_journal) {
3989 jbd2_journal_destroy(sbi->s_journal);
3990 sbi->s_journal = NULL;
3991 }
ac27a0ec 3992failed_mount3:
04496411 3993 del_timer(&sbi->s_err_report);
9933fc0a
TT		 3994 if (sbi->s_flex_groups)
3995 ext4_kvfree(sbi->s_flex_groups);
57042651 3996 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 3997 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3998 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 3999 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 4000 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
c5e06d10
JL		 4001 if (sbi->s_mmp_tsk)
4002 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4003failed_mount2:
4004 for (i = 0; i < db_count; i++)
4005 brelse(sbi->s_group_desc[i]);
f18a5f21 4006 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4007failed_mount:
0441984a
DW		 4008 if (sbi->s_chksum_driver)
4009 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4010 if (sbi->s_proc) {
66acdcf4 4011 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4012 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4013 }
ac27a0ec
DK		 4014#ifdef CONFIG_QUOTA
4015 for (i = 0; i < MAXQUOTAS; i++)
4016 kfree(sbi->s_qf_names[i]);
4017#endif
617ba13b 4018 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4019 brelse(bh);
4020out_fail:
4021 sb->s_fs_info = NULL;
f6830165 4022 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4023 kfree(sbi);
dcc7dae3 4024out_free_orig:
d4c402d9 4025 kfree(orig_data);
07aa2ea1 4026 return err ? err : ret;
ac27a0ec
DK		 4027}
4028
4029/*
4030 * Setup any per-fs journal parameters now. We'll do this both on
4031 * initial mount, once the journal has been initialised but before we've
4032 * done any recovery; and again on any subsequent remount.
4033 */
617ba13b 4034static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4035{
617ba13b 4036 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4037
30773840
TT		 4038 journal->j_commit_interval = sbi->s_commit_interval;
4039 journal->j_min_batch_time = sbi->s_min_batch_time;
4040 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4041
a931da6a 4042 write_lock(&journal->j_state_lock);
ac27a0ec 4043 if (test_opt(sb, BARRIER))
dab291af 4044 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4045 else
dab291af 4046 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4047 if (test_opt(sb, DATA_ERR_ABORT))
4048 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4049 else
4050 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4051 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4052}
4053
617ba13b 4054static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4055 unsigned int journal_inum)
4056{
4057 struct inode *journal_inode;
4058 journal_t *journal;
4059
0390131b
FM		 4060 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4061
ac27a0ec
DK		 4062 /* First, test for the existence of a valid inode on disk. Bad
4063 * things happen if we iget() an unused inode, as the subsequent
4064 * iput() will try to delete it. */
4065
1d1fe1ee
DH		 4066 journal_inode = ext4_iget(sb, journal_inum);
4067 if (IS_ERR(journal_inode)) {
b31e1552 4068 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4069 return NULL;
4070 }
4071 if (!journal_inode->i_nlink) {
4072 make_bad_inode(journal_inode);
4073 iput(journal_inode);
b31e1552 4074 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4075 return NULL;
4076 }
4077
e5f8eab8 4078 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4079 journal_inode, journal_inode->i_size);
1d1fe1ee 4080 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4081 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4082 iput(journal_inode);
4083 return NULL;
4084 }
4085
dab291af 4086 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4087 if (!journal) {
b31e1552 4088 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4089 iput(journal_inode);
4090 return NULL;
4091 }
4092 journal->j_private = sb;
617ba13b 4093 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4094 return journal;
4095}
4096
617ba13b 4097static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4098 dev_t j_dev)
4099{
2b2d6d01 4100 struct buffer_head *bh;
ac27a0ec 4101 journal_t *journal;
617ba13b
MC		 4102 ext4_fsblk_t start;
4103 ext4_fsblk_t len;
ac27a0ec 4104 int hblock, blocksize;
617ba13b 4105 ext4_fsblk_t sb_block;
ac27a0ec 4106 unsigned long offset;
2b2d6d01 4107 struct ext4_super_block *es;
ac27a0ec
DK		 4108 struct block_device *bdev;
4109
0390131b
FM		 4110 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4111
b31e1552 4112 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4113 if (bdev == NULL)
4114 return NULL;
4115
ac27a0ec 4116 blocksize = sb->s_blocksize;
e1defc4f 4117 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4118 if (blocksize < hblock) {
b31e1552
ES		 4119 ext4_msg(sb, KERN_ERR,
4120 "blocksize too small for journal device");
ac27a0ec
DK		 4121 goto out_bdev;
4122 }
4123
617ba13b
MC		 4124 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4125 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4126 set_blocksize(bdev, blocksize);
4127 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4128 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4129 "external journal");
ac27a0ec
DK		 4130 goto out_bdev;
4131 }
4132
2716b802 4133 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4134 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4135 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4136 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4137 ext4_msg(sb, KERN_ERR, "external journal has "
4138 "bad superblock");
ac27a0ec
DK		 4139 brelse(bh);
4140 goto out_bdev;
4141 }
4142
617ba13b 4143 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4144 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4145 brelse(bh);
4146 goto out_bdev;
4147 }
4148
bd81d8ee 4149 len = ext4_blocks_count(es);
ac27a0ec
DK		 4150 start = sb_block + 1;
4151 brelse(bh); /* we're done with the superblock */
4152
dab291af 4153 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4154 start, len, blocksize);
4155 if (!journal) {
b31e1552 4156 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4157 goto out_bdev;
4158 }
4159 journal->j_private = sb;
4160 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4161 wait_on_buffer(journal->j_sb_buffer);
4162 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4163 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4164 goto out_journal;
4165 }
4166 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4167 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4168 "user (unsupported) - %d",
ac27a0ec
DK		 4169 be32_to_cpu(journal->j_superblock->s_nr_users));
4170 goto out_journal;
4171 }
617ba13b
MC		 4172 EXT4_SB(sb)->journal_bdev = bdev;
4173 ext4_init_journal_params(sb, journal);
ac27a0ec 4174 return journal;
0b8e58a1 4175
ac27a0ec 4176out_journal:
dab291af 4177 jbd2_journal_destroy(journal);
ac27a0ec 4178out_bdev:
617ba13b 4179 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4180 return NULL;
4181}
4182
617ba13b
MC		 4183static int ext4_load_journal(struct super_block *sb,
4184 struct ext4_super_block *es,
ac27a0ec
DK		 4185 unsigned long journal_devnum)
4186{
4187 journal_t *journal;
4188 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4189 dev_t journal_dev;
4190 int err = 0;
4191 int really_read_only;
4192
0390131b
FM		 4193 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4194
ac27a0ec
DK		 4195 if (journal_devnum &&
4196 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4197 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4198 "numbers have changed");
ac27a0ec
DK		 4199 journal_dev = new_decode_dev(journal_devnum);
4200 } else
4201 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4202
4203 really_read_only = bdev_read_only(sb->s_bdev);
4204
4205 /*
4206 * Are we loading a blank journal or performing recovery after a
4207 * crash? For recovery, we need to check in advance whether we
4208 * can get read-write access to the device.
4209 */
617ba13b 4210 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4211 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4212 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4213 "required on readonly filesystem");
ac27a0ec 4214 if (really_read_only) {
b31e1552
ES		 4215 ext4_msg(sb, KERN_ERR, "write access "
4216 "unavailable, cannot proceed");
ac27a0ec
DK		 4217 return -EROFS;
4218 }
b31e1552
ES		 4219 ext4_msg(sb, KERN_INFO, "write access will "
4220 "be enabled during recovery");
ac27a0ec
DK		 4221 }
4222 }
4223
4224 if (journal_inum && journal_dev) {
b31e1552
ES		 4225 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4226 "and inode journals!");
ac27a0ec
DK		 4227 return -EINVAL;
4228 }
4229
4230 if (journal_inum) {
617ba13b 4231 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4232 return -EINVAL;
4233 } else {
617ba13b 4234 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4235 return -EINVAL;
4236 }
4237
90576c0b 4238 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4239 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4240
617ba13b 4241 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4242 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4243 if (!err) {
4244 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4245 if (save)
4246 memcpy(save, ((char *) es) +
4247 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4248 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4249 if (save)
4250 memcpy(((char *) es) + EXT4_S_ERR_START,
4251 save, EXT4_S_ERR_LEN);
4252 kfree(save);
4253 }
ac27a0ec
DK		 4254
4255 if (err) {
b31e1552 4256 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4257 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4258 return err;
4259 }
4260
617ba13b
MC		 4261 EXT4_SB(sb)->s_journal = journal;
4262 ext4_clear_journal_err(sb, es);
ac27a0ec 4263
c41303ce 4264 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4265 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4266 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4267
4268 /* Make sure we flush the recovery flag to disk. */
e2d67052 4269 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4270 }
4271
4272 return 0;
4273}
4274
e2d67052 4275static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4276{
e2d67052 4277 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4278 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4279 int error = 0;
ac27a0ec 4280
7c2e7087 4281 if (!sbh || block_device_ejected(sb))
c4be0c1d 4282 return error;
914258bf
TT		 4283 if (buffer_write_io_error(sbh)) {
4284 /*
4285 * Oh, dear. A previous attempt to write the
4286 * superblock failed. This could happen because the
4287 * USB device was yanked out. Or it could happen to
4288 * be a transient write error and maybe the block will
4289 * be remapped. Nothing we can do but to retry the
4290 * write and hope for the best.
4291 */
b31e1552
ES		 4292 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4293 "superblock detected");
914258bf
TT		 4294 clear_buffer_write_io_error(sbh);
4295 set_buffer_uptodate(sbh);
4296 }
71290b36
TT		 4297 /*
4298 * If the file system is mounted read-only, don't update the
4299 * superblock write time. This avoids updating the superblock
4300 * write time when we are mounting the root file system
4301 * read/only but we need to replay the journal; at that point,
4302 * for people who are east of GMT and who make their clock
4303 * tick in localtime for Windows bug-for-bug compatibility,
4304 * the clock is set in the future, and this will cause e2fsck
4305 * to complain and force a full file system check.
4306 */
4307 if (!(sb->s_flags & MS_RDONLY))
4308 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4309 if (sb->s_bdev->bd_part)
4310 es->s_kbytes_written =
4311 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4312 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4313 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4314 else
4315 es->s_kbytes_written =
4316 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4317 ext4_free_blocks_count_set(es,
4318 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4319 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4320 es->s_free_inodes_count =
4321 cpu_to_le32(percpu_counter_sum_positive(
4322 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4323 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4324 ext4_superblock_csum_set(sb);
ac27a0ec 4325 mark_buffer_dirty(sbh);
914258bf 4326 if (sync) {
c4be0c1d
TS		 4327 error = sync_dirty_buffer(sbh);
4328 if (error)
4329 return error;
4330
4331 error = buffer_write_io_error(sbh);
4332 if (error) {
b31e1552
ES		 4333 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4334 "superblock");
914258bf
TT		 4335 clear_buffer_write_io_error(sbh);
4336 set_buffer_uptodate(sbh);
4337 }
4338 }
c4be0c1d 4339 return error;
ac27a0ec
DK		 4340}
4341
ac27a0ec
DK		 4342/*
4343 * Have we just finished recovery? If so, and if we are mounting (or
4344 * remounting) the filesystem readonly, then we will end up with a
4345 * consistent fs on disk. Record that fact.
4346 */
2b2d6d01
TT		 4347static void ext4_mark_recovery_complete(struct super_block *sb,
4348 struct ext4_super_block *es)
ac27a0ec 4349{
617ba13b 4350 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4351
0390131b
FM		 4352 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4353 BUG_ON(journal != NULL);
4354 return;
4355 }
dab291af 4356 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4357 if (jbd2_journal_flush(journal) < 0)
4358 goto out;
4359
617ba13b 4360 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4361 sb->s_flags & MS_RDONLY) {
617ba13b 4362 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4363 ext4_commit_super(sb, 1);
ac27a0ec 4364 }
7ffe1ea8
HK		 4365
4366out:
dab291af 4367 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4368}
4369
4370/*
4371 * If we are mounting (or read-write remounting) a filesystem whose journal
4372 * has recorded an error from a previous lifetime, move that error to the
4373 * main filesystem now.
4374 */
2b2d6d01
TT		 4375static void ext4_clear_journal_err(struct super_block *sb,
4376 struct ext4_super_block *es)
ac27a0ec
DK		 4377{
4378 journal_t *journal;
4379 int j_errno;
4380 const char *errstr;
4381
0390131b
FM		 4382 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4383
617ba13b 4384 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4385
4386 /*
4387 * Now check for any error status which may have been recorded in the
617ba13b 4388 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4389 */
4390
dab291af 4391 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4392 if (j_errno) {
4393 char nbuf[16];
4394
617ba13b 4395 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4396 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4397 "from previous mount: %s", errstr);
12062ddd 4398 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4399
617ba13b
MC		 4400 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4401 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4402 ext4_commit_super(sb, 1);
ac27a0ec 4403
dab291af 4404 jbd2_journal_clear_err(journal);
d796c52e 4405 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4406 }
4407}
4408
4409/*
4410 * Force the running and committing transactions to commit,
4411 * and wait on the commit.
4412 */
617ba13b 4413int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4414{
4415 journal_t *journal;
ac27a0ec
DK		 4416
4417 if (sb->s_flags & MS_RDONLY)
4418 return 0;
4419
617ba13b 4420 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4421 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4422}
4423
617ba13b 4424static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4425{
14ce0cb4 4426 int ret = 0;
9eddacf9 4427 tid_t target;
8d5d02e6 4428 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4429
9bffad1e 4430 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4431 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4432 /*
4433 * Writeback quota in non-journalled quota case - journalled quota has
4434 * no dirty dquots
4435 */
4436 dquot_writeback_dquots(sb, -1);
8d5d02e6 4437 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4438 if (wait)
8d5d02e6 4439 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4440 }
14ce0cb4 4441 return ret;
ac27a0ec
DK		 4442}
4443
4444/*
4445 * LVM calls this function before a (read-only) snapshot is created. This
4446 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4447 *
4448 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4449 * state independently. It relies on upper layer to stop all data & metadata
4450 * modifications.
ac27a0ec 4451 */
c4be0c1d 4452static int ext4_freeze(struct super_block *sb)
ac27a0ec 4453{
c4be0c1d
TS		 4454 int error = 0;
4455 journal_t *journal;
ac27a0ec 4456
9ca92389
TT		 4457 if (sb->s_flags & MS_RDONLY)
4458 return 0;
ac27a0ec 4459
9ca92389 4460 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4461
9ca92389
TT		 4462 /* Now we set up the journal barrier. */
4463 jbd2_journal_lock_updates(journal);
ac27a0ec 4464
9ca92389
TT		 4465 /*
4466 * Don't clear the needs_recovery flag if we failed to flush
4467 * the journal.
4468 */
4469 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4470 if (error < 0)
4471 goto out;
9ca92389
TT		 4472
4473 /* Journal blocked and flushed, clear needs_recovery flag. */
4474 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4475 error = ext4_commit_super(sb, 1);
6b0310fb 4476out:
8e8ad8a5 4477 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4478 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4479 return error;
ac27a0ec
DK		 4480}
4481
4482/*
4483 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4484 * flag here, even though the filesystem is not technically dirty yet.
4485 */
c4be0c1d 4486static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4487{
9ca92389
TT		 4488 if (sb->s_flags & MS_RDONLY)
4489 return 0;
4490
9ca92389
TT		 4491 /* Reset the needs_recovery flag before the fs is unlocked. */
4492 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4493 ext4_commit_super(sb, 1);
c4be0c1d 4494 return 0;
ac27a0ec
DK		 4495}
4496
673c6100
TT		 4497/*
4498 * Structure to save mount options for ext4_remount's benefit
4499 */
4500struct ext4_mount_options {
4501 unsigned long s_mount_opt;
a2595b8a 4502 unsigned long s_mount_opt2;
08cefc7a
EB		 4503 kuid_t s_resuid;
4504 kgid_t s_resgid;
673c6100
TT		 4505 unsigned long s_commit_interval;
4506 u32 s_min_batch_time, s_max_batch_time;
4507#ifdef CONFIG_QUOTA
4508 int s_jquota_fmt;
4509 char *s_qf_names[MAXQUOTAS];
4510#endif
4511};
4512
2b2d6d01 4513static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4514{
2b2d6d01 4515 struct ext4_super_block *es;
617ba13b 4516 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4517 unsigned long old_sb_flags;
617ba13b 4518 struct ext4_mount_options old_opts;
c79d967d 4519 int enable_quota = 0;
8a266467 4520 ext4_group_t g;
b3881f74 4521 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4522 int err = 0;
ac27a0ec 4523#ifdef CONFIG_QUOTA
03dafb5f 4524 int i, j;
ac27a0ec 4525#endif
d4c402d9 4526 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4527
4528 /* Store the original options */
4529 old_sb_flags = sb->s_flags;
4530 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4531 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4532 old_opts.s_resuid = sbi->s_resuid;
4533 old_opts.s_resgid = sbi->s_resgid;
4534 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4535 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4536 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4537#ifdef CONFIG_QUOTA
4538 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4539 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f
CG		 4540 if (sbi->s_qf_names[i]) {
4541 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4542 GFP_KERNEL);
4543 if (!old_opts.s_qf_names[i]) {
4544 for (j = 0; j < i; j++)
4545 kfree(old_opts.s_qf_names[j]);
3e36a163 4546 kfree(orig_data);
03dafb5f
CG		 4547 return -ENOMEM;
4548 }
4549 } else
4550 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4551#endif
b3881f74
TT		 4552 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4553 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4554
4555 /*
4556 * Allow the "check" option to be passed as a remount option.
4557 */
661aa520 4558 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4559 err = -EINVAL;
4560 goto restore_opts;
4561 }
4562
4ab2f15b 4563 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4564 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4565
4566 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4567 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4568
4569 es = sbi->s_es;
4570
b3881f74 4571 if (sbi->s_journal) {
0390131b 4572 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4573 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4574 }
ac27a0ec 4575
661aa520 4576 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4577 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4578 err = -EROFS;
4579 goto restore_opts;
4580 }
4581
4582 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4583 err = dquot_suspend(sb, -1);
4584 if (err < 0)
c79d967d 4585 goto restore_opts;
c79d967d 4586
ac27a0ec
DK		 4587 /*
4588 * First of all, the unconditional stuff we have to do
4589 * to disable replay of the journal when we next remount
4590 */
4591 sb->s_flags |= MS_RDONLY;
4592
4593 /*
4594 * OK, test if we are remounting a valid rw partition
4595 * readonly, and if so set the rdonly flag and then
4596 * mark the partition as valid again.
4597 */
617ba13b
MC		 4598 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4599 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4600 es->s_state = cpu_to_le16(sbi->s_mount_state);
4601
a63c9eb2 4602 if (sbi->s_journal)
0390131b 4603 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4604 } else {
a13fb1a4
ES		 4605 /* Make sure we can mount this feature set readwrite */
4606 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4607 err = -EROFS;
4608 goto restore_opts;
4609 }
8a266467
TT		 4610 /*
4611 * Make sure the group descriptor checksums
0b8e58a1 4612 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4613 */
4614 for (g = 0; g < sbi->s_groups_count; g++) {
4615 struct ext4_group_desc *gdp =
4616 ext4_get_group_desc(sb, g, NULL);
4617
feb0ab32 4618 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4619 ext4_msg(sb, KERN_ERR,
4620 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4621 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4622 le16_to_cpu(gdp->bg_checksum));
4623 err = -EINVAL;
4624 goto restore_opts;
4625 }
4626 }
4627
ead6596b
ES		 4628 /*
4629 * If we have an unprocessed orphan list hanging
4630 * around from a previously readonly bdev mount,
4631 * require a full umount/remount for now.
4632 */
4633 if (es->s_last_orphan) {
b31e1552 4634 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4635 "remount RDWR because of unprocessed "
4636 "orphan inode list. Please "
b31e1552 4637 "umount/remount instead");
ead6596b
ES		 4638 err = -EINVAL;
4639 goto restore_opts;
4640 }
4641
ac27a0ec
DK		 4642 /*
4643 * Mounting a RDONLY partition read-write, so reread
4644 * and store the current valid flag. (It may have
4645 * been changed by e2fsck since we originally mounted
4646 * the partition.)
4647 */
0390131b
FM		 4648 if (sbi->s_journal)
4649 ext4_clear_journal_err(sb, es);
ac27a0ec 4650 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4651 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4652 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4653 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4654 EXT4_FEATURE_INCOMPAT_MMP))
4655 if (ext4_multi_mount_protect(sb,
4656 le64_to_cpu(es->s_mmp_block))) {
4657 err = -EROFS;
4658 goto restore_opts;
4659 }
c79d967d 4660 enable_quota = 1;
ac27a0ec
DK		 4661 }
4662 }
bfff6873
LC		 4663
4664 /*
4665 * Reinitialize lazy itable initialization thread based on
4666 * current settings
4667 */
4668 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4669 ext4_unregister_li_request(sb);
4670 else {
4671 ext4_group_t first_not_zeroed;
4672 first_not_zeroed = ext4_has_uninit_itable(sb);
4673 ext4_register_li_request(sb, first_not_zeroed);
4674 }
4675
6fd058f7 4676 ext4_setup_system_zone(sb);
d096ad0f 4677 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4678 ext4_commit_super(sb, 1);
0390131b 4679
ac27a0ec
DK		 4680#ifdef CONFIG_QUOTA
4681 /* Release old quota file names */
4682 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f 4683 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4684 if (enable_quota) {
4685 if (sb_any_quota_suspended(sb))
4686 dquot_resume(sb, -1);
4687 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4688 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4689 err = ext4_enable_quotas(sb);
07724f98 4690 if (err)
7c319d32 4691 goto restore_opts;
7c319d32
AK		 4692 }
4693 }
ac27a0ec 4694#endif
d4c402d9
CW		 4695
4696 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4697 kfree(orig_data);
ac27a0ec 4698 return 0;
0b8e58a1 4699
ac27a0ec
DK		 4700restore_opts:
4701 sb->s_flags = old_sb_flags;
4702 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4703 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4704 sbi->s_resuid = old_opts.s_resuid;
4705 sbi->s_resgid = old_opts.s_resgid;
4706 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4707 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4708 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4709#ifdef CONFIG_QUOTA
4710 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4711 for (i = 0; i < MAXQUOTAS; i++) {
03dafb5f 4712 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4713 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4714 }
4715#endif
d4c402d9 4716 kfree(orig_data);
ac27a0ec
DK		 4717 return err;
4718}
4719
2b2d6d01 4720static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4721{
4722 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4723 struct ext4_sb_info *sbi = EXT4_SB(sb);
4724 struct ext4_super_block *es = sbi->s_es;
952fc18e 4725 ext4_fsblk_t overhead = 0;
960cc398 4726 u64 fsid;
d02a9391 4727 s64 bfree;
ac27a0ec 4728
952fc18e
TT		 4729 if (!test_opt(sb, MINIX_DF))
4730 overhead = sbi->s_overhead;
ac27a0ec 4731
617ba13b 4732 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4733 buf->f_bsize = sb->s_blocksize;
b72f78cb 4734 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4735 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4736 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4737 /* prevent underflow in case that few free space is available */
57042651 4738 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4739 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4740 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4741 buf->f_bavail = 0;
4742 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4743 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4744 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4745 fsid = le64_to_cpup((void *)es->s_uuid) ^
4746 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4747 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4748 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4749
ac27a0ec
DK		 4750 return 0;
4751}
4752
0b8e58a1
AD		 4753/* Helper function for writing quotas on sync - we need to start transaction
4754 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4755 * Process 1 Process 2
617ba13b 4756 * ext4_create() quota_sync()
a269eb18 4757 * jbd2_journal_start() write_dquot()
871a2931 4758 * dquot_initialize() down(dqio_mutex)
dab291af 4759 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4760 *
4761 */
4762
4763#ifdef CONFIG_QUOTA
4764
4765static inline struct inode *dquot_to_inode(struct dquot *dquot)
4766{
4c376dca 4767 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4768}
4769
617ba13b 4770static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4771{
4772 int ret, err;
4773 handle_t *handle;
4774 struct inode *inode;
4775
4776 inode = dquot_to_inode(dquot);
9924a92a 4777 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 4778 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4779 if (IS_ERR(handle))
4780 return PTR_ERR(handle);
4781 ret = dquot_commit(dquot);
617ba13b 4782 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4783 if (!ret)
4784 ret = err;
4785 return ret;
4786}
4787
617ba13b 4788static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4789{
4790 int ret, err;
4791 handle_t *handle;
4792
9924a92a 4793 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4794 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4795 if (IS_ERR(handle))
4796 return PTR_ERR(handle);
4797 ret = dquot_acquire(dquot);
617ba13b 4798 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4799 if (!ret)
4800 ret = err;
4801 return ret;
4802}
4803
617ba13b 4804static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4805{
4806 int ret, err;
4807 handle_t *handle;
4808
9924a92a 4809 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 4810 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4811 if (IS_ERR(handle)) {
4812 /* Release dquot anyway to avoid endless cycle in dqput() */
4813 dquot_release(dquot);
ac27a0ec 4814 return PTR_ERR(handle);
9c3013e9 4815 }
ac27a0ec 4816 ret = dquot_release(dquot);
617ba13b 4817 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4818 if (!ret)
4819 ret = err;
4820 return ret;
4821}
4822
617ba13b 4823static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4824{
2c8be6b2 4825 /* Are we journaling quotas? */
617ba13b
MC		 4826 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4827 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4828 dquot_mark_dquot_dirty(dquot);
617ba13b 4829 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4830 } else {
4831 return dquot_mark_dquot_dirty(dquot);
4832 }
4833}
4834
617ba13b 4835static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4836{
4837 int ret, err;
4838 handle_t *handle;
4839
4840 /* Data block + inode block */
9924a92a 4841 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 4842 if (IS_ERR(handle))
4843 return PTR_ERR(handle);
4844 ret = dquot_commit_info(sb, type);
617ba13b 4845 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4846 if (!ret)
4847 ret = err;
4848 return ret;
4849}
4850
4851/*
4852 * Turn on quotas during mount time - we need to find
4853 * the quota file and such...
4854 */
617ba13b 4855static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4856{
287a8095
CH		 4857 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4858 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4859}
4860
4861/*
4862 * Standard function to be called on quota_on
4863 */
617ba13b 4864static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4865 struct path *path)
ac27a0ec
DK		 4866{
4867 int err;
ac27a0ec
DK		 4868
4869 if (!test_opt(sb, QUOTA))
4870 return -EINVAL;
0623543b 4871
ac27a0ec 4872 /* Quotafile not on the same filesystem? */
d8c9584e 4873 if (path->dentry->d_sb != sb)
ac27a0ec 4874 return -EXDEV;
0623543b
JK		 4875 /* Journaling quota? */
4876 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4877 /* Quotafile not in fs root? */
f00c9e44 4878 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4879 ext4_msg(sb, KERN_WARNING,
4880 "Quota file not on filesystem root. "
4881 "Journaled quota will not work");
2b2d6d01 4882 }
0623543b
JK		 4883
4884 /*
4885 * When we journal data on quota file, we have to flush journal to see
4886 * all updates to the file when we bypass pagecache...
4887 */
0390131b 4888 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4889 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4890 /*
4891 * We don't need to lock updates but journal_flush() could
4892 * otherwise be livelocked...
4893 */
4894 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4895 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4896 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4897 if (err)
7ffe1ea8 4898 return err;
0623543b
JK		 4899 }
4900
f00c9e44 4901 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4902}
4903
7c319d32
AK		 4904static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4905 unsigned int flags)
4906{
4907 int err;
4908 struct inode *qf_inode;
4909 unsigned long qf_inums[MAXQUOTAS] = {
4910 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4911 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4912 };
4913
4914 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4915
4916 if (!qf_inums[type])
4917 return -EPERM;
4918
4919 qf_inode = ext4_iget(sb, qf_inums[type]);
4920 if (IS_ERR(qf_inode)) {
4921 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4922 return PTR_ERR(qf_inode);
4923 }
4924
4925 err = dquot_enable(qf_inode, type, format_id, flags);
4926 iput(qf_inode);
4927
4928 return err;
4929}
4930
4931/* Enable usage tracking for all quota types. */
4932static int ext4_enable_quotas(struct super_block *sb)
4933{
4934 int type, err = 0;
4935 unsigned long qf_inums[MAXQUOTAS] = {
4936 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4937 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4938 };
4939
4940 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4941 for (type = 0; type < MAXQUOTAS; type++) {
4942 if (qf_inums[type]) {
4943 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4944 DQUOT_USAGE_ENABLED);
4945 if (err) {
4946 ext4_warning(sb,
72ba7450
TT		 4947 "Failed to enable quota tracking "
4948 "(type=%d, err=%d). Please run "
4949 "e2fsck to fix.", type, err);
7c319d32
AK		 4950 return err;
4951 }
4952 }
4953 }
4954 return 0;
4955}
4956
4957/*
4958 * quota_on function that is used when QUOTA feature is set.
4959 */
4960static int ext4_quota_on_sysfile(struct super_block *sb, int type,
4961 int format_id)
4962{
4963 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4964 return -EINVAL;
4965
4966 /*
4967 * USAGE was enabled at mount time. Only need to enable LIMITS now.
4968 */
4969 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
4970}
4971
ca0e05e4
DM		 4972static int ext4_quota_off(struct super_block *sb, int type)
4973{
21f97697
JK		 4974 struct inode *inode = sb_dqopt(sb)->files[type];
4975 handle_t *handle;
4976
87009d86
DM		 4977 /* Force all delayed allocation blocks to be allocated.
4978 * Caller already holds s_umount sem */
4979 if (test_opt(sb, DELALLOC))
ca0e05e4 4980 sync_filesystem(sb);
ca0e05e4 4981
0b268590
AG		 4982 if (!inode)
4983 goto out;
4984
21f97697
JK		 4985 /* Update modification times of quota files when userspace can
4986 * start looking at them */
9924a92a 4987 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 4988 if (IS_ERR(handle))
4989 goto out;
4990 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4991 ext4_mark_inode_dirty(handle, inode);
4992 ext4_journal_stop(handle);
4993
4994out:
ca0e05e4
DM		 4995 return dquot_quota_off(sb, type);
4996}
4997
7c319d32
AK		 4998/*
4999 * quota_off function that is used when QUOTA feature is set.
5000 */
5001static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5002{
5003 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5004 return -EINVAL;
5005
5006 /* Disable only the limits. */
5007 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5008}
5009
ac27a0ec
DK		 5010/* Read data from quotafile - avoid pagecache and such because we cannot afford
5011 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5012 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5013 * we don't have to be afraid of races */
617ba13b 5014static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5015 size_t len, loff_t off)
5016{
5017 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5018 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5019 int err = 0;
5020 int offset = off & (sb->s_blocksize - 1);
5021 int tocopy;
5022 size_t toread;
5023 struct buffer_head *bh;
5024 loff_t i_size = i_size_read(inode);
5025
5026 if (off > i_size)
5027 return 0;
5028 if (off+len > i_size)
5029 len = i_size-off;
5030 toread = len;
5031 while (toread > 0) {
5032 tocopy = sb->s_blocksize - offset < toread ?
5033 sb->s_blocksize - offset : toread;
617ba13b 5034 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5035 if (err)
5036 return err;
5037 if (!bh) /* A hole? */
5038 memset(data, 0, tocopy);
5039 else
5040 memcpy(data, bh->b_data+offset, tocopy);
5041 brelse(bh);
5042 offset = 0;
5043 toread -= tocopy;
5044 data += tocopy;
5045 blk++;
5046 }
5047 return len;
5048}
5049
5050/* Write to quotafile (we know the transaction is already started and has
5051 * enough credits) */
617ba13b 5052static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5053 const char *data, size_t len, loff_t off)
5054{
5055 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5056 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5057 int err = 0;
5058 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5059 struct buffer_head *bh;
5060 handle_t *handle = journal_current_handle();
5061
0390131b 5062 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5063 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5064 " cancelled because transaction is not started",
9c3013e9
JK		 5065 (unsigned long long)off, (unsigned long long)len);
5066 return -EIO;
5067 }
67eeb568
DM		 5068 /*
5069 * Since we account only one data block in transaction credits,
5070 * then it is impossible to cross a block boundary.
5071 */
5072 if (sb->s_blocksize - offset < len) {
5073 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5074 " cancelled because not block aligned",
5075 (unsigned long long)off, (unsigned long long)len);
5076 return -EIO;
5077 }
5078
67eeb568
DM		 5079 bh = ext4_bread(handle, inode, blk, 1, &err);
5080 if (!bh)
5081 goto out;
62d2b5f2
JK		 5082 err = ext4_journal_get_write_access(handle, bh);
5083 if (err) {
5084 brelse(bh);
5085 goto out;
ac27a0ec 5086 }
67eeb568
DM		 5087 lock_buffer(bh);
5088 memcpy(bh->b_data+offset, data, len);
5089 flush_dcache_page(bh->b_page);
5090 unlock_buffer(bh);
62d2b5f2 5091 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5092 brelse(bh);
ac27a0ec 5093out:
0b7f7cef 5094 if (err)
ac27a0ec 5095 return err;
67eeb568
DM		 5096 if (inode->i_size < off + len) {
5097 i_size_write(inode, off + len);
617ba13b 5098 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5099 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5100 }
67eeb568 5101 return len;
ac27a0ec
DK		 5102}
5103
5104#endif
5105
152a0836
AV		 5106static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5107 const char *dev_name, void *data)
ac27a0ec 5108{
152a0836 5109 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5110}
5111
37f328eb 5112#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5113static inline void register_as_ext2(void)
5114{
5115 int err = register_filesystem(&ext2_fs_type);
5116 if (err)
5117 printk(KERN_WARNING
5118 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5119}
5120
5121static inline void unregister_as_ext2(void)
5122{
5123 unregister_filesystem(&ext2_fs_type);
5124}
2035e776
TT		 5125
5126static inline int ext2_feature_set_ok(struct super_block *sb)
5127{
5128 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5129 return 0;
5130 if (sb->s_flags & MS_RDONLY)
5131 return 1;
5132 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5133 return 0;
5134 return 1;
5135}
51b7e3c9 5136MODULE_ALIAS("ext2");
24b58424
TT		 5137#else
5138static inline void register_as_ext2(void) { }
5139static inline void unregister_as_ext2(void) { }
2035e776 5140static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5141#endif
5142
37f328eb 5143#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5144static inline void register_as_ext3(void)
5145{
5146 int err = register_filesystem(&ext3_fs_type);
5147 if (err)
5148 printk(KERN_WARNING
5149 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5150}
5151
5152static inline void unregister_as_ext3(void)
5153{
5154 unregister_filesystem(&ext3_fs_type);
5155}
2035e776
TT		 5156
5157static inline int ext3_feature_set_ok(struct super_block *sb)
5158{
5159 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5160 return 0;
5161 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5162 return 0;
5163 if (sb->s_flags & MS_RDONLY)
5164 return 1;
5165 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5166 return 0;
5167 return 1;
5168}
51b7e3c9 5169MODULE_ALIAS("ext3");
24b58424
TT		 5170#else
5171static inline void register_as_ext3(void) { }
5172static inline void unregister_as_ext3(void) { }
2035e776 5173static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5174#endif
5175
03010a33
TT		 5176static struct file_system_type ext4_fs_type = {
5177 .owner = THIS_MODULE,
5178 .name = "ext4",
152a0836 5179 .mount = ext4_mount,
03010a33
TT		 5180 .kill_sb = kill_block_super,
5181 .fs_flags = FS_REQUIRES_DEV,
5182};
5183
8f021222 5184static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5185{
5186 struct ext4_features *ef;
5187 int ret = -ENOMEM;
5188
5189 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5190 if (!ef)
5191 goto out;
5192
5193 ef->f_kobj.kset = ext4_kset;
5194 init_completion(&ef->f_kobj_unregister);
5195 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5196 "features");
5197 if (ret) {
5198 kfree(ef);
5199 goto out;
5200 }
5201
5202 ext4_feat = ef;
5203 ret = 0;
5204out:
5205 return ret;
5206}
5207
8f021222
LC		 5208static void ext4_exit_feat_adverts(void)
5209{
5210 kobject_put(&ext4_feat->f_kobj);
5211 wait_for_completion(&ext4_feat->f_kobj_unregister);
5212 kfree(ext4_feat);
5213}
5214
e9e3bcec
ES		 5215/* Shared across all ext4 file systems */
5216wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5217struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5218
5dabfc78 5219static int __init ext4_init_fs(void)
ac27a0ec 5220{
e9e3bcec 5221 int i, err;
c9de560d 5222
07c0c5d8
AV		 5223 ext4_li_info = NULL;
5224 mutex_init(&ext4_li_mtx);
5225
9a4c8019 5226 /* Build-time check for flags consistency */
12e9b892 5227 ext4_check_flag_values();
e9e3bcec
ES		 5228
5229 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5230 mutex_init(&ext4__aio_mutex[i]);
5231 init_waitqueue_head(&ext4__ioend_wq[i]);
5232 }
5233
51865fda 5234 err = ext4_init_es();
6fd058f7
TT		 5235 if (err)
5236 return err;
51865fda
ZL		 5237
5238 err = ext4_init_pageio();
5239 if (err)
5240 goto out7;
5241
5dabfc78 5242 err = ext4_init_system_zone();
bd2d0210 5243 if (err)
d44651d0 5244 goto out6;
3197ebdb 5245 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5246 if (!ext4_kset) {
5247 err = -ENOMEM;
dd68314c 5248 goto out5;
0e376b1e 5249 }
d44651d0 5250 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5251
5252 err = ext4_init_feat_adverts();
dd68314c
TT		 5253 if (err)
5254 goto out4;
857ac889 5255
5dabfc78 5256 err = ext4_init_mballoc();
ac27a0ec 5257 if (err)
6fd058f7 5258 goto out3;
c9de560d 5259
5dabfc78 5260 err = ext4_init_xattr();
c9de560d
AT		 5261 if (err)
5262 goto out2;
ac27a0ec
DK		 5263 err = init_inodecache();
5264 if (err)
5265 goto out1;
24b58424 5266 register_as_ext3();
2035e776 5267 register_as_ext2();
03010a33 5268 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5269 if (err)
5270 goto out;
bfff6873 5271
ac27a0ec
DK		 5272 return 0;
5273out:
24b58424
TT		 5274 unregister_as_ext2();
5275 unregister_as_ext3();
ac27a0ec
DK		 5276 destroy_inodecache();
5277out1:
5dabfc78 5278 ext4_exit_xattr();
c9de560d 5279out2:
5dabfc78 5280 ext4_exit_mballoc();
6fd058f7 5281out3:
8f021222 5282 ext4_exit_feat_adverts();
dd68314c 5283out4:
d44651d0
FJ		 5284 if (ext4_proc_root)
5285 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5286 kset_unregister(ext4_kset);
d44651d0 5287out5:
5dabfc78 5288 ext4_exit_system_zone();
d44651d0 5289out6:
5dabfc78 5290 ext4_exit_pageio();
51865fda
ZL		 5291out7:
5292 ext4_exit_es();
5293
ac27a0ec
DK		 5294 return err;
5295}
5296
5dabfc78 5297static void __exit ext4_exit_fs(void)
ac27a0ec 5298{
bfff6873 5299 ext4_destroy_lazyinit_thread();
24b58424
TT		 5300 unregister_as_ext2();
5301 unregister_as_ext3();
03010a33 5302 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5303 destroy_inodecache();
5dabfc78
TT		 5304 ext4_exit_xattr();
5305 ext4_exit_mballoc();
8f021222 5306 ext4_exit_feat_adverts();
9f6200bb 5307 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5308 kset_unregister(ext4_kset);
5dabfc78
TT		 5309 ext4_exit_system_zone();
5310 ext4_exit_pageio();
ac27a0ec
DK		 5311}
5312
5313MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5314MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5315MODULE_LICENSE("GPL");
5dabfc78
TT		 5316module_init(ext4_init_fs)
5317module_exit(ext4_exit_fs)
Linux 6.x block layer and io_uring tree(s)