]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: read-only support for bigalloc file systems
[mirror_ubuntu-bionic-kernel.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
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
CH		 47#include "ext4.h"
48#include "ext4_jbd2.h"
ac27a0ec
DK		 49#include "xattr.h"
50#include "acl.h"
3661d286 51#include "mballoc.h"
ac27a0ec 52
9bffad1e
TT		 53#define CREATE_TRACE_POINTS
54#include <trace/events/ext4.h>
55
1f109d5a 56static struct proc_dir_entry *ext4_proc_root;
3197ebdb 57static struct kset *ext4_kset;
0b75a840
LC		 58static struct ext4_lazy_init *ext4_li_info;
59static struct mutex ext4_li_mtx;
60static struct ext4_features *ext4_feat;
9f6200bb 61
617ba13b 62static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 63 unsigned long journal_devnum);
e2d67052 64static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 65static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
617ba13b 69static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 70static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 71 char nbuf[16]);
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);
2b2d6d01 75static void ext4_write_super(struct super_block *sb);
c4be0c1d 76static int ext4_freeze(struct super_block *sb);
152a0836
AV		 77static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
2035e776
TT		 79static inline int ext2_feature_set_ok(struct super_block *sb);
80static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 81static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 82static void ext4_destroy_lazyinit_thread(void);
83static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 84static void ext4_clear_request_list(void);
ac27a0ec 85
2035e776
TT		 86#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87static struct file_system_type ext2_fs_type = {
88 .owner = THIS_MODULE,
89 .name = "ext2",
90 .mount = ext4_mount,
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
93};
94#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
95#else
96#define IS_EXT2_SB(sb) (0)
97#endif
98
99
ba69f9ab
JK		 100#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
101static struct file_system_type ext3_fs_type = {
102 .owner = THIS_MODULE,
103 .name = "ext3",
152a0836 104 .mount = ext4_mount,
ba69f9ab
JK		 105 .kill_sb = kill_block_super,
106 .fs_flags = FS_REQUIRES_DEV,
107};
108#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
109#else
110#define IS_EXT3_SB(sb) (0)
111#endif
bd81d8ee 112
9933fc0a
TT		 113void *ext4_kvmalloc(size_t size, gfp_t flags)
114{
115 void *ret;
116
117 ret = kmalloc(size, flags);
118 if (!ret)
119 ret = __vmalloc(size, flags, PAGE_KERNEL);
120 return ret;
121}
122
123void *ext4_kvzalloc(size_t size, gfp_t flags)
124{
125 void *ret;
126
db9481c0 127 ret = kzalloc(size, flags);
9933fc0a
TT		 128 if (!ret)
129 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
130 return ret;
131}
132
133void ext4_kvfree(void *ptr)
134{
135 if (is_vmalloc_addr(ptr))
136 vfree(ptr);
137 else
138 kfree(ptr);
139
140}
141
8fadc143
AR		 142ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
143 struct ext4_group_desc *bg)
bd81d8ee 144{
3a14589c 145 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 146 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 147 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 148}
149
8fadc143
AR		 150ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
151 struct ext4_group_desc *bg)
bd81d8ee 152{
5272f837 153 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 154 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 155 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 156}
157
8fadc143
AR		 158ext4_fsblk_t ext4_inode_table(struct super_block *sb,
159 struct ext4_group_desc *bg)
bd81d8ee 160{
5272f837 161 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 162 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 163 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 164}
165
560671a0
AK		 166__u32 ext4_free_blks_count(struct super_block *sb,
167 struct ext4_group_desc *bg)
168{
169 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
170 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 171 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 172}
173
174__u32 ext4_free_inodes_count(struct super_block *sb,
175 struct ext4_group_desc *bg)
176{
177 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
178 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 179 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 180}
181
182__u32 ext4_used_dirs_count(struct super_block *sb,
183 struct ext4_group_desc *bg)
184{
185 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
186 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 187 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 188}
189
190__u32 ext4_itable_unused_count(struct super_block *sb,
191 struct ext4_group_desc *bg)
192{
193 return le16_to_cpu(bg->bg_itable_unused_lo) |
194 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 195 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 196}
197
8fadc143
AR		 198void ext4_block_bitmap_set(struct super_block *sb,
199 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 200{
3a14589c 201 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 202 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
203 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 204}
205
8fadc143
AR		 206void ext4_inode_bitmap_set(struct super_block *sb,
207 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 208{
5272f837 209 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 210 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
211 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 212}
213
8fadc143
AR		 214void ext4_inode_table_set(struct super_block *sb,
215 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 216{
5272f837 217 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 218 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
219 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 220}
221
560671a0
AK		 222void ext4_free_blks_set(struct super_block *sb,
223 struct ext4_group_desc *bg, __u32 count)
224{
225 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
226 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
227 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
228}
229
230void ext4_free_inodes_set(struct super_block *sb,
231 struct ext4_group_desc *bg, __u32 count)
232{
233 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
234 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
235 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
236}
237
238void ext4_used_dirs_set(struct super_block *sb,
239 struct ext4_group_desc *bg, __u32 count)
240{
241 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
242 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
243 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
244}
245
246void ext4_itable_unused_set(struct super_block *sb,
247 struct ext4_group_desc *bg, __u32 count)
248{
249 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
250 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
251 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
252}
253
d3d1faf6
CW		 254
255/* Just increment the non-pointer handle value */
256static handle_t *ext4_get_nojournal(void)
257{
258 handle_t *handle = current->journal_info;
259 unsigned long ref_cnt = (unsigned long)handle;
260
261 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
262
263 ref_cnt++;
264 handle = (handle_t *)ref_cnt;
265
266 current->journal_info = handle;
267 return handle;
268}
269
270
271/* Decrement the non-pointer handle value */
272static void ext4_put_nojournal(handle_t *handle)
273{
274 unsigned long ref_cnt = (unsigned long)handle;
275
276 BUG_ON(ref_cnt == 0);
277
278 ref_cnt--;
279 handle = (handle_t *)ref_cnt;
280
281 current->journal_info = handle;
282}
283
ac27a0ec 284/*
dab291af 285 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 286 *
287 * The only special thing we need to do here is to make sure that all
288 * journal_end calls result in the superblock being marked dirty, so
289 * that sync() will call the filesystem's write_super callback if
290 * appropriate.
be4f27d3
YY		 291 *
292 * To avoid j_barrier hold in userspace when a user calls freeze(),
293 * ext4 prevents a new handle from being started by s_frozen, which
294 * is in an upper layer.
ac27a0ec 295 */
617ba13b 296handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 297{
298 journal_t *journal;
be4f27d3 299 handle_t *handle;
ac27a0ec 300
12706394 301 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 302 if (sb->s_flags & MS_RDONLY)
303 return ERR_PTR(-EROFS);
304
617ba13b 305 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 306 handle = ext4_journal_current_handle();
307
308 /*
309 * If a handle has been started, it should be allowed to
310 * finish, otherwise deadlock could happen between freeze
311 * and others(e.g. truncate) due to the restart of the
312 * journal handle if the filesystem is forzen and active
313 * handles are not stopped.
314 */
315 if (!handle)
316 vfs_check_frozen(sb, SB_FREEZE_TRANS);
317
318 if (!journal)
319 return ext4_get_nojournal();
320 /*
321 * Special case here: if the journal has aborted behind our
322 * backs (eg. EIO in the commit thread), then we still need to
323 * take the FS itself readonly cleanly.
324 */
325 if (is_journal_aborted(journal)) {
326 ext4_abort(sb, "Detected aborted journal");
327 return ERR_PTR(-EROFS);
ac27a0ec 328 }
be4f27d3 329 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 330}
331
332/*
333 * The only special thing we need to do here is to make sure that all
dab291af 334 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 335 * that sync() will call the filesystem's write_super callback if
336 * appropriate.
337 */
c398eda0 338int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 339{
340 struct super_block *sb;
341 int err;
342 int rc;
343
0390131b 344 if (!ext4_handle_valid(handle)) {
d3d1faf6 345 ext4_put_nojournal(handle);
0390131b
FM		 346 return 0;
347 }
ac27a0ec
DK		 348 sb = handle->h_transaction->t_journal->j_private;
349 err = handle->h_err;
dab291af 350 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 351
352 if (!err)
353 err = rc;
354 if (err)
c398eda0 355 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 356 return err;
357}
358
90c7201b
TT		 359void ext4_journal_abort_handle(const char *caller, unsigned int line,
360 const char *err_fn, struct buffer_head *bh,
361 handle_t *handle, int err)
ac27a0ec
DK		 362{
363 char nbuf[16];
617ba13b 364 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 365
0390131b
FM		 366 BUG_ON(!ext4_handle_valid(handle));
367
ac27a0ec
DK		 368 if (bh)
369 BUFFER_TRACE(bh, "abort");
370
371 if (!handle->h_err)
372 handle->h_err = err;
373
374 if (is_handle_aborted(handle))
375 return;
376
90c7201b
TT		 377 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
378 caller, line, errstr, err_fn);
ac27a0ec 379
dab291af 380 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 381}
382
1c13d5c0
TT		 383static void __save_error_info(struct super_block *sb, const char *func,
384 unsigned int line)
385{
386 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
387
388 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
389 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
390 es->s_last_error_time = cpu_to_le32(get_seconds());
391 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
392 es->s_last_error_line = cpu_to_le32(line);
393 if (!es->s_first_error_time) {
394 es->s_first_error_time = es->s_last_error_time;
395 strncpy(es->s_first_error_func, func,
396 sizeof(es->s_first_error_func));
397 es->s_first_error_line = cpu_to_le32(line);
398 es->s_first_error_ino = es->s_last_error_ino;
399 es->s_first_error_block = es->s_last_error_block;
400 }
66e61a9e
TT		 401 /*
402 * Start the daily error reporting function if it hasn't been
403 * started already
404 */
405 if (!es->s_error_count)
406 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 407 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
408}
409
410static void save_error_info(struct super_block *sb, const char *func,
411 unsigned int line)
412{
413 __save_error_info(sb, func, line);
414 ext4_commit_super(sb, 1);
415}
416
7c2e7087
TT		 417/*
418 * The del_gendisk() function uninitializes the disk-specific data
419 * structures, including the bdi structure, without telling anyone
420 * else. Once this happens, any attempt to call mark_buffer_dirty()
421 * (for example, by ext4_commit_super), will cause a kernel OOPS.
422 * This is a kludge to prevent these oops until we can put in a proper
423 * hook in del_gendisk() to inform the VFS and file system layers.
424 */
425static int block_device_ejected(struct super_block *sb)
426{
427 struct inode *bd_inode = sb->s_bdev->bd_inode;
428 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
429
430 return bdi->dev == NULL;
431}
432
1c13d5c0 433
ac27a0ec
DK		 434/* Deal with the reporting of failure conditions on a filesystem such as
435 * inconsistencies detected or read IO failures.
436 *
437 * On ext2, we can store the error state of the filesystem in the
617ba13b 438 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 439 * write ordering constraints on the superblock which prevent us from
440 * writing it out straight away; and given that the journal is about to
441 * be aborted, we can't rely on the current, or future, transactions to
442 * write out the superblock safely.
443 *
dab291af 444 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 445 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 446 * that error until we've noted it down and cleared it.
447 */
448
617ba13b 449static void ext4_handle_error(struct super_block *sb)
ac27a0ec 450{
ac27a0ec
DK		 451 if (sb->s_flags & MS_RDONLY)
452 return;
453
2b2d6d01 454 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 455 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 456
4ab2f15b 457 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 458 if (journal)
dab291af 459 jbd2_journal_abort(journal, -EIO);
ac27a0ec 460 }
2b2d6d01 461 if (test_opt(sb, ERRORS_RO)) {
b31e1552 462 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 463 sb->s_flags |= MS_RDONLY;
464 }
ac27a0ec 465 if (test_opt(sb, ERRORS_PANIC))
617ba13b 466 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 467 sb->s_id);
468}
469
12062ddd 470void __ext4_error(struct super_block *sb, const char *function,
c398eda0 471 unsigned int line, const char *fmt, ...)
ac27a0ec 472{
0ff2ea7d 473 struct va_format vaf;
ac27a0ec
DK		 474 va_list args;
475
476 va_start(args, fmt);
0ff2ea7d
JP		 477 vaf.fmt = fmt;
478 vaf.va = &args;
479 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
480 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec
DK		 481 va_end(args);
482
617ba13b 483 ext4_handle_error(sb);
ac27a0ec
DK		 484}
485
c398eda0
TT		 486void ext4_error_inode(struct inode *inode, const char *function,
487 unsigned int line, ext4_fsblk_t block,
273df556
FM		 488 const char *fmt, ...)
489{
490 va_list args;
f7c21177 491 struct va_format vaf;
1c13d5c0 492 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 493
1c13d5c0
TT		 494 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
495 es->s_last_error_block = cpu_to_le64(block);
496 save_error_info(inode->i_sb, function, line);
273df556 497 va_start(args, fmt);
f7c21177
TT		 498 vaf.fmt = fmt;
499 vaf.va = &args;
c398eda0
TT		 500 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
501 inode->i_sb->s_id, function, line, inode->i_ino);
502 if (block)
f7c21177
TT		 503 printk(KERN_CONT "block %llu: ", block);
504 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
273df556
FM		 505 va_end(args);
506
507 ext4_handle_error(inode->i_sb);
508}
509
c398eda0 510void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 511 unsigned int line, ext4_fsblk_t block,
512 const char *fmt, ...)
273df556
FM		 513{
514 va_list args;
f7c21177 515 struct va_format vaf;
1c13d5c0 516 struct ext4_super_block *es;
273df556
FM		 517 struct inode *inode = file->f_dentry->d_inode;
518 char pathname[80], *path;
519
1c13d5c0
TT		 520 es = EXT4_SB(inode->i_sb)->s_es;
521 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
522 save_error_info(inode->i_sb, function, line);
273df556 523 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 524 if (IS_ERR(path))
273df556
FM		 525 path = "(unknown)";
526 printk(KERN_CRIT
f7c21177
TT		 527 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
528 inode->i_sb->s_id, function, line, inode->i_ino);
529 if (block)
530 printk(KERN_CONT "block %llu: ", block);
531 va_start(args, fmt);
532 vaf.fmt = fmt;
533 vaf.va = &args;
534 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
273df556
FM		 535 va_end(args);
536
537 ext4_handle_error(inode->i_sb);
538}
539
2b2d6d01 540static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 541 char nbuf[16])
542{
543 char *errstr = NULL;
544
545 switch (errno) {
546 case -EIO:
547 errstr = "IO failure";
548 break;
549 case -ENOMEM:
550 errstr = "Out of memory";
551 break;
552 case -EROFS:
78f1ddbb
TT		 553 if (!sb || (EXT4_SB(sb)->s_journal &&
554 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 555 errstr = "Journal has aborted";
556 else
557 errstr = "Readonly filesystem";
558 break;
559 default:
560 /* If the caller passed in an extra buffer for unknown
561 * errors, textualise them now. Else we just return
562 * NULL. */
563 if (nbuf) {
564 /* Check for truncated error codes... */
565 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
566 errstr = nbuf;
567 }
568 break;
569 }
570
571 return errstr;
572}
573
617ba13b 574/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 575 * automatically and invokes the appropriate error response. */
576
c398eda0
TT		 577void __ext4_std_error(struct super_block *sb, const char *function,
578 unsigned int line, int errno)
ac27a0ec
DK		 579{
580 char nbuf[16];
581 const char *errstr;
582
583 /* Special case: if the error is EROFS, and we're not already
584 * inside a transaction, then there's really no point in logging
585 * an error. */
586 if (errno == -EROFS && journal_current_handle() == NULL &&
587 (sb->s_flags & MS_RDONLY))
588 return;
589
617ba13b 590 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 591 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
592 sb->s_id, function, line, errstr);
1c13d5c0 593 save_error_info(sb, function, line);
ac27a0ec 594
617ba13b 595 ext4_handle_error(sb);
ac27a0ec
DK		 596}
597
598/*
617ba13b 599 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 600 * abort function may be used to deal with unrecoverable failures such
601 * as journal IO errors or ENOMEM at a critical moment in log management.
602 *
603 * We unconditionally force the filesystem into an ABORT|READONLY state,
604 * unless the error response on the fs has been set to panic in which
605 * case we take the easy way out and panic immediately.
606 */
607
c67d859e 608void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 609 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 610{
611 va_list args;
612
1c13d5c0 613 save_error_info(sb, function, line);
ac27a0ec 614 va_start(args, fmt);
c398eda0
TT		 615 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
616 function, line);
ac27a0ec
DK		 617 vprintk(fmt, args);
618 printk("\n");
619 va_end(args);
620
1c13d5c0
TT		 621 if ((sb->s_flags & MS_RDONLY) == 0) {
622 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
623 sb->s_flags |= MS_RDONLY;
624 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
625 if (EXT4_SB(sb)->s_journal)
626 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
627 save_error_info(sb, function, line);
628 }
ac27a0ec 629 if (test_opt(sb, ERRORS_PANIC))
617ba13b 630 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 631}
632
0ff2ea7d 633void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 634{
0ff2ea7d 635 struct va_format vaf;
b31e1552
ES		 636 va_list args;
637
638 va_start(args, fmt);
0ff2ea7d
JP		 639 vaf.fmt = fmt;
640 vaf.va = &args;
641 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 642 va_end(args);
643}
644
12062ddd 645void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 646 unsigned int line, const char *fmt, ...)
ac27a0ec 647{
0ff2ea7d 648 struct va_format vaf;
ac27a0ec
DK		 649 va_list args;
650
651 va_start(args, fmt);
0ff2ea7d
JP		 652 vaf.fmt = fmt;
653 vaf.va = &args;
654 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
655 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 656 va_end(args);
657}
658
e29136f8
TT		 659void __ext4_grp_locked_error(const char *function, unsigned int line,
660 struct super_block *sb, ext4_group_t grp,
661 unsigned long ino, ext4_fsblk_t block,
662 const char *fmt, ...)
5d1b1b3f
AK		 663__releases(bitlock)
664__acquires(bitlock)
665{
0ff2ea7d 666 struct va_format vaf;
5d1b1b3f
AK		 667 va_list args;
668 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
669
1c13d5c0
TT		 670 es->s_last_error_ino = cpu_to_le32(ino);
671 es->s_last_error_block = cpu_to_le64(block);
672 __save_error_info(sb, function, line);
0ff2ea7d 673
5d1b1b3f 674 va_start(args, fmt);
0ff2ea7d
JP		 675
676 vaf.fmt = fmt;
677 vaf.va = &args;
21149d61 678 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 679 sb->s_id, function, line, grp);
680 if (ino)
0ff2ea7d 681 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 682 if (block)
0ff2ea7d
JP		 683 printk(KERN_CONT "block %llu:", (unsigned long long) block);
684 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 685 va_end(args);
686
687 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 688 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 689 return;
690 }
1c13d5c0 691
5d1b1b3f
AK		 692 ext4_unlock_group(sb, grp);
693 ext4_handle_error(sb);
694 /*
695 * We only get here in the ERRORS_RO case; relocking the group
696 * may be dangerous, but nothing bad will happen since the
697 * filesystem will have already been marked read/only and the
698 * journal has been aborted. We return 1 as a hint to callers
699 * who might what to use the return value from
25985edc 700 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 701 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
702 * aggressively from the ext4 function in question, with a
703 * more appropriate error code.
704 */
705 ext4_lock_group(sb, grp);
706 return;
707}
708
617ba13b 709void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 710{
617ba13b 711 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 712
617ba13b 713 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 714 return;
715
12062ddd 716 ext4_warning(sb,
ac27a0ec
DK		 717 "updating to rev %d because of new feature flag, "
718 "running e2fsck is recommended",
617ba13b 719 EXT4_DYNAMIC_REV);
ac27a0ec 720
617ba13b
MC		 721 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
722 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
723 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 724 /* leave es->s_feature_*compat flags alone */
725 /* es->s_uuid will be set by e2fsck if empty */
726
727 /*
728 * The rest of the superblock fields should be zero, and if not it
729 * means they are likely already in use, so leave them alone. We
730 * can leave it up to e2fsck to clean up any inconsistencies there.
731 */
732}
733
734/*
735 * Open the external journal device
736 */
b31e1552 737static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 738{
739 struct block_device *bdev;
740 char b[BDEVNAME_SIZE];
741
d4d77629 742 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 743 if (IS_ERR(bdev))
744 goto fail;
745 return bdev;
746
747fail:
b31e1552 748 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 749 __bdevname(dev, b), PTR_ERR(bdev));
750 return NULL;
751}
752
753/*
754 * Release the journal device
755 */
617ba13b 756static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 757{
e525fd89 758 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 759}
760
617ba13b 761static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 762{
763 struct block_device *bdev;
764 int ret = -ENODEV;
765
766 bdev = sbi->journal_bdev;
767 if (bdev) {
617ba13b 768 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 769 sbi->journal_bdev = NULL;
770 }
771 return ret;
772}
773
774static inline struct inode *orphan_list_entry(struct list_head *l)
775{
617ba13b 776 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 777}
778
617ba13b 779static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 780{
781 struct list_head *l;
782
b31e1552
ES		 783 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
784 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 785
786 printk(KERN_ERR "sb_info orphan list:\n");
787 list_for_each(l, &sbi->s_orphan) {
788 struct inode *inode = orphan_list_entry(l);
789 printk(KERN_ERR " "
790 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
791 inode->i_sb->s_id, inode->i_ino, inode,
792 inode->i_mode, inode->i_nlink,
793 NEXT_ORPHAN(inode));
794 }
795}
796
2b2d6d01 797static void ext4_put_super(struct super_block *sb)
ac27a0ec 798{
617ba13b
MC		 799 struct ext4_sb_info *sbi = EXT4_SB(sb);
800 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 801 int i, err;
ac27a0ec 802
857ac889 803 ext4_unregister_li_request(sb);
e0ccfd95
CH		 804 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
805
4c0425ff
MC		 806 flush_workqueue(sbi->dio_unwritten_wq);
807 destroy_workqueue(sbi->dio_unwritten_wq);
808
a9e220f8 809 lock_super(sb);
8c85e125 810 if (sb->s_dirt)
ebc1ac16 811 ext4_commit_super(sb, 1);
8c85e125 812
0390131b
FM		 813 if (sbi->s_journal) {
814 err = jbd2_journal_destroy(sbi->s_journal);
815 sbi->s_journal = NULL;
816 if (err < 0)
c67d859e 817 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 818 }
d4edac31 819
a1c6c569 820 del_timer(&sbi->s_err_report);
d4edac31
JB		 821 ext4_release_system_zone(sb);
822 ext4_mb_release(sb);
823 ext4_ext_release(sb);
824 ext4_xattr_put_super(sb);
825
ac27a0ec 826 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 827 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 828 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 829 ext4_commit_super(sb, 1);
ac27a0ec 830 }
240799cd 831 if (sbi->s_proc) {
9f6200bb 832 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 833 }
3197ebdb 834 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 835
836 for (i = 0; i < sbi->s_gdb_count; i++)
837 brelse(sbi->s_group_desc[i]);
f18a5f21 838 ext4_kvfree(sbi->s_group_desc);
9933fc0a 839 ext4_kvfree(sbi->s_flex_groups);
ac27a0ec
DK		 840 percpu_counter_destroy(&sbi->s_freeblocks_counter);
841 percpu_counter_destroy(&sbi->s_freeinodes_counter);
842 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 843 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 844 brelse(sbi->s_sbh);
845#ifdef CONFIG_QUOTA
846 for (i = 0; i < MAXQUOTAS; i++)
847 kfree(sbi->s_qf_names[i]);
848#endif
849
850 /* Debugging code just in case the in-memory inode orphan list
851 * isn't empty. The on-disk one can be non-empty if we've
852 * detected an error and taken the fs readonly, but the
853 * in-memory list had better be clean by this point. */
854 if (!list_empty(&sbi->s_orphan))
855 dump_orphan_list(sb, sbi);
856 J_ASSERT(list_empty(&sbi->s_orphan));
857
f98393a6 858 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 859 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
860 /*
861 * Invalidate the journal device's buffers. We don't want them
862 * floating about in memory - the physical journal device may
863 * hotswapped, and it breaks the `ro-after' testing code.
864 */
865 sync_blockdev(sbi->journal_bdev);
f98393a6 866 invalidate_bdev(sbi->journal_bdev);
617ba13b 867 ext4_blkdev_remove(sbi);
ac27a0ec 868 }
c5e06d10
JL		 869 if (sbi->s_mmp_tsk)
870 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 871 sb->s_fs_info = NULL;
3197ebdb
TT		 872 /*
873 * Now that we are completely done shutting down the
874 * superblock, we need to actually destroy the kobject.
875 */
3197ebdb
TT		 876 unlock_super(sb);
877 kobject_put(&sbi->s_kobj);
878 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 879 kfree(sbi->s_blockgroup_lock);
ac27a0ec 880 kfree(sbi);
ac27a0ec
DK		 881}
882
e18b890b 883static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 884
885/*
886 * Called inside transaction, so use GFP_NOFS
887 */
617ba13b 888static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 889{
617ba13b 890 struct ext4_inode_info *ei;
ac27a0ec 891
e6b4f8da 892 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 893 if (!ei)
894 return NULL;
0b8e58a1 895
ac27a0ec 896 ei->vfs_inode.i_version = 1;
91246c00 897 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 898 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 899 INIT_LIST_HEAD(&ei->i_prealloc_list);
900 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 901 ei->i_reserved_data_blocks = 0;
902 ei->i_reserved_meta_blocks = 0;
903 ei->i_allocated_meta_blocks = 0;
9d0be502 904 ei->i_da_metadata_calc_len = 0;
d2a17637 905 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 906#ifdef CONFIG_QUOTA
907 ei->i_reserved_quota = 0;
908#endif
8aefcd55 909 ei->jinode = NULL;
c7064ef1 910 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 911 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 912 ei->cur_aio_dio = NULL;
b436b9be
JK		 913 ei->i_sync_tid = 0;
914 ei->i_datasync_tid = 0;
f7ad6d2e 915 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 916 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 917
ac27a0ec
DK		 918 return &ei->vfs_inode;
919}
920
7ff9c073
TT		 921static int ext4_drop_inode(struct inode *inode)
922{
923 int drop = generic_drop_inode(inode);
924
925 trace_ext4_drop_inode(inode, drop);
926 return drop;
927}
928
fa0d7e3d
NP		 929static void ext4_i_callback(struct rcu_head *head)
930{
931 struct inode *inode = container_of(head, struct inode, i_rcu);
932 INIT_LIST_HEAD(&inode->i_dentry);
933 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
934}
935
617ba13b 936static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 937{
9f7dd93d 938 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 939 ext4_msg(inode->i_sb, KERN_ERR,
940 "Inode %lu (%p): orphan list check failed!",
941 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 942 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
943 EXT4_I(inode), sizeof(struct ext4_inode_info),
944 true);
945 dump_stack();
946 }
fa0d7e3d 947 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 948}
949
51cc5068 950static void init_once(void *foo)
ac27a0ec 951{
617ba13b 952 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 953
a35afb83 954 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 955#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 956 init_rwsem(&ei->xattr_sem);
ac27a0ec 957#endif
0e855ac8 958 init_rwsem(&ei->i_data_sem);
a35afb83 959 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 960}
961
962static int init_inodecache(void)
963{
617ba13b
MC		 964 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
965 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 966 0, (SLAB_RECLAIM_ACCOUNT|
967 SLAB_MEM_SPREAD),
20c2df83 968 init_once);
617ba13b 969 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 970 return -ENOMEM;
971 return 0;
972}
973
974static void destroy_inodecache(void)
975{
617ba13b 976 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 977}
978
0930fcc1 979void ext4_clear_inode(struct inode *inode)
ac27a0ec 980{
0930fcc1
AV		 981 invalidate_inode_buffers(inode);
982 end_writeback(inode);
9f754758 983 dquot_drop(inode);
c2ea3fde 984 ext4_discard_preallocations(inode);
8aefcd55
TT		 985 if (EXT4_I(inode)->jinode) {
986 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
987 EXT4_I(inode)->jinode);
988 jbd2_free_inode(EXT4_I(inode)->jinode);
989 EXT4_I(inode)->jinode = NULL;
990 }
ac27a0ec
DK		 991}
992
2b2d6d01
TT		 993static inline void ext4_show_quota_options(struct seq_file *seq,
994 struct super_block *sb)
ac27a0ec
DK		 995{
996#if defined(CONFIG_QUOTA)
617ba13b 997 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 998
5a20bdfc
JK		 999 if (sbi->s_jquota_fmt) {
1000 char *fmtname = "";
1001
1002 switch (sbi->s_jquota_fmt) {
1003 case QFMT_VFS_OLD:
1004 fmtname = "vfsold";
1005 break;
1006 case QFMT_VFS_V0:
1007 fmtname = "vfsv0";
1008 break;
1009 case QFMT_VFS_V1:
1010 fmtname = "vfsv1";
1011 break;
1012 }
1013 seq_printf(seq, ",jqfmt=%s", fmtname);
1014 }
ac27a0ec
DK		 1015
1016 if (sbi->s_qf_names[USRQUOTA])
1017 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1018
1019 if (sbi->s_qf_names[GRPQUOTA])
1020 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1021
482a7425 1022 if (test_opt(sb, USRQUOTA))
ac27a0ec
DK		 1023 seq_puts(seq, ",usrquota");
1024
482a7425 1025 if (test_opt(sb, GRPQUOTA))
ac27a0ec
DK		 1026 seq_puts(seq, ",grpquota");
1027#endif
1028}
1029
d9c9bef1
MS		 1030/*
1031 * Show an option if
1032 * - it's set to a non-default value OR
1033 * - if the per-sb default is different from the global default
1034 */
617ba13b 1035static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 1036{
aa22df2c
AK		 1037 int def_errors;
1038 unsigned long def_mount_opts;
ac27a0ec 1039 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 1040 struct ext4_sb_info *sbi = EXT4_SB(sb);
1041 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 1042
1043 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 1044 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 1045
1046 if (sbi->s_sb_block != 1)
1047 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1048 if (test_opt(sb, MINIX_DF))
1049 seq_puts(seq, ",minixdf");
aa22df2c 1050 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 1051 seq_puts(seq, ",grpid");
1052 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1053 seq_puts(seq, ",nogrpid");
1054 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1055 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1056 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1057 }
1058 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1059 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1060 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1061 }
bb4f397a 1062 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 1063 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 1064 def_errors == EXT4_ERRORS_CONTINUE) {
1065 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 1066 }
1067 }
aa22df2c 1068 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 1069 seq_puts(seq, ",errors=continue");
aa22df2c 1070 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 1071 seq_puts(seq, ",errors=panic");
aa22df2c 1072 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 1073 seq_puts(seq, ",nouid32");
aa22df2c 1074 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 1075 seq_puts(seq, ",debug");
1076 if (test_opt(sb, OLDALLOC))
1077 seq_puts(seq, ",oldalloc");
03010a33 1078#ifdef CONFIG_EXT4_FS_XATTR
ea663336 1079 if (test_opt(sb, XATTR_USER))
d9c9bef1 1080 seq_puts(seq, ",user_xattr");
ea663336 1081 if (!test_opt(sb, XATTR_USER))
d9c9bef1 1082 seq_puts(seq, ",nouser_xattr");
d9c9bef1 1083#endif
03010a33 1084#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 1085 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 1086 seq_puts(seq, ",acl");
1087 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1088 seq_puts(seq, ",noacl");
1089#endif
30773840 1090 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 1091 seq_printf(seq, ",commit=%u",
1092 (unsigned) (sbi->s_commit_interval / HZ));
1093 }
30773840
TT		 1094 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1095 seq_printf(seq, ",min_batch_time=%u",
1096 (unsigned) sbi->s_min_batch_time);
1097 }
1098 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1099 seq_printf(seq, ",max_batch_time=%u",
1100 (unsigned) sbi->s_min_batch_time);
1101 }
1102
571640ca
ES		 1103 /*
1104 * We're changing the default of barrier mount option, so
1105 * let's always display its mount state so it's clear what its
1106 * status is.
1107 */
1108 seq_puts(seq, ",barrier=");
1109 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 1110 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1111 seq_puts(seq, ",journal_async_commit");
39a4bade
JK		 1112 else if (test_opt(sb, JOURNAL_CHECKSUM))
1113 seq_puts(seq, ",journal_checksum");
25ec56b5
JNC		 1114 if (test_opt(sb, I_VERSION))
1115 seq_puts(seq, ",i_version");
8b67f04a
TT		 1116 if (!test_opt(sb, DELALLOC) &&
1117 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
dd919b98
AK		 1118 seq_puts(seq, ",nodelalloc");
1119
6fd7a467
TT		 1120 if (!test_opt(sb, MBLK_IO_SUBMIT))
1121 seq_puts(seq, ",nomblk_io_submit");
cb45bbe4
MS		 1122 if (sbi->s_stripe)
1123 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 1124 /*
1125 * journal mode get enabled in different ways
1126 * So just print the value even if we didn't specify it
1127 */
617ba13b 1128 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 1129 seq_puts(seq, ",data=journal");
617ba13b 1130 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 1131 seq_puts(seq, ",data=ordered");
617ba13b 1132 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 1133 seq_puts(seq, ",data=writeback");
1134
240799cd
TT		 1135 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1136 seq_printf(seq, ",inode_readahead_blks=%u",
1137 sbi->s_inode_readahead_blks);
1138
5bf5683a
HK		 1139 if (test_opt(sb, DATA_ERR_ABORT))
1140 seq_puts(seq, ",data_err=abort");
1141
afd4672d 1142 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 1143 seq_puts(seq, ",noauto_da_alloc");
afd4672d 1144
8b67f04a 1145 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
5328e635
ES		 1146 seq_puts(seq, ",discard");
1147
e3bb52ae
ES		 1148 if (test_opt(sb, NOLOAD))
1149 seq_puts(seq, ",norecovery");
1150
744692dc
JZ		 1151 if (test_opt(sb, DIOREAD_NOLOCK))
1152 seq_puts(seq, ",dioread_nolock");
1153
8b67f04a
TT		 1154 if (test_opt(sb, BLOCK_VALIDITY) &&
1155 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1156 seq_puts(seq, ",block_validity");
1157
bfff6873
LC		 1158 if (!test_opt(sb, INIT_INODE_TABLE))
1159 seq_puts(seq, ",noinit_inode_table");
51ce6511 1160 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
bfff6873
LC		 1161 seq_printf(seq, ",init_inode_table=%u",
1162 (unsigned) sbi->s_li_wait_mult);
1163
617ba13b 1164 ext4_show_quota_options(seq, sb);
0b8e58a1 1165
ac27a0ec
DK		 1166 return 0;
1167}
1168
1b961ac0 1169static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1170 u64 ino, u32 generation)
ac27a0ec 1171{
ac27a0ec 1172 struct inode *inode;
ac27a0ec 1173
617ba13b 1174 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1175 return ERR_PTR(-ESTALE);
617ba13b 1176 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1177 return ERR_PTR(-ESTALE);
1178
1179 /* iget isn't really right if the inode is currently unallocated!!
1180 *
617ba13b 1181 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1182 * deleted, so we should be safe.
1183 *
1184 * Currently we don't know the generation for parent directory, so
1185 * a generation of 0 means "accept any"
1186 */
1d1fe1ee
DH		 1187 inode = ext4_iget(sb, ino);
1188 if (IS_ERR(inode))
1189 return ERR_CAST(inode);
1190 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1191 iput(inode);
1192 return ERR_PTR(-ESTALE);
1193 }
1b961ac0
CH		 1194
1195 return inode;
1196}
1197
1198static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1199 int fh_len, int fh_type)
1b961ac0
CH		 1200{
1201 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1202 ext4_nfs_get_inode);
1203}
1204
1205static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1206 int fh_len, int fh_type)
1b961ac0
CH		 1207{
1208 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1209 ext4_nfs_get_inode);
ac27a0ec
DK		 1210}
1211
c39a7f84
TO		 1212/*
1213 * Try to release metadata pages (indirect blocks, directories) which are
1214 * mapped via the block device. Since these pages could have journal heads
1215 * which would prevent try_to_free_buffers() from freeing them, we must use
1216 * jbd2 layer's try_to_free_buffers() function to release them.
1217 */
0b8e58a1
AD		 1218static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1219 gfp_t wait)
c39a7f84
TO		 1220{
1221 journal_t *journal = EXT4_SB(sb)->s_journal;
1222
1223 WARN_ON(PageChecked(page));
1224 if (!page_has_buffers(page))
1225 return 0;
1226 if (journal)
1227 return jbd2_journal_try_to_free_buffers(journal, page,
1228 wait & ~__GFP_WAIT);
1229 return try_to_free_buffers(page);
1230}
1231
ac27a0ec 1232#ifdef CONFIG_QUOTA
af5bc92d 1233#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1234#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1235
617ba13b
MC		 1236static int ext4_write_dquot(struct dquot *dquot);
1237static int ext4_acquire_dquot(struct dquot *dquot);
1238static int ext4_release_dquot(struct dquot *dquot);
1239static int ext4_mark_dquot_dirty(struct dquot *dquot);
1240static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1241static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1242 struct path *path);
ca0e05e4 1243static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1244static int ext4_quota_on_mount(struct super_block *sb, int type);
1245static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1246 size_t len, loff_t off);
617ba13b 1247static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1248 const char *data, size_t len, loff_t off);
1249
61e225dc 1250static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1251 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1252 .write_dquot = ext4_write_dquot,
1253 .acquire_dquot = ext4_acquire_dquot,
1254 .release_dquot = ext4_release_dquot,
1255 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1256 .write_info = ext4_write_info,
1257 .alloc_dquot = dquot_alloc,
1258 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1259};
1260
0d54b217 1261static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1262 .quota_on = ext4_quota_on,
ca0e05e4 1263 .quota_off = ext4_quota_off,
287a8095
CH		 1264 .quota_sync = dquot_quota_sync,
1265 .get_info = dquot_get_dqinfo,
1266 .set_info = dquot_set_dqinfo,
1267 .get_dqblk = dquot_get_dqblk,
1268 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1269};
1270#endif
1271
ee9b6d61 1272static const struct super_operations ext4_sops = {
617ba13b
MC		 1273 .alloc_inode = ext4_alloc_inode,
1274 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1275 .write_inode = ext4_write_inode,
1276 .dirty_inode = ext4_dirty_inode,
7ff9c073 1277 .drop_inode = ext4_drop_inode,
0930fcc1 1278 .evict_inode = ext4_evict_inode,
617ba13b 1279 .put_super = ext4_put_super,
617ba13b 1280 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1281 .freeze_fs = ext4_freeze,
1282 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1283 .statfs = ext4_statfs,
1284 .remount_fs = ext4_remount,
617ba13b 1285 .show_options = ext4_show_options,
ac27a0ec 1286#ifdef CONFIG_QUOTA
617ba13b
MC		 1287 .quota_read = ext4_quota_read,
1288 .quota_write = ext4_quota_write,
ac27a0ec 1289#endif
c39a7f84 1290 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1291};
1292
9ca92389
TT		 1293static const struct super_operations ext4_nojournal_sops = {
1294 .alloc_inode = ext4_alloc_inode,
1295 .destroy_inode = ext4_destroy_inode,
1296 .write_inode = ext4_write_inode,
1297 .dirty_inode = ext4_dirty_inode,
7ff9c073 1298 .drop_inode = ext4_drop_inode,
0930fcc1 1299 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1300 .write_super = ext4_write_super,
1301 .put_super = ext4_put_super,
1302 .statfs = ext4_statfs,
1303 .remount_fs = ext4_remount,
9ca92389
TT		 1304 .show_options = ext4_show_options,
1305#ifdef CONFIG_QUOTA
1306 .quota_read = ext4_quota_read,
1307 .quota_write = ext4_quota_write,
1308#endif
1309 .bdev_try_to_free_page = bdev_try_to_free_page,
1310};
1311
39655164 1312static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1313 .fh_to_dentry = ext4_fh_to_dentry,
1314 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1315 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1316};
1317
1318enum {
1319 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1320 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1321 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1322 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1323 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1324 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1325 Opt_journal_update, Opt_journal_dev,
818d276c 1326 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1327 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1328 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1329 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1330 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1331 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1332 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1333 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1334 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1335 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1336 Opt_dioread_nolock, Opt_dioread_lock,
5328e635 1337 Opt_discard, Opt_nodiscard,
bfff6873 1338 Opt_init_inode_table, Opt_noinit_inode_table,
ac27a0ec
DK		 1339};
1340
a447c093 1341static const match_table_t tokens = {
ac27a0ec
DK		 1342 {Opt_bsd_df, "bsddf"},
1343 {Opt_minix_df, "minixdf"},
1344 {Opt_grpid, "grpid"},
1345 {Opt_grpid, "bsdgroups"},
1346 {Opt_nogrpid, "nogrpid"},
1347 {Opt_nogrpid, "sysvgroups"},
1348 {Opt_resgid, "resgid=%u"},
1349 {Opt_resuid, "resuid=%u"},
1350 {Opt_sb, "sb=%u"},
1351 {Opt_err_cont, "errors=continue"},
1352 {Opt_err_panic, "errors=panic"},
1353 {Opt_err_ro, "errors=remount-ro"},
1354 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1355 {Opt_debug, "debug"},
1356 {Opt_oldalloc, "oldalloc"},
1357 {Opt_orlov, "orlov"},
1358 {Opt_user_xattr, "user_xattr"},
1359 {Opt_nouser_xattr, "nouser_xattr"},
1360 {Opt_acl, "acl"},
1361 {Opt_noacl, "noacl"},
ac27a0ec 1362 {Opt_noload, "noload"},
e3bb52ae 1363 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1364 {Opt_nobh, "nobh"},
1365 {Opt_bh, "bh"},
1366 {Opt_commit, "commit=%u"},
30773840
TT		 1367 {Opt_min_batch_time, "min_batch_time=%u"},
1368 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1369 {Opt_journal_update, "journal=update"},
ac27a0ec 1370 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1371 {Opt_journal_checksum, "journal_checksum"},
1372 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1373 {Opt_abort, "abort"},
1374 {Opt_data_journal, "data=journal"},
1375 {Opt_data_ordered, "data=ordered"},
1376 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1377 {Opt_data_err_abort, "data_err=abort"},
1378 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1379 {Opt_offusrjquota, "usrjquota="},
1380 {Opt_usrjquota, "usrjquota=%s"},
1381 {Opt_offgrpjquota, "grpjquota="},
1382 {Opt_grpjquota, "grpjquota=%s"},
1383 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1384 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1385 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1386 {Opt_grpquota, "grpquota"},
1387 {Opt_noquota, "noquota"},
1388 {Opt_quota, "quota"},
1389 {Opt_usrquota, "usrquota"},
1390 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1391 {Opt_barrier, "barrier"},
1392 {Opt_nobarrier, "nobarrier"},
25ec56b5 1393 {Opt_i_version, "i_version"},
c9de560d 1394 {Opt_stripe, "stripe=%u"},
ac27a0ec 1395 {Opt_resize, "resize"},
64769240 1396 {Opt_delalloc, "delalloc"},
dd919b98 1397 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1398 {Opt_mblk_io_submit, "mblk_io_submit"},
1399 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1400 {Opt_block_validity, "block_validity"},
1401 {Opt_noblock_validity, "noblock_validity"},
240799cd 1402 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1403 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1404 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1405 {Opt_auto_da_alloc, "auto_da_alloc"},
1406 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1407 {Opt_dioread_nolock, "dioread_nolock"},
1408 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1409 {Opt_discard, "discard"},
1410 {Opt_nodiscard, "nodiscard"},
bfff6873
LC		 1411 {Opt_init_inode_table, "init_itable=%u"},
1412 {Opt_init_inode_table, "init_itable"},
1413 {Opt_noinit_inode_table, "noinit_itable"},
f3f12faa 1414 {Opt_err, NULL},
ac27a0ec
DK		 1415};
1416
617ba13b 1417static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1418{
617ba13b 1419 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1420 char *options = (char *) *data;
1421
1422 if (!options || strncmp(options, "sb=", 3) != 0)
1423 return 1; /* Default location */
0b8e58a1 1424
ac27a0ec 1425 options += 3;
0b8e58a1 1426 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1427 sb_block = simple_strtoul(options, &options, 0);
1428 if (*options && *options != ',') {
4776004f 1429 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1430 (char *) *data);
1431 return 1;
1432 }
1433 if (*options == ',')
1434 options++;
1435 *data = (void *) options;
0b8e58a1 1436
ac27a0ec
DK		 1437 return sb_block;
1438}
1439
b3881f74 1440#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1441static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1442 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1443
56c50f11
DM		 1444#ifdef CONFIG_QUOTA
1445static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1446{
1447 struct ext4_sb_info *sbi = EXT4_SB(sb);
1448 char *qname;
1449
1450 if (sb_any_quota_loaded(sb) &&
1451 !sbi->s_qf_names[qtype]) {
1452 ext4_msg(sb, KERN_ERR,
1453 "Cannot change journaled "
1454 "quota options when quota turned on");
1455 return 0;
1456 }
1457 qname = match_strdup(args);
1458 if (!qname) {
1459 ext4_msg(sb, KERN_ERR,
1460 "Not enough memory for storing quotafile name");
1461 return 0;
1462 }
1463 if (sbi->s_qf_names[qtype] &&
1464 strcmp(sbi->s_qf_names[qtype], qname)) {
1465 ext4_msg(sb, KERN_ERR,
1466 "%s quota file already specified", QTYPE2NAME(qtype));
1467 kfree(qname);
1468 return 0;
1469 }
1470 sbi->s_qf_names[qtype] = qname;
1471 if (strchr(sbi->s_qf_names[qtype], '/')) {
1472 ext4_msg(sb, KERN_ERR,
1473 "quotafile must be on filesystem root");
1474 kfree(sbi->s_qf_names[qtype]);
1475 sbi->s_qf_names[qtype] = NULL;
1476 return 0;
1477 }
fd8c37ec 1478 set_opt(sb, QUOTA);
56c50f11
DM		 1479 return 1;
1480}
1481
1482static int clear_qf_name(struct super_block *sb, int qtype)
1483{
1484
1485 struct ext4_sb_info *sbi = EXT4_SB(sb);
1486
1487 if (sb_any_quota_loaded(sb) &&
1488 sbi->s_qf_names[qtype]) {
1489 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1490 " when quota turned on");
1491 return 0;
1492 }
1493 /*
1494 * The space will be released later when all options are confirmed
1495 * to be correct
1496 */
1497 sbi->s_qf_names[qtype] = NULL;
1498 return 1;
1499}
1500#endif
1501
2b2d6d01 1502static int parse_options(char *options, struct super_block *sb,
c3191067 1503 unsigned long *journal_devnum,
b3881f74 1504 unsigned int *journal_ioprio,
2b2d6d01 1505 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1506{
617ba13b 1507 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1508 char *p;
ac27a0ec
DK		 1509 substring_t args[MAX_OPT_ARGS];
1510 int data_opt = 0;
1511 int option;
1512#ifdef CONFIG_QUOTA
56c50f11 1513 int qfmt;
ac27a0ec
DK		 1514#endif
1515
1516 if (!options)
1517 return 1;
1518
2b2d6d01 1519 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1520 int token;
1521 if (!*p)
1522 continue;
1523
15121c18
ES		 1524 /*
1525 * Initialize args struct so we know whether arg was
1526 * found; some options take optional arguments.
1527 */
7dc57615 1528 args[0].to = args[0].from = NULL;
ac27a0ec
DK		 1529 token = match_token(p, tokens, args);
1530 switch (token) {
1531 case Opt_bsd_df:
437ca0fd 1532 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1533 clear_opt(sb, MINIX_DF);
ac27a0ec
DK		 1534 break;
1535 case Opt_minix_df:
437ca0fd 1536 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1537 set_opt(sb, MINIX_DF);
437ca0fd 1538
ac27a0ec
DK		 1539 break;
1540 case Opt_grpid:
437ca0fd 1541 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1542 set_opt(sb, GRPID);
437ca0fd 1543
ac27a0ec
DK		 1544 break;
1545 case Opt_nogrpid:
437ca0fd 1546 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1547 clear_opt(sb, GRPID);
437ca0fd 1548
ac27a0ec
DK		 1549 break;
1550 case Opt_resuid:
1551 if (match_int(&args[0], &option))
1552 return 0;
1553 sbi->s_resuid = option;
1554 break;
1555 case Opt_resgid:
1556 if (match_int(&args[0], &option))
1557 return 0;
1558 sbi->s_resgid = option;
1559 break;
1560 case Opt_sb:
1561 /* handled by get_sb_block() instead of here */
1562 /* *sb_block = match_int(&args[0]); */
1563 break;
1564 case Opt_err_panic:
fd8c37ec
TT		 1565 clear_opt(sb, ERRORS_CONT);
1566 clear_opt(sb, ERRORS_RO);
1567 set_opt(sb, ERRORS_PANIC);
ac27a0ec
DK		 1568 break;
1569 case Opt_err_ro:
fd8c37ec
TT		 1570 clear_opt(sb, ERRORS_CONT);
1571 clear_opt(sb, ERRORS_PANIC);
1572 set_opt(sb, ERRORS_RO);
ac27a0ec
DK		 1573 break;
1574 case Opt_err_cont:
fd8c37ec
TT		 1575 clear_opt(sb, ERRORS_RO);
1576 clear_opt(sb, ERRORS_PANIC);
1577 set_opt(sb, ERRORS_CONT);
ac27a0ec
DK		 1578 break;
1579 case Opt_nouid32:
fd8c37ec 1580 set_opt(sb, NO_UID32);
ac27a0ec 1581 break;
ac27a0ec 1582 case Opt_debug:
fd8c37ec 1583 set_opt(sb, DEBUG);
ac27a0ec
DK		 1584 break;
1585 case Opt_oldalloc:
fd8c37ec 1586 set_opt(sb, OLDALLOC);
ac27a0ec
DK		 1587 break;
1588 case Opt_orlov:
fd8c37ec 1589 clear_opt(sb, OLDALLOC);
ac27a0ec 1590 break;
03010a33 1591#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1592 case Opt_user_xattr:
fd8c37ec 1593 set_opt(sb, XATTR_USER);
ac27a0ec
DK		 1594 break;
1595 case Opt_nouser_xattr:
fd8c37ec 1596 clear_opt(sb, XATTR_USER);
ac27a0ec
DK		 1597 break;
1598#else
1599 case Opt_user_xattr:
1600 case Opt_nouser_xattr:
b31e1552 1601 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1602 break;
1603#endif
03010a33 1604#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec 1605 case Opt_acl:
fd8c37ec 1606 set_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1607 break;
1608 case Opt_noacl:
fd8c37ec 1609 clear_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1610 break;
1611#else
1612 case Opt_acl:
1613 case Opt_noacl:
b31e1552 1614 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1615 break;
1616#endif
ac27a0ec
DK		 1617 case Opt_journal_update:
1618 /* @@@ FIXME */
1619 /* Eventually we will want to be able to create
1620 a journal file here. For now, only allow the
1621 user to specify an existing inode to be the
1622 journal file. */
1623 if (is_remount) {
b31e1552
ES		 1624 ext4_msg(sb, KERN_ERR,
1625 "Cannot specify journal on remount");
ac27a0ec
DK		 1626 return 0;
1627 }
fd8c37ec 1628 set_opt(sb, UPDATE_JOURNAL);
ac27a0ec 1629 break;
ac27a0ec
DK		 1630 case Opt_journal_dev:
1631 if (is_remount) {
b31e1552
ES		 1632 ext4_msg(sb, KERN_ERR,
1633 "Cannot specify journal on remount");
ac27a0ec
DK		 1634 return 0;
1635 }
1636 if (match_int(&args[0], &option))
1637 return 0;
1638 *journal_devnum = option;
1639 break;
818d276c 1640 case Opt_journal_checksum:
fd8c37ec 1641 set_opt(sb, JOURNAL_CHECKSUM);
d4da6c9c 1642 break;
818d276c 1643 case Opt_journal_async_commit:
fd8c37ec
TT		 1644 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1645 set_opt(sb, JOURNAL_CHECKSUM);
818d276c 1646 break;
ac27a0ec 1647 case Opt_noload:
fd8c37ec 1648 set_opt(sb, NOLOAD);
ac27a0ec
DK		 1649 break;
1650 case Opt_commit:
1651 if (match_int(&args[0], &option))
1652 return 0;
1653 if (option < 0)
1654 return 0;
1655 if (option == 0)
cd02ff0b 1656 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1657 sbi->s_commit_interval = HZ * option;
1658 break;
30773840
TT		 1659 case Opt_max_batch_time:
1660 if (match_int(&args[0], &option))
1661 return 0;
1662 if (option < 0)
1663 return 0;
1664 if (option == 0)
1665 option = EXT4_DEF_MAX_BATCH_TIME;
1666 sbi->s_max_batch_time = option;
1667 break;
1668 case Opt_min_batch_time:
1669 if (match_int(&args[0], &option))
1670 return 0;
1671 if (option < 0)
1672 return 0;
1673 sbi->s_min_batch_time = option;
1674 break;
ac27a0ec 1675 case Opt_data_journal:
617ba13b 1676 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1677 goto datacheck;
1678 case Opt_data_ordered:
617ba13b 1679 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1680 goto datacheck;
1681 case Opt_data_writeback:
617ba13b 1682 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1683 datacheck:
1684 if (is_remount) {
482a7425 1685 if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1686 ext4_msg(sb, KERN_ERR,
1687 "Cannot change data mode on remount");
ac27a0ec
DK		 1688 return 0;
1689 }
1690 } else {
fd8c37ec 1691 clear_opt(sb, DATA_FLAGS);
ac27a0ec
DK		 1692 sbi->s_mount_opt |= data_opt;
1693 }
1694 break;
5bf5683a 1695 case Opt_data_err_abort:
fd8c37ec 1696 set_opt(sb, DATA_ERR_ABORT);
5bf5683a
HK		 1697 break;
1698 case Opt_data_err_ignore:
fd8c37ec 1699 clear_opt(sb, DATA_ERR_ABORT);
5bf5683a 1700 break;
ac27a0ec
DK		 1701#ifdef CONFIG_QUOTA
1702 case Opt_usrjquota:
56c50f11 1703 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1704 return 0;
56c50f11
DM		 1705 break;
1706 case Opt_grpjquota:
1707 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1708 return 0;
ac27a0ec
DK		 1709 break;
1710 case Opt_offusrjquota:
56c50f11
DM		 1711 if (!clear_qf_name(sb, USRQUOTA))
1712 return 0;
1713 break;
ac27a0ec 1714 case Opt_offgrpjquota:
56c50f11 1715 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1716 return 0;
ac27a0ec 1717 break;
56c50f11 1718
ac27a0ec 1719 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1720 qfmt = QFMT_VFS_OLD;
1721 goto set_qf_format;
ac27a0ec 1722 case Opt_jqfmt_vfsv0:
dfc5d03f 1723 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1724 goto set_qf_format;
1725 case Opt_jqfmt_vfsv1:
1726 qfmt = QFMT_VFS_V1;
dfc5d03f 1727set_qf_format:
17bd13b3 1728 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1729 sbi->s_jquota_fmt != qfmt) {
b31e1552 1730 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1731 "journaled quota options when "
b31e1552 1732 "quota turned on");
dfc5d03f
JK		 1733 return 0;
1734 }
1735 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1736 break;
1737 case Opt_quota:
1738 case Opt_usrquota:
fd8c37ec
TT		 1739 set_opt(sb, QUOTA);
1740 set_opt(sb, USRQUOTA);
ac27a0ec
DK		 1741 break;
1742 case Opt_grpquota:
fd8c37ec
TT		 1743 set_opt(sb, QUOTA);
1744 set_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1745 break;
1746 case Opt_noquota:
17bd13b3 1747 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1748 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1749 "options when quota turned on");
ac27a0ec
DK		 1750 return 0;
1751 }
fd8c37ec
TT		 1752 clear_opt(sb, QUOTA);
1753 clear_opt(sb, USRQUOTA);
1754 clear_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1755 break;
1756#else
1757 case Opt_quota:
1758 case Opt_usrquota:
1759 case Opt_grpquota:
b31e1552
ES		 1760 ext4_msg(sb, KERN_ERR,
1761 "quota options not supported");
cd59e7b9 1762 break;
ac27a0ec
DK		 1763 case Opt_usrjquota:
1764 case Opt_grpjquota:
1765 case Opt_offusrjquota:
1766 case Opt_offgrpjquota:
1767 case Opt_jqfmt_vfsold:
1768 case Opt_jqfmt_vfsv0:
5a20bdfc 1769 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1770 ext4_msg(sb, KERN_ERR,
1771 "journaled quota options not supported");
ac27a0ec
DK		 1772 break;
1773 case Opt_noquota:
1774 break;
1775#endif
1776 case Opt_abort:
4ab2f15b 1777 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1778 break;
06705bff 1779 case Opt_nobarrier:
fd8c37ec 1780 clear_opt(sb, BARRIER);
06705bff 1781 break;
ac27a0ec 1782 case Opt_barrier:
15121c18
ES		 1783 if (args[0].from) {
1784 if (match_int(&args[0], &option))
1785 return 0;
1786 } else
1787 option = 1; /* No argument, default to 1 */
ac27a0ec 1788 if (option)
fd8c37ec 1789 set_opt(sb, BARRIER);
ac27a0ec 1790 else
fd8c37ec 1791 clear_opt(sb, BARRIER);
ac27a0ec
DK		 1792 break;
1793 case Opt_ignore:
1794 break;
1795 case Opt_resize:
1796 if (!is_remount) {
b31e1552
ES		 1797 ext4_msg(sb, KERN_ERR,
1798 "resize option only available "
1799 "for remount");
ac27a0ec
DK		 1800 return 0;
1801 }
1802 if (match_int(&args[0], &option) != 0)
1803 return 0;
1804 *n_blocks_count = option;
1805 break;
1806 case Opt_nobh:
206f7ab4
CH		 1807 ext4_msg(sb, KERN_WARNING,
1808 "Ignoring deprecated nobh option");
ac27a0ec
DK		 1809 break;
1810 case Opt_bh:
206f7ab4
CH		 1811 ext4_msg(sb, KERN_WARNING,
1812 "Ignoring deprecated bh option");
ac27a0ec 1813 break;
25ec56b5 1814 case Opt_i_version:
fd8c37ec 1815 set_opt(sb, I_VERSION);
25ec56b5
JNC		 1816 sb->s_flags |= MS_I_VERSION;
1817 break;
dd919b98 1818 case Opt_nodelalloc:
fd8c37ec 1819 clear_opt(sb, DELALLOC);
56889787 1820 clear_opt2(sb, EXPLICIT_DELALLOC);
dd919b98 1821 break;
1449032b 1822 case Opt_mblk_io_submit:
fd8c37ec 1823 set_opt(sb, MBLK_IO_SUBMIT);
1449032b
TT		 1824 break;
1825 case Opt_nomblk_io_submit:
fd8c37ec 1826 clear_opt(sb, MBLK_IO_SUBMIT);
1449032b 1827 break;
c9de560d
AT		 1828 case Opt_stripe:
1829 if (match_int(&args[0], &option))
1830 return 0;
1831 if (option < 0)
1832 return 0;
1833 sbi->s_stripe = option;
1834 break;
64769240 1835 case Opt_delalloc:
fd8c37ec 1836 set_opt(sb, DELALLOC);
56889787 1837 set_opt2(sb, EXPLICIT_DELALLOC);
64769240 1838 break;
6fd058f7 1839 case Opt_block_validity:
fd8c37ec 1840 set_opt(sb, BLOCK_VALIDITY);
6fd058f7
TT		 1841 break;
1842 case Opt_noblock_validity:
fd8c37ec 1843 clear_opt(sb, BLOCK_VALIDITY);
6fd058f7 1844 break;
240799cd
TT		 1845 case Opt_inode_readahead_blks:
1846 if (match_int(&args[0], &option))
1847 return 0;
1848 if (option < 0 || option > (1 << 30))
1849 return 0;
5dbd571d 1850 if (option && !is_power_of_2(option)) {
b31e1552
ES		 1851 ext4_msg(sb, KERN_ERR,
1852 "EXT4-fs: inode_readahead_blks"
1853 " must be a power of 2");
3197ebdb
TT		 1854 return 0;
1855 }
240799cd
TT		 1856 sbi->s_inode_readahead_blks = option;
1857 break;
b3881f74
TT		 1858 case Opt_journal_ioprio:
1859 if (match_int(&args[0], &option))
1860 return 0;
1861 if (option < 0 || option > 7)
1862 break;
1863 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1864 option);
1865 break;
06705bff 1866 case Opt_noauto_da_alloc:
fd8c37ec 1867 set_opt(sb, NO_AUTO_DA_ALLOC);
06705bff 1868 break;
afd4672d 1869 case Opt_auto_da_alloc:
15121c18
ES		 1870 if (args[0].from) {
1871 if (match_int(&args[0], &option))
1872 return 0;
1873 } else
1874 option = 1; /* No argument, default to 1 */
afd4672d 1875 if (option)
fd8c37ec 1876 clear_opt(sb, NO_AUTO_DA_ALLOC);
afd4672d 1877 else
fd8c37ec 1878 set_opt(sb,NO_AUTO_DA_ALLOC);
afd4672d 1879 break;
5328e635 1880 case Opt_discard:
fd8c37ec 1881 set_opt(sb, DISCARD);
5328e635
ES		 1882 break;
1883 case Opt_nodiscard:
fd8c37ec 1884 clear_opt(sb, DISCARD);
5328e635 1885 break;
744692dc 1886 case Opt_dioread_nolock:
fd8c37ec 1887 set_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 1888 break;
1889 case Opt_dioread_lock:
fd8c37ec 1890 clear_opt(sb, DIOREAD_NOLOCK);
744692dc 1891 break;
bfff6873 1892 case Opt_init_inode_table:
fd8c37ec 1893 set_opt(sb, INIT_INODE_TABLE);
bfff6873
LC		 1894 if (args[0].from) {
1895 if (match_int(&args[0], &option))
1896 return 0;
1897 } else
1898 option = EXT4_DEF_LI_WAIT_MULT;
1899 if (option < 0)
1900 return 0;
1901 sbi->s_li_wait_mult = option;
1902 break;
1903 case Opt_noinit_inode_table:
fd8c37ec 1904 clear_opt(sb, INIT_INODE_TABLE);
bfff6873 1905 break;
ac27a0ec 1906 default:
b31e1552
ES		 1907 ext4_msg(sb, KERN_ERR,
1908 "Unrecognized mount option \"%s\" "
1909 "or missing value", p);
ac27a0ec
DK		 1910 return 0;
1911 }
1912 }
1913#ifdef CONFIG_QUOTA
1914 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1915 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1916 clear_opt(sb, USRQUOTA);
ac27a0ec 1917
482a7425 1918 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1919 clear_opt(sb, GRPQUOTA);
ac27a0ec 1920
56c50f11 1921 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1922 ext4_msg(sb, KERN_ERR, "old and new quota "
1923 "format mixing");
ac27a0ec
DK		 1924 return 0;
1925 }
1926
1927 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1928 ext4_msg(sb, KERN_ERR, "journaled quota format "
1929 "not specified");
ac27a0ec
DK		 1930 return 0;
1931 }
1932 } else {
1933 if (sbi->s_jquota_fmt) {
b31e1552 1934 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1935 "specified with no journaling "
b31e1552 1936 "enabled");
ac27a0ec
DK		 1937 return 0;
1938 }
1939 }
1940#endif
1941 return 1;
1942}
1943
617ba13b 1944static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1945 int read_only)
1946{
617ba13b 1947 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1948 int res = 0;
1949
617ba13b 1950 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1951 ext4_msg(sb, KERN_ERR, "revision level too high, "
1952 "forcing read-only mode");
ac27a0ec
DK		 1953 res = MS_RDONLY;
1954 }
1955 if (read_only)
281b5995 1956 goto done;
617ba13b 1957 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1958 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1959 "running e2fsck is recommended");
617ba13b 1960 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1961 ext4_msg(sb, KERN_WARNING,
1962 "warning: mounting fs with errors, "
1963 "running e2fsck is recommended");
ed3ce80a 1964 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1965 le16_to_cpu(es->s_mnt_count) >=
1966 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1967 ext4_msg(sb, KERN_WARNING,
1968 "warning: maximal mount count reached, "
1969 "running e2fsck is recommended");
ac27a0ec
DK		 1970 else if (le32_to_cpu(es->s_checkinterval) &&
1971 (le32_to_cpu(es->s_lastcheck) +
1972 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1973 ext4_msg(sb, KERN_WARNING,
1974 "warning: checktime reached, "
1975 "running e2fsck is recommended");
0b8e58a1 1976 if (!sbi->s_journal)
0390131b 1977 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1978 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1979 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1980 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1981 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1982 ext4_update_dynamic_rev(sb);
0390131b
FM		 1983 if (sbi->s_journal)
1984 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1985
e2d67052 1986 ext4_commit_super(sb, 1);
281b5995 1987done:
ac27a0ec 1988 if (test_opt(sb, DEBUG))
a9df9a49 1989 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1990 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1991 sb->s_blocksize,
1992 sbi->s_groups_count,
617ba13b
MC		 1993 EXT4_BLOCKS_PER_GROUP(sb),
1994 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1995 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1996
7abc52c2 1997 cleancache_init_fs(sb);
ac27a0ec
DK		 1998 return res;
1999}
2000
772cb7c8
JS		 2001static int ext4_fill_flex_info(struct super_block *sb)
2002{
2003 struct ext4_sb_info *sbi = EXT4_SB(sb);
2004 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 2005 ext4_group_t flex_group_count;
2006 ext4_group_t flex_group;
2007 int groups_per_flex = 0;
c5ca7c76 2008 size_t size;
772cb7c8
JS		 2009 int i;
2010
503358ae
TT		 2011 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2012 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
2013
2014 if (groups_per_flex < 2) {
772cb7c8
JS		 2015 sbi->s_log_groups_per_flex = 0;
2016 return 1;
2017 }
2018
c62a11fd
FB		 2019 /* We allocate both existing and potentially added groups */
2020 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 2021 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2022 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 2023 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 2024 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 2025 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 2026 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2027 flex_group_count);
2028 goto failed;
772cb7c8 2029 }
772cb7c8 2030
772cb7c8 2031 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2032 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2033
2034 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2035 atomic_add(ext4_free_inodes_count(sb, gdp),
2036 &sbi->s_flex_groups[flex_group].free_inodes);
2037 atomic_add(ext4_free_blks_count(sb, gdp),
2038 &sbi->s_flex_groups[flex_group].free_blocks);
2039 atomic_add(ext4_used_dirs_count(sb, gdp),
2040 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2041 }
2042
2043 return 1;
2044failed:
2045 return 0;
2046}
2047
717d50e4
AD		 2048__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2049 struct ext4_group_desc *gdp)
2050{
2051 __u16 crc = 0;
2052
2053 if (sbi->s_es->s_feature_ro_compat &
2054 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2055 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2056 __le32 le_group = cpu_to_le32(block_group);
2057
2058 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2059 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2060 crc = crc16(crc, (__u8 *)gdp, offset);
2061 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2062 /* for checksum of struct ext4_group_desc do the rest...*/
2063 if ((sbi->s_es->s_feature_incompat &
2064 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2065 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2066 crc = crc16(crc, (__u8 *)gdp + offset,
2067 le16_to_cpu(sbi->s_es->s_desc_size) -
2068 offset);
2069 }
2070
2071 return cpu_to_le16(crc);
2072}
2073
2074int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2075 struct ext4_group_desc *gdp)
2076{
2077 if ((sbi->s_es->s_feature_ro_compat &
2078 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2079 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2080 return 0;
2081
2082 return 1;
2083}
2084
ac27a0ec 2085/* Called at mount-time, super-block is locked */
bfff6873
LC		 2086static int ext4_check_descriptors(struct super_block *sb,
2087 ext4_group_t *first_not_zeroed)
ac27a0ec 2088{
617ba13b
MC		 2089 struct ext4_sb_info *sbi = EXT4_SB(sb);
2090 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2091 ext4_fsblk_t last_block;
bd81d8ee
LV		 2092 ext4_fsblk_t block_bitmap;
2093 ext4_fsblk_t inode_bitmap;
2094 ext4_fsblk_t inode_table;
ce421581 2095 int flexbg_flag = 0;
bfff6873 2096 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2097
ce421581
JS		 2098 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2099 flexbg_flag = 1;
2100
af5bc92d 2101 ext4_debug("Checking group descriptors");
ac27a0ec 2102
197cd65a
AM		 2103 for (i = 0; i < sbi->s_groups_count; i++) {
2104 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2105
ce421581 2106 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2107 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2108 else
2109 last_block = first_block +
617ba13b 2110 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2111
bfff6873
LC		 2112 if ((grp == sbi->s_groups_count) &&
2113 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2114 grp = i;
2115
8fadc143 2116 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2117 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2118 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2119 "Block bitmap for group %u not in group "
b31e1552 2120 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2121 return 0;
2122 }
8fadc143 2123 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2124 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2125 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2126 "Inode bitmap for group %u not in group "
b31e1552 2127 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2128 return 0;
2129 }
8fadc143 2130 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2131 if (inode_table < first_block ||
2b2d6d01 2132 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2133 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2134 "Inode table for group %u not in group "
b31e1552 2135 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2136 return 0;
2137 }
955ce5f5 2138 ext4_lock_group(sb, i);
717d50e4 2139 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2140 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2141 "Checksum for group %u failed (%u!=%u)",
2142 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2143 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2144 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2145 ext4_unlock_group(sb, i);
8a266467 2146 return 0;
7ee1ec4c 2147 }
717d50e4 2148 }
955ce5f5 2149 ext4_unlock_group(sb, i);
ce421581
JS		 2150 if (!flexbg_flag)
2151 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2152 }
bfff6873
LC		 2153 if (NULL != first_not_zeroed)
2154 *first_not_zeroed = grp;
ac27a0ec 2155
bd81d8ee 2156 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 2157 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2158 return 1;
2159}
2160
617ba13b 2161/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2162 * the superblock) which were deleted from all directories, but held open by
2163 * a process at the time of a crash. We walk the list and try to delete these
2164 * inodes at recovery time (only with a read-write filesystem).
2165 *
2166 * In order to keep the orphan inode chain consistent during traversal (in
2167 * case of crash during recovery), we link each inode into the superblock
2168 * orphan list_head and handle it the same way as an inode deletion during
2169 * normal operation (which journals the operations for us).
2170 *
2171 * We only do an iget() and an iput() on each inode, which is very safe if we
2172 * accidentally point at an in-use or already deleted inode. The worst that
2173 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2174 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2175 * e2fsck was run on this filesystem, and it must have already done the orphan
2176 * inode cleanup for us, so we can safely abort without any further action.
2177 */
2b2d6d01
TT		 2178static void ext4_orphan_cleanup(struct super_block *sb,
2179 struct ext4_super_block *es)
ac27a0ec
DK		 2180{
2181 unsigned int s_flags = sb->s_flags;
2182 int nr_orphans = 0, nr_truncates = 0;
2183#ifdef CONFIG_QUOTA
2184 int i;
2185#endif
2186 if (!es->s_last_orphan) {
2187 jbd_debug(4, "no orphan inodes to clean up\n");
2188 return;
2189 }
2190
a8f48a95 2191 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2192 ext4_msg(sb, KERN_ERR, "write access "
2193 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2194 return;
2195 }
2196
d39195c3
AG		 2197 /* Check if feature set would not allow a r/w mount */
2198 if (!ext4_feature_set_ok(sb, 0)) {
2199 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2200 "unknown ROCOMPAT features");
2201 return;
2202 }
2203
617ba13b 2204 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2205 if (es->s_last_orphan)
2206 jbd_debug(1, "Errors on filesystem, "
2207 "clearing orphan list.\n");
2208 es->s_last_orphan = 0;
2209 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2210 return;
2211 }
2212
2213 if (s_flags & MS_RDONLY) {
b31e1552 2214 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2215 sb->s_flags &= ~MS_RDONLY;
2216 }
2217#ifdef CONFIG_QUOTA
2218 /* Needed for iput() to work correctly and not trash data */
2219 sb->s_flags |= MS_ACTIVE;
2220 /* Turn on quotas so that they are updated correctly */
2221 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2222 if (EXT4_SB(sb)->s_qf_names[i]) {
2223 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2224 if (ret < 0)
b31e1552
ES		 2225 ext4_msg(sb, KERN_ERR,
2226 "Cannot turn on journaled "
2227 "quota: error %d", ret);
ac27a0ec
DK		 2228 }
2229 }
2230#endif
2231
2232 while (es->s_last_orphan) {
2233 struct inode *inode;
2234
97bd42b9
JB		 2235 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2236 if (IS_ERR(inode)) {
ac27a0ec
DK		 2237 es->s_last_orphan = 0;
2238 break;
2239 }
2240
617ba13b 2241 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2242 dquot_initialize(inode);
ac27a0ec 2243 if (inode->i_nlink) {
b31e1552
ES		 2244 ext4_msg(sb, KERN_DEBUG,
2245 "%s: truncating inode %lu to %lld bytes",
46e665e9 2246 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2247 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2248 inode->i_ino, inode->i_size);
617ba13b 2249 ext4_truncate(inode);
ac27a0ec
DK		 2250 nr_truncates++;
2251 } else {
b31e1552
ES		 2252 ext4_msg(sb, KERN_DEBUG,
2253 "%s: deleting unreferenced inode %lu",
46e665e9 2254 __func__, inode->i_ino);
ac27a0ec
DK		 2255 jbd_debug(2, "deleting unreferenced inode %lu\n",
2256 inode->i_ino);
2257 nr_orphans++;
2258 }
2259 iput(inode); /* The delete magic happens here! */
2260 }
2261
2b2d6d01 2262#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2263
2264 if (nr_orphans)
b31e1552
ES		 2265 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2266 PLURAL(nr_orphans));
ac27a0ec 2267 if (nr_truncates)
b31e1552
ES		 2268 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2269 PLURAL(nr_truncates));
ac27a0ec
DK		 2270#ifdef CONFIG_QUOTA
2271 /* Turn quotas off */
2272 for (i = 0; i < MAXQUOTAS; i++) {
2273 if (sb_dqopt(sb)->files[i])
287a8095 2274 dquot_quota_off(sb, i);
ac27a0ec
DK		 2275 }
2276#endif
2277 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2278}
0b8e58a1 2279
cd2291a4
ES		 2280/*
2281 * Maximal extent format file size.
2282 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2283 * extent format containers, within a sector_t, and within i_blocks
2284 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2285 * so that won't be a limiting factor.
2286 *
f17722f9
LC		 2287 * However there is other limiting factor. We do store extents in the form
2288 * of starting block and length, hence the resulting length of the extent
2289 * covering maximum file size must fit into on-disk format containers as
2290 * well. Given that length is always by 1 unit bigger than max unit (because
2291 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2292 *
cd2291a4
ES		 2293 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2294 */
f287a1a5 2295static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2296{
2297 loff_t res;
2298 loff_t upper_limit = MAX_LFS_FILESIZE;
2299
2300 /* small i_blocks in vfs inode? */
f287a1a5 2301 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2302 /*
90c699a9 2303 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2304 * i_block represent total blocks in 512 bytes
2305 * 32 == size of vfs inode i_blocks * 8
2306 */
2307 upper_limit = (1LL << 32) - 1;
2308
2309 /* total blocks in file system block size */
2310 upper_limit >>= (blkbits - 9);
2311 upper_limit <<= blkbits;
2312 }
2313
f17722f9
LC		 2314 /*
2315 * 32-bit extent-start container, ee_block. We lower the maxbytes
2316 * by one fs block, so ee_len can cover the extent of maximum file
2317 * size
2318 */
2319 res = (1LL << 32) - 1;
cd2291a4 2320 res <<= blkbits;
cd2291a4
ES		 2321
2322 /* Sanity check against vm- & vfs- imposed limits */
2323 if (res > upper_limit)
2324 res = upper_limit;
2325
2326 return res;
2327}
ac27a0ec 2328
ac27a0ec 2329/*
cd2291a4 2330 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2331 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2332 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2333 */
f287a1a5 2334static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2335{
617ba13b 2336 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2337 int meta_blocks;
2338 loff_t upper_limit;
0b8e58a1
AD		 2339 /* This is calculated to be the largest file size for a dense, block
2340 * mapped file such that the file's total number of 512-byte sectors,
2341 * including data and all indirect blocks, does not exceed (2^48 - 1).
2342 *
2343 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2344 * number of 512-byte sectors of the file.
0fc1b451
AK		 2345 */
2346
f287a1a5 2347 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2348 /*
90c699a9 2349 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2350 * the inode i_block field represents total file blocks in
2351 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2352 */
2353 upper_limit = (1LL << 32) - 1;
2354
2355 /* total blocks in file system block size */
2356 upper_limit >>= (bits - 9);
2357
2358 } else {
8180a562
AK		 2359 /*
2360 * We use 48 bit ext4_inode i_blocks
2361 * With EXT4_HUGE_FILE_FL set the i_blocks
2362 * represent total number of blocks in
2363 * file system block size
2364 */
0fc1b451
AK		 2365 upper_limit = (1LL << 48) - 1;
2366
0fc1b451
AK		 2367 }
2368
2369 /* indirect blocks */
2370 meta_blocks = 1;
2371 /* double indirect blocks */
2372 meta_blocks += 1 + (1LL << (bits-2));
2373 /* tripple indirect blocks */
2374 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2375
2376 upper_limit -= meta_blocks;
2377 upper_limit <<= bits;
ac27a0ec
DK		 2378
2379 res += 1LL << (bits-2);
2380 res += 1LL << (2*(bits-2));
2381 res += 1LL << (3*(bits-2));
2382 res <<= bits;
2383 if (res > upper_limit)
2384 res = upper_limit;
0fc1b451
AK		 2385
2386 if (res > MAX_LFS_FILESIZE)
2387 res = MAX_LFS_FILESIZE;
2388
ac27a0ec
DK		 2389 return res;
2390}
2391
617ba13b 2392static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2393 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2394{
617ba13b 2395 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2396 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2397 int has_super = 0;
2398
2399 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2400
617ba13b 2401 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2402 nr < first_meta_bg)
70bbb3e0 2403 return logical_sb_block + nr + 1;
ac27a0ec 2404 bg = sbi->s_desc_per_block * nr;
617ba13b 2405 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2406 has_super = 1;
0b8e58a1 2407
617ba13b 2408 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2409}
2410
c9de560d
AT		 2411/**
2412 * ext4_get_stripe_size: Get the stripe size.
2413 * @sbi: In memory super block info
2414 *
2415 * If we have specified it via mount option, then
2416 * use the mount option value. If the value specified at mount time is
2417 * greater than the blocks per group use the super block value.
2418 * If the super block value is greater than blocks per group return 0.
2419 * Allocator needs it be less than blocks per group.
2420 *
2421 */
2422static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2423{
2424 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2425 unsigned long stripe_width =
2426 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2427 int ret;
c9de560d
AT		 2428
2429 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2430 ret = sbi->s_stripe;
2431 else if (stripe_width <= sbi->s_blocks_per_group)
2432 ret = stripe_width;
2433 else if (stride <= sbi->s_blocks_per_group)
2434 ret = stride;
2435 else
2436 ret = 0;
c9de560d 2437
3eb08658
DE		 2438 /*
2439 * If the stripe width is 1, this makes no sense and
2440 * we set it to 0 to turn off stripe handling code.
2441 */
2442 if (ret <= 1)
2443 ret = 0;
c9de560d 2444
3eb08658 2445 return ret;
c9de560d 2446}
ac27a0ec 2447
3197ebdb
TT		 2448/* sysfs supprt */
2449
2450struct ext4_attr {
2451 struct attribute attr;
2452 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2453 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2454 const char *, size_t);
2455 int offset;
2456};
2457
2458static int parse_strtoul(const char *buf,
2459 unsigned long max, unsigned long *value)
2460{
2461 char *endp;
2462
e7d2860b
AGR		 2463 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2464 endp = skip_spaces(endp);
3197ebdb
TT		 2465 if (*endp || *value > max)
2466 return -EINVAL;
2467
2468 return 0;
2469}
2470
2471static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2472 struct ext4_sb_info *sbi,
2473 char *buf)
2474{
2475 return snprintf(buf, PAGE_SIZE, "%llu\n",
2476 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2477}
2478
2479static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2480 struct ext4_sb_info *sbi, char *buf)
2481{
2482 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2483
f613dfcb
TT		 2484 if (!sb->s_bdev->bd_part)
2485 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2486 return snprintf(buf, PAGE_SIZE, "%lu\n",
2487 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2488 sbi->s_sectors_written_start) >> 1);
2489}
2490
2491static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2492 struct ext4_sb_info *sbi, char *buf)
2493{
2494 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2495
f613dfcb
TT		 2496 if (!sb->s_bdev->bd_part)
2497 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2498 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2499 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2500 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2501 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2502}
2503
77f4135f
VH		 2504static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2505 struct ext4_sb_info *sbi, char *buf)
2506{
2507 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2508}
2509
2510static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2511 struct ext4_sb_info *sbi, char *buf)
2512{
2513 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2514}
2515
3197ebdb
TT		 2516static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2517 struct ext4_sb_info *sbi,
2518 const char *buf, size_t count)
2519{
2520 unsigned long t;
2521
2522 if (parse_strtoul(buf, 0x40000000, &t))
2523 return -EINVAL;
2524
5dbd571d 2525 if (t && !is_power_of_2(t))
3197ebdb
TT		 2526 return -EINVAL;
2527
2528 sbi->s_inode_readahead_blks = t;
2529 return count;
2530}
2531
2532static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2533 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2534{
2535 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2536
2537 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2538}
2539
2540static ssize_t sbi_ui_store(struct ext4_attr *a,
2541 struct ext4_sb_info *sbi,
2542 const char *buf, size_t count)
2543{
2544 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2545 unsigned long t;
2546
2547 if (parse_strtoul(buf, 0xffffffff, &t))
2548 return -EINVAL;
2549 *ui = t;
2550 return count;
2551}
2552
2553#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2554static struct ext4_attr ext4_attr_##_name = { \
2555 .attr = {.name = __stringify(_name), .mode = _mode }, \
2556 .show = _show, \
2557 .store = _store, \
2558 .offset = offsetof(struct ext4_sb_info, _elname), \
2559}
2560#define EXT4_ATTR(name, mode, show, store) \
2561static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2562
857ac889 2563#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2564#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2565#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2566#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2567 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2568#define ATTR_LIST(name) &ext4_attr_##name.attr
2569
2570EXT4_RO_ATTR(delayed_allocation_blocks);
2571EXT4_RO_ATTR(session_write_kbytes);
2572EXT4_RO_ATTR(lifetime_write_kbytes);
77f4135f
VH		 2573EXT4_RO_ATTR(extent_cache_hits);
2574EXT4_RO_ATTR(extent_cache_misses);
3197ebdb
TT		 2575EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2576 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2577EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2578EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2579EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2580EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2581EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2582EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2583EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2584EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2585
2586static struct attribute *ext4_attrs[] = {
2587 ATTR_LIST(delayed_allocation_blocks),
2588 ATTR_LIST(session_write_kbytes),
2589 ATTR_LIST(lifetime_write_kbytes),
77f4135f
VH		 2590 ATTR_LIST(extent_cache_hits),
2591 ATTR_LIST(extent_cache_misses),
3197ebdb 2592 ATTR_LIST(inode_readahead_blks),
11013911 2593 ATTR_LIST(inode_goal),
3197ebdb
TT		 2594 ATTR_LIST(mb_stats),
2595 ATTR_LIST(mb_max_to_scan),
2596 ATTR_LIST(mb_min_to_scan),
2597 ATTR_LIST(mb_order2_req),
2598 ATTR_LIST(mb_stream_req),
2599 ATTR_LIST(mb_group_prealloc),
55138e0b 2600 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2601 NULL,
2602};
2603
857ac889
LC		 2604/* Features this copy of ext4 supports */
2605EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2606EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2607
2608static struct attribute *ext4_feat_attrs[] = {
2609 ATTR_LIST(lazy_itable_init),
27ee40df 2610 ATTR_LIST(batched_discard),
857ac889
LC		 2611 NULL,
2612};
2613
3197ebdb
TT		 2614static ssize_t ext4_attr_show(struct kobject *kobj,
2615 struct attribute *attr, char *buf)
2616{
2617 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2618 s_kobj);
2619 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2620
2621 return a->show ? a->show(a, sbi, buf) : 0;
2622}
2623
2624static ssize_t ext4_attr_store(struct kobject *kobj,
2625 struct attribute *attr,
2626 const char *buf, size_t len)
2627{
2628 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2629 s_kobj);
2630 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2631
2632 return a->store ? a->store(a, sbi, buf, len) : 0;
2633}
2634
2635static void ext4_sb_release(struct kobject *kobj)
2636{
2637 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2638 s_kobj);
2639 complete(&sbi->s_kobj_unregister);
2640}
2641
52cf25d0 2642static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2643 .show = ext4_attr_show,
2644 .store = ext4_attr_store,
2645};
2646
2647static struct kobj_type ext4_ktype = {
2648 .default_attrs = ext4_attrs,
2649 .sysfs_ops = &ext4_attr_ops,
2650 .release = ext4_sb_release,
2651};
2652
857ac889
LC		 2653static void ext4_feat_release(struct kobject *kobj)
2654{
2655 complete(&ext4_feat->f_kobj_unregister);
2656}
2657
2658static struct kobj_type ext4_feat_ktype = {
2659 .default_attrs = ext4_feat_attrs,
2660 .sysfs_ops = &ext4_attr_ops,
2661 .release = ext4_feat_release,
2662};
2663
a13fb1a4
ES		 2664/*
2665 * Check whether this filesystem can be mounted based on
2666 * the features present and the RDONLY/RDWR mount requested.
2667 * Returns 1 if this filesystem can be mounted as requested,
2668 * 0 if it cannot be.
2669 */
2670static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2671{
2672 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2673 ext4_msg(sb, KERN_ERR,
2674 "Couldn't mount because of "
2675 "unsupported optional features (%x)",
2676 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2677 ~EXT4_FEATURE_INCOMPAT_SUPP));
2678 return 0;
2679 }
2680
2681 if (readonly)
2682 return 1;
2683
2684 /* Check that feature set is OK for a read-write mount */
2685 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2686 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2687 "unsupported optional features (%x)",
2688 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2689 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2690 return 0;
2691 }
2692 /*
2693 * Large file size enabled file system can only be mounted
2694 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2695 */
2696 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2697 if (sizeof(blkcnt_t) < sizeof(u64)) {
2698 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2699 "cannot be mounted RDWR without "
2700 "CONFIG_LBDAF");
2701 return 0;
2702 }
2703 }
2704 return 1;
2705}
2706
66e61a9e
TT		 2707/*
2708 * This function is called once a day if we have errors logged
2709 * on the file system
2710 */
2711static void print_daily_error_info(unsigned long arg)
2712{
2713 struct super_block *sb = (struct super_block *) arg;
2714 struct ext4_sb_info *sbi;
2715 struct ext4_super_block *es;
2716
2717 sbi = EXT4_SB(sb);
2718 es = sbi->s_es;
2719
2720 if (es->s_error_count)
2721 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2722 le32_to_cpu(es->s_error_count));
2723 if (es->s_first_error_time) {
2724 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2725 sb->s_id, le32_to_cpu(es->s_first_error_time),
2726 (int) sizeof(es->s_first_error_func),
2727 es->s_first_error_func,
2728 le32_to_cpu(es->s_first_error_line));
2729 if (es->s_first_error_ino)
2730 printk(": inode %u",
2731 le32_to_cpu(es->s_first_error_ino));
2732 if (es->s_first_error_block)
2733 printk(": block %llu", (unsigned long long)
2734 le64_to_cpu(es->s_first_error_block));
2735 printk("\n");
2736 }
2737 if (es->s_last_error_time) {
2738 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2739 sb->s_id, le32_to_cpu(es->s_last_error_time),
2740 (int) sizeof(es->s_last_error_func),
2741 es->s_last_error_func,
2742 le32_to_cpu(es->s_last_error_line));
2743 if (es->s_last_error_ino)
2744 printk(": inode %u",
2745 le32_to_cpu(es->s_last_error_ino));
2746 if (es->s_last_error_block)
2747 printk(": block %llu", (unsigned long long)
2748 le64_to_cpu(es->s_last_error_block));
2749 printk("\n");
2750 }
2751 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2752}
2753
bfff6873
LC		 2754/* Find next suitable group and run ext4_init_inode_table */
2755static int ext4_run_li_request(struct ext4_li_request *elr)
2756{
2757 struct ext4_group_desc *gdp = NULL;
2758 ext4_group_t group, ngroups;
2759 struct super_block *sb;
2760 unsigned long timeout = 0;
2761 int ret = 0;
2762
2763 sb = elr->lr_super;
2764 ngroups = EXT4_SB(sb)->s_groups_count;
2765
2766 for (group = elr->lr_next_group; group < ngroups; group++) {
2767 gdp = ext4_get_group_desc(sb, group, NULL);
2768 if (!gdp) {
2769 ret = 1;
2770 break;
2771 }
2772
2773 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2774 break;
2775 }
2776
2777 if (group == ngroups)
2778 ret = 1;
2779
2780 if (!ret) {
2781 timeout = jiffies;
2782 ret = ext4_init_inode_table(sb, group,
2783 elr->lr_timeout ? 0 : 1);
2784 if (elr->lr_timeout == 0) {
51ce6511
LC		 2785 timeout = (jiffies - timeout) *
2786 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2787 elr->lr_timeout = timeout;
2788 }
2789 elr->lr_next_sched = jiffies + elr->lr_timeout;
2790 elr->lr_next_group = group + 1;
2791 }
2792
2793 return ret;
2794}
2795
2796/*
2797 * Remove lr_request from the list_request and free the
4ed5c033 2798 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2799 */
2800static void ext4_remove_li_request(struct ext4_li_request *elr)
2801{
2802 struct ext4_sb_info *sbi;
2803
2804 if (!elr)
2805 return;
2806
2807 sbi = elr->lr_sbi;
2808
2809 list_del(&elr->lr_request);
2810 sbi->s_li_request = NULL;
2811 kfree(elr);
2812}
2813
2814static void ext4_unregister_li_request(struct super_block *sb)
2815{
1bb933fb
LC		 2816 mutex_lock(&ext4_li_mtx);
2817 if (!ext4_li_info) {
2818 mutex_unlock(&ext4_li_mtx);
bfff6873 2819 return;
1bb933fb 2820 }
bfff6873
LC		 2821
2822 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2823 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2824 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2825 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2826}
2827
8f1f7453
ES		 2828static struct task_struct *ext4_lazyinit_task;
2829
bfff6873
LC		 2830/*
2831 * This is the function where ext4lazyinit thread lives. It walks
2832 * through the request list searching for next scheduled filesystem.
2833 * When such a fs is found, run the lazy initialization request
2834 * (ext4_rn_li_request) and keep track of the time spend in this
2835 * function. Based on that time we compute next schedule time of
2836 * the request. When walking through the list is complete, compute
2837 * next waking time and put itself into sleep.
2838 */
2839static int ext4_lazyinit_thread(void *arg)
2840{
2841 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2842 struct list_head *pos, *n;
2843 struct ext4_li_request *elr;
4ed5c033 2844 unsigned long next_wakeup, cur;
bfff6873
LC		 2845
2846 BUG_ON(NULL == eli);
2847
bfff6873
LC		 2848cont_thread:
2849 while (true) {
2850 next_wakeup = MAX_JIFFY_OFFSET;
2851
2852 mutex_lock(&eli->li_list_mtx);
2853 if (list_empty(&eli->li_request_list)) {
2854 mutex_unlock(&eli->li_list_mtx);
2855 goto exit_thread;
2856 }
2857
2858 list_for_each_safe(pos, n, &eli->li_request_list) {
2859 elr = list_entry(pos, struct ext4_li_request,
2860 lr_request);
2861
b2c78cd0
TT		 2862 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2863 if (ext4_run_li_request(elr) != 0) {
2864 /* error, remove the lazy_init job */
2865 ext4_remove_li_request(elr);
2866 continue;
2867 }
bfff6873
LC		 2868 }
2869
2870 if (time_before(elr->lr_next_sched, next_wakeup))
2871 next_wakeup = elr->lr_next_sched;
2872 }
2873 mutex_unlock(&eli->li_list_mtx);
2874
2875 if (freezing(current))
2876 refrigerator();
2877
4ed5c033
LC		 2878 cur = jiffies;
2879 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2880 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2881 cond_resched();
2882 continue;
2883 }
2884
4ed5c033
LC		 2885 schedule_timeout_interruptible(next_wakeup - cur);
2886
8f1f7453
ES		 2887 if (kthread_should_stop()) {
2888 ext4_clear_request_list();
2889 goto exit_thread;
2890 }
bfff6873
LC		 2891 }
2892
2893exit_thread:
2894 /*
2895 * It looks like the request list is empty, but we need
2896 * to check it under the li_list_mtx lock, to prevent any
2897 * additions into it, and of course we should lock ext4_li_mtx
2898 * to atomically free the list and ext4_li_info, because at
2899 * this point another ext4 filesystem could be registering
2900 * new one.
2901 */
2902 mutex_lock(&ext4_li_mtx);
2903 mutex_lock(&eli->li_list_mtx);
2904 if (!list_empty(&eli->li_request_list)) {
2905 mutex_unlock(&eli->li_list_mtx);
2906 mutex_unlock(&ext4_li_mtx);
2907 goto cont_thread;
2908 }
2909 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2910 kfree(ext4_li_info);
2911 ext4_li_info = NULL;
2912 mutex_unlock(&ext4_li_mtx);
2913
2914 return 0;
2915}
2916
2917static void ext4_clear_request_list(void)
2918{
2919 struct list_head *pos, *n;
2920 struct ext4_li_request *elr;
2921
2922 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2923 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2924 elr = list_entry(pos, struct ext4_li_request,
2925 lr_request);
2926 ext4_remove_li_request(elr);
2927 }
2928 mutex_unlock(&ext4_li_info->li_list_mtx);
2929}
2930
2931static int ext4_run_lazyinit_thread(void)
2932{
8f1f7453
ES		 2933 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2934 ext4_li_info, "ext4lazyinit");
2935 if (IS_ERR(ext4_lazyinit_task)) {
2936 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2937 ext4_clear_request_list();
bfff6873
LC		 2938 kfree(ext4_li_info);
2939 ext4_li_info = NULL;
2940 printk(KERN_CRIT "EXT4: error %d creating inode table "
2941 "initialization thread\n",
2942 err);
2943 return err;
2944 }
2945 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2946 return 0;
2947}
2948
2949/*
2950 * Check whether it make sense to run itable init. thread or not.
2951 * If there is at least one uninitialized inode table, return
2952 * corresponding group number, else the loop goes through all
2953 * groups and return total number of groups.
2954 */
2955static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2956{
2957 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2958 struct ext4_group_desc *gdp = NULL;
2959
2960 for (group = 0; group < ngroups; group++) {
2961 gdp = ext4_get_group_desc(sb, group, NULL);
2962 if (!gdp)
2963 continue;
2964
2965 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2966 break;
2967 }
2968
2969 return group;
2970}
2971
2972static int ext4_li_info_new(void)
2973{
2974 struct ext4_lazy_init *eli = NULL;
2975
2976 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2977 if (!eli)
2978 return -ENOMEM;
2979
bfff6873
LC		 2980 INIT_LIST_HEAD(&eli->li_request_list);
2981 mutex_init(&eli->li_list_mtx);
2982
bfff6873
LC		 2983 eli->li_state |= EXT4_LAZYINIT_QUIT;
2984
2985 ext4_li_info = eli;
2986
2987 return 0;
2988}
2989
2990static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2991 ext4_group_t start)
2992{
2993 struct ext4_sb_info *sbi = EXT4_SB(sb);
2994 struct ext4_li_request *elr;
2995 unsigned long rnd;
2996
2997 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2998 if (!elr)
2999 return NULL;
3000
3001 elr->lr_super = sb;
3002 elr->lr_sbi = sbi;
3003 elr->lr_next_group = start;
3004
3005 /*
3006 * Randomize first schedule time of the request to
3007 * spread the inode table initialization requests
3008 * better.
3009 */
3010 get_random_bytes(&rnd, sizeof(rnd));
3011 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3012 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3013
3014 return elr;
3015}
3016
3017static int ext4_register_li_request(struct super_block *sb,
3018 ext4_group_t first_not_zeroed)
3019{
3020 struct ext4_sb_info *sbi = EXT4_SB(sb);
3021 struct ext4_li_request *elr;
3022 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3023 int ret = 0;
bfff6873 3024
51ce6511
LC		 3025 if (sbi->s_li_request != NULL) {
3026 /*
3027 * Reset timeout so it can be computed again, because
3028 * s_li_wait_mult might have changed.
3029 */
3030 sbi->s_li_request->lr_timeout = 0;
beed5ecb 3031 return 0;
51ce6511 3032 }
bfff6873
LC		 3033
3034 if (first_not_zeroed == ngroups ||
3035 (sb->s_flags & MS_RDONLY) ||
55ff3840 3036 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 3037 return 0;
bfff6873
LC		 3038
3039 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 3040 if (!elr)
3041 return -ENOMEM;
bfff6873
LC		 3042
3043 mutex_lock(&ext4_li_mtx);
3044
3045 if (NULL == ext4_li_info) {
3046 ret = ext4_li_info_new();
3047 if (ret)
3048 goto out;
3049 }
3050
3051 mutex_lock(&ext4_li_info->li_list_mtx);
3052 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3053 mutex_unlock(&ext4_li_info->li_list_mtx);
3054
3055 sbi->s_li_request = elr;
46e4690b
TM		 3056 /*
3057 * set elr to NULL here since it has been inserted to
3058 * the request_list and the removal and free of it is
3059 * handled by ext4_clear_request_list from now on.
3060 */
3061 elr = NULL;
bfff6873
LC		 3062
3063 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3064 ret = ext4_run_lazyinit_thread();
3065 if (ret)
3066 goto out;
3067 }
bfff6873 3068out:
beed5ecb
NK		 3069 mutex_unlock(&ext4_li_mtx);
3070 if (ret)
bfff6873 3071 kfree(elr);
bfff6873
LC		 3072 return ret;
3073}
3074
3075/*
3076 * We do not need to lock anything since this is called on
3077 * module unload.
3078 */
3079static void ext4_destroy_lazyinit_thread(void)
3080{
3081 /*
3082 * If thread exited earlier
3083 * there's nothing to be done.
3084 */
8f1f7453 3085 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3086 return;
3087
8f1f7453 3088 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3089}
3090
2b2d6d01 3091static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 3092 __releases(kernel_lock)
3093 __acquires(kernel_lock)
ac27a0ec 3094{
d4c402d9 3095 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3096 struct buffer_head *bh;
617ba13b
MC		 3097 struct ext4_super_block *es = NULL;
3098 struct ext4_sb_info *sbi;
3099 ext4_fsblk_t block;
3100 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3101 ext4_fsblk_t logical_sb_block;
ac27a0ec 3102 unsigned long offset = 0;
ac27a0ec
DK		 3103 unsigned long journal_devnum = 0;
3104 unsigned long def_mount_opts;
3105 struct inode *root;
9f6200bb 3106 char *cp;
0390131b 3107 const char *descr;
dcc7dae3 3108 int ret = -ENOMEM;
281b5995 3109 int blocksize, clustersize;
4ec11028
TT		 3110 unsigned int db_count;
3111 unsigned int i;
281b5995 3112 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3113 __u64 blocks_count;
833f4077 3114 int err;
b3881f74 3115 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3116 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3117
3118 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3119 if (!sbi)
dcc7dae3 3120 goto out_free_orig;
705895b6
PE		 3121
3122 sbi->s_blockgroup_lock =
3123 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3124 if (!sbi->s_blockgroup_lock) {
3125 kfree(sbi);
dcc7dae3 3126 goto out_free_orig;
705895b6 3127 }
ac27a0ec
DK		 3128 sb->s_fs_info = sbi;
3129 sbi->s_mount_opt = 0;
617ba13b
MC		 3130 sbi->s_resuid = EXT4_DEF_RESUID;
3131 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 3132 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3133 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3134 if (sb->s_bdev->bd_part)
3135 sbi->s_sectors_written_start =
3136 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3137
9f6200bb
TT		 3138 /* Cleanup superblock name */
3139 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3140 *cp = '!';
3141
dcc7dae3 3142 ret = -EINVAL;
617ba13b 3143 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3144 if (!blocksize) {
b31e1552 3145 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3146 goto out_fail;
3147 }
3148
3149 /*
617ba13b 3150 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3151 * block sizes. We need to calculate the offset from buffer start.
3152 */
617ba13b 3153 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3154 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3155 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3156 } else {
70bbb3e0 3157 logical_sb_block = sb_block;
ac27a0ec
DK		 3158 }
3159
70bbb3e0 3160 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3161 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3162 goto out_fail;
3163 }
3164 /*
3165 * Note: s_es must be initialized as soon as possible because
617ba13b 3166 * some ext4 macro-instructions depend on its value
ac27a0ec 3167 */
617ba13b 3168 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3169 sbi->s_es = es;
3170 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3171 if (sb->s_magic != EXT4_SUPER_MAGIC)
3172 goto cantfind_ext4;
afc32f7e 3173 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 3174
3175 /* Set defaults before we parse the mount options */
3176 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3177 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3178 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3179 set_opt(sb, DEBUG);
437ca0fd
DM		 3180 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3181 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3182 "2.6.38");
fd8c37ec 3183 set_opt(sb, GRPID);
437ca0fd 3184 }
617ba13b 3185 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3186 set_opt(sb, NO_UID32);
ea663336 3187 /* xattr user namespace & acls are now defaulted on */
03010a33 3188#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3189 set_opt(sb, XATTR_USER);
2e7842b8 3190#endif
03010a33 3191#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3192 set_opt(sb, POSIX_ACL);
2e7842b8 3193#endif
6fd7a467 3194 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3195 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3196 set_opt(sb, JOURNAL_DATA);
617ba13b 3197 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3198 set_opt(sb, ORDERED_DATA);
617ba13b 3199 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3200 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3201
3202 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3203 set_opt(sb, ERRORS_PANIC);
bb4f397a 3204 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3205 set_opt(sb, ERRORS_CONT);
bb4f397a 3206 else
fd8c37ec 3207 set_opt(sb, ERRORS_RO);
8b67f04a 3208 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3209 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3210 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3211 set_opt(sb, DISCARD);
ac27a0ec
DK		 3212
3213 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3214 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3215 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3216 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3217 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3218
8b67f04a 3219 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3220 set_opt(sb, BARRIER);
ac27a0ec 3221
dd919b98
AK		 3222 /*
3223 * enable delayed allocation by default
3224 * Use -o nodelalloc to turn it off
3225 */
8b67f04a
TT		 3226 if (!IS_EXT3_SB(sb) &&
3227 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3228 set_opt(sb, DELALLOC);
dd919b98 3229
51ce6511
LC		 3230 /*
3231 * set default s_li_wait_mult for lazyinit, for the case there is
3232 * no mount option specified.
3233 */
3234 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3235
8b67f04a
TT		 3236 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3237 &journal_devnum, &journal_ioprio, NULL, 0)) {
3238 ext4_msg(sb, KERN_WARNING,
3239 "failed to parse options in superblock: %s",
3240 sbi->s_es->s_mount_opts);
3241 }
b3881f74
TT		 3242 if (!parse_options((char *) data, sb, &journal_devnum,
3243 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 3244 goto failed_mount;
3245
56889787
TT		 3246 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3247 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3248 "with data=journal disables delayed "
3249 "allocation and O_DIRECT support!\n");
3250 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3251 ext4_msg(sb, KERN_ERR, "can't mount with "
3252 "both data=journal and delalloc");
3253 goto failed_mount;
3254 }
3255 if (test_opt(sb, DIOREAD_NOLOCK)) {
3256 ext4_msg(sb, KERN_ERR, "can't mount with "
3257 "both data=journal and delalloc");
3258 goto failed_mount;
3259 }
3260 if (test_opt(sb, DELALLOC))
3261 clear_opt(sb, DELALLOC);
3262 }
3263
3264 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3265 if (test_opt(sb, DIOREAD_NOLOCK)) {
3266 if (blocksize < PAGE_SIZE) {
3267 ext4_msg(sb, KERN_ERR, "can't mount with "
3268 "dioread_nolock if block size != PAGE_SIZE");
3269 goto failed_mount;
3270 }
3271 }
3272
ac27a0ec 3273 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3274 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3275
617ba13b
MC		 3276 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3277 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3278 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3279 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3280 ext4_msg(sb, KERN_WARNING,
3281 "feature flags set on rev 0 fs, "
3282 "running e2fsck is recommended");
469108ff 3283
2035e776
TT		 3284 if (IS_EXT2_SB(sb)) {
3285 if (ext2_feature_set_ok(sb))
3286 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3287 "using the ext4 subsystem");
3288 else {
3289 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3290 "to feature incompatibilities");
3291 goto failed_mount;
3292 }
3293 }
3294
3295 if (IS_EXT3_SB(sb)) {
3296 if (ext3_feature_set_ok(sb))
3297 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3298 "using the ext4 subsystem");
3299 else {
3300 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3301 "to feature incompatibilities");
3302 goto failed_mount;
3303 }
3304 }
3305
ac27a0ec
DK		 3306 /*
3307 * Check feature flags regardless of the revision level, since we
3308 * previously didn't change the revision level when setting the flags,
3309 * so there is a chance incompat flags are set on a rev 0 filesystem.
3310 */
a13fb1a4 3311 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3312 goto failed_mount;
a13fb1a4 3313
617ba13b
MC		 3314 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3315 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3316 ext4_msg(sb, KERN_ERR,
3317 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3318 goto failed_mount;
3319 }
3320
ac27a0ec 3321 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3322 /* Validate the filesystem blocksize */
3323 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3324 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3325 blocksize);
ac27a0ec
DK		 3326 goto failed_mount;
3327 }
3328
2b2d6d01 3329 brelse(bh);
70bbb3e0
AM		 3330 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3331 offset = do_div(logical_sb_block, blocksize);
3332 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3333 if (!bh) {
b31e1552
ES		 3334 ext4_msg(sb, KERN_ERR,
3335 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3336 goto failed_mount;
3337 }
617ba13b 3338 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3339 sbi->s_es = es;
617ba13b 3340 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3341 ext4_msg(sb, KERN_ERR,
3342 "Magic mismatch, very weird!");
ac27a0ec
DK		 3343 goto failed_mount;
3344 }
3345 }
3346
a13fb1a4
ES		 3347 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3348 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3349 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3350 has_huge_files);
3351 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3352
617ba13b
MC		 3353 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3354 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3355 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3356 } else {
3357 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3358 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3359 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3360 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3361 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3362 ext4_msg(sb, KERN_ERR,
3363 "unsupported inode size: %d",
2b2d6d01 3364 sbi->s_inode_size);
ac27a0ec
DK		 3365 goto failed_mount;
3366 }
ef7f3835
KS		 3367 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3368 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3369 }
0b8e58a1 3370
0d1ee42f
AR		 3371 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3372 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3373 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3374 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3375 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3376 ext4_msg(sb, KERN_ERR,
3377 "unsupported descriptor size %lu",
0d1ee42f
AR		 3378 sbi->s_desc_size);
3379 goto failed_mount;
3380 }
3381 } else
3382 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3383
ac27a0ec 3384 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3385 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3386 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3387 goto cantfind_ext4;
0b8e58a1 3388
617ba13b 3389 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3390 if (sbi->s_inodes_per_block == 0)
617ba13b 3391 goto cantfind_ext4;
ac27a0ec
DK		 3392 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3393 sbi->s_inodes_per_block;
0d1ee42f 3394 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3395 sbi->s_sbh = bh;
3396 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3397 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3398 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3399
2b2d6d01 3400 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3401 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3402 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3403 i = le32_to_cpu(es->s_flags);
3404 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3405 sbi->s_hash_unsigned = 3;
3406 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3407#ifdef __CHAR_UNSIGNED__
3408 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3409 sbi->s_hash_unsigned = 3;
3410#else
3411 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3412#endif
3413 sb->s_dirt = 1;
3414 }
ac27a0ec 3415
281b5995
TT		 3416 /* Handle clustersize */
3417 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3418 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3419 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3420 if (has_bigalloc) {
3421 if (clustersize < blocksize) {
3422 ext4_msg(sb, KERN_ERR,
3423 "cluster size (%d) smaller than "
3424 "block size (%d)", clustersize, blocksize);
3425 goto failed_mount;
3426 }
3427 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3428 le32_to_cpu(es->s_log_block_size);
3429 sbi->s_clusters_per_group =
3430 le32_to_cpu(es->s_clusters_per_group);
3431 if (sbi->s_clusters_per_group > blocksize * 8) {
3432 ext4_msg(sb, KERN_ERR,
3433 "#clusters per group too big: %lu",
3434 sbi->s_clusters_per_group);
3435 goto failed_mount;
3436 }
3437 if (sbi->s_blocks_per_group !=
3438 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3439 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3440 "clusters per group (%lu) inconsistent",
3441 sbi->s_blocks_per_group,
3442 sbi->s_clusters_per_group);
3443 goto failed_mount;
3444 }
3445 } else {
3446 if (clustersize != blocksize) {
3447 ext4_warning(sb, "fragment/cluster size (%d) != "
3448 "block size (%d)", clustersize,
3449 blocksize);
3450 clustersize = blocksize;
3451 }
3452 if (sbi->s_blocks_per_group > blocksize * 8) {
3453 ext4_msg(sb, KERN_ERR,
3454 "#blocks per group too big: %lu",
3455 sbi->s_blocks_per_group);
3456 goto failed_mount;
3457 }
3458 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3459 sbi->s_cluster_bits = 0;
ac27a0ec 3460 }
281b5995
TT		 3461 sbi->s_cluster_ratio = clustersize / blocksize;
3462
ac27a0ec 3463 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3464 ext4_msg(sb, KERN_ERR,
3465 "#inodes per group too big: %lu",
2b2d6d01 3466 sbi->s_inodes_per_group);
ac27a0ec
DK		 3467 goto failed_mount;
3468 }
3469
bf43d84b
ES		 3470 /*
3471 * Test whether we have more sectors than will fit in sector_t,
3472 * and whether the max offset is addressable by the page cache.
3473 */
5a9ae68a 3474 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3475 ext4_blocks_count(es));
5a9ae68a 3476 if (err) {
b31e1552 3477 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3478 " too large to mount safely on this system");
ac27a0ec 3479 if (sizeof(sector_t) < 8)
90c699a9 3480 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3481 ret = err;
ac27a0ec
DK		 3482 goto failed_mount;
3483 }
3484
617ba13b
MC		 3485 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3486 goto cantfind_ext4;
e7c95593 3487
0f2ddca6
FTN		 3488 /* check blocks count against device size */
3489 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3490 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3491 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3492 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3493 ext4_blocks_count(es), blocks_count);
3494 goto failed_mount;
3495 }
3496
0b8e58a1
AD		 3497 /*
3498 * It makes no sense for the first data block to be beyond the end
3499 * of the filesystem.
3500 */
3501 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 3502 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3503 "block %u is beyond end of filesystem (%llu)",
3504 le32_to_cpu(es->s_first_data_block),
3505 ext4_blocks_count(es));
e7c95593
ES		 3506 goto failed_mount;
3507 }
bd81d8ee
LV		 3508 blocks_count = (ext4_blocks_count(es) -
3509 le32_to_cpu(es->s_first_data_block) +
3510 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3511 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3512 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3513 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3514 "(block count %llu, first data block %u, "
b31e1552 3515 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3516 ext4_blocks_count(es),
3517 le32_to_cpu(es->s_first_data_block),
3518 EXT4_BLOCKS_PER_GROUP(sb));
3519 goto failed_mount;
3520 }
bd81d8ee 3521 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3522 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3523 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3524 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3525 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3526 sbi->s_group_desc = ext4_kvmalloc(db_count *
3527 sizeof(struct buffer_head *),
3528 GFP_KERNEL);
ac27a0ec 3529 if (sbi->s_group_desc == NULL) {
b31e1552 3530 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3531 goto failed_mount;
3532 }
3533
3244fcb1 3534#ifdef CONFIG_PROC_FS
9f6200bb
TT		 3535 if (ext4_proc_root)
3536 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 3537#endif
240799cd 3538
705895b6 3539 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3540
3541 for (i = 0; i < db_count; i++) {
70bbb3e0 3542 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3543 sbi->s_group_desc[i] = sb_bread(sb, block);
3544 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3545 ext4_msg(sb, KERN_ERR,
3546 "can't read group descriptor %d", i);
ac27a0ec
DK		 3547 db_count = i;
3548 goto failed_mount2;
3549 }
3550 }
bfff6873 3551 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3552 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3553 goto failed_mount2;
3554 }
772cb7c8
JS		 3555 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3556 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3557 ext4_msg(sb, KERN_ERR,
3558 "unable to initialize "
3559 "flex_bg meta info!");
772cb7c8
JS		 3560 goto failed_mount2;
3561 }
3562
ac27a0ec
DK		 3563 sbi->s_gdb_count = db_count;
3564 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3565 spin_lock_init(&sbi->s_next_gen_lock);
3566
04496411
TM		 3567 init_timer(&sbi->s_err_report);
3568 sbi->s_err_report.function = print_daily_error_info;
3569 sbi->s_err_report.data = (unsigned long) sb;
3570
ce7e010a
TT		 3571 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3572 ext4_count_free_blocks(sb));
3573 if (!err) {
3574 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3575 ext4_count_free_inodes(sb));
3576 }
3577 if (!err) {
3578 err = percpu_counter_init(&sbi->s_dirs_counter,
3579 ext4_count_dirs(sb));
3580 }
3581 if (!err) {
3582 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3583 }
3584 if (err) {
3585 ext4_msg(sb, KERN_ERR, "insufficient memory");
3586 goto failed_mount3;
3587 }
3588
c9de560d 3589 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3590 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3591
ac27a0ec
DK		 3592 /*
3593 * set up enough so that it can read an inode
3594 */
9ca92389
TT		 3595 if (!test_opt(sb, NOLOAD) &&
3596 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3597 sb->s_op = &ext4_sops;
3598 else
3599 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3600 sb->s_export_op = &ext4_export_ops;
3601 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3602#ifdef CONFIG_QUOTA
617ba13b
MC		 3603 sb->s_qcop = &ext4_qctl_operations;
3604 sb->dq_op = &ext4_quota_operations;
ac27a0ec 3605#endif
f2fa2ffc
AK		 3606 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3607
ac27a0ec 3608 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3609 mutex_init(&sbi->s_orphan_lock);
8f82f840 3610 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3611
3612 sb->s_root = NULL;
3613
3614 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3615 EXT4_HAS_INCOMPAT_FEATURE(sb,
3616 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3617
c5e06d10
JL		 3618 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3619 !(sb->s_flags & MS_RDONLY))
3620 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3621 goto failed_mount3;
3622
ac27a0ec
DK		 3623 /*
3624 * The first inode we look at is the journal inode. Don't try
3625 * root first: it may be modified in the journal!
3626 */
3627 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3628 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3629 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3630 goto failed_mount3;
0390131b
FM		 3631 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3632 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3633 ext4_msg(sb, KERN_ERR, "required journal recovery "
3634 "suppressed and not mounted read-only");
744692dc 3635 goto failed_mount_wq;
ac27a0ec 3636 } else {
fd8c37ec 3637 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3638 sbi->s_journal = NULL;
3639 needs_recovery = 0;
3640 goto no_journal;
ac27a0ec
DK		 3641 }
3642
eb40a09c
JS		 3643 if (ext4_blocks_count(es) > 0xffffffffULL &&
3644 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3645 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3646 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3647 goto failed_mount_wq;
eb40a09c
JS		 3648 }
3649
d4da6c9c
LT		 3650 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3651 jbd2_journal_set_features(sbi->s_journal,
3652 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3653 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3654 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3655 jbd2_journal_set_features(sbi->s_journal,
3656 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3657 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3658 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3659 } else {
3660 jbd2_journal_clear_features(sbi->s_journal,
3661 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3662 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3663 }
818d276c 3664
ac27a0ec
DK		 3665 /* We have now updated the journal if required, so we can
3666 * validate the data journaling mode. */
3667 switch (test_opt(sb, DATA_FLAGS)) {
3668 case 0:
3669 /* No mode set, assume a default based on the journal
63f57933
AM		 3670 * capabilities: ORDERED_DATA if the journal can
3671 * cope, else JOURNAL_DATA
3672 */
dab291af
MC		 3673 if (jbd2_journal_check_available_features
3674 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3675 set_opt(sb, ORDERED_DATA);
ac27a0ec 3676 else
fd8c37ec 3677 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3678 break;
3679
617ba13b
MC		 3680 case EXT4_MOUNT_ORDERED_DATA:
3681 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3682 if (!jbd2_journal_check_available_features
3683 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3684 ext4_msg(sb, KERN_ERR, "Journal does not support "
3685 "requested data journaling mode");
744692dc 3686 goto failed_mount_wq;
ac27a0ec
DK		 3687 }
3688 default:
3689 break;
3690 }
b3881f74 3691 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3692
ce7e010a
TT		 3693 /*
3694 * The journal may have updated the bg summary counts, so we
3695 * need to update the global counters.
3696 */
3697 percpu_counter_set(&sbi->s_freeblocks_counter,
3698 ext4_count_free_blocks(sb));
3699 percpu_counter_set(&sbi->s_freeinodes_counter,
3700 ext4_count_free_inodes(sb));
3701 percpu_counter_set(&sbi->s_dirs_counter,
3702 ext4_count_dirs(sb));
3703 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
206f7ab4 3704
ce7e010a 3705no_journal:
fd89d5f2
TH		 3706 /*
3707 * The maximum number of concurrent works can be high and
3708 * concurrency isn't really necessary. Limit it to 1.
3709 */
3710 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3711 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3712 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3713 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3714 goto failed_mount_wq;
3715 }
3716
ac27a0ec 3717 /*
dab291af 3718 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3719 * so we can safely mount the rest of the filesystem now.
3720 */
3721
1d1fe1ee
DH		 3722 root = ext4_iget(sb, EXT4_ROOT_INO);
3723 if (IS_ERR(root)) {
b31e1552 3724 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3725 ret = PTR_ERR(root);
32a9bb57 3726 root = NULL;
ac27a0ec
DK		 3727 goto failed_mount4;
3728 }
3729 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3730 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 3731 goto failed_mount4;
3732 }
1d1fe1ee
DH		 3733 sb->s_root = d_alloc_root(root);
3734 if (!sb->s_root) {
b31e1552 3735 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3736 ret = -ENOMEM;
3737 goto failed_mount4;
3738 }
ac27a0ec 3739
2b2d6d01 3740 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3741
3742 /* determine the minimum size of new large inodes, if present */
3743 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3744 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3745 EXT4_GOOD_OLD_INODE_SIZE;
3746 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3747 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3748 if (sbi->s_want_extra_isize <
3749 le16_to_cpu(es->s_want_extra_isize))
3750 sbi->s_want_extra_isize =
3751 le16_to_cpu(es->s_want_extra_isize);
3752 if (sbi->s_want_extra_isize <
3753 le16_to_cpu(es->s_min_extra_isize))
3754 sbi->s_want_extra_isize =
3755 le16_to_cpu(es->s_min_extra_isize);
3756 }
3757 }
3758 /* Check if enough inode space is available */
3759 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3760 sbi->s_inode_size) {
3761 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3762 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3763 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3764 "available");
ef7f3835
KS		 3765 }
3766
6fd058f7
TT		 3767 err = ext4_setup_system_zone(sb);
3768 if (err) {
b31e1552 3769 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3770 "zone (%d)", err);
6fd058f7
TT		 3771 goto failed_mount4;
3772 }
3773
c2774d84
AK		 3774 ext4_ext_init(sb);
3775 err = ext4_mb_init(sb, needs_recovery);
3776 if (err) {
421f91d2 3777 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3778 err);
c2774d84
AK		 3779 goto failed_mount4;
3780 }
3781
bfff6873
LC		 3782 err = ext4_register_li_request(sb, first_not_zeroed);
3783 if (err)
3784 goto failed_mount4;
3785
3197ebdb
TT		 3786 sbi->s_kobj.kset = ext4_kset;
3787 init_completion(&sbi->s_kobj_unregister);
3788 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3789 "%s", sb->s_id);
3790 if (err) {
3791 ext4_mb_release(sb);
3792 ext4_ext_release(sb);
3793 goto failed_mount4;
3794 };
3795
617ba13b
MC		 3796 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3797 ext4_orphan_cleanup(sb, es);
3798 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3799 if (needs_recovery) {
b31e1552 3800 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3801 ext4_mark_recovery_complete(sb, es);
3802 }
3803 if (EXT4_SB(sb)->s_journal) {
3804 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3805 descr = " journalled data mode";
3806 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3807 descr = " ordered data mode";
3808 else
3809 descr = " writeback data mode";
3810 } else
3811 descr = "out journal";
3812
d4c402d9 3813 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3814 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3815 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3816
66e61a9e
TT		 3817 if (es->s_error_count)
3818 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3819
d4c402d9 3820 kfree(orig_data);
ac27a0ec
DK		 3821 return 0;
3822
617ba13b 3823cantfind_ext4:
ac27a0ec 3824 if (!silent)
b31e1552 3825 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3826 goto failed_mount;
3827
3828failed_mount4:
32a9bb57
MK		 3829 iput(root);
3830 sb->s_root = NULL;
b31e1552 3831 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3832 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3833failed_mount_wq:
6fd058f7 3834 ext4_release_system_zone(sb);
0390131b
FM		 3835 if (sbi->s_journal) {
3836 jbd2_journal_destroy(sbi->s_journal);
3837 sbi->s_journal = NULL;
3838 }
ac27a0ec 3839failed_mount3:
04496411 3840 del_timer(&sbi->s_err_report);
9933fc0a
TT		 3841 if (sbi->s_flex_groups)
3842 ext4_kvfree(sbi->s_flex_groups);
ce7e010a
TT		 3843 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3844 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3845 percpu_counter_destroy(&sbi->s_dirs_counter);
3846 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
c5e06d10
JL		 3847 if (sbi->s_mmp_tsk)
3848 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3849failed_mount2:
3850 for (i = 0; i < db_count; i++)
3851 brelse(sbi->s_group_desc[i]);
f18a5f21 3852 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 3853failed_mount:
240799cd 3854 if (sbi->s_proc) {
9f6200bb 3855 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3856 }
ac27a0ec
DK		 3857#ifdef CONFIG_QUOTA
3858 for (i = 0; i < MAXQUOTAS; i++)
3859 kfree(sbi->s_qf_names[i]);
3860#endif
617ba13b 3861 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3862 brelse(bh);
3863out_fail:
3864 sb->s_fs_info = NULL;
f6830165 3865 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3866 kfree(sbi);
dcc7dae3 3867out_free_orig:
d4c402d9 3868 kfree(orig_data);
1d1fe1ee 3869 return ret;
ac27a0ec
DK		 3870}
3871
3872/*
3873 * Setup any per-fs journal parameters now. We'll do this both on
3874 * initial mount, once the journal has been initialised but before we've
3875 * done any recovery; and again on any subsequent remount.
3876 */
617ba13b 3877static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3878{
617ba13b 3879 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3880
30773840
TT		 3881 journal->j_commit_interval = sbi->s_commit_interval;
3882 journal->j_min_batch_time = sbi->s_min_batch_time;
3883 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3884
a931da6a 3885 write_lock(&journal->j_state_lock);
ac27a0ec 3886 if (test_opt(sb, BARRIER))
dab291af 3887 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3888 else
dab291af 3889 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3890 if (test_opt(sb, DATA_ERR_ABORT))
3891 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3892 else
3893 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3894 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3895}
3896
617ba13b 3897static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3898 unsigned int journal_inum)
3899{
3900 struct inode *journal_inode;
3901 journal_t *journal;
3902
0390131b
FM		 3903 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3904
ac27a0ec
DK		 3905 /* First, test for the existence of a valid inode on disk. Bad
3906 * things happen if we iget() an unused inode, as the subsequent
3907 * iput() will try to delete it. */
3908
1d1fe1ee
DH		 3909 journal_inode = ext4_iget(sb, journal_inum);
3910 if (IS_ERR(journal_inode)) {
b31e1552 3911 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3912 return NULL;
3913 }
3914 if (!journal_inode->i_nlink) {
3915 make_bad_inode(journal_inode);
3916 iput(journal_inode);
b31e1552 3917 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3918 return NULL;
3919 }
3920
e5f8eab8 3921 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3922 journal_inode, journal_inode->i_size);
1d1fe1ee 3923 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3924 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3925 iput(journal_inode);
3926 return NULL;
3927 }
3928
dab291af 3929 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3930 if (!journal) {
b31e1552 3931 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3932 iput(journal_inode);
3933 return NULL;
3934 }
3935 journal->j_private = sb;
617ba13b 3936 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3937 return journal;
3938}
3939
617ba13b 3940static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3941 dev_t j_dev)
3942{
2b2d6d01 3943 struct buffer_head *bh;
ac27a0ec 3944 journal_t *journal;
617ba13b
MC		 3945 ext4_fsblk_t start;
3946 ext4_fsblk_t len;
ac27a0ec 3947 int hblock, blocksize;
617ba13b 3948 ext4_fsblk_t sb_block;
ac27a0ec 3949 unsigned long offset;
2b2d6d01 3950 struct ext4_super_block *es;
ac27a0ec
DK		 3951 struct block_device *bdev;
3952
0390131b
FM		 3953 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3954
b31e1552 3955 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3956 if (bdev == NULL)
3957 return NULL;
3958
ac27a0ec 3959 blocksize = sb->s_blocksize;
e1defc4f 3960 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3961 if (blocksize < hblock) {
b31e1552
ES		 3962 ext4_msg(sb, KERN_ERR,
3963 "blocksize too small for journal device");
ac27a0ec
DK		 3964 goto out_bdev;
3965 }
3966
617ba13b
MC		 3967 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3968 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3969 set_blocksize(bdev, blocksize);
3970 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3971 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3972 "external journal");
ac27a0ec
DK		 3973 goto out_bdev;
3974 }
3975
617ba13b
MC		 3976 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3977 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3978 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3979 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3980 ext4_msg(sb, KERN_ERR, "external journal has "
3981 "bad superblock");
ac27a0ec
DK		 3982 brelse(bh);
3983 goto out_bdev;
3984 }
3985
617ba13b 3986 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3987 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3988 brelse(bh);
3989 goto out_bdev;
3990 }
3991
bd81d8ee 3992 len = ext4_blocks_count(es);
ac27a0ec
DK		 3993 start = sb_block + 1;
3994 brelse(bh); /* we're done with the superblock */
3995
dab291af 3996 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3997 start, len, blocksize);
3998 if (!journal) {
b31e1552 3999 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4000 goto out_bdev;
4001 }
4002 journal->j_private = sb;
4003 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4004 wait_on_buffer(journal->j_sb_buffer);
4005 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4006 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4007 goto out_journal;
4008 }
4009 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4010 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4011 "user (unsupported) - %d",
ac27a0ec
DK		 4012 be32_to_cpu(journal->j_superblock->s_nr_users));
4013 goto out_journal;
4014 }
617ba13b
MC		 4015 EXT4_SB(sb)->journal_bdev = bdev;
4016 ext4_init_journal_params(sb, journal);
ac27a0ec 4017 return journal;
0b8e58a1 4018
ac27a0ec 4019out_journal:
dab291af 4020 jbd2_journal_destroy(journal);
ac27a0ec 4021out_bdev:
617ba13b 4022 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4023 return NULL;
4024}
4025
617ba13b
MC		 4026static int ext4_load_journal(struct super_block *sb,
4027 struct ext4_super_block *es,
ac27a0ec
DK		 4028 unsigned long journal_devnum)
4029{
4030 journal_t *journal;
4031 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4032 dev_t journal_dev;
4033 int err = 0;
4034 int really_read_only;
4035
0390131b
FM		 4036 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4037
ac27a0ec
DK		 4038 if (journal_devnum &&
4039 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4040 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4041 "numbers have changed");
ac27a0ec
DK		 4042 journal_dev = new_decode_dev(journal_devnum);
4043 } else
4044 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4045
4046 really_read_only = bdev_read_only(sb->s_bdev);
4047
4048 /*
4049 * Are we loading a blank journal or performing recovery after a
4050 * crash? For recovery, we need to check in advance whether we
4051 * can get read-write access to the device.
4052 */
617ba13b 4053 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4054 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4055 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4056 "required on readonly filesystem");
ac27a0ec 4057 if (really_read_only) {
b31e1552
ES		 4058 ext4_msg(sb, KERN_ERR, "write access "
4059 "unavailable, cannot proceed");
ac27a0ec
DK		 4060 return -EROFS;
4061 }
b31e1552
ES		 4062 ext4_msg(sb, KERN_INFO, "write access will "
4063 "be enabled during recovery");
ac27a0ec
DK		 4064 }
4065 }
4066
4067 if (journal_inum && journal_dev) {
b31e1552
ES		 4068 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4069 "and inode journals!");
ac27a0ec
DK		 4070 return -EINVAL;
4071 }
4072
4073 if (journal_inum) {
617ba13b 4074 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4075 return -EINVAL;
4076 } else {
617ba13b 4077 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4078 return -EINVAL;
4079 }
4080
90576c0b 4081 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4082 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4083
ac27a0ec 4084 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 4085 err = jbd2_journal_update_format(journal);
ac27a0ec 4086 if (err) {
b31e1552 4087 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 4088 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4089 return err;
4090 }
4091 }
4092
617ba13b 4093 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4094 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4095 if (!err) {
4096 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4097 if (save)
4098 memcpy(save, ((char *) es) +
4099 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4100 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4101 if (save)
4102 memcpy(((char *) es) + EXT4_S_ERR_START,
4103 save, EXT4_S_ERR_LEN);
4104 kfree(save);
4105 }
ac27a0ec
DK		 4106
4107 if (err) {
b31e1552 4108 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4109 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4110 return err;
4111 }
4112
617ba13b
MC		 4113 EXT4_SB(sb)->s_journal = journal;
4114 ext4_clear_journal_err(sb, es);
ac27a0ec 4115
c41303ce 4116 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4117 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4118 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4119
4120 /* Make sure we flush the recovery flag to disk. */
e2d67052 4121 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4122 }
4123
4124 return 0;
4125}
4126
e2d67052 4127static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4128{
e2d67052 4129 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4130 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4131 int error = 0;
ac27a0ec 4132
7c2e7087 4133 if (!sbh || block_device_ejected(sb))
c4be0c1d 4134 return error;
914258bf
TT		 4135 if (buffer_write_io_error(sbh)) {
4136 /*
4137 * Oh, dear. A previous attempt to write the
4138 * superblock failed. This could happen because the
4139 * USB device was yanked out. Or it could happen to
4140 * be a transient write error and maybe the block will
4141 * be remapped. Nothing we can do but to retry the
4142 * write and hope for the best.
4143 */
b31e1552
ES		 4144 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4145 "superblock detected");
914258bf
TT		 4146 clear_buffer_write_io_error(sbh);
4147 set_buffer_uptodate(sbh);
4148 }
71290b36
TT		 4149 /*
4150 * If the file system is mounted read-only, don't update the
4151 * superblock write time. This avoids updating the superblock
4152 * write time when we are mounting the root file system
4153 * read/only but we need to replay the journal; at that point,
4154 * for people who are east of GMT and who make their clock
4155 * tick in localtime for Windows bug-for-bug compatibility,
4156 * the clock is set in the future, and this will cause e2fsck
4157 * to complain and force a full file system check.
4158 */
4159 if (!(sb->s_flags & MS_RDONLY))
4160 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4161 if (sb->s_bdev->bd_part)
4162 es->s_kbytes_written =
4163 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4164 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4165 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4166 else
4167 es->s_kbytes_written =
4168 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
ce7e010a
TT		 4169 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
4170 &EXT4_SB(sb)->s_freeblocks_counter));
4171 es->s_free_inodes_count =
4172 cpu_to_le32(percpu_counter_sum_positive(
4173 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 4174 sb->s_dirt = 0;
ac27a0ec
DK		 4175 BUFFER_TRACE(sbh, "marking dirty");
4176 mark_buffer_dirty(sbh);
914258bf 4177 if (sync) {
c4be0c1d
TS		 4178 error = sync_dirty_buffer(sbh);
4179 if (error)
4180 return error;
4181
4182 error = buffer_write_io_error(sbh);
4183 if (error) {
b31e1552
ES		 4184 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4185 "superblock");
914258bf
TT		 4186 clear_buffer_write_io_error(sbh);
4187 set_buffer_uptodate(sbh);
4188 }
4189 }
c4be0c1d 4190 return error;
ac27a0ec
DK		 4191}
4192
ac27a0ec
DK		 4193/*
4194 * Have we just finished recovery? If so, and if we are mounting (or
4195 * remounting) the filesystem readonly, then we will end up with a
4196 * consistent fs on disk. Record that fact.
4197 */
2b2d6d01
TT		 4198static void ext4_mark_recovery_complete(struct super_block *sb,
4199 struct ext4_super_block *es)
ac27a0ec 4200{
617ba13b 4201 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4202
0390131b
FM		 4203 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4204 BUG_ON(journal != NULL);
4205 return;
4206 }
dab291af 4207 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4208 if (jbd2_journal_flush(journal) < 0)
4209 goto out;
4210
617ba13b 4211 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4212 sb->s_flags & MS_RDONLY) {
617ba13b 4213 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4214 ext4_commit_super(sb, 1);
ac27a0ec 4215 }
7ffe1ea8
HK		 4216
4217out:
dab291af 4218 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4219}
4220
4221/*
4222 * If we are mounting (or read-write remounting) a filesystem whose journal
4223 * has recorded an error from a previous lifetime, move that error to the
4224 * main filesystem now.
4225 */
2b2d6d01
TT		 4226static void ext4_clear_journal_err(struct super_block *sb,
4227 struct ext4_super_block *es)
ac27a0ec
DK		 4228{
4229 journal_t *journal;
4230 int j_errno;
4231 const char *errstr;
4232
0390131b
FM		 4233 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4234
617ba13b 4235 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4236
4237 /*
4238 * Now check for any error status which may have been recorded in the
617ba13b 4239 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4240 */
4241
dab291af 4242 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4243 if (j_errno) {
4244 char nbuf[16];
4245
617ba13b 4246 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4247 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4248 "from previous mount: %s", errstr);
12062ddd 4249 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4250
617ba13b
MC		 4251 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4252 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4253 ext4_commit_super(sb, 1);
ac27a0ec 4254
dab291af 4255 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4256 }
4257}
4258
4259/*
4260 * Force the running and committing transactions to commit,
4261 * and wait on the commit.
4262 */
617ba13b 4263int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4264{
4265 journal_t *journal;
0390131b 4266 int ret = 0;
ac27a0ec
DK		 4267
4268 if (sb->s_flags & MS_RDONLY)
4269 return 0;
4270
617ba13b 4271 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4272 if (journal) {
437f88cc 4273 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4274 ret = ext4_journal_force_commit(journal);
6b0310fb 4275 }
0390131b 4276
ac27a0ec
DK		 4277 return ret;
4278}
4279
2b2d6d01 4280static void ext4_write_super(struct super_block *sb)
ac27a0ec 4281{
ebc1ac16 4282 lock_super(sb);
9ca92389 4283 ext4_commit_super(sb, 1);
ebc1ac16 4284 unlock_super(sb);
ac27a0ec
DK		 4285}
4286
617ba13b 4287static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4288{
14ce0cb4 4289 int ret = 0;
9eddacf9 4290 tid_t target;
8d5d02e6 4291 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4292
9bffad1e 4293 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4294 flush_workqueue(sbi->dio_unwritten_wq);
4295 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4296 if (wait)
8d5d02e6 4297 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4298 }
14ce0cb4 4299 return ret;
ac27a0ec
DK		 4300}
4301
4302/*
4303 * LVM calls this function before a (read-only) snapshot is created. This
4304 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4305 *
4306 * Note that only this function cannot bring a filesystem to be in a clean
4307 * state independently, because ext4 prevents a new handle from being started
4308 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4309 * the upper layer.
ac27a0ec 4310 */
c4be0c1d 4311static int ext4_freeze(struct super_block *sb)
ac27a0ec 4312{
c4be0c1d
TS		 4313 int error = 0;
4314 journal_t *journal;
ac27a0ec 4315
9ca92389
TT		 4316 if (sb->s_flags & MS_RDONLY)
4317 return 0;
ac27a0ec 4318
9ca92389 4319 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4320
9ca92389
TT		 4321 /* Now we set up the journal barrier. */
4322 jbd2_journal_lock_updates(journal);
ac27a0ec 4323
9ca92389
TT		 4324 /*
4325 * Don't clear the needs_recovery flag if we failed to flush
4326 * the journal.
4327 */
4328 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4329 if (error < 0)
4330 goto out;
9ca92389
TT		 4331
4332 /* Journal blocked and flushed, clear needs_recovery flag. */
4333 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4334 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4335out:
4336 /* we rely on s_frozen to stop further updates */
4337 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4338 return error;
ac27a0ec
DK		 4339}
4340
4341/*
4342 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4343 * flag here, even though the filesystem is not technically dirty yet.
4344 */
c4be0c1d 4345static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4346{
9ca92389
TT		 4347 if (sb->s_flags & MS_RDONLY)
4348 return 0;
4349
4350 lock_super(sb);
4351 /* Reset the needs_recovery flag before the fs is unlocked. */
4352 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4353 ext4_commit_super(sb, 1);
4354 unlock_super(sb);
c4be0c1d 4355 return 0;
ac27a0ec
DK		 4356}
4357
673c6100
TT		 4358/*
4359 * Structure to save mount options for ext4_remount's benefit
4360 */
4361struct ext4_mount_options {
4362 unsigned long s_mount_opt;
a2595b8a 4363 unsigned long s_mount_opt2;
673c6100
TT		 4364 uid_t s_resuid;
4365 gid_t s_resgid;
4366 unsigned long s_commit_interval;
4367 u32 s_min_batch_time, s_max_batch_time;
4368#ifdef CONFIG_QUOTA
4369 int s_jquota_fmt;
4370 char *s_qf_names[MAXQUOTAS];
4371#endif
4372};
4373
2b2d6d01 4374static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4375{
2b2d6d01 4376 struct ext4_super_block *es;
617ba13b
MC		 4377 struct ext4_sb_info *sbi = EXT4_SB(sb);
4378 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 4379 unsigned long old_sb_flags;
617ba13b 4380 struct ext4_mount_options old_opts;
c79d967d 4381 int enable_quota = 0;
8a266467 4382 ext4_group_t g;
b3881f74 4383 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4384 int err = 0;
ac27a0ec
DK		 4385#ifdef CONFIG_QUOTA
4386 int i;
4387#endif
d4c402d9 4388 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4389
4390 /* Store the original options */
bbd6851a 4391 lock_super(sb);
ac27a0ec
DK		 4392 old_sb_flags = sb->s_flags;
4393 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4394 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4395 old_opts.s_resuid = sbi->s_resuid;
4396 old_opts.s_resgid = sbi->s_resgid;
4397 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4398 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4399 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4400#ifdef CONFIG_QUOTA
4401 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4402 for (i = 0; i < MAXQUOTAS; i++)
4403 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4404#endif
b3881f74
TT		 4405 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4406 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4407
4408 /*
4409 * Allow the "check" option to be passed as a remount option.
4410 */
b3881f74
TT		 4411 if (!parse_options(data, sb, NULL, &journal_ioprio,
4412 &n_blocks_count, 1)) {
ac27a0ec
DK		 4413 err = -EINVAL;
4414 goto restore_opts;
4415 }
4416
4ab2f15b 4417 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4418 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4419
4420 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4421 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4422
4423 es = sbi->s_es;
4424
b3881f74 4425 if (sbi->s_journal) {
0390131b 4426 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4427 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4428 }
ac27a0ec
DK		 4429
4430 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 4431 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 4432 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4433 err = -EROFS;
4434 goto restore_opts;
4435 }
4436
4437 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4438 err = dquot_suspend(sb, -1);
4439 if (err < 0)
c79d967d 4440 goto restore_opts;
c79d967d 4441
ac27a0ec
DK		 4442 /*
4443 * First of all, the unconditional stuff we have to do
4444 * to disable replay of the journal when we next remount
4445 */
4446 sb->s_flags |= MS_RDONLY;
4447
4448 /*
4449 * OK, test if we are remounting a valid rw partition
4450 * readonly, and if so set the rdonly flag and then
4451 * mark the partition as valid again.
4452 */
617ba13b
MC		 4453 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4454 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4455 es->s_state = cpu_to_le16(sbi->s_mount_state);
4456
a63c9eb2 4457 if (sbi->s_journal)
0390131b 4458 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4459 } else {
a13fb1a4
ES		 4460 /* Make sure we can mount this feature set readwrite */
4461 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4462 err = -EROFS;
4463 goto restore_opts;
4464 }
8a266467
TT		 4465 /*
4466 * Make sure the group descriptor checksums
0b8e58a1 4467 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4468 */
4469 for (g = 0; g < sbi->s_groups_count; g++) {
4470 struct ext4_group_desc *gdp =
4471 ext4_get_group_desc(sb, g, NULL);
4472
4473 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4474 ext4_msg(sb, KERN_ERR,
4475 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4476 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4477 le16_to_cpu(gdp->bg_checksum));
4478 err = -EINVAL;
4479 goto restore_opts;
4480 }
4481 }
4482
ead6596b
ES		 4483 /*
4484 * If we have an unprocessed orphan list hanging
4485 * around from a previously readonly bdev mount,
4486 * require a full umount/remount for now.
4487 */
4488 if (es->s_last_orphan) {
b31e1552 4489 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4490 "remount RDWR because of unprocessed "
4491 "orphan inode list. Please "
b31e1552 4492 "umount/remount instead");
ead6596b
ES		 4493 err = -EINVAL;
4494 goto restore_opts;
4495 }
4496
ac27a0ec
DK		 4497 /*
4498 * Mounting a RDONLY partition read-write, so reread
4499 * and store the current valid flag. (It may have
4500 * been changed by e2fsck since we originally mounted
4501 * the partition.)
4502 */
0390131b
FM		 4503 if (sbi->s_journal)
4504 ext4_clear_journal_err(sb, es);
ac27a0ec 4505 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 4506 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 4507 goto restore_opts;
2b2d6d01 4508 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4509 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4510 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4511 EXT4_FEATURE_INCOMPAT_MMP))
4512 if (ext4_multi_mount_protect(sb,
4513 le64_to_cpu(es->s_mmp_block))) {
4514 err = -EROFS;
4515 goto restore_opts;
4516 }
c79d967d 4517 enable_quota = 1;
ac27a0ec
DK		 4518 }
4519 }
bfff6873
LC		 4520
4521 /*
4522 * Reinitialize lazy itable initialization thread based on
4523 * current settings
4524 */
4525 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4526 ext4_unregister_li_request(sb);
4527 else {
4528 ext4_group_t first_not_zeroed;
4529 first_not_zeroed = ext4_has_uninit_itable(sb);
4530 ext4_register_li_request(sb, first_not_zeroed);
4531 }
4532
6fd058f7 4533 ext4_setup_system_zone(sb);
0390131b 4534 if (sbi->s_journal == NULL)
e2d67052 4535 ext4_commit_super(sb, 1);
0390131b 4536
ac27a0ec
DK		 4537#ifdef CONFIG_QUOTA
4538 /* Release old quota file names */
4539 for (i = 0; i < MAXQUOTAS; i++)
4540 if (old_opts.s_qf_names[i] &&
4541 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4542 kfree(old_opts.s_qf_names[i]);
4543#endif
bbd6851a 4544 unlock_super(sb);
c79d967d 4545 if (enable_quota)
0f0dd62f 4546 dquot_resume(sb, -1);
d4c402d9
CW		 4547
4548 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4549 kfree(orig_data);
ac27a0ec 4550 return 0;
0b8e58a1 4551
ac27a0ec
DK		 4552restore_opts:
4553 sb->s_flags = old_sb_flags;
4554 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4555 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4556 sbi->s_resuid = old_opts.s_resuid;
4557 sbi->s_resgid = old_opts.s_resgid;
4558 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4559 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4560 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4561#ifdef CONFIG_QUOTA
4562 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4563 for (i = 0; i < MAXQUOTAS; i++) {
4564 if (sbi->s_qf_names[i] &&
4565 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4566 kfree(sbi->s_qf_names[i]);
4567 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4568 }
4569#endif
bbd6851a 4570 unlock_super(sb);
d4c402d9 4571 kfree(orig_data);
ac27a0ec
DK		 4572 return err;
4573}
4574
2b2d6d01 4575static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4576{
4577 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4578 struct ext4_sb_info *sbi = EXT4_SB(sb);
4579 struct ext4_super_block *es = sbi->s_es;
960cc398 4580 u64 fsid;
d02a9391 4581 s64 bfree;
ac27a0ec 4582
5e70030d
BP		 4583 if (test_opt(sb, MINIX_DF)) {
4584 sbi->s_overhead_last = 0;
6bc9feff 4585 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4586 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4587 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4588
4589 /*
5e70030d
BP		 4590 * Compute the overhead (FS structures). This is constant
4591 * for a given filesystem unless the number of block groups
4592 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4593 */
4594
4595 /*
4596 * All of the blocks before first_data_block are
4597 * overhead
4598 */
4599 overhead = le32_to_cpu(es->s_first_data_block);
4600
4601 /*
4602 * Add the overhead attributed to the superblock and
4603 * block group descriptors. If the sparse superblocks
4604 * feature is turned on, then not all groups have this.
4605 */
4606 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 4607 overhead += ext4_bg_has_super(sb, i) +
4608 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 4609 cond_resched();
4610 }
4611
4612 /*
4613 * Every block group has an inode bitmap, a block
4614 * bitmap, and an inode table.
4615 */
5e70030d
BP		 4616 overhead += ngroups * (2 + sbi->s_itb_per_group);
4617 sbi->s_overhead_last = overhead;
4618 smp_wmb();
6bc9feff 4619 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4620 }
4621
617ba13b 4622 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4623 buf->f_bsize = sb->s_blocksize;
5e70030d 4624 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
d02a9391 4625 bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
6bc6e63f 4626 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
d02a9391
KM		 4627 /* prevent underflow in case that few free space is available */
4628 buf->f_bfree = max_t(s64, bfree, 0);
bd81d8ee
LV		 4629 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4630 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4631 buf->f_bavail = 0;
4632 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4633 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4634 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4635 fsid = le64_to_cpup((void *)es->s_uuid) ^
4636 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4637 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4638 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4639
ac27a0ec
DK		 4640 return 0;
4641}
4642
0b8e58a1
AD		 4643/* Helper function for writing quotas on sync - we need to start transaction
4644 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4645 * Process 1 Process 2
617ba13b 4646 * ext4_create() quota_sync()
a269eb18 4647 * jbd2_journal_start() write_dquot()
871a2931 4648 * dquot_initialize() down(dqio_mutex)
dab291af 4649 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4650 *
4651 */
4652
4653#ifdef CONFIG_QUOTA
4654
4655static inline struct inode *dquot_to_inode(struct dquot *dquot)
4656{
4657 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4658}
4659
617ba13b 4660static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4661{
4662 int ret, err;
4663 handle_t *handle;
4664 struct inode *inode;
4665
4666 inode = dquot_to_inode(dquot);
617ba13b 4667 handle = ext4_journal_start(inode,
0b8e58a1 4668 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4669 if (IS_ERR(handle))
4670 return PTR_ERR(handle);
4671 ret = dquot_commit(dquot);
617ba13b 4672 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4673 if (!ret)
4674 ret = err;
4675 return ret;
4676}
4677
617ba13b 4678static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4679{
4680 int ret, err;
4681 handle_t *handle;
4682
617ba13b 4683 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4684 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4685 if (IS_ERR(handle))
4686 return PTR_ERR(handle);
4687 ret = dquot_acquire(dquot);
617ba13b 4688 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4689 if (!ret)
4690 ret = err;
4691 return ret;
4692}
4693
617ba13b 4694static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4695{
4696 int ret, err;
4697 handle_t *handle;
4698
617ba13b 4699 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4700 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4701 if (IS_ERR(handle)) {
4702 /* Release dquot anyway to avoid endless cycle in dqput() */
4703 dquot_release(dquot);
ac27a0ec 4704 return PTR_ERR(handle);
9c3013e9 4705 }
ac27a0ec 4706 ret = dquot_release(dquot);
617ba13b 4707 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4708 if (!ret)
4709 ret = err;
4710 return ret;
4711}
4712
617ba13b 4713static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4714{
2c8be6b2 4715 /* Are we journaling quotas? */
617ba13b
MC		 4716 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4717 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4718 dquot_mark_dquot_dirty(dquot);
617ba13b 4719 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4720 } else {
4721 return dquot_mark_dquot_dirty(dquot);
4722 }
4723}
4724
617ba13b 4725static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4726{
4727 int ret, err;
4728 handle_t *handle;
4729
4730 /* Data block + inode block */
617ba13b 4731 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4732 if (IS_ERR(handle))
4733 return PTR_ERR(handle);
4734 ret = dquot_commit_info(sb, type);
617ba13b 4735 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4736 if (!ret)
4737 ret = err;
4738 return ret;
4739}
4740
4741/*
4742 * Turn on quotas during mount time - we need to find
4743 * the quota file and such...
4744 */
617ba13b 4745static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4746{
287a8095
CH		 4747 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4748 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4749}
4750
4751/*
4752 * Standard function to be called on quota_on
4753 */
617ba13b 4754static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4755 struct path *path)
ac27a0ec
DK		 4756{
4757 int err;
ac27a0ec
DK		 4758
4759 if (!test_opt(sb, QUOTA))
4760 return -EINVAL;
0623543b 4761
ac27a0ec 4762 /* Quotafile not on the same filesystem? */
f00c9e44 4763 if (path->mnt->mnt_sb != sb)
ac27a0ec 4764 return -EXDEV;
0623543b
JK		 4765 /* Journaling quota? */
4766 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4767 /* Quotafile not in fs root? */
f00c9e44 4768 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4769 ext4_msg(sb, KERN_WARNING,
4770 "Quota file not on filesystem root. "
4771 "Journaled quota will not work");
2b2d6d01 4772 }
0623543b
JK		 4773
4774 /*
4775 * When we journal data on quota file, we have to flush journal to see
4776 * all updates to the file when we bypass pagecache...
4777 */
0390131b 4778 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4779 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4780 /*
4781 * We don't need to lock updates but journal_flush() could
4782 * otherwise be livelocked...
4783 */
4784 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4785 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4786 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4787 if (err)
7ffe1ea8 4788 return err;
0623543b
JK		 4789 }
4790
f00c9e44 4791 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4792}
4793
ca0e05e4
DM		 4794static int ext4_quota_off(struct super_block *sb, int type)
4795{
21f97697
JK		 4796 struct inode *inode = sb_dqopt(sb)->files[type];
4797 handle_t *handle;
4798
87009d86
DM		 4799 /* Force all delayed allocation blocks to be allocated.
4800 * Caller already holds s_umount sem */
4801 if (test_opt(sb, DELALLOC))
ca0e05e4 4802 sync_filesystem(sb);
ca0e05e4 4803
0b268590
AG		 4804 if (!inode)
4805 goto out;
4806
21f97697
JK		 4807 /* Update modification times of quota files when userspace can
4808 * start looking at them */
4809 handle = ext4_journal_start(inode, 1);
4810 if (IS_ERR(handle))
4811 goto out;
4812 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4813 ext4_mark_inode_dirty(handle, inode);
4814 ext4_journal_stop(handle);
4815
4816out:
ca0e05e4
DM		 4817 return dquot_quota_off(sb, type);
4818}
4819
ac27a0ec
DK		 4820/* Read data from quotafile - avoid pagecache and such because we cannot afford
4821 * acquiring the locks... As quota files are never truncated and quota code
25985edc 4822 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 4823 * we don't have to be afraid of races */
617ba13b 4824static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4825 size_t len, loff_t off)
4826{
4827 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4828 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4829 int err = 0;
4830 int offset = off & (sb->s_blocksize - 1);
4831 int tocopy;
4832 size_t toread;
4833 struct buffer_head *bh;
4834 loff_t i_size = i_size_read(inode);
4835
4836 if (off > i_size)
4837 return 0;
4838 if (off+len > i_size)
4839 len = i_size-off;
4840 toread = len;
4841 while (toread > 0) {
4842 tocopy = sb->s_blocksize - offset < toread ?
4843 sb->s_blocksize - offset : toread;
617ba13b 4844 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4845 if (err)
4846 return err;
4847 if (!bh) /* A hole? */
4848 memset(data, 0, tocopy);
4849 else
4850 memcpy(data, bh->b_data+offset, tocopy);
4851 brelse(bh);
4852 offset = 0;
4853 toread -= tocopy;
4854 data += tocopy;
4855 blk++;
4856 }
4857 return len;
4858}
4859
4860/* Write to quotafile (we know the transaction is already started and has
4861 * enough credits) */
617ba13b 4862static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4863 const char *data, size_t len, loff_t off)
4864{
4865 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4866 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4867 int err = 0;
4868 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4869 struct buffer_head *bh;
4870 handle_t *handle = journal_current_handle();
4871
0390131b 4872 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4873 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4874 " cancelled because transaction is not started",
9c3013e9
JK		 4875 (unsigned long long)off, (unsigned long long)len);
4876 return -EIO;
4877 }
67eeb568
DM		 4878 /*
4879 * Since we account only one data block in transaction credits,
4880 * then it is impossible to cross a block boundary.
4881 */
4882 if (sb->s_blocksize - offset < len) {
4883 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4884 " cancelled because not block aligned",
4885 (unsigned long long)off, (unsigned long long)len);
4886 return -EIO;
4887 }
4888
ac27a0ec 4889 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4890 bh = ext4_bread(handle, inode, blk, 1, &err);
4891 if (!bh)
4892 goto out;
62d2b5f2
JK		 4893 err = ext4_journal_get_write_access(handle, bh);
4894 if (err) {
4895 brelse(bh);
4896 goto out;
ac27a0ec 4897 }
67eeb568
DM		 4898 lock_buffer(bh);
4899 memcpy(bh->b_data+offset, data, len);
4900 flush_dcache_page(bh->b_page);
4901 unlock_buffer(bh);
62d2b5f2 4902 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4903 brelse(bh);
ac27a0ec 4904out:
67eeb568 4905 if (err) {
4d04e4fb 4906 mutex_unlock(&inode->i_mutex);
ac27a0ec 4907 return err;
4d04e4fb 4908 }
67eeb568
DM		 4909 if (inode->i_size < off + len) {
4910 i_size_write(inode, off + len);
617ba13b 4911 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 4912 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4913 }
ac27a0ec 4914 mutex_unlock(&inode->i_mutex);
67eeb568 4915 return len;
ac27a0ec
DK		 4916}
4917
4918#endif
4919
152a0836
AV		 4920static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4921 const char *dev_name, void *data)
ac27a0ec 4922{
152a0836 4923 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4924}
4925
37f328eb 4926#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4927static inline void register_as_ext2(void)
4928{
4929 int err = register_filesystem(&ext2_fs_type);
4930 if (err)
4931 printk(KERN_WARNING
4932 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4933}
4934
4935static inline void unregister_as_ext2(void)
4936{
4937 unregister_filesystem(&ext2_fs_type);
4938}
2035e776
TT		 4939
4940static inline int ext2_feature_set_ok(struct super_block *sb)
4941{
4942 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4943 return 0;
4944 if (sb->s_flags & MS_RDONLY)
4945 return 1;
4946 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4947 return 0;
4948 return 1;
4949}
51b7e3c9 4950MODULE_ALIAS("ext2");
24b58424
TT		 4951#else
4952static inline void register_as_ext2(void) { }
4953static inline void unregister_as_ext2(void) { }
2035e776 4954static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4955#endif
4956
37f328eb 4957#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4958static inline void register_as_ext3(void)
4959{
4960 int err = register_filesystem(&ext3_fs_type);
4961 if (err)
4962 printk(KERN_WARNING
4963 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4964}
4965
4966static inline void unregister_as_ext3(void)
4967{
4968 unregister_filesystem(&ext3_fs_type);
4969}
2035e776
TT		 4970
4971static inline int ext3_feature_set_ok(struct super_block *sb)
4972{
4973 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4974 return 0;
4975 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4976 return 0;
4977 if (sb->s_flags & MS_RDONLY)
4978 return 1;
4979 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4980 return 0;
4981 return 1;
4982}
51b7e3c9 4983MODULE_ALIAS("ext3");
24b58424
TT		 4984#else
4985static inline void register_as_ext3(void) { }
4986static inline void unregister_as_ext3(void) { }
2035e776 4987static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4988#endif
4989
03010a33
TT		 4990static struct file_system_type ext4_fs_type = {
4991 .owner = THIS_MODULE,
4992 .name = "ext4",
152a0836 4993 .mount = ext4_mount,
03010a33
TT		 4994 .kill_sb = kill_block_super,
4995 .fs_flags = FS_REQUIRES_DEV,
4996};
4997
8f021222 4998static int __init ext4_init_feat_adverts(void)
857ac889
LC		 4999{
5000 struct ext4_features *ef;
5001 int ret = -ENOMEM;
5002
5003 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5004 if (!ef)
5005 goto out;
5006
5007 ef->f_kobj.kset = ext4_kset;
5008 init_completion(&ef->f_kobj_unregister);
5009 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5010 "features");
5011 if (ret) {
5012 kfree(ef);
5013 goto out;
5014 }
5015
5016 ext4_feat = ef;
5017 ret = 0;
5018out:
5019 return ret;
5020}
5021
8f021222
LC		 5022static void ext4_exit_feat_adverts(void)
5023{
5024 kobject_put(&ext4_feat->f_kobj);
5025 wait_for_completion(&ext4_feat->f_kobj_unregister);
5026 kfree(ext4_feat);
5027}
5028
e9e3bcec
ES		 5029/* Shared across all ext4 file systems */
5030wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5031struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5032
5dabfc78 5033static int __init ext4_init_fs(void)
ac27a0ec 5034{
e9e3bcec 5035 int i, err;
c9de560d 5036
12e9b892 5037 ext4_check_flag_values();
e9e3bcec
ES		 5038
5039 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5040 mutex_init(&ext4__aio_mutex[i]);
5041 init_waitqueue_head(&ext4__ioend_wq[i]);
5042 }
5043
5dabfc78 5044 err = ext4_init_pageio();
6fd058f7
TT		 5045 if (err)
5046 return err;
5dabfc78 5047 err = ext4_init_system_zone();
bd2d0210 5048 if (err)
dd68314c 5049 goto out7;
3197ebdb
TT		 5050 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5051 if (!ext4_kset)
dd68314c 5052 goto out6;
9f6200bb 5053 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
dd68314c
TT		 5054 if (!ext4_proc_root)
5055 goto out5;
857ac889
LC		 5056
5057 err = ext4_init_feat_adverts();
dd68314c
TT		 5058 if (err)
5059 goto out4;
857ac889 5060
5dabfc78 5061 err = ext4_init_mballoc();
ac27a0ec 5062 if (err)
6fd058f7 5063 goto out3;
c9de560d 5064
5dabfc78 5065 err = ext4_init_xattr();
c9de560d
AT		 5066 if (err)
5067 goto out2;
ac27a0ec
DK		 5068 err = init_inodecache();
5069 if (err)
5070 goto out1;
24b58424 5071 register_as_ext3();
2035e776 5072 register_as_ext2();
03010a33 5073 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5074 if (err)
5075 goto out;
bfff6873
LC		 5076
5077 ext4_li_info = NULL;
5078 mutex_init(&ext4_li_mtx);
ac27a0ec
DK		 5079 return 0;
5080out:
24b58424
TT		 5081 unregister_as_ext2();
5082 unregister_as_ext3();
ac27a0ec
DK		 5083 destroy_inodecache();
5084out1:
5dabfc78 5085 ext4_exit_xattr();
c9de560d 5086out2:
5dabfc78 5087 ext4_exit_mballoc();
6fd058f7 5088out3:
8f021222 5089 ext4_exit_feat_adverts();
dd68314c 5090out4:
6fd058f7 5091 remove_proc_entry("fs/ext4", NULL);
dd68314c 5092out5:
6fd058f7 5093 kset_unregister(ext4_kset);
dd68314c 5094out6:
5dabfc78 5095 ext4_exit_system_zone();
dd68314c 5096out7:
5dabfc78 5097 ext4_exit_pageio();
ac27a0ec
DK		 5098 return err;
5099}
5100
5dabfc78 5101static void __exit ext4_exit_fs(void)
ac27a0ec 5102{
bfff6873 5103 ext4_destroy_lazyinit_thread();
24b58424
TT		 5104 unregister_as_ext2();
5105 unregister_as_ext3();
03010a33 5106 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5107 destroy_inodecache();
5dabfc78
TT		 5108 ext4_exit_xattr();
5109 ext4_exit_mballoc();
8f021222 5110 ext4_exit_feat_adverts();
9f6200bb 5111 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5112 kset_unregister(ext4_kset);
5dabfc78
TT		 5113 ext4_exit_system_zone();
5114 ext4_exit_pageio();
ac27a0ec
DK		 5115}
5116
5117MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5118MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5119MODULE_LICENSE("GPL");
5dabfc78
TT		 5120module_init(ext4_init_fs)
5121module_exit(ext4_exit_fs)
ubuntu-bionic-kernel mirror