xref: /linux/fs/ext4/super.c (revision d89dffa976bcd13fd87eb76e02e3b71c3a7868e3)
1 /*
2  *  linux/fs/ext4/super.c
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>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
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>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
43 
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
46 
47 #include "ext4.h"
48 #include "ext4_extents.h"
49 #include "ext4_jbd2.h"
50 #include "xattr.h"
51 #include "acl.h"
52 #include "mballoc.h"
53 
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
56 
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
62 
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 			     unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 					struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 				   struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static const char *ext4_decode_error(struct super_block *sb, int errno,
73 				     char nbuf[16]);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 		       const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
89 	.owner		= THIS_MODULE,
90 	.name		= "ext2",
91 	.mount		= ext4_mount,
92 	.kill_sb	= kill_block_super,
93 	.fs_flags	= FS_REQUIRES_DEV,
94 };
95 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96 #else
97 #define IS_EXT2_SB(sb) (0)
98 #endif
99 
100 
101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102 static struct file_system_type ext3_fs_type = {
103 	.owner		= THIS_MODULE,
104 	.name		= "ext3",
105 	.mount		= ext4_mount,
106 	.kill_sb	= kill_block_super,
107 	.fs_flags	= FS_REQUIRES_DEV,
108 };
109 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110 #else
111 #define IS_EXT3_SB(sb) (0)
112 #endif
113 
114 static int ext4_verify_csum_type(struct super_block *sb,
115 				 struct ext4_super_block *es)
116 {
117 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
118 					EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
119 		return 1;
120 
121 	return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
122 }
123 
124 static __le32 ext4_superblock_csum(struct super_block *sb,
125 				   struct ext4_super_block *es)
126 {
127 	struct ext4_sb_info *sbi = EXT4_SB(sb);
128 	int offset = offsetof(struct ext4_super_block, s_checksum);
129 	__u32 csum;
130 
131 	csum = ext4_chksum(sbi, ~0, (char *)es, offset);
132 
133 	return cpu_to_le32(csum);
134 }
135 
136 int ext4_superblock_csum_verify(struct super_block *sb,
137 				struct ext4_super_block *es)
138 {
139 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
140 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
141 		return 1;
142 
143 	return es->s_checksum == ext4_superblock_csum(sb, es);
144 }
145 
146 void ext4_superblock_csum_set(struct super_block *sb,
147 			      struct ext4_super_block *es)
148 {
149 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
150 		EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
151 		return;
152 
153 	es->s_checksum = ext4_superblock_csum(sb, es);
154 }
155 
156 void *ext4_kvmalloc(size_t size, gfp_t flags)
157 {
158 	void *ret;
159 
160 	ret = kmalloc(size, flags);
161 	if (!ret)
162 		ret = __vmalloc(size, flags, PAGE_KERNEL);
163 	return ret;
164 }
165 
166 void *ext4_kvzalloc(size_t size, gfp_t flags)
167 {
168 	void *ret;
169 
170 	ret = kzalloc(size, flags);
171 	if (!ret)
172 		ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
173 	return ret;
174 }
175 
176 void ext4_kvfree(void *ptr)
177 {
178 	if (is_vmalloc_addr(ptr))
179 		vfree(ptr);
180 	else
181 		kfree(ptr);
182 
183 }
184 
185 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
186 			       struct ext4_group_desc *bg)
187 {
188 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
189 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
190 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
191 }
192 
193 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
194 			       struct ext4_group_desc *bg)
195 {
196 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
197 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
198 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
199 }
200 
201 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
202 			      struct ext4_group_desc *bg)
203 {
204 	return le32_to_cpu(bg->bg_inode_table_lo) |
205 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
206 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
207 }
208 
209 __u32 ext4_free_group_clusters(struct super_block *sb,
210 			       struct ext4_group_desc *bg)
211 {
212 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
213 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
214 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
215 }
216 
217 __u32 ext4_free_inodes_count(struct super_block *sb,
218 			      struct ext4_group_desc *bg)
219 {
220 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
221 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
222 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
223 }
224 
225 __u32 ext4_used_dirs_count(struct super_block *sb,
226 			      struct ext4_group_desc *bg)
227 {
228 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
229 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
230 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
231 }
232 
233 __u32 ext4_itable_unused_count(struct super_block *sb,
234 			      struct ext4_group_desc *bg)
235 {
236 	return le16_to_cpu(bg->bg_itable_unused_lo) |
237 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
238 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
239 }
240 
241 void ext4_block_bitmap_set(struct super_block *sb,
242 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
243 {
244 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
245 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
246 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
247 }
248 
249 void ext4_inode_bitmap_set(struct super_block *sb,
250 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
251 {
252 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
253 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
254 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
255 }
256 
257 void ext4_inode_table_set(struct super_block *sb,
258 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
259 {
260 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
261 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
263 }
264 
265 void ext4_free_group_clusters_set(struct super_block *sb,
266 				  struct ext4_group_desc *bg, __u32 count)
267 {
268 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
269 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
271 }
272 
273 void ext4_free_inodes_set(struct super_block *sb,
274 			  struct ext4_group_desc *bg, __u32 count)
275 {
276 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
277 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
279 }
280 
281 void ext4_used_dirs_set(struct super_block *sb,
282 			  struct ext4_group_desc *bg, __u32 count)
283 {
284 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
285 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
287 }
288 
289 void ext4_itable_unused_set(struct super_block *sb,
290 			  struct ext4_group_desc *bg, __u32 count)
291 {
292 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
293 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
295 }
296 
297 
298 /* Just increment the non-pointer handle value */
299 static handle_t *ext4_get_nojournal(void)
300 {
301 	handle_t *handle = current->journal_info;
302 	unsigned long ref_cnt = (unsigned long)handle;
303 
304 	BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
305 
306 	ref_cnt++;
307 	handle = (handle_t *)ref_cnt;
308 
309 	current->journal_info = handle;
310 	return handle;
311 }
312 
313 
314 /* Decrement the non-pointer handle value */
315 static void ext4_put_nojournal(handle_t *handle)
316 {
317 	unsigned long ref_cnt = (unsigned long)handle;
318 
319 	BUG_ON(ref_cnt == 0);
320 
321 	ref_cnt--;
322 	handle = (handle_t *)ref_cnt;
323 
324 	current->journal_info = handle;
325 }
326 
327 /*
328  * Wrappers for jbd2_journal_start/end.
329  *
330  * The only special thing we need to do here is to make sure that all
331  * journal_end calls result in the superblock being marked dirty, so
332  * that sync() will call the filesystem's write_super callback if
333  * appropriate.
334  *
335  * To avoid j_barrier hold in userspace when a user calls freeze(),
336  * ext4 prevents a new handle from being started by s_frozen, which
337  * is in an upper layer.
338  */
339 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
340 {
341 	journal_t *journal;
342 	handle_t  *handle;
343 
344 	trace_ext4_journal_start(sb, nblocks, _RET_IP_);
345 	if (sb->s_flags & MS_RDONLY)
346 		return ERR_PTR(-EROFS);
347 
348 	journal = EXT4_SB(sb)->s_journal;
349 	handle = ext4_journal_current_handle();
350 
351 	/*
352 	 * If a handle has been started, it should be allowed to
353 	 * finish, otherwise deadlock could happen between freeze
354 	 * and others(e.g. truncate) due to the restart of the
355 	 * journal handle if the filesystem is forzen and active
356 	 * handles are not stopped.
357 	 */
358 	if (!handle)
359 		vfs_check_frozen(sb, SB_FREEZE_TRANS);
360 
361 	if (!journal)
362 		return ext4_get_nojournal();
363 	/*
364 	 * Special case here: if the journal has aborted behind our
365 	 * backs (eg. EIO in the commit thread), then we still need to
366 	 * take the FS itself readonly cleanly.
367 	 */
368 	if (is_journal_aborted(journal)) {
369 		ext4_abort(sb, "Detected aborted journal");
370 		return ERR_PTR(-EROFS);
371 	}
372 	return jbd2_journal_start(journal, nblocks);
373 }
374 
375 /*
376  * The only special thing we need to do here is to make sure that all
377  * jbd2_journal_stop calls result in the superblock being marked dirty, so
378  * that sync() will call the filesystem's write_super callback if
379  * appropriate.
380  */
381 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
382 {
383 	struct super_block *sb;
384 	int err;
385 	int rc;
386 
387 	if (!ext4_handle_valid(handle)) {
388 		ext4_put_nojournal(handle);
389 		return 0;
390 	}
391 	sb = handle->h_transaction->t_journal->j_private;
392 	err = handle->h_err;
393 	rc = jbd2_journal_stop(handle);
394 
395 	if (!err)
396 		err = rc;
397 	if (err)
398 		__ext4_std_error(sb, where, line, err);
399 	return err;
400 }
401 
402 void ext4_journal_abort_handle(const char *caller, unsigned int line,
403 			       const char *err_fn, struct buffer_head *bh,
404 			       handle_t *handle, int err)
405 {
406 	char nbuf[16];
407 	const char *errstr = ext4_decode_error(NULL, err, nbuf);
408 
409 	BUG_ON(!ext4_handle_valid(handle));
410 
411 	if (bh)
412 		BUFFER_TRACE(bh, "abort");
413 
414 	if (!handle->h_err)
415 		handle->h_err = err;
416 
417 	if (is_handle_aborted(handle))
418 		return;
419 
420 	printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
421 	       caller, line, errstr, err_fn);
422 
423 	jbd2_journal_abort_handle(handle);
424 }
425 
426 static void __save_error_info(struct super_block *sb, const char *func,
427 			    unsigned int line)
428 {
429 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
430 
431 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
432 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
433 	es->s_last_error_time = cpu_to_le32(get_seconds());
434 	strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
435 	es->s_last_error_line = cpu_to_le32(line);
436 	if (!es->s_first_error_time) {
437 		es->s_first_error_time = es->s_last_error_time;
438 		strncpy(es->s_first_error_func, func,
439 			sizeof(es->s_first_error_func));
440 		es->s_first_error_line = cpu_to_le32(line);
441 		es->s_first_error_ino = es->s_last_error_ino;
442 		es->s_first_error_block = es->s_last_error_block;
443 	}
444 	/*
445 	 * Start the daily error reporting function if it hasn't been
446 	 * started already
447 	 */
448 	if (!es->s_error_count)
449 		mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
450 	es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
451 }
452 
453 static void save_error_info(struct super_block *sb, const char *func,
454 			    unsigned int line)
455 {
456 	__save_error_info(sb, func, line);
457 	ext4_commit_super(sb, 1);
458 }
459 
460 /*
461  * The del_gendisk() function uninitializes the disk-specific data
462  * structures, including the bdi structure, without telling anyone
463  * else.  Once this happens, any attempt to call mark_buffer_dirty()
464  * (for example, by ext4_commit_super), will cause a kernel OOPS.
465  * This is a kludge to prevent these oops until we can put in a proper
466  * hook in del_gendisk() to inform the VFS and file system layers.
467  */
468 static int block_device_ejected(struct super_block *sb)
469 {
470 	struct inode *bd_inode = sb->s_bdev->bd_inode;
471 	struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
472 
473 	return bdi->dev == NULL;
474 }
475 
476 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
477 {
478 	struct super_block		*sb = journal->j_private;
479 	struct ext4_sb_info		*sbi = EXT4_SB(sb);
480 	int				error = is_journal_aborted(journal);
481 	struct ext4_journal_cb_entry	*jce, *tmp;
482 
483 	spin_lock(&sbi->s_md_lock);
484 	list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
485 		list_del_init(&jce->jce_list);
486 		spin_unlock(&sbi->s_md_lock);
487 		jce->jce_func(sb, jce, error);
488 		spin_lock(&sbi->s_md_lock);
489 	}
490 	spin_unlock(&sbi->s_md_lock);
491 }
492 
493 /* Deal with the reporting of failure conditions on a filesystem such as
494  * inconsistencies detected or read IO failures.
495  *
496  * On ext2, we can store the error state of the filesystem in the
497  * superblock.  That is not possible on ext4, because we may have other
498  * write ordering constraints on the superblock which prevent us from
499  * writing it out straight away; and given that the journal is about to
500  * be aborted, we can't rely on the current, or future, transactions to
501  * write out the superblock safely.
502  *
503  * We'll just use the jbd2_journal_abort() error code to record an error in
504  * the journal instead.  On recovery, the journal will complain about
505  * that error until we've noted it down and cleared it.
506  */
507 
508 static void ext4_handle_error(struct super_block *sb)
509 {
510 	if (sb->s_flags & MS_RDONLY)
511 		return;
512 
513 	if (!test_opt(sb, ERRORS_CONT)) {
514 		journal_t *journal = EXT4_SB(sb)->s_journal;
515 
516 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
517 		if (journal)
518 			jbd2_journal_abort(journal, -EIO);
519 	}
520 	if (test_opt(sb, ERRORS_RO)) {
521 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
522 		sb->s_flags |= MS_RDONLY;
523 	}
524 	if (test_opt(sb, ERRORS_PANIC))
525 		panic("EXT4-fs (device %s): panic forced after error\n",
526 			sb->s_id);
527 }
528 
529 void __ext4_error(struct super_block *sb, const char *function,
530 		  unsigned int line, const char *fmt, ...)
531 {
532 	struct va_format vaf;
533 	va_list args;
534 
535 	va_start(args, fmt);
536 	vaf.fmt = fmt;
537 	vaf.va = &args;
538 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
539 	       sb->s_id, function, line, current->comm, &vaf);
540 	va_end(args);
541 	save_error_info(sb, function, line);
542 
543 	ext4_handle_error(sb);
544 }
545 
546 void ext4_error_inode(struct inode *inode, const char *function,
547 		      unsigned int line, ext4_fsblk_t block,
548 		      const char *fmt, ...)
549 {
550 	va_list args;
551 	struct va_format vaf;
552 	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
553 
554 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
555 	es->s_last_error_block = cpu_to_le64(block);
556 	save_error_info(inode->i_sb, function, line);
557 	va_start(args, fmt);
558 	vaf.fmt = fmt;
559 	vaf.va = &args;
560 	if (block)
561 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
562 		       "inode #%lu: block %llu: comm %s: %pV\n",
563 		       inode->i_sb->s_id, function, line, inode->i_ino,
564 		       block, current->comm, &vaf);
565 	else
566 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
567 		       "inode #%lu: comm %s: %pV\n",
568 		       inode->i_sb->s_id, function, line, inode->i_ino,
569 		       current->comm, &vaf);
570 	va_end(args);
571 
572 	ext4_handle_error(inode->i_sb);
573 }
574 
575 void ext4_error_file(struct file *file, const char *function,
576 		     unsigned int line, ext4_fsblk_t block,
577 		     const char *fmt, ...)
578 {
579 	va_list args;
580 	struct va_format vaf;
581 	struct ext4_super_block *es;
582 	struct inode *inode = file->f_dentry->d_inode;
583 	char pathname[80], *path;
584 
585 	es = EXT4_SB(inode->i_sb)->s_es;
586 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
587 	save_error_info(inode->i_sb, function, line);
588 	path = d_path(&(file->f_path), pathname, sizeof(pathname));
589 	if (IS_ERR(path))
590 		path = "(unknown)";
591 	va_start(args, fmt);
592 	vaf.fmt = fmt;
593 	vaf.va = &args;
594 	if (block)
595 		printk(KERN_CRIT
596 		       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
597 		       "block %llu: comm %s: path %s: %pV\n",
598 		       inode->i_sb->s_id, function, line, inode->i_ino,
599 		       block, current->comm, path, &vaf);
600 	else
601 		printk(KERN_CRIT
602 		       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
603 		       "comm %s: path %s: %pV\n",
604 		       inode->i_sb->s_id, function, line, inode->i_ino,
605 		       current->comm, path, &vaf);
606 	va_end(args);
607 
608 	ext4_handle_error(inode->i_sb);
609 }
610 
611 static const char *ext4_decode_error(struct super_block *sb, int errno,
612 				     char nbuf[16])
613 {
614 	char *errstr = NULL;
615 
616 	switch (errno) {
617 	case -EIO:
618 		errstr = "IO failure";
619 		break;
620 	case -ENOMEM:
621 		errstr = "Out of memory";
622 		break;
623 	case -EROFS:
624 		if (!sb || (EXT4_SB(sb)->s_journal &&
625 			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
626 			errstr = "Journal has aborted";
627 		else
628 			errstr = "Readonly filesystem";
629 		break;
630 	default:
631 		/* If the caller passed in an extra buffer for unknown
632 		 * errors, textualise them now.  Else we just return
633 		 * NULL. */
634 		if (nbuf) {
635 			/* Check for truncated error codes... */
636 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
637 				errstr = nbuf;
638 		}
639 		break;
640 	}
641 
642 	return errstr;
643 }
644 
645 /* __ext4_std_error decodes expected errors from journaling functions
646  * automatically and invokes the appropriate error response.  */
647 
648 void __ext4_std_error(struct super_block *sb, const char *function,
649 		      unsigned int line, int errno)
650 {
651 	char nbuf[16];
652 	const char *errstr;
653 
654 	/* Special case: if the error is EROFS, and we're not already
655 	 * inside a transaction, then there's really no point in logging
656 	 * an error. */
657 	if (errno == -EROFS && journal_current_handle() == NULL &&
658 	    (sb->s_flags & MS_RDONLY))
659 		return;
660 
661 	errstr = ext4_decode_error(sb, errno, nbuf);
662 	printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
663 	       sb->s_id, function, line, errstr);
664 	save_error_info(sb, function, line);
665 
666 	ext4_handle_error(sb);
667 }
668 
669 /*
670  * ext4_abort is a much stronger failure handler than ext4_error.  The
671  * abort function may be used to deal with unrecoverable failures such
672  * as journal IO errors or ENOMEM at a critical moment in log management.
673  *
674  * We unconditionally force the filesystem into an ABORT|READONLY state,
675  * unless the error response on the fs has been set to panic in which
676  * case we take the easy way out and panic immediately.
677  */
678 
679 void __ext4_abort(struct super_block *sb, const char *function,
680 		unsigned int line, const char *fmt, ...)
681 {
682 	va_list args;
683 
684 	save_error_info(sb, function, line);
685 	va_start(args, fmt);
686 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
687 	       function, line);
688 	vprintk(fmt, args);
689 	printk("\n");
690 	va_end(args);
691 
692 	if ((sb->s_flags & MS_RDONLY) == 0) {
693 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
694 		sb->s_flags |= MS_RDONLY;
695 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
696 		if (EXT4_SB(sb)->s_journal)
697 			jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
698 		save_error_info(sb, function, line);
699 	}
700 	if (test_opt(sb, ERRORS_PANIC))
701 		panic("EXT4-fs panic from previous error\n");
702 }
703 
704 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
705 {
706 	struct va_format vaf;
707 	va_list args;
708 
709 	va_start(args, fmt);
710 	vaf.fmt = fmt;
711 	vaf.va = &args;
712 	printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
713 	va_end(args);
714 }
715 
716 void __ext4_warning(struct super_block *sb, const char *function,
717 		    unsigned int line, const char *fmt, ...)
718 {
719 	struct va_format vaf;
720 	va_list args;
721 
722 	va_start(args, fmt);
723 	vaf.fmt = fmt;
724 	vaf.va = &args;
725 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
726 	       sb->s_id, function, line, &vaf);
727 	va_end(args);
728 }
729 
730 void __ext4_grp_locked_error(const char *function, unsigned int line,
731 			     struct super_block *sb, ext4_group_t grp,
732 			     unsigned long ino, ext4_fsblk_t block,
733 			     const char *fmt, ...)
734 __releases(bitlock)
735 __acquires(bitlock)
736 {
737 	struct va_format vaf;
738 	va_list args;
739 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
740 
741 	es->s_last_error_ino = cpu_to_le32(ino);
742 	es->s_last_error_block = cpu_to_le64(block);
743 	__save_error_info(sb, function, line);
744 
745 	va_start(args, fmt);
746 
747 	vaf.fmt = fmt;
748 	vaf.va = &args;
749 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
750 	       sb->s_id, function, line, grp);
751 	if (ino)
752 		printk(KERN_CONT "inode %lu: ", ino);
753 	if (block)
754 		printk(KERN_CONT "block %llu:", (unsigned long long) block);
755 	printk(KERN_CONT "%pV\n", &vaf);
756 	va_end(args);
757 
758 	if (test_opt(sb, ERRORS_CONT)) {
759 		ext4_commit_super(sb, 0);
760 		return;
761 	}
762 
763 	ext4_unlock_group(sb, grp);
764 	ext4_handle_error(sb);
765 	/*
766 	 * We only get here in the ERRORS_RO case; relocking the group
767 	 * may be dangerous, but nothing bad will happen since the
768 	 * filesystem will have already been marked read/only and the
769 	 * journal has been aborted.  We return 1 as a hint to callers
770 	 * who might what to use the return value from
771 	 * ext4_grp_locked_error() to distinguish between the
772 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
773 	 * aggressively from the ext4 function in question, with a
774 	 * more appropriate error code.
775 	 */
776 	ext4_lock_group(sb, grp);
777 	return;
778 }
779 
780 void ext4_update_dynamic_rev(struct super_block *sb)
781 {
782 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
783 
784 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
785 		return;
786 
787 	ext4_warning(sb,
788 		     "updating to rev %d because of new feature flag, "
789 		     "running e2fsck is recommended",
790 		     EXT4_DYNAMIC_REV);
791 
792 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
793 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
794 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
795 	/* leave es->s_feature_*compat flags alone */
796 	/* es->s_uuid will be set by e2fsck if empty */
797 
798 	/*
799 	 * The rest of the superblock fields should be zero, and if not it
800 	 * means they are likely already in use, so leave them alone.  We
801 	 * can leave it up to e2fsck to clean up any inconsistencies there.
802 	 */
803 }
804 
805 /*
806  * Open the external journal device
807  */
808 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
809 {
810 	struct block_device *bdev;
811 	char b[BDEVNAME_SIZE];
812 
813 	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
814 	if (IS_ERR(bdev))
815 		goto fail;
816 	return bdev;
817 
818 fail:
819 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
820 			__bdevname(dev, b), PTR_ERR(bdev));
821 	return NULL;
822 }
823 
824 /*
825  * Release the journal device
826  */
827 static int ext4_blkdev_put(struct block_device *bdev)
828 {
829 	return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
830 }
831 
832 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
833 {
834 	struct block_device *bdev;
835 	int ret = -ENODEV;
836 
837 	bdev = sbi->journal_bdev;
838 	if (bdev) {
839 		ret = ext4_blkdev_put(bdev);
840 		sbi->journal_bdev = NULL;
841 	}
842 	return ret;
843 }
844 
845 static inline struct inode *orphan_list_entry(struct list_head *l)
846 {
847 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
848 }
849 
850 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
851 {
852 	struct list_head *l;
853 
854 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
855 		 le32_to_cpu(sbi->s_es->s_last_orphan));
856 
857 	printk(KERN_ERR "sb_info orphan list:\n");
858 	list_for_each(l, &sbi->s_orphan) {
859 		struct inode *inode = orphan_list_entry(l);
860 		printk(KERN_ERR "  "
861 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
862 		       inode->i_sb->s_id, inode->i_ino, inode,
863 		       inode->i_mode, inode->i_nlink,
864 		       NEXT_ORPHAN(inode));
865 	}
866 }
867 
868 static void ext4_put_super(struct super_block *sb)
869 {
870 	struct ext4_sb_info *sbi = EXT4_SB(sb);
871 	struct ext4_super_block *es = sbi->s_es;
872 	int i, err;
873 
874 	ext4_unregister_li_request(sb);
875 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
876 
877 	flush_workqueue(sbi->dio_unwritten_wq);
878 	destroy_workqueue(sbi->dio_unwritten_wq);
879 
880 	lock_super(sb);
881 	if (sbi->s_journal) {
882 		err = jbd2_journal_destroy(sbi->s_journal);
883 		sbi->s_journal = NULL;
884 		if (err < 0)
885 			ext4_abort(sb, "Couldn't clean up the journal");
886 	}
887 
888 	del_timer(&sbi->s_err_report);
889 	ext4_release_system_zone(sb);
890 	ext4_mb_release(sb);
891 	ext4_ext_release(sb);
892 	ext4_xattr_put_super(sb);
893 
894 	if (!(sb->s_flags & MS_RDONLY)) {
895 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
896 		es->s_state = cpu_to_le16(sbi->s_mount_state);
897 	}
898 	if (!(sb->s_flags & MS_RDONLY))
899 		ext4_commit_super(sb, 1);
900 
901 	if (sbi->s_proc) {
902 		remove_proc_entry("options", sbi->s_proc);
903 		remove_proc_entry(sb->s_id, ext4_proc_root);
904 	}
905 	kobject_del(&sbi->s_kobj);
906 
907 	for (i = 0; i < sbi->s_gdb_count; i++)
908 		brelse(sbi->s_group_desc[i]);
909 	ext4_kvfree(sbi->s_group_desc);
910 	ext4_kvfree(sbi->s_flex_groups);
911 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
912 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
913 	percpu_counter_destroy(&sbi->s_dirs_counter);
914 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
915 	brelse(sbi->s_sbh);
916 #ifdef CONFIG_QUOTA
917 	for (i = 0; i < MAXQUOTAS; i++)
918 		kfree(sbi->s_qf_names[i]);
919 #endif
920 
921 	/* Debugging code just in case the in-memory inode orphan list
922 	 * isn't empty.  The on-disk one can be non-empty if we've
923 	 * detected an error and taken the fs readonly, but the
924 	 * in-memory list had better be clean by this point. */
925 	if (!list_empty(&sbi->s_orphan))
926 		dump_orphan_list(sb, sbi);
927 	J_ASSERT(list_empty(&sbi->s_orphan));
928 
929 	invalidate_bdev(sb->s_bdev);
930 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
931 		/*
932 		 * Invalidate the journal device's buffers.  We don't want them
933 		 * floating about in memory - the physical journal device may
934 		 * hotswapped, and it breaks the `ro-after' testing code.
935 		 */
936 		sync_blockdev(sbi->journal_bdev);
937 		invalidate_bdev(sbi->journal_bdev);
938 		ext4_blkdev_remove(sbi);
939 	}
940 	if (sbi->s_mmp_tsk)
941 		kthread_stop(sbi->s_mmp_tsk);
942 	sb->s_fs_info = NULL;
943 	/*
944 	 * Now that we are completely done shutting down the
945 	 * superblock, we need to actually destroy the kobject.
946 	 */
947 	unlock_super(sb);
948 	kobject_put(&sbi->s_kobj);
949 	wait_for_completion(&sbi->s_kobj_unregister);
950 	if (sbi->s_chksum_driver)
951 		crypto_free_shash(sbi->s_chksum_driver);
952 	kfree(sbi->s_blockgroup_lock);
953 	kfree(sbi);
954 }
955 
956 static struct kmem_cache *ext4_inode_cachep;
957 
958 /*
959  * Called inside transaction, so use GFP_NOFS
960  */
961 static struct inode *ext4_alloc_inode(struct super_block *sb)
962 {
963 	struct ext4_inode_info *ei;
964 
965 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
966 	if (!ei)
967 		return NULL;
968 
969 	ei->vfs_inode.i_version = 1;
970 	ei->vfs_inode.i_data.writeback_index = 0;
971 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
972 	INIT_LIST_HEAD(&ei->i_prealloc_list);
973 	spin_lock_init(&ei->i_prealloc_lock);
974 	ei->i_reserved_data_blocks = 0;
975 	ei->i_reserved_meta_blocks = 0;
976 	ei->i_allocated_meta_blocks = 0;
977 	ei->i_da_metadata_calc_len = 0;
978 	spin_lock_init(&(ei->i_block_reservation_lock));
979 #ifdef CONFIG_QUOTA
980 	ei->i_reserved_quota = 0;
981 #endif
982 	ei->jinode = NULL;
983 	INIT_LIST_HEAD(&ei->i_completed_io_list);
984 	spin_lock_init(&ei->i_completed_io_lock);
985 	ei->cur_aio_dio = NULL;
986 	ei->i_sync_tid = 0;
987 	ei->i_datasync_tid = 0;
988 	atomic_set(&ei->i_ioend_count, 0);
989 	atomic_set(&ei->i_aiodio_unwritten, 0);
990 
991 	return &ei->vfs_inode;
992 }
993 
994 static int ext4_drop_inode(struct inode *inode)
995 {
996 	int drop = generic_drop_inode(inode);
997 
998 	trace_ext4_drop_inode(inode, drop);
999 	return drop;
1000 }
1001 
1002 static void ext4_i_callback(struct rcu_head *head)
1003 {
1004 	struct inode *inode = container_of(head, struct inode, i_rcu);
1005 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1006 }
1007 
1008 static void ext4_destroy_inode(struct inode *inode)
1009 {
1010 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1011 		ext4_msg(inode->i_sb, KERN_ERR,
1012 			 "Inode %lu (%p): orphan list check failed!",
1013 			 inode->i_ino, EXT4_I(inode));
1014 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1015 				EXT4_I(inode), sizeof(struct ext4_inode_info),
1016 				true);
1017 		dump_stack();
1018 	}
1019 	call_rcu(&inode->i_rcu, ext4_i_callback);
1020 }
1021 
1022 static void init_once(void *foo)
1023 {
1024 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1025 
1026 	INIT_LIST_HEAD(&ei->i_orphan);
1027 #ifdef CONFIG_EXT4_FS_XATTR
1028 	init_rwsem(&ei->xattr_sem);
1029 #endif
1030 	init_rwsem(&ei->i_data_sem);
1031 	inode_init_once(&ei->vfs_inode);
1032 }
1033 
1034 static int init_inodecache(void)
1035 {
1036 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1037 					     sizeof(struct ext4_inode_info),
1038 					     0, (SLAB_RECLAIM_ACCOUNT|
1039 						SLAB_MEM_SPREAD),
1040 					     init_once);
1041 	if (ext4_inode_cachep == NULL)
1042 		return -ENOMEM;
1043 	return 0;
1044 }
1045 
1046 static void destroy_inodecache(void)
1047 {
1048 	kmem_cache_destroy(ext4_inode_cachep);
1049 }
1050 
1051 void ext4_clear_inode(struct inode *inode)
1052 {
1053 	invalidate_inode_buffers(inode);
1054 	clear_inode(inode);
1055 	dquot_drop(inode);
1056 	ext4_discard_preallocations(inode);
1057 	if (EXT4_I(inode)->jinode) {
1058 		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1059 					       EXT4_I(inode)->jinode);
1060 		jbd2_free_inode(EXT4_I(inode)->jinode);
1061 		EXT4_I(inode)->jinode = NULL;
1062 	}
1063 }
1064 
1065 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1066 					u64 ino, u32 generation)
1067 {
1068 	struct inode *inode;
1069 
1070 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1071 		return ERR_PTR(-ESTALE);
1072 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1073 		return ERR_PTR(-ESTALE);
1074 
1075 	/* iget isn't really right if the inode is currently unallocated!!
1076 	 *
1077 	 * ext4_read_inode will return a bad_inode if the inode had been
1078 	 * deleted, so we should be safe.
1079 	 *
1080 	 * Currently we don't know the generation for parent directory, so
1081 	 * a generation of 0 means "accept any"
1082 	 */
1083 	inode = ext4_iget(sb, ino);
1084 	if (IS_ERR(inode))
1085 		return ERR_CAST(inode);
1086 	if (generation && inode->i_generation != generation) {
1087 		iput(inode);
1088 		return ERR_PTR(-ESTALE);
1089 	}
1090 
1091 	return inode;
1092 }
1093 
1094 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1095 					int fh_len, int fh_type)
1096 {
1097 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1098 				    ext4_nfs_get_inode);
1099 }
1100 
1101 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1102 					int fh_len, int fh_type)
1103 {
1104 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1105 				    ext4_nfs_get_inode);
1106 }
1107 
1108 /*
1109  * Try to release metadata pages (indirect blocks, directories) which are
1110  * mapped via the block device.  Since these pages could have journal heads
1111  * which would prevent try_to_free_buffers() from freeing them, we must use
1112  * jbd2 layer's try_to_free_buffers() function to release them.
1113  */
1114 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1115 				 gfp_t wait)
1116 {
1117 	journal_t *journal = EXT4_SB(sb)->s_journal;
1118 
1119 	WARN_ON(PageChecked(page));
1120 	if (!page_has_buffers(page))
1121 		return 0;
1122 	if (journal)
1123 		return jbd2_journal_try_to_free_buffers(journal, page,
1124 							wait & ~__GFP_WAIT);
1125 	return try_to_free_buffers(page);
1126 }
1127 
1128 #ifdef CONFIG_QUOTA
1129 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1130 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1131 
1132 static int ext4_write_dquot(struct dquot *dquot);
1133 static int ext4_acquire_dquot(struct dquot *dquot);
1134 static int ext4_release_dquot(struct dquot *dquot);
1135 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1136 static int ext4_write_info(struct super_block *sb, int type);
1137 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1138 			 struct path *path);
1139 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1140 				 int format_id);
1141 static int ext4_quota_off(struct super_block *sb, int type);
1142 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1143 static int ext4_quota_on_mount(struct super_block *sb, int type);
1144 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1145 			       size_t len, loff_t off);
1146 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1147 				const char *data, size_t len, loff_t off);
1148 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1149 			     unsigned int flags);
1150 static int ext4_enable_quotas(struct super_block *sb);
1151 
1152 static const struct dquot_operations ext4_quota_operations = {
1153 	.get_reserved_space = ext4_get_reserved_space,
1154 	.write_dquot	= ext4_write_dquot,
1155 	.acquire_dquot	= ext4_acquire_dquot,
1156 	.release_dquot	= ext4_release_dquot,
1157 	.mark_dirty	= ext4_mark_dquot_dirty,
1158 	.write_info	= ext4_write_info,
1159 	.alloc_dquot	= dquot_alloc,
1160 	.destroy_dquot	= dquot_destroy,
1161 };
1162 
1163 static const struct quotactl_ops ext4_qctl_operations = {
1164 	.quota_on	= ext4_quota_on,
1165 	.quota_off	= ext4_quota_off,
1166 	.quota_sync	= dquot_quota_sync,
1167 	.get_info	= dquot_get_dqinfo,
1168 	.set_info	= dquot_set_dqinfo,
1169 	.get_dqblk	= dquot_get_dqblk,
1170 	.set_dqblk	= dquot_set_dqblk
1171 };
1172 
1173 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1174 	.quota_on_meta	= ext4_quota_on_sysfile,
1175 	.quota_off	= ext4_quota_off_sysfile,
1176 	.quota_sync	= dquot_quota_sync,
1177 	.get_info	= dquot_get_dqinfo,
1178 	.set_info	= dquot_set_dqinfo,
1179 	.get_dqblk	= dquot_get_dqblk,
1180 	.set_dqblk	= dquot_set_dqblk
1181 };
1182 #endif
1183 
1184 static const struct super_operations ext4_sops = {
1185 	.alloc_inode	= ext4_alloc_inode,
1186 	.destroy_inode	= ext4_destroy_inode,
1187 	.write_inode	= ext4_write_inode,
1188 	.dirty_inode	= ext4_dirty_inode,
1189 	.drop_inode	= ext4_drop_inode,
1190 	.evict_inode	= ext4_evict_inode,
1191 	.put_super	= ext4_put_super,
1192 	.sync_fs	= ext4_sync_fs,
1193 	.freeze_fs	= ext4_freeze,
1194 	.unfreeze_fs	= ext4_unfreeze,
1195 	.statfs		= ext4_statfs,
1196 	.remount_fs	= ext4_remount,
1197 	.show_options	= ext4_show_options,
1198 #ifdef CONFIG_QUOTA
1199 	.quota_read	= ext4_quota_read,
1200 	.quota_write	= ext4_quota_write,
1201 #endif
1202 	.bdev_try_to_free_page = bdev_try_to_free_page,
1203 };
1204 
1205 static const struct super_operations ext4_nojournal_sops = {
1206 	.alloc_inode	= ext4_alloc_inode,
1207 	.destroy_inode	= ext4_destroy_inode,
1208 	.write_inode	= ext4_write_inode,
1209 	.dirty_inode	= ext4_dirty_inode,
1210 	.drop_inode	= ext4_drop_inode,
1211 	.evict_inode	= ext4_evict_inode,
1212 	.put_super	= ext4_put_super,
1213 	.statfs		= ext4_statfs,
1214 	.remount_fs	= ext4_remount,
1215 	.show_options	= ext4_show_options,
1216 #ifdef CONFIG_QUOTA
1217 	.quota_read	= ext4_quota_read,
1218 	.quota_write	= ext4_quota_write,
1219 #endif
1220 	.bdev_try_to_free_page = bdev_try_to_free_page,
1221 };
1222 
1223 static const struct export_operations ext4_export_ops = {
1224 	.fh_to_dentry = ext4_fh_to_dentry,
1225 	.fh_to_parent = ext4_fh_to_parent,
1226 	.get_parent = ext4_get_parent,
1227 };
1228 
1229 enum {
1230 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1231 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1232 	Opt_nouid32, Opt_debug, Opt_removed,
1233 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1234 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1235 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1236 	Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1237 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1238 	Opt_data_err_abort, Opt_data_err_ignore,
1239 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1240 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1241 	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1242 	Opt_usrquota, Opt_grpquota, Opt_i_version,
1243 	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1244 	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1245 	Opt_inode_readahead_blks, Opt_journal_ioprio,
1246 	Opt_dioread_nolock, Opt_dioread_lock,
1247 	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1248 };
1249 
1250 static const match_table_t tokens = {
1251 	{Opt_bsd_df, "bsddf"},
1252 	{Opt_minix_df, "minixdf"},
1253 	{Opt_grpid, "grpid"},
1254 	{Opt_grpid, "bsdgroups"},
1255 	{Opt_nogrpid, "nogrpid"},
1256 	{Opt_nogrpid, "sysvgroups"},
1257 	{Opt_resgid, "resgid=%u"},
1258 	{Opt_resuid, "resuid=%u"},
1259 	{Opt_sb, "sb=%u"},
1260 	{Opt_err_cont, "errors=continue"},
1261 	{Opt_err_panic, "errors=panic"},
1262 	{Opt_err_ro, "errors=remount-ro"},
1263 	{Opt_nouid32, "nouid32"},
1264 	{Opt_debug, "debug"},
1265 	{Opt_removed, "oldalloc"},
1266 	{Opt_removed, "orlov"},
1267 	{Opt_user_xattr, "user_xattr"},
1268 	{Opt_nouser_xattr, "nouser_xattr"},
1269 	{Opt_acl, "acl"},
1270 	{Opt_noacl, "noacl"},
1271 	{Opt_noload, "norecovery"},
1272 	{Opt_noload, "noload"},
1273 	{Opt_removed, "nobh"},
1274 	{Opt_removed, "bh"},
1275 	{Opt_commit, "commit=%u"},
1276 	{Opt_min_batch_time, "min_batch_time=%u"},
1277 	{Opt_max_batch_time, "max_batch_time=%u"},
1278 	{Opt_journal_dev, "journal_dev=%u"},
1279 	{Opt_journal_checksum, "journal_checksum"},
1280 	{Opt_journal_async_commit, "journal_async_commit"},
1281 	{Opt_abort, "abort"},
1282 	{Opt_data_journal, "data=journal"},
1283 	{Opt_data_ordered, "data=ordered"},
1284 	{Opt_data_writeback, "data=writeback"},
1285 	{Opt_data_err_abort, "data_err=abort"},
1286 	{Opt_data_err_ignore, "data_err=ignore"},
1287 	{Opt_offusrjquota, "usrjquota="},
1288 	{Opt_usrjquota, "usrjquota=%s"},
1289 	{Opt_offgrpjquota, "grpjquota="},
1290 	{Opt_grpjquota, "grpjquota=%s"},
1291 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1292 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1293 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1294 	{Opt_grpquota, "grpquota"},
1295 	{Opt_noquota, "noquota"},
1296 	{Opt_quota, "quota"},
1297 	{Opt_usrquota, "usrquota"},
1298 	{Opt_barrier, "barrier=%u"},
1299 	{Opt_barrier, "barrier"},
1300 	{Opt_nobarrier, "nobarrier"},
1301 	{Opt_i_version, "i_version"},
1302 	{Opt_stripe, "stripe=%u"},
1303 	{Opt_delalloc, "delalloc"},
1304 	{Opt_nodelalloc, "nodelalloc"},
1305 	{Opt_mblk_io_submit, "mblk_io_submit"},
1306 	{Opt_nomblk_io_submit, "nomblk_io_submit"},
1307 	{Opt_block_validity, "block_validity"},
1308 	{Opt_noblock_validity, "noblock_validity"},
1309 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1310 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1311 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1312 	{Opt_auto_da_alloc, "auto_da_alloc"},
1313 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1314 	{Opt_dioread_nolock, "dioread_nolock"},
1315 	{Opt_dioread_lock, "dioread_lock"},
1316 	{Opt_discard, "discard"},
1317 	{Opt_nodiscard, "nodiscard"},
1318 	{Opt_init_itable, "init_itable=%u"},
1319 	{Opt_init_itable, "init_itable"},
1320 	{Opt_noinit_itable, "noinit_itable"},
1321 	{Opt_removed, "check=none"},	/* mount option from ext2/3 */
1322 	{Opt_removed, "nocheck"},	/* mount option from ext2/3 */
1323 	{Opt_removed, "reservation"},	/* mount option from ext2/3 */
1324 	{Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1325 	{Opt_removed, "journal=%u"},	/* mount option from ext2/3 */
1326 	{Opt_err, NULL},
1327 };
1328 
1329 static ext4_fsblk_t get_sb_block(void **data)
1330 {
1331 	ext4_fsblk_t	sb_block;
1332 	char		*options = (char *) *data;
1333 
1334 	if (!options || strncmp(options, "sb=", 3) != 0)
1335 		return 1;	/* Default location */
1336 
1337 	options += 3;
1338 	/* TODO: use simple_strtoll with >32bit ext4 */
1339 	sb_block = simple_strtoul(options, &options, 0);
1340 	if (*options && *options != ',') {
1341 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1342 		       (char *) *data);
1343 		return 1;
1344 	}
1345 	if (*options == ',')
1346 		options++;
1347 	*data = (void *) options;
1348 
1349 	return sb_block;
1350 }
1351 
1352 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1353 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1354 	"Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1355 
1356 #ifdef CONFIG_QUOTA
1357 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1358 {
1359 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1360 	char *qname;
1361 
1362 	if (sb_any_quota_loaded(sb) &&
1363 		!sbi->s_qf_names[qtype]) {
1364 		ext4_msg(sb, KERN_ERR,
1365 			"Cannot change journaled "
1366 			"quota options when quota turned on");
1367 		return -1;
1368 	}
1369 	qname = match_strdup(args);
1370 	if (!qname) {
1371 		ext4_msg(sb, KERN_ERR,
1372 			"Not enough memory for storing quotafile name");
1373 		return -1;
1374 	}
1375 	if (sbi->s_qf_names[qtype] &&
1376 		strcmp(sbi->s_qf_names[qtype], qname)) {
1377 		ext4_msg(sb, KERN_ERR,
1378 			"%s quota file already specified", QTYPE2NAME(qtype));
1379 		kfree(qname);
1380 		return -1;
1381 	}
1382 	sbi->s_qf_names[qtype] = qname;
1383 	if (strchr(sbi->s_qf_names[qtype], '/')) {
1384 		ext4_msg(sb, KERN_ERR,
1385 			"quotafile must be on filesystem root");
1386 		kfree(sbi->s_qf_names[qtype]);
1387 		sbi->s_qf_names[qtype] = NULL;
1388 		return -1;
1389 	}
1390 	set_opt(sb, QUOTA);
1391 	return 1;
1392 }
1393 
1394 static int clear_qf_name(struct super_block *sb, int qtype)
1395 {
1396 
1397 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1398 
1399 	if (sb_any_quota_loaded(sb) &&
1400 		sbi->s_qf_names[qtype]) {
1401 		ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1402 			" when quota turned on");
1403 		return -1;
1404 	}
1405 	/*
1406 	 * The space will be released later when all options are confirmed
1407 	 * to be correct
1408 	 */
1409 	sbi->s_qf_names[qtype] = NULL;
1410 	return 1;
1411 }
1412 #endif
1413 
1414 #define MOPT_SET	0x0001
1415 #define MOPT_CLEAR	0x0002
1416 #define MOPT_NOSUPPORT	0x0004
1417 #define MOPT_EXPLICIT	0x0008
1418 #define MOPT_CLEAR_ERR	0x0010
1419 #define MOPT_GTE0	0x0020
1420 #ifdef CONFIG_QUOTA
1421 #define MOPT_Q		0
1422 #define MOPT_QFMT	0x0040
1423 #else
1424 #define MOPT_Q		MOPT_NOSUPPORT
1425 #define MOPT_QFMT	MOPT_NOSUPPORT
1426 #endif
1427 #define MOPT_DATAJ	0x0080
1428 
1429 static const struct mount_opts {
1430 	int	token;
1431 	int	mount_opt;
1432 	int	flags;
1433 } ext4_mount_opts[] = {
1434 	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1435 	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1436 	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1437 	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1438 	{Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1439 	{Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1440 	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1441 	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1442 	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1443 	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1444 	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1445 	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1446 	{Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1447 	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1448 	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1449 	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1450 				    EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1451 	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1452 	{Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1453 	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1454 	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1455 	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1456 	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1457 	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1458 	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1459 	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1460 	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1461 	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1462 	{Opt_commit, 0, MOPT_GTE0},
1463 	{Opt_max_batch_time, 0, MOPT_GTE0},
1464 	{Opt_min_batch_time, 0, MOPT_GTE0},
1465 	{Opt_inode_readahead_blks, 0, MOPT_GTE0},
1466 	{Opt_init_itable, 0, MOPT_GTE0},
1467 	{Opt_stripe, 0, MOPT_GTE0},
1468 	{Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1469 	{Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1470 	{Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1471 #ifdef CONFIG_EXT4_FS_XATTR
1472 	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1473 	{Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1474 #else
1475 	{Opt_user_xattr, 0, MOPT_NOSUPPORT},
1476 	{Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
1477 #endif
1478 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1479 	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1480 	{Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1481 #else
1482 	{Opt_acl, 0, MOPT_NOSUPPORT},
1483 	{Opt_noacl, 0, MOPT_NOSUPPORT},
1484 #endif
1485 	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1486 	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1487 	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1488 	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1489 							MOPT_SET | MOPT_Q},
1490 	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1491 							MOPT_SET | MOPT_Q},
1492 	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1493 		       EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1494 	{Opt_usrjquota, 0, MOPT_Q},
1495 	{Opt_grpjquota, 0, MOPT_Q},
1496 	{Opt_offusrjquota, 0, MOPT_Q},
1497 	{Opt_offgrpjquota, 0, MOPT_Q},
1498 	{Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1499 	{Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1500 	{Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1501 	{Opt_err, 0, 0}
1502 };
1503 
1504 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1505 			    substring_t *args, unsigned long *journal_devnum,
1506 			    unsigned int *journal_ioprio, int is_remount)
1507 {
1508 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1509 	const struct mount_opts *m;
1510 	kuid_t uid;
1511 	kgid_t gid;
1512 	int arg = 0;
1513 
1514 #ifdef CONFIG_QUOTA
1515 	if (token == Opt_usrjquota)
1516 		return set_qf_name(sb, USRQUOTA, &args[0]);
1517 	else if (token == Opt_grpjquota)
1518 		return set_qf_name(sb, GRPQUOTA, &args[0]);
1519 	else if (token == Opt_offusrjquota)
1520 		return clear_qf_name(sb, USRQUOTA);
1521 	else if (token == Opt_offgrpjquota)
1522 		return clear_qf_name(sb, GRPQUOTA);
1523 #endif
1524 	if (args->from && match_int(args, &arg))
1525 		return -1;
1526 	switch (token) {
1527 	case Opt_noacl:
1528 	case Opt_nouser_xattr:
1529 		ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1530 		break;
1531 	case Opt_sb:
1532 		return 1;	/* handled by get_sb_block() */
1533 	case Opt_removed:
1534 		ext4_msg(sb, KERN_WARNING,
1535 			 "Ignoring removed %s option", opt);
1536 		return 1;
1537 	case Opt_resuid:
1538 		uid = make_kuid(current_user_ns(), arg);
1539 		if (!uid_valid(uid)) {
1540 			ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1541 			return -1;
1542 		}
1543 		sbi->s_resuid = uid;
1544 		return 1;
1545 	case Opt_resgid:
1546 		gid = make_kgid(current_user_ns(), arg);
1547 		if (!gid_valid(gid)) {
1548 			ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1549 			return -1;
1550 		}
1551 		sbi->s_resgid = gid;
1552 		return 1;
1553 	case Opt_abort:
1554 		sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1555 		return 1;
1556 	case Opt_i_version:
1557 		sb->s_flags |= MS_I_VERSION;
1558 		return 1;
1559 	case Opt_journal_dev:
1560 		if (is_remount) {
1561 			ext4_msg(sb, KERN_ERR,
1562 				 "Cannot specify journal on remount");
1563 			return -1;
1564 		}
1565 		*journal_devnum = arg;
1566 		return 1;
1567 	case Opt_journal_ioprio:
1568 		if (arg < 0 || arg > 7)
1569 			return -1;
1570 		*journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1571 		return 1;
1572 	}
1573 
1574 	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1575 		if (token != m->token)
1576 			continue;
1577 		if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1578 			return -1;
1579 		if (m->flags & MOPT_EXPLICIT)
1580 			set_opt2(sb, EXPLICIT_DELALLOC);
1581 		if (m->flags & MOPT_CLEAR_ERR)
1582 			clear_opt(sb, ERRORS_MASK);
1583 		if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1584 			ext4_msg(sb, KERN_ERR, "Cannot change quota "
1585 				 "options when quota turned on");
1586 			return -1;
1587 		}
1588 
1589 		if (m->flags & MOPT_NOSUPPORT) {
1590 			ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1591 		} else if (token == Opt_commit) {
1592 			if (arg == 0)
1593 				arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1594 			sbi->s_commit_interval = HZ * arg;
1595 		} else if (token == Opt_max_batch_time) {
1596 			if (arg == 0)
1597 				arg = EXT4_DEF_MAX_BATCH_TIME;
1598 			sbi->s_max_batch_time = arg;
1599 		} else if (token == Opt_min_batch_time) {
1600 			sbi->s_min_batch_time = arg;
1601 		} else if (token == Opt_inode_readahead_blks) {
1602 			if (arg > (1 << 30))
1603 				return -1;
1604 			if (arg && !is_power_of_2(arg)) {
1605 				ext4_msg(sb, KERN_ERR,
1606 					 "EXT4-fs: inode_readahead_blks"
1607 					 " must be a power of 2");
1608 				return -1;
1609 			}
1610 			sbi->s_inode_readahead_blks = arg;
1611 		} else if (token == Opt_init_itable) {
1612 			set_opt(sb, INIT_INODE_TABLE);
1613 			if (!args->from)
1614 				arg = EXT4_DEF_LI_WAIT_MULT;
1615 			sbi->s_li_wait_mult = arg;
1616 		} else if (token == Opt_stripe) {
1617 			sbi->s_stripe = arg;
1618 		} else if (m->flags & MOPT_DATAJ) {
1619 			if (is_remount) {
1620 				if (!sbi->s_journal)
1621 					ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1622 				else if (test_opt(sb, DATA_FLAGS) !=
1623 					 m->mount_opt) {
1624 					ext4_msg(sb, KERN_ERR,
1625 					 "Cannot change data mode on remount");
1626 					return -1;
1627 				}
1628 			} else {
1629 				clear_opt(sb, DATA_FLAGS);
1630 				sbi->s_mount_opt |= m->mount_opt;
1631 			}
1632 #ifdef CONFIG_QUOTA
1633 		} else if (m->flags & MOPT_QFMT) {
1634 			if (sb_any_quota_loaded(sb) &&
1635 			    sbi->s_jquota_fmt != m->mount_opt) {
1636 				ext4_msg(sb, KERN_ERR, "Cannot "
1637 					 "change journaled quota options "
1638 					 "when quota turned on");
1639 				return -1;
1640 			}
1641 			sbi->s_jquota_fmt = m->mount_opt;
1642 #endif
1643 		} else {
1644 			if (!args->from)
1645 				arg = 1;
1646 			if (m->flags & MOPT_CLEAR)
1647 				arg = !arg;
1648 			else if (unlikely(!(m->flags & MOPT_SET))) {
1649 				ext4_msg(sb, KERN_WARNING,
1650 					 "buggy handling of option %s", opt);
1651 				WARN_ON(1);
1652 				return -1;
1653 			}
1654 			if (arg != 0)
1655 				sbi->s_mount_opt |= m->mount_opt;
1656 			else
1657 				sbi->s_mount_opt &= ~m->mount_opt;
1658 		}
1659 		return 1;
1660 	}
1661 	ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1662 		 "or missing value", opt);
1663 	return -1;
1664 }
1665 
1666 static int parse_options(char *options, struct super_block *sb,
1667 			 unsigned long *journal_devnum,
1668 			 unsigned int *journal_ioprio,
1669 			 int is_remount)
1670 {
1671 #ifdef CONFIG_QUOTA
1672 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1673 #endif
1674 	char *p;
1675 	substring_t args[MAX_OPT_ARGS];
1676 	int token;
1677 
1678 	if (!options)
1679 		return 1;
1680 
1681 	while ((p = strsep(&options, ",")) != NULL) {
1682 		if (!*p)
1683 			continue;
1684 		/*
1685 		 * Initialize args struct so we know whether arg was
1686 		 * found; some options take optional arguments.
1687 		 */
1688 		args[0].to = args[0].from = 0;
1689 		token = match_token(p, tokens, args);
1690 		if (handle_mount_opt(sb, p, token, args, journal_devnum,
1691 				     journal_ioprio, is_remount) < 0)
1692 			return 0;
1693 	}
1694 #ifdef CONFIG_QUOTA
1695 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1696 		if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1697 			clear_opt(sb, USRQUOTA);
1698 
1699 		if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1700 			clear_opt(sb, GRPQUOTA);
1701 
1702 		if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1703 			ext4_msg(sb, KERN_ERR, "old and new quota "
1704 					"format mixing");
1705 			return 0;
1706 		}
1707 
1708 		if (!sbi->s_jquota_fmt) {
1709 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1710 					"not specified");
1711 			return 0;
1712 		}
1713 	} else {
1714 		if (sbi->s_jquota_fmt) {
1715 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1716 					"specified with no journaling "
1717 					"enabled");
1718 			return 0;
1719 		}
1720 	}
1721 #endif
1722 	return 1;
1723 }
1724 
1725 static inline void ext4_show_quota_options(struct seq_file *seq,
1726 					   struct super_block *sb)
1727 {
1728 #if defined(CONFIG_QUOTA)
1729 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1730 
1731 	if (sbi->s_jquota_fmt) {
1732 		char *fmtname = "";
1733 
1734 		switch (sbi->s_jquota_fmt) {
1735 		case QFMT_VFS_OLD:
1736 			fmtname = "vfsold";
1737 			break;
1738 		case QFMT_VFS_V0:
1739 			fmtname = "vfsv0";
1740 			break;
1741 		case QFMT_VFS_V1:
1742 			fmtname = "vfsv1";
1743 			break;
1744 		}
1745 		seq_printf(seq, ",jqfmt=%s", fmtname);
1746 	}
1747 
1748 	if (sbi->s_qf_names[USRQUOTA])
1749 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1750 
1751 	if (sbi->s_qf_names[GRPQUOTA])
1752 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1753 
1754 	if (test_opt(sb, USRQUOTA))
1755 		seq_puts(seq, ",usrquota");
1756 
1757 	if (test_opt(sb, GRPQUOTA))
1758 		seq_puts(seq, ",grpquota");
1759 #endif
1760 }
1761 
1762 static const char *token2str(int token)
1763 {
1764 	static const struct match_token *t;
1765 
1766 	for (t = tokens; t->token != Opt_err; t++)
1767 		if (t->token == token && !strchr(t->pattern, '='))
1768 			break;
1769 	return t->pattern;
1770 }
1771 
1772 /*
1773  * Show an option if
1774  *  - it's set to a non-default value OR
1775  *  - if the per-sb default is different from the global default
1776  */
1777 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1778 			      int nodefs)
1779 {
1780 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1781 	struct ext4_super_block *es = sbi->s_es;
1782 	int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1783 	const struct mount_opts *m;
1784 	char sep = nodefs ? '\n' : ',';
1785 
1786 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1787 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1788 
1789 	if (sbi->s_sb_block != 1)
1790 		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1791 
1792 	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1793 		int want_set = m->flags & MOPT_SET;
1794 		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1795 		    (m->flags & MOPT_CLEAR_ERR))
1796 			continue;
1797 		if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1798 			continue; /* skip if same as the default */
1799 		if ((want_set &&
1800 		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1801 		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1802 			continue; /* select Opt_noFoo vs Opt_Foo */
1803 		SEQ_OPTS_PRINT("%s", token2str(m->token));
1804 	}
1805 
1806 	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1807 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1808 		SEQ_OPTS_PRINT("resuid=%u",
1809 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
1810 	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1811 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1812 		SEQ_OPTS_PRINT("resgid=%u",
1813 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
1814 	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1815 	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1816 		SEQ_OPTS_PUTS("errors=remount-ro");
1817 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1818 		SEQ_OPTS_PUTS("errors=continue");
1819 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1820 		SEQ_OPTS_PUTS("errors=panic");
1821 	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1822 		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1823 	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1824 		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1825 	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1826 		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1827 	if (sb->s_flags & MS_I_VERSION)
1828 		SEQ_OPTS_PUTS("i_version");
1829 	if (nodefs || sbi->s_stripe)
1830 		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1831 	if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1832 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1833 			SEQ_OPTS_PUTS("data=journal");
1834 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1835 			SEQ_OPTS_PUTS("data=ordered");
1836 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1837 			SEQ_OPTS_PUTS("data=writeback");
1838 	}
1839 	if (nodefs ||
1840 	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1841 		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1842 			       sbi->s_inode_readahead_blks);
1843 
1844 	if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1845 		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1846 		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1847 
1848 	ext4_show_quota_options(seq, sb);
1849 	return 0;
1850 }
1851 
1852 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1853 {
1854 	return _ext4_show_options(seq, root->d_sb, 0);
1855 }
1856 
1857 static int options_seq_show(struct seq_file *seq, void *offset)
1858 {
1859 	struct super_block *sb = seq->private;
1860 	int rc;
1861 
1862 	seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1863 	rc = _ext4_show_options(seq, sb, 1);
1864 	seq_puts(seq, "\n");
1865 	return rc;
1866 }
1867 
1868 static int options_open_fs(struct inode *inode, struct file *file)
1869 {
1870 	return single_open(file, options_seq_show, PDE(inode)->data);
1871 }
1872 
1873 static const struct file_operations ext4_seq_options_fops = {
1874 	.owner = THIS_MODULE,
1875 	.open = options_open_fs,
1876 	.read = seq_read,
1877 	.llseek = seq_lseek,
1878 	.release = single_release,
1879 };
1880 
1881 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1882 			    int read_only)
1883 {
1884 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1885 	int res = 0;
1886 
1887 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1888 		ext4_msg(sb, KERN_ERR, "revision level too high, "
1889 			 "forcing read-only mode");
1890 		res = MS_RDONLY;
1891 	}
1892 	if (read_only)
1893 		goto done;
1894 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1895 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1896 			 "running e2fsck is recommended");
1897 	else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1898 		ext4_msg(sb, KERN_WARNING,
1899 			 "warning: mounting fs with errors, "
1900 			 "running e2fsck is recommended");
1901 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1902 		 le16_to_cpu(es->s_mnt_count) >=
1903 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1904 		ext4_msg(sb, KERN_WARNING,
1905 			 "warning: maximal mount count reached, "
1906 			 "running e2fsck is recommended");
1907 	else if (le32_to_cpu(es->s_checkinterval) &&
1908 		(le32_to_cpu(es->s_lastcheck) +
1909 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1910 		ext4_msg(sb, KERN_WARNING,
1911 			 "warning: checktime reached, "
1912 			 "running e2fsck is recommended");
1913 	if (!sbi->s_journal)
1914 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1915 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1916 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1917 	le16_add_cpu(&es->s_mnt_count, 1);
1918 	es->s_mtime = cpu_to_le32(get_seconds());
1919 	ext4_update_dynamic_rev(sb);
1920 	if (sbi->s_journal)
1921 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1922 
1923 	ext4_commit_super(sb, 1);
1924 done:
1925 	if (test_opt(sb, DEBUG))
1926 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1927 				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1928 			sb->s_blocksize,
1929 			sbi->s_groups_count,
1930 			EXT4_BLOCKS_PER_GROUP(sb),
1931 			EXT4_INODES_PER_GROUP(sb),
1932 			sbi->s_mount_opt, sbi->s_mount_opt2);
1933 
1934 	cleancache_init_fs(sb);
1935 	return res;
1936 }
1937 
1938 static int ext4_fill_flex_info(struct super_block *sb)
1939 {
1940 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1941 	struct ext4_group_desc *gdp = NULL;
1942 	ext4_group_t flex_group_count;
1943 	ext4_group_t flex_group;
1944 	unsigned int groups_per_flex = 0;
1945 	size_t size;
1946 	int i;
1947 
1948 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1949 	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1950 		sbi->s_log_groups_per_flex = 0;
1951 		return 1;
1952 	}
1953 	groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1954 
1955 	/* We allocate both existing and potentially added groups */
1956 	flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1957 			((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1958 			      EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1959 	size = flex_group_count * sizeof(struct flex_groups);
1960 	sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
1961 	if (sbi->s_flex_groups == NULL) {
1962 		ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1963 			 flex_group_count);
1964 		goto failed;
1965 	}
1966 
1967 	for (i = 0; i < sbi->s_groups_count; i++) {
1968 		gdp = ext4_get_group_desc(sb, i, NULL);
1969 
1970 		flex_group = ext4_flex_group(sbi, i);
1971 		atomic_add(ext4_free_inodes_count(sb, gdp),
1972 			   &sbi->s_flex_groups[flex_group].free_inodes);
1973 		atomic_add(ext4_free_group_clusters(sb, gdp),
1974 			   &sbi->s_flex_groups[flex_group].free_clusters);
1975 		atomic_add(ext4_used_dirs_count(sb, gdp),
1976 			   &sbi->s_flex_groups[flex_group].used_dirs);
1977 	}
1978 
1979 	return 1;
1980 failed:
1981 	return 0;
1982 }
1983 
1984 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1985 				   struct ext4_group_desc *gdp)
1986 {
1987 	int offset;
1988 	__u16 crc = 0;
1989 	__le32 le_group = cpu_to_le32(block_group);
1990 
1991 	if ((sbi->s_es->s_feature_ro_compat &
1992 	     cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1993 		/* Use new metadata_csum algorithm */
1994 		__u16 old_csum;
1995 		__u32 csum32;
1996 
1997 		old_csum = gdp->bg_checksum;
1998 		gdp->bg_checksum = 0;
1999 		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2000 				     sizeof(le_group));
2001 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2002 				     sbi->s_desc_size);
2003 		gdp->bg_checksum = old_csum;
2004 
2005 		crc = csum32 & 0xFFFF;
2006 		goto out;
2007 	}
2008 
2009 	/* old crc16 code */
2010 	offset = offsetof(struct ext4_group_desc, bg_checksum);
2011 
2012 	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2013 	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2014 	crc = crc16(crc, (__u8 *)gdp, offset);
2015 	offset += sizeof(gdp->bg_checksum); /* skip checksum */
2016 	/* for checksum of struct ext4_group_desc do the rest...*/
2017 	if ((sbi->s_es->s_feature_incompat &
2018 	     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2019 	    offset < le16_to_cpu(sbi->s_es->s_desc_size))
2020 		crc = crc16(crc, (__u8 *)gdp + offset,
2021 			    le16_to_cpu(sbi->s_es->s_desc_size) -
2022 				offset);
2023 
2024 out:
2025 	return cpu_to_le16(crc);
2026 }
2027 
2028 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2029 				struct ext4_group_desc *gdp)
2030 {
2031 	if (ext4_has_group_desc_csum(sb) &&
2032 	    (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2033 						      block_group, gdp)))
2034 		return 0;
2035 
2036 	return 1;
2037 }
2038 
2039 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2040 			      struct ext4_group_desc *gdp)
2041 {
2042 	if (!ext4_has_group_desc_csum(sb))
2043 		return;
2044 	gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2045 }
2046 
2047 /* Called at mount-time, super-block is locked */
2048 static int ext4_check_descriptors(struct super_block *sb,
2049 				  ext4_group_t *first_not_zeroed)
2050 {
2051 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2052 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2053 	ext4_fsblk_t last_block;
2054 	ext4_fsblk_t block_bitmap;
2055 	ext4_fsblk_t inode_bitmap;
2056 	ext4_fsblk_t inode_table;
2057 	int flexbg_flag = 0;
2058 	ext4_group_t i, grp = sbi->s_groups_count;
2059 
2060 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2061 		flexbg_flag = 1;
2062 
2063 	ext4_debug("Checking group descriptors");
2064 
2065 	for (i = 0; i < sbi->s_groups_count; i++) {
2066 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2067 
2068 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
2069 			last_block = ext4_blocks_count(sbi->s_es) - 1;
2070 		else
2071 			last_block = first_block +
2072 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
2073 
2074 		if ((grp == sbi->s_groups_count) &&
2075 		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2076 			grp = i;
2077 
2078 		block_bitmap = ext4_block_bitmap(sb, gdp);
2079 		if (block_bitmap < first_block || block_bitmap > last_block) {
2080 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2081 			       "Block bitmap for group %u not in group "
2082 			       "(block %llu)!", i, block_bitmap);
2083 			return 0;
2084 		}
2085 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
2086 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
2087 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2088 			       "Inode bitmap for group %u not in group "
2089 			       "(block %llu)!", i, inode_bitmap);
2090 			return 0;
2091 		}
2092 		inode_table = ext4_inode_table(sb, gdp);
2093 		if (inode_table < first_block ||
2094 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2095 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2096 			       "Inode table for group %u not in group "
2097 			       "(block %llu)!", i, inode_table);
2098 			return 0;
2099 		}
2100 		ext4_lock_group(sb, i);
2101 		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2102 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2103 				 "Checksum for group %u failed (%u!=%u)",
2104 				 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2105 				     gdp)), le16_to_cpu(gdp->bg_checksum));
2106 			if (!(sb->s_flags & MS_RDONLY)) {
2107 				ext4_unlock_group(sb, i);
2108 				return 0;
2109 			}
2110 		}
2111 		ext4_unlock_group(sb, i);
2112 		if (!flexbg_flag)
2113 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
2114 	}
2115 	if (NULL != first_not_zeroed)
2116 		*first_not_zeroed = grp;
2117 
2118 	ext4_free_blocks_count_set(sbi->s_es,
2119 				   EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2120 	sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2121 	return 1;
2122 }
2123 
2124 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2125  * the superblock) which were deleted from all directories, but held open by
2126  * a process at the time of a crash.  We walk the list and try to delete these
2127  * inodes at recovery time (only with a read-write filesystem).
2128  *
2129  * In order to keep the orphan inode chain consistent during traversal (in
2130  * case of crash during recovery), we link each inode into the superblock
2131  * orphan list_head and handle it the same way as an inode deletion during
2132  * normal operation (which journals the operations for us).
2133  *
2134  * We only do an iget() and an iput() on each inode, which is very safe if we
2135  * accidentally point at an in-use or already deleted inode.  The worst that
2136  * can happen in this case is that we get a "bit already cleared" message from
2137  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2138  * e2fsck was run on this filesystem, and it must have already done the orphan
2139  * inode cleanup for us, so we can safely abort without any further action.
2140  */
2141 static void ext4_orphan_cleanup(struct super_block *sb,
2142 				struct ext4_super_block *es)
2143 {
2144 	unsigned int s_flags = sb->s_flags;
2145 	int nr_orphans = 0, nr_truncates = 0;
2146 #ifdef CONFIG_QUOTA
2147 	int i;
2148 #endif
2149 	if (!es->s_last_orphan) {
2150 		jbd_debug(4, "no orphan inodes to clean up\n");
2151 		return;
2152 	}
2153 
2154 	if (bdev_read_only(sb->s_bdev)) {
2155 		ext4_msg(sb, KERN_ERR, "write access "
2156 			"unavailable, skipping orphan cleanup");
2157 		return;
2158 	}
2159 
2160 	/* Check if feature set would not allow a r/w mount */
2161 	if (!ext4_feature_set_ok(sb, 0)) {
2162 		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2163 			 "unknown ROCOMPAT features");
2164 		return;
2165 	}
2166 
2167 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2168 		if (es->s_last_orphan)
2169 			jbd_debug(1, "Errors on filesystem, "
2170 				  "clearing orphan list.\n");
2171 		es->s_last_orphan = 0;
2172 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2173 		return;
2174 	}
2175 
2176 	if (s_flags & MS_RDONLY) {
2177 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2178 		sb->s_flags &= ~MS_RDONLY;
2179 	}
2180 #ifdef CONFIG_QUOTA
2181 	/* Needed for iput() to work correctly and not trash data */
2182 	sb->s_flags |= MS_ACTIVE;
2183 	/* Turn on quotas so that they are updated correctly */
2184 	for (i = 0; i < MAXQUOTAS; i++) {
2185 		if (EXT4_SB(sb)->s_qf_names[i]) {
2186 			int ret = ext4_quota_on_mount(sb, i);
2187 			if (ret < 0)
2188 				ext4_msg(sb, KERN_ERR,
2189 					"Cannot turn on journaled "
2190 					"quota: error %d", ret);
2191 		}
2192 	}
2193 #endif
2194 
2195 	while (es->s_last_orphan) {
2196 		struct inode *inode;
2197 
2198 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2199 		if (IS_ERR(inode)) {
2200 			es->s_last_orphan = 0;
2201 			break;
2202 		}
2203 
2204 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2205 		dquot_initialize(inode);
2206 		if (inode->i_nlink) {
2207 			ext4_msg(sb, KERN_DEBUG,
2208 				"%s: truncating inode %lu to %lld bytes",
2209 				__func__, inode->i_ino, inode->i_size);
2210 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2211 				  inode->i_ino, inode->i_size);
2212 			ext4_truncate(inode);
2213 			nr_truncates++;
2214 		} else {
2215 			ext4_msg(sb, KERN_DEBUG,
2216 				"%s: deleting unreferenced inode %lu",
2217 				__func__, inode->i_ino);
2218 			jbd_debug(2, "deleting unreferenced inode %lu\n",
2219 				  inode->i_ino);
2220 			nr_orphans++;
2221 		}
2222 		iput(inode);  /* The delete magic happens here! */
2223 	}
2224 
2225 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2226 
2227 	if (nr_orphans)
2228 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2229 		       PLURAL(nr_orphans));
2230 	if (nr_truncates)
2231 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2232 		       PLURAL(nr_truncates));
2233 #ifdef CONFIG_QUOTA
2234 	/* Turn quotas off */
2235 	for (i = 0; i < MAXQUOTAS; i++) {
2236 		if (sb_dqopt(sb)->files[i])
2237 			dquot_quota_off(sb, i);
2238 	}
2239 #endif
2240 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2241 }
2242 
2243 /*
2244  * Maximal extent format file size.
2245  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2246  * extent format containers, within a sector_t, and within i_blocks
2247  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2248  * so that won't be a limiting factor.
2249  *
2250  * However there is other limiting factor. We do store extents in the form
2251  * of starting block and length, hence the resulting length of the extent
2252  * covering maximum file size must fit into on-disk format containers as
2253  * well. Given that length is always by 1 unit bigger than max unit (because
2254  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2255  *
2256  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2257  */
2258 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2259 {
2260 	loff_t res;
2261 	loff_t upper_limit = MAX_LFS_FILESIZE;
2262 
2263 	/* small i_blocks in vfs inode? */
2264 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2265 		/*
2266 		 * CONFIG_LBDAF is not enabled implies the inode
2267 		 * i_block represent total blocks in 512 bytes
2268 		 * 32 == size of vfs inode i_blocks * 8
2269 		 */
2270 		upper_limit = (1LL << 32) - 1;
2271 
2272 		/* total blocks in file system block size */
2273 		upper_limit >>= (blkbits - 9);
2274 		upper_limit <<= blkbits;
2275 	}
2276 
2277 	/*
2278 	 * 32-bit extent-start container, ee_block. We lower the maxbytes
2279 	 * by one fs block, so ee_len can cover the extent of maximum file
2280 	 * size
2281 	 */
2282 	res = (1LL << 32) - 1;
2283 	res <<= blkbits;
2284 
2285 	/* Sanity check against vm- & vfs- imposed limits */
2286 	if (res > upper_limit)
2287 		res = upper_limit;
2288 
2289 	return res;
2290 }
2291 
2292 /*
2293  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2294  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2295  * We need to be 1 filesystem block less than the 2^48 sector limit.
2296  */
2297 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2298 {
2299 	loff_t res = EXT4_NDIR_BLOCKS;
2300 	int meta_blocks;
2301 	loff_t upper_limit;
2302 	/* This is calculated to be the largest file size for a dense, block
2303 	 * mapped file such that the file's total number of 512-byte sectors,
2304 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
2305 	 *
2306 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2307 	 * number of 512-byte sectors of the file.
2308 	 */
2309 
2310 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2311 		/*
2312 		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2313 		 * the inode i_block field represents total file blocks in
2314 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2315 		 */
2316 		upper_limit = (1LL << 32) - 1;
2317 
2318 		/* total blocks in file system block size */
2319 		upper_limit >>= (bits - 9);
2320 
2321 	} else {
2322 		/*
2323 		 * We use 48 bit ext4_inode i_blocks
2324 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2325 		 * represent total number of blocks in
2326 		 * file system block size
2327 		 */
2328 		upper_limit = (1LL << 48) - 1;
2329 
2330 	}
2331 
2332 	/* indirect blocks */
2333 	meta_blocks = 1;
2334 	/* double indirect blocks */
2335 	meta_blocks += 1 + (1LL << (bits-2));
2336 	/* tripple indirect blocks */
2337 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2338 
2339 	upper_limit -= meta_blocks;
2340 	upper_limit <<= bits;
2341 
2342 	res += 1LL << (bits-2);
2343 	res += 1LL << (2*(bits-2));
2344 	res += 1LL << (3*(bits-2));
2345 	res <<= bits;
2346 	if (res > upper_limit)
2347 		res = upper_limit;
2348 
2349 	if (res > MAX_LFS_FILESIZE)
2350 		res = MAX_LFS_FILESIZE;
2351 
2352 	return res;
2353 }
2354 
2355 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2356 				   ext4_fsblk_t logical_sb_block, int nr)
2357 {
2358 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2359 	ext4_group_t bg, first_meta_bg;
2360 	int has_super = 0;
2361 
2362 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2363 
2364 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2365 	    nr < first_meta_bg)
2366 		return logical_sb_block + nr + 1;
2367 	bg = sbi->s_desc_per_block * nr;
2368 	if (ext4_bg_has_super(sb, bg))
2369 		has_super = 1;
2370 
2371 	return (has_super + ext4_group_first_block_no(sb, bg));
2372 }
2373 
2374 /**
2375  * ext4_get_stripe_size: Get the stripe size.
2376  * @sbi: In memory super block info
2377  *
2378  * If we have specified it via mount option, then
2379  * use the mount option value. If the value specified at mount time is
2380  * greater than the blocks per group use the super block value.
2381  * If the super block value is greater than blocks per group return 0.
2382  * Allocator needs it be less than blocks per group.
2383  *
2384  */
2385 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2386 {
2387 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2388 	unsigned long stripe_width =
2389 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2390 	int ret;
2391 
2392 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2393 		ret = sbi->s_stripe;
2394 	else if (stripe_width <= sbi->s_blocks_per_group)
2395 		ret = stripe_width;
2396 	else if (stride <= sbi->s_blocks_per_group)
2397 		ret = stride;
2398 	else
2399 		ret = 0;
2400 
2401 	/*
2402 	 * If the stripe width is 1, this makes no sense and
2403 	 * we set it to 0 to turn off stripe handling code.
2404 	 */
2405 	if (ret <= 1)
2406 		ret = 0;
2407 
2408 	return ret;
2409 }
2410 
2411 /* sysfs supprt */
2412 
2413 struct ext4_attr {
2414 	struct attribute attr;
2415 	ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2416 	ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2417 			 const char *, size_t);
2418 	int offset;
2419 };
2420 
2421 static int parse_strtoul(const char *buf,
2422 		unsigned long max, unsigned long *value)
2423 {
2424 	char *endp;
2425 
2426 	*value = simple_strtoul(skip_spaces(buf), &endp, 0);
2427 	endp = skip_spaces(endp);
2428 	if (*endp || *value > max)
2429 		return -EINVAL;
2430 
2431 	return 0;
2432 }
2433 
2434 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2435 					      struct ext4_sb_info *sbi,
2436 					      char *buf)
2437 {
2438 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2439 		(s64) EXT4_C2B(sbi,
2440 			percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2441 }
2442 
2443 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2444 					 struct ext4_sb_info *sbi, char *buf)
2445 {
2446 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2447 
2448 	if (!sb->s_bdev->bd_part)
2449 		return snprintf(buf, PAGE_SIZE, "0\n");
2450 	return snprintf(buf, PAGE_SIZE, "%lu\n",
2451 			(part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2452 			 sbi->s_sectors_written_start) >> 1);
2453 }
2454 
2455 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2456 					  struct ext4_sb_info *sbi, char *buf)
2457 {
2458 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2459 
2460 	if (!sb->s_bdev->bd_part)
2461 		return snprintf(buf, PAGE_SIZE, "0\n");
2462 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2463 			(unsigned long long)(sbi->s_kbytes_written +
2464 			((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2465 			  EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2466 }
2467 
2468 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2469 					  struct ext4_sb_info *sbi,
2470 					  const char *buf, size_t count)
2471 {
2472 	unsigned long t;
2473 
2474 	if (parse_strtoul(buf, 0x40000000, &t))
2475 		return -EINVAL;
2476 
2477 	if (t && !is_power_of_2(t))
2478 		return -EINVAL;
2479 
2480 	sbi->s_inode_readahead_blks = t;
2481 	return count;
2482 }
2483 
2484 static ssize_t sbi_ui_show(struct ext4_attr *a,
2485 			   struct ext4_sb_info *sbi, char *buf)
2486 {
2487 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2488 
2489 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2490 }
2491 
2492 static ssize_t sbi_ui_store(struct ext4_attr *a,
2493 			    struct ext4_sb_info *sbi,
2494 			    const char *buf, size_t count)
2495 {
2496 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2497 	unsigned long t;
2498 
2499 	if (parse_strtoul(buf, 0xffffffff, &t))
2500 		return -EINVAL;
2501 	*ui = t;
2502 	return count;
2503 }
2504 
2505 static ssize_t trigger_test_error(struct ext4_attr *a,
2506 				  struct ext4_sb_info *sbi,
2507 				  const char *buf, size_t count)
2508 {
2509 	int len = count;
2510 
2511 	if (!capable(CAP_SYS_ADMIN))
2512 		return -EPERM;
2513 
2514 	if (len && buf[len-1] == '\n')
2515 		len--;
2516 
2517 	if (len)
2518 		ext4_error(sbi->s_sb, "%.*s", len, buf);
2519 	return count;
2520 }
2521 
2522 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2523 static struct ext4_attr ext4_attr_##_name = {			\
2524 	.attr = {.name = __stringify(_name), .mode = _mode },	\
2525 	.show	= _show,					\
2526 	.store	= _store,					\
2527 	.offset = offsetof(struct ext4_sb_info, _elname),	\
2528 }
2529 #define EXT4_ATTR(name, mode, show, store) \
2530 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2531 
2532 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2533 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2534 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2535 #define EXT4_RW_ATTR_SBI_UI(name, elname)	\
2536 	EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2537 #define ATTR_LIST(name) &ext4_attr_##name.attr
2538 
2539 EXT4_RO_ATTR(delayed_allocation_blocks);
2540 EXT4_RO_ATTR(session_write_kbytes);
2541 EXT4_RO_ATTR(lifetime_write_kbytes);
2542 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2543 		 inode_readahead_blks_store, s_inode_readahead_blks);
2544 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2545 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2546 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2547 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2548 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2549 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2550 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2551 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2552 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2553 
2554 static struct attribute *ext4_attrs[] = {
2555 	ATTR_LIST(delayed_allocation_blocks),
2556 	ATTR_LIST(session_write_kbytes),
2557 	ATTR_LIST(lifetime_write_kbytes),
2558 	ATTR_LIST(inode_readahead_blks),
2559 	ATTR_LIST(inode_goal),
2560 	ATTR_LIST(mb_stats),
2561 	ATTR_LIST(mb_max_to_scan),
2562 	ATTR_LIST(mb_min_to_scan),
2563 	ATTR_LIST(mb_order2_req),
2564 	ATTR_LIST(mb_stream_req),
2565 	ATTR_LIST(mb_group_prealloc),
2566 	ATTR_LIST(max_writeback_mb_bump),
2567 	ATTR_LIST(trigger_fs_error),
2568 	NULL,
2569 };
2570 
2571 /* Features this copy of ext4 supports */
2572 EXT4_INFO_ATTR(lazy_itable_init);
2573 EXT4_INFO_ATTR(batched_discard);
2574 
2575 static struct attribute *ext4_feat_attrs[] = {
2576 	ATTR_LIST(lazy_itable_init),
2577 	ATTR_LIST(batched_discard),
2578 	NULL,
2579 };
2580 
2581 static ssize_t ext4_attr_show(struct kobject *kobj,
2582 			      struct attribute *attr, char *buf)
2583 {
2584 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2585 						s_kobj);
2586 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2587 
2588 	return a->show ? a->show(a, sbi, buf) : 0;
2589 }
2590 
2591 static ssize_t ext4_attr_store(struct kobject *kobj,
2592 			       struct attribute *attr,
2593 			       const char *buf, size_t len)
2594 {
2595 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2596 						s_kobj);
2597 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2598 
2599 	return a->store ? a->store(a, sbi, buf, len) : 0;
2600 }
2601 
2602 static void ext4_sb_release(struct kobject *kobj)
2603 {
2604 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2605 						s_kobj);
2606 	complete(&sbi->s_kobj_unregister);
2607 }
2608 
2609 static const struct sysfs_ops ext4_attr_ops = {
2610 	.show	= ext4_attr_show,
2611 	.store	= ext4_attr_store,
2612 };
2613 
2614 static struct kobj_type ext4_ktype = {
2615 	.default_attrs	= ext4_attrs,
2616 	.sysfs_ops	= &ext4_attr_ops,
2617 	.release	= ext4_sb_release,
2618 };
2619 
2620 static void ext4_feat_release(struct kobject *kobj)
2621 {
2622 	complete(&ext4_feat->f_kobj_unregister);
2623 }
2624 
2625 static struct kobj_type ext4_feat_ktype = {
2626 	.default_attrs	= ext4_feat_attrs,
2627 	.sysfs_ops	= &ext4_attr_ops,
2628 	.release	= ext4_feat_release,
2629 };
2630 
2631 /*
2632  * Check whether this filesystem can be mounted based on
2633  * the features present and the RDONLY/RDWR mount requested.
2634  * Returns 1 if this filesystem can be mounted as requested,
2635  * 0 if it cannot be.
2636  */
2637 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2638 {
2639 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2640 		ext4_msg(sb, KERN_ERR,
2641 			"Couldn't mount because of "
2642 			"unsupported optional features (%x)",
2643 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2644 			~EXT4_FEATURE_INCOMPAT_SUPP));
2645 		return 0;
2646 	}
2647 
2648 	if (readonly)
2649 		return 1;
2650 
2651 	/* Check that feature set is OK for a read-write mount */
2652 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2653 		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2654 			 "unsupported optional features (%x)",
2655 			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2656 				~EXT4_FEATURE_RO_COMPAT_SUPP));
2657 		return 0;
2658 	}
2659 	/*
2660 	 * Large file size enabled file system can only be mounted
2661 	 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2662 	 */
2663 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2664 		if (sizeof(blkcnt_t) < sizeof(u64)) {
2665 			ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2666 				 "cannot be mounted RDWR without "
2667 				 "CONFIG_LBDAF");
2668 			return 0;
2669 		}
2670 	}
2671 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2672 	    !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2673 		ext4_msg(sb, KERN_ERR,
2674 			 "Can't support bigalloc feature without "
2675 			 "extents feature\n");
2676 		return 0;
2677 	}
2678 
2679 #ifndef CONFIG_QUOTA
2680 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2681 	    !readonly) {
2682 		ext4_msg(sb, KERN_ERR,
2683 			 "Filesystem with quota feature cannot be mounted RDWR "
2684 			 "without CONFIG_QUOTA");
2685 		return 0;
2686 	}
2687 #endif  /* CONFIG_QUOTA */
2688 	return 1;
2689 }
2690 
2691 /*
2692  * This function is called once a day if we have errors logged
2693  * on the file system
2694  */
2695 static void print_daily_error_info(unsigned long arg)
2696 {
2697 	struct super_block *sb = (struct super_block *) arg;
2698 	struct ext4_sb_info *sbi;
2699 	struct ext4_super_block *es;
2700 
2701 	sbi = EXT4_SB(sb);
2702 	es = sbi->s_es;
2703 
2704 	if (es->s_error_count)
2705 		ext4_msg(sb, KERN_NOTICE, "error count: %u",
2706 			 le32_to_cpu(es->s_error_count));
2707 	if (es->s_first_error_time) {
2708 		printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2709 		       sb->s_id, le32_to_cpu(es->s_first_error_time),
2710 		       (int) sizeof(es->s_first_error_func),
2711 		       es->s_first_error_func,
2712 		       le32_to_cpu(es->s_first_error_line));
2713 		if (es->s_first_error_ino)
2714 			printk(": inode %u",
2715 			       le32_to_cpu(es->s_first_error_ino));
2716 		if (es->s_first_error_block)
2717 			printk(": block %llu", (unsigned long long)
2718 			       le64_to_cpu(es->s_first_error_block));
2719 		printk("\n");
2720 	}
2721 	if (es->s_last_error_time) {
2722 		printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2723 		       sb->s_id, le32_to_cpu(es->s_last_error_time),
2724 		       (int) sizeof(es->s_last_error_func),
2725 		       es->s_last_error_func,
2726 		       le32_to_cpu(es->s_last_error_line));
2727 		if (es->s_last_error_ino)
2728 			printk(": inode %u",
2729 			       le32_to_cpu(es->s_last_error_ino));
2730 		if (es->s_last_error_block)
2731 			printk(": block %llu", (unsigned long long)
2732 			       le64_to_cpu(es->s_last_error_block));
2733 		printk("\n");
2734 	}
2735 	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2736 }
2737 
2738 /* Find next suitable group and run ext4_init_inode_table */
2739 static int ext4_run_li_request(struct ext4_li_request *elr)
2740 {
2741 	struct ext4_group_desc *gdp = NULL;
2742 	ext4_group_t group, ngroups;
2743 	struct super_block *sb;
2744 	unsigned long timeout = 0;
2745 	int ret = 0;
2746 
2747 	sb = elr->lr_super;
2748 	ngroups = EXT4_SB(sb)->s_groups_count;
2749 
2750 	for (group = elr->lr_next_group; group < ngroups; group++) {
2751 		gdp = ext4_get_group_desc(sb, group, NULL);
2752 		if (!gdp) {
2753 			ret = 1;
2754 			break;
2755 		}
2756 
2757 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2758 			break;
2759 	}
2760 
2761 	if (group == ngroups)
2762 		ret = 1;
2763 
2764 	if (!ret) {
2765 		timeout = jiffies;
2766 		ret = ext4_init_inode_table(sb, group,
2767 					    elr->lr_timeout ? 0 : 1);
2768 		if (elr->lr_timeout == 0) {
2769 			timeout = (jiffies - timeout) *
2770 				  elr->lr_sbi->s_li_wait_mult;
2771 			elr->lr_timeout = timeout;
2772 		}
2773 		elr->lr_next_sched = jiffies + elr->lr_timeout;
2774 		elr->lr_next_group = group + 1;
2775 	}
2776 
2777 	return ret;
2778 }
2779 
2780 /*
2781  * Remove lr_request from the list_request and free the
2782  * request structure. Should be called with li_list_mtx held
2783  */
2784 static void ext4_remove_li_request(struct ext4_li_request *elr)
2785 {
2786 	struct ext4_sb_info *sbi;
2787 
2788 	if (!elr)
2789 		return;
2790 
2791 	sbi = elr->lr_sbi;
2792 
2793 	list_del(&elr->lr_request);
2794 	sbi->s_li_request = NULL;
2795 	kfree(elr);
2796 }
2797 
2798 static void ext4_unregister_li_request(struct super_block *sb)
2799 {
2800 	mutex_lock(&ext4_li_mtx);
2801 	if (!ext4_li_info) {
2802 		mutex_unlock(&ext4_li_mtx);
2803 		return;
2804 	}
2805 
2806 	mutex_lock(&ext4_li_info->li_list_mtx);
2807 	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2808 	mutex_unlock(&ext4_li_info->li_list_mtx);
2809 	mutex_unlock(&ext4_li_mtx);
2810 }
2811 
2812 static struct task_struct *ext4_lazyinit_task;
2813 
2814 /*
2815  * This is the function where ext4lazyinit thread lives. It walks
2816  * through the request list searching for next scheduled filesystem.
2817  * When such a fs is found, run the lazy initialization request
2818  * (ext4_rn_li_request) and keep track of the time spend in this
2819  * function. Based on that time we compute next schedule time of
2820  * the request. When walking through the list is complete, compute
2821  * next waking time and put itself into sleep.
2822  */
2823 static int ext4_lazyinit_thread(void *arg)
2824 {
2825 	struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2826 	struct list_head *pos, *n;
2827 	struct ext4_li_request *elr;
2828 	unsigned long next_wakeup, cur;
2829 
2830 	BUG_ON(NULL == eli);
2831 
2832 cont_thread:
2833 	while (true) {
2834 		next_wakeup = MAX_JIFFY_OFFSET;
2835 
2836 		mutex_lock(&eli->li_list_mtx);
2837 		if (list_empty(&eli->li_request_list)) {
2838 			mutex_unlock(&eli->li_list_mtx);
2839 			goto exit_thread;
2840 		}
2841 
2842 		list_for_each_safe(pos, n, &eli->li_request_list) {
2843 			elr = list_entry(pos, struct ext4_li_request,
2844 					 lr_request);
2845 
2846 			if (time_after_eq(jiffies, elr->lr_next_sched)) {
2847 				if (ext4_run_li_request(elr) != 0) {
2848 					/* error, remove the lazy_init job */
2849 					ext4_remove_li_request(elr);
2850 					continue;
2851 				}
2852 			}
2853 
2854 			if (time_before(elr->lr_next_sched, next_wakeup))
2855 				next_wakeup = elr->lr_next_sched;
2856 		}
2857 		mutex_unlock(&eli->li_list_mtx);
2858 
2859 		try_to_freeze();
2860 
2861 		cur = jiffies;
2862 		if ((time_after_eq(cur, next_wakeup)) ||
2863 		    (MAX_JIFFY_OFFSET == next_wakeup)) {
2864 			cond_resched();
2865 			continue;
2866 		}
2867 
2868 		schedule_timeout_interruptible(next_wakeup - cur);
2869 
2870 		if (kthread_should_stop()) {
2871 			ext4_clear_request_list();
2872 			goto exit_thread;
2873 		}
2874 	}
2875 
2876 exit_thread:
2877 	/*
2878 	 * It looks like the request list is empty, but we need
2879 	 * to check it under the li_list_mtx lock, to prevent any
2880 	 * additions into it, and of course we should lock ext4_li_mtx
2881 	 * to atomically free the list and ext4_li_info, because at
2882 	 * this point another ext4 filesystem could be registering
2883 	 * new one.
2884 	 */
2885 	mutex_lock(&ext4_li_mtx);
2886 	mutex_lock(&eli->li_list_mtx);
2887 	if (!list_empty(&eli->li_request_list)) {
2888 		mutex_unlock(&eli->li_list_mtx);
2889 		mutex_unlock(&ext4_li_mtx);
2890 		goto cont_thread;
2891 	}
2892 	mutex_unlock(&eli->li_list_mtx);
2893 	kfree(ext4_li_info);
2894 	ext4_li_info = NULL;
2895 	mutex_unlock(&ext4_li_mtx);
2896 
2897 	return 0;
2898 }
2899 
2900 static void ext4_clear_request_list(void)
2901 {
2902 	struct list_head *pos, *n;
2903 	struct ext4_li_request *elr;
2904 
2905 	mutex_lock(&ext4_li_info->li_list_mtx);
2906 	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2907 		elr = list_entry(pos, struct ext4_li_request,
2908 				 lr_request);
2909 		ext4_remove_li_request(elr);
2910 	}
2911 	mutex_unlock(&ext4_li_info->li_list_mtx);
2912 }
2913 
2914 static int ext4_run_lazyinit_thread(void)
2915 {
2916 	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2917 					 ext4_li_info, "ext4lazyinit");
2918 	if (IS_ERR(ext4_lazyinit_task)) {
2919 		int err = PTR_ERR(ext4_lazyinit_task);
2920 		ext4_clear_request_list();
2921 		kfree(ext4_li_info);
2922 		ext4_li_info = NULL;
2923 		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2924 				 "initialization thread\n",
2925 				 err);
2926 		return err;
2927 	}
2928 	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2929 	return 0;
2930 }
2931 
2932 /*
2933  * Check whether it make sense to run itable init. thread or not.
2934  * If there is at least one uninitialized inode table, return
2935  * corresponding group number, else the loop goes through all
2936  * groups and return total number of groups.
2937  */
2938 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2939 {
2940 	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2941 	struct ext4_group_desc *gdp = NULL;
2942 
2943 	for (group = 0; group < ngroups; group++) {
2944 		gdp = ext4_get_group_desc(sb, group, NULL);
2945 		if (!gdp)
2946 			continue;
2947 
2948 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2949 			break;
2950 	}
2951 
2952 	return group;
2953 }
2954 
2955 static int ext4_li_info_new(void)
2956 {
2957 	struct ext4_lazy_init *eli = NULL;
2958 
2959 	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2960 	if (!eli)
2961 		return -ENOMEM;
2962 
2963 	INIT_LIST_HEAD(&eli->li_request_list);
2964 	mutex_init(&eli->li_list_mtx);
2965 
2966 	eli->li_state |= EXT4_LAZYINIT_QUIT;
2967 
2968 	ext4_li_info = eli;
2969 
2970 	return 0;
2971 }
2972 
2973 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2974 					    ext4_group_t start)
2975 {
2976 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2977 	struct ext4_li_request *elr;
2978 	unsigned long rnd;
2979 
2980 	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2981 	if (!elr)
2982 		return NULL;
2983 
2984 	elr->lr_super = sb;
2985 	elr->lr_sbi = sbi;
2986 	elr->lr_next_group = start;
2987 
2988 	/*
2989 	 * Randomize first schedule time of the request to
2990 	 * spread the inode table initialization requests
2991 	 * better.
2992 	 */
2993 	get_random_bytes(&rnd, sizeof(rnd));
2994 	elr->lr_next_sched = jiffies + (unsigned long)rnd %
2995 			     (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2996 
2997 	return elr;
2998 }
2999 
3000 static int ext4_register_li_request(struct super_block *sb,
3001 				    ext4_group_t first_not_zeroed)
3002 {
3003 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3004 	struct ext4_li_request *elr;
3005 	ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3006 	int ret = 0;
3007 
3008 	if (sbi->s_li_request != NULL) {
3009 		/*
3010 		 * Reset timeout so it can be computed again, because
3011 		 * s_li_wait_mult might have changed.
3012 		 */
3013 		sbi->s_li_request->lr_timeout = 0;
3014 		return 0;
3015 	}
3016 
3017 	if (first_not_zeroed == ngroups ||
3018 	    (sb->s_flags & MS_RDONLY) ||
3019 	    !test_opt(sb, INIT_INODE_TABLE))
3020 		return 0;
3021 
3022 	elr = ext4_li_request_new(sb, first_not_zeroed);
3023 	if (!elr)
3024 		return -ENOMEM;
3025 
3026 	mutex_lock(&ext4_li_mtx);
3027 
3028 	if (NULL == ext4_li_info) {
3029 		ret = ext4_li_info_new();
3030 		if (ret)
3031 			goto out;
3032 	}
3033 
3034 	mutex_lock(&ext4_li_info->li_list_mtx);
3035 	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3036 	mutex_unlock(&ext4_li_info->li_list_mtx);
3037 
3038 	sbi->s_li_request = elr;
3039 	/*
3040 	 * set elr to NULL here since it has been inserted to
3041 	 * the request_list and the removal and free of it is
3042 	 * handled by ext4_clear_request_list from now on.
3043 	 */
3044 	elr = NULL;
3045 
3046 	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3047 		ret = ext4_run_lazyinit_thread();
3048 		if (ret)
3049 			goto out;
3050 	}
3051 out:
3052 	mutex_unlock(&ext4_li_mtx);
3053 	if (ret)
3054 		kfree(elr);
3055 	return ret;
3056 }
3057 
3058 /*
3059  * We do not need to lock anything since this is called on
3060  * module unload.
3061  */
3062 static void ext4_destroy_lazyinit_thread(void)
3063 {
3064 	/*
3065 	 * If thread exited earlier
3066 	 * there's nothing to be done.
3067 	 */
3068 	if (!ext4_li_info || !ext4_lazyinit_task)
3069 		return;
3070 
3071 	kthread_stop(ext4_lazyinit_task);
3072 }
3073 
3074 static int set_journal_csum_feature_set(struct super_block *sb)
3075 {
3076 	int ret = 1;
3077 	int compat, incompat;
3078 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3079 
3080 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3081 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3082 		/* journal checksum v2 */
3083 		compat = 0;
3084 		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3085 	} else {
3086 		/* journal checksum v1 */
3087 		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3088 		incompat = 0;
3089 	}
3090 
3091 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3092 		ret = jbd2_journal_set_features(sbi->s_journal,
3093 				compat, 0,
3094 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3095 				incompat);
3096 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3097 		ret = jbd2_journal_set_features(sbi->s_journal,
3098 				compat, 0,
3099 				incompat);
3100 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3101 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3102 	} else {
3103 		jbd2_journal_clear_features(sbi->s_journal,
3104 				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3105 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3106 				JBD2_FEATURE_INCOMPAT_CSUM_V2);
3107 	}
3108 
3109 	return ret;
3110 }
3111 
3112 /*
3113  * Note: calculating the overhead so we can be compatible with
3114  * historical BSD practice is quite difficult in the face of
3115  * clusters/bigalloc.  This is because multiple metadata blocks from
3116  * different block group can end up in the same allocation cluster.
3117  * Calculating the exact overhead in the face of clustered allocation
3118  * requires either O(all block bitmaps) in memory or O(number of block
3119  * groups**2) in time.  We will still calculate the superblock for
3120  * older file systems --- and if we come across with a bigalloc file
3121  * system with zero in s_overhead_clusters the estimate will be close to
3122  * correct especially for very large cluster sizes --- but for newer
3123  * file systems, it's better to calculate this figure once at mkfs
3124  * time, and store it in the superblock.  If the superblock value is
3125  * present (even for non-bigalloc file systems), we will use it.
3126  */
3127 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3128 			  char *buf)
3129 {
3130 	struct ext4_sb_info	*sbi = EXT4_SB(sb);
3131 	struct ext4_group_desc	*gdp;
3132 	ext4_fsblk_t		first_block, last_block, b;
3133 	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
3134 	int			s, j, count = 0;
3135 
3136 	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3137 		(grp * EXT4_BLOCKS_PER_GROUP(sb));
3138 	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3139 	for (i = 0; i < ngroups; i++) {
3140 		gdp = ext4_get_group_desc(sb, i, NULL);
3141 		b = ext4_block_bitmap(sb, gdp);
3142 		if (b >= first_block && b <= last_block) {
3143 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3144 			count++;
3145 		}
3146 		b = ext4_inode_bitmap(sb, gdp);
3147 		if (b >= first_block && b <= last_block) {
3148 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3149 			count++;
3150 		}
3151 		b = ext4_inode_table(sb, gdp);
3152 		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3153 			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3154 				int c = EXT4_B2C(sbi, b - first_block);
3155 				ext4_set_bit(c, buf);
3156 				count++;
3157 			}
3158 		if (i != grp)
3159 			continue;
3160 		s = 0;
3161 		if (ext4_bg_has_super(sb, grp)) {
3162 			ext4_set_bit(s++, buf);
3163 			count++;
3164 		}
3165 		for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3166 			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3167 			count++;
3168 		}
3169 	}
3170 	if (!count)
3171 		return 0;
3172 	return EXT4_CLUSTERS_PER_GROUP(sb) -
3173 		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3174 }
3175 
3176 /*
3177  * Compute the overhead and stash it in sbi->s_overhead
3178  */
3179 int ext4_calculate_overhead(struct super_block *sb)
3180 {
3181 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3182 	struct ext4_super_block *es = sbi->s_es;
3183 	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3184 	ext4_fsblk_t overhead = 0;
3185 	char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3186 
3187 	memset(buf, 0, PAGE_SIZE);
3188 	if (!buf)
3189 		return -ENOMEM;
3190 
3191 	/*
3192 	 * Compute the overhead (FS structures).  This is constant
3193 	 * for a given filesystem unless the number of block groups
3194 	 * changes so we cache the previous value until it does.
3195 	 */
3196 
3197 	/*
3198 	 * All of the blocks before first_data_block are overhead
3199 	 */
3200 	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3201 
3202 	/*
3203 	 * Add the overhead found in each block group
3204 	 */
3205 	for (i = 0; i < ngroups; i++) {
3206 		int blks;
3207 
3208 		blks = count_overhead(sb, i, buf);
3209 		overhead += blks;
3210 		if (blks)
3211 			memset(buf, 0, PAGE_SIZE);
3212 		cond_resched();
3213 	}
3214 	sbi->s_overhead = overhead;
3215 	smp_wmb();
3216 	free_page((unsigned long) buf);
3217 	return 0;
3218 }
3219 
3220 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3221 {
3222 	char *orig_data = kstrdup(data, GFP_KERNEL);
3223 	struct buffer_head *bh;
3224 	struct ext4_super_block *es = NULL;
3225 	struct ext4_sb_info *sbi;
3226 	ext4_fsblk_t block;
3227 	ext4_fsblk_t sb_block = get_sb_block(&data);
3228 	ext4_fsblk_t logical_sb_block;
3229 	unsigned long offset = 0;
3230 	unsigned long journal_devnum = 0;
3231 	unsigned long def_mount_opts;
3232 	struct inode *root;
3233 	char *cp;
3234 	const char *descr;
3235 	int ret = -ENOMEM;
3236 	int blocksize, clustersize;
3237 	unsigned int db_count;
3238 	unsigned int i;
3239 	int needs_recovery, has_huge_files, has_bigalloc;
3240 	__u64 blocks_count;
3241 	int err;
3242 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3243 	ext4_group_t first_not_zeroed;
3244 
3245 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3246 	if (!sbi)
3247 		goto out_free_orig;
3248 
3249 	sbi->s_blockgroup_lock =
3250 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3251 	if (!sbi->s_blockgroup_lock) {
3252 		kfree(sbi);
3253 		goto out_free_orig;
3254 	}
3255 	sb->s_fs_info = sbi;
3256 	sbi->s_sb = sb;
3257 	sbi->s_mount_opt = 0;
3258 	sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3259 	sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
3260 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3261 	sbi->s_sb_block = sb_block;
3262 	if (sb->s_bdev->bd_part)
3263 		sbi->s_sectors_written_start =
3264 			part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3265 
3266 	/* Cleanup superblock name */
3267 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3268 		*cp = '!';
3269 
3270 	ret = -EINVAL;
3271 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3272 	if (!blocksize) {
3273 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3274 		goto out_fail;
3275 	}
3276 
3277 	/*
3278 	 * The ext4 superblock will not be buffer aligned for other than 1kB
3279 	 * block sizes.  We need to calculate the offset from buffer start.
3280 	 */
3281 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3282 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3283 		offset = do_div(logical_sb_block, blocksize);
3284 	} else {
3285 		logical_sb_block = sb_block;
3286 	}
3287 
3288 	if (!(bh = sb_bread(sb, logical_sb_block))) {
3289 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
3290 		goto out_fail;
3291 	}
3292 	/*
3293 	 * Note: s_es must be initialized as soon as possible because
3294 	 *       some ext4 macro-instructions depend on its value
3295 	 */
3296 	es = (struct ext4_super_block *) (bh->b_data + offset);
3297 	sbi->s_es = es;
3298 	sb->s_magic = le16_to_cpu(es->s_magic);
3299 	if (sb->s_magic != EXT4_SUPER_MAGIC)
3300 		goto cantfind_ext4;
3301 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3302 
3303 	/* Warn if metadata_csum and gdt_csum are both set. */
3304 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3305 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3306 	    EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3307 		ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3308 			     "redundant flags; please run fsck.");
3309 
3310 	/* Check for a known checksum algorithm */
3311 	if (!ext4_verify_csum_type(sb, es)) {
3312 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3313 			 "unknown checksum algorithm.");
3314 		silent = 1;
3315 		goto cantfind_ext4;
3316 	}
3317 
3318 	/* Load the checksum driver */
3319 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3320 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3321 		sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3322 		if (IS_ERR(sbi->s_chksum_driver)) {
3323 			ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3324 			ret = PTR_ERR(sbi->s_chksum_driver);
3325 			sbi->s_chksum_driver = NULL;
3326 			goto failed_mount;
3327 		}
3328 	}
3329 
3330 	/* Check superblock checksum */
3331 	if (!ext4_superblock_csum_verify(sb, es)) {
3332 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3333 			 "invalid superblock checksum.  Run e2fsck?");
3334 		silent = 1;
3335 		goto cantfind_ext4;
3336 	}
3337 
3338 	/* Precompute checksum seed for all metadata */
3339 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3340 			EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3341 		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3342 					       sizeof(es->s_uuid));
3343 
3344 	/* Set defaults before we parse the mount options */
3345 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3346 	set_opt(sb, INIT_INODE_TABLE);
3347 	if (def_mount_opts & EXT4_DEFM_DEBUG)
3348 		set_opt(sb, DEBUG);
3349 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3350 		set_opt(sb, GRPID);
3351 	if (def_mount_opts & EXT4_DEFM_UID16)
3352 		set_opt(sb, NO_UID32);
3353 	/* xattr user namespace & acls are now defaulted on */
3354 #ifdef CONFIG_EXT4_FS_XATTR
3355 	set_opt(sb, XATTR_USER);
3356 #endif
3357 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3358 	set_opt(sb, POSIX_ACL);
3359 #endif
3360 	set_opt(sb, MBLK_IO_SUBMIT);
3361 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3362 		set_opt(sb, JOURNAL_DATA);
3363 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3364 		set_opt(sb, ORDERED_DATA);
3365 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3366 		set_opt(sb, WRITEBACK_DATA);
3367 
3368 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3369 		set_opt(sb, ERRORS_PANIC);
3370 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3371 		set_opt(sb, ERRORS_CONT);
3372 	else
3373 		set_opt(sb, ERRORS_RO);
3374 	if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3375 		set_opt(sb, BLOCK_VALIDITY);
3376 	if (def_mount_opts & EXT4_DEFM_DISCARD)
3377 		set_opt(sb, DISCARD);
3378 
3379 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3380 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3381 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3382 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3383 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3384 
3385 	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3386 		set_opt(sb, BARRIER);
3387 
3388 	/*
3389 	 * enable delayed allocation by default
3390 	 * Use -o nodelalloc to turn it off
3391 	 */
3392 	if (!IS_EXT3_SB(sb) &&
3393 	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3394 		set_opt(sb, DELALLOC);
3395 
3396 	/*
3397 	 * set default s_li_wait_mult for lazyinit, for the case there is
3398 	 * no mount option specified.
3399 	 */
3400 	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3401 
3402 	if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3403 			   &journal_devnum, &journal_ioprio, 0)) {
3404 		ext4_msg(sb, KERN_WARNING,
3405 			 "failed to parse options in superblock: %s",
3406 			 sbi->s_es->s_mount_opts);
3407 	}
3408 	sbi->s_def_mount_opt = sbi->s_mount_opt;
3409 	if (!parse_options((char *) data, sb, &journal_devnum,
3410 			   &journal_ioprio, 0))
3411 		goto failed_mount;
3412 
3413 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3414 		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3415 			    "with data=journal disables delayed "
3416 			    "allocation and O_DIRECT support!\n");
3417 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3418 			ext4_msg(sb, KERN_ERR, "can't mount with "
3419 				 "both data=journal and delalloc");
3420 			goto failed_mount;
3421 		}
3422 		if (test_opt(sb, DIOREAD_NOLOCK)) {
3423 			ext4_msg(sb, KERN_ERR, "can't mount with "
3424 				 "both data=journal and delalloc");
3425 			goto failed_mount;
3426 		}
3427 		if (test_opt(sb, DELALLOC))
3428 			clear_opt(sb, DELALLOC);
3429 	}
3430 
3431 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3432 	if (test_opt(sb, DIOREAD_NOLOCK)) {
3433 		if (blocksize < PAGE_SIZE) {
3434 			ext4_msg(sb, KERN_ERR, "can't mount with "
3435 				 "dioread_nolock if block size != PAGE_SIZE");
3436 			goto failed_mount;
3437 		}
3438 	}
3439 
3440 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3441 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3442 
3443 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3444 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3445 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3446 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3447 		ext4_msg(sb, KERN_WARNING,
3448 		       "feature flags set on rev 0 fs, "
3449 		       "running e2fsck is recommended");
3450 
3451 	if (IS_EXT2_SB(sb)) {
3452 		if (ext2_feature_set_ok(sb))
3453 			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3454 				 "using the ext4 subsystem");
3455 		else {
3456 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3457 				 "to feature incompatibilities");
3458 			goto failed_mount;
3459 		}
3460 	}
3461 
3462 	if (IS_EXT3_SB(sb)) {
3463 		if (ext3_feature_set_ok(sb))
3464 			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3465 				 "using the ext4 subsystem");
3466 		else {
3467 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3468 				 "to feature incompatibilities");
3469 			goto failed_mount;
3470 		}
3471 	}
3472 
3473 	/*
3474 	 * Check feature flags regardless of the revision level, since we
3475 	 * previously didn't change the revision level when setting the flags,
3476 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
3477 	 */
3478 	if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3479 		goto failed_mount;
3480 
3481 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3482 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
3483 		ext4_msg(sb, KERN_ERR,
3484 		       "Unsupported filesystem blocksize %d", blocksize);
3485 		goto failed_mount;
3486 	}
3487 
3488 	if (sb->s_blocksize != blocksize) {
3489 		/* Validate the filesystem blocksize */
3490 		if (!sb_set_blocksize(sb, blocksize)) {
3491 			ext4_msg(sb, KERN_ERR, "bad block size %d",
3492 					blocksize);
3493 			goto failed_mount;
3494 		}
3495 
3496 		brelse(bh);
3497 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3498 		offset = do_div(logical_sb_block, blocksize);
3499 		bh = sb_bread(sb, logical_sb_block);
3500 		if (!bh) {
3501 			ext4_msg(sb, KERN_ERR,
3502 			       "Can't read superblock on 2nd try");
3503 			goto failed_mount;
3504 		}
3505 		es = (struct ext4_super_block *)(bh->b_data + offset);
3506 		sbi->s_es = es;
3507 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3508 			ext4_msg(sb, KERN_ERR,
3509 			       "Magic mismatch, very weird!");
3510 			goto failed_mount;
3511 		}
3512 	}
3513 
3514 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3515 				EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3516 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3517 						      has_huge_files);
3518 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3519 
3520 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3521 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3522 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3523 	} else {
3524 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3525 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3526 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3527 		    (!is_power_of_2(sbi->s_inode_size)) ||
3528 		    (sbi->s_inode_size > blocksize)) {
3529 			ext4_msg(sb, KERN_ERR,
3530 			       "unsupported inode size: %d",
3531 			       sbi->s_inode_size);
3532 			goto failed_mount;
3533 		}
3534 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3535 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3536 	}
3537 
3538 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3539 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3540 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3541 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3542 		    !is_power_of_2(sbi->s_desc_size)) {
3543 			ext4_msg(sb, KERN_ERR,
3544 			       "unsupported descriptor size %lu",
3545 			       sbi->s_desc_size);
3546 			goto failed_mount;
3547 		}
3548 	} else
3549 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3550 
3551 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3552 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3553 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3554 		goto cantfind_ext4;
3555 
3556 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3557 	if (sbi->s_inodes_per_block == 0)
3558 		goto cantfind_ext4;
3559 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
3560 					sbi->s_inodes_per_block;
3561 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3562 	sbi->s_sbh = bh;
3563 	sbi->s_mount_state = le16_to_cpu(es->s_state);
3564 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3565 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3566 
3567 	for (i = 0; i < 4; i++)
3568 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3569 	sbi->s_def_hash_version = es->s_def_hash_version;
3570 	i = le32_to_cpu(es->s_flags);
3571 	if (i & EXT2_FLAGS_UNSIGNED_HASH)
3572 		sbi->s_hash_unsigned = 3;
3573 	else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3574 #ifdef __CHAR_UNSIGNED__
3575 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3576 		sbi->s_hash_unsigned = 3;
3577 #else
3578 		es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3579 #endif
3580 	}
3581 
3582 	/* Handle clustersize */
3583 	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3584 	has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3585 				EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3586 	if (has_bigalloc) {
3587 		if (clustersize < blocksize) {
3588 			ext4_msg(sb, KERN_ERR,
3589 				 "cluster size (%d) smaller than "
3590 				 "block size (%d)", clustersize, blocksize);
3591 			goto failed_mount;
3592 		}
3593 		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3594 			le32_to_cpu(es->s_log_block_size);
3595 		sbi->s_clusters_per_group =
3596 			le32_to_cpu(es->s_clusters_per_group);
3597 		if (sbi->s_clusters_per_group > blocksize * 8) {
3598 			ext4_msg(sb, KERN_ERR,
3599 				 "#clusters per group too big: %lu",
3600 				 sbi->s_clusters_per_group);
3601 			goto failed_mount;
3602 		}
3603 		if (sbi->s_blocks_per_group !=
3604 		    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3605 			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3606 				 "clusters per group (%lu) inconsistent",
3607 				 sbi->s_blocks_per_group,
3608 				 sbi->s_clusters_per_group);
3609 			goto failed_mount;
3610 		}
3611 	} else {
3612 		if (clustersize != blocksize) {
3613 			ext4_warning(sb, "fragment/cluster size (%d) != "
3614 				     "block size (%d)", clustersize,
3615 				     blocksize);
3616 			clustersize = blocksize;
3617 		}
3618 		if (sbi->s_blocks_per_group > blocksize * 8) {
3619 			ext4_msg(sb, KERN_ERR,
3620 				 "#blocks per group too big: %lu",
3621 				 sbi->s_blocks_per_group);
3622 			goto failed_mount;
3623 		}
3624 		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3625 		sbi->s_cluster_bits = 0;
3626 	}
3627 	sbi->s_cluster_ratio = clustersize / blocksize;
3628 
3629 	if (sbi->s_inodes_per_group > blocksize * 8) {
3630 		ext4_msg(sb, KERN_ERR,
3631 		       "#inodes per group too big: %lu",
3632 		       sbi->s_inodes_per_group);
3633 		goto failed_mount;
3634 	}
3635 
3636 	/*
3637 	 * Test whether we have more sectors than will fit in sector_t,
3638 	 * and whether the max offset is addressable by the page cache.
3639 	 */
3640 	err = generic_check_addressable(sb->s_blocksize_bits,
3641 					ext4_blocks_count(es));
3642 	if (err) {
3643 		ext4_msg(sb, KERN_ERR, "filesystem"
3644 			 " too large to mount safely on this system");
3645 		if (sizeof(sector_t) < 8)
3646 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3647 		ret = err;
3648 		goto failed_mount;
3649 	}
3650 
3651 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3652 		goto cantfind_ext4;
3653 
3654 	/* check blocks count against device size */
3655 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3656 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3657 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3658 		       "exceeds size of device (%llu blocks)",
3659 		       ext4_blocks_count(es), blocks_count);
3660 		goto failed_mount;
3661 	}
3662 
3663 	/*
3664 	 * It makes no sense for the first data block to be beyond the end
3665 	 * of the filesystem.
3666 	 */
3667 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3668 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3669 			 "block %u is beyond end of filesystem (%llu)",
3670 			 le32_to_cpu(es->s_first_data_block),
3671 			 ext4_blocks_count(es));
3672 		goto failed_mount;
3673 	}
3674 	blocks_count = (ext4_blocks_count(es) -
3675 			le32_to_cpu(es->s_first_data_block) +
3676 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
3677 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3678 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3679 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3680 		       "(block count %llu, first data block %u, "
3681 		       "blocks per group %lu)", sbi->s_groups_count,
3682 		       ext4_blocks_count(es),
3683 		       le32_to_cpu(es->s_first_data_block),
3684 		       EXT4_BLOCKS_PER_GROUP(sb));
3685 		goto failed_mount;
3686 	}
3687 	sbi->s_groups_count = blocks_count;
3688 	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3689 			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3690 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3691 		   EXT4_DESC_PER_BLOCK(sb);
3692 	sbi->s_group_desc = ext4_kvmalloc(db_count *
3693 					  sizeof(struct buffer_head *),
3694 					  GFP_KERNEL);
3695 	if (sbi->s_group_desc == NULL) {
3696 		ext4_msg(sb, KERN_ERR, "not enough memory");
3697 		ret = -ENOMEM;
3698 		goto failed_mount;
3699 	}
3700 
3701 	if (ext4_proc_root)
3702 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3703 
3704 	if (sbi->s_proc)
3705 		proc_create_data("options", S_IRUGO, sbi->s_proc,
3706 				 &ext4_seq_options_fops, sb);
3707 
3708 	bgl_lock_init(sbi->s_blockgroup_lock);
3709 
3710 	for (i = 0; i < db_count; i++) {
3711 		block = descriptor_loc(sb, logical_sb_block, i);
3712 		sbi->s_group_desc[i] = sb_bread(sb, block);
3713 		if (!sbi->s_group_desc[i]) {
3714 			ext4_msg(sb, KERN_ERR,
3715 			       "can't read group descriptor %d", i);
3716 			db_count = i;
3717 			goto failed_mount2;
3718 		}
3719 	}
3720 	if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3721 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3722 		goto failed_mount2;
3723 	}
3724 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3725 		if (!ext4_fill_flex_info(sb)) {
3726 			ext4_msg(sb, KERN_ERR,
3727 			       "unable to initialize "
3728 			       "flex_bg meta info!");
3729 			goto failed_mount2;
3730 		}
3731 
3732 	sbi->s_gdb_count = db_count;
3733 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3734 	spin_lock_init(&sbi->s_next_gen_lock);
3735 
3736 	init_timer(&sbi->s_err_report);
3737 	sbi->s_err_report.function = print_daily_error_info;
3738 	sbi->s_err_report.data = (unsigned long) sb;
3739 
3740 	err = percpu_counter_init(&sbi->s_freeclusters_counter,
3741 			ext4_count_free_clusters(sb));
3742 	if (!err) {
3743 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
3744 				ext4_count_free_inodes(sb));
3745 	}
3746 	if (!err) {
3747 		err = percpu_counter_init(&sbi->s_dirs_counter,
3748 				ext4_count_dirs(sb));
3749 	}
3750 	if (!err) {
3751 		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3752 	}
3753 	if (err) {
3754 		ext4_msg(sb, KERN_ERR, "insufficient memory");
3755 		ret = err;
3756 		goto failed_mount3;
3757 	}
3758 
3759 	sbi->s_stripe = ext4_get_stripe_size(sbi);
3760 	sbi->s_max_writeback_mb_bump = 128;
3761 
3762 	/*
3763 	 * set up enough so that it can read an inode
3764 	 */
3765 	if (!test_opt(sb, NOLOAD) &&
3766 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3767 		sb->s_op = &ext4_sops;
3768 	else
3769 		sb->s_op = &ext4_nojournal_sops;
3770 	sb->s_export_op = &ext4_export_ops;
3771 	sb->s_xattr = ext4_xattr_handlers;
3772 #ifdef CONFIG_QUOTA
3773 	sb->s_qcop = &ext4_qctl_operations;
3774 	sb->dq_op = &ext4_quota_operations;
3775 
3776 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3777 		/* Use qctl operations for hidden quota files. */
3778 		sb->s_qcop = &ext4_qctl_sysfile_operations;
3779 	}
3780 #endif
3781 	memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3782 
3783 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3784 	mutex_init(&sbi->s_orphan_lock);
3785 	sbi->s_resize_flags = 0;
3786 
3787 	sb->s_root = NULL;
3788 
3789 	needs_recovery = (es->s_last_orphan != 0 ||
3790 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
3791 				    EXT4_FEATURE_INCOMPAT_RECOVER));
3792 
3793 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3794 	    !(sb->s_flags & MS_RDONLY))
3795 		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3796 			goto failed_mount3;
3797 
3798 	/*
3799 	 * The first inode we look at is the journal inode.  Don't try
3800 	 * root first: it may be modified in the journal!
3801 	 */
3802 	if (!test_opt(sb, NOLOAD) &&
3803 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3804 		if (ext4_load_journal(sb, es, journal_devnum))
3805 			goto failed_mount3;
3806 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3807 	      EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3808 		ext4_msg(sb, KERN_ERR, "required journal recovery "
3809 		       "suppressed and not mounted read-only");
3810 		goto failed_mount_wq;
3811 	} else {
3812 		clear_opt(sb, DATA_FLAGS);
3813 		sbi->s_journal = NULL;
3814 		needs_recovery = 0;
3815 		goto no_journal;
3816 	}
3817 
3818 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3819 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3820 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
3821 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3822 		goto failed_mount_wq;
3823 	}
3824 
3825 	if (!set_journal_csum_feature_set(sb)) {
3826 		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3827 			 "feature set");
3828 		goto failed_mount_wq;
3829 	}
3830 
3831 	/* We have now updated the journal if required, so we can
3832 	 * validate the data journaling mode. */
3833 	switch (test_opt(sb, DATA_FLAGS)) {
3834 	case 0:
3835 		/* No mode set, assume a default based on the journal
3836 		 * capabilities: ORDERED_DATA if the journal can
3837 		 * cope, else JOURNAL_DATA
3838 		 */
3839 		if (jbd2_journal_check_available_features
3840 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3841 			set_opt(sb, ORDERED_DATA);
3842 		else
3843 			set_opt(sb, JOURNAL_DATA);
3844 		break;
3845 
3846 	case EXT4_MOUNT_ORDERED_DATA:
3847 	case EXT4_MOUNT_WRITEBACK_DATA:
3848 		if (!jbd2_journal_check_available_features
3849 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3850 			ext4_msg(sb, KERN_ERR, "Journal does not support "
3851 			       "requested data journaling mode");
3852 			goto failed_mount_wq;
3853 		}
3854 	default:
3855 		break;
3856 	}
3857 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3858 
3859 	sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3860 
3861 	/*
3862 	 * The journal may have updated the bg summary counts, so we
3863 	 * need to update the global counters.
3864 	 */
3865 	percpu_counter_set(&sbi->s_freeclusters_counter,
3866 			   ext4_count_free_clusters(sb));
3867 	percpu_counter_set(&sbi->s_freeinodes_counter,
3868 			   ext4_count_free_inodes(sb));
3869 	percpu_counter_set(&sbi->s_dirs_counter,
3870 			   ext4_count_dirs(sb));
3871 	percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3872 
3873 no_journal:
3874 	/*
3875 	 * Get the # of file system overhead blocks from the
3876 	 * superblock if present.
3877 	 */
3878 	if (es->s_overhead_clusters)
3879 		sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3880 	else {
3881 		ret = ext4_calculate_overhead(sb);
3882 		if (ret)
3883 			goto failed_mount_wq;
3884 	}
3885 
3886 	/*
3887 	 * The maximum number of concurrent works can be high and
3888 	 * concurrency isn't really necessary.  Limit it to 1.
3889 	 */
3890 	EXT4_SB(sb)->dio_unwritten_wq =
3891 		alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3892 	if (!EXT4_SB(sb)->dio_unwritten_wq) {
3893 		printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3894 		goto failed_mount_wq;
3895 	}
3896 
3897 	/*
3898 	 * The jbd2_journal_load will have done any necessary log recovery,
3899 	 * so we can safely mount the rest of the filesystem now.
3900 	 */
3901 
3902 	root = ext4_iget(sb, EXT4_ROOT_INO);
3903 	if (IS_ERR(root)) {
3904 		ext4_msg(sb, KERN_ERR, "get root inode failed");
3905 		ret = PTR_ERR(root);
3906 		root = NULL;
3907 		goto failed_mount4;
3908 	}
3909 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3910 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3911 		iput(root);
3912 		goto failed_mount4;
3913 	}
3914 	sb->s_root = d_make_root(root);
3915 	if (!sb->s_root) {
3916 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
3917 		ret = -ENOMEM;
3918 		goto failed_mount4;
3919 	}
3920 
3921 	if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3922 		sb->s_flags |= MS_RDONLY;
3923 
3924 	/* determine the minimum size of new large inodes, if present */
3925 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3926 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3927 						     EXT4_GOOD_OLD_INODE_SIZE;
3928 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3929 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3930 			if (sbi->s_want_extra_isize <
3931 			    le16_to_cpu(es->s_want_extra_isize))
3932 				sbi->s_want_extra_isize =
3933 					le16_to_cpu(es->s_want_extra_isize);
3934 			if (sbi->s_want_extra_isize <
3935 			    le16_to_cpu(es->s_min_extra_isize))
3936 				sbi->s_want_extra_isize =
3937 					le16_to_cpu(es->s_min_extra_isize);
3938 		}
3939 	}
3940 	/* Check if enough inode space is available */
3941 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3942 							sbi->s_inode_size) {
3943 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3944 						       EXT4_GOOD_OLD_INODE_SIZE;
3945 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
3946 			 "available");
3947 	}
3948 
3949 	err = ext4_setup_system_zone(sb);
3950 	if (err) {
3951 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
3952 			 "zone (%d)", err);
3953 		goto failed_mount4a;
3954 	}
3955 
3956 	ext4_ext_init(sb);
3957 	err = ext4_mb_init(sb);
3958 	if (err) {
3959 		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3960 			 err);
3961 		goto failed_mount5;
3962 	}
3963 
3964 	err = ext4_register_li_request(sb, first_not_zeroed);
3965 	if (err)
3966 		goto failed_mount6;
3967 
3968 	sbi->s_kobj.kset = ext4_kset;
3969 	init_completion(&sbi->s_kobj_unregister);
3970 	err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3971 				   "%s", sb->s_id);
3972 	if (err)
3973 		goto failed_mount7;
3974 
3975 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3976 	ext4_orphan_cleanup(sb, es);
3977 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3978 	if (needs_recovery) {
3979 		ext4_msg(sb, KERN_INFO, "recovery complete");
3980 		ext4_mark_recovery_complete(sb, es);
3981 	}
3982 	if (EXT4_SB(sb)->s_journal) {
3983 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3984 			descr = " journalled data mode";
3985 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3986 			descr = " ordered data mode";
3987 		else
3988 			descr = " writeback data mode";
3989 	} else
3990 		descr = "out journal";
3991 
3992 #ifdef CONFIG_QUOTA
3993 	/* Enable quota usage during mount. */
3994 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3995 	    !(sb->s_flags & MS_RDONLY)) {
3996 		ret = ext4_enable_quotas(sb);
3997 		if (ret)
3998 			goto failed_mount7;
3999 	}
4000 #endif  /* CONFIG_QUOTA */
4001 
4002 	ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4003 		 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4004 		 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4005 
4006 	if (es->s_error_count)
4007 		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4008 
4009 	kfree(orig_data);
4010 	return 0;
4011 
4012 cantfind_ext4:
4013 	if (!silent)
4014 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4015 	goto failed_mount;
4016 
4017 failed_mount7:
4018 	ext4_unregister_li_request(sb);
4019 failed_mount6:
4020 	ext4_mb_release(sb);
4021 failed_mount5:
4022 	ext4_ext_release(sb);
4023 	ext4_release_system_zone(sb);
4024 failed_mount4a:
4025 	dput(sb->s_root);
4026 	sb->s_root = NULL;
4027 failed_mount4:
4028 	ext4_msg(sb, KERN_ERR, "mount failed");
4029 	destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4030 failed_mount_wq:
4031 	if (sbi->s_journal) {
4032 		jbd2_journal_destroy(sbi->s_journal);
4033 		sbi->s_journal = NULL;
4034 	}
4035 failed_mount3:
4036 	del_timer(&sbi->s_err_report);
4037 	if (sbi->s_flex_groups)
4038 		ext4_kvfree(sbi->s_flex_groups);
4039 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
4040 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
4041 	percpu_counter_destroy(&sbi->s_dirs_counter);
4042 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4043 	if (sbi->s_mmp_tsk)
4044 		kthread_stop(sbi->s_mmp_tsk);
4045 failed_mount2:
4046 	for (i = 0; i < db_count; i++)
4047 		brelse(sbi->s_group_desc[i]);
4048 	ext4_kvfree(sbi->s_group_desc);
4049 failed_mount:
4050 	if (sbi->s_chksum_driver)
4051 		crypto_free_shash(sbi->s_chksum_driver);
4052 	if (sbi->s_proc) {
4053 		remove_proc_entry("options", sbi->s_proc);
4054 		remove_proc_entry(sb->s_id, ext4_proc_root);
4055 	}
4056 #ifdef CONFIG_QUOTA
4057 	for (i = 0; i < MAXQUOTAS; i++)
4058 		kfree(sbi->s_qf_names[i]);
4059 #endif
4060 	ext4_blkdev_remove(sbi);
4061 	brelse(bh);
4062 out_fail:
4063 	sb->s_fs_info = NULL;
4064 	kfree(sbi->s_blockgroup_lock);
4065 	kfree(sbi);
4066 out_free_orig:
4067 	kfree(orig_data);
4068 	return ret;
4069 }
4070 
4071 /*
4072  * Setup any per-fs journal parameters now.  We'll do this both on
4073  * initial mount, once the journal has been initialised but before we've
4074  * done any recovery; and again on any subsequent remount.
4075  */
4076 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4077 {
4078 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4079 
4080 	journal->j_commit_interval = sbi->s_commit_interval;
4081 	journal->j_min_batch_time = sbi->s_min_batch_time;
4082 	journal->j_max_batch_time = sbi->s_max_batch_time;
4083 
4084 	write_lock(&journal->j_state_lock);
4085 	if (test_opt(sb, BARRIER))
4086 		journal->j_flags |= JBD2_BARRIER;
4087 	else
4088 		journal->j_flags &= ~JBD2_BARRIER;
4089 	if (test_opt(sb, DATA_ERR_ABORT))
4090 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4091 	else
4092 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4093 	write_unlock(&journal->j_state_lock);
4094 }
4095 
4096 static journal_t *ext4_get_journal(struct super_block *sb,
4097 				   unsigned int journal_inum)
4098 {
4099 	struct inode *journal_inode;
4100 	journal_t *journal;
4101 
4102 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4103 
4104 	/* First, test for the existence of a valid inode on disk.  Bad
4105 	 * things happen if we iget() an unused inode, as the subsequent
4106 	 * iput() will try to delete it. */
4107 
4108 	journal_inode = ext4_iget(sb, journal_inum);
4109 	if (IS_ERR(journal_inode)) {
4110 		ext4_msg(sb, KERN_ERR, "no journal found");
4111 		return NULL;
4112 	}
4113 	if (!journal_inode->i_nlink) {
4114 		make_bad_inode(journal_inode);
4115 		iput(journal_inode);
4116 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4117 		return NULL;
4118 	}
4119 
4120 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4121 		  journal_inode, journal_inode->i_size);
4122 	if (!S_ISREG(journal_inode->i_mode)) {
4123 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
4124 		iput(journal_inode);
4125 		return NULL;
4126 	}
4127 
4128 	journal = jbd2_journal_init_inode(journal_inode);
4129 	if (!journal) {
4130 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4131 		iput(journal_inode);
4132 		return NULL;
4133 	}
4134 	journal->j_private = sb;
4135 	ext4_init_journal_params(sb, journal);
4136 	return journal;
4137 }
4138 
4139 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4140 				       dev_t j_dev)
4141 {
4142 	struct buffer_head *bh;
4143 	journal_t *journal;
4144 	ext4_fsblk_t start;
4145 	ext4_fsblk_t len;
4146 	int hblock, blocksize;
4147 	ext4_fsblk_t sb_block;
4148 	unsigned long offset;
4149 	struct ext4_super_block *es;
4150 	struct block_device *bdev;
4151 
4152 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4153 
4154 	bdev = ext4_blkdev_get(j_dev, sb);
4155 	if (bdev == NULL)
4156 		return NULL;
4157 
4158 	blocksize = sb->s_blocksize;
4159 	hblock = bdev_logical_block_size(bdev);
4160 	if (blocksize < hblock) {
4161 		ext4_msg(sb, KERN_ERR,
4162 			"blocksize too small for journal device");
4163 		goto out_bdev;
4164 	}
4165 
4166 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4167 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4168 	set_blocksize(bdev, blocksize);
4169 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
4170 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4171 		       "external journal");
4172 		goto out_bdev;
4173 	}
4174 
4175 	es = (struct ext4_super_block *) (bh->b_data + offset);
4176 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4177 	    !(le32_to_cpu(es->s_feature_incompat) &
4178 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4179 		ext4_msg(sb, KERN_ERR, "external journal has "
4180 					"bad superblock");
4181 		brelse(bh);
4182 		goto out_bdev;
4183 	}
4184 
4185 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4186 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4187 		brelse(bh);
4188 		goto out_bdev;
4189 	}
4190 
4191 	len = ext4_blocks_count(es);
4192 	start = sb_block + 1;
4193 	brelse(bh);	/* we're done with the superblock */
4194 
4195 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4196 					start, len, blocksize);
4197 	if (!journal) {
4198 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
4199 		goto out_bdev;
4200 	}
4201 	journal->j_private = sb;
4202 	ll_rw_block(READ, 1, &journal->j_sb_buffer);
4203 	wait_on_buffer(journal->j_sb_buffer);
4204 	if (!buffer_uptodate(journal->j_sb_buffer)) {
4205 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4206 		goto out_journal;
4207 	}
4208 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4209 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
4210 					"user (unsupported) - %d",
4211 			be32_to_cpu(journal->j_superblock->s_nr_users));
4212 		goto out_journal;
4213 	}
4214 	EXT4_SB(sb)->journal_bdev = bdev;
4215 	ext4_init_journal_params(sb, journal);
4216 	return journal;
4217 
4218 out_journal:
4219 	jbd2_journal_destroy(journal);
4220 out_bdev:
4221 	ext4_blkdev_put(bdev);
4222 	return NULL;
4223 }
4224 
4225 static int ext4_load_journal(struct super_block *sb,
4226 			     struct ext4_super_block *es,
4227 			     unsigned long journal_devnum)
4228 {
4229 	journal_t *journal;
4230 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4231 	dev_t journal_dev;
4232 	int err = 0;
4233 	int really_read_only;
4234 
4235 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4236 
4237 	if (journal_devnum &&
4238 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4239 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4240 			"numbers have changed");
4241 		journal_dev = new_decode_dev(journal_devnum);
4242 	} else
4243 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4244 
4245 	really_read_only = bdev_read_only(sb->s_bdev);
4246 
4247 	/*
4248 	 * Are we loading a blank journal or performing recovery after a
4249 	 * crash?  For recovery, we need to check in advance whether we
4250 	 * can get read-write access to the device.
4251 	 */
4252 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4253 		if (sb->s_flags & MS_RDONLY) {
4254 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
4255 					"required on readonly filesystem");
4256 			if (really_read_only) {
4257 				ext4_msg(sb, KERN_ERR, "write access "
4258 					"unavailable, cannot proceed");
4259 				return -EROFS;
4260 			}
4261 			ext4_msg(sb, KERN_INFO, "write access will "
4262 			       "be enabled during recovery");
4263 		}
4264 	}
4265 
4266 	if (journal_inum && journal_dev) {
4267 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4268 		       "and inode journals!");
4269 		return -EINVAL;
4270 	}
4271 
4272 	if (journal_inum) {
4273 		if (!(journal = ext4_get_journal(sb, journal_inum)))
4274 			return -EINVAL;
4275 	} else {
4276 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4277 			return -EINVAL;
4278 	}
4279 
4280 	if (!(journal->j_flags & JBD2_BARRIER))
4281 		ext4_msg(sb, KERN_INFO, "barriers disabled");
4282 
4283 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4284 		err = jbd2_journal_wipe(journal, !really_read_only);
4285 	if (!err) {
4286 		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4287 		if (save)
4288 			memcpy(save, ((char *) es) +
4289 			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4290 		err = jbd2_journal_load(journal);
4291 		if (save)
4292 			memcpy(((char *) es) + EXT4_S_ERR_START,
4293 			       save, EXT4_S_ERR_LEN);
4294 		kfree(save);
4295 	}
4296 
4297 	if (err) {
4298 		ext4_msg(sb, KERN_ERR, "error loading journal");
4299 		jbd2_journal_destroy(journal);
4300 		return err;
4301 	}
4302 
4303 	EXT4_SB(sb)->s_journal = journal;
4304 	ext4_clear_journal_err(sb, es);
4305 
4306 	if (!really_read_only && journal_devnum &&
4307 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4308 		es->s_journal_dev = cpu_to_le32(journal_devnum);
4309 
4310 		/* Make sure we flush the recovery flag to disk. */
4311 		ext4_commit_super(sb, 1);
4312 	}
4313 
4314 	return 0;
4315 }
4316 
4317 static int ext4_commit_super(struct super_block *sb, int sync)
4318 {
4319 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4320 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4321 	int error = 0;
4322 
4323 	if (!sbh || block_device_ejected(sb))
4324 		return error;
4325 	if (buffer_write_io_error(sbh)) {
4326 		/*
4327 		 * Oh, dear.  A previous attempt to write the
4328 		 * superblock failed.  This could happen because the
4329 		 * USB device was yanked out.  Or it could happen to
4330 		 * be a transient write error and maybe the block will
4331 		 * be remapped.  Nothing we can do but to retry the
4332 		 * write and hope for the best.
4333 		 */
4334 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
4335 		       "superblock detected");
4336 		clear_buffer_write_io_error(sbh);
4337 		set_buffer_uptodate(sbh);
4338 	}
4339 	/*
4340 	 * If the file system is mounted read-only, don't update the
4341 	 * superblock write time.  This avoids updating the superblock
4342 	 * write time when we are mounting the root file system
4343 	 * read/only but we need to replay the journal; at that point,
4344 	 * for people who are east of GMT and who make their clock
4345 	 * tick in localtime for Windows bug-for-bug compatibility,
4346 	 * the clock is set in the future, and this will cause e2fsck
4347 	 * to complain and force a full file system check.
4348 	 */
4349 	if (!(sb->s_flags & MS_RDONLY))
4350 		es->s_wtime = cpu_to_le32(get_seconds());
4351 	if (sb->s_bdev->bd_part)
4352 		es->s_kbytes_written =
4353 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4354 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4355 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
4356 	else
4357 		es->s_kbytes_written =
4358 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4359 	ext4_free_blocks_count_set(es,
4360 			EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4361 				&EXT4_SB(sb)->s_freeclusters_counter)));
4362 	es->s_free_inodes_count =
4363 		cpu_to_le32(percpu_counter_sum_positive(
4364 				&EXT4_SB(sb)->s_freeinodes_counter));
4365 	BUFFER_TRACE(sbh, "marking dirty");
4366 	ext4_superblock_csum_set(sb, es);
4367 	mark_buffer_dirty(sbh);
4368 	if (sync) {
4369 		error = sync_dirty_buffer(sbh);
4370 		if (error)
4371 			return error;
4372 
4373 		error = buffer_write_io_error(sbh);
4374 		if (error) {
4375 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
4376 			       "superblock");
4377 			clear_buffer_write_io_error(sbh);
4378 			set_buffer_uptodate(sbh);
4379 		}
4380 	}
4381 	return error;
4382 }
4383 
4384 /*
4385  * Have we just finished recovery?  If so, and if we are mounting (or
4386  * remounting) the filesystem readonly, then we will end up with a
4387  * consistent fs on disk.  Record that fact.
4388  */
4389 static void ext4_mark_recovery_complete(struct super_block *sb,
4390 					struct ext4_super_block *es)
4391 {
4392 	journal_t *journal = EXT4_SB(sb)->s_journal;
4393 
4394 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4395 		BUG_ON(journal != NULL);
4396 		return;
4397 	}
4398 	jbd2_journal_lock_updates(journal);
4399 	if (jbd2_journal_flush(journal) < 0)
4400 		goto out;
4401 
4402 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4403 	    sb->s_flags & MS_RDONLY) {
4404 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4405 		ext4_commit_super(sb, 1);
4406 	}
4407 
4408 out:
4409 	jbd2_journal_unlock_updates(journal);
4410 }
4411 
4412 /*
4413  * If we are mounting (or read-write remounting) a filesystem whose journal
4414  * has recorded an error from a previous lifetime, move that error to the
4415  * main filesystem now.
4416  */
4417 static void ext4_clear_journal_err(struct super_block *sb,
4418 				   struct ext4_super_block *es)
4419 {
4420 	journal_t *journal;
4421 	int j_errno;
4422 	const char *errstr;
4423 
4424 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4425 
4426 	journal = EXT4_SB(sb)->s_journal;
4427 
4428 	/*
4429 	 * Now check for any error status which may have been recorded in the
4430 	 * journal by a prior ext4_error() or ext4_abort()
4431 	 */
4432 
4433 	j_errno = jbd2_journal_errno(journal);
4434 	if (j_errno) {
4435 		char nbuf[16];
4436 
4437 		errstr = ext4_decode_error(sb, j_errno, nbuf);
4438 		ext4_warning(sb, "Filesystem error recorded "
4439 			     "from previous mount: %s", errstr);
4440 		ext4_warning(sb, "Marking fs in need of filesystem check.");
4441 
4442 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4443 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4444 		ext4_commit_super(sb, 1);
4445 
4446 		jbd2_journal_clear_err(journal);
4447 	}
4448 }
4449 
4450 /*
4451  * Force the running and committing transactions to commit,
4452  * and wait on the commit.
4453  */
4454 int ext4_force_commit(struct super_block *sb)
4455 {
4456 	journal_t *journal;
4457 	int ret = 0;
4458 
4459 	if (sb->s_flags & MS_RDONLY)
4460 		return 0;
4461 
4462 	journal = EXT4_SB(sb)->s_journal;
4463 	if (journal) {
4464 		vfs_check_frozen(sb, SB_FREEZE_TRANS);
4465 		ret = ext4_journal_force_commit(journal);
4466 	}
4467 
4468 	return ret;
4469 }
4470 
4471 static int ext4_sync_fs(struct super_block *sb, int wait)
4472 {
4473 	int ret = 0;
4474 	tid_t target;
4475 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4476 
4477 	trace_ext4_sync_fs(sb, wait);
4478 	flush_workqueue(sbi->dio_unwritten_wq);
4479 	/*
4480 	 * Writeback quota in non-journalled quota case - journalled quota has
4481 	 * no dirty dquots
4482 	 */
4483 	dquot_writeback_dquots(sb, -1);
4484 	if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4485 		if (wait)
4486 			jbd2_log_wait_commit(sbi->s_journal, target);
4487 	}
4488 	return ret;
4489 }
4490 
4491 /*
4492  * LVM calls this function before a (read-only) snapshot is created.  This
4493  * gives us a chance to flush the journal completely and mark the fs clean.
4494  *
4495  * Note that only this function cannot bring a filesystem to be in a clean
4496  * state independently, because ext4 prevents a new handle from being started
4497  * by @sb->s_frozen, which stays in an upper layer.  It thus needs help from
4498  * the upper layer.
4499  */
4500 static int ext4_freeze(struct super_block *sb)
4501 {
4502 	int error = 0;
4503 	journal_t *journal;
4504 
4505 	if (sb->s_flags & MS_RDONLY)
4506 		return 0;
4507 
4508 	journal = EXT4_SB(sb)->s_journal;
4509 
4510 	/* Now we set up the journal barrier. */
4511 	jbd2_journal_lock_updates(journal);
4512 
4513 	/*
4514 	 * Don't clear the needs_recovery flag if we failed to flush
4515 	 * the journal.
4516 	 */
4517 	error = jbd2_journal_flush(journal);
4518 	if (error < 0)
4519 		goto out;
4520 
4521 	/* Journal blocked and flushed, clear needs_recovery flag. */
4522 	EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4523 	error = ext4_commit_super(sb, 1);
4524 out:
4525 	/* we rely on s_frozen to stop further updates */
4526 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4527 	return error;
4528 }
4529 
4530 /*
4531  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4532  * flag here, even though the filesystem is not technically dirty yet.
4533  */
4534 static int ext4_unfreeze(struct super_block *sb)
4535 {
4536 	if (sb->s_flags & MS_RDONLY)
4537 		return 0;
4538 
4539 	lock_super(sb);
4540 	/* Reset the needs_recovery flag before the fs is unlocked. */
4541 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4542 	ext4_commit_super(sb, 1);
4543 	unlock_super(sb);
4544 	return 0;
4545 }
4546 
4547 /*
4548  * Structure to save mount options for ext4_remount's benefit
4549  */
4550 struct ext4_mount_options {
4551 	unsigned long s_mount_opt;
4552 	unsigned long s_mount_opt2;
4553 	kuid_t s_resuid;
4554 	kgid_t s_resgid;
4555 	unsigned long s_commit_interval;
4556 	u32 s_min_batch_time, s_max_batch_time;
4557 #ifdef CONFIG_QUOTA
4558 	int s_jquota_fmt;
4559 	char *s_qf_names[MAXQUOTAS];
4560 #endif
4561 };
4562 
4563 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4564 {
4565 	struct ext4_super_block *es;
4566 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4567 	unsigned long old_sb_flags;
4568 	struct ext4_mount_options old_opts;
4569 	int enable_quota = 0;
4570 	ext4_group_t g;
4571 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4572 	int err = 0;
4573 #ifdef CONFIG_QUOTA
4574 	int i;
4575 #endif
4576 	char *orig_data = kstrdup(data, GFP_KERNEL);
4577 
4578 	/* Store the original options */
4579 	lock_super(sb);
4580 	old_sb_flags = sb->s_flags;
4581 	old_opts.s_mount_opt = sbi->s_mount_opt;
4582 	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4583 	old_opts.s_resuid = sbi->s_resuid;
4584 	old_opts.s_resgid = sbi->s_resgid;
4585 	old_opts.s_commit_interval = sbi->s_commit_interval;
4586 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
4587 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
4588 #ifdef CONFIG_QUOTA
4589 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4590 	for (i = 0; i < MAXQUOTAS; i++)
4591 		old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4592 #endif
4593 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4594 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4595 
4596 	/*
4597 	 * Allow the "check" option to be passed as a remount option.
4598 	 */
4599 	if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4600 		err = -EINVAL;
4601 		goto restore_opts;
4602 	}
4603 
4604 	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4605 		ext4_abort(sb, "Abort forced by user");
4606 
4607 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4608 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4609 
4610 	es = sbi->s_es;
4611 
4612 	if (sbi->s_journal) {
4613 		ext4_init_journal_params(sb, sbi->s_journal);
4614 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4615 	}
4616 
4617 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4618 		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4619 			err = -EROFS;
4620 			goto restore_opts;
4621 		}
4622 
4623 		if (*flags & MS_RDONLY) {
4624 			err = dquot_suspend(sb, -1);
4625 			if (err < 0)
4626 				goto restore_opts;
4627 
4628 			/*
4629 			 * First of all, the unconditional stuff we have to do
4630 			 * to disable replay of the journal when we next remount
4631 			 */
4632 			sb->s_flags |= MS_RDONLY;
4633 
4634 			/*
4635 			 * OK, test if we are remounting a valid rw partition
4636 			 * readonly, and if so set the rdonly flag and then
4637 			 * mark the partition as valid again.
4638 			 */
4639 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4640 			    (sbi->s_mount_state & EXT4_VALID_FS))
4641 				es->s_state = cpu_to_le16(sbi->s_mount_state);
4642 
4643 			if (sbi->s_journal)
4644 				ext4_mark_recovery_complete(sb, es);
4645 		} else {
4646 			/* Make sure we can mount this feature set readwrite */
4647 			if (!ext4_feature_set_ok(sb, 0)) {
4648 				err = -EROFS;
4649 				goto restore_opts;
4650 			}
4651 			/*
4652 			 * Make sure the group descriptor checksums
4653 			 * are sane.  If they aren't, refuse to remount r/w.
4654 			 */
4655 			for (g = 0; g < sbi->s_groups_count; g++) {
4656 				struct ext4_group_desc *gdp =
4657 					ext4_get_group_desc(sb, g, NULL);
4658 
4659 				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4660 					ext4_msg(sb, KERN_ERR,
4661 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
4662 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4663 					       le16_to_cpu(gdp->bg_checksum));
4664 					err = -EINVAL;
4665 					goto restore_opts;
4666 				}
4667 			}
4668 
4669 			/*
4670 			 * If we have an unprocessed orphan list hanging
4671 			 * around from a previously readonly bdev mount,
4672 			 * require a full umount/remount for now.
4673 			 */
4674 			if (es->s_last_orphan) {
4675 				ext4_msg(sb, KERN_WARNING, "Couldn't "
4676 				       "remount RDWR because of unprocessed "
4677 				       "orphan inode list.  Please "
4678 				       "umount/remount instead");
4679 				err = -EINVAL;
4680 				goto restore_opts;
4681 			}
4682 
4683 			/*
4684 			 * Mounting a RDONLY partition read-write, so reread
4685 			 * and store the current valid flag.  (It may have
4686 			 * been changed by e2fsck since we originally mounted
4687 			 * the partition.)
4688 			 */
4689 			if (sbi->s_journal)
4690 				ext4_clear_journal_err(sb, es);
4691 			sbi->s_mount_state = le16_to_cpu(es->s_state);
4692 			if (!ext4_setup_super(sb, es, 0))
4693 				sb->s_flags &= ~MS_RDONLY;
4694 			if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4695 						     EXT4_FEATURE_INCOMPAT_MMP))
4696 				if (ext4_multi_mount_protect(sb,
4697 						le64_to_cpu(es->s_mmp_block))) {
4698 					err = -EROFS;
4699 					goto restore_opts;
4700 				}
4701 			enable_quota = 1;
4702 		}
4703 	}
4704 
4705 	/*
4706 	 * Reinitialize lazy itable initialization thread based on
4707 	 * current settings
4708 	 */
4709 	if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4710 		ext4_unregister_li_request(sb);
4711 	else {
4712 		ext4_group_t first_not_zeroed;
4713 		first_not_zeroed = ext4_has_uninit_itable(sb);
4714 		ext4_register_li_request(sb, first_not_zeroed);
4715 	}
4716 
4717 	ext4_setup_system_zone(sb);
4718 	if (sbi->s_journal == NULL)
4719 		ext4_commit_super(sb, 1);
4720 
4721 	unlock_super(sb);
4722 #ifdef CONFIG_QUOTA
4723 	/* Release old quota file names */
4724 	for (i = 0; i < MAXQUOTAS; i++)
4725 		if (old_opts.s_qf_names[i] &&
4726 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4727 			kfree(old_opts.s_qf_names[i]);
4728 	if (enable_quota) {
4729 		if (sb_any_quota_suspended(sb))
4730 			dquot_resume(sb, -1);
4731 		else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4732 					EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4733 			err = ext4_enable_quotas(sb);
4734 			if (err) {
4735 				lock_super(sb);
4736 				goto restore_opts;
4737 			}
4738 		}
4739 	}
4740 #endif
4741 
4742 	ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4743 	kfree(orig_data);
4744 	return 0;
4745 
4746 restore_opts:
4747 	sb->s_flags = old_sb_flags;
4748 	sbi->s_mount_opt = old_opts.s_mount_opt;
4749 	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4750 	sbi->s_resuid = old_opts.s_resuid;
4751 	sbi->s_resgid = old_opts.s_resgid;
4752 	sbi->s_commit_interval = old_opts.s_commit_interval;
4753 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
4754 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
4755 #ifdef CONFIG_QUOTA
4756 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4757 	for (i = 0; i < MAXQUOTAS; i++) {
4758 		if (sbi->s_qf_names[i] &&
4759 		    old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4760 			kfree(sbi->s_qf_names[i]);
4761 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4762 	}
4763 #endif
4764 	unlock_super(sb);
4765 	kfree(orig_data);
4766 	return err;
4767 }
4768 
4769 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4770 {
4771 	struct super_block *sb = dentry->d_sb;
4772 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4773 	struct ext4_super_block *es = sbi->s_es;
4774 	ext4_fsblk_t overhead = 0;
4775 	u64 fsid;
4776 	s64 bfree;
4777 
4778 	if (!test_opt(sb, MINIX_DF))
4779 		overhead = sbi->s_overhead;
4780 
4781 	buf->f_type = EXT4_SUPER_MAGIC;
4782 	buf->f_bsize = sb->s_blocksize;
4783 	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
4784 	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4785 		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4786 	/* prevent underflow in case that few free space is available */
4787 	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4788 	buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4789 	if (buf->f_bfree < ext4_r_blocks_count(es))
4790 		buf->f_bavail = 0;
4791 	buf->f_files = le32_to_cpu(es->s_inodes_count);
4792 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4793 	buf->f_namelen = EXT4_NAME_LEN;
4794 	fsid = le64_to_cpup((void *)es->s_uuid) ^
4795 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4796 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4797 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4798 
4799 	return 0;
4800 }
4801 
4802 /* Helper function for writing quotas on sync - we need to start transaction
4803  * before quota file is locked for write. Otherwise the are possible deadlocks:
4804  * Process 1                         Process 2
4805  * ext4_create()                     quota_sync()
4806  *   jbd2_journal_start()                  write_dquot()
4807  *   dquot_initialize()                         down(dqio_mutex)
4808  *     down(dqio_mutex)                    jbd2_journal_start()
4809  *
4810  */
4811 
4812 #ifdef CONFIG_QUOTA
4813 
4814 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4815 {
4816 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4817 }
4818 
4819 static int ext4_write_dquot(struct dquot *dquot)
4820 {
4821 	int ret, err;
4822 	handle_t *handle;
4823 	struct inode *inode;
4824 
4825 	inode = dquot_to_inode(dquot);
4826 	handle = ext4_journal_start(inode,
4827 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4828 	if (IS_ERR(handle))
4829 		return PTR_ERR(handle);
4830 	ret = dquot_commit(dquot);
4831 	err = ext4_journal_stop(handle);
4832 	if (!ret)
4833 		ret = err;
4834 	return ret;
4835 }
4836 
4837 static int ext4_acquire_dquot(struct dquot *dquot)
4838 {
4839 	int ret, err;
4840 	handle_t *handle;
4841 
4842 	handle = ext4_journal_start(dquot_to_inode(dquot),
4843 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4844 	if (IS_ERR(handle))
4845 		return PTR_ERR(handle);
4846 	ret = dquot_acquire(dquot);
4847 	err = ext4_journal_stop(handle);
4848 	if (!ret)
4849 		ret = err;
4850 	return ret;
4851 }
4852 
4853 static int ext4_release_dquot(struct dquot *dquot)
4854 {
4855 	int ret, err;
4856 	handle_t *handle;
4857 
4858 	handle = ext4_journal_start(dquot_to_inode(dquot),
4859 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4860 	if (IS_ERR(handle)) {
4861 		/* Release dquot anyway to avoid endless cycle in dqput() */
4862 		dquot_release(dquot);
4863 		return PTR_ERR(handle);
4864 	}
4865 	ret = dquot_release(dquot);
4866 	err = ext4_journal_stop(handle);
4867 	if (!ret)
4868 		ret = err;
4869 	return ret;
4870 }
4871 
4872 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4873 {
4874 	/* Are we journaling quotas? */
4875 	if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4876 	    EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4877 		dquot_mark_dquot_dirty(dquot);
4878 		return ext4_write_dquot(dquot);
4879 	} else {
4880 		return dquot_mark_dquot_dirty(dquot);
4881 	}
4882 }
4883 
4884 static int ext4_write_info(struct super_block *sb, int type)
4885 {
4886 	int ret, err;
4887 	handle_t *handle;
4888 
4889 	/* Data block + inode block */
4890 	handle = ext4_journal_start(sb->s_root->d_inode, 2);
4891 	if (IS_ERR(handle))
4892 		return PTR_ERR(handle);
4893 	ret = dquot_commit_info(sb, type);
4894 	err = ext4_journal_stop(handle);
4895 	if (!ret)
4896 		ret = err;
4897 	return ret;
4898 }
4899 
4900 /*
4901  * Turn on quotas during mount time - we need to find
4902  * the quota file and such...
4903  */
4904 static int ext4_quota_on_mount(struct super_block *sb, int type)
4905 {
4906 	return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4907 					EXT4_SB(sb)->s_jquota_fmt, type);
4908 }
4909 
4910 /*
4911  * Standard function to be called on quota_on
4912  */
4913 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4914 			 struct path *path)
4915 {
4916 	int err;
4917 
4918 	if (!test_opt(sb, QUOTA))
4919 		return -EINVAL;
4920 
4921 	/* Quotafile not on the same filesystem? */
4922 	if (path->dentry->d_sb != sb)
4923 		return -EXDEV;
4924 	/* Journaling quota? */
4925 	if (EXT4_SB(sb)->s_qf_names[type]) {
4926 		/* Quotafile not in fs root? */
4927 		if (path->dentry->d_parent != sb->s_root)
4928 			ext4_msg(sb, KERN_WARNING,
4929 				"Quota file not on filesystem root. "
4930 				"Journaled quota will not work");
4931 	}
4932 
4933 	/*
4934 	 * When we journal data on quota file, we have to flush journal to see
4935 	 * all updates to the file when we bypass pagecache...
4936 	 */
4937 	if (EXT4_SB(sb)->s_journal &&
4938 	    ext4_should_journal_data(path->dentry->d_inode)) {
4939 		/*
4940 		 * We don't need to lock updates but journal_flush() could
4941 		 * otherwise be livelocked...
4942 		 */
4943 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4944 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4945 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4946 		if (err)
4947 			return err;
4948 	}
4949 
4950 	return dquot_quota_on(sb, type, format_id, path);
4951 }
4952 
4953 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4954 			     unsigned int flags)
4955 {
4956 	int err;
4957 	struct inode *qf_inode;
4958 	unsigned long qf_inums[MAXQUOTAS] = {
4959 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4960 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4961 	};
4962 
4963 	BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4964 
4965 	if (!qf_inums[type])
4966 		return -EPERM;
4967 
4968 	qf_inode = ext4_iget(sb, qf_inums[type]);
4969 	if (IS_ERR(qf_inode)) {
4970 		ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4971 		return PTR_ERR(qf_inode);
4972 	}
4973 
4974 	err = dquot_enable(qf_inode, type, format_id, flags);
4975 	iput(qf_inode);
4976 
4977 	return err;
4978 }
4979 
4980 /* Enable usage tracking for all quota types. */
4981 static int ext4_enable_quotas(struct super_block *sb)
4982 {
4983 	int type, err = 0;
4984 	unsigned long qf_inums[MAXQUOTAS] = {
4985 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4986 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4987 	};
4988 
4989 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4990 	for (type = 0; type < MAXQUOTAS; type++) {
4991 		if (qf_inums[type]) {
4992 			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4993 						DQUOT_USAGE_ENABLED);
4994 			if (err) {
4995 				ext4_warning(sb,
4996 					"Failed to enable quota (type=%d) "
4997 					"tracking. Please run e2fsck to fix.",
4998 					type);
4999 				return err;
5000 			}
5001 		}
5002 	}
5003 	return 0;
5004 }
5005 
5006 /*
5007  * quota_on function that is used when QUOTA feature is set.
5008  */
5009 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5010 				 int format_id)
5011 {
5012 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5013 		return -EINVAL;
5014 
5015 	/*
5016 	 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5017 	 */
5018 	return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5019 }
5020 
5021 static int ext4_quota_off(struct super_block *sb, int type)
5022 {
5023 	struct inode *inode = sb_dqopt(sb)->files[type];
5024 	handle_t *handle;
5025 
5026 	/* Force all delayed allocation blocks to be allocated.
5027 	 * Caller already holds s_umount sem */
5028 	if (test_opt(sb, DELALLOC))
5029 		sync_filesystem(sb);
5030 
5031 	if (!inode)
5032 		goto out;
5033 
5034 	/* Update modification times of quota files when userspace can
5035 	 * start looking at them */
5036 	handle = ext4_journal_start(inode, 1);
5037 	if (IS_ERR(handle))
5038 		goto out;
5039 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5040 	ext4_mark_inode_dirty(handle, inode);
5041 	ext4_journal_stop(handle);
5042 
5043 out:
5044 	return dquot_quota_off(sb, type);
5045 }
5046 
5047 /*
5048  * quota_off function that is used when QUOTA feature is set.
5049  */
5050 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5051 {
5052 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5053 		return -EINVAL;
5054 
5055 	/* Disable only the limits. */
5056 	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5057 }
5058 
5059 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5060  * acquiring the locks... As quota files are never truncated and quota code
5061  * itself serializes the operations (and no one else should touch the files)
5062  * we don't have to be afraid of races */
5063 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5064 			       size_t len, loff_t off)
5065 {
5066 	struct inode *inode = sb_dqopt(sb)->files[type];
5067 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5068 	int err = 0;
5069 	int offset = off & (sb->s_blocksize - 1);
5070 	int tocopy;
5071 	size_t toread;
5072 	struct buffer_head *bh;
5073 	loff_t i_size = i_size_read(inode);
5074 
5075 	if (off > i_size)
5076 		return 0;
5077 	if (off+len > i_size)
5078 		len = i_size-off;
5079 	toread = len;
5080 	while (toread > 0) {
5081 		tocopy = sb->s_blocksize - offset < toread ?
5082 				sb->s_blocksize - offset : toread;
5083 		bh = ext4_bread(NULL, inode, blk, 0, &err);
5084 		if (err)
5085 			return err;
5086 		if (!bh)	/* A hole? */
5087 			memset(data, 0, tocopy);
5088 		else
5089 			memcpy(data, bh->b_data+offset, tocopy);
5090 		brelse(bh);
5091 		offset = 0;
5092 		toread -= tocopy;
5093 		data += tocopy;
5094 		blk++;
5095 	}
5096 	return len;
5097 }
5098 
5099 /* Write to quotafile (we know the transaction is already started and has
5100  * enough credits) */
5101 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5102 				const char *data, size_t len, loff_t off)
5103 {
5104 	struct inode *inode = sb_dqopt(sb)->files[type];
5105 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5106 	int err = 0;
5107 	int offset = off & (sb->s_blocksize - 1);
5108 	struct buffer_head *bh;
5109 	handle_t *handle = journal_current_handle();
5110 
5111 	if (EXT4_SB(sb)->s_journal && !handle) {
5112 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5113 			" cancelled because transaction is not started",
5114 			(unsigned long long)off, (unsigned long long)len);
5115 		return -EIO;
5116 	}
5117 	/*
5118 	 * Since we account only one data block in transaction credits,
5119 	 * then it is impossible to cross a block boundary.
5120 	 */
5121 	if (sb->s_blocksize - offset < len) {
5122 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5123 			" cancelled because not block aligned",
5124 			(unsigned long long)off, (unsigned long long)len);
5125 		return -EIO;
5126 	}
5127 
5128 	bh = ext4_bread(handle, inode, blk, 1, &err);
5129 	if (!bh)
5130 		goto out;
5131 	err = ext4_journal_get_write_access(handle, bh);
5132 	if (err) {
5133 		brelse(bh);
5134 		goto out;
5135 	}
5136 	lock_buffer(bh);
5137 	memcpy(bh->b_data+offset, data, len);
5138 	flush_dcache_page(bh->b_page);
5139 	unlock_buffer(bh);
5140 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
5141 	brelse(bh);
5142 out:
5143 	if (err)
5144 		return err;
5145 	if (inode->i_size < off + len) {
5146 		i_size_write(inode, off + len);
5147 		EXT4_I(inode)->i_disksize = inode->i_size;
5148 		ext4_mark_inode_dirty(handle, inode);
5149 	}
5150 	return len;
5151 }
5152 
5153 #endif
5154 
5155 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5156 		       const char *dev_name, void *data)
5157 {
5158 	return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5159 }
5160 
5161 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5162 static inline void register_as_ext2(void)
5163 {
5164 	int err = register_filesystem(&ext2_fs_type);
5165 	if (err)
5166 		printk(KERN_WARNING
5167 		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5168 }
5169 
5170 static inline void unregister_as_ext2(void)
5171 {
5172 	unregister_filesystem(&ext2_fs_type);
5173 }
5174 
5175 static inline int ext2_feature_set_ok(struct super_block *sb)
5176 {
5177 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5178 		return 0;
5179 	if (sb->s_flags & MS_RDONLY)
5180 		return 1;
5181 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5182 		return 0;
5183 	return 1;
5184 }
5185 MODULE_ALIAS("ext2");
5186 #else
5187 static inline void register_as_ext2(void) { }
5188 static inline void unregister_as_ext2(void) { }
5189 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5190 #endif
5191 
5192 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5193 static inline void register_as_ext3(void)
5194 {
5195 	int err = register_filesystem(&ext3_fs_type);
5196 	if (err)
5197 		printk(KERN_WARNING
5198 		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5199 }
5200 
5201 static inline void unregister_as_ext3(void)
5202 {
5203 	unregister_filesystem(&ext3_fs_type);
5204 }
5205 
5206 static inline int ext3_feature_set_ok(struct super_block *sb)
5207 {
5208 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5209 		return 0;
5210 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5211 		return 0;
5212 	if (sb->s_flags & MS_RDONLY)
5213 		return 1;
5214 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5215 		return 0;
5216 	return 1;
5217 }
5218 MODULE_ALIAS("ext3");
5219 #else
5220 static inline void register_as_ext3(void) { }
5221 static inline void unregister_as_ext3(void) { }
5222 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5223 #endif
5224 
5225 static struct file_system_type ext4_fs_type = {
5226 	.owner		= THIS_MODULE,
5227 	.name		= "ext4",
5228 	.mount		= ext4_mount,
5229 	.kill_sb	= kill_block_super,
5230 	.fs_flags	= FS_REQUIRES_DEV,
5231 };
5232 
5233 static int __init ext4_init_feat_adverts(void)
5234 {
5235 	struct ext4_features *ef;
5236 	int ret = -ENOMEM;
5237 
5238 	ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5239 	if (!ef)
5240 		goto out;
5241 
5242 	ef->f_kobj.kset = ext4_kset;
5243 	init_completion(&ef->f_kobj_unregister);
5244 	ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5245 				   "features");
5246 	if (ret) {
5247 		kfree(ef);
5248 		goto out;
5249 	}
5250 
5251 	ext4_feat = ef;
5252 	ret = 0;
5253 out:
5254 	return ret;
5255 }
5256 
5257 static void ext4_exit_feat_adverts(void)
5258 {
5259 	kobject_put(&ext4_feat->f_kobj);
5260 	wait_for_completion(&ext4_feat->f_kobj_unregister);
5261 	kfree(ext4_feat);
5262 }
5263 
5264 /* Shared across all ext4 file systems */
5265 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5266 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5267 
5268 static int __init ext4_init_fs(void)
5269 {
5270 	int i, err;
5271 
5272 	ext4_li_info = NULL;
5273 	mutex_init(&ext4_li_mtx);
5274 
5275 	ext4_check_flag_values();
5276 
5277 	for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5278 		mutex_init(&ext4__aio_mutex[i]);
5279 		init_waitqueue_head(&ext4__ioend_wq[i]);
5280 	}
5281 
5282 	err = ext4_init_pageio();
5283 	if (err)
5284 		return err;
5285 	err = ext4_init_system_zone();
5286 	if (err)
5287 		goto out6;
5288 	ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5289 	if (!ext4_kset)
5290 		goto out5;
5291 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5292 
5293 	err = ext4_init_feat_adverts();
5294 	if (err)
5295 		goto out4;
5296 
5297 	err = ext4_init_mballoc();
5298 	if (err)
5299 		goto out3;
5300 
5301 	err = ext4_init_xattr();
5302 	if (err)
5303 		goto out2;
5304 	err = init_inodecache();
5305 	if (err)
5306 		goto out1;
5307 	register_as_ext3();
5308 	register_as_ext2();
5309 	err = register_filesystem(&ext4_fs_type);
5310 	if (err)
5311 		goto out;
5312 
5313 	return 0;
5314 out:
5315 	unregister_as_ext2();
5316 	unregister_as_ext3();
5317 	destroy_inodecache();
5318 out1:
5319 	ext4_exit_xattr();
5320 out2:
5321 	ext4_exit_mballoc();
5322 out3:
5323 	ext4_exit_feat_adverts();
5324 out4:
5325 	if (ext4_proc_root)
5326 		remove_proc_entry("fs/ext4", NULL);
5327 	kset_unregister(ext4_kset);
5328 out5:
5329 	ext4_exit_system_zone();
5330 out6:
5331 	ext4_exit_pageio();
5332 	return err;
5333 }
5334 
5335 static void __exit ext4_exit_fs(void)
5336 {
5337 	ext4_destroy_lazyinit_thread();
5338 	unregister_as_ext2();
5339 	unregister_as_ext3();
5340 	unregister_filesystem(&ext4_fs_type);
5341 	destroy_inodecache();
5342 	ext4_exit_xattr();
5343 	ext4_exit_mballoc();
5344 	ext4_exit_feat_adverts();
5345 	remove_proc_entry("fs/ext4", NULL);
5346 	kset_unregister(ext4_kset);
5347 	ext4_exit_system_zone();
5348 	ext4_exit_pageio();
5349 }
5350 
5351 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5352 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5353 MODULE_LICENSE("GPL");
5354 module_init(ext4_init_fs)
5355 module_exit(ext4_exit_fs)
5356