xref: /linux/fs/ext4/super.c (revision 4949009eb8d40a441dcddcd96e101e77d31cf1b2)
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"	/* Needed for trace points definition */
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 static int ext4_mballoc_ready;
63 
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 			     unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69 					struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71 				   struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 		       const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
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 MODULE_ALIAS_FS("ext2");
96 MODULE_ALIAS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #else
99 #define IS_EXT2_SB(sb) (0)
100 #endif
101 
102 
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type = {
105 	.owner		= THIS_MODULE,
106 	.name		= "ext3",
107 	.mount		= ext4_mount,
108 	.kill_sb	= kill_block_super,
109 	.fs_flags	= FS_REQUIRES_DEV,
110 };
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114 #else
115 #define IS_EXT3_SB(sb) (0)
116 #endif
117 
118 static int ext4_verify_csum_type(struct super_block *sb,
119 				 struct ext4_super_block *es)
120 {
121 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 					EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
123 		return 1;
124 
125 	return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
126 }
127 
128 static __le32 ext4_superblock_csum(struct super_block *sb,
129 				   struct ext4_super_block *es)
130 {
131 	struct ext4_sb_info *sbi = EXT4_SB(sb);
132 	int offset = offsetof(struct ext4_super_block, s_checksum);
133 	__u32 csum;
134 
135 	csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136 
137 	return cpu_to_le32(csum);
138 }
139 
140 static int ext4_superblock_csum_verify(struct super_block *sb,
141 				       struct ext4_super_block *es)
142 {
143 	if (!ext4_has_metadata_csum(sb))
144 		return 1;
145 
146 	return es->s_checksum == ext4_superblock_csum(sb, es);
147 }
148 
149 void ext4_superblock_csum_set(struct super_block *sb)
150 {
151 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152 
153 	if (!ext4_has_metadata_csum(sb))
154 		return;
155 
156 	es->s_checksum = ext4_superblock_csum(sb, es);
157 }
158 
159 void *ext4_kvmalloc(size_t size, gfp_t flags)
160 {
161 	void *ret;
162 
163 	ret = kmalloc(size, flags | __GFP_NOWARN);
164 	if (!ret)
165 		ret = __vmalloc(size, flags, PAGE_KERNEL);
166 	return ret;
167 }
168 
169 void *ext4_kvzalloc(size_t size, gfp_t flags)
170 {
171 	void *ret;
172 
173 	ret = kzalloc(size, flags | __GFP_NOWARN);
174 	if (!ret)
175 		ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
176 	return ret;
177 }
178 
179 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
180 			       struct ext4_group_desc *bg)
181 {
182 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
183 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
184 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
185 }
186 
187 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
188 			       struct ext4_group_desc *bg)
189 {
190 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
191 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
192 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
193 }
194 
195 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
196 			      struct ext4_group_desc *bg)
197 {
198 	return le32_to_cpu(bg->bg_inode_table_lo) |
199 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
200 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
201 }
202 
203 __u32 ext4_free_group_clusters(struct super_block *sb,
204 			       struct ext4_group_desc *bg)
205 {
206 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
207 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
209 }
210 
211 __u32 ext4_free_inodes_count(struct super_block *sb,
212 			      struct ext4_group_desc *bg)
213 {
214 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
215 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
217 }
218 
219 __u32 ext4_used_dirs_count(struct super_block *sb,
220 			      struct ext4_group_desc *bg)
221 {
222 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
223 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
225 }
226 
227 __u32 ext4_itable_unused_count(struct super_block *sb,
228 			      struct ext4_group_desc *bg)
229 {
230 	return le16_to_cpu(bg->bg_itable_unused_lo) |
231 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
233 }
234 
235 void ext4_block_bitmap_set(struct super_block *sb,
236 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
237 {
238 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
239 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
240 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
241 }
242 
243 void ext4_inode_bitmap_set(struct super_block *sb,
244 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
245 {
246 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
247 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
248 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
249 }
250 
251 void ext4_inode_table_set(struct super_block *sb,
252 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
253 {
254 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
255 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
256 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
257 }
258 
259 void ext4_free_group_clusters_set(struct super_block *sb,
260 				  struct ext4_group_desc *bg, __u32 count)
261 {
262 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
263 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
265 }
266 
267 void ext4_free_inodes_set(struct super_block *sb,
268 			  struct ext4_group_desc *bg, __u32 count)
269 {
270 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
271 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
273 }
274 
275 void ext4_used_dirs_set(struct super_block *sb,
276 			  struct ext4_group_desc *bg, __u32 count)
277 {
278 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
279 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
281 }
282 
283 void ext4_itable_unused_set(struct super_block *sb,
284 			  struct ext4_group_desc *bg, __u32 count)
285 {
286 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
287 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
289 }
290 
291 
292 static void __save_error_info(struct super_block *sb, const char *func,
293 			    unsigned int line)
294 {
295 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
296 
297 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
298 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
299 	es->s_last_error_time = cpu_to_le32(get_seconds());
300 	strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
301 	es->s_last_error_line = cpu_to_le32(line);
302 	if (!es->s_first_error_time) {
303 		es->s_first_error_time = es->s_last_error_time;
304 		strncpy(es->s_first_error_func, func,
305 			sizeof(es->s_first_error_func));
306 		es->s_first_error_line = cpu_to_le32(line);
307 		es->s_first_error_ino = es->s_last_error_ino;
308 		es->s_first_error_block = es->s_last_error_block;
309 	}
310 	/*
311 	 * Start the daily error reporting function if it hasn't been
312 	 * started already
313 	 */
314 	if (!es->s_error_count)
315 		mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
316 	le32_add_cpu(&es->s_error_count, 1);
317 }
318 
319 static void save_error_info(struct super_block *sb, const char *func,
320 			    unsigned int line)
321 {
322 	__save_error_info(sb, func, line);
323 	ext4_commit_super(sb, 1);
324 }
325 
326 /*
327  * The del_gendisk() function uninitializes the disk-specific data
328  * structures, including the bdi structure, without telling anyone
329  * else.  Once this happens, any attempt to call mark_buffer_dirty()
330  * (for example, by ext4_commit_super), will cause a kernel OOPS.
331  * This is a kludge to prevent these oops until we can put in a proper
332  * hook in del_gendisk() to inform the VFS and file system layers.
333  */
334 static int block_device_ejected(struct super_block *sb)
335 {
336 	struct inode *bd_inode = sb->s_bdev->bd_inode;
337 	struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
338 
339 	return bdi->dev == NULL;
340 }
341 
342 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
343 {
344 	struct super_block		*sb = journal->j_private;
345 	struct ext4_sb_info		*sbi = EXT4_SB(sb);
346 	int				error = is_journal_aborted(journal);
347 	struct ext4_journal_cb_entry	*jce;
348 
349 	BUG_ON(txn->t_state == T_FINISHED);
350 	spin_lock(&sbi->s_md_lock);
351 	while (!list_empty(&txn->t_private_list)) {
352 		jce = list_entry(txn->t_private_list.next,
353 				 struct ext4_journal_cb_entry, jce_list);
354 		list_del_init(&jce->jce_list);
355 		spin_unlock(&sbi->s_md_lock);
356 		jce->jce_func(sb, jce, error);
357 		spin_lock(&sbi->s_md_lock);
358 	}
359 	spin_unlock(&sbi->s_md_lock);
360 }
361 
362 /* Deal with the reporting of failure conditions on a filesystem such as
363  * inconsistencies detected or read IO failures.
364  *
365  * On ext2, we can store the error state of the filesystem in the
366  * superblock.  That is not possible on ext4, because we may have other
367  * write ordering constraints on the superblock which prevent us from
368  * writing it out straight away; and given that the journal is about to
369  * be aborted, we can't rely on the current, or future, transactions to
370  * write out the superblock safely.
371  *
372  * We'll just use the jbd2_journal_abort() error code to record an error in
373  * the journal instead.  On recovery, the journal will complain about
374  * that error until we've noted it down and cleared it.
375  */
376 
377 static void ext4_handle_error(struct super_block *sb)
378 {
379 	if (sb->s_flags & MS_RDONLY)
380 		return;
381 
382 	if (!test_opt(sb, ERRORS_CONT)) {
383 		journal_t *journal = EXT4_SB(sb)->s_journal;
384 
385 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
386 		if (journal)
387 			jbd2_journal_abort(journal, -EIO);
388 	}
389 	if (test_opt(sb, ERRORS_RO)) {
390 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
391 		/*
392 		 * Make sure updated value of ->s_mount_flags will be visible
393 		 * before ->s_flags update
394 		 */
395 		smp_wmb();
396 		sb->s_flags |= MS_RDONLY;
397 	}
398 	if (test_opt(sb, ERRORS_PANIC))
399 		panic("EXT4-fs (device %s): panic forced after error\n",
400 			sb->s_id);
401 }
402 
403 #define ext4_error_ratelimit(sb)					\
404 		___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),	\
405 			     "EXT4-fs error")
406 
407 void __ext4_error(struct super_block *sb, const char *function,
408 		  unsigned int line, const char *fmt, ...)
409 {
410 	struct va_format vaf;
411 	va_list args;
412 
413 	if (ext4_error_ratelimit(sb)) {
414 		va_start(args, fmt);
415 		vaf.fmt = fmt;
416 		vaf.va = &args;
417 		printk(KERN_CRIT
418 		       "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
419 		       sb->s_id, function, line, current->comm, &vaf);
420 		va_end(args);
421 	}
422 	save_error_info(sb, function, line);
423 	ext4_handle_error(sb);
424 }
425 
426 void __ext4_error_inode(struct inode *inode, const char *function,
427 			unsigned int line, ext4_fsblk_t block,
428 			const char *fmt, ...)
429 {
430 	va_list args;
431 	struct va_format vaf;
432 	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
433 
434 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
435 	es->s_last_error_block = cpu_to_le64(block);
436 	if (ext4_error_ratelimit(inode->i_sb)) {
437 		va_start(args, fmt);
438 		vaf.fmt = fmt;
439 		vaf.va = &args;
440 		if (block)
441 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
442 			       "inode #%lu: block %llu: comm %s: %pV\n",
443 			       inode->i_sb->s_id, function, line, inode->i_ino,
444 			       block, current->comm, &vaf);
445 		else
446 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
447 			       "inode #%lu: comm %s: %pV\n",
448 			       inode->i_sb->s_id, function, line, inode->i_ino,
449 			       current->comm, &vaf);
450 		va_end(args);
451 	}
452 	save_error_info(inode->i_sb, function, line);
453 	ext4_handle_error(inode->i_sb);
454 }
455 
456 void __ext4_error_file(struct file *file, const char *function,
457 		       unsigned int line, ext4_fsblk_t block,
458 		       const char *fmt, ...)
459 {
460 	va_list args;
461 	struct va_format vaf;
462 	struct ext4_super_block *es;
463 	struct inode *inode = file_inode(file);
464 	char pathname[80], *path;
465 
466 	es = EXT4_SB(inode->i_sb)->s_es;
467 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
468 	if (ext4_error_ratelimit(inode->i_sb)) {
469 		path = d_path(&(file->f_path), pathname, sizeof(pathname));
470 		if (IS_ERR(path))
471 			path = "(unknown)";
472 		va_start(args, fmt);
473 		vaf.fmt = fmt;
474 		vaf.va = &args;
475 		if (block)
476 			printk(KERN_CRIT
477 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
478 			       "block %llu: comm %s: path %s: %pV\n",
479 			       inode->i_sb->s_id, function, line, inode->i_ino,
480 			       block, current->comm, path, &vaf);
481 		else
482 			printk(KERN_CRIT
483 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
484 			       "comm %s: path %s: %pV\n",
485 			       inode->i_sb->s_id, function, line, inode->i_ino,
486 			       current->comm, path, &vaf);
487 		va_end(args);
488 	}
489 	save_error_info(inode->i_sb, function, line);
490 	ext4_handle_error(inode->i_sb);
491 }
492 
493 const char *ext4_decode_error(struct super_block *sb, int errno,
494 			      char nbuf[16])
495 {
496 	char *errstr = NULL;
497 
498 	switch (errno) {
499 	case -EIO:
500 		errstr = "IO failure";
501 		break;
502 	case -ENOMEM:
503 		errstr = "Out of memory";
504 		break;
505 	case -EROFS:
506 		if (!sb || (EXT4_SB(sb)->s_journal &&
507 			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
508 			errstr = "Journal has aborted";
509 		else
510 			errstr = "Readonly filesystem";
511 		break;
512 	default:
513 		/* If the caller passed in an extra buffer for unknown
514 		 * errors, textualise them now.  Else we just return
515 		 * NULL. */
516 		if (nbuf) {
517 			/* Check for truncated error codes... */
518 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
519 				errstr = nbuf;
520 		}
521 		break;
522 	}
523 
524 	return errstr;
525 }
526 
527 /* __ext4_std_error decodes expected errors from journaling functions
528  * automatically and invokes the appropriate error response.  */
529 
530 void __ext4_std_error(struct super_block *sb, const char *function,
531 		      unsigned int line, int errno)
532 {
533 	char nbuf[16];
534 	const char *errstr;
535 
536 	/* Special case: if the error is EROFS, and we're not already
537 	 * inside a transaction, then there's really no point in logging
538 	 * an error. */
539 	if (errno == -EROFS && journal_current_handle() == NULL &&
540 	    (sb->s_flags & MS_RDONLY))
541 		return;
542 
543 	if (ext4_error_ratelimit(sb)) {
544 		errstr = ext4_decode_error(sb, errno, nbuf);
545 		printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
546 		       sb->s_id, function, line, errstr);
547 	}
548 
549 	save_error_info(sb, function, line);
550 	ext4_handle_error(sb);
551 }
552 
553 /*
554  * ext4_abort is a much stronger failure handler than ext4_error.  The
555  * abort function may be used to deal with unrecoverable failures such
556  * as journal IO errors or ENOMEM at a critical moment in log management.
557  *
558  * We unconditionally force the filesystem into an ABORT|READONLY state,
559  * unless the error response on the fs has been set to panic in which
560  * case we take the easy way out and panic immediately.
561  */
562 
563 void __ext4_abort(struct super_block *sb, const char *function,
564 		unsigned int line, const char *fmt, ...)
565 {
566 	va_list args;
567 
568 	save_error_info(sb, function, line);
569 	va_start(args, fmt);
570 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
571 	       function, line);
572 	vprintk(fmt, args);
573 	printk("\n");
574 	va_end(args);
575 
576 	if ((sb->s_flags & MS_RDONLY) == 0) {
577 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
578 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
579 		/*
580 		 * Make sure updated value of ->s_mount_flags will be visible
581 		 * before ->s_flags update
582 		 */
583 		smp_wmb();
584 		sb->s_flags |= MS_RDONLY;
585 		if (EXT4_SB(sb)->s_journal)
586 			jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
587 		save_error_info(sb, function, line);
588 	}
589 	if (test_opt(sb, ERRORS_PANIC))
590 		panic("EXT4-fs panic from previous error\n");
591 }
592 
593 void __ext4_msg(struct super_block *sb,
594 		const char *prefix, const char *fmt, ...)
595 {
596 	struct va_format vaf;
597 	va_list args;
598 
599 	if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
600 		return;
601 
602 	va_start(args, fmt);
603 	vaf.fmt = fmt;
604 	vaf.va = &args;
605 	printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
606 	va_end(args);
607 }
608 
609 void __ext4_warning(struct super_block *sb, const char *function,
610 		    unsigned int line, const char *fmt, ...)
611 {
612 	struct va_format vaf;
613 	va_list args;
614 
615 	if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
616 			  "EXT4-fs warning"))
617 		return;
618 
619 	va_start(args, fmt);
620 	vaf.fmt = fmt;
621 	vaf.va = &args;
622 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
623 	       sb->s_id, function, line, &vaf);
624 	va_end(args);
625 }
626 
627 void __ext4_grp_locked_error(const char *function, unsigned int line,
628 			     struct super_block *sb, ext4_group_t grp,
629 			     unsigned long ino, ext4_fsblk_t block,
630 			     const char *fmt, ...)
631 __releases(bitlock)
632 __acquires(bitlock)
633 {
634 	struct va_format vaf;
635 	va_list args;
636 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
637 
638 	es->s_last_error_ino = cpu_to_le32(ino);
639 	es->s_last_error_block = cpu_to_le64(block);
640 	__save_error_info(sb, function, line);
641 
642 	if (ext4_error_ratelimit(sb)) {
643 		va_start(args, fmt);
644 		vaf.fmt = fmt;
645 		vaf.va = &args;
646 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
647 		       sb->s_id, function, line, grp);
648 		if (ino)
649 			printk(KERN_CONT "inode %lu: ", ino);
650 		if (block)
651 			printk(KERN_CONT "block %llu:",
652 			       (unsigned long long) block);
653 		printk(KERN_CONT "%pV\n", &vaf);
654 		va_end(args);
655 	}
656 
657 	if (test_opt(sb, ERRORS_CONT)) {
658 		ext4_commit_super(sb, 0);
659 		return;
660 	}
661 
662 	ext4_unlock_group(sb, grp);
663 	ext4_handle_error(sb);
664 	/*
665 	 * We only get here in the ERRORS_RO case; relocking the group
666 	 * may be dangerous, but nothing bad will happen since the
667 	 * filesystem will have already been marked read/only and the
668 	 * journal has been aborted.  We return 1 as a hint to callers
669 	 * who might what to use the return value from
670 	 * ext4_grp_locked_error() to distinguish between the
671 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
672 	 * aggressively from the ext4 function in question, with a
673 	 * more appropriate error code.
674 	 */
675 	ext4_lock_group(sb, grp);
676 	return;
677 }
678 
679 void ext4_update_dynamic_rev(struct super_block *sb)
680 {
681 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
682 
683 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
684 		return;
685 
686 	ext4_warning(sb,
687 		     "updating to rev %d because of new feature flag, "
688 		     "running e2fsck is recommended",
689 		     EXT4_DYNAMIC_REV);
690 
691 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
692 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
693 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
694 	/* leave es->s_feature_*compat flags alone */
695 	/* es->s_uuid will be set by e2fsck if empty */
696 
697 	/*
698 	 * The rest of the superblock fields should be zero, and if not it
699 	 * means they are likely already in use, so leave them alone.  We
700 	 * can leave it up to e2fsck to clean up any inconsistencies there.
701 	 */
702 }
703 
704 /*
705  * Open the external journal device
706  */
707 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
708 {
709 	struct block_device *bdev;
710 	char b[BDEVNAME_SIZE];
711 
712 	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
713 	if (IS_ERR(bdev))
714 		goto fail;
715 	return bdev;
716 
717 fail:
718 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
719 			__bdevname(dev, b), PTR_ERR(bdev));
720 	return NULL;
721 }
722 
723 /*
724  * Release the journal device
725  */
726 static void ext4_blkdev_put(struct block_device *bdev)
727 {
728 	blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
729 }
730 
731 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
732 {
733 	struct block_device *bdev;
734 	bdev = sbi->journal_bdev;
735 	if (bdev) {
736 		ext4_blkdev_put(bdev);
737 		sbi->journal_bdev = NULL;
738 	}
739 }
740 
741 static inline struct inode *orphan_list_entry(struct list_head *l)
742 {
743 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
744 }
745 
746 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
747 {
748 	struct list_head *l;
749 
750 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
751 		 le32_to_cpu(sbi->s_es->s_last_orphan));
752 
753 	printk(KERN_ERR "sb_info orphan list:\n");
754 	list_for_each(l, &sbi->s_orphan) {
755 		struct inode *inode = orphan_list_entry(l);
756 		printk(KERN_ERR "  "
757 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
758 		       inode->i_sb->s_id, inode->i_ino, inode,
759 		       inode->i_mode, inode->i_nlink,
760 		       NEXT_ORPHAN(inode));
761 	}
762 }
763 
764 static void ext4_put_super(struct super_block *sb)
765 {
766 	struct ext4_sb_info *sbi = EXT4_SB(sb);
767 	struct ext4_super_block *es = sbi->s_es;
768 	int i, err;
769 
770 	ext4_unregister_li_request(sb);
771 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
772 
773 	flush_workqueue(sbi->rsv_conversion_wq);
774 	destroy_workqueue(sbi->rsv_conversion_wq);
775 
776 	if (sbi->s_journal) {
777 		err = jbd2_journal_destroy(sbi->s_journal);
778 		sbi->s_journal = NULL;
779 		if (err < 0)
780 			ext4_abort(sb, "Couldn't clean up the journal");
781 	}
782 
783 	ext4_es_unregister_shrinker(sbi);
784 	del_timer_sync(&sbi->s_err_report);
785 	ext4_release_system_zone(sb);
786 	ext4_mb_release(sb);
787 	ext4_ext_release(sb);
788 	ext4_xattr_put_super(sb);
789 
790 	if (!(sb->s_flags & MS_RDONLY)) {
791 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
792 		es->s_state = cpu_to_le16(sbi->s_mount_state);
793 	}
794 	if (!(sb->s_flags & MS_RDONLY))
795 		ext4_commit_super(sb, 1);
796 
797 	if (sbi->s_proc) {
798 		remove_proc_entry("options", sbi->s_proc);
799 		remove_proc_entry(sb->s_id, ext4_proc_root);
800 	}
801 	kobject_del(&sbi->s_kobj);
802 
803 	for (i = 0; i < sbi->s_gdb_count; i++)
804 		brelse(sbi->s_group_desc[i]);
805 	kvfree(sbi->s_group_desc);
806 	kvfree(sbi->s_flex_groups);
807 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
808 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
809 	percpu_counter_destroy(&sbi->s_dirs_counter);
810 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
811 	brelse(sbi->s_sbh);
812 #ifdef CONFIG_QUOTA
813 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
814 		kfree(sbi->s_qf_names[i]);
815 #endif
816 
817 	/* Debugging code just in case the in-memory inode orphan list
818 	 * isn't empty.  The on-disk one can be non-empty if we've
819 	 * detected an error and taken the fs readonly, but the
820 	 * in-memory list had better be clean by this point. */
821 	if (!list_empty(&sbi->s_orphan))
822 		dump_orphan_list(sb, sbi);
823 	J_ASSERT(list_empty(&sbi->s_orphan));
824 
825 	invalidate_bdev(sb->s_bdev);
826 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
827 		/*
828 		 * Invalidate the journal device's buffers.  We don't want them
829 		 * floating about in memory - the physical journal device may
830 		 * hotswapped, and it breaks the `ro-after' testing code.
831 		 */
832 		sync_blockdev(sbi->journal_bdev);
833 		invalidate_bdev(sbi->journal_bdev);
834 		ext4_blkdev_remove(sbi);
835 	}
836 	if (sbi->s_mb_cache) {
837 		ext4_xattr_destroy_cache(sbi->s_mb_cache);
838 		sbi->s_mb_cache = NULL;
839 	}
840 	if (sbi->s_mmp_tsk)
841 		kthread_stop(sbi->s_mmp_tsk);
842 	sb->s_fs_info = NULL;
843 	/*
844 	 * Now that we are completely done shutting down the
845 	 * superblock, we need to actually destroy the kobject.
846 	 */
847 	kobject_put(&sbi->s_kobj);
848 	wait_for_completion(&sbi->s_kobj_unregister);
849 	if (sbi->s_chksum_driver)
850 		crypto_free_shash(sbi->s_chksum_driver);
851 	kfree(sbi->s_blockgroup_lock);
852 	kfree(sbi);
853 }
854 
855 static struct kmem_cache *ext4_inode_cachep;
856 
857 /*
858  * Called inside transaction, so use GFP_NOFS
859  */
860 static struct inode *ext4_alloc_inode(struct super_block *sb)
861 {
862 	struct ext4_inode_info *ei;
863 
864 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
865 	if (!ei)
866 		return NULL;
867 
868 	ei->vfs_inode.i_version = 1;
869 	spin_lock_init(&ei->i_raw_lock);
870 	INIT_LIST_HEAD(&ei->i_prealloc_list);
871 	spin_lock_init(&ei->i_prealloc_lock);
872 	ext4_es_init_tree(&ei->i_es_tree);
873 	rwlock_init(&ei->i_es_lock);
874 	INIT_LIST_HEAD(&ei->i_es_list);
875 	ei->i_es_all_nr = 0;
876 	ei->i_es_shk_nr = 0;
877 	ei->i_es_shrink_lblk = 0;
878 	ei->i_reserved_data_blocks = 0;
879 	ei->i_reserved_meta_blocks = 0;
880 	ei->i_allocated_meta_blocks = 0;
881 	ei->i_da_metadata_calc_len = 0;
882 	ei->i_da_metadata_calc_last_lblock = 0;
883 	spin_lock_init(&(ei->i_block_reservation_lock));
884 #ifdef CONFIG_QUOTA
885 	ei->i_reserved_quota = 0;
886 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
887 #endif
888 	ei->jinode = NULL;
889 	INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
890 	spin_lock_init(&ei->i_completed_io_lock);
891 	ei->i_sync_tid = 0;
892 	ei->i_datasync_tid = 0;
893 	atomic_set(&ei->i_ioend_count, 0);
894 	atomic_set(&ei->i_unwritten, 0);
895 	INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
896 
897 	return &ei->vfs_inode;
898 }
899 
900 static int ext4_drop_inode(struct inode *inode)
901 {
902 	int drop = generic_drop_inode(inode);
903 
904 	trace_ext4_drop_inode(inode, drop);
905 	return drop;
906 }
907 
908 static void ext4_i_callback(struct rcu_head *head)
909 {
910 	struct inode *inode = container_of(head, struct inode, i_rcu);
911 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
912 }
913 
914 static void ext4_destroy_inode(struct inode *inode)
915 {
916 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
917 		ext4_msg(inode->i_sb, KERN_ERR,
918 			 "Inode %lu (%p): orphan list check failed!",
919 			 inode->i_ino, EXT4_I(inode));
920 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
921 				EXT4_I(inode), sizeof(struct ext4_inode_info),
922 				true);
923 		dump_stack();
924 	}
925 	call_rcu(&inode->i_rcu, ext4_i_callback);
926 }
927 
928 static void init_once(void *foo)
929 {
930 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
931 
932 	INIT_LIST_HEAD(&ei->i_orphan);
933 	init_rwsem(&ei->xattr_sem);
934 	init_rwsem(&ei->i_data_sem);
935 	inode_init_once(&ei->vfs_inode);
936 }
937 
938 static int __init init_inodecache(void)
939 {
940 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
941 					     sizeof(struct ext4_inode_info),
942 					     0, (SLAB_RECLAIM_ACCOUNT|
943 						SLAB_MEM_SPREAD),
944 					     init_once);
945 	if (ext4_inode_cachep == NULL)
946 		return -ENOMEM;
947 	return 0;
948 }
949 
950 static void destroy_inodecache(void)
951 {
952 	/*
953 	 * Make sure all delayed rcu free inodes are flushed before we
954 	 * destroy cache.
955 	 */
956 	rcu_barrier();
957 	kmem_cache_destroy(ext4_inode_cachep);
958 }
959 
960 void ext4_clear_inode(struct inode *inode)
961 {
962 	invalidate_inode_buffers(inode);
963 	clear_inode(inode);
964 	dquot_drop(inode);
965 	ext4_discard_preallocations(inode);
966 	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
967 	if (EXT4_I(inode)->jinode) {
968 		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
969 					       EXT4_I(inode)->jinode);
970 		jbd2_free_inode(EXT4_I(inode)->jinode);
971 		EXT4_I(inode)->jinode = NULL;
972 	}
973 }
974 
975 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
976 					u64 ino, u32 generation)
977 {
978 	struct inode *inode;
979 
980 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
981 		return ERR_PTR(-ESTALE);
982 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
983 		return ERR_PTR(-ESTALE);
984 
985 	/* iget isn't really right if the inode is currently unallocated!!
986 	 *
987 	 * ext4_read_inode will return a bad_inode if the inode had been
988 	 * deleted, so we should be safe.
989 	 *
990 	 * Currently we don't know the generation for parent directory, so
991 	 * a generation of 0 means "accept any"
992 	 */
993 	inode = ext4_iget_normal(sb, ino);
994 	if (IS_ERR(inode))
995 		return ERR_CAST(inode);
996 	if (generation && inode->i_generation != generation) {
997 		iput(inode);
998 		return ERR_PTR(-ESTALE);
999 	}
1000 
1001 	return inode;
1002 }
1003 
1004 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1005 					int fh_len, int fh_type)
1006 {
1007 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1008 				    ext4_nfs_get_inode);
1009 }
1010 
1011 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1012 					int fh_len, int fh_type)
1013 {
1014 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1015 				    ext4_nfs_get_inode);
1016 }
1017 
1018 /*
1019  * Try to release metadata pages (indirect blocks, directories) which are
1020  * mapped via the block device.  Since these pages could have journal heads
1021  * which would prevent try_to_free_buffers() from freeing them, we must use
1022  * jbd2 layer's try_to_free_buffers() function to release them.
1023  */
1024 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1025 				 gfp_t wait)
1026 {
1027 	journal_t *journal = EXT4_SB(sb)->s_journal;
1028 
1029 	WARN_ON(PageChecked(page));
1030 	if (!page_has_buffers(page))
1031 		return 0;
1032 	if (journal)
1033 		return jbd2_journal_try_to_free_buffers(journal, page,
1034 							wait & ~__GFP_WAIT);
1035 	return try_to_free_buffers(page);
1036 }
1037 
1038 #ifdef CONFIG_QUOTA
1039 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1040 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1041 
1042 static int ext4_write_dquot(struct dquot *dquot);
1043 static int ext4_acquire_dquot(struct dquot *dquot);
1044 static int ext4_release_dquot(struct dquot *dquot);
1045 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1046 static int ext4_write_info(struct super_block *sb, int type);
1047 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1048 			 struct path *path);
1049 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1050 				 int format_id);
1051 static int ext4_quota_off(struct super_block *sb, int type);
1052 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1053 static int ext4_quota_on_mount(struct super_block *sb, int type);
1054 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1055 			       size_t len, loff_t off);
1056 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1057 				const char *data, size_t len, loff_t off);
1058 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1059 			     unsigned int flags);
1060 static int ext4_enable_quotas(struct super_block *sb);
1061 
1062 static struct dquot **ext4_get_dquots(struct inode *inode)
1063 {
1064 	return EXT4_I(inode)->i_dquot;
1065 }
1066 
1067 static const struct dquot_operations ext4_quota_operations = {
1068 	.get_reserved_space = ext4_get_reserved_space,
1069 	.write_dquot	= ext4_write_dquot,
1070 	.acquire_dquot	= ext4_acquire_dquot,
1071 	.release_dquot	= ext4_release_dquot,
1072 	.mark_dirty	= ext4_mark_dquot_dirty,
1073 	.write_info	= ext4_write_info,
1074 	.alloc_dquot	= dquot_alloc,
1075 	.destroy_dquot	= dquot_destroy,
1076 };
1077 
1078 static const struct quotactl_ops ext4_qctl_operations = {
1079 	.quota_on	= ext4_quota_on,
1080 	.quota_off	= ext4_quota_off,
1081 	.quota_sync	= dquot_quota_sync,
1082 	.get_info	= dquot_get_dqinfo,
1083 	.set_info	= dquot_set_dqinfo,
1084 	.get_dqblk	= dquot_get_dqblk,
1085 	.set_dqblk	= dquot_set_dqblk
1086 };
1087 
1088 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1089 	.quota_on_meta	= ext4_quota_on_sysfile,
1090 	.quota_off	= ext4_quota_off_sysfile,
1091 	.quota_sync	= dquot_quota_sync,
1092 	.get_info	= dquot_get_dqinfo,
1093 	.set_info	= dquot_set_dqinfo,
1094 	.get_dqblk	= dquot_get_dqblk,
1095 	.set_dqblk	= dquot_set_dqblk
1096 };
1097 #endif
1098 
1099 static const struct super_operations ext4_sops = {
1100 	.alloc_inode	= ext4_alloc_inode,
1101 	.destroy_inode	= ext4_destroy_inode,
1102 	.write_inode	= ext4_write_inode,
1103 	.dirty_inode	= ext4_dirty_inode,
1104 	.drop_inode	= ext4_drop_inode,
1105 	.evict_inode	= ext4_evict_inode,
1106 	.put_super	= ext4_put_super,
1107 	.sync_fs	= ext4_sync_fs,
1108 	.freeze_fs	= ext4_freeze,
1109 	.unfreeze_fs	= ext4_unfreeze,
1110 	.statfs		= ext4_statfs,
1111 	.remount_fs	= ext4_remount,
1112 	.show_options	= ext4_show_options,
1113 #ifdef CONFIG_QUOTA
1114 	.quota_read	= ext4_quota_read,
1115 	.quota_write	= ext4_quota_write,
1116 	.get_dquots	= ext4_get_dquots,
1117 #endif
1118 	.bdev_try_to_free_page = bdev_try_to_free_page,
1119 };
1120 
1121 static const struct export_operations ext4_export_ops = {
1122 	.fh_to_dentry = ext4_fh_to_dentry,
1123 	.fh_to_parent = ext4_fh_to_parent,
1124 	.get_parent = ext4_get_parent,
1125 };
1126 
1127 enum {
1128 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1129 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1130 	Opt_nouid32, Opt_debug, Opt_removed,
1131 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1132 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1133 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1134 	Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1135 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1136 	Opt_data_err_abort, Opt_data_err_ignore,
1137 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1138 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1139 	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1140 	Opt_usrquota, Opt_grpquota, Opt_i_version,
1141 	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1142 	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1143 	Opt_inode_readahead_blks, Opt_journal_ioprio,
1144 	Opt_dioread_nolock, Opt_dioread_lock,
1145 	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1146 	Opt_max_dir_size_kb, Opt_nojournal_checksum,
1147 };
1148 
1149 static const match_table_t tokens = {
1150 	{Opt_bsd_df, "bsddf"},
1151 	{Opt_minix_df, "minixdf"},
1152 	{Opt_grpid, "grpid"},
1153 	{Opt_grpid, "bsdgroups"},
1154 	{Opt_nogrpid, "nogrpid"},
1155 	{Opt_nogrpid, "sysvgroups"},
1156 	{Opt_resgid, "resgid=%u"},
1157 	{Opt_resuid, "resuid=%u"},
1158 	{Opt_sb, "sb=%u"},
1159 	{Opt_err_cont, "errors=continue"},
1160 	{Opt_err_panic, "errors=panic"},
1161 	{Opt_err_ro, "errors=remount-ro"},
1162 	{Opt_nouid32, "nouid32"},
1163 	{Opt_debug, "debug"},
1164 	{Opt_removed, "oldalloc"},
1165 	{Opt_removed, "orlov"},
1166 	{Opt_user_xattr, "user_xattr"},
1167 	{Opt_nouser_xattr, "nouser_xattr"},
1168 	{Opt_acl, "acl"},
1169 	{Opt_noacl, "noacl"},
1170 	{Opt_noload, "norecovery"},
1171 	{Opt_noload, "noload"},
1172 	{Opt_removed, "nobh"},
1173 	{Opt_removed, "bh"},
1174 	{Opt_commit, "commit=%u"},
1175 	{Opt_min_batch_time, "min_batch_time=%u"},
1176 	{Opt_max_batch_time, "max_batch_time=%u"},
1177 	{Opt_journal_dev, "journal_dev=%u"},
1178 	{Opt_journal_path, "journal_path=%s"},
1179 	{Opt_journal_checksum, "journal_checksum"},
1180 	{Opt_nojournal_checksum, "nojournal_checksum"},
1181 	{Opt_journal_async_commit, "journal_async_commit"},
1182 	{Opt_abort, "abort"},
1183 	{Opt_data_journal, "data=journal"},
1184 	{Opt_data_ordered, "data=ordered"},
1185 	{Opt_data_writeback, "data=writeback"},
1186 	{Opt_data_err_abort, "data_err=abort"},
1187 	{Opt_data_err_ignore, "data_err=ignore"},
1188 	{Opt_offusrjquota, "usrjquota="},
1189 	{Opt_usrjquota, "usrjquota=%s"},
1190 	{Opt_offgrpjquota, "grpjquota="},
1191 	{Opt_grpjquota, "grpjquota=%s"},
1192 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1193 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1194 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1195 	{Opt_grpquota, "grpquota"},
1196 	{Opt_noquota, "noquota"},
1197 	{Opt_quota, "quota"},
1198 	{Opt_usrquota, "usrquota"},
1199 	{Opt_barrier, "barrier=%u"},
1200 	{Opt_barrier, "barrier"},
1201 	{Opt_nobarrier, "nobarrier"},
1202 	{Opt_i_version, "i_version"},
1203 	{Opt_stripe, "stripe=%u"},
1204 	{Opt_delalloc, "delalloc"},
1205 	{Opt_nodelalloc, "nodelalloc"},
1206 	{Opt_removed, "mblk_io_submit"},
1207 	{Opt_removed, "nomblk_io_submit"},
1208 	{Opt_block_validity, "block_validity"},
1209 	{Opt_noblock_validity, "noblock_validity"},
1210 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1211 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1212 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1213 	{Opt_auto_da_alloc, "auto_da_alloc"},
1214 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1215 	{Opt_dioread_nolock, "dioread_nolock"},
1216 	{Opt_dioread_lock, "dioread_lock"},
1217 	{Opt_discard, "discard"},
1218 	{Opt_nodiscard, "nodiscard"},
1219 	{Opt_init_itable, "init_itable=%u"},
1220 	{Opt_init_itable, "init_itable"},
1221 	{Opt_noinit_itable, "noinit_itable"},
1222 	{Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1223 	{Opt_removed, "check=none"},	/* mount option from ext2/3 */
1224 	{Opt_removed, "nocheck"},	/* mount option from ext2/3 */
1225 	{Opt_removed, "reservation"},	/* mount option from ext2/3 */
1226 	{Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1227 	{Opt_removed, "journal=%u"},	/* mount option from ext2/3 */
1228 	{Opt_err, NULL},
1229 };
1230 
1231 static ext4_fsblk_t get_sb_block(void **data)
1232 {
1233 	ext4_fsblk_t	sb_block;
1234 	char		*options = (char *) *data;
1235 
1236 	if (!options || strncmp(options, "sb=", 3) != 0)
1237 		return 1;	/* Default location */
1238 
1239 	options += 3;
1240 	/* TODO: use simple_strtoll with >32bit ext4 */
1241 	sb_block = simple_strtoul(options, &options, 0);
1242 	if (*options && *options != ',') {
1243 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1244 		       (char *) *data);
1245 		return 1;
1246 	}
1247 	if (*options == ',')
1248 		options++;
1249 	*data = (void *) options;
1250 
1251 	return sb_block;
1252 }
1253 
1254 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1255 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1256 	"Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1257 
1258 #ifdef CONFIG_QUOTA
1259 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1260 {
1261 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1262 	char *qname;
1263 	int ret = -1;
1264 
1265 	if (sb_any_quota_loaded(sb) &&
1266 		!sbi->s_qf_names[qtype]) {
1267 		ext4_msg(sb, KERN_ERR,
1268 			"Cannot change journaled "
1269 			"quota options when quota turned on");
1270 		return -1;
1271 	}
1272 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1273 		ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1274 			 "when QUOTA feature is enabled");
1275 		return -1;
1276 	}
1277 	qname = match_strdup(args);
1278 	if (!qname) {
1279 		ext4_msg(sb, KERN_ERR,
1280 			"Not enough memory for storing quotafile name");
1281 		return -1;
1282 	}
1283 	if (sbi->s_qf_names[qtype]) {
1284 		if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1285 			ret = 1;
1286 		else
1287 			ext4_msg(sb, KERN_ERR,
1288 				 "%s quota file already specified",
1289 				 QTYPE2NAME(qtype));
1290 		goto errout;
1291 	}
1292 	if (strchr(qname, '/')) {
1293 		ext4_msg(sb, KERN_ERR,
1294 			"quotafile must be on filesystem root");
1295 		goto errout;
1296 	}
1297 	sbi->s_qf_names[qtype] = qname;
1298 	set_opt(sb, QUOTA);
1299 	return 1;
1300 errout:
1301 	kfree(qname);
1302 	return ret;
1303 }
1304 
1305 static int clear_qf_name(struct super_block *sb, int qtype)
1306 {
1307 
1308 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1309 
1310 	if (sb_any_quota_loaded(sb) &&
1311 		sbi->s_qf_names[qtype]) {
1312 		ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1313 			" when quota turned on");
1314 		return -1;
1315 	}
1316 	kfree(sbi->s_qf_names[qtype]);
1317 	sbi->s_qf_names[qtype] = NULL;
1318 	return 1;
1319 }
1320 #endif
1321 
1322 #define MOPT_SET	0x0001
1323 #define MOPT_CLEAR	0x0002
1324 #define MOPT_NOSUPPORT	0x0004
1325 #define MOPT_EXPLICIT	0x0008
1326 #define MOPT_CLEAR_ERR	0x0010
1327 #define MOPT_GTE0	0x0020
1328 #ifdef CONFIG_QUOTA
1329 #define MOPT_Q		0
1330 #define MOPT_QFMT	0x0040
1331 #else
1332 #define MOPT_Q		MOPT_NOSUPPORT
1333 #define MOPT_QFMT	MOPT_NOSUPPORT
1334 #endif
1335 #define MOPT_DATAJ	0x0080
1336 #define MOPT_NO_EXT2	0x0100
1337 #define MOPT_NO_EXT3	0x0200
1338 #define MOPT_EXT4_ONLY	(MOPT_NO_EXT2 | MOPT_NO_EXT3)
1339 #define MOPT_STRING	0x0400
1340 
1341 static const struct mount_opts {
1342 	int	token;
1343 	int	mount_opt;
1344 	int	flags;
1345 } ext4_mount_opts[] = {
1346 	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1347 	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1348 	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1349 	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1350 	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1351 	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1352 	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1353 	 MOPT_EXT4_ONLY | MOPT_SET},
1354 	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1355 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1356 	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1357 	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1358 	{Opt_delalloc, EXT4_MOUNT_DELALLOC,
1359 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1360 	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1361 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1362 	{Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1363 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1364 	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1365 	 MOPT_EXT4_ONLY | MOPT_SET},
1366 	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1367 				    EXT4_MOUNT_JOURNAL_CHECKSUM),
1368 	 MOPT_EXT4_ONLY | MOPT_SET},
1369 	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1370 	{Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1371 	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1372 	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1373 	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1374 	 MOPT_NO_EXT2 | MOPT_SET},
1375 	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1376 	 MOPT_NO_EXT2 | MOPT_CLEAR},
1377 	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1378 	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1379 	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1380 	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1381 	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1382 	{Opt_commit, 0, MOPT_GTE0},
1383 	{Opt_max_batch_time, 0, MOPT_GTE0},
1384 	{Opt_min_batch_time, 0, MOPT_GTE0},
1385 	{Opt_inode_readahead_blks, 0, MOPT_GTE0},
1386 	{Opt_init_itable, 0, MOPT_GTE0},
1387 	{Opt_stripe, 0, MOPT_GTE0},
1388 	{Opt_resuid, 0, MOPT_GTE0},
1389 	{Opt_resgid, 0, MOPT_GTE0},
1390 	{Opt_journal_dev, 0, MOPT_GTE0},
1391 	{Opt_journal_path, 0, MOPT_STRING},
1392 	{Opt_journal_ioprio, 0, MOPT_GTE0},
1393 	{Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1394 	{Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1395 	{Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1396 	 MOPT_NO_EXT2 | MOPT_DATAJ},
1397 	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1398 	{Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1399 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1400 	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1401 	{Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1402 #else
1403 	{Opt_acl, 0, MOPT_NOSUPPORT},
1404 	{Opt_noacl, 0, MOPT_NOSUPPORT},
1405 #endif
1406 	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1407 	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1408 	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1409 	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1410 							MOPT_SET | MOPT_Q},
1411 	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1412 							MOPT_SET | MOPT_Q},
1413 	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1414 		       EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1415 	{Opt_usrjquota, 0, MOPT_Q},
1416 	{Opt_grpjquota, 0, MOPT_Q},
1417 	{Opt_offusrjquota, 0, MOPT_Q},
1418 	{Opt_offgrpjquota, 0, MOPT_Q},
1419 	{Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1420 	{Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1421 	{Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1422 	{Opt_max_dir_size_kb, 0, MOPT_GTE0},
1423 	{Opt_err, 0, 0}
1424 };
1425 
1426 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1427 			    substring_t *args, unsigned long *journal_devnum,
1428 			    unsigned int *journal_ioprio, int is_remount)
1429 {
1430 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1431 	const struct mount_opts *m;
1432 	kuid_t uid;
1433 	kgid_t gid;
1434 	int arg = 0;
1435 
1436 #ifdef CONFIG_QUOTA
1437 	if (token == Opt_usrjquota)
1438 		return set_qf_name(sb, USRQUOTA, &args[0]);
1439 	else if (token == Opt_grpjquota)
1440 		return set_qf_name(sb, GRPQUOTA, &args[0]);
1441 	else if (token == Opt_offusrjquota)
1442 		return clear_qf_name(sb, USRQUOTA);
1443 	else if (token == Opt_offgrpjquota)
1444 		return clear_qf_name(sb, GRPQUOTA);
1445 #endif
1446 	switch (token) {
1447 	case Opt_noacl:
1448 	case Opt_nouser_xattr:
1449 		ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1450 		break;
1451 	case Opt_sb:
1452 		return 1;	/* handled by get_sb_block() */
1453 	case Opt_removed:
1454 		ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1455 		return 1;
1456 	case Opt_abort:
1457 		sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1458 		return 1;
1459 	case Opt_i_version:
1460 		sb->s_flags |= MS_I_VERSION;
1461 		return 1;
1462 	}
1463 
1464 	for (m = ext4_mount_opts; m->token != Opt_err; m++)
1465 		if (token == m->token)
1466 			break;
1467 
1468 	if (m->token == Opt_err) {
1469 		ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1470 			 "or missing value", opt);
1471 		return -1;
1472 	}
1473 
1474 	if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1475 		ext4_msg(sb, KERN_ERR,
1476 			 "Mount option \"%s\" incompatible with ext2", opt);
1477 		return -1;
1478 	}
1479 	if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1480 		ext4_msg(sb, KERN_ERR,
1481 			 "Mount option \"%s\" incompatible with ext3", opt);
1482 		return -1;
1483 	}
1484 
1485 	if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1486 		return -1;
1487 	if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1488 		return -1;
1489 	if (m->flags & MOPT_EXPLICIT)
1490 		set_opt2(sb, EXPLICIT_DELALLOC);
1491 	if (m->flags & MOPT_CLEAR_ERR)
1492 		clear_opt(sb, ERRORS_MASK);
1493 	if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1494 		ext4_msg(sb, KERN_ERR, "Cannot change quota "
1495 			 "options when quota turned on");
1496 		return -1;
1497 	}
1498 
1499 	if (m->flags & MOPT_NOSUPPORT) {
1500 		ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1501 	} else if (token == Opt_commit) {
1502 		if (arg == 0)
1503 			arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1504 		sbi->s_commit_interval = HZ * arg;
1505 	} else if (token == Opt_max_batch_time) {
1506 		sbi->s_max_batch_time = arg;
1507 	} else if (token == Opt_min_batch_time) {
1508 		sbi->s_min_batch_time = arg;
1509 	} else if (token == Opt_inode_readahead_blks) {
1510 		if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1511 			ext4_msg(sb, KERN_ERR,
1512 				 "EXT4-fs: inode_readahead_blks must be "
1513 				 "0 or a power of 2 smaller than 2^31");
1514 			return -1;
1515 		}
1516 		sbi->s_inode_readahead_blks = arg;
1517 	} else if (token == Opt_init_itable) {
1518 		set_opt(sb, INIT_INODE_TABLE);
1519 		if (!args->from)
1520 			arg = EXT4_DEF_LI_WAIT_MULT;
1521 		sbi->s_li_wait_mult = arg;
1522 	} else if (token == Opt_max_dir_size_kb) {
1523 		sbi->s_max_dir_size_kb = arg;
1524 	} else if (token == Opt_stripe) {
1525 		sbi->s_stripe = arg;
1526 	} else if (token == Opt_resuid) {
1527 		uid = make_kuid(current_user_ns(), arg);
1528 		if (!uid_valid(uid)) {
1529 			ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1530 			return -1;
1531 		}
1532 		sbi->s_resuid = uid;
1533 	} else if (token == Opt_resgid) {
1534 		gid = make_kgid(current_user_ns(), arg);
1535 		if (!gid_valid(gid)) {
1536 			ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1537 			return -1;
1538 		}
1539 		sbi->s_resgid = gid;
1540 	} else if (token == Opt_journal_dev) {
1541 		if (is_remount) {
1542 			ext4_msg(sb, KERN_ERR,
1543 				 "Cannot specify journal on remount");
1544 			return -1;
1545 		}
1546 		*journal_devnum = arg;
1547 	} else if (token == Opt_journal_path) {
1548 		char *journal_path;
1549 		struct inode *journal_inode;
1550 		struct path path;
1551 		int error;
1552 
1553 		if (is_remount) {
1554 			ext4_msg(sb, KERN_ERR,
1555 				 "Cannot specify journal on remount");
1556 			return -1;
1557 		}
1558 		journal_path = match_strdup(&args[0]);
1559 		if (!journal_path) {
1560 			ext4_msg(sb, KERN_ERR, "error: could not dup "
1561 				"journal device string");
1562 			return -1;
1563 		}
1564 
1565 		error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1566 		if (error) {
1567 			ext4_msg(sb, KERN_ERR, "error: could not find "
1568 				"journal device path: error %d", error);
1569 			kfree(journal_path);
1570 			return -1;
1571 		}
1572 
1573 		journal_inode = path.dentry->d_inode;
1574 		if (!S_ISBLK(journal_inode->i_mode)) {
1575 			ext4_msg(sb, KERN_ERR, "error: journal path %s "
1576 				"is not a block device", journal_path);
1577 			path_put(&path);
1578 			kfree(journal_path);
1579 			return -1;
1580 		}
1581 
1582 		*journal_devnum = new_encode_dev(journal_inode->i_rdev);
1583 		path_put(&path);
1584 		kfree(journal_path);
1585 	} else if (token == Opt_journal_ioprio) {
1586 		if (arg > 7) {
1587 			ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1588 				 " (must be 0-7)");
1589 			return -1;
1590 		}
1591 		*journal_ioprio =
1592 			IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1593 	} else if (m->flags & MOPT_DATAJ) {
1594 		if (is_remount) {
1595 			if (!sbi->s_journal)
1596 				ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1597 			else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1598 				ext4_msg(sb, KERN_ERR,
1599 					 "Cannot change data mode on remount");
1600 				return -1;
1601 			}
1602 		} else {
1603 			clear_opt(sb, DATA_FLAGS);
1604 			sbi->s_mount_opt |= m->mount_opt;
1605 		}
1606 #ifdef CONFIG_QUOTA
1607 	} else if (m->flags & MOPT_QFMT) {
1608 		if (sb_any_quota_loaded(sb) &&
1609 		    sbi->s_jquota_fmt != m->mount_opt) {
1610 			ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1611 				 "quota options when quota turned on");
1612 			return -1;
1613 		}
1614 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1615 					       EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1616 			ext4_msg(sb, KERN_ERR,
1617 				 "Cannot set journaled quota options "
1618 				 "when QUOTA feature is enabled");
1619 			return -1;
1620 		}
1621 		sbi->s_jquota_fmt = m->mount_opt;
1622 #endif
1623 	} else {
1624 		if (!args->from)
1625 			arg = 1;
1626 		if (m->flags & MOPT_CLEAR)
1627 			arg = !arg;
1628 		else if (unlikely(!(m->flags & MOPT_SET))) {
1629 			ext4_msg(sb, KERN_WARNING,
1630 				 "buggy handling of option %s", opt);
1631 			WARN_ON(1);
1632 			return -1;
1633 		}
1634 		if (arg != 0)
1635 			sbi->s_mount_opt |= m->mount_opt;
1636 		else
1637 			sbi->s_mount_opt &= ~m->mount_opt;
1638 	}
1639 	return 1;
1640 }
1641 
1642 static int parse_options(char *options, struct super_block *sb,
1643 			 unsigned long *journal_devnum,
1644 			 unsigned int *journal_ioprio,
1645 			 int is_remount)
1646 {
1647 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1648 	char *p;
1649 	substring_t args[MAX_OPT_ARGS];
1650 	int token;
1651 
1652 	if (!options)
1653 		return 1;
1654 
1655 	while ((p = strsep(&options, ",")) != NULL) {
1656 		if (!*p)
1657 			continue;
1658 		/*
1659 		 * Initialize args struct so we know whether arg was
1660 		 * found; some options take optional arguments.
1661 		 */
1662 		args[0].to = args[0].from = NULL;
1663 		token = match_token(p, tokens, args);
1664 		if (handle_mount_opt(sb, p, token, args, journal_devnum,
1665 				     journal_ioprio, is_remount) < 0)
1666 			return 0;
1667 	}
1668 #ifdef CONFIG_QUOTA
1669 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1670 	    (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1671 		ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1672 			 "feature is enabled");
1673 		return 0;
1674 	}
1675 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1676 		if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1677 			clear_opt(sb, USRQUOTA);
1678 
1679 		if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1680 			clear_opt(sb, GRPQUOTA);
1681 
1682 		if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1683 			ext4_msg(sb, KERN_ERR, "old and new quota "
1684 					"format mixing");
1685 			return 0;
1686 		}
1687 
1688 		if (!sbi->s_jquota_fmt) {
1689 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1690 					"not specified");
1691 			return 0;
1692 		}
1693 	}
1694 #endif
1695 	if (test_opt(sb, DIOREAD_NOLOCK)) {
1696 		int blocksize =
1697 			BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1698 
1699 		if (blocksize < PAGE_CACHE_SIZE) {
1700 			ext4_msg(sb, KERN_ERR, "can't mount with "
1701 				 "dioread_nolock if block size != PAGE_SIZE");
1702 			return 0;
1703 		}
1704 	}
1705 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1706 	    test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1707 		ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1708 			 "in data=ordered mode");
1709 		return 0;
1710 	}
1711 	return 1;
1712 }
1713 
1714 static inline void ext4_show_quota_options(struct seq_file *seq,
1715 					   struct super_block *sb)
1716 {
1717 #if defined(CONFIG_QUOTA)
1718 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1719 
1720 	if (sbi->s_jquota_fmt) {
1721 		char *fmtname = "";
1722 
1723 		switch (sbi->s_jquota_fmt) {
1724 		case QFMT_VFS_OLD:
1725 			fmtname = "vfsold";
1726 			break;
1727 		case QFMT_VFS_V0:
1728 			fmtname = "vfsv0";
1729 			break;
1730 		case QFMT_VFS_V1:
1731 			fmtname = "vfsv1";
1732 			break;
1733 		}
1734 		seq_printf(seq, ",jqfmt=%s", fmtname);
1735 	}
1736 
1737 	if (sbi->s_qf_names[USRQUOTA])
1738 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1739 
1740 	if (sbi->s_qf_names[GRPQUOTA])
1741 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1742 #endif
1743 }
1744 
1745 static const char *token2str(int token)
1746 {
1747 	const struct match_token *t;
1748 
1749 	for (t = tokens; t->token != Opt_err; t++)
1750 		if (t->token == token && !strchr(t->pattern, '='))
1751 			break;
1752 	return t->pattern;
1753 }
1754 
1755 /*
1756  * Show an option if
1757  *  - it's set to a non-default value OR
1758  *  - if the per-sb default is different from the global default
1759  */
1760 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1761 			      int nodefs)
1762 {
1763 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1764 	struct ext4_super_block *es = sbi->s_es;
1765 	int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1766 	const struct mount_opts *m;
1767 	char sep = nodefs ? '\n' : ',';
1768 
1769 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1770 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1771 
1772 	if (sbi->s_sb_block != 1)
1773 		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1774 
1775 	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1776 		int want_set = m->flags & MOPT_SET;
1777 		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1778 		    (m->flags & MOPT_CLEAR_ERR))
1779 			continue;
1780 		if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1781 			continue; /* skip if same as the default */
1782 		if ((want_set &&
1783 		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1784 		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1785 			continue; /* select Opt_noFoo vs Opt_Foo */
1786 		SEQ_OPTS_PRINT("%s", token2str(m->token));
1787 	}
1788 
1789 	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1790 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1791 		SEQ_OPTS_PRINT("resuid=%u",
1792 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
1793 	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1794 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1795 		SEQ_OPTS_PRINT("resgid=%u",
1796 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
1797 	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1798 	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1799 		SEQ_OPTS_PUTS("errors=remount-ro");
1800 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1801 		SEQ_OPTS_PUTS("errors=continue");
1802 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1803 		SEQ_OPTS_PUTS("errors=panic");
1804 	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1805 		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1806 	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1807 		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1808 	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1809 		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1810 	if (sb->s_flags & MS_I_VERSION)
1811 		SEQ_OPTS_PUTS("i_version");
1812 	if (nodefs || sbi->s_stripe)
1813 		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1814 	if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1815 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1816 			SEQ_OPTS_PUTS("data=journal");
1817 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1818 			SEQ_OPTS_PUTS("data=ordered");
1819 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1820 			SEQ_OPTS_PUTS("data=writeback");
1821 	}
1822 	if (nodefs ||
1823 	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1824 		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1825 			       sbi->s_inode_readahead_blks);
1826 
1827 	if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1828 		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1829 		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1830 	if (nodefs || sbi->s_max_dir_size_kb)
1831 		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1832 
1833 	ext4_show_quota_options(seq, sb);
1834 	return 0;
1835 }
1836 
1837 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1838 {
1839 	return _ext4_show_options(seq, root->d_sb, 0);
1840 }
1841 
1842 static int options_seq_show(struct seq_file *seq, void *offset)
1843 {
1844 	struct super_block *sb = seq->private;
1845 	int rc;
1846 
1847 	seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1848 	rc = _ext4_show_options(seq, sb, 1);
1849 	seq_puts(seq, "\n");
1850 	return rc;
1851 }
1852 
1853 static int options_open_fs(struct inode *inode, struct file *file)
1854 {
1855 	return single_open(file, options_seq_show, PDE_DATA(inode));
1856 }
1857 
1858 static const struct file_operations ext4_seq_options_fops = {
1859 	.owner = THIS_MODULE,
1860 	.open = options_open_fs,
1861 	.read = seq_read,
1862 	.llseek = seq_lseek,
1863 	.release = single_release,
1864 };
1865 
1866 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1867 			    int read_only)
1868 {
1869 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1870 	int res = 0;
1871 
1872 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1873 		ext4_msg(sb, KERN_ERR, "revision level too high, "
1874 			 "forcing read-only mode");
1875 		res = MS_RDONLY;
1876 	}
1877 	if (read_only)
1878 		goto done;
1879 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1880 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1881 			 "running e2fsck is recommended");
1882 	else if (sbi->s_mount_state & EXT4_ERROR_FS)
1883 		ext4_msg(sb, KERN_WARNING,
1884 			 "warning: mounting fs with errors, "
1885 			 "running e2fsck is recommended");
1886 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1887 		 le16_to_cpu(es->s_mnt_count) >=
1888 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1889 		ext4_msg(sb, KERN_WARNING,
1890 			 "warning: maximal mount count reached, "
1891 			 "running e2fsck is recommended");
1892 	else if (le32_to_cpu(es->s_checkinterval) &&
1893 		(le32_to_cpu(es->s_lastcheck) +
1894 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1895 		ext4_msg(sb, KERN_WARNING,
1896 			 "warning: checktime reached, "
1897 			 "running e2fsck is recommended");
1898 	if (!sbi->s_journal)
1899 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1900 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1901 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1902 	le16_add_cpu(&es->s_mnt_count, 1);
1903 	es->s_mtime = cpu_to_le32(get_seconds());
1904 	ext4_update_dynamic_rev(sb);
1905 	if (sbi->s_journal)
1906 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1907 
1908 	ext4_commit_super(sb, 1);
1909 done:
1910 	if (test_opt(sb, DEBUG))
1911 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1912 				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1913 			sb->s_blocksize,
1914 			sbi->s_groups_count,
1915 			EXT4_BLOCKS_PER_GROUP(sb),
1916 			EXT4_INODES_PER_GROUP(sb),
1917 			sbi->s_mount_opt, sbi->s_mount_opt2);
1918 
1919 	cleancache_init_fs(sb);
1920 	return res;
1921 }
1922 
1923 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1924 {
1925 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1926 	struct flex_groups *new_groups;
1927 	int size;
1928 
1929 	if (!sbi->s_log_groups_per_flex)
1930 		return 0;
1931 
1932 	size = ext4_flex_group(sbi, ngroup - 1) + 1;
1933 	if (size <= sbi->s_flex_groups_allocated)
1934 		return 0;
1935 
1936 	size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1937 	new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1938 	if (!new_groups) {
1939 		ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1940 			 size / (int) sizeof(struct flex_groups));
1941 		return -ENOMEM;
1942 	}
1943 
1944 	if (sbi->s_flex_groups) {
1945 		memcpy(new_groups, sbi->s_flex_groups,
1946 		       (sbi->s_flex_groups_allocated *
1947 			sizeof(struct flex_groups)));
1948 		kvfree(sbi->s_flex_groups);
1949 	}
1950 	sbi->s_flex_groups = new_groups;
1951 	sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1952 	return 0;
1953 }
1954 
1955 static int ext4_fill_flex_info(struct super_block *sb)
1956 {
1957 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1958 	struct ext4_group_desc *gdp = NULL;
1959 	ext4_group_t flex_group;
1960 	int i, err;
1961 
1962 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1963 	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1964 		sbi->s_log_groups_per_flex = 0;
1965 		return 1;
1966 	}
1967 
1968 	err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1969 	if (err)
1970 		goto failed;
1971 
1972 	for (i = 0; i < sbi->s_groups_count; i++) {
1973 		gdp = ext4_get_group_desc(sb, i, NULL);
1974 
1975 		flex_group = ext4_flex_group(sbi, i);
1976 		atomic_add(ext4_free_inodes_count(sb, gdp),
1977 			   &sbi->s_flex_groups[flex_group].free_inodes);
1978 		atomic64_add(ext4_free_group_clusters(sb, gdp),
1979 			     &sbi->s_flex_groups[flex_group].free_clusters);
1980 		atomic_add(ext4_used_dirs_count(sb, gdp),
1981 			   &sbi->s_flex_groups[flex_group].used_dirs);
1982 	}
1983 
1984 	return 1;
1985 failed:
1986 	return 0;
1987 }
1988 
1989 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1990 				   struct ext4_group_desc *gdp)
1991 {
1992 	int offset;
1993 	__u16 crc = 0;
1994 	__le32 le_group = cpu_to_le32(block_group);
1995 
1996 	if (ext4_has_metadata_csum(sbi->s_sb)) {
1997 		/* Use new metadata_csum algorithm */
1998 		__le16 save_csum;
1999 		__u32 csum32;
2000 
2001 		save_csum = gdp->bg_checksum;
2002 		gdp->bg_checksum = 0;
2003 		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2004 				     sizeof(le_group));
2005 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2006 				     sbi->s_desc_size);
2007 		gdp->bg_checksum = save_csum;
2008 
2009 		crc = csum32 & 0xFFFF;
2010 		goto out;
2011 	}
2012 
2013 	/* old crc16 code */
2014 	if (!(sbi->s_es->s_feature_ro_compat &
2015 	      cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2016 		return 0;
2017 
2018 	offset = offsetof(struct ext4_group_desc, bg_checksum);
2019 
2020 	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2021 	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2022 	crc = crc16(crc, (__u8 *)gdp, offset);
2023 	offset += sizeof(gdp->bg_checksum); /* skip checksum */
2024 	/* for checksum of struct ext4_group_desc do the rest...*/
2025 	if ((sbi->s_es->s_feature_incompat &
2026 	     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2027 	    offset < le16_to_cpu(sbi->s_es->s_desc_size))
2028 		crc = crc16(crc, (__u8 *)gdp + offset,
2029 			    le16_to_cpu(sbi->s_es->s_desc_size) -
2030 				offset);
2031 
2032 out:
2033 	return cpu_to_le16(crc);
2034 }
2035 
2036 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2037 				struct ext4_group_desc *gdp)
2038 {
2039 	if (ext4_has_group_desc_csum(sb) &&
2040 	    (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2041 						      block_group, gdp)))
2042 		return 0;
2043 
2044 	return 1;
2045 }
2046 
2047 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2048 			      struct ext4_group_desc *gdp)
2049 {
2050 	if (!ext4_has_group_desc_csum(sb))
2051 		return;
2052 	gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2053 }
2054 
2055 /* Called at mount-time, super-block is locked */
2056 static int ext4_check_descriptors(struct super_block *sb,
2057 				  ext4_group_t *first_not_zeroed)
2058 {
2059 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2060 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2061 	ext4_fsblk_t last_block;
2062 	ext4_fsblk_t block_bitmap;
2063 	ext4_fsblk_t inode_bitmap;
2064 	ext4_fsblk_t inode_table;
2065 	int flexbg_flag = 0;
2066 	ext4_group_t i, grp = sbi->s_groups_count;
2067 
2068 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2069 		flexbg_flag = 1;
2070 
2071 	ext4_debug("Checking group descriptors");
2072 
2073 	for (i = 0; i < sbi->s_groups_count; i++) {
2074 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2075 
2076 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
2077 			last_block = ext4_blocks_count(sbi->s_es) - 1;
2078 		else
2079 			last_block = first_block +
2080 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
2081 
2082 		if ((grp == sbi->s_groups_count) &&
2083 		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2084 			grp = i;
2085 
2086 		block_bitmap = ext4_block_bitmap(sb, gdp);
2087 		if (block_bitmap < first_block || block_bitmap > last_block) {
2088 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2089 			       "Block bitmap for group %u not in group "
2090 			       "(block %llu)!", i, block_bitmap);
2091 			return 0;
2092 		}
2093 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
2094 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
2095 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2096 			       "Inode bitmap for group %u not in group "
2097 			       "(block %llu)!", i, inode_bitmap);
2098 			return 0;
2099 		}
2100 		inode_table = ext4_inode_table(sb, gdp);
2101 		if (inode_table < first_block ||
2102 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2103 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2104 			       "Inode table for group %u not in group "
2105 			       "(block %llu)!", i, inode_table);
2106 			return 0;
2107 		}
2108 		ext4_lock_group(sb, i);
2109 		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2110 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2111 				 "Checksum for group %u failed (%u!=%u)",
2112 				 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2113 				     gdp)), le16_to_cpu(gdp->bg_checksum));
2114 			if (!(sb->s_flags & MS_RDONLY)) {
2115 				ext4_unlock_group(sb, i);
2116 				return 0;
2117 			}
2118 		}
2119 		ext4_unlock_group(sb, i);
2120 		if (!flexbg_flag)
2121 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
2122 	}
2123 	if (NULL != first_not_zeroed)
2124 		*first_not_zeroed = grp;
2125 	return 1;
2126 }
2127 
2128 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2129  * the superblock) which were deleted from all directories, but held open by
2130  * a process at the time of a crash.  We walk the list and try to delete these
2131  * inodes at recovery time (only with a read-write filesystem).
2132  *
2133  * In order to keep the orphan inode chain consistent during traversal (in
2134  * case of crash during recovery), we link each inode into the superblock
2135  * orphan list_head and handle it the same way as an inode deletion during
2136  * normal operation (which journals the operations for us).
2137  *
2138  * We only do an iget() and an iput() on each inode, which is very safe if we
2139  * accidentally point at an in-use or already deleted inode.  The worst that
2140  * can happen in this case is that we get a "bit already cleared" message from
2141  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2142  * e2fsck was run on this filesystem, and it must have already done the orphan
2143  * inode cleanup for us, so we can safely abort without any further action.
2144  */
2145 static void ext4_orphan_cleanup(struct super_block *sb,
2146 				struct ext4_super_block *es)
2147 {
2148 	unsigned int s_flags = sb->s_flags;
2149 	int nr_orphans = 0, nr_truncates = 0;
2150 #ifdef CONFIG_QUOTA
2151 	int i;
2152 #endif
2153 	if (!es->s_last_orphan) {
2154 		jbd_debug(4, "no orphan inodes to clean up\n");
2155 		return;
2156 	}
2157 
2158 	if (bdev_read_only(sb->s_bdev)) {
2159 		ext4_msg(sb, KERN_ERR, "write access "
2160 			"unavailable, skipping orphan cleanup");
2161 		return;
2162 	}
2163 
2164 	/* Check if feature set would not allow a r/w mount */
2165 	if (!ext4_feature_set_ok(sb, 0)) {
2166 		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2167 			 "unknown ROCOMPAT features");
2168 		return;
2169 	}
2170 
2171 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2172 		/* don't clear list on RO mount w/ errors */
2173 		if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2174 			ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2175 				  "clearing orphan list.\n");
2176 			es->s_last_orphan = 0;
2177 		}
2178 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2179 		return;
2180 	}
2181 
2182 	if (s_flags & MS_RDONLY) {
2183 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2184 		sb->s_flags &= ~MS_RDONLY;
2185 	}
2186 #ifdef CONFIG_QUOTA
2187 	/* Needed for iput() to work correctly and not trash data */
2188 	sb->s_flags |= MS_ACTIVE;
2189 	/* Turn on quotas so that they are updated correctly */
2190 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2191 		if (EXT4_SB(sb)->s_qf_names[i]) {
2192 			int ret = ext4_quota_on_mount(sb, i);
2193 			if (ret < 0)
2194 				ext4_msg(sb, KERN_ERR,
2195 					"Cannot turn on journaled "
2196 					"quota: error %d", ret);
2197 		}
2198 	}
2199 #endif
2200 
2201 	while (es->s_last_orphan) {
2202 		struct inode *inode;
2203 
2204 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2205 		if (IS_ERR(inode)) {
2206 			es->s_last_orphan = 0;
2207 			break;
2208 		}
2209 
2210 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2211 		dquot_initialize(inode);
2212 		if (inode->i_nlink) {
2213 			if (test_opt(sb, DEBUG))
2214 				ext4_msg(sb, KERN_DEBUG,
2215 					"%s: truncating inode %lu to %lld bytes",
2216 					__func__, inode->i_ino, inode->i_size);
2217 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2218 				  inode->i_ino, inode->i_size);
2219 			mutex_lock(&inode->i_mutex);
2220 			truncate_inode_pages(inode->i_mapping, inode->i_size);
2221 			ext4_truncate(inode);
2222 			mutex_unlock(&inode->i_mutex);
2223 			nr_truncates++;
2224 		} else {
2225 			if (test_opt(sb, DEBUG))
2226 				ext4_msg(sb, KERN_DEBUG,
2227 					"%s: deleting unreferenced inode %lu",
2228 					__func__, inode->i_ino);
2229 			jbd_debug(2, "deleting unreferenced inode %lu\n",
2230 				  inode->i_ino);
2231 			nr_orphans++;
2232 		}
2233 		iput(inode);  /* The delete magic happens here! */
2234 	}
2235 
2236 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2237 
2238 	if (nr_orphans)
2239 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2240 		       PLURAL(nr_orphans));
2241 	if (nr_truncates)
2242 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2243 		       PLURAL(nr_truncates));
2244 #ifdef CONFIG_QUOTA
2245 	/* Turn quotas off */
2246 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2247 		if (sb_dqopt(sb)->files[i])
2248 			dquot_quota_off(sb, i);
2249 	}
2250 #endif
2251 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2252 }
2253 
2254 /*
2255  * Maximal extent format file size.
2256  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2257  * extent format containers, within a sector_t, and within i_blocks
2258  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2259  * so that won't be a limiting factor.
2260  *
2261  * However there is other limiting factor. We do store extents in the form
2262  * of starting block and length, hence the resulting length of the extent
2263  * covering maximum file size must fit into on-disk format containers as
2264  * well. Given that length is always by 1 unit bigger than max unit (because
2265  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2266  *
2267  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2268  */
2269 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2270 {
2271 	loff_t res;
2272 	loff_t upper_limit = MAX_LFS_FILESIZE;
2273 
2274 	/* small i_blocks in vfs inode? */
2275 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2276 		/*
2277 		 * CONFIG_LBDAF is not enabled implies the inode
2278 		 * i_block represent total blocks in 512 bytes
2279 		 * 32 == size of vfs inode i_blocks * 8
2280 		 */
2281 		upper_limit = (1LL << 32) - 1;
2282 
2283 		/* total blocks in file system block size */
2284 		upper_limit >>= (blkbits - 9);
2285 		upper_limit <<= blkbits;
2286 	}
2287 
2288 	/*
2289 	 * 32-bit extent-start container, ee_block. We lower the maxbytes
2290 	 * by one fs block, so ee_len can cover the extent of maximum file
2291 	 * size
2292 	 */
2293 	res = (1LL << 32) - 1;
2294 	res <<= blkbits;
2295 
2296 	/* Sanity check against vm- & vfs- imposed limits */
2297 	if (res > upper_limit)
2298 		res = upper_limit;
2299 
2300 	return res;
2301 }
2302 
2303 /*
2304  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2305  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2306  * We need to be 1 filesystem block less than the 2^48 sector limit.
2307  */
2308 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2309 {
2310 	loff_t res = EXT4_NDIR_BLOCKS;
2311 	int meta_blocks;
2312 	loff_t upper_limit;
2313 	/* This is calculated to be the largest file size for a dense, block
2314 	 * mapped file such that the file's total number of 512-byte sectors,
2315 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
2316 	 *
2317 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2318 	 * number of 512-byte sectors of the file.
2319 	 */
2320 
2321 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2322 		/*
2323 		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2324 		 * the inode i_block field represents total file blocks in
2325 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2326 		 */
2327 		upper_limit = (1LL << 32) - 1;
2328 
2329 		/* total blocks in file system block size */
2330 		upper_limit >>= (bits - 9);
2331 
2332 	} else {
2333 		/*
2334 		 * We use 48 bit ext4_inode i_blocks
2335 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2336 		 * represent total number of blocks in
2337 		 * file system block size
2338 		 */
2339 		upper_limit = (1LL << 48) - 1;
2340 
2341 	}
2342 
2343 	/* indirect blocks */
2344 	meta_blocks = 1;
2345 	/* double indirect blocks */
2346 	meta_blocks += 1 + (1LL << (bits-2));
2347 	/* tripple indirect blocks */
2348 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2349 
2350 	upper_limit -= meta_blocks;
2351 	upper_limit <<= bits;
2352 
2353 	res += 1LL << (bits-2);
2354 	res += 1LL << (2*(bits-2));
2355 	res += 1LL << (3*(bits-2));
2356 	res <<= bits;
2357 	if (res > upper_limit)
2358 		res = upper_limit;
2359 
2360 	if (res > MAX_LFS_FILESIZE)
2361 		res = MAX_LFS_FILESIZE;
2362 
2363 	return res;
2364 }
2365 
2366 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2367 				   ext4_fsblk_t logical_sb_block, int nr)
2368 {
2369 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2370 	ext4_group_t bg, first_meta_bg;
2371 	int has_super = 0;
2372 
2373 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2374 
2375 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2376 	    nr < first_meta_bg)
2377 		return logical_sb_block + nr + 1;
2378 	bg = sbi->s_desc_per_block * nr;
2379 	if (ext4_bg_has_super(sb, bg))
2380 		has_super = 1;
2381 
2382 	/*
2383 	 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2384 	 * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
2385 	 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2386 	 * compensate.
2387 	 */
2388 	if (sb->s_blocksize == 1024 && nr == 0 &&
2389 	    le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2390 		has_super++;
2391 
2392 	return (has_super + ext4_group_first_block_no(sb, bg));
2393 }
2394 
2395 /**
2396  * ext4_get_stripe_size: Get the stripe size.
2397  * @sbi: In memory super block info
2398  *
2399  * If we have specified it via mount option, then
2400  * use the mount option value. If the value specified at mount time is
2401  * greater than the blocks per group use the super block value.
2402  * If the super block value is greater than blocks per group return 0.
2403  * Allocator needs it be less than blocks per group.
2404  *
2405  */
2406 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2407 {
2408 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2409 	unsigned long stripe_width =
2410 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2411 	int ret;
2412 
2413 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2414 		ret = sbi->s_stripe;
2415 	else if (stripe_width <= sbi->s_blocks_per_group)
2416 		ret = stripe_width;
2417 	else if (stride <= sbi->s_blocks_per_group)
2418 		ret = stride;
2419 	else
2420 		ret = 0;
2421 
2422 	/*
2423 	 * If the stripe width is 1, this makes no sense and
2424 	 * we set it to 0 to turn off stripe handling code.
2425 	 */
2426 	if (ret <= 1)
2427 		ret = 0;
2428 
2429 	return ret;
2430 }
2431 
2432 /* sysfs supprt */
2433 
2434 struct ext4_attr {
2435 	struct attribute attr;
2436 	ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2437 	ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2438 			 const char *, size_t);
2439 	union {
2440 		int offset;
2441 		int deprecated_val;
2442 	} u;
2443 };
2444 
2445 static int parse_strtoull(const char *buf,
2446 		unsigned long long max, unsigned long long *value)
2447 {
2448 	int ret;
2449 
2450 	ret = kstrtoull(skip_spaces(buf), 0, value);
2451 	if (!ret && *value > max)
2452 		ret = -EINVAL;
2453 	return ret;
2454 }
2455 
2456 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2457 					      struct ext4_sb_info *sbi,
2458 					      char *buf)
2459 {
2460 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2461 		(s64) EXT4_C2B(sbi,
2462 			percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2463 }
2464 
2465 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2466 					 struct ext4_sb_info *sbi, char *buf)
2467 {
2468 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2469 
2470 	if (!sb->s_bdev->bd_part)
2471 		return snprintf(buf, PAGE_SIZE, "0\n");
2472 	return snprintf(buf, PAGE_SIZE, "%lu\n",
2473 			(part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2474 			 sbi->s_sectors_written_start) >> 1);
2475 }
2476 
2477 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2478 					  struct ext4_sb_info *sbi, char *buf)
2479 {
2480 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2481 
2482 	if (!sb->s_bdev->bd_part)
2483 		return snprintf(buf, PAGE_SIZE, "0\n");
2484 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2485 			(unsigned long long)(sbi->s_kbytes_written +
2486 			((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2487 			  EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2488 }
2489 
2490 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2491 					  struct ext4_sb_info *sbi,
2492 					  const char *buf, size_t count)
2493 {
2494 	unsigned long t;
2495 	int ret;
2496 
2497 	ret = kstrtoul(skip_spaces(buf), 0, &t);
2498 	if (ret)
2499 		return ret;
2500 
2501 	if (t && (!is_power_of_2(t) || t > 0x40000000))
2502 		return -EINVAL;
2503 
2504 	sbi->s_inode_readahead_blks = t;
2505 	return count;
2506 }
2507 
2508 static ssize_t sbi_ui_show(struct ext4_attr *a,
2509 			   struct ext4_sb_info *sbi, char *buf)
2510 {
2511 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2512 
2513 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2514 }
2515 
2516 static ssize_t sbi_ui_store(struct ext4_attr *a,
2517 			    struct ext4_sb_info *sbi,
2518 			    const char *buf, size_t count)
2519 {
2520 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2521 	unsigned long t;
2522 	int ret;
2523 
2524 	ret = kstrtoul(skip_spaces(buf), 0, &t);
2525 	if (ret)
2526 		return ret;
2527 	*ui = t;
2528 	return count;
2529 }
2530 
2531 static ssize_t es_ui_show(struct ext4_attr *a,
2532 			   struct ext4_sb_info *sbi, char *buf)
2533 {
2534 
2535 	unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2536 			   a->u.offset);
2537 
2538 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2539 }
2540 
2541 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2542 				  struct ext4_sb_info *sbi, char *buf)
2543 {
2544 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2545 		(unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2546 }
2547 
2548 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2549 				   struct ext4_sb_info *sbi,
2550 				   const char *buf, size_t count)
2551 {
2552 	unsigned long long val;
2553 	int ret;
2554 
2555 	if (parse_strtoull(buf, -1ULL, &val))
2556 		return -EINVAL;
2557 	ret = ext4_reserve_clusters(sbi, val);
2558 
2559 	return ret ? ret : count;
2560 }
2561 
2562 static ssize_t trigger_test_error(struct ext4_attr *a,
2563 				  struct ext4_sb_info *sbi,
2564 				  const char *buf, size_t count)
2565 {
2566 	int len = count;
2567 
2568 	if (!capable(CAP_SYS_ADMIN))
2569 		return -EPERM;
2570 
2571 	if (len && buf[len-1] == '\n')
2572 		len--;
2573 
2574 	if (len)
2575 		ext4_error(sbi->s_sb, "%.*s", len, buf);
2576 	return count;
2577 }
2578 
2579 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2580 				   struct ext4_sb_info *sbi, char *buf)
2581 {
2582 	return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2583 }
2584 
2585 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2586 static struct ext4_attr ext4_attr_##_name = {			\
2587 	.attr = {.name = __stringify(_name), .mode = _mode },	\
2588 	.show	= _show,					\
2589 	.store	= _store,					\
2590 	.u = {							\
2591 		.offset = offsetof(struct ext4_sb_info, _elname),\
2592 	},							\
2593 }
2594 
2595 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname)		\
2596 static struct ext4_attr ext4_attr_##_name = {				\
2597 	.attr = {.name = __stringify(_name), .mode = _mode },		\
2598 	.show	= _show,						\
2599 	.store	= _store,						\
2600 	.u = {								\
2601 		.offset = offsetof(struct ext4_super_block, _elname),	\
2602 	},								\
2603 }
2604 
2605 #define EXT4_ATTR(name, mode, show, store) \
2606 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2607 
2608 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2609 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2610 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2611 
2612 #define EXT4_RO_ATTR_ES_UI(name, elname)	\
2613 	EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2614 #define EXT4_RW_ATTR_SBI_UI(name, elname)	\
2615 	EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2616 
2617 #define ATTR_LIST(name) &ext4_attr_##name.attr
2618 #define EXT4_DEPRECATED_ATTR(_name, _val)	\
2619 static struct ext4_attr ext4_attr_##_name = {			\
2620 	.attr = {.name = __stringify(_name), .mode = 0444 },	\
2621 	.show	= sbi_deprecated_show,				\
2622 	.u = {							\
2623 		.deprecated_val = _val,				\
2624 	},							\
2625 }
2626 
2627 EXT4_RO_ATTR(delayed_allocation_blocks);
2628 EXT4_RO_ATTR(session_write_kbytes);
2629 EXT4_RO_ATTR(lifetime_write_kbytes);
2630 EXT4_RW_ATTR(reserved_clusters);
2631 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2632 		 inode_readahead_blks_store, s_inode_readahead_blks);
2633 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2634 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2635 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2636 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2637 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2638 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2639 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2640 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2641 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2642 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2643 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2644 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2645 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2646 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2647 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2648 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2649 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2650 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2651 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2652 
2653 static struct attribute *ext4_attrs[] = {
2654 	ATTR_LIST(delayed_allocation_blocks),
2655 	ATTR_LIST(session_write_kbytes),
2656 	ATTR_LIST(lifetime_write_kbytes),
2657 	ATTR_LIST(reserved_clusters),
2658 	ATTR_LIST(inode_readahead_blks),
2659 	ATTR_LIST(inode_goal),
2660 	ATTR_LIST(mb_stats),
2661 	ATTR_LIST(mb_max_to_scan),
2662 	ATTR_LIST(mb_min_to_scan),
2663 	ATTR_LIST(mb_order2_req),
2664 	ATTR_LIST(mb_stream_req),
2665 	ATTR_LIST(mb_group_prealloc),
2666 	ATTR_LIST(max_writeback_mb_bump),
2667 	ATTR_LIST(extent_max_zeroout_kb),
2668 	ATTR_LIST(trigger_fs_error),
2669 	ATTR_LIST(err_ratelimit_interval_ms),
2670 	ATTR_LIST(err_ratelimit_burst),
2671 	ATTR_LIST(warning_ratelimit_interval_ms),
2672 	ATTR_LIST(warning_ratelimit_burst),
2673 	ATTR_LIST(msg_ratelimit_interval_ms),
2674 	ATTR_LIST(msg_ratelimit_burst),
2675 	ATTR_LIST(errors_count),
2676 	ATTR_LIST(first_error_time),
2677 	ATTR_LIST(last_error_time),
2678 	NULL,
2679 };
2680 
2681 /* Features this copy of ext4 supports */
2682 EXT4_INFO_ATTR(lazy_itable_init);
2683 EXT4_INFO_ATTR(batched_discard);
2684 EXT4_INFO_ATTR(meta_bg_resize);
2685 
2686 static struct attribute *ext4_feat_attrs[] = {
2687 	ATTR_LIST(lazy_itable_init),
2688 	ATTR_LIST(batched_discard),
2689 	ATTR_LIST(meta_bg_resize),
2690 	NULL,
2691 };
2692 
2693 static ssize_t ext4_attr_show(struct kobject *kobj,
2694 			      struct attribute *attr, char *buf)
2695 {
2696 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2697 						s_kobj);
2698 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2699 
2700 	return a->show ? a->show(a, sbi, buf) : 0;
2701 }
2702 
2703 static ssize_t ext4_attr_store(struct kobject *kobj,
2704 			       struct attribute *attr,
2705 			       const char *buf, size_t len)
2706 {
2707 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2708 						s_kobj);
2709 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2710 
2711 	return a->store ? a->store(a, sbi, buf, len) : 0;
2712 }
2713 
2714 static void ext4_sb_release(struct kobject *kobj)
2715 {
2716 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2717 						s_kobj);
2718 	complete(&sbi->s_kobj_unregister);
2719 }
2720 
2721 static const struct sysfs_ops ext4_attr_ops = {
2722 	.show	= ext4_attr_show,
2723 	.store	= ext4_attr_store,
2724 };
2725 
2726 static struct kobj_type ext4_ktype = {
2727 	.default_attrs	= ext4_attrs,
2728 	.sysfs_ops	= &ext4_attr_ops,
2729 	.release	= ext4_sb_release,
2730 };
2731 
2732 static void ext4_feat_release(struct kobject *kobj)
2733 {
2734 	complete(&ext4_feat->f_kobj_unregister);
2735 }
2736 
2737 static ssize_t ext4_feat_show(struct kobject *kobj,
2738 			      struct attribute *attr, char *buf)
2739 {
2740 	return snprintf(buf, PAGE_SIZE, "supported\n");
2741 }
2742 
2743 /*
2744  * We can not use ext4_attr_show/store because it relies on the kobject
2745  * being embedded in the ext4_sb_info structure which is definitely not
2746  * true in this case.
2747  */
2748 static const struct sysfs_ops ext4_feat_ops = {
2749 	.show	= ext4_feat_show,
2750 	.store	= NULL,
2751 };
2752 
2753 static struct kobj_type ext4_feat_ktype = {
2754 	.default_attrs	= ext4_feat_attrs,
2755 	.sysfs_ops	= &ext4_feat_ops,
2756 	.release	= ext4_feat_release,
2757 };
2758 
2759 /*
2760  * Check whether this filesystem can be mounted based on
2761  * the features present and the RDONLY/RDWR mount requested.
2762  * Returns 1 if this filesystem can be mounted as requested,
2763  * 0 if it cannot be.
2764  */
2765 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2766 {
2767 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2768 		ext4_msg(sb, KERN_ERR,
2769 			"Couldn't mount because of "
2770 			"unsupported optional features (%x)",
2771 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2772 			~EXT4_FEATURE_INCOMPAT_SUPP));
2773 		return 0;
2774 	}
2775 
2776 	if (readonly)
2777 		return 1;
2778 
2779 	/* Check that feature set is OK for a read-write mount */
2780 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2781 		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2782 			 "unsupported optional features (%x)",
2783 			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2784 				~EXT4_FEATURE_RO_COMPAT_SUPP));
2785 		return 0;
2786 	}
2787 	/*
2788 	 * Large file size enabled file system can only be mounted
2789 	 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2790 	 */
2791 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2792 		if (sizeof(blkcnt_t) < sizeof(u64)) {
2793 			ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2794 				 "cannot be mounted RDWR without "
2795 				 "CONFIG_LBDAF");
2796 			return 0;
2797 		}
2798 	}
2799 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2800 	    !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2801 		ext4_msg(sb, KERN_ERR,
2802 			 "Can't support bigalloc feature without "
2803 			 "extents feature\n");
2804 		return 0;
2805 	}
2806 
2807 #ifndef CONFIG_QUOTA
2808 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2809 	    !readonly) {
2810 		ext4_msg(sb, KERN_ERR,
2811 			 "Filesystem with quota feature cannot be mounted RDWR "
2812 			 "without CONFIG_QUOTA");
2813 		return 0;
2814 	}
2815 #endif  /* CONFIG_QUOTA */
2816 	return 1;
2817 }
2818 
2819 /*
2820  * This function is called once a day if we have errors logged
2821  * on the file system
2822  */
2823 static void print_daily_error_info(unsigned long arg)
2824 {
2825 	struct super_block *sb = (struct super_block *) arg;
2826 	struct ext4_sb_info *sbi;
2827 	struct ext4_super_block *es;
2828 
2829 	sbi = EXT4_SB(sb);
2830 	es = sbi->s_es;
2831 
2832 	if (es->s_error_count)
2833 		/* fsck newer than v1.41.13 is needed to clean this condition. */
2834 		ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2835 			 le32_to_cpu(es->s_error_count));
2836 	if (es->s_first_error_time) {
2837 		printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2838 		       sb->s_id, le32_to_cpu(es->s_first_error_time),
2839 		       (int) sizeof(es->s_first_error_func),
2840 		       es->s_first_error_func,
2841 		       le32_to_cpu(es->s_first_error_line));
2842 		if (es->s_first_error_ino)
2843 			printk(": inode %u",
2844 			       le32_to_cpu(es->s_first_error_ino));
2845 		if (es->s_first_error_block)
2846 			printk(": block %llu", (unsigned long long)
2847 			       le64_to_cpu(es->s_first_error_block));
2848 		printk("\n");
2849 	}
2850 	if (es->s_last_error_time) {
2851 		printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2852 		       sb->s_id, le32_to_cpu(es->s_last_error_time),
2853 		       (int) sizeof(es->s_last_error_func),
2854 		       es->s_last_error_func,
2855 		       le32_to_cpu(es->s_last_error_line));
2856 		if (es->s_last_error_ino)
2857 			printk(": inode %u",
2858 			       le32_to_cpu(es->s_last_error_ino));
2859 		if (es->s_last_error_block)
2860 			printk(": block %llu", (unsigned long long)
2861 			       le64_to_cpu(es->s_last_error_block));
2862 		printk("\n");
2863 	}
2864 	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2865 }
2866 
2867 /* Find next suitable group and run ext4_init_inode_table */
2868 static int ext4_run_li_request(struct ext4_li_request *elr)
2869 {
2870 	struct ext4_group_desc *gdp = NULL;
2871 	ext4_group_t group, ngroups;
2872 	struct super_block *sb;
2873 	unsigned long timeout = 0;
2874 	int ret = 0;
2875 
2876 	sb = elr->lr_super;
2877 	ngroups = EXT4_SB(sb)->s_groups_count;
2878 
2879 	sb_start_write(sb);
2880 	for (group = elr->lr_next_group; group < ngroups; group++) {
2881 		gdp = ext4_get_group_desc(sb, group, NULL);
2882 		if (!gdp) {
2883 			ret = 1;
2884 			break;
2885 		}
2886 
2887 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2888 			break;
2889 	}
2890 
2891 	if (group >= ngroups)
2892 		ret = 1;
2893 
2894 	if (!ret) {
2895 		timeout = jiffies;
2896 		ret = ext4_init_inode_table(sb, group,
2897 					    elr->lr_timeout ? 0 : 1);
2898 		if (elr->lr_timeout == 0) {
2899 			timeout = (jiffies - timeout) *
2900 				  elr->lr_sbi->s_li_wait_mult;
2901 			elr->lr_timeout = timeout;
2902 		}
2903 		elr->lr_next_sched = jiffies + elr->lr_timeout;
2904 		elr->lr_next_group = group + 1;
2905 	}
2906 	sb_end_write(sb);
2907 
2908 	return ret;
2909 }
2910 
2911 /*
2912  * Remove lr_request from the list_request and free the
2913  * request structure. Should be called with li_list_mtx held
2914  */
2915 static void ext4_remove_li_request(struct ext4_li_request *elr)
2916 {
2917 	struct ext4_sb_info *sbi;
2918 
2919 	if (!elr)
2920 		return;
2921 
2922 	sbi = elr->lr_sbi;
2923 
2924 	list_del(&elr->lr_request);
2925 	sbi->s_li_request = NULL;
2926 	kfree(elr);
2927 }
2928 
2929 static void ext4_unregister_li_request(struct super_block *sb)
2930 {
2931 	mutex_lock(&ext4_li_mtx);
2932 	if (!ext4_li_info) {
2933 		mutex_unlock(&ext4_li_mtx);
2934 		return;
2935 	}
2936 
2937 	mutex_lock(&ext4_li_info->li_list_mtx);
2938 	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2939 	mutex_unlock(&ext4_li_info->li_list_mtx);
2940 	mutex_unlock(&ext4_li_mtx);
2941 }
2942 
2943 static struct task_struct *ext4_lazyinit_task;
2944 
2945 /*
2946  * This is the function where ext4lazyinit thread lives. It walks
2947  * through the request list searching for next scheduled filesystem.
2948  * When such a fs is found, run the lazy initialization request
2949  * (ext4_rn_li_request) and keep track of the time spend in this
2950  * function. Based on that time we compute next schedule time of
2951  * the request. When walking through the list is complete, compute
2952  * next waking time and put itself into sleep.
2953  */
2954 static int ext4_lazyinit_thread(void *arg)
2955 {
2956 	struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2957 	struct list_head *pos, *n;
2958 	struct ext4_li_request *elr;
2959 	unsigned long next_wakeup, cur;
2960 
2961 	BUG_ON(NULL == eli);
2962 
2963 cont_thread:
2964 	while (true) {
2965 		next_wakeup = MAX_JIFFY_OFFSET;
2966 
2967 		mutex_lock(&eli->li_list_mtx);
2968 		if (list_empty(&eli->li_request_list)) {
2969 			mutex_unlock(&eli->li_list_mtx);
2970 			goto exit_thread;
2971 		}
2972 
2973 		list_for_each_safe(pos, n, &eli->li_request_list) {
2974 			elr = list_entry(pos, struct ext4_li_request,
2975 					 lr_request);
2976 
2977 			if (time_after_eq(jiffies, elr->lr_next_sched)) {
2978 				if (ext4_run_li_request(elr) != 0) {
2979 					/* error, remove the lazy_init job */
2980 					ext4_remove_li_request(elr);
2981 					continue;
2982 				}
2983 			}
2984 
2985 			if (time_before(elr->lr_next_sched, next_wakeup))
2986 				next_wakeup = elr->lr_next_sched;
2987 		}
2988 		mutex_unlock(&eli->li_list_mtx);
2989 
2990 		try_to_freeze();
2991 
2992 		cur = jiffies;
2993 		if ((time_after_eq(cur, next_wakeup)) ||
2994 		    (MAX_JIFFY_OFFSET == next_wakeup)) {
2995 			cond_resched();
2996 			continue;
2997 		}
2998 
2999 		schedule_timeout_interruptible(next_wakeup - cur);
3000 
3001 		if (kthread_should_stop()) {
3002 			ext4_clear_request_list();
3003 			goto exit_thread;
3004 		}
3005 	}
3006 
3007 exit_thread:
3008 	/*
3009 	 * It looks like the request list is empty, but we need
3010 	 * to check it under the li_list_mtx lock, to prevent any
3011 	 * additions into it, and of course we should lock ext4_li_mtx
3012 	 * to atomically free the list and ext4_li_info, because at
3013 	 * this point another ext4 filesystem could be registering
3014 	 * new one.
3015 	 */
3016 	mutex_lock(&ext4_li_mtx);
3017 	mutex_lock(&eli->li_list_mtx);
3018 	if (!list_empty(&eli->li_request_list)) {
3019 		mutex_unlock(&eli->li_list_mtx);
3020 		mutex_unlock(&ext4_li_mtx);
3021 		goto cont_thread;
3022 	}
3023 	mutex_unlock(&eli->li_list_mtx);
3024 	kfree(ext4_li_info);
3025 	ext4_li_info = NULL;
3026 	mutex_unlock(&ext4_li_mtx);
3027 
3028 	return 0;
3029 }
3030 
3031 static void ext4_clear_request_list(void)
3032 {
3033 	struct list_head *pos, *n;
3034 	struct ext4_li_request *elr;
3035 
3036 	mutex_lock(&ext4_li_info->li_list_mtx);
3037 	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3038 		elr = list_entry(pos, struct ext4_li_request,
3039 				 lr_request);
3040 		ext4_remove_li_request(elr);
3041 	}
3042 	mutex_unlock(&ext4_li_info->li_list_mtx);
3043 }
3044 
3045 static int ext4_run_lazyinit_thread(void)
3046 {
3047 	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3048 					 ext4_li_info, "ext4lazyinit");
3049 	if (IS_ERR(ext4_lazyinit_task)) {
3050 		int err = PTR_ERR(ext4_lazyinit_task);
3051 		ext4_clear_request_list();
3052 		kfree(ext4_li_info);
3053 		ext4_li_info = NULL;
3054 		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3055 				 "initialization thread\n",
3056 				 err);
3057 		return err;
3058 	}
3059 	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3060 	return 0;
3061 }
3062 
3063 /*
3064  * Check whether it make sense to run itable init. thread or not.
3065  * If there is at least one uninitialized inode table, return
3066  * corresponding group number, else the loop goes through all
3067  * groups and return total number of groups.
3068  */
3069 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3070 {
3071 	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3072 	struct ext4_group_desc *gdp = NULL;
3073 
3074 	for (group = 0; group < ngroups; group++) {
3075 		gdp = ext4_get_group_desc(sb, group, NULL);
3076 		if (!gdp)
3077 			continue;
3078 
3079 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3080 			break;
3081 	}
3082 
3083 	return group;
3084 }
3085 
3086 static int ext4_li_info_new(void)
3087 {
3088 	struct ext4_lazy_init *eli = NULL;
3089 
3090 	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3091 	if (!eli)
3092 		return -ENOMEM;
3093 
3094 	INIT_LIST_HEAD(&eli->li_request_list);
3095 	mutex_init(&eli->li_list_mtx);
3096 
3097 	eli->li_state |= EXT4_LAZYINIT_QUIT;
3098 
3099 	ext4_li_info = eli;
3100 
3101 	return 0;
3102 }
3103 
3104 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3105 					    ext4_group_t start)
3106 {
3107 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3108 	struct ext4_li_request *elr;
3109 
3110 	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3111 	if (!elr)
3112 		return NULL;
3113 
3114 	elr->lr_super = sb;
3115 	elr->lr_sbi = sbi;
3116 	elr->lr_next_group = start;
3117 
3118 	/*
3119 	 * Randomize first schedule time of the request to
3120 	 * spread the inode table initialization requests
3121 	 * better.
3122 	 */
3123 	elr->lr_next_sched = jiffies + (prandom_u32() %
3124 				(EXT4_DEF_LI_MAX_START_DELAY * HZ));
3125 	return elr;
3126 }
3127 
3128 int ext4_register_li_request(struct super_block *sb,
3129 			     ext4_group_t first_not_zeroed)
3130 {
3131 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3132 	struct ext4_li_request *elr = NULL;
3133 	ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3134 	int ret = 0;
3135 
3136 	mutex_lock(&ext4_li_mtx);
3137 	if (sbi->s_li_request != NULL) {
3138 		/*
3139 		 * Reset timeout so it can be computed again, because
3140 		 * s_li_wait_mult might have changed.
3141 		 */
3142 		sbi->s_li_request->lr_timeout = 0;
3143 		goto out;
3144 	}
3145 
3146 	if (first_not_zeroed == ngroups ||
3147 	    (sb->s_flags & MS_RDONLY) ||
3148 	    !test_opt(sb, INIT_INODE_TABLE))
3149 		goto out;
3150 
3151 	elr = ext4_li_request_new(sb, first_not_zeroed);
3152 	if (!elr) {
3153 		ret = -ENOMEM;
3154 		goto out;
3155 	}
3156 
3157 	if (NULL == ext4_li_info) {
3158 		ret = ext4_li_info_new();
3159 		if (ret)
3160 			goto out;
3161 	}
3162 
3163 	mutex_lock(&ext4_li_info->li_list_mtx);
3164 	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3165 	mutex_unlock(&ext4_li_info->li_list_mtx);
3166 
3167 	sbi->s_li_request = elr;
3168 	/*
3169 	 * set elr to NULL here since it has been inserted to
3170 	 * the request_list and the removal and free of it is
3171 	 * handled by ext4_clear_request_list from now on.
3172 	 */
3173 	elr = NULL;
3174 
3175 	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3176 		ret = ext4_run_lazyinit_thread();
3177 		if (ret)
3178 			goto out;
3179 	}
3180 out:
3181 	mutex_unlock(&ext4_li_mtx);
3182 	if (ret)
3183 		kfree(elr);
3184 	return ret;
3185 }
3186 
3187 /*
3188  * We do not need to lock anything since this is called on
3189  * module unload.
3190  */
3191 static void ext4_destroy_lazyinit_thread(void)
3192 {
3193 	/*
3194 	 * If thread exited earlier
3195 	 * there's nothing to be done.
3196 	 */
3197 	if (!ext4_li_info || !ext4_lazyinit_task)
3198 		return;
3199 
3200 	kthread_stop(ext4_lazyinit_task);
3201 }
3202 
3203 static int set_journal_csum_feature_set(struct super_block *sb)
3204 {
3205 	int ret = 1;
3206 	int compat, incompat;
3207 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3208 
3209 	if (ext4_has_metadata_csum(sb)) {
3210 		/* journal checksum v3 */
3211 		compat = 0;
3212 		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3213 	} else {
3214 		/* journal checksum v1 */
3215 		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3216 		incompat = 0;
3217 	}
3218 
3219 	jbd2_journal_clear_features(sbi->s_journal,
3220 			JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3221 			JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3222 			JBD2_FEATURE_INCOMPAT_CSUM_V2);
3223 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3224 		ret = jbd2_journal_set_features(sbi->s_journal,
3225 				compat, 0,
3226 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3227 				incompat);
3228 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3229 		ret = jbd2_journal_set_features(sbi->s_journal,
3230 				compat, 0,
3231 				incompat);
3232 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3233 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3234 	} else {
3235 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3236 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3237 	}
3238 
3239 	return ret;
3240 }
3241 
3242 /*
3243  * Note: calculating the overhead so we can be compatible with
3244  * historical BSD practice is quite difficult in the face of
3245  * clusters/bigalloc.  This is because multiple metadata blocks from
3246  * different block group can end up in the same allocation cluster.
3247  * Calculating the exact overhead in the face of clustered allocation
3248  * requires either O(all block bitmaps) in memory or O(number of block
3249  * groups**2) in time.  We will still calculate the superblock for
3250  * older file systems --- and if we come across with a bigalloc file
3251  * system with zero in s_overhead_clusters the estimate will be close to
3252  * correct especially for very large cluster sizes --- but for newer
3253  * file systems, it's better to calculate this figure once at mkfs
3254  * time, and store it in the superblock.  If the superblock value is
3255  * present (even for non-bigalloc file systems), we will use it.
3256  */
3257 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3258 			  char *buf)
3259 {
3260 	struct ext4_sb_info	*sbi = EXT4_SB(sb);
3261 	struct ext4_group_desc	*gdp;
3262 	ext4_fsblk_t		first_block, last_block, b;
3263 	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
3264 	int			s, j, count = 0;
3265 
3266 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3267 		return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3268 			sbi->s_itb_per_group + 2);
3269 
3270 	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3271 		(grp * EXT4_BLOCKS_PER_GROUP(sb));
3272 	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3273 	for (i = 0; i < ngroups; i++) {
3274 		gdp = ext4_get_group_desc(sb, i, NULL);
3275 		b = ext4_block_bitmap(sb, gdp);
3276 		if (b >= first_block && b <= last_block) {
3277 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3278 			count++;
3279 		}
3280 		b = ext4_inode_bitmap(sb, gdp);
3281 		if (b >= first_block && b <= last_block) {
3282 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3283 			count++;
3284 		}
3285 		b = ext4_inode_table(sb, gdp);
3286 		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3287 			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3288 				int c = EXT4_B2C(sbi, b - first_block);
3289 				ext4_set_bit(c, buf);
3290 				count++;
3291 			}
3292 		if (i != grp)
3293 			continue;
3294 		s = 0;
3295 		if (ext4_bg_has_super(sb, grp)) {
3296 			ext4_set_bit(s++, buf);
3297 			count++;
3298 		}
3299 		for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3300 			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3301 			count++;
3302 		}
3303 	}
3304 	if (!count)
3305 		return 0;
3306 	return EXT4_CLUSTERS_PER_GROUP(sb) -
3307 		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3308 }
3309 
3310 /*
3311  * Compute the overhead and stash it in sbi->s_overhead
3312  */
3313 int ext4_calculate_overhead(struct super_block *sb)
3314 {
3315 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3316 	struct ext4_super_block *es = sbi->s_es;
3317 	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3318 	ext4_fsblk_t overhead = 0;
3319 	char *buf = (char *) get_zeroed_page(GFP_NOFS);
3320 
3321 	if (!buf)
3322 		return -ENOMEM;
3323 
3324 	/*
3325 	 * Compute the overhead (FS structures).  This is constant
3326 	 * for a given filesystem unless the number of block groups
3327 	 * changes so we cache the previous value until it does.
3328 	 */
3329 
3330 	/*
3331 	 * All of the blocks before first_data_block are overhead
3332 	 */
3333 	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3334 
3335 	/*
3336 	 * Add the overhead found in each block group
3337 	 */
3338 	for (i = 0; i < ngroups; i++) {
3339 		int blks;
3340 
3341 		blks = count_overhead(sb, i, buf);
3342 		overhead += blks;
3343 		if (blks)
3344 			memset(buf, 0, PAGE_SIZE);
3345 		cond_resched();
3346 	}
3347 	/* Add the internal journal blocks as well */
3348 	if (sbi->s_journal && !sbi->journal_bdev)
3349 		overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3350 
3351 	sbi->s_overhead = overhead;
3352 	smp_wmb();
3353 	free_page((unsigned long) buf);
3354 	return 0;
3355 }
3356 
3357 
3358 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3359 {
3360 	ext4_fsblk_t resv_clusters;
3361 
3362 	/*
3363 	 * There's no need to reserve anything when we aren't using extents.
3364 	 * The space estimates are exact, there are no unwritten extents,
3365 	 * hole punching doesn't need new metadata... This is needed especially
3366 	 * to keep ext2/3 backward compatibility.
3367 	 */
3368 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3369 		return 0;
3370 	/*
3371 	 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3372 	 * This should cover the situations where we can not afford to run
3373 	 * out of space like for example punch hole, or converting
3374 	 * unwritten extents in delalloc path. In most cases such
3375 	 * allocation would require 1, or 2 blocks, higher numbers are
3376 	 * very rare.
3377 	 */
3378 	resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3379 			EXT4_SB(sb)->s_cluster_bits;
3380 
3381 	do_div(resv_clusters, 50);
3382 	resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3383 
3384 	return resv_clusters;
3385 }
3386 
3387 
3388 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3389 {
3390 	ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3391 				sbi->s_cluster_bits;
3392 
3393 	if (count >= clusters)
3394 		return -EINVAL;
3395 
3396 	atomic64_set(&sbi->s_resv_clusters, count);
3397 	return 0;
3398 }
3399 
3400 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3401 {
3402 	char *orig_data = kstrdup(data, GFP_KERNEL);
3403 	struct buffer_head *bh;
3404 	struct ext4_super_block *es = NULL;
3405 	struct ext4_sb_info *sbi;
3406 	ext4_fsblk_t block;
3407 	ext4_fsblk_t sb_block = get_sb_block(&data);
3408 	ext4_fsblk_t logical_sb_block;
3409 	unsigned long offset = 0;
3410 	unsigned long journal_devnum = 0;
3411 	unsigned long def_mount_opts;
3412 	struct inode *root;
3413 	char *cp;
3414 	const char *descr;
3415 	int ret = -ENOMEM;
3416 	int blocksize, clustersize;
3417 	unsigned int db_count;
3418 	unsigned int i;
3419 	int needs_recovery, has_huge_files, has_bigalloc;
3420 	__u64 blocks_count;
3421 	int err = 0;
3422 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3423 	ext4_group_t first_not_zeroed;
3424 
3425 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3426 	if (!sbi)
3427 		goto out_free_orig;
3428 
3429 	sbi->s_blockgroup_lock =
3430 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3431 	if (!sbi->s_blockgroup_lock) {
3432 		kfree(sbi);
3433 		goto out_free_orig;
3434 	}
3435 	sb->s_fs_info = sbi;
3436 	sbi->s_sb = sb;
3437 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3438 	sbi->s_sb_block = sb_block;
3439 	if (sb->s_bdev->bd_part)
3440 		sbi->s_sectors_written_start =
3441 			part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3442 
3443 	/* Cleanup superblock name */
3444 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3445 		*cp = '!';
3446 
3447 	/* -EINVAL is default */
3448 	ret = -EINVAL;
3449 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3450 	if (!blocksize) {
3451 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3452 		goto out_fail;
3453 	}
3454 
3455 	/*
3456 	 * The ext4 superblock will not be buffer aligned for other than 1kB
3457 	 * block sizes.  We need to calculate the offset from buffer start.
3458 	 */
3459 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3460 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3461 		offset = do_div(logical_sb_block, blocksize);
3462 	} else {
3463 		logical_sb_block = sb_block;
3464 	}
3465 
3466 	if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3467 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
3468 		goto out_fail;
3469 	}
3470 	/*
3471 	 * Note: s_es must be initialized as soon as possible because
3472 	 *       some ext4 macro-instructions depend on its value
3473 	 */
3474 	es = (struct ext4_super_block *) (bh->b_data + offset);
3475 	sbi->s_es = es;
3476 	sb->s_magic = le16_to_cpu(es->s_magic);
3477 	if (sb->s_magic != EXT4_SUPER_MAGIC)
3478 		goto cantfind_ext4;
3479 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3480 
3481 	/* Warn if metadata_csum and gdt_csum are both set. */
3482 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3483 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3484 	    EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3485 		ext4_warning(sb, "metadata_csum and uninit_bg are "
3486 			     "redundant flags; please run fsck.");
3487 
3488 	/* Check for a known checksum algorithm */
3489 	if (!ext4_verify_csum_type(sb, es)) {
3490 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3491 			 "unknown checksum algorithm.");
3492 		silent = 1;
3493 		goto cantfind_ext4;
3494 	}
3495 
3496 	/* Load the checksum driver */
3497 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3498 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3499 		sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3500 		if (IS_ERR(sbi->s_chksum_driver)) {
3501 			ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3502 			ret = PTR_ERR(sbi->s_chksum_driver);
3503 			sbi->s_chksum_driver = NULL;
3504 			goto failed_mount;
3505 		}
3506 	}
3507 
3508 	/* Check superblock checksum */
3509 	if (!ext4_superblock_csum_verify(sb, es)) {
3510 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3511 			 "invalid superblock checksum.  Run e2fsck?");
3512 		silent = 1;
3513 		goto cantfind_ext4;
3514 	}
3515 
3516 	/* Precompute checksum seed for all metadata */
3517 	if (ext4_has_metadata_csum(sb))
3518 		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3519 					       sizeof(es->s_uuid));
3520 
3521 	/* Set defaults before we parse the mount options */
3522 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3523 	set_opt(sb, INIT_INODE_TABLE);
3524 	if (def_mount_opts & EXT4_DEFM_DEBUG)
3525 		set_opt(sb, DEBUG);
3526 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3527 		set_opt(sb, GRPID);
3528 	if (def_mount_opts & EXT4_DEFM_UID16)
3529 		set_opt(sb, NO_UID32);
3530 	/* xattr user namespace & acls are now defaulted on */
3531 	set_opt(sb, XATTR_USER);
3532 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3533 	set_opt(sb, POSIX_ACL);
3534 #endif
3535 	/* don't forget to enable journal_csum when metadata_csum is enabled. */
3536 	if (ext4_has_metadata_csum(sb))
3537 		set_opt(sb, JOURNAL_CHECKSUM);
3538 
3539 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3540 		set_opt(sb, JOURNAL_DATA);
3541 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3542 		set_opt(sb, ORDERED_DATA);
3543 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3544 		set_opt(sb, WRITEBACK_DATA);
3545 
3546 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3547 		set_opt(sb, ERRORS_PANIC);
3548 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3549 		set_opt(sb, ERRORS_CONT);
3550 	else
3551 		set_opt(sb, ERRORS_RO);
3552 	/* block_validity enabled by default; disable with noblock_validity */
3553 	set_opt(sb, BLOCK_VALIDITY);
3554 	if (def_mount_opts & EXT4_DEFM_DISCARD)
3555 		set_opt(sb, DISCARD);
3556 
3557 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3558 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3559 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3560 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3561 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3562 
3563 	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3564 		set_opt(sb, BARRIER);
3565 
3566 	/*
3567 	 * enable delayed allocation by default
3568 	 * Use -o nodelalloc to turn it off
3569 	 */
3570 	if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3571 	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3572 		set_opt(sb, DELALLOC);
3573 
3574 	/*
3575 	 * set default s_li_wait_mult for lazyinit, for the case there is
3576 	 * no mount option specified.
3577 	 */
3578 	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3579 
3580 	if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3581 			   &journal_devnum, &journal_ioprio, 0)) {
3582 		ext4_msg(sb, KERN_WARNING,
3583 			 "failed to parse options in superblock: %s",
3584 			 sbi->s_es->s_mount_opts);
3585 	}
3586 	sbi->s_def_mount_opt = sbi->s_mount_opt;
3587 	if (!parse_options((char *) data, sb, &journal_devnum,
3588 			   &journal_ioprio, 0))
3589 		goto failed_mount;
3590 
3591 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3592 		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3593 			    "with data=journal disables delayed "
3594 			    "allocation and O_DIRECT support!\n");
3595 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3596 			ext4_msg(sb, KERN_ERR, "can't mount with "
3597 				 "both data=journal and delalloc");
3598 			goto failed_mount;
3599 		}
3600 		if (test_opt(sb, DIOREAD_NOLOCK)) {
3601 			ext4_msg(sb, KERN_ERR, "can't mount with "
3602 				 "both data=journal and dioread_nolock");
3603 			goto failed_mount;
3604 		}
3605 		if (test_opt(sb, DELALLOC))
3606 			clear_opt(sb, DELALLOC);
3607 	}
3608 
3609 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3610 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3611 
3612 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3613 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3614 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3615 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3616 		ext4_msg(sb, KERN_WARNING,
3617 		       "feature flags set on rev 0 fs, "
3618 		       "running e2fsck is recommended");
3619 
3620 	if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3621 		set_opt2(sb, HURD_COMPAT);
3622 		if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3623 					      EXT4_FEATURE_INCOMPAT_64BIT)) {
3624 			ext4_msg(sb, KERN_ERR,
3625 				 "The Hurd can't support 64-bit file systems");
3626 			goto failed_mount;
3627 		}
3628 	}
3629 
3630 	if (IS_EXT2_SB(sb)) {
3631 		if (ext2_feature_set_ok(sb))
3632 			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3633 				 "using the ext4 subsystem");
3634 		else {
3635 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3636 				 "to feature incompatibilities");
3637 			goto failed_mount;
3638 		}
3639 	}
3640 
3641 	if (IS_EXT3_SB(sb)) {
3642 		if (ext3_feature_set_ok(sb))
3643 			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3644 				 "using the ext4 subsystem");
3645 		else {
3646 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3647 				 "to feature incompatibilities");
3648 			goto failed_mount;
3649 		}
3650 	}
3651 
3652 	/*
3653 	 * Check feature flags regardless of the revision level, since we
3654 	 * previously didn't change the revision level when setting the flags,
3655 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
3656 	 */
3657 	if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3658 		goto failed_mount;
3659 
3660 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3661 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3662 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
3663 		ext4_msg(sb, KERN_ERR,
3664 		       "Unsupported filesystem blocksize %d", blocksize);
3665 		goto failed_mount;
3666 	}
3667 
3668 	if (sb->s_blocksize != blocksize) {
3669 		/* Validate the filesystem blocksize */
3670 		if (!sb_set_blocksize(sb, blocksize)) {
3671 			ext4_msg(sb, KERN_ERR, "bad block size %d",
3672 					blocksize);
3673 			goto failed_mount;
3674 		}
3675 
3676 		brelse(bh);
3677 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3678 		offset = do_div(logical_sb_block, blocksize);
3679 		bh = sb_bread_unmovable(sb, logical_sb_block);
3680 		if (!bh) {
3681 			ext4_msg(sb, KERN_ERR,
3682 			       "Can't read superblock on 2nd try");
3683 			goto failed_mount;
3684 		}
3685 		es = (struct ext4_super_block *)(bh->b_data + offset);
3686 		sbi->s_es = es;
3687 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3688 			ext4_msg(sb, KERN_ERR,
3689 			       "Magic mismatch, very weird!");
3690 			goto failed_mount;
3691 		}
3692 	}
3693 
3694 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3695 				EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3696 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3697 						      has_huge_files);
3698 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3699 
3700 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3701 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3702 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3703 	} else {
3704 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3705 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3706 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3707 		    (!is_power_of_2(sbi->s_inode_size)) ||
3708 		    (sbi->s_inode_size > blocksize)) {
3709 			ext4_msg(sb, KERN_ERR,
3710 			       "unsupported inode size: %d",
3711 			       sbi->s_inode_size);
3712 			goto failed_mount;
3713 		}
3714 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3715 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3716 	}
3717 
3718 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3719 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3720 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3721 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3722 		    !is_power_of_2(sbi->s_desc_size)) {
3723 			ext4_msg(sb, KERN_ERR,
3724 			       "unsupported descriptor size %lu",
3725 			       sbi->s_desc_size);
3726 			goto failed_mount;
3727 		}
3728 	} else
3729 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3730 
3731 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3732 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3733 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3734 		goto cantfind_ext4;
3735 
3736 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3737 	if (sbi->s_inodes_per_block == 0)
3738 		goto cantfind_ext4;
3739 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
3740 					sbi->s_inodes_per_block;
3741 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3742 	sbi->s_sbh = bh;
3743 	sbi->s_mount_state = le16_to_cpu(es->s_state);
3744 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3745 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3746 
3747 	for (i = 0; i < 4; i++)
3748 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3749 	sbi->s_def_hash_version = es->s_def_hash_version;
3750 	if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3751 		i = le32_to_cpu(es->s_flags);
3752 		if (i & EXT2_FLAGS_UNSIGNED_HASH)
3753 			sbi->s_hash_unsigned = 3;
3754 		else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3755 #ifdef __CHAR_UNSIGNED__
3756 			if (!(sb->s_flags & MS_RDONLY))
3757 				es->s_flags |=
3758 					cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3759 			sbi->s_hash_unsigned = 3;
3760 #else
3761 			if (!(sb->s_flags & MS_RDONLY))
3762 				es->s_flags |=
3763 					cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3764 #endif
3765 		}
3766 	}
3767 
3768 	/* Handle clustersize */
3769 	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3770 	has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3771 				EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3772 	if (has_bigalloc) {
3773 		if (clustersize < blocksize) {
3774 			ext4_msg(sb, KERN_ERR,
3775 				 "cluster size (%d) smaller than "
3776 				 "block size (%d)", clustersize, blocksize);
3777 			goto failed_mount;
3778 		}
3779 		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3780 			le32_to_cpu(es->s_log_block_size);
3781 		sbi->s_clusters_per_group =
3782 			le32_to_cpu(es->s_clusters_per_group);
3783 		if (sbi->s_clusters_per_group > blocksize * 8) {
3784 			ext4_msg(sb, KERN_ERR,
3785 				 "#clusters per group too big: %lu",
3786 				 sbi->s_clusters_per_group);
3787 			goto failed_mount;
3788 		}
3789 		if (sbi->s_blocks_per_group !=
3790 		    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3791 			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3792 				 "clusters per group (%lu) inconsistent",
3793 				 sbi->s_blocks_per_group,
3794 				 sbi->s_clusters_per_group);
3795 			goto failed_mount;
3796 		}
3797 	} else {
3798 		if (clustersize != blocksize) {
3799 			ext4_warning(sb, "fragment/cluster size (%d) != "
3800 				     "block size (%d)", clustersize,
3801 				     blocksize);
3802 			clustersize = blocksize;
3803 		}
3804 		if (sbi->s_blocks_per_group > blocksize * 8) {
3805 			ext4_msg(sb, KERN_ERR,
3806 				 "#blocks per group too big: %lu",
3807 				 sbi->s_blocks_per_group);
3808 			goto failed_mount;
3809 		}
3810 		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3811 		sbi->s_cluster_bits = 0;
3812 	}
3813 	sbi->s_cluster_ratio = clustersize / blocksize;
3814 
3815 	if (sbi->s_inodes_per_group > blocksize * 8) {
3816 		ext4_msg(sb, KERN_ERR,
3817 		       "#inodes per group too big: %lu",
3818 		       sbi->s_inodes_per_group);
3819 		goto failed_mount;
3820 	}
3821 
3822 	/* Do we have standard group size of clustersize * 8 blocks ? */
3823 	if (sbi->s_blocks_per_group == clustersize << 3)
3824 		set_opt2(sb, STD_GROUP_SIZE);
3825 
3826 	/*
3827 	 * Test whether we have more sectors than will fit in sector_t,
3828 	 * and whether the max offset is addressable by the page cache.
3829 	 */
3830 	err = generic_check_addressable(sb->s_blocksize_bits,
3831 					ext4_blocks_count(es));
3832 	if (err) {
3833 		ext4_msg(sb, KERN_ERR, "filesystem"
3834 			 " too large to mount safely on this system");
3835 		if (sizeof(sector_t) < 8)
3836 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3837 		goto failed_mount;
3838 	}
3839 
3840 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3841 		goto cantfind_ext4;
3842 
3843 	/* check blocks count against device size */
3844 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3845 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3846 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3847 		       "exceeds size of device (%llu blocks)",
3848 		       ext4_blocks_count(es), blocks_count);
3849 		goto failed_mount;
3850 	}
3851 
3852 	/*
3853 	 * It makes no sense for the first data block to be beyond the end
3854 	 * of the filesystem.
3855 	 */
3856 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3857 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3858 			 "block %u is beyond end of filesystem (%llu)",
3859 			 le32_to_cpu(es->s_first_data_block),
3860 			 ext4_blocks_count(es));
3861 		goto failed_mount;
3862 	}
3863 	blocks_count = (ext4_blocks_count(es) -
3864 			le32_to_cpu(es->s_first_data_block) +
3865 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
3866 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3867 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3868 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3869 		       "(block count %llu, first data block %u, "
3870 		       "blocks per group %lu)", sbi->s_groups_count,
3871 		       ext4_blocks_count(es),
3872 		       le32_to_cpu(es->s_first_data_block),
3873 		       EXT4_BLOCKS_PER_GROUP(sb));
3874 		goto failed_mount;
3875 	}
3876 	sbi->s_groups_count = blocks_count;
3877 	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3878 			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3879 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3880 		   EXT4_DESC_PER_BLOCK(sb);
3881 	sbi->s_group_desc = ext4_kvmalloc(db_count *
3882 					  sizeof(struct buffer_head *),
3883 					  GFP_KERNEL);
3884 	if (sbi->s_group_desc == NULL) {
3885 		ext4_msg(sb, KERN_ERR, "not enough memory");
3886 		ret = -ENOMEM;
3887 		goto failed_mount;
3888 	}
3889 
3890 	if (ext4_proc_root)
3891 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3892 
3893 	if (sbi->s_proc)
3894 		proc_create_data("options", S_IRUGO, sbi->s_proc,
3895 				 &ext4_seq_options_fops, sb);
3896 
3897 	bgl_lock_init(sbi->s_blockgroup_lock);
3898 
3899 	for (i = 0; i < db_count; i++) {
3900 		block = descriptor_loc(sb, logical_sb_block, i);
3901 		sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3902 		if (!sbi->s_group_desc[i]) {
3903 			ext4_msg(sb, KERN_ERR,
3904 			       "can't read group descriptor %d", i);
3905 			db_count = i;
3906 			goto failed_mount2;
3907 		}
3908 	}
3909 	if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3910 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3911 		goto failed_mount2;
3912 	}
3913 
3914 	sbi->s_gdb_count = db_count;
3915 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3916 	spin_lock_init(&sbi->s_next_gen_lock);
3917 
3918 	init_timer(&sbi->s_err_report);
3919 	sbi->s_err_report.function = print_daily_error_info;
3920 	sbi->s_err_report.data = (unsigned long) sb;
3921 
3922 	/* Register extent status tree shrinker */
3923 	if (ext4_es_register_shrinker(sbi))
3924 		goto failed_mount3;
3925 
3926 	sbi->s_stripe = ext4_get_stripe_size(sbi);
3927 	sbi->s_extent_max_zeroout_kb = 32;
3928 
3929 	/*
3930 	 * set up enough so that it can read an inode
3931 	 */
3932 	sb->s_op = &ext4_sops;
3933 	sb->s_export_op = &ext4_export_ops;
3934 	sb->s_xattr = ext4_xattr_handlers;
3935 #ifdef CONFIG_QUOTA
3936 	sb->dq_op = &ext4_quota_operations;
3937 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3938 		sb->s_qcop = &ext4_qctl_sysfile_operations;
3939 	else
3940 		sb->s_qcop = &ext4_qctl_operations;
3941 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3942 #endif
3943 	memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3944 
3945 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3946 	mutex_init(&sbi->s_orphan_lock);
3947 
3948 	sb->s_root = NULL;
3949 
3950 	needs_recovery = (es->s_last_orphan != 0 ||
3951 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
3952 				    EXT4_FEATURE_INCOMPAT_RECOVER));
3953 
3954 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3955 	    !(sb->s_flags & MS_RDONLY))
3956 		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3957 			goto failed_mount3a;
3958 
3959 	/*
3960 	 * The first inode we look at is the journal inode.  Don't try
3961 	 * root first: it may be modified in the journal!
3962 	 */
3963 	if (!test_opt(sb, NOLOAD) &&
3964 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3965 		if (ext4_load_journal(sb, es, journal_devnum))
3966 			goto failed_mount3a;
3967 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3968 	      EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3969 		ext4_msg(sb, KERN_ERR, "required journal recovery "
3970 		       "suppressed and not mounted read-only");
3971 		goto failed_mount_wq;
3972 	} else {
3973 		clear_opt(sb, DATA_FLAGS);
3974 		sbi->s_journal = NULL;
3975 		needs_recovery = 0;
3976 		goto no_journal;
3977 	}
3978 
3979 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3980 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3981 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
3982 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3983 		goto failed_mount_wq;
3984 	}
3985 
3986 	if (!set_journal_csum_feature_set(sb)) {
3987 		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3988 			 "feature set");
3989 		goto failed_mount_wq;
3990 	}
3991 
3992 	/* We have now updated the journal if required, so we can
3993 	 * validate the data journaling mode. */
3994 	switch (test_opt(sb, DATA_FLAGS)) {
3995 	case 0:
3996 		/* No mode set, assume a default based on the journal
3997 		 * capabilities: ORDERED_DATA if the journal can
3998 		 * cope, else JOURNAL_DATA
3999 		 */
4000 		if (jbd2_journal_check_available_features
4001 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4002 			set_opt(sb, ORDERED_DATA);
4003 		else
4004 			set_opt(sb, JOURNAL_DATA);
4005 		break;
4006 
4007 	case EXT4_MOUNT_ORDERED_DATA:
4008 	case EXT4_MOUNT_WRITEBACK_DATA:
4009 		if (!jbd2_journal_check_available_features
4010 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4011 			ext4_msg(sb, KERN_ERR, "Journal does not support "
4012 			       "requested data journaling mode");
4013 			goto failed_mount_wq;
4014 		}
4015 	default:
4016 		break;
4017 	}
4018 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4019 
4020 	sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4021 
4022 no_journal:
4023 	if (ext4_mballoc_ready) {
4024 		sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4025 		if (!sbi->s_mb_cache) {
4026 			ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4027 			goto failed_mount_wq;
4028 		}
4029 	}
4030 
4031 	/*
4032 	 * Get the # of file system overhead blocks from the
4033 	 * superblock if present.
4034 	 */
4035 	if (es->s_overhead_clusters)
4036 		sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4037 	else {
4038 		err = ext4_calculate_overhead(sb);
4039 		if (err)
4040 			goto failed_mount_wq;
4041 	}
4042 
4043 	/*
4044 	 * The maximum number of concurrent works can be high and
4045 	 * concurrency isn't really necessary.  Limit it to 1.
4046 	 */
4047 	EXT4_SB(sb)->rsv_conversion_wq =
4048 		alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4049 	if (!EXT4_SB(sb)->rsv_conversion_wq) {
4050 		printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4051 		ret = -ENOMEM;
4052 		goto failed_mount4;
4053 	}
4054 
4055 	/*
4056 	 * The jbd2_journal_load will have done any necessary log recovery,
4057 	 * so we can safely mount the rest of the filesystem now.
4058 	 */
4059 
4060 	root = ext4_iget(sb, EXT4_ROOT_INO);
4061 	if (IS_ERR(root)) {
4062 		ext4_msg(sb, KERN_ERR, "get root inode failed");
4063 		ret = PTR_ERR(root);
4064 		root = NULL;
4065 		goto failed_mount4;
4066 	}
4067 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4068 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4069 		iput(root);
4070 		goto failed_mount4;
4071 	}
4072 	sb->s_root = d_make_root(root);
4073 	if (!sb->s_root) {
4074 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
4075 		ret = -ENOMEM;
4076 		goto failed_mount4;
4077 	}
4078 
4079 	if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4080 		sb->s_flags |= MS_RDONLY;
4081 
4082 	/* determine the minimum size of new large inodes, if present */
4083 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4084 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4085 						     EXT4_GOOD_OLD_INODE_SIZE;
4086 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4087 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4088 			if (sbi->s_want_extra_isize <
4089 			    le16_to_cpu(es->s_want_extra_isize))
4090 				sbi->s_want_extra_isize =
4091 					le16_to_cpu(es->s_want_extra_isize);
4092 			if (sbi->s_want_extra_isize <
4093 			    le16_to_cpu(es->s_min_extra_isize))
4094 				sbi->s_want_extra_isize =
4095 					le16_to_cpu(es->s_min_extra_isize);
4096 		}
4097 	}
4098 	/* Check if enough inode space is available */
4099 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4100 							sbi->s_inode_size) {
4101 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4102 						       EXT4_GOOD_OLD_INODE_SIZE;
4103 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
4104 			 "available");
4105 	}
4106 
4107 	err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4108 	if (err) {
4109 		ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4110 			 "reserved pool", ext4_calculate_resv_clusters(sb));
4111 		goto failed_mount4a;
4112 	}
4113 
4114 	err = ext4_setup_system_zone(sb);
4115 	if (err) {
4116 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
4117 			 "zone (%d)", err);
4118 		goto failed_mount4a;
4119 	}
4120 
4121 	ext4_ext_init(sb);
4122 	err = ext4_mb_init(sb);
4123 	if (err) {
4124 		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4125 			 err);
4126 		goto failed_mount5;
4127 	}
4128 
4129 	block = ext4_count_free_clusters(sb);
4130 	ext4_free_blocks_count_set(sbi->s_es,
4131 				   EXT4_C2B(sbi, block));
4132 	err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4133 				  GFP_KERNEL);
4134 	if (!err) {
4135 		unsigned long freei = ext4_count_free_inodes(sb);
4136 		sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4137 		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4138 					  GFP_KERNEL);
4139 	}
4140 	if (!err)
4141 		err = percpu_counter_init(&sbi->s_dirs_counter,
4142 					  ext4_count_dirs(sb), GFP_KERNEL);
4143 	if (!err)
4144 		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4145 					  GFP_KERNEL);
4146 	if (err) {
4147 		ext4_msg(sb, KERN_ERR, "insufficient memory");
4148 		goto failed_mount6;
4149 	}
4150 
4151 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4152 		if (!ext4_fill_flex_info(sb)) {
4153 			ext4_msg(sb, KERN_ERR,
4154 			       "unable to initialize "
4155 			       "flex_bg meta info!");
4156 			goto failed_mount6;
4157 		}
4158 
4159 	err = ext4_register_li_request(sb, first_not_zeroed);
4160 	if (err)
4161 		goto failed_mount6;
4162 
4163 	sbi->s_kobj.kset = ext4_kset;
4164 	init_completion(&sbi->s_kobj_unregister);
4165 	err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4166 				   "%s", sb->s_id);
4167 	if (err)
4168 		goto failed_mount7;
4169 
4170 #ifdef CONFIG_QUOTA
4171 	/* Enable quota usage during mount. */
4172 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4173 	    !(sb->s_flags & MS_RDONLY)) {
4174 		err = ext4_enable_quotas(sb);
4175 		if (err)
4176 			goto failed_mount8;
4177 	}
4178 #endif  /* CONFIG_QUOTA */
4179 
4180 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4181 	ext4_orphan_cleanup(sb, es);
4182 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4183 	if (needs_recovery) {
4184 		ext4_msg(sb, KERN_INFO, "recovery complete");
4185 		ext4_mark_recovery_complete(sb, es);
4186 	}
4187 	if (EXT4_SB(sb)->s_journal) {
4188 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4189 			descr = " journalled data mode";
4190 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4191 			descr = " ordered data mode";
4192 		else
4193 			descr = " writeback data mode";
4194 	} else
4195 		descr = "out journal";
4196 
4197 	if (test_opt(sb, DISCARD)) {
4198 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
4199 		if (!blk_queue_discard(q))
4200 			ext4_msg(sb, KERN_WARNING,
4201 				 "mounting with \"discard\" option, but "
4202 				 "the device does not support discard");
4203 	}
4204 
4205 	ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4206 		 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4207 		 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4208 
4209 	if (es->s_error_count)
4210 		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4211 
4212 	/* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4213 	ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4214 	ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4215 	ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4216 
4217 	kfree(orig_data);
4218 	return 0;
4219 
4220 cantfind_ext4:
4221 	if (!silent)
4222 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4223 	goto failed_mount;
4224 
4225 #ifdef CONFIG_QUOTA
4226 failed_mount8:
4227 	kobject_del(&sbi->s_kobj);
4228 #endif
4229 failed_mount7:
4230 	ext4_unregister_li_request(sb);
4231 failed_mount6:
4232 	ext4_mb_release(sb);
4233 	if (sbi->s_flex_groups)
4234 		kvfree(sbi->s_flex_groups);
4235 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
4236 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
4237 	percpu_counter_destroy(&sbi->s_dirs_counter);
4238 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4239 failed_mount5:
4240 	ext4_ext_release(sb);
4241 	ext4_release_system_zone(sb);
4242 failed_mount4a:
4243 	dput(sb->s_root);
4244 	sb->s_root = NULL;
4245 failed_mount4:
4246 	ext4_msg(sb, KERN_ERR, "mount failed");
4247 	if (EXT4_SB(sb)->rsv_conversion_wq)
4248 		destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4249 failed_mount_wq:
4250 	if (sbi->s_journal) {
4251 		jbd2_journal_destroy(sbi->s_journal);
4252 		sbi->s_journal = NULL;
4253 	}
4254 failed_mount3a:
4255 	ext4_es_unregister_shrinker(sbi);
4256 failed_mount3:
4257 	del_timer_sync(&sbi->s_err_report);
4258 	if (sbi->s_mmp_tsk)
4259 		kthread_stop(sbi->s_mmp_tsk);
4260 failed_mount2:
4261 	for (i = 0; i < db_count; i++)
4262 		brelse(sbi->s_group_desc[i]);
4263 	kvfree(sbi->s_group_desc);
4264 failed_mount:
4265 	if (sbi->s_chksum_driver)
4266 		crypto_free_shash(sbi->s_chksum_driver);
4267 	if (sbi->s_proc) {
4268 		remove_proc_entry("options", sbi->s_proc);
4269 		remove_proc_entry(sb->s_id, ext4_proc_root);
4270 	}
4271 #ifdef CONFIG_QUOTA
4272 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4273 		kfree(sbi->s_qf_names[i]);
4274 #endif
4275 	ext4_blkdev_remove(sbi);
4276 	brelse(bh);
4277 out_fail:
4278 	sb->s_fs_info = NULL;
4279 	kfree(sbi->s_blockgroup_lock);
4280 	kfree(sbi);
4281 out_free_orig:
4282 	kfree(orig_data);
4283 	return err ? err : ret;
4284 }
4285 
4286 /*
4287  * Setup any per-fs journal parameters now.  We'll do this both on
4288  * initial mount, once the journal has been initialised but before we've
4289  * done any recovery; and again on any subsequent remount.
4290  */
4291 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4292 {
4293 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4294 
4295 	journal->j_commit_interval = sbi->s_commit_interval;
4296 	journal->j_min_batch_time = sbi->s_min_batch_time;
4297 	journal->j_max_batch_time = sbi->s_max_batch_time;
4298 
4299 	write_lock(&journal->j_state_lock);
4300 	if (test_opt(sb, BARRIER))
4301 		journal->j_flags |= JBD2_BARRIER;
4302 	else
4303 		journal->j_flags &= ~JBD2_BARRIER;
4304 	if (test_opt(sb, DATA_ERR_ABORT))
4305 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4306 	else
4307 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4308 	write_unlock(&journal->j_state_lock);
4309 }
4310 
4311 static journal_t *ext4_get_journal(struct super_block *sb,
4312 				   unsigned int journal_inum)
4313 {
4314 	struct inode *journal_inode;
4315 	journal_t *journal;
4316 
4317 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4318 
4319 	/* First, test for the existence of a valid inode on disk.  Bad
4320 	 * things happen if we iget() an unused inode, as the subsequent
4321 	 * iput() will try to delete it. */
4322 
4323 	journal_inode = ext4_iget(sb, journal_inum);
4324 	if (IS_ERR(journal_inode)) {
4325 		ext4_msg(sb, KERN_ERR, "no journal found");
4326 		return NULL;
4327 	}
4328 	if (!journal_inode->i_nlink) {
4329 		make_bad_inode(journal_inode);
4330 		iput(journal_inode);
4331 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4332 		return NULL;
4333 	}
4334 
4335 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4336 		  journal_inode, journal_inode->i_size);
4337 	if (!S_ISREG(journal_inode->i_mode)) {
4338 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
4339 		iput(journal_inode);
4340 		return NULL;
4341 	}
4342 
4343 	journal = jbd2_journal_init_inode(journal_inode);
4344 	if (!journal) {
4345 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4346 		iput(journal_inode);
4347 		return NULL;
4348 	}
4349 	journal->j_private = sb;
4350 	ext4_init_journal_params(sb, journal);
4351 	return journal;
4352 }
4353 
4354 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4355 				       dev_t j_dev)
4356 {
4357 	struct buffer_head *bh;
4358 	journal_t *journal;
4359 	ext4_fsblk_t start;
4360 	ext4_fsblk_t len;
4361 	int hblock, blocksize;
4362 	ext4_fsblk_t sb_block;
4363 	unsigned long offset;
4364 	struct ext4_super_block *es;
4365 	struct block_device *bdev;
4366 
4367 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4368 
4369 	bdev = ext4_blkdev_get(j_dev, sb);
4370 	if (bdev == NULL)
4371 		return NULL;
4372 
4373 	blocksize = sb->s_blocksize;
4374 	hblock = bdev_logical_block_size(bdev);
4375 	if (blocksize < hblock) {
4376 		ext4_msg(sb, KERN_ERR,
4377 			"blocksize too small for journal device");
4378 		goto out_bdev;
4379 	}
4380 
4381 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4382 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4383 	set_blocksize(bdev, blocksize);
4384 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
4385 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4386 		       "external journal");
4387 		goto out_bdev;
4388 	}
4389 
4390 	es = (struct ext4_super_block *) (bh->b_data + offset);
4391 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4392 	    !(le32_to_cpu(es->s_feature_incompat) &
4393 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4394 		ext4_msg(sb, KERN_ERR, "external journal has "
4395 					"bad superblock");
4396 		brelse(bh);
4397 		goto out_bdev;
4398 	}
4399 
4400 	if ((le32_to_cpu(es->s_feature_ro_compat) &
4401 	     EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4402 	    es->s_checksum != ext4_superblock_csum(sb, es)) {
4403 		ext4_msg(sb, KERN_ERR, "external journal has "
4404 				       "corrupt superblock");
4405 		brelse(bh);
4406 		goto out_bdev;
4407 	}
4408 
4409 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4410 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4411 		brelse(bh);
4412 		goto out_bdev;
4413 	}
4414 
4415 	len = ext4_blocks_count(es);
4416 	start = sb_block + 1;
4417 	brelse(bh);	/* we're done with the superblock */
4418 
4419 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4420 					start, len, blocksize);
4421 	if (!journal) {
4422 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
4423 		goto out_bdev;
4424 	}
4425 	journal->j_private = sb;
4426 	ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4427 	wait_on_buffer(journal->j_sb_buffer);
4428 	if (!buffer_uptodate(journal->j_sb_buffer)) {
4429 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4430 		goto out_journal;
4431 	}
4432 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4433 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
4434 					"user (unsupported) - %d",
4435 			be32_to_cpu(journal->j_superblock->s_nr_users));
4436 		goto out_journal;
4437 	}
4438 	EXT4_SB(sb)->journal_bdev = bdev;
4439 	ext4_init_journal_params(sb, journal);
4440 	return journal;
4441 
4442 out_journal:
4443 	jbd2_journal_destroy(journal);
4444 out_bdev:
4445 	ext4_blkdev_put(bdev);
4446 	return NULL;
4447 }
4448 
4449 static int ext4_load_journal(struct super_block *sb,
4450 			     struct ext4_super_block *es,
4451 			     unsigned long journal_devnum)
4452 {
4453 	journal_t *journal;
4454 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4455 	dev_t journal_dev;
4456 	int err = 0;
4457 	int really_read_only;
4458 
4459 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4460 
4461 	if (journal_devnum &&
4462 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4463 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4464 			"numbers have changed");
4465 		journal_dev = new_decode_dev(journal_devnum);
4466 	} else
4467 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4468 
4469 	really_read_only = bdev_read_only(sb->s_bdev);
4470 
4471 	/*
4472 	 * Are we loading a blank journal or performing recovery after a
4473 	 * crash?  For recovery, we need to check in advance whether we
4474 	 * can get read-write access to the device.
4475 	 */
4476 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4477 		if (sb->s_flags & MS_RDONLY) {
4478 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
4479 					"required on readonly filesystem");
4480 			if (really_read_only) {
4481 				ext4_msg(sb, KERN_ERR, "write access "
4482 					"unavailable, cannot proceed");
4483 				return -EROFS;
4484 			}
4485 			ext4_msg(sb, KERN_INFO, "write access will "
4486 			       "be enabled during recovery");
4487 		}
4488 	}
4489 
4490 	if (journal_inum && journal_dev) {
4491 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4492 		       "and inode journals!");
4493 		return -EINVAL;
4494 	}
4495 
4496 	if (journal_inum) {
4497 		if (!(journal = ext4_get_journal(sb, journal_inum)))
4498 			return -EINVAL;
4499 	} else {
4500 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4501 			return -EINVAL;
4502 	}
4503 
4504 	if (!(journal->j_flags & JBD2_BARRIER))
4505 		ext4_msg(sb, KERN_INFO, "barriers disabled");
4506 
4507 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4508 		err = jbd2_journal_wipe(journal, !really_read_only);
4509 	if (!err) {
4510 		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4511 		if (save)
4512 			memcpy(save, ((char *) es) +
4513 			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4514 		err = jbd2_journal_load(journal);
4515 		if (save)
4516 			memcpy(((char *) es) + EXT4_S_ERR_START,
4517 			       save, EXT4_S_ERR_LEN);
4518 		kfree(save);
4519 	}
4520 
4521 	if (err) {
4522 		ext4_msg(sb, KERN_ERR, "error loading journal");
4523 		jbd2_journal_destroy(journal);
4524 		return err;
4525 	}
4526 
4527 	EXT4_SB(sb)->s_journal = journal;
4528 	ext4_clear_journal_err(sb, es);
4529 
4530 	if (!really_read_only && journal_devnum &&
4531 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4532 		es->s_journal_dev = cpu_to_le32(journal_devnum);
4533 
4534 		/* Make sure we flush the recovery flag to disk. */
4535 		ext4_commit_super(sb, 1);
4536 	}
4537 
4538 	return 0;
4539 }
4540 
4541 static int ext4_commit_super(struct super_block *sb, int sync)
4542 {
4543 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4544 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4545 	int error = 0;
4546 
4547 	if (!sbh || block_device_ejected(sb))
4548 		return error;
4549 	if (buffer_write_io_error(sbh)) {
4550 		/*
4551 		 * Oh, dear.  A previous attempt to write the
4552 		 * superblock failed.  This could happen because the
4553 		 * USB device was yanked out.  Or it could happen to
4554 		 * be a transient write error and maybe the block will
4555 		 * be remapped.  Nothing we can do but to retry the
4556 		 * write and hope for the best.
4557 		 */
4558 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
4559 		       "superblock detected");
4560 		clear_buffer_write_io_error(sbh);
4561 		set_buffer_uptodate(sbh);
4562 	}
4563 	/*
4564 	 * If the file system is mounted read-only, don't update the
4565 	 * superblock write time.  This avoids updating the superblock
4566 	 * write time when we are mounting the root file system
4567 	 * read/only but we need to replay the journal; at that point,
4568 	 * for people who are east of GMT and who make their clock
4569 	 * tick in localtime for Windows bug-for-bug compatibility,
4570 	 * the clock is set in the future, and this will cause e2fsck
4571 	 * to complain and force a full file system check.
4572 	 */
4573 	if (!(sb->s_flags & MS_RDONLY))
4574 		es->s_wtime = cpu_to_le32(get_seconds());
4575 	if (sb->s_bdev->bd_part)
4576 		es->s_kbytes_written =
4577 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4578 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4579 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
4580 	else
4581 		es->s_kbytes_written =
4582 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4583 	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4584 		ext4_free_blocks_count_set(es,
4585 			EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4586 				&EXT4_SB(sb)->s_freeclusters_counter)));
4587 	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4588 		es->s_free_inodes_count =
4589 			cpu_to_le32(percpu_counter_sum_positive(
4590 				&EXT4_SB(sb)->s_freeinodes_counter));
4591 	BUFFER_TRACE(sbh, "marking dirty");
4592 	ext4_superblock_csum_set(sb);
4593 	mark_buffer_dirty(sbh);
4594 	if (sync) {
4595 		error = sync_dirty_buffer(sbh);
4596 		if (error)
4597 			return error;
4598 
4599 		error = buffer_write_io_error(sbh);
4600 		if (error) {
4601 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
4602 			       "superblock");
4603 			clear_buffer_write_io_error(sbh);
4604 			set_buffer_uptodate(sbh);
4605 		}
4606 	}
4607 	return error;
4608 }
4609 
4610 /*
4611  * Have we just finished recovery?  If so, and if we are mounting (or
4612  * remounting) the filesystem readonly, then we will end up with a
4613  * consistent fs on disk.  Record that fact.
4614  */
4615 static void ext4_mark_recovery_complete(struct super_block *sb,
4616 					struct ext4_super_block *es)
4617 {
4618 	journal_t *journal = EXT4_SB(sb)->s_journal;
4619 
4620 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4621 		BUG_ON(journal != NULL);
4622 		return;
4623 	}
4624 	jbd2_journal_lock_updates(journal);
4625 	if (jbd2_journal_flush(journal) < 0)
4626 		goto out;
4627 
4628 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4629 	    sb->s_flags & MS_RDONLY) {
4630 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4631 		ext4_commit_super(sb, 1);
4632 	}
4633 
4634 out:
4635 	jbd2_journal_unlock_updates(journal);
4636 }
4637 
4638 /*
4639  * If we are mounting (or read-write remounting) a filesystem whose journal
4640  * has recorded an error from a previous lifetime, move that error to the
4641  * main filesystem now.
4642  */
4643 static void ext4_clear_journal_err(struct super_block *sb,
4644 				   struct ext4_super_block *es)
4645 {
4646 	journal_t *journal;
4647 	int j_errno;
4648 	const char *errstr;
4649 
4650 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4651 
4652 	journal = EXT4_SB(sb)->s_journal;
4653 
4654 	/*
4655 	 * Now check for any error status which may have been recorded in the
4656 	 * journal by a prior ext4_error() or ext4_abort()
4657 	 */
4658 
4659 	j_errno = jbd2_journal_errno(journal);
4660 	if (j_errno) {
4661 		char nbuf[16];
4662 
4663 		errstr = ext4_decode_error(sb, j_errno, nbuf);
4664 		ext4_warning(sb, "Filesystem error recorded "
4665 			     "from previous mount: %s", errstr);
4666 		ext4_warning(sb, "Marking fs in need of filesystem check.");
4667 
4668 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4669 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4670 		ext4_commit_super(sb, 1);
4671 
4672 		jbd2_journal_clear_err(journal);
4673 		jbd2_journal_update_sb_errno(journal);
4674 	}
4675 }
4676 
4677 /*
4678  * Force the running and committing transactions to commit,
4679  * and wait on the commit.
4680  */
4681 int ext4_force_commit(struct super_block *sb)
4682 {
4683 	journal_t *journal;
4684 
4685 	if (sb->s_flags & MS_RDONLY)
4686 		return 0;
4687 
4688 	journal = EXT4_SB(sb)->s_journal;
4689 	return ext4_journal_force_commit(journal);
4690 }
4691 
4692 static int ext4_sync_fs(struct super_block *sb, int wait)
4693 {
4694 	int ret = 0;
4695 	tid_t target;
4696 	bool needs_barrier = false;
4697 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4698 
4699 	trace_ext4_sync_fs(sb, wait);
4700 	flush_workqueue(sbi->rsv_conversion_wq);
4701 	/*
4702 	 * Writeback quota in non-journalled quota case - journalled quota has
4703 	 * no dirty dquots
4704 	 */
4705 	dquot_writeback_dquots(sb, -1);
4706 	/*
4707 	 * Data writeback is possible w/o journal transaction, so barrier must
4708 	 * being sent at the end of the function. But we can skip it if
4709 	 * transaction_commit will do it for us.
4710 	 */
4711 	if (sbi->s_journal) {
4712 		target = jbd2_get_latest_transaction(sbi->s_journal);
4713 		if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4714 		    !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4715 			needs_barrier = true;
4716 
4717 		if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4718 			if (wait)
4719 				ret = jbd2_log_wait_commit(sbi->s_journal,
4720 							   target);
4721 		}
4722 	} else if (wait && test_opt(sb, BARRIER))
4723 		needs_barrier = true;
4724 	if (needs_barrier) {
4725 		int err;
4726 		err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4727 		if (!ret)
4728 			ret = err;
4729 	}
4730 
4731 	return ret;
4732 }
4733 
4734 /*
4735  * LVM calls this function before a (read-only) snapshot is created.  This
4736  * gives us a chance to flush the journal completely and mark the fs clean.
4737  *
4738  * Note that only this function cannot bring a filesystem to be in a clean
4739  * state independently. It relies on upper layer to stop all data & metadata
4740  * modifications.
4741  */
4742 static int ext4_freeze(struct super_block *sb)
4743 {
4744 	int error = 0;
4745 	journal_t *journal;
4746 
4747 	if (sb->s_flags & MS_RDONLY)
4748 		return 0;
4749 
4750 	journal = EXT4_SB(sb)->s_journal;
4751 
4752 	if (journal) {
4753 		/* Now we set up the journal barrier. */
4754 		jbd2_journal_lock_updates(journal);
4755 
4756 		/*
4757 		 * Don't clear the needs_recovery flag if we failed to
4758 		 * flush the journal.
4759 		 */
4760 		error = jbd2_journal_flush(journal);
4761 		if (error < 0)
4762 			goto out;
4763 	}
4764 
4765 	/* Journal blocked and flushed, clear needs_recovery flag. */
4766 	EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4767 	error = ext4_commit_super(sb, 1);
4768 out:
4769 	if (journal)
4770 		/* we rely on upper layer to stop further updates */
4771 		jbd2_journal_unlock_updates(journal);
4772 	return error;
4773 }
4774 
4775 /*
4776  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4777  * flag here, even though the filesystem is not technically dirty yet.
4778  */
4779 static int ext4_unfreeze(struct super_block *sb)
4780 {
4781 	if (sb->s_flags & MS_RDONLY)
4782 		return 0;
4783 
4784 	/* Reset the needs_recovery flag before the fs is unlocked. */
4785 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4786 	ext4_commit_super(sb, 1);
4787 	return 0;
4788 }
4789 
4790 /*
4791  * Structure to save mount options for ext4_remount's benefit
4792  */
4793 struct ext4_mount_options {
4794 	unsigned long s_mount_opt;
4795 	unsigned long s_mount_opt2;
4796 	kuid_t s_resuid;
4797 	kgid_t s_resgid;
4798 	unsigned long s_commit_interval;
4799 	u32 s_min_batch_time, s_max_batch_time;
4800 #ifdef CONFIG_QUOTA
4801 	int s_jquota_fmt;
4802 	char *s_qf_names[EXT4_MAXQUOTAS];
4803 #endif
4804 };
4805 
4806 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4807 {
4808 	struct ext4_super_block *es;
4809 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4810 	unsigned long old_sb_flags;
4811 	struct ext4_mount_options old_opts;
4812 	int enable_quota = 0;
4813 	ext4_group_t g;
4814 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4815 	int err = 0;
4816 #ifdef CONFIG_QUOTA
4817 	int i, j;
4818 #endif
4819 	char *orig_data = kstrdup(data, GFP_KERNEL);
4820 
4821 	/* Store the original options */
4822 	old_sb_flags = sb->s_flags;
4823 	old_opts.s_mount_opt = sbi->s_mount_opt;
4824 	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4825 	old_opts.s_resuid = sbi->s_resuid;
4826 	old_opts.s_resgid = sbi->s_resgid;
4827 	old_opts.s_commit_interval = sbi->s_commit_interval;
4828 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
4829 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
4830 #ifdef CONFIG_QUOTA
4831 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4832 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4833 		if (sbi->s_qf_names[i]) {
4834 			old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4835 							 GFP_KERNEL);
4836 			if (!old_opts.s_qf_names[i]) {
4837 				for (j = 0; j < i; j++)
4838 					kfree(old_opts.s_qf_names[j]);
4839 				kfree(orig_data);
4840 				return -ENOMEM;
4841 			}
4842 		} else
4843 			old_opts.s_qf_names[i] = NULL;
4844 #endif
4845 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4846 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4847 
4848 	/*
4849 	 * Allow the "check" option to be passed as a remount option.
4850 	 */
4851 	if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4852 		err = -EINVAL;
4853 		goto restore_opts;
4854 	}
4855 
4856 	if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4857 	    test_opt(sb, JOURNAL_CHECKSUM)) {
4858 		ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4859 			 "during remount not supported");
4860 		err = -EINVAL;
4861 		goto restore_opts;
4862 	}
4863 
4864 	if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4865 	    test_opt(sb, JOURNAL_CHECKSUM)) {
4866 		ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4867 			 "during remount not supported");
4868 		err = -EINVAL;
4869 		goto restore_opts;
4870 	}
4871 
4872 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4873 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4874 			ext4_msg(sb, KERN_ERR, "can't mount with "
4875 				 "both data=journal and delalloc");
4876 			err = -EINVAL;
4877 			goto restore_opts;
4878 		}
4879 		if (test_opt(sb, DIOREAD_NOLOCK)) {
4880 			ext4_msg(sb, KERN_ERR, "can't mount with "
4881 				 "both data=journal and dioread_nolock");
4882 			err = -EINVAL;
4883 			goto restore_opts;
4884 		}
4885 	}
4886 
4887 	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4888 		ext4_abort(sb, "Abort forced by user");
4889 
4890 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4891 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4892 
4893 	es = sbi->s_es;
4894 
4895 	if (sbi->s_journal) {
4896 		ext4_init_journal_params(sb, sbi->s_journal);
4897 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4898 	}
4899 
4900 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4901 		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4902 			err = -EROFS;
4903 			goto restore_opts;
4904 		}
4905 
4906 		if (*flags & MS_RDONLY) {
4907 			err = sync_filesystem(sb);
4908 			if (err < 0)
4909 				goto restore_opts;
4910 			err = dquot_suspend(sb, -1);
4911 			if (err < 0)
4912 				goto restore_opts;
4913 
4914 			/*
4915 			 * First of all, the unconditional stuff we have to do
4916 			 * to disable replay of the journal when we next remount
4917 			 */
4918 			sb->s_flags |= MS_RDONLY;
4919 
4920 			/*
4921 			 * OK, test if we are remounting a valid rw partition
4922 			 * readonly, and if so set the rdonly flag and then
4923 			 * mark the partition as valid again.
4924 			 */
4925 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4926 			    (sbi->s_mount_state & EXT4_VALID_FS))
4927 				es->s_state = cpu_to_le16(sbi->s_mount_state);
4928 
4929 			if (sbi->s_journal)
4930 				ext4_mark_recovery_complete(sb, es);
4931 		} else {
4932 			/* Make sure we can mount this feature set readwrite */
4933 			if (!ext4_feature_set_ok(sb, 0)) {
4934 				err = -EROFS;
4935 				goto restore_opts;
4936 			}
4937 			/*
4938 			 * Make sure the group descriptor checksums
4939 			 * are sane.  If they aren't, refuse to remount r/w.
4940 			 */
4941 			for (g = 0; g < sbi->s_groups_count; g++) {
4942 				struct ext4_group_desc *gdp =
4943 					ext4_get_group_desc(sb, g, NULL);
4944 
4945 				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4946 					ext4_msg(sb, KERN_ERR,
4947 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
4948 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4949 					       le16_to_cpu(gdp->bg_checksum));
4950 					err = -EINVAL;
4951 					goto restore_opts;
4952 				}
4953 			}
4954 
4955 			/*
4956 			 * If we have an unprocessed orphan list hanging
4957 			 * around from a previously readonly bdev mount,
4958 			 * require a full umount/remount for now.
4959 			 */
4960 			if (es->s_last_orphan) {
4961 				ext4_msg(sb, KERN_WARNING, "Couldn't "
4962 				       "remount RDWR because of unprocessed "
4963 				       "orphan inode list.  Please "
4964 				       "umount/remount instead");
4965 				err = -EINVAL;
4966 				goto restore_opts;
4967 			}
4968 
4969 			/*
4970 			 * Mounting a RDONLY partition read-write, so reread
4971 			 * and store the current valid flag.  (It may have
4972 			 * been changed by e2fsck since we originally mounted
4973 			 * the partition.)
4974 			 */
4975 			if (sbi->s_journal)
4976 				ext4_clear_journal_err(sb, es);
4977 			sbi->s_mount_state = le16_to_cpu(es->s_state);
4978 			if (!ext4_setup_super(sb, es, 0))
4979 				sb->s_flags &= ~MS_RDONLY;
4980 			if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4981 						     EXT4_FEATURE_INCOMPAT_MMP))
4982 				if (ext4_multi_mount_protect(sb,
4983 						le64_to_cpu(es->s_mmp_block))) {
4984 					err = -EROFS;
4985 					goto restore_opts;
4986 				}
4987 			enable_quota = 1;
4988 		}
4989 	}
4990 
4991 	/*
4992 	 * Reinitialize lazy itable initialization thread based on
4993 	 * current settings
4994 	 */
4995 	if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4996 		ext4_unregister_li_request(sb);
4997 	else {
4998 		ext4_group_t first_not_zeroed;
4999 		first_not_zeroed = ext4_has_uninit_itable(sb);
5000 		ext4_register_li_request(sb, first_not_zeroed);
5001 	}
5002 
5003 	ext4_setup_system_zone(sb);
5004 	if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5005 		ext4_commit_super(sb, 1);
5006 
5007 #ifdef CONFIG_QUOTA
5008 	/* Release old quota file names */
5009 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
5010 		kfree(old_opts.s_qf_names[i]);
5011 	if (enable_quota) {
5012 		if (sb_any_quota_suspended(sb))
5013 			dquot_resume(sb, -1);
5014 		else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5015 					EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5016 			err = ext4_enable_quotas(sb);
5017 			if (err)
5018 				goto restore_opts;
5019 		}
5020 	}
5021 #endif
5022 
5023 	ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5024 	kfree(orig_data);
5025 	return 0;
5026 
5027 restore_opts:
5028 	sb->s_flags = old_sb_flags;
5029 	sbi->s_mount_opt = old_opts.s_mount_opt;
5030 	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5031 	sbi->s_resuid = old_opts.s_resuid;
5032 	sbi->s_resgid = old_opts.s_resgid;
5033 	sbi->s_commit_interval = old_opts.s_commit_interval;
5034 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
5035 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
5036 #ifdef CONFIG_QUOTA
5037 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5038 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5039 		kfree(sbi->s_qf_names[i]);
5040 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5041 	}
5042 #endif
5043 	kfree(orig_data);
5044 	return err;
5045 }
5046 
5047 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5048 {
5049 	struct super_block *sb = dentry->d_sb;
5050 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5051 	struct ext4_super_block *es = sbi->s_es;
5052 	ext4_fsblk_t overhead = 0, resv_blocks;
5053 	u64 fsid;
5054 	s64 bfree;
5055 	resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5056 
5057 	if (!test_opt(sb, MINIX_DF))
5058 		overhead = sbi->s_overhead;
5059 
5060 	buf->f_type = EXT4_SUPER_MAGIC;
5061 	buf->f_bsize = sb->s_blocksize;
5062 	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5063 	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5064 		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5065 	/* prevent underflow in case that few free space is available */
5066 	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5067 	buf->f_bavail = buf->f_bfree -
5068 			(ext4_r_blocks_count(es) + resv_blocks);
5069 	if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5070 		buf->f_bavail = 0;
5071 	buf->f_files = le32_to_cpu(es->s_inodes_count);
5072 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5073 	buf->f_namelen = EXT4_NAME_LEN;
5074 	fsid = le64_to_cpup((void *)es->s_uuid) ^
5075 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5076 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5077 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5078 
5079 	return 0;
5080 }
5081 
5082 /* Helper function for writing quotas on sync - we need to start transaction
5083  * before quota file is locked for write. Otherwise the are possible deadlocks:
5084  * Process 1                         Process 2
5085  * ext4_create()                     quota_sync()
5086  *   jbd2_journal_start()                  write_dquot()
5087  *   dquot_initialize()                         down(dqio_mutex)
5088  *     down(dqio_mutex)                    jbd2_journal_start()
5089  *
5090  */
5091 
5092 #ifdef CONFIG_QUOTA
5093 
5094 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5095 {
5096 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5097 }
5098 
5099 static int ext4_write_dquot(struct dquot *dquot)
5100 {
5101 	int ret, err;
5102 	handle_t *handle;
5103 	struct inode *inode;
5104 
5105 	inode = dquot_to_inode(dquot);
5106 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5107 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5108 	if (IS_ERR(handle))
5109 		return PTR_ERR(handle);
5110 	ret = dquot_commit(dquot);
5111 	err = ext4_journal_stop(handle);
5112 	if (!ret)
5113 		ret = err;
5114 	return ret;
5115 }
5116 
5117 static int ext4_acquire_dquot(struct dquot *dquot)
5118 {
5119 	int ret, err;
5120 	handle_t *handle;
5121 
5122 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5123 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5124 	if (IS_ERR(handle))
5125 		return PTR_ERR(handle);
5126 	ret = dquot_acquire(dquot);
5127 	err = ext4_journal_stop(handle);
5128 	if (!ret)
5129 		ret = err;
5130 	return ret;
5131 }
5132 
5133 static int ext4_release_dquot(struct dquot *dquot)
5134 {
5135 	int ret, err;
5136 	handle_t *handle;
5137 
5138 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5139 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5140 	if (IS_ERR(handle)) {
5141 		/* Release dquot anyway to avoid endless cycle in dqput() */
5142 		dquot_release(dquot);
5143 		return PTR_ERR(handle);
5144 	}
5145 	ret = dquot_release(dquot);
5146 	err = ext4_journal_stop(handle);
5147 	if (!ret)
5148 		ret = err;
5149 	return ret;
5150 }
5151 
5152 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5153 {
5154 	struct super_block *sb = dquot->dq_sb;
5155 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5156 
5157 	/* Are we journaling quotas? */
5158 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5159 	    sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5160 		dquot_mark_dquot_dirty(dquot);
5161 		return ext4_write_dquot(dquot);
5162 	} else {
5163 		return dquot_mark_dquot_dirty(dquot);
5164 	}
5165 }
5166 
5167 static int ext4_write_info(struct super_block *sb, int type)
5168 {
5169 	int ret, err;
5170 	handle_t *handle;
5171 
5172 	/* Data block + inode block */
5173 	handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5174 	if (IS_ERR(handle))
5175 		return PTR_ERR(handle);
5176 	ret = dquot_commit_info(sb, type);
5177 	err = ext4_journal_stop(handle);
5178 	if (!ret)
5179 		ret = err;
5180 	return ret;
5181 }
5182 
5183 /*
5184  * Turn on quotas during mount time - we need to find
5185  * the quota file and such...
5186  */
5187 static int ext4_quota_on_mount(struct super_block *sb, int type)
5188 {
5189 	return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5190 					EXT4_SB(sb)->s_jquota_fmt, type);
5191 }
5192 
5193 /*
5194  * Standard function to be called on quota_on
5195  */
5196 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5197 			 struct path *path)
5198 {
5199 	int err;
5200 
5201 	if (!test_opt(sb, QUOTA))
5202 		return -EINVAL;
5203 
5204 	/* Quotafile not on the same filesystem? */
5205 	if (path->dentry->d_sb != sb)
5206 		return -EXDEV;
5207 	/* Journaling quota? */
5208 	if (EXT4_SB(sb)->s_qf_names[type]) {
5209 		/* Quotafile not in fs root? */
5210 		if (path->dentry->d_parent != sb->s_root)
5211 			ext4_msg(sb, KERN_WARNING,
5212 				"Quota file not on filesystem root. "
5213 				"Journaled quota will not work");
5214 	}
5215 
5216 	/*
5217 	 * When we journal data on quota file, we have to flush journal to see
5218 	 * all updates to the file when we bypass pagecache...
5219 	 */
5220 	if (EXT4_SB(sb)->s_journal &&
5221 	    ext4_should_journal_data(path->dentry->d_inode)) {
5222 		/*
5223 		 * We don't need to lock updates but journal_flush() could
5224 		 * otherwise be livelocked...
5225 		 */
5226 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5227 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5228 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5229 		if (err)
5230 			return err;
5231 	}
5232 
5233 	return dquot_quota_on(sb, type, format_id, path);
5234 }
5235 
5236 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5237 			     unsigned int flags)
5238 {
5239 	int err;
5240 	struct inode *qf_inode;
5241 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5242 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5243 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5244 	};
5245 
5246 	BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5247 
5248 	if (!qf_inums[type])
5249 		return -EPERM;
5250 
5251 	qf_inode = ext4_iget(sb, qf_inums[type]);
5252 	if (IS_ERR(qf_inode)) {
5253 		ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5254 		return PTR_ERR(qf_inode);
5255 	}
5256 
5257 	/* Don't account quota for quota files to avoid recursion */
5258 	qf_inode->i_flags |= S_NOQUOTA;
5259 	err = dquot_enable(qf_inode, type, format_id, flags);
5260 	iput(qf_inode);
5261 
5262 	return err;
5263 }
5264 
5265 /* Enable usage tracking for all quota types. */
5266 static int ext4_enable_quotas(struct super_block *sb)
5267 {
5268 	int type, err = 0;
5269 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5270 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5271 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5272 	};
5273 
5274 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5275 	for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5276 		if (qf_inums[type]) {
5277 			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5278 						DQUOT_USAGE_ENABLED);
5279 			if (err) {
5280 				ext4_warning(sb,
5281 					"Failed to enable quota tracking "
5282 					"(type=%d, err=%d). Please run "
5283 					"e2fsck to fix.", type, err);
5284 				return err;
5285 			}
5286 		}
5287 	}
5288 	return 0;
5289 }
5290 
5291 /*
5292  * quota_on function that is used when QUOTA feature is set.
5293  */
5294 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5295 				 int format_id)
5296 {
5297 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5298 		return -EINVAL;
5299 
5300 	/*
5301 	 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5302 	 */
5303 	return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5304 }
5305 
5306 static int ext4_quota_off(struct super_block *sb, int type)
5307 {
5308 	struct inode *inode = sb_dqopt(sb)->files[type];
5309 	handle_t *handle;
5310 
5311 	/* Force all delayed allocation blocks to be allocated.
5312 	 * Caller already holds s_umount sem */
5313 	if (test_opt(sb, DELALLOC))
5314 		sync_filesystem(sb);
5315 
5316 	if (!inode)
5317 		goto out;
5318 
5319 	/* Update modification times of quota files when userspace can
5320 	 * start looking at them */
5321 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5322 	if (IS_ERR(handle))
5323 		goto out;
5324 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5325 	ext4_mark_inode_dirty(handle, inode);
5326 	ext4_journal_stop(handle);
5327 
5328 out:
5329 	return dquot_quota_off(sb, type);
5330 }
5331 
5332 /*
5333  * quota_off function that is used when QUOTA feature is set.
5334  */
5335 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5336 {
5337 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5338 		return -EINVAL;
5339 
5340 	/* Disable only the limits. */
5341 	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5342 }
5343 
5344 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5345  * acquiring the locks... As quota files are never truncated and quota code
5346  * itself serializes the operations (and no one else should touch the files)
5347  * we don't have to be afraid of races */
5348 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5349 			       size_t len, loff_t off)
5350 {
5351 	struct inode *inode = sb_dqopt(sb)->files[type];
5352 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5353 	int offset = off & (sb->s_blocksize - 1);
5354 	int tocopy;
5355 	size_t toread;
5356 	struct buffer_head *bh;
5357 	loff_t i_size = i_size_read(inode);
5358 
5359 	if (off > i_size)
5360 		return 0;
5361 	if (off+len > i_size)
5362 		len = i_size-off;
5363 	toread = len;
5364 	while (toread > 0) {
5365 		tocopy = sb->s_blocksize - offset < toread ?
5366 				sb->s_blocksize - offset : toread;
5367 		bh = ext4_bread(NULL, inode, blk, 0);
5368 		if (IS_ERR(bh))
5369 			return PTR_ERR(bh);
5370 		if (!bh)	/* A hole? */
5371 			memset(data, 0, tocopy);
5372 		else
5373 			memcpy(data, bh->b_data+offset, tocopy);
5374 		brelse(bh);
5375 		offset = 0;
5376 		toread -= tocopy;
5377 		data += tocopy;
5378 		blk++;
5379 	}
5380 	return len;
5381 }
5382 
5383 /* Write to quotafile (we know the transaction is already started and has
5384  * enough credits) */
5385 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5386 				const char *data, size_t len, loff_t off)
5387 {
5388 	struct inode *inode = sb_dqopt(sb)->files[type];
5389 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5390 	int err, offset = off & (sb->s_blocksize - 1);
5391 	struct buffer_head *bh;
5392 	handle_t *handle = journal_current_handle();
5393 
5394 	if (EXT4_SB(sb)->s_journal && !handle) {
5395 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5396 			" cancelled because transaction is not started",
5397 			(unsigned long long)off, (unsigned long long)len);
5398 		return -EIO;
5399 	}
5400 	/*
5401 	 * Since we account only one data block in transaction credits,
5402 	 * then it is impossible to cross a block boundary.
5403 	 */
5404 	if (sb->s_blocksize - offset < len) {
5405 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5406 			" cancelled because not block aligned",
5407 			(unsigned long long)off, (unsigned long long)len);
5408 		return -EIO;
5409 	}
5410 
5411 	bh = ext4_bread(handle, inode, blk, 1);
5412 	if (IS_ERR(bh))
5413 		return PTR_ERR(bh);
5414 	if (!bh)
5415 		goto out;
5416 	BUFFER_TRACE(bh, "get write access");
5417 	err = ext4_journal_get_write_access(handle, bh);
5418 	if (err) {
5419 		brelse(bh);
5420 		return err;
5421 	}
5422 	lock_buffer(bh);
5423 	memcpy(bh->b_data+offset, data, len);
5424 	flush_dcache_page(bh->b_page);
5425 	unlock_buffer(bh);
5426 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
5427 	brelse(bh);
5428 out:
5429 	if (inode->i_size < off + len) {
5430 		i_size_write(inode, off + len);
5431 		EXT4_I(inode)->i_disksize = inode->i_size;
5432 		ext4_mark_inode_dirty(handle, inode);
5433 	}
5434 	return len;
5435 }
5436 
5437 #endif
5438 
5439 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5440 		       const char *dev_name, void *data)
5441 {
5442 	return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5443 }
5444 
5445 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5446 static inline void register_as_ext2(void)
5447 {
5448 	int err = register_filesystem(&ext2_fs_type);
5449 	if (err)
5450 		printk(KERN_WARNING
5451 		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5452 }
5453 
5454 static inline void unregister_as_ext2(void)
5455 {
5456 	unregister_filesystem(&ext2_fs_type);
5457 }
5458 
5459 static inline int ext2_feature_set_ok(struct super_block *sb)
5460 {
5461 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5462 		return 0;
5463 	if (sb->s_flags & MS_RDONLY)
5464 		return 1;
5465 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5466 		return 0;
5467 	return 1;
5468 }
5469 #else
5470 static inline void register_as_ext2(void) { }
5471 static inline void unregister_as_ext2(void) { }
5472 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5473 #endif
5474 
5475 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5476 static inline void register_as_ext3(void)
5477 {
5478 	int err = register_filesystem(&ext3_fs_type);
5479 	if (err)
5480 		printk(KERN_WARNING
5481 		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5482 }
5483 
5484 static inline void unregister_as_ext3(void)
5485 {
5486 	unregister_filesystem(&ext3_fs_type);
5487 }
5488 
5489 static inline int ext3_feature_set_ok(struct super_block *sb)
5490 {
5491 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5492 		return 0;
5493 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5494 		return 0;
5495 	if (sb->s_flags & MS_RDONLY)
5496 		return 1;
5497 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5498 		return 0;
5499 	return 1;
5500 }
5501 #else
5502 static inline void register_as_ext3(void) { }
5503 static inline void unregister_as_ext3(void) { }
5504 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5505 #endif
5506 
5507 static struct file_system_type ext4_fs_type = {
5508 	.owner		= THIS_MODULE,
5509 	.name		= "ext4",
5510 	.mount		= ext4_mount,
5511 	.kill_sb	= kill_block_super,
5512 	.fs_flags	= FS_REQUIRES_DEV,
5513 };
5514 MODULE_ALIAS_FS("ext4");
5515 
5516 static int __init ext4_init_feat_adverts(void)
5517 {
5518 	struct ext4_features *ef;
5519 	int ret = -ENOMEM;
5520 
5521 	ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5522 	if (!ef)
5523 		goto out;
5524 
5525 	ef->f_kobj.kset = ext4_kset;
5526 	init_completion(&ef->f_kobj_unregister);
5527 	ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5528 				   "features");
5529 	if (ret) {
5530 		kfree(ef);
5531 		goto out;
5532 	}
5533 
5534 	ext4_feat = ef;
5535 	ret = 0;
5536 out:
5537 	return ret;
5538 }
5539 
5540 static void ext4_exit_feat_adverts(void)
5541 {
5542 	kobject_put(&ext4_feat->f_kobj);
5543 	wait_for_completion(&ext4_feat->f_kobj_unregister);
5544 	kfree(ext4_feat);
5545 }
5546 
5547 /* Shared across all ext4 file systems */
5548 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5549 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5550 
5551 static int __init ext4_init_fs(void)
5552 {
5553 	int i, err;
5554 
5555 	ext4_li_info = NULL;
5556 	mutex_init(&ext4_li_mtx);
5557 
5558 	/* Build-time check for flags consistency */
5559 	ext4_check_flag_values();
5560 
5561 	for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5562 		mutex_init(&ext4__aio_mutex[i]);
5563 		init_waitqueue_head(&ext4__ioend_wq[i]);
5564 	}
5565 
5566 	err = ext4_init_es();
5567 	if (err)
5568 		return err;
5569 
5570 	err = ext4_init_pageio();
5571 	if (err)
5572 		goto out7;
5573 
5574 	err = ext4_init_system_zone();
5575 	if (err)
5576 		goto out6;
5577 	ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5578 	if (!ext4_kset) {
5579 		err = -ENOMEM;
5580 		goto out5;
5581 	}
5582 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5583 
5584 	err = ext4_init_feat_adverts();
5585 	if (err)
5586 		goto out4;
5587 
5588 	err = ext4_init_mballoc();
5589 	if (err)
5590 		goto out2;
5591 	else
5592 		ext4_mballoc_ready = 1;
5593 	err = init_inodecache();
5594 	if (err)
5595 		goto out1;
5596 	register_as_ext3();
5597 	register_as_ext2();
5598 	err = register_filesystem(&ext4_fs_type);
5599 	if (err)
5600 		goto out;
5601 
5602 	return 0;
5603 out:
5604 	unregister_as_ext2();
5605 	unregister_as_ext3();
5606 	destroy_inodecache();
5607 out1:
5608 	ext4_mballoc_ready = 0;
5609 	ext4_exit_mballoc();
5610 out2:
5611 	ext4_exit_feat_adverts();
5612 out4:
5613 	if (ext4_proc_root)
5614 		remove_proc_entry("fs/ext4", NULL);
5615 	kset_unregister(ext4_kset);
5616 out5:
5617 	ext4_exit_system_zone();
5618 out6:
5619 	ext4_exit_pageio();
5620 out7:
5621 	ext4_exit_es();
5622 
5623 	return err;
5624 }
5625 
5626 static void __exit ext4_exit_fs(void)
5627 {
5628 	ext4_destroy_lazyinit_thread();
5629 	unregister_as_ext2();
5630 	unregister_as_ext3();
5631 	unregister_filesystem(&ext4_fs_type);
5632 	destroy_inodecache();
5633 	ext4_exit_mballoc();
5634 	ext4_exit_feat_adverts();
5635 	remove_proc_entry("fs/ext4", NULL);
5636 	kset_unregister(ext4_kset);
5637 	ext4_exit_system_zone();
5638 	ext4_exit_pageio();
5639 	ext4_exit_es();
5640 }
5641 
5642 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5643 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5644 MODULE_LICENSE("GPL");
5645 module_init(ext4_init_fs)
5646 module_exit(ext4_exit_fs)
5647