xref: /openbmc/linux/fs/ext4/super.c (revision 0f4630f3)
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/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.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/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <asm/uaccess.h>
42 
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
45 
46 #include "ext4.h"
47 #include "ext4_extents.h"	/* Needed for trace points definition */
48 #include "ext4_jbd2.h"
49 #include "xattr.h"
50 #include "acl.h"
51 #include "mballoc.h"
52 
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
55 
56 static struct ext4_lazy_init *ext4_li_info;
57 static struct mutex ext4_li_mtx;
58 static int ext4_mballoc_ready;
59 static struct ratelimit_state ext4_mount_msg_ratelimit;
60 
61 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
62 			     unsigned long journal_devnum);
63 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
64 static int ext4_commit_super(struct super_block *sb, int sync);
65 static void ext4_mark_recovery_complete(struct super_block *sb,
66 					struct ext4_super_block *es);
67 static void ext4_clear_journal_err(struct super_block *sb,
68 				   struct ext4_super_block *es);
69 static int ext4_sync_fs(struct super_block *sb, int wait);
70 static int ext4_remount(struct super_block *sb, int *flags, char *data);
71 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
72 static int ext4_unfreeze(struct super_block *sb);
73 static int ext4_freeze(struct super_block *sb);
74 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
75 		       const char *dev_name, void *data);
76 static inline int ext2_feature_set_ok(struct super_block *sb);
77 static inline int ext3_feature_set_ok(struct super_block *sb);
78 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
79 static void ext4_destroy_lazyinit_thread(void);
80 static void ext4_unregister_li_request(struct super_block *sb);
81 static void ext4_clear_request_list(void);
82 
83 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
84 static struct file_system_type ext2_fs_type = {
85 	.owner		= THIS_MODULE,
86 	.name		= "ext2",
87 	.mount		= ext4_mount,
88 	.kill_sb	= kill_block_super,
89 	.fs_flags	= FS_REQUIRES_DEV,
90 };
91 MODULE_ALIAS_FS("ext2");
92 MODULE_ALIAS("ext2");
93 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
94 #else
95 #define IS_EXT2_SB(sb) (0)
96 #endif
97 
98 
99 static struct file_system_type ext3_fs_type = {
100 	.owner		= THIS_MODULE,
101 	.name		= "ext3",
102 	.mount		= ext4_mount,
103 	.kill_sb	= kill_block_super,
104 	.fs_flags	= FS_REQUIRES_DEV,
105 };
106 MODULE_ALIAS_FS("ext3");
107 MODULE_ALIAS("ext3");
108 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
109 
110 static int ext4_verify_csum_type(struct super_block *sb,
111 				 struct ext4_super_block *es)
112 {
113 	if (!ext4_has_feature_metadata_csum(sb))
114 		return 1;
115 
116 	return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
117 }
118 
119 static __le32 ext4_superblock_csum(struct super_block *sb,
120 				   struct ext4_super_block *es)
121 {
122 	struct ext4_sb_info *sbi = EXT4_SB(sb);
123 	int offset = offsetof(struct ext4_super_block, s_checksum);
124 	__u32 csum;
125 
126 	csum = ext4_chksum(sbi, ~0, (char *)es, offset);
127 
128 	return cpu_to_le32(csum);
129 }
130 
131 static int ext4_superblock_csum_verify(struct super_block *sb,
132 				       struct ext4_super_block *es)
133 {
134 	if (!ext4_has_metadata_csum(sb))
135 		return 1;
136 
137 	return es->s_checksum == ext4_superblock_csum(sb, es);
138 }
139 
140 void ext4_superblock_csum_set(struct super_block *sb)
141 {
142 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
143 
144 	if (!ext4_has_metadata_csum(sb))
145 		return;
146 
147 	es->s_checksum = ext4_superblock_csum(sb, es);
148 }
149 
150 void *ext4_kvmalloc(size_t size, gfp_t flags)
151 {
152 	void *ret;
153 
154 	ret = kmalloc(size, flags | __GFP_NOWARN);
155 	if (!ret)
156 		ret = __vmalloc(size, flags, PAGE_KERNEL);
157 	return ret;
158 }
159 
160 void *ext4_kvzalloc(size_t size, gfp_t flags)
161 {
162 	void *ret;
163 
164 	ret = kzalloc(size, flags | __GFP_NOWARN);
165 	if (!ret)
166 		ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
167 	return ret;
168 }
169 
170 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
171 			       struct ext4_group_desc *bg)
172 {
173 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
174 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
175 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
176 }
177 
178 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
179 			       struct ext4_group_desc *bg)
180 {
181 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
182 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
183 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
184 }
185 
186 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
187 			      struct ext4_group_desc *bg)
188 {
189 	return le32_to_cpu(bg->bg_inode_table_lo) |
190 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
192 }
193 
194 __u32 ext4_free_group_clusters(struct super_block *sb,
195 			       struct ext4_group_desc *bg)
196 {
197 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
198 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
200 }
201 
202 __u32 ext4_free_inodes_count(struct super_block *sb,
203 			      struct ext4_group_desc *bg)
204 {
205 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
206 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
208 }
209 
210 __u32 ext4_used_dirs_count(struct super_block *sb,
211 			      struct ext4_group_desc *bg)
212 {
213 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
214 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
216 }
217 
218 __u32 ext4_itable_unused_count(struct super_block *sb,
219 			      struct ext4_group_desc *bg)
220 {
221 	return le16_to_cpu(bg->bg_itable_unused_lo) |
222 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
224 }
225 
226 void ext4_block_bitmap_set(struct super_block *sb,
227 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
228 {
229 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
230 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
231 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
232 }
233 
234 void ext4_inode_bitmap_set(struct super_block *sb,
235 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
236 {
237 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
238 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
240 }
241 
242 void ext4_inode_table_set(struct super_block *sb,
243 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
244 {
245 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
246 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
248 }
249 
250 void ext4_free_group_clusters_set(struct super_block *sb,
251 				  struct ext4_group_desc *bg, __u32 count)
252 {
253 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
254 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
256 }
257 
258 void ext4_free_inodes_set(struct super_block *sb,
259 			  struct ext4_group_desc *bg, __u32 count)
260 {
261 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
262 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
264 }
265 
266 void ext4_used_dirs_set(struct super_block *sb,
267 			  struct ext4_group_desc *bg, __u32 count)
268 {
269 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
270 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
272 }
273 
274 void ext4_itable_unused_set(struct super_block *sb,
275 			  struct ext4_group_desc *bg, __u32 count)
276 {
277 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
278 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
280 }
281 
282 
283 static void __save_error_info(struct super_block *sb, const char *func,
284 			    unsigned int line)
285 {
286 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
287 
288 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
289 	if (bdev_read_only(sb->s_bdev))
290 		return;
291 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
292 	es->s_last_error_time = cpu_to_le32(get_seconds());
293 	strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
294 	es->s_last_error_line = cpu_to_le32(line);
295 	if (!es->s_first_error_time) {
296 		es->s_first_error_time = es->s_last_error_time;
297 		strncpy(es->s_first_error_func, func,
298 			sizeof(es->s_first_error_func));
299 		es->s_first_error_line = cpu_to_le32(line);
300 		es->s_first_error_ino = es->s_last_error_ino;
301 		es->s_first_error_block = es->s_last_error_block;
302 	}
303 	/*
304 	 * Start the daily error reporting function if it hasn't been
305 	 * started already
306 	 */
307 	if (!es->s_error_count)
308 		mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
309 	le32_add_cpu(&es->s_error_count, 1);
310 }
311 
312 static void save_error_info(struct super_block *sb, const char *func,
313 			    unsigned int line)
314 {
315 	__save_error_info(sb, func, line);
316 	ext4_commit_super(sb, 1);
317 }
318 
319 /*
320  * The del_gendisk() function uninitializes the disk-specific data
321  * structures, including the bdi structure, without telling anyone
322  * else.  Once this happens, any attempt to call mark_buffer_dirty()
323  * (for example, by ext4_commit_super), will cause a kernel OOPS.
324  * This is a kludge to prevent these oops until we can put in a proper
325  * hook in del_gendisk() to inform the VFS and file system layers.
326  */
327 static int block_device_ejected(struct super_block *sb)
328 {
329 	struct inode *bd_inode = sb->s_bdev->bd_inode;
330 	struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
331 
332 	return bdi->dev == NULL;
333 }
334 
335 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
336 {
337 	struct super_block		*sb = journal->j_private;
338 	struct ext4_sb_info		*sbi = EXT4_SB(sb);
339 	int				error = is_journal_aborted(journal);
340 	struct ext4_journal_cb_entry	*jce;
341 
342 	BUG_ON(txn->t_state == T_FINISHED);
343 	spin_lock(&sbi->s_md_lock);
344 	while (!list_empty(&txn->t_private_list)) {
345 		jce = list_entry(txn->t_private_list.next,
346 				 struct ext4_journal_cb_entry, jce_list);
347 		list_del_init(&jce->jce_list);
348 		spin_unlock(&sbi->s_md_lock);
349 		jce->jce_func(sb, jce, error);
350 		spin_lock(&sbi->s_md_lock);
351 	}
352 	spin_unlock(&sbi->s_md_lock);
353 }
354 
355 /* Deal with the reporting of failure conditions on a filesystem such as
356  * inconsistencies detected or read IO failures.
357  *
358  * On ext2, we can store the error state of the filesystem in the
359  * superblock.  That is not possible on ext4, because we may have other
360  * write ordering constraints on the superblock which prevent us from
361  * writing it out straight away; and given that the journal is about to
362  * be aborted, we can't rely on the current, or future, transactions to
363  * write out the superblock safely.
364  *
365  * We'll just use the jbd2_journal_abort() error code to record an error in
366  * the journal instead.  On recovery, the journal will complain about
367  * that error until we've noted it down and cleared it.
368  */
369 
370 static void ext4_handle_error(struct super_block *sb)
371 {
372 	if (sb->s_flags & MS_RDONLY)
373 		return;
374 
375 	if (!test_opt(sb, ERRORS_CONT)) {
376 		journal_t *journal = EXT4_SB(sb)->s_journal;
377 
378 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
379 		if (journal)
380 			jbd2_journal_abort(journal, -EIO);
381 	}
382 	if (test_opt(sb, ERRORS_RO)) {
383 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
384 		/*
385 		 * Make sure updated value of ->s_mount_flags will be visible
386 		 * before ->s_flags update
387 		 */
388 		smp_wmb();
389 		sb->s_flags |= MS_RDONLY;
390 	}
391 	if (test_opt(sb, ERRORS_PANIC)) {
392 		if (EXT4_SB(sb)->s_journal &&
393 		  !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
394 			return;
395 		panic("EXT4-fs (device %s): panic forced after error\n",
396 			sb->s_id);
397 	}
398 }
399 
400 #define ext4_error_ratelimit(sb)					\
401 		___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),	\
402 			     "EXT4-fs error")
403 
404 void __ext4_error(struct super_block *sb, const char *function,
405 		  unsigned int line, const char *fmt, ...)
406 {
407 	struct va_format vaf;
408 	va_list args;
409 
410 	if (ext4_error_ratelimit(sb)) {
411 		va_start(args, fmt);
412 		vaf.fmt = fmt;
413 		vaf.va = &args;
414 		printk(KERN_CRIT
415 		       "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
416 		       sb->s_id, function, line, current->comm, &vaf);
417 		va_end(args);
418 	}
419 	save_error_info(sb, function, line);
420 	ext4_handle_error(sb);
421 }
422 
423 void __ext4_error_inode(struct inode *inode, const char *function,
424 			unsigned int line, ext4_fsblk_t block,
425 			const char *fmt, ...)
426 {
427 	va_list args;
428 	struct va_format vaf;
429 	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
430 
431 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
432 	es->s_last_error_block = cpu_to_le64(block);
433 	if (ext4_error_ratelimit(inode->i_sb)) {
434 		va_start(args, fmt);
435 		vaf.fmt = fmt;
436 		vaf.va = &args;
437 		if (block)
438 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
439 			       "inode #%lu: block %llu: comm %s: %pV\n",
440 			       inode->i_sb->s_id, function, line, inode->i_ino,
441 			       block, current->comm, &vaf);
442 		else
443 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
444 			       "inode #%lu: comm %s: %pV\n",
445 			       inode->i_sb->s_id, function, line, inode->i_ino,
446 			       current->comm, &vaf);
447 		va_end(args);
448 	}
449 	save_error_info(inode->i_sb, function, line);
450 	ext4_handle_error(inode->i_sb);
451 }
452 
453 void __ext4_error_file(struct file *file, const char *function,
454 		       unsigned int line, ext4_fsblk_t block,
455 		       const char *fmt, ...)
456 {
457 	va_list args;
458 	struct va_format vaf;
459 	struct ext4_super_block *es;
460 	struct inode *inode = file_inode(file);
461 	char pathname[80], *path;
462 
463 	es = EXT4_SB(inode->i_sb)->s_es;
464 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
465 	if (ext4_error_ratelimit(inode->i_sb)) {
466 		path = file_path(file, pathname, sizeof(pathname));
467 		if (IS_ERR(path))
468 			path = "(unknown)";
469 		va_start(args, fmt);
470 		vaf.fmt = fmt;
471 		vaf.va = &args;
472 		if (block)
473 			printk(KERN_CRIT
474 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
475 			       "block %llu: comm %s: path %s: %pV\n",
476 			       inode->i_sb->s_id, function, line, inode->i_ino,
477 			       block, current->comm, path, &vaf);
478 		else
479 			printk(KERN_CRIT
480 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
481 			       "comm %s: path %s: %pV\n",
482 			       inode->i_sb->s_id, function, line, inode->i_ino,
483 			       current->comm, path, &vaf);
484 		va_end(args);
485 	}
486 	save_error_info(inode->i_sb, function, line);
487 	ext4_handle_error(inode->i_sb);
488 }
489 
490 const char *ext4_decode_error(struct super_block *sb, int errno,
491 			      char nbuf[16])
492 {
493 	char *errstr = NULL;
494 
495 	switch (errno) {
496 	case -EFSCORRUPTED:
497 		errstr = "Corrupt filesystem";
498 		break;
499 	case -EFSBADCRC:
500 		errstr = "Filesystem failed CRC";
501 		break;
502 	case -EIO:
503 		errstr = "IO failure";
504 		break;
505 	case -ENOMEM:
506 		errstr = "Out of memory";
507 		break;
508 	case -EROFS:
509 		if (!sb || (EXT4_SB(sb)->s_journal &&
510 			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
511 			errstr = "Journal has aborted";
512 		else
513 			errstr = "Readonly filesystem";
514 		break;
515 	default:
516 		/* If the caller passed in an extra buffer for unknown
517 		 * errors, textualise them now.  Else we just return
518 		 * NULL. */
519 		if (nbuf) {
520 			/* Check for truncated error codes... */
521 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
522 				errstr = nbuf;
523 		}
524 		break;
525 	}
526 
527 	return errstr;
528 }
529 
530 /* __ext4_std_error decodes expected errors from journaling functions
531  * automatically and invokes the appropriate error response.  */
532 
533 void __ext4_std_error(struct super_block *sb, const char *function,
534 		      unsigned int line, int errno)
535 {
536 	char nbuf[16];
537 	const char *errstr;
538 
539 	/* Special case: if the error is EROFS, and we're not already
540 	 * inside a transaction, then there's really no point in logging
541 	 * an error. */
542 	if (errno == -EROFS && journal_current_handle() == NULL &&
543 	    (sb->s_flags & MS_RDONLY))
544 		return;
545 
546 	if (ext4_error_ratelimit(sb)) {
547 		errstr = ext4_decode_error(sb, errno, nbuf);
548 		printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
549 		       sb->s_id, function, line, errstr);
550 	}
551 
552 	save_error_info(sb, function, line);
553 	ext4_handle_error(sb);
554 }
555 
556 /*
557  * ext4_abort is a much stronger failure handler than ext4_error.  The
558  * abort function may be used to deal with unrecoverable failures such
559  * as journal IO errors or ENOMEM at a critical moment in log management.
560  *
561  * We unconditionally force the filesystem into an ABORT|READONLY state,
562  * unless the error response on the fs has been set to panic in which
563  * case we take the easy way out and panic immediately.
564  */
565 
566 void __ext4_abort(struct super_block *sb, const char *function,
567 		unsigned int line, const char *fmt, ...)
568 {
569 	va_list args;
570 
571 	save_error_info(sb, function, line);
572 	va_start(args, fmt);
573 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
574 	       function, line);
575 	vprintk(fmt, args);
576 	printk("\n");
577 	va_end(args);
578 
579 	if ((sb->s_flags & MS_RDONLY) == 0) {
580 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
581 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
582 		/*
583 		 * Make sure updated value of ->s_mount_flags will be visible
584 		 * before ->s_flags update
585 		 */
586 		smp_wmb();
587 		sb->s_flags |= MS_RDONLY;
588 		if (EXT4_SB(sb)->s_journal)
589 			jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
590 		save_error_info(sb, function, line);
591 	}
592 	if (test_opt(sb, ERRORS_PANIC)) {
593 		if (EXT4_SB(sb)->s_journal &&
594 		  !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
595 			return;
596 		panic("EXT4-fs panic from previous error\n");
597 	}
598 }
599 
600 void __ext4_msg(struct super_block *sb,
601 		const char *prefix, const char *fmt, ...)
602 {
603 	struct va_format vaf;
604 	va_list args;
605 
606 	if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
607 		return;
608 
609 	va_start(args, fmt);
610 	vaf.fmt = fmt;
611 	vaf.va = &args;
612 	printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
613 	va_end(args);
614 }
615 
616 #define ext4_warning_ratelimit(sb)					\
617 		___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),	\
618 			     "EXT4-fs warning")
619 
620 void __ext4_warning(struct super_block *sb, const char *function,
621 		    unsigned int line, const char *fmt, ...)
622 {
623 	struct va_format vaf;
624 	va_list args;
625 
626 	if (!ext4_warning_ratelimit(sb))
627 		return;
628 
629 	va_start(args, fmt);
630 	vaf.fmt = fmt;
631 	vaf.va = &args;
632 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
633 	       sb->s_id, function, line, &vaf);
634 	va_end(args);
635 }
636 
637 void __ext4_warning_inode(const struct inode *inode, const char *function,
638 			  unsigned int line, const char *fmt, ...)
639 {
640 	struct va_format vaf;
641 	va_list args;
642 
643 	if (!ext4_warning_ratelimit(inode->i_sb))
644 		return;
645 
646 	va_start(args, fmt);
647 	vaf.fmt = fmt;
648 	vaf.va = &args;
649 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
650 	       "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
651 	       function, line, inode->i_ino, current->comm, &vaf);
652 	va_end(args);
653 }
654 
655 void __ext4_grp_locked_error(const char *function, unsigned int line,
656 			     struct super_block *sb, ext4_group_t grp,
657 			     unsigned long ino, ext4_fsblk_t block,
658 			     const char *fmt, ...)
659 __releases(bitlock)
660 __acquires(bitlock)
661 {
662 	struct va_format vaf;
663 	va_list args;
664 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
665 
666 	es->s_last_error_ino = cpu_to_le32(ino);
667 	es->s_last_error_block = cpu_to_le64(block);
668 	__save_error_info(sb, function, line);
669 
670 	if (ext4_error_ratelimit(sb)) {
671 		va_start(args, fmt);
672 		vaf.fmt = fmt;
673 		vaf.va = &args;
674 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
675 		       sb->s_id, function, line, grp);
676 		if (ino)
677 			printk(KERN_CONT "inode %lu: ", ino);
678 		if (block)
679 			printk(KERN_CONT "block %llu:",
680 			       (unsigned long long) block);
681 		printk(KERN_CONT "%pV\n", &vaf);
682 		va_end(args);
683 	}
684 
685 	if (test_opt(sb, ERRORS_CONT)) {
686 		ext4_commit_super(sb, 0);
687 		return;
688 	}
689 
690 	ext4_unlock_group(sb, grp);
691 	ext4_handle_error(sb);
692 	/*
693 	 * We only get here in the ERRORS_RO case; relocking the group
694 	 * may be dangerous, but nothing bad will happen since the
695 	 * filesystem will have already been marked read/only and the
696 	 * journal has been aborted.  We return 1 as a hint to callers
697 	 * who might what to use the return value from
698 	 * ext4_grp_locked_error() to distinguish between the
699 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
700 	 * aggressively from the ext4 function in question, with a
701 	 * more appropriate error code.
702 	 */
703 	ext4_lock_group(sb, grp);
704 	return;
705 }
706 
707 void ext4_update_dynamic_rev(struct super_block *sb)
708 {
709 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
710 
711 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
712 		return;
713 
714 	ext4_warning(sb,
715 		     "updating to rev %d because of new feature flag, "
716 		     "running e2fsck is recommended",
717 		     EXT4_DYNAMIC_REV);
718 
719 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
720 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
721 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
722 	/* leave es->s_feature_*compat flags alone */
723 	/* es->s_uuid will be set by e2fsck if empty */
724 
725 	/*
726 	 * The rest of the superblock fields should be zero, and if not it
727 	 * means they are likely already in use, so leave them alone.  We
728 	 * can leave it up to e2fsck to clean up any inconsistencies there.
729 	 */
730 }
731 
732 /*
733  * Open the external journal device
734  */
735 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
736 {
737 	struct block_device *bdev;
738 	char b[BDEVNAME_SIZE];
739 
740 	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
741 	if (IS_ERR(bdev))
742 		goto fail;
743 	return bdev;
744 
745 fail:
746 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
747 			__bdevname(dev, b), PTR_ERR(bdev));
748 	return NULL;
749 }
750 
751 /*
752  * Release the journal device
753  */
754 static void ext4_blkdev_put(struct block_device *bdev)
755 {
756 	blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
757 }
758 
759 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
760 {
761 	struct block_device *bdev;
762 	bdev = sbi->journal_bdev;
763 	if (bdev) {
764 		ext4_blkdev_put(bdev);
765 		sbi->journal_bdev = NULL;
766 	}
767 }
768 
769 static inline struct inode *orphan_list_entry(struct list_head *l)
770 {
771 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
772 }
773 
774 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
775 {
776 	struct list_head *l;
777 
778 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
779 		 le32_to_cpu(sbi->s_es->s_last_orphan));
780 
781 	printk(KERN_ERR "sb_info orphan list:\n");
782 	list_for_each(l, &sbi->s_orphan) {
783 		struct inode *inode = orphan_list_entry(l);
784 		printk(KERN_ERR "  "
785 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
786 		       inode->i_sb->s_id, inode->i_ino, inode,
787 		       inode->i_mode, inode->i_nlink,
788 		       NEXT_ORPHAN(inode));
789 	}
790 }
791 
792 static void ext4_put_super(struct super_block *sb)
793 {
794 	struct ext4_sb_info *sbi = EXT4_SB(sb);
795 	struct ext4_super_block *es = sbi->s_es;
796 	int i, err;
797 
798 	ext4_unregister_li_request(sb);
799 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
800 
801 	flush_workqueue(sbi->rsv_conversion_wq);
802 	destroy_workqueue(sbi->rsv_conversion_wq);
803 
804 	if (sbi->s_journal) {
805 		err = jbd2_journal_destroy(sbi->s_journal);
806 		sbi->s_journal = NULL;
807 		if (err < 0)
808 			ext4_abort(sb, "Couldn't clean up the journal");
809 	}
810 
811 	ext4_unregister_sysfs(sb);
812 	ext4_es_unregister_shrinker(sbi);
813 	del_timer_sync(&sbi->s_err_report);
814 	ext4_release_system_zone(sb);
815 	ext4_mb_release(sb);
816 	ext4_ext_release(sb);
817 	ext4_xattr_put_super(sb);
818 
819 	if (!(sb->s_flags & MS_RDONLY)) {
820 		ext4_clear_feature_journal_needs_recovery(sb);
821 		es->s_state = cpu_to_le16(sbi->s_mount_state);
822 	}
823 	if (!(sb->s_flags & MS_RDONLY))
824 		ext4_commit_super(sb, 1);
825 
826 	for (i = 0; i < sbi->s_gdb_count; i++)
827 		brelse(sbi->s_group_desc[i]);
828 	kvfree(sbi->s_group_desc);
829 	kvfree(sbi->s_flex_groups);
830 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
831 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
832 	percpu_counter_destroy(&sbi->s_dirs_counter);
833 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
834 	brelse(sbi->s_sbh);
835 #ifdef CONFIG_QUOTA
836 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
837 		kfree(sbi->s_qf_names[i]);
838 #endif
839 
840 	/* Debugging code just in case the in-memory inode orphan list
841 	 * isn't empty.  The on-disk one can be non-empty if we've
842 	 * detected an error and taken the fs readonly, but the
843 	 * in-memory list had better be clean by this point. */
844 	if (!list_empty(&sbi->s_orphan))
845 		dump_orphan_list(sb, sbi);
846 	J_ASSERT(list_empty(&sbi->s_orphan));
847 
848 	sync_blockdev(sb->s_bdev);
849 	invalidate_bdev(sb->s_bdev);
850 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
851 		/*
852 		 * Invalidate the journal device's buffers.  We don't want them
853 		 * floating about in memory - the physical journal device may
854 		 * hotswapped, and it breaks the `ro-after' testing code.
855 		 */
856 		sync_blockdev(sbi->journal_bdev);
857 		invalidate_bdev(sbi->journal_bdev);
858 		ext4_blkdev_remove(sbi);
859 	}
860 	if (sbi->s_mb_cache) {
861 		ext4_xattr_destroy_cache(sbi->s_mb_cache);
862 		sbi->s_mb_cache = NULL;
863 	}
864 	if (sbi->s_mmp_tsk)
865 		kthread_stop(sbi->s_mmp_tsk);
866 	sb->s_fs_info = NULL;
867 	/*
868 	 * Now that we are completely done shutting down the
869 	 * superblock, we need to actually destroy the kobject.
870 	 */
871 	kobject_put(&sbi->s_kobj);
872 	wait_for_completion(&sbi->s_kobj_unregister);
873 	if (sbi->s_chksum_driver)
874 		crypto_free_shash(sbi->s_chksum_driver);
875 	kfree(sbi->s_blockgroup_lock);
876 	kfree(sbi);
877 }
878 
879 static struct kmem_cache *ext4_inode_cachep;
880 
881 /*
882  * Called inside transaction, so use GFP_NOFS
883  */
884 static struct inode *ext4_alloc_inode(struct super_block *sb)
885 {
886 	struct ext4_inode_info *ei;
887 
888 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
889 	if (!ei)
890 		return NULL;
891 
892 	ei->vfs_inode.i_version = 1;
893 	spin_lock_init(&ei->i_raw_lock);
894 	INIT_LIST_HEAD(&ei->i_prealloc_list);
895 	spin_lock_init(&ei->i_prealloc_lock);
896 	ext4_es_init_tree(&ei->i_es_tree);
897 	rwlock_init(&ei->i_es_lock);
898 	INIT_LIST_HEAD(&ei->i_es_list);
899 	ei->i_es_all_nr = 0;
900 	ei->i_es_shk_nr = 0;
901 	ei->i_es_shrink_lblk = 0;
902 	ei->i_reserved_data_blocks = 0;
903 	ei->i_reserved_meta_blocks = 0;
904 	ei->i_allocated_meta_blocks = 0;
905 	ei->i_da_metadata_calc_len = 0;
906 	ei->i_da_metadata_calc_last_lblock = 0;
907 	spin_lock_init(&(ei->i_block_reservation_lock));
908 #ifdef CONFIG_QUOTA
909 	ei->i_reserved_quota = 0;
910 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
911 #endif
912 	ei->jinode = NULL;
913 	INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
914 	spin_lock_init(&ei->i_completed_io_lock);
915 	ei->i_sync_tid = 0;
916 	ei->i_datasync_tid = 0;
917 	atomic_set(&ei->i_ioend_count, 0);
918 	atomic_set(&ei->i_unwritten, 0);
919 	INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
920 #ifdef CONFIG_EXT4_FS_ENCRYPTION
921 	ei->i_crypt_info = NULL;
922 #endif
923 	return &ei->vfs_inode;
924 }
925 
926 static int ext4_drop_inode(struct inode *inode)
927 {
928 	int drop = generic_drop_inode(inode);
929 
930 	trace_ext4_drop_inode(inode, drop);
931 	return drop;
932 }
933 
934 static void ext4_i_callback(struct rcu_head *head)
935 {
936 	struct inode *inode = container_of(head, struct inode, i_rcu);
937 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
938 }
939 
940 static void ext4_destroy_inode(struct inode *inode)
941 {
942 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
943 		ext4_msg(inode->i_sb, KERN_ERR,
944 			 "Inode %lu (%p): orphan list check failed!",
945 			 inode->i_ino, EXT4_I(inode));
946 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
947 				EXT4_I(inode), sizeof(struct ext4_inode_info),
948 				true);
949 		dump_stack();
950 	}
951 	call_rcu(&inode->i_rcu, ext4_i_callback);
952 }
953 
954 static void init_once(void *foo)
955 {
956 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
957 
958 	INIT_LIST_HEAD(&ei->i_orphan);
959 	init_rwsem(&ei->xattr_sem);
960 	init_rwsem(&ei->i_data_sem);
961 	inode_init_once(&ei->vfs_inode);
962 }
963 
964 static int __init init_inodecache(void)
965 {
966 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
967 					     sizeof(struct ext4_inode_info),
968 					     0, (SLAB_RECLAIM_ACCOUNT|
969 						SLAB_MEM_SPREAD),
970 					     init_once);
971 	if (ext4_inode_cachep == NULL)
972 		return -ENOMEM;
973 	return 0;
974 }
975 
976 static void destroy_inodecache(void)
977 {
978 	/*
979 	 * Make sure all delayed rcu free inodes are flushed before we
980 	 * destroy cache.
981 	 */
982 	rcu_barrier();
983 	kmem_cache_destroy(ext4_inode_cachep);
984 }
985 
986 void ext4_clear_inode(struct inode *inode)
987 {
988 	invalidate_inode_buffers(inode);
989 	clear_inode(inode);
990 	dquot_drop(inode);
991 	ext4_discard_preallocations(inode);
992 	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
993 	if (EXT4_I(inode)->jinode) {
994 		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
995 					       EXT4_I(inode)->jinode);
996 		jbd2_free_inode(EXT4_I(inode)->jinode);
997 		EXT4_I(inode)->jinode = NULL;
998 	}
999 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1000 	if (EXT4_I(inode)->i_crypt_info)
1001 		ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
1002 #endif
1003 }
1004 
1005 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1006 					u64 ino, u32 generation)
1007 {
1008 	struct inode *inode;
1009 
1010 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1011 		return ERR_PTR(-ESTALE);
1012 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1013 		return ERR_PTR(-ESTALE);
1014 
1015 	/* iget isn't really right if the inode is currently unallocated!!
1016 	 *
1017 	 * ext4_read_inode will return a bad_inode if the inode had been
1018 	 * deleted, so we should be safe.
1019 	 *
1020 	 * Currently we don't know the generation for parent directory, so
1021 	 * a generation of 0 means "accept any"
1022 	 */
1023 	inode = ext4_iget_normal(sb, ino);
1024 	if (IS_ERR(inode))
1025 		return ERR_CAST(inode);
1026 	if (generation && inode->i_generation != generation) {
1027 		iput(inode);
1028 		return ERR_PTR(-ESTALE);
1029 	}
1030 
1031 	return inode;
1032 }
1033 
1034 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1035 					int fh_len, int fh_type)
1036 {
1037 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1038 				    ext4_nfs_get_inode);
1039 }
1040 
1041 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1042 					int fh_len, int fh_type)
1043 {
1044 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1045 				    ext4_nfs_get_inode);
1046 }
1047 
1048 /*
1049  * Try to release metadata pages (indirect blocks, directories) which are
1050  * mapped via the block device.  Since these pages could have journal heads
1051  * which would prevent try_to_free_buffers() from freeing them, we must use
1052  * jbd2 layer's try_to_free_buffers() function to release them.
1053  */
1054 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1055 				 gfp_t wait)
1056 {
1057 	journal_t *journal = EXT4_SB(sb)->s_journal;
1058 
1059 	WARN_ON(PageChecked(page));
1060 	if (!page_has_buffers(page))
1061 		return 0;
1062 	if (journal)
1063 		return jbd2_journal_try_to_free_buffers(journal, page,
1064 						wait & ~__GFP_DIRECT_RECLAIM);
1065 	return try_to_free_buffers(page);
1066 }
1067 
1068 #ifdef CONFIG_QUOTA
1069 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1070 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1071 
1072 static int ext4_write_dquot(struct dquot *dquot);
1073 static int ext4_acquire_dquot(struct dquot *dquot);
1074 static int ext4_release_dquot(struct dquot *dquot);
1075 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1076 static int ext4_write_info(struct super_block *sb, int type);
1077 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1078 			 struct path *path);
1079 static int ext4_quota_off(struct super_block *sb, int type);
1080 static int ext4_quota_on_mount(struct super_block *sb, int type);
1081 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1082 			       size_t len, loff_t off);
1083 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1084 				const char *data, size_t len, loff_t off);
1085 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1086 			     unsigned int flags);
1087 static int ext4_enable_quotas(struct super_block *sb);
1088 
1089 static struct dquot **ext4_get_dquots(struct inode *inode)
1090 {
1091 	return EXT4_I(inode)->i_dquot;
1092 }
1093 
1094 static const struct dquot_operations ext4_quota_operations = {
1095 	.get_reserved_space = ext4_get_reserved_space,
1096 	.write_dquot	= ext4_write_dquot,
1097 	.acquire_dquot	= ext4_acquire_dquot,
1098 	.release_dquot	= ext4_release_dquot,
1099 	.mark_dirty	= ext4_mark_dquot_dirty,
1100 	.write_info	= ext4_write_info,
1101 	.alloc_dquot	= dquot_alloc,
1102 	.destroy_dquot	= dquot_destroy,
1103 };
1104 
1105 static const struct quotactl_ops ext4_qctl_operations = {
1106 	.quota_on	= ext4_quota_on,
1107 	.quota_off	= ext4_quota_off,
1108 	.quota_sync	= dquot_quota_sync,
1109 	.get_state	= dquot_get_state,
1110 	.set_info	= dquot_set_dqinfo,
1111 	.get_dqblk	= dquot_get_dqblk,
1112 	.set_dqblk	= dquot_set_dqblk
1113 };
1114 #endif
1115 
1116 static const struct super_operations ext4_sops = {
1117 	.alloc_inode	= ext4_alloc_inode,
1118 	.destroy_inode	= ext4_destroy_inode,
1119 	.write_inode	= ext4_write_inode,
1120 	.dirty_inode	= ext4_dirty_inode,
1121 	.drop_inode	= ext4_drop_inode,
1122 	.evict_inode	= ext4_evict_inode,
1123 	.put_super	= ext4_put_super,
1124 	.sync_fs	= ext4_sync_fs,
1125 	.freeze_fs	= ext4_freeze,
1126 	.unfreeze_fs	= ext4_unfreeze,
1127 	.statfs		= ext4_statfs,
1128 	.remount_fs	= ext4_remount,
1129 	.show_options	= ext4_show_options,
1130 #ifdef CONFIG_QUOTA
1131 	.quota_read	= ext4_quota_read,
1132 	.quota_write	= ext4_quota_write,
1133 	.get_dquots	= ext4_get_dquots,
1134 #endif
1135 	.bdev_try_to_free_page = bdev_try_to_free_page,
1136 };
1137 
1138 static const struct export_operations ext4_export_ops = {
1139 	.fh_to_dentry = ext4_fh_to_dentry,
1140 	.fh_to_parent = ext4_fh_to_parent,
1141 	.get_parent = ext4_get_parent,
1142 };
1143 
1144 enum {
1145 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1146 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1147 	Opt_nouid32, Opt_debug, Opt_removed,
1148 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1149 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1150 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1151 	Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1152 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1153 	Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1154 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1155 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1156 	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1157 	Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1158 	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1159 	Opt_lazytime, Opt_nolazytime,
1160 	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1161 	Opt_inode_readahead_blks, Opt_journal_ioprio,
1162 	Opt_dioread_nolock, Opt_dioread_lock,
1163 	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1164 	Opt_max_dir_size_kb, Opt_nojournal_checksum,
1165 };
1166 
1167 static const match_table_t tokens = {
1168 	{Opt_bsd_df, "bsddf"},
1169 	{Opt_minix_df, "minixdf"},
1170 	{Opt_grpid, "grpid"},
1171 	{Opt_grpid, "bsdgroups"},
1172 	{Opt_nogrpid, "nogrpid"},
1173 	{Opt_nogrpid, "sysvgroups"},
1174 	{Opt_resgid, "resgid=%u"},
1175 	{Opt_resuid, "resuid=%u"},
1176 	{Opt_sb, "sb=%u"},
1177 	{Opt_err_cont, "errors=continue"},
1178 	{Opt_err_panic, "errors=panic"},
1179 	{Opt_err_ro, "errors=remount-ro"},
1180 	{Opt_nouid32, "nouid32"},
1181 	{Opt_debug, "debug"},
1182 	{Opt_removed, "oldalloc"},
1183 	{Opt_removed, "orlov"},
1184 	{Opt_user_xattr, "user_xattr"},
1185 	{Opt_nouser_xattr, "nouser_xattr"},
1186 	{Opt_acl, "acl"},
1187 	{Opt_noacl, "noacl"},
1188 	{Opt_noload, "norecovery"},
1189 	{Opt_noload, "noload"},
1190 	{Opt_removed, "nobh"},
1191 	{Opt_removed, "bh"},
1192 	{Opt_commit, "commit=%u"},
1193 	{Opt_min_batch_time, "min_batch_time=%u"},
1194 	{Opt_max_batch_time, "max_batch_time=%u"},
1195 	{Opt_journal_dev, "journal_dev=%u"},
1196 	{Opt_journal_path, "journal_path=%s"},
1197 	{Opt_journal_checksum, "journal_checksum"},
1198 	{Opt_nojournal_checksum, "nojournal_checksum"},
1199 	{Opt_journal_async_commit, "journal_async_commit"},
1200 	{Opt_abort, "abort"},
1201 	{Opt_data_journal, "data=journal"},
1202 	{Opt_data_ordered, "data=ordered"},
1203 	{Opt_data_writeback, "data=writeback"},
1204 	{Opt_data_err_abort, "data_err=abort"},
1205 	{Opt_data_err_ignore, "data_err=ignore"},
1206 	{Opt_offusrjquota, "usrjquota="},
1207 	{Opt_usrjquota, "usrjquota=%s"},
1208 	{Opt_offgrpjquota, "grpjquota="},
1209 	{Opt_grpjquota, "grpjquota=%s"},
1210 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1211 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1212 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1213 	{Opt_grpquota, "grpquota"},
1214 	{Opt_noquota, "noquota"},
1215 	{Opt_quota, "quota"},
1216 	{Opt_usrquota, "usrquota"},
1217 	{Opt_barrier, "barrier=%u"},
1218 	{Opt_barrier, "barrier"},
1219 	{Opt_nobarrier, "nobarrier"},
1220 	{Opt_i_version, "i_version"},
1221 	{Opt_dax, "dax"},
1222 	{Opt_stripe, "stripe=%u"},
1223 	{Opt_delalloc, "delalloc"},
1224 	{Opt_lazytime, "lazytime"},
1225 	{Opt_nolazytime, "nolazytime"},
1226 	{Opt_nodelalloc, "nodelalloc"},
1227 	{Opt_removed, "mblk_io_submit"},
1228 	{Opt_removed, "nomblk_io_submit"},
1229 	{Opt_block_validity, "block_validity"},
1230 	{Opt_noblock_validity, "noblock_validity"},
1231 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1232 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1233 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1234 	{Opt_auto_da_alloc, "auto_da_alloc"},
1235 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1236 	{Opt_dioread_nolock, "dioread_nolock"},
1237 	{Opt_dioread_lock, "dioread_lock"},
1238 	{Opt_discard, "discard"},
1239 	{Opt_nodiscard, "nodiscard"},
1240 	{Opt_init_itable, "init_itable=%u"},
1241 	{Opt_init_itable, "init_itable"},
1242 	{Opt_noinit_itable, "noinit_itable"},
1243 	{Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1244 	{Opt_test_dummy_encryption, "test_dummy_encryption"},
1245 	{Opt_removed, "check=none"},	/* mount option from ext2/3 */
1246 	{Opt_removed, "nocheck"},	/* mount option from ext2/3 */
1247 	{Opt_removed, "reservation"},	/* mount option from ext2/3 */
1248 	{Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1249 	{Opt_removed, "journal=%u"},	/* mount option from ext2/3 */
1250 	{Opt_err, NULL},
1251 };
1252 
1253 static ext4_fsblk_t get_sb_block(void **data)
1254 {
1255 	ext4_fsblk_t	sb_block;
1256 	char		*options = (char *) *data;
1257 
1258 	if (!options || strncmp(options, "sb=", 3) != 0)
1259 		return 1;	/* Default location */
1260 
1261 	options += 3;
1262 	/* TODO: use simple_strtoll with >32bit ext4 */
1263 	sb_block = simple_strtoul(options, &options, 0);
1264 	if (*options && *options != ',') {
1265 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1266 		       (char *) *data);
1267 		return 1;
1268 	}
1269 	if (*options == ',')
1270 		options++;
1271 	*data = (void *) options;
1272 
1273 	return sb_block;
1274 }
1275 
1276 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1277 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1278 	"Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1279 
1280 #ifdef CONFIG_QUOTA
1281 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1282 {
1283 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1284 	char *qname;
1285 	int ret = -1;
1286 
1287 	if (sb_any_quota_loaded(sb) &&
1288 		!sbi->s_qf_names[qtype]) {
1289 		ext4_msg(sb, KERN_ERR,
1290 			"Cannot change journaled "
1291 			"quota options when quota turned on");
1292 		return -1;
1293 	}
1294 	if (ext4_has_feature_quota(sb)) {
1295 		ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1296 			 "when QUOTA feature is enabled");
1297 		return -1;
1298 	}
1299 	qname = match_strdup(args);
1300 	if (!qname) {
1301 		ext4_msg(sb, KERN_ERR,
1302 			"Not enough memory for storing quotafile name");
1303 		return -1;
1304 	}
1305 	if (sbi->s_qf_names[qtype]) {
1306 		if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1307 			ret = 1;
1308 		else
1309 			ext4_msg(sb, KERN_ERR,
1310 				 "%s quota file already specified",
1311 				 QTYPE2NAME(qtype));
1312 		goto errout;
1313 	}
1314 	if (strchr(qname, '/')) {
1315 		ext4_msg(sb, KERN_ERR,
1316 			"quotafile must be on filesystem root");
1317 		goto errout;
1318 	}
1319 	sbi->s_qf_names[qtype] = qname;
1320 	set_opt(sb, QUOTA);
1321 	return 1;
1322 errout:
1323 	kfree(qname);
1324 	return ret;
1325 }
1326 
1327 static int clear_qf_name(struct super_block *sb, int qtype)
1328 {
1329 
1330 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1331 
1332 	if (sb_any_quota_loaded(sb) &&
1333 		sbi->s_qf_names[qtype]) {
1334 		ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1335 			" when quota turned on");
1336 		return -1;
1337 	}
1338 	kfree(sbi->s_qf_names[qtype]);
1339 	sbi->s_qf_names[qtype] = NULL;
1340 	return 1;
1341 }
1342 #endif
1343 
1344 #define MOPT_SET	0x0001
1345 #define MOPT_CLEAR	0x0002
1346 #define MOPT_NOSUPPORT	0x0004
1347 #define MOPT_EXPLICIT	0x0008
1348 #define MOPT_CLEAR_ERR	0x0010
1349 #define MOPT_GTE0	0x0020
1350 #ifdef CONFIG_QUOTA
1351 #define MOPT_Q		0
1352 #define MOPT_QFMT	0x0040
1353 #else
1354 #define MOPT_Q		MOPT_NOSUPPORT
1355 #define MOPT_QFMT	MOPT_NOSUPPORT
1356 #endif
1357 #define MOPT_DATAJ	0x0080
1358 #define MOPT_NO_EXT2	0x0100
1359 #define MOPT_NO_EXT3	0x0200
1360 #define MOPT_EXT4_ONLY	(MOPT_NO_EXT2 | MOPT_NO_EXT3)
1361 #define MOPT_STRING	0x0400
1362 
1363 static const struct mount_opts {
1364 	int	token;
1365 	int	mount_opt;
1366 	int	flags;
1367 } ext4_mount_opts[] = {
1368 	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1369 	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1370 	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1371 	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1372 	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1373 	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1374 	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1375 	 MOPT_EXT4_ONLY | MOPT_SET},
1376 	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1377 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1378 	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1379 	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1380 	{Opt_delalloc, EXT4_MOUNT_DELALLOC,
1381 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1382 	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1383 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1384 	{Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1385 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1386 	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1387 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1388 	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1389 				    EXT4_MOUNT_JOURNAL_CHECKSUM),
1390 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1391 	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1392 	{Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1393 	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1394 	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1395 	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1396 	 MOPT_NO_EXT2 | MOPT_SET},
1397 	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1398 	 MOPT_NO_EXT2 | MOPT_CLEAR},
1399 	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1400 	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1401 	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1402 	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1403 	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1404 	{Opt_commit, 0, MOPT_GTE0},
1405 	{Opt_max_batch_time, 0, MOPT_GTE0},
1406 	{Opt_min_batch_time, 0, MOPT_GTE0},
1407 	{Opt_inode_readahead_blks, 0, MOPT_GTE0},
1408 	{Opt_init_itable, 0, MOPT_GTE0},
1409 	{Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1410 	{Opt_stripe, 0, MOPT_GTE0},
1411 	{Opt_resuid, 0, MOPT_GTE0},
1412 	{Opt_resgid, 0, MOPT_GTE0},
1413 	{Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1414 	{Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1415 	{Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1416 	{Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1417 	{Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1418 	{Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1419 	 MOPT_NO_EXT2 | MOPT_DATAJ},
1420 	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1421 	{Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1422 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1423 	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1424 	{Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1425 #else
1426 	{Opt_acl, 0, MOPT_NOSUPPORT},
1427 	{Opt_noacl, 0, MOPT_NOSUPPORT},
1428 #endif
1429 	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1430 	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1431 	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1432 	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1433 							MOPT_SET | MOPT_Q},
1434 	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1435 							MOPT_SET | MOPT_Q},
1436 	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1437 		       EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1438 	{Opt_usrjquota, 0, MOPT_Q},
1439 	{Opt_grpjquota, 0, MOPT_Q},
1440 	{Opt_offusrjquota, 0, MOPT_Q},
1441 	{Opt_offgrpjquota, 0, MOPT_Q},
1442 	{Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1443 	{Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1444 	{Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1445 	{Opt_max_dir_size_kb, 0, MOPT_GTE0},
1446 	{Opt_test_dummy_encryption, 0, MOPT_GTE0},
1447 	{Opt_err, 0, 0}
1448 };
1449 
1450 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1451 			    substring_t *args, unsigned long *journal_devnum,
1452 			    unsigned int *journal_ioprio, int is_remount)
1453 {
1454 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1455 	const struct mount_opts *m;
1456 	kuid_t uid;
1457 	kgid_t gid;
1458 	int arg = 0;
1459 
1460 #ifdef CONFIG_QUOTA
1461 	if (token == Opt_usrjquota)
1462 		return set_qf_name(sb, USRQUOTA, &args[0]);
1463 	else if (token == Opt_grpjquota)
1464 		return set_qf_name(sb, GRPQUOTA, &args[0]);
1465 	else if (token == Opt_offusrjquota)
1466 		return clear_qf_name(sb, USRQUOTA);
1467 	else if (token == Opt_offgrpjquota)
1468 		return clear_qf_name(sb, GRPQUOTA);
1469 #endif
1470 	switch (token) {
1471 	case Opt_noacl:
1472 	case Opt_nouser_xattr:
1473 		ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1474 		break;
1475 	case Opt_sb:
1476 		return 1;	/* handled by get_sb_block() */
1477 	case Opt_removed:
1478 		ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1479 		return 1;
1480 	case Opt_abort:
1481 		sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1482 		return 1;
1483 	case Opt_i_version:
1484 		sb->s_flags |= MS_I_VERSION;
1485 		return 1;
1486 	case Opt_lazytime:
1487 		sb->s_flags |= MS_LAZYTIME;
1488 		return 1;
1489 	case Opt_nolazytime:
1490 		sb->s_flags &= ~MS_LAZYTIME;
1491 		return 1;
1492 	}
1493 
1494 	for (m = ext4_mount_opts; m->token != Opt_err; m++)
1495 		if (token == m->token)
1496 			break;
1497 
1498 	if (m->token == Opt_err) {
1499 		ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1500 			 "or missing value", opt);
1501 		return -1;
1502 	}
1503 
1504 	if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1505 		ext4_msg(sb, KERN_ERR,
1506 			 "Mount option \"%s\" incompatible with ext2", opt);
1507 		return -1;
1508 	}
1509 	if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1510 		ext4_msg(sb, KERN_ERR,
1511 			 "Mount option \"%s\" incompatible with ext3", opt);
1512 		return -1;
1513 	}
1514 
1515 	if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1516 		return -1;
1517 	if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1518 		return -1;
1519 	if (m->flags & MOPT_EXPLICIT) {
1520 		if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1521 			set_opt2(sb, EXPLICIT_DELALLOC);
1522 		} else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1523 			set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1524 		} else
1525 			return -1;
1526 	}
1527 	if (m->flags & MOPT_CLEAR_ERR)
1528 		clear_opt(sb, ERRORS_MASK);
1529 	if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1530 		ext4_msg(sb, KERN_ERR, "Cannot change quota "
1531 			 "options when quota turned on");
1532 		return -1;
1533 	}
1534 
1535 	if (m->flags & MOPT_NOSUPPORT) {
1536 		ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1537 	} else if (token == Opt_commit) {
1538 		if (arg == 0)
1539 			arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1540 		sbi->s_commit_interval = HZ * arg;
1541 	} else if (token == Opt_max_batch_time) {
1542 		sbi->s_max_batch_time = arg;
1543 	} else if (token == Opt_min_batch_time) {
1544 		sbi->s_min_batch_time = arg;
1545 	} else if (token == Opt_inode_readahead_blks) {
1546 		if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1547 			ext4_msg(sb, KERN_ERR,
1548 				 "EXT4-fs: inode_readahead_blks must be "
1549 				 "0 or a power of 2 smaller than 2^31");
1550 			return -1;
1551 		}
1552 		sbi->s_inode_readahead_blks = arg;
1553 	} else if (token == Opt_init_itable) {
1554 		set_opt(sb, INIT_INODE_TABLE);
1555 		if (!args->from)
1556 			arg = EXT4_DEF_LI_WAIT_MULT;
1557 		sbi->s_li_wait_mult = arg;
1558 	} else if (token == Opt_max_dir_size_kb) {
1559 		sbi->s_max_dir_size_kb = arg;
1560 	} else if (token == Opt_stripe) {
1561 		sbi->s_stripe = arg;
1562 	} else if (token == Opt_resuid) {
1563 		uid = make_kuid(current_user_ns(), arg);
1564 		if (!uid_valid(uid)) {
1565 			ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1566 			return -1;
1567 		}
1568 		sbi->s_resuid = uid;
1569 	} else if (token == Opt_resgid) {
1570 		gid = make_kgid(current_user_ns(), arg);
1571 		if (!gid_valid(gid)) {
1572 			ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1573 			return -1;
1574 		}
1575 		sbi->s_resgid = gid;
1576 	} else if (token == Opt_journal_dev) {
1577 		if (is_remount) {
1578 			ext4_msg(sb, KERN_ERR,
1579 				 "Cannot specify journal on remount");
1580 			return -1;
1581 		}
1582 		*journal_devnum = arg;
1583 	} else if (token == Opt_journal_path) {
1584 		char *journal_path;
1585 		struct inode *journal_inode;
1586 		struct path path;
1587 		int error;
1588 
1589 		if (is_remount) {
1590 			ext4_msg(sb, KERN_ERR,
1591 				 "Cannot specify journal on remount");
1592 			return -1;
1593 		}
1594 		journal_path = match_strdup(&args[0]);
1595 		if (!journal_path) {
1596 			ext4_msg(sb, KERN_ERR, "error: could not dup "
1597 				"journal device string");
1598 			return -1;
1599 		}
1600 
1601 		error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1602 		if (error) {
1603 			ext4_msg(sb, KERN_ERR, "error: could not find "
1604 				"journal device path: error %d", error);
1605 			kfree(journal_path);
1606 			return -1;
1607 		}
1608 
1609 		journal_inode = d_inode(path.dentry);
1610 		if (!S_ISBLK(journal_inode->i_mode)) {
1611 			ext4_msg(sb, KERN_ERR, "error: journal path %s "
1612 				"is not a block device", journal_path);
1613 			path_put(&path);
1614 			kfree(journal_path);
1615 			return -1;
1616 		}
1617 
1618 		*journal_devnum = new_encode_dev(journal_inode->i_rdev);
1619 		path_put(&path);
1620 		kfree(journal_path);
1621 	} else if (token == Opt_journal_ioprio) {
1622 		if (arg > 7) {
1623 			ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1624 				 " (must be 0-7)");
1625 			return -1;
1626 		}
1627 		*journal_ioprio =
1628 			IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1629 	} else if (token == Opt_test_dummy_encryption) {
1630 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1631 		sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1632 		ext4_msg(sb, KERN_WARNING,
1633 			 "Test dummy encryption mode enabled");
1634 #else
1635 		ext4_msg(sb, KERN_WARNING,
1636 			 "Test dummy encryption mount option ignored");
1637 #endif
1638 	} else if (m->flags & MOPT_DATAJ) {
1639 		if (is_remount) {
1640 			if (!sbi->s_journal)
1641 				ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1642 			else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1643 				ext4_msg(sb, KERN_ERR,
1644 					 "Cannot change data mode on remount");
1645 				return -1;
1646 			}
1647 		} else {
1648 			clear_opt(sb, DATA_FLAGS);
1649 			sbi->s_mount_opt |= m->mount_opt;
1650 		}
1651 #ifdef CONFIG_QUOTA
1652 	} else if (m->flags & MOPT_QFMT) {
1653 		if (sb_any_quota_loaded(sb) &&
1654 		    sbi->s_jquota_fmt != m->mount_opt) {
1655 			ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1656 				 "quota options when quota turned on");
1657 			return -1;
1658 		}
1659 		if (ext4_has_feature_quota(sb)) {
1660 			ext4_msg(sb, KERN_ERR,
1661 				 "Cannot set journaled quota options "
1662 				 "when QUOTA feature is enabled");
1663 			return -1;
1664 		}
1665 		sbi->s_jquota_fmt = m->mount_opt;
1666 #endif
1667 	} else if (token == Opt_dax) {
1668 #ifdef CONFIG_FS_DAX
1669 		ext4_msg(sb, KERN_WARNING,
1670 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1671 			sbi->s_mount_opt |= m->mount_opt;
1672 #else
1673 		ext4_msg(sb, KERN_INFO, "dax option not supported");
1674 		return -1;
1675 #endif
1676 	} else {
1677 		if (!args->from)
1678 			arg = 1;
1679 		if (m->flags & MOPT_CLEAR)
1680 			arg = !arg;
1681 		else if (unlikely(!(m->flags & MOPT_SET))) {
1682 			ext4_msg(sb, KERN_WARNING,
1683 				 "buggy handling of option %s", opt);
1684 			WARN_ON(1);
1685 			return -1;
1686 		}
1687 		if (arg != 0)
1688 			sbi->s_mount_opt |= m->mount_opt;
1689 		else
1690 			sbi->s_mount_opt &= ~m->mount_opt;
1691 	}
1692 	return 1;
1693 }
1694 
1695 static int parse_options(char *options, struct super_block *sb,
1696 			 unsigned long *journal_devnum,
1697 			 unsigned int *journal_ioprio,
1698 			 int is_remount)
1699 {
1700 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1701 	char *p;
1702 	substring_t args[MAX_OPT_ARGS];
1703 	int token;
1704 
1705 	if (!options)
1706 		return 1;
1707 
1708 	while ((p = strsep(&options, ",")) != NULL) {
1709 		if (!*p)
1710 			continue;
1711 		/*
1712 		 * Initialize args struct so we know whether arg was
1713 		 * found; some options take optional arguments.
1714 		 */
1715 		args[0].to = args[0].from = NULL;
1716 		token = match_token(p, tokens, args);
1717 		if (handle_mount_opt(sb, p, token, args, journal_devnum,
1718 				     journal_ioprio, is_remount) < 0)
1719 			return 0;
1720 	}
1721 #ifdef CONFIG_QUOTA
1722 	if (ext4_has_feature_quota(sb) &&
1723 	    (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1724 		ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1725 			 "feature is enabled");
1726 		return 0;
1727 	}
1728 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1729 		if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1730 			clear_opt(sb, USRQUOTA);
1731 
1732 		if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1733 			clear_opt(sb, GRPQUOTA);
1734 
1735 		if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1736 			ext4_msg(sb, KERN_ERR, "old and new quota "
1737 					"format mixing");
1738 			return 0;
1739 		}
1740 
1741 		if (!sbi->s_jquota_fmt) {
1742 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1743 					"not specified");
1744 			return 0;
1745 		}
1746 	}
1747 #endif
1748 	if (test_opt(sb, DIOREAD_NOLOCK)) {
1749 		int blocksize =
1750 			BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1751 
1752 		if (blocksize < PAGE_CACHE_SIZE) {
1753 			ext4_msg(sb, KERN_ERR, "can't mount with "
1754 				 "dioread_nolock if block size != PAGE_SIZE");
1755 			return 0;
1756 		}
1757 	}
1758 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1759 	    test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1760 		ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1761 			 "in data=ordered mode");
1762 		return 0;
1763 	}
1764 	return 1;
1765 }
1766 
1767 static inline void ext4_show_quota_options(struct seq_file *seq,
1768 					   struct super_block *sb)
1769 {
1770 #if defined(CONFIG_QUOTA)
1771 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1772 
1773 	if (sbi->s_jquota_fmt) {
1774 		char *fmtname = "";
1775 
1776 		switch (sbi->s_jquota_fmt) {
1777 		case QFMT_VFS_OLD:
1778 			fmtname = "vfsold";
1779 			break;
1780 		case QFMT_VFS_V0:
1781 			fmtname = "vfsv0";
1782 			break;
1783 		case QFMT_VFS_V1:
1784 			fmtname = "vfsv1";
1785 			break;
1786 		}
1787 		seq_printf(seq, ",jqfmt=%s", fmtname);
1788 	}
1789 
1790 	if (sbi->s_qf_names[USRQUOTA])
1791 		seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1792 
1793 	if (sbi->s_qf_names[GRPQUOTA])
1794 		seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1795 #endif
1796 }
1797 
1798 static const char *token2str(int token)
1799 {
1800 	const struct match_token *t;
1801 
1802 	for (t = tokens; t->token != Opt_err; t++)
1803 		if (t->token == token && !strchr(t->pattern, '='))
1804 			break;
1805 	return t->pattern;
1806 }
1807 
1808 /*
1809  * Show an option if
1810  *  - it's set to a non-default value OR
1811  *  - if the per-sb default is different from the global default
1812  */
1813 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1814 			      int nodefs)
1815 {
1816 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1817 	struct ext4_super_block *es = sbi->s_es;
1818 	int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1819 	const struct mount_opts *m;
1820 	char sep = nodefs ? '\n' : ',';
1821 
1822 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1823 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1824 
1825 	if (sbi->s_sb_block != 1)
1826 		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1827 
1828 	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1829 		int want_set = m->flags & MOPT_SET;
1830 		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1831 		    (m->flags & MOPT_CLEAR_ERR))
1832 			continue;
1833 		if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1834 			continue; /* skip if same as the default */
1835 		if ((want_set &&
1836 		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1837 		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1838 			continue; /* select Opt_noFoo vs Opt_Foo */
1839 		SEQ_OPTS_PRINT("%s", token2str(m->token));
1840 	}
1841 
1842 	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1843 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1844 		SEQ_OPTS_PRINT("resuid=%u",
1845 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
1846 	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1847 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1848 		SEQ_OPTS_PRINT("resgid=%u",
1849 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
1850 	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1851 	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1852 		SEQ_OPTS_PUTS("errors=remount-ro");
1853 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1854 		SEQ_OPTS_PUTS("errors=continue");
1855 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1856 		SEQ_OPTS_PUTS("errors=panic");
1857 	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1858 		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1859 	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1860 		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1861 	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1862 		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1863 	if (sb->s_flags & MS_I_VERSION)
1864 		SEQ_OPTS_PUTS("i_version");
1865 	if (nodefs || sbi->s_stripe)
1866 		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1867 	if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1868 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1869 			SEQ_OPTS_PUTS("data=journal");
1870 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1871 			SEQ_OPTS_PUTS("data=ordered");
1872 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1873 			SEQ_OPTS_PUTS("data=writeback");
1874 	}
1875 	if (nodefs ||
1876 	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1877 		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1878 			       sbi->s_inode_readahead_blks);
1879 
1880 	if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1881 		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1882 		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1883 	if (nodefs || sbi->s_max_dir_size_kb)
1884 		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1885 
1886 	ext4_show_quota_options(seq, sb);
1887 	return 0;
1888 }
1889 
1890 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1891 {
1892 	return _ext4_show_options(seq, root->d_sb, 0);
1893 }
1894 
1895 int ext4_seq_options_show(struct seq_file *seq, void *offset)
1896 {
1897 	struct super_block *sb = seq->private;
1898 	int rc;
1899 
1900 	seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1901 	rc = _ext4_show_options(seq, sb, 1);
1902 	seq_puts(seq, "\n");
1903 	return rc;
1904 }
1905 
1906 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1907 			    int read_only)
1908 {
1909 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1910 	int res = 0;
1911 
1912 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1913 		ext4_msg(sb, KERN_ERR, "revision level too high, "
1914 			 "forcing read-only mode");
1915 		res = MS_RDONLY;
1916 	}
1917 	if (read_only)
1918 		goto done;
1919 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1920 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1921 			 "running e2fsck is recommended");
1922 	else if (sbi->s_mount_state & EXT4_ERROR_FS)
1923 		ext4_msg(sb, KERN_WARNING,
1924 			 "warning: mounting fs with errors, "
1925 			 "running e2fsck is recommended");
1926 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1927 		 le16_to_cpu(es->s_mnt_count) >=
1928 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1929 		ext4_msg(sb, KERN_WARNING,
1930 			 "warning: maximal mount count reached, "
1931 			 "running e2fsck is recommended");
1932 	else if (le32_to_cpu(es->s_checkinterval) &&
1933 		(le32_to_cpu(es->s_lastcheck) +
1934 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1935 		ext4_msg(sb, KERN_WARNING,
1936 			 "warning: checktime reached, "
1937 			 "running e2fsck is recommended");
1938 	if (!sbi->s_journal)
1939 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1940 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1941 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1942 	le16_add_cpu(&es->s_mnt_count, 1);
1943 	es->s_mtime = cpu_to_le32(get_seconds());
1944 	ext4_update_dynamic_rev(sb);
1945 	if (sbi->s_journal)
1946 		ext4_set_feature_journal_needs_recovery(sb);
1947 
1948 	ext4_commit_super(sb, 1);
1949 done:
1950 	if (test_opt(sb, DEBUG))
1951 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1952 				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1953 			sb->s_blocksize,
1954 			sbi->s_groups_count,
1955 			EXT4_BLOCKS_PER_GROUP(sb),
1956 			EXT4_INODES_PER_GROUP(sb),
1957 			sbi->s_mount_opt, sbi->s_mount_opt2);
1958 
1959 	cleancache_init_fs(sb);
1960 	return res;
1961 }
1962 
1963 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1964 {
1965 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1966 	struct flex_groups *new_groups;
1967 	int size;
1968 
1969 	if (!sbi->s_log_groups_per_flex)
1970 		return 0;
1971 
1972 	size = ext4_flex_group(sbi, ngroup - 1) + 1;
1973 	if (size <= sbi->s_flex_groups_allocated)
1974 		return 0;
1975 
1976 	size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1977 	new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1978 	if (!new_groups) {
1979 		ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1980 			 size / (int) sizeof(struct flex_groups));
1981 		return -ENOMEM;
1982 	}
1983 
1984 	if (sbi->s_flex_groups) {
1985 		memcpy(new_groups, sbi->s_flex_groups,
1986 		       (sbi->s_flex_groups_allocated *
1987 			sizeof(struct flex_groups)));
1988 		kvfree(sbi->s_flex_groups);
1989 	}
1990 	sbi->s_flex_groups = new_groups;
1991 	sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1992 	return 0;
1993 }
1994 
1995 static int ext4_fill_flex_info(struct super_block *sb)
1996 {
1997 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1998 	struct ext4_group_desc *gdp = NULL;
1999 	ext4_group_t flex_group;
2000 	int i, err;
2001 
2002 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2003 	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2004 		sbi->s_log_groups_per_flex = 0;
2005 		return 1;
2006 	}
2007 
2008 	err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2009 	if (err)
2010 		goto failed;
2011 
2012 	for (i = 0; i < sbi->s_groups_count; i++) {
2013 		gdp = ext4_get_group_desc(sb, i, NULL);
2014 
2015 		flex_group = ext4_flex_group(sbi, i);
2016 		atomic_add(ext4_free_inodes_count(sb, gdp),
2017 			   &sbi->s_flex_groups[flex_group].free_inodes);
2018 		atomic64_add(ext4_free_group_clusters(sb, gdp),
2019 			     &sbi->s_flex_groups[flex_group].free_clusters);
2020 		atomic_add(ext4_used_dirs_count(sb, gdp),
2021 			   &sbi->s_flex_groups[flex_group].used_dirs);
2022 	}
2023 
2024 	return 1;
2025 failed:
2026 	return 0;
2027 }
2028 
2029 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2030 				   struct ext4_group_desc *gdp)
2031 {
2032 	int offset;
2033 	__u16 crc = 0;
2034 	__le32 le_group = cpu_to_le32(block_group);
2035 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2036 
2037 	if (ext4_has_metadata_csum(sbi->s_sb)) {
2038 		/* Use new metadata_csum algorithm */
2039 		__le16 save_csum;
2040 		__u32 csum32;
2041 
2042 		save_csum = gdp->bg_checksum;
2043 		gdp->bg_checksum = 0;
2044 		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2045 				     sizeof(le_group));
2046 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2047 				     sbi->s_desc_size);
2048 		gdp->bg_checksum = save_csum;
2049 
2050 		crc = csum32 & 0xFFFF;
2051 		goto out;
2052 	}
2053 
2054 	/* old crc16 code */
2055 	if (!ext4_has_feature_gdt_csum(sb))
2056 		return 0;
2057 
2058 	offset = offsetof(struct ext4_group_desc, bg_checksum);
2059 
2060 	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2061 	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2062 	crc = crc16(crc, (__u8 *)gdp, offset);
2063 	offset += sizeof(gdp->bg_checksum); /* skip checksum */
2064 	/* for checksum of struct ext4_group_desc do the rest...*/
2065 	if (ext4_has_feature_64bit(sb) &&
2066 	    offset < le16_to_cpu(sbi->s_es->s_desc_size))
2067 		crc = crc16(crc, (__u8 *)gdp + offset,
2068 			    le16_to_cpu(sbi->s_es->s_desc_size) -
2069 				offset);
2070 
2071 out:
2072 	return cpu_to_le16(crc);
2073 }
2074 
2075 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2076 				struct ext4_group_desc *gdp)
2077 {
2078 	if (ext4_has_group_desc_csum(sb) &&
2079 	    (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2080 		return 0;
2081 
2082 	return 1;
2083 }
2084 
2085 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2086 			      struct ext4_group_desc *gdp)
2087 {
2088 	if (!ext4_has_group_desc_csum(sb))
2089 		return;
2090 	gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2091 }
2092 
2093 /* Called at mount-time, super-block is locked */
2094 static int ext4_check_descriptors(struct super_block *sb,
2095 				  ext4_group_t *first_not_zeroed)
2096 {
2097 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2098 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2099 	ext4_fsblk_t last_block;
2100 	ext4_fsblk_t block_bitmap;
2101 	ext4_fsblk_t inode_bitmap;
2102 	ext4_fsblk_t inode_table;
2103 	int flexbg_flag = 0;
2104 	ext4_group_t i, grp = sbi->s_groups_count;
2105 
2106 	if (ext4_has_feature_flex_bg(sb))
2107 		flexbg_flag = 1;
2108 
2109 	ext4_debug("Checking group descriptors");
2110 
2111 	for (i = 0; i < sbi->s_groups_count; i++) {
2112 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2113 
2114 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
2115 			last_block = ext4_blocks_count(sbi->s_es) - 1;
2116 		else
2117 			last_block = first_block +
2118 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
2119 
2120 		if ((grp == sbi->s_groups_count) &&
2121 		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2122 			grp = i;
2123 
2124 		block_bitmap = ext4_block_bitmap(sb, gdp);
2125 		if (block_bitmap < first_block || block_bitmap > last_block) {
2126 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2127 			       "Block bitmap for group %u not in group "
2128 			       "(block %llu)!", i, block_bitmap);
2129 			return 0;
2130 		}
2131 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
2132 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
2133 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2134 			       "Inode bitmap for group %u not in group "
2135 			       "(block %llu)!", i, inode_bitmap);
2136 			return 0;
2137 		}
2138 		inode_table = ext4_inode_table(sb, gdp);
2139 		if (inode_table < first_block ||
2140 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2141 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2142 			       "Inode table for group %u not in group "
2143 			       "(block %llu)!", i, inode_table);
2144 			return 0;
2145 		}
2146 		ext4_lock_group(sb, i);
2147 		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2148 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2149 				 "Checksum for group %u failed (%u!=%u)",
2150 				 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2151 				     gdp)), le16_to_cpu(gdp->bg_checksum));
2152 			if (!(sb->s_flags & MS_RDONLY)) {
2153 				ext4_unlock_group(sb, i);
2154 				return 0;
2155 			}
2156 		}
2157 		ext4_unlock_group(sb, i);
2158 		if (!flexbg_flag)
2159 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
2160 	}
2161 	if (NULL != first_not_zeroed)
2162 		*first_not_zeroed = grp;
2163 	return 1;
2164 }
2165 
2166 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2167  * the superblock) which were deleted from all directories, but held open by
2168  * a process at the time of a crash.  We walk the list and try to delete these
2169  * inodes at recovery time (only with a read-write filesystem).
2170  *
2171  * In order to keep the orphan inode chain consistent during traversal (in
2172  * case of crash during recovery), we link each inode into the superblock
2173  * orphan list_head and handle it the same way as an inode deletion during
2174  * normal operation (which journals the operations for us).
2175  *
2176  * We only do an iget() and an iput() on each inode, which is very safe if we
2177  * accidentally point at an in-use or already deleted inode.  The worst that
2178  * can happen in this case is that we get a "bit already cleared" message from
2179  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2180  * e2fsck was run on this filesystem, and it must have already done the orphan
2181  * inode cleanup for us, so we can safely abort without any further action.
2182  */
2183 static void ext4_orphan_cleanup(struct super_block *sb,
2184 				struct ext4_super_block *es)
2185 {
2186 	unsigned int s_flags = sb->s_flags;
2187 	int nr_orphans = 0, nr_truncates = 0;
2188 #ifdef CONFIG_QUOTA
2189 	int i;
2190 #endif
2191 	if (!es->s_last_orphan) {
2192 		jbd_debug(4, "no orphan inodes to clean up\n");
2193 		return;
2194 	}
2195 
2196 	if (bdev_read_only(sb->s_bdev)) {
2197 		ext4_msg(sb, KERN_ERR, "write access "
2198 			"unavailable, skipping orphan cleanup");
2199 		return;
2200 	}
2201 
2202 	/* Check if feature set would not allow a r/w mount */
2203 	if (!ext4_feature_set_ok(sb, 0)) {
2204 		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2205 			 "unknown ROCOMPAT features");
2206 		return;
2207 	}
2208 
2209 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2210 		/* don't clear list on RO mount w/ errors */
2211 		if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2212 			ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2213 				  "clearing orphan list.\n");
2214 			es->s_last_orphan = 0;
2215 		}
2216 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2217 		return;
2218 	}
2219 
2220 	if (s_flags & MS_RDONLY) {
2221 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2222 		sb->s_flags &= ~MS_RDONLY;
2223 	}
2224 #ifdef CONFIG_QUOTA
2225 	/* Needed for iput() to work correctly and not trash data */
2226 	sb->s_flags |= MS_ACTIVE;
2227 	/* Turn on quotas so that they are updated correctly */
2228 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2229 		if (EXT4_SB(sb)->s_qf_names[i]) {
2230 			int ret = ext4_quota_on_mount(sb, i);
2231 			if (ret < 0)
2232 				ext4_msg(sb, KERN_ERR,
2233 					"Cannot turn on journaled "
2234 					"quota: error %d", ret);
2235 		}
2236 	}
2237 #endif
2238 
2239 	while (es->s_last_orphan) {
2240 		struct inode *inode;
2241 
2242 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2243 		if (IS_ERR(inode)) {
2244 			es->s_last_orphan = 0;
2245 			break;
2246 		}
2247 
2248 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2249 		dquot_initialize(inode);
2250 		if (inode->i_nlink) {
2251 			if (test_opt(sb, DEBUG))
2252 				ext4_msg(sb, KERN_DEBUG,
2253 					"%s: truncating inode %lu to %lld bytes",
2254 					__func__, inode->i_ino, inode->i_size);
2255 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2256 				  inode->i_ino, inode->i_size);
2257 			mutex_lock(&inode->i_mutex);
2258 			truncate_inode_pages(inode->i_mapping, inode->i_size);
2259 			ext4_truncate(inode);
2260 			mutex_unlock(&inode->i_mutex);
2261 			nr_truncates++;
2262 		} else {
2263 			if (test_opt(sb, DEBUG))
2264 				ext4_msg(sb, KERN_DEBUG,
2265 					"%s: deleting unreferenced inode %lu",
2266 					__func__, inode->i_ino);
2267 			jbd_debug(2, "deleting unreferenced inode %lu\n",
2268 				  inode->i_ino);
2269 			nr_orphans++;
2270 		}
2271 		iput(inode);  /* The delete magic happens here! */
2272 	}
2273 
2274 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2275 
2276 	if (nr_orphans)
2277 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2278 		       PLURAL(nr_orphans));
2279 	if (nr_truncates)
2280 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2281 		       PLURAL(nr_truncates));
2282 #ifdef CONFIG_QUOTA
2283 	/* Turn quotas off */
2284 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2285 		if (sb_dqopt(sb)->files[i])
2286 			dquot_quota_off(sb, i);
2287 	}
2288 #endif
2289 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2290 }
2291 
2292 /*
2293  * Maximal extent format file size.
2294  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2295  * extent format containers, within a sector_t, and within i_blocks
2296  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2297  * so that won't be a limiting factor.
2298  *
2299  * However there is other limiting factor. We do store extents in the form
2300  * of starting block and length, hence the resulting length of the extent
2301  * covering maximum file size must fit into on-disk format containers as
2302  * well. Given that length is always by 1 unit bigger than max unit (because
2303  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2304  *
2305  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2306  */
2307 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2308 {
2309 	loff_t res;
2310 	loff_t upper_limit = MAX_LFS_FILESIZE;
2311 
2312 	/* small i_blocks in vfs inode? */
2313 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2314 		/*
2315 		 * CONFIG_LBDAF is not enabled implies the inode
2316 		 * i_block represent total blocks in 512 bytes
2317 		 * 32 == size of vfs inode i_blocks * 8
2318 		 */
2319 		upper_limit = (1LL << 32) - 1;
2320 
2321 		/* total blocks in file system block size */
2322 		upper_limit >>= (blkbits - 9);
2323 		upper_limit <<= blkbits;
2324 	}
2325 
2326 	/*
2327 	 * 32-bit extent-start container, ee_block. We lower the maxbytes
2328 	 * by one fs block, so ee_len can cover the extent of maximum file
2329 	 * size
2330 	 */
2331 	res = (1LL << 32) - 1;
2332 	res <<= blkbits;
2333 
2334 	/* Sanity check against vm- & vfs- imposed limits */
2335 	if (res > upper_limit)
2336 		res = upper_limit;
2337 
2338 	return res;
2339 }
2340 
2341 /*
2342  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2343  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2344  * We need to be 1 filesystem block less than the 2^48 sector limit.
2345  */
2346 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2347 {
2348 	loff_t res = EXT4_NDIR_BLOCKS;
2349 	int meta_blocks;
2350 	loff_t upper_limit;
2351 	/* This is calculated to be the largest file size for a dense, block
2352 	 * mapped file such that the file's total number of 512-byte sectors,
2353 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
2354 	 *
2355 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2356 	 * number of 512-byte sectors of the file.
2357 	 */
2358 
2359 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2360 		/*
2361 		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2362 		 * the inode i_block field represents total file blocks in
2363 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2364 		 */
2365 		upper_limit = (1LL << 32) - 1;
2366 
2367 		/* total blocks in file system block size */
2368 		upper_limit >>= (bits - 9);
2369 
2370 	} else {
2371 		/*
2372 		 * We use 48 bit ext4_inode i_blocks
2373 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2374 		 * represent total number of blocks in
2375 		 * file system block size
2376 		 */
2377 		upper_limit = (1LL << 48) - 1;
2378 
2379 	}
2380 
2381 	/* indirect blocks */
2382 	meta_blocks = 1;
2383 	/* double indirect blocks */
2384 	meta_blocks += 1 + (1LL << (bits-2));
2385 	/* tripple indirect blocks */
2386 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2387 
2388 	upper_limit -= meta_blocks;
2389 	upper_limit <<= bits;
2390 
2391 	res += 1LL << (bits-2);
2392 	res += 1LL << (2*(bits-2));
2393 	res += 1LL << (3*(bits-2));
2394 	res <<= bits;
2395 	if (res > upper_limit)
2396 		res = upper_limit;
2397 
2398 	if (res > MAX_LFS_FILESIZE)
2399 		res = MAX_LFS_FILESIZE;
2400 
2401 	return res;
2402 }
2403 
2404 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2405 				   ext4_fsblk_t logical_sb_block, int nr)
2406 {
2407 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2408 	ext4_group_t bg, first_meta_bg;
2409 	int has_super = 0;
2410 
2411 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2412 
2413 	if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2414 		return logical_sb_block + nr + 1;
2415 	bg = sbi->s_desc_per_block * nr;
2416 	if (ext4_bg_has_super(sb, bg))
2417 		has_super = 1;
2418 
2419 	/*
2420 	 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2421 	 * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
2422 	 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2423 	 * compensate.
2424 	 */
2425 	if (sb->s_blocksize == 1024 && nr == 0 &&
2426 	    le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2427 		has_super++;
2428 
2429 	return (has_super + ext4_group_first_block_no(sb, bg));
2430 }
2431 
2432 /**
2433  * ext4_get_stripe_size: Get the stripe size.
2434  * @sbi: In memory super block info
2435  *
2436  * If we have specified it via mount option, then
2437  * use the mount option value. If the value specified at mount time is
2438  * greater than the blocks per group use the super block value.
2439  * If the super block value is greater than blocks per group return 0.
2440  * Allocator needs it be less than blocks per group.
2441  *
2442  */
2443 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2444 {
2445 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2446 	unsigned long stripe_width =
2447 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2448 	int ret;
2449 
2450 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2451 		ret = sbi->s_stripe;
2452 	else if (stripe_width <= sbi->s_blocks_per_group)
2453 		ret = stripe_width;
2454 	else if (stride <= sbi->s_blocks_per_group)
2455 		ret = stride;
2456 	else
2457 		ret = 0;
2458 
2459 	/*
2460 	 * If the stripe width is 1, this makes no sense and
2461 	 * we set it to 0 to turn off stripe handling code.
2462 	 */
2463 	if (ret <= 1)
2464 		ret = 0;
2465 
2466 	return ret;
2467 }
2468 
2469 /*
2470  * Check whether this filesystem can be mounted based on
2471  * the features present and the RDONLY/RDWR mount requested.
2472  * Returns 1 if this filesystem can be mounted as requested,
2473  * 0 if it cannot be.
2474  */
2475 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2476 {
2477 	if (ext4_has_unknown_ext4_incompat_features(sb)) {
2478 		ext4_msg(sb, KERN_ERR,
2479 			"Couldn't mount because of "
2480 			"unsupported optional features (%x)",
2481 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2482 			~EXT4_FEATURE_INCOMPAT_SUPP));
2483 		return 0;
2484 	}
2485 
2486 	if (readonly)
2487 		return 1;
2488 
2489 	if (ext4_has_feature_readonly(sb)) {
2490 		ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2491 		sb->s_flags |= MS_RDONLY;
2492 		return 1;
2493 	}
2494 
2495 	/* Check that feature set is OK for a read-write mount */
2496 	if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2497 		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2498 			 "unsupported optional features (%x)",
2499 			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2500 				~EXT4_FEATURE_RO_COMPAT_SUPP));
2501 		return 0;
2502 	}
2503 	/*
2504 	 * Large file size enabled file system can only be mounted
2505 	 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2506 	 */
2507 	if (ext4_has_feature_huge_file(sb)) {
2508 		if (sizeof(blkcnt_t) < sizeof(u64)) {
2509 			ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2510 				 "cannot be mounted RDWR without "
2511 				 "CONFIG_LBDAF");
2512 			return 0;
2513 		}
2514 	}
2515 	if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2516 		ext4_msg(sb, KERN_ERR,
2517 			 "Can't support bigalloc feature without "
2518 			 "extents feature\n");
2519 		return 0;
2520 	}
2521 
2522 #ifndef CONFIG_QUOTA
2523 	if (ext4_has_feature_quota(sb) && !readonly) {
2524 		ext4_msg(sb, KERN_ERR,
2525 			 "Filesystem with quota feature cannot be mounted RDWR "
2526 			 "without CONFIG_QUOTA");
2527 		return 0;
2528 	}
2529 #endif  /* CONFIG_QUOTA */
2530 	return 1;
2531 }
2532 
2533 /*
2534  * This function is called once a day if we have errors logged
2535  * on the file system
2536  */
2537 static void print_daily_error_info(unsigned long arg)
2538 {
2539 	struct super_block *sb = (struct super_block *) arg;
2540 	struct ext4_sb_info *sbi;
2541 	struct ext4_super_block *es;
2542 
2543 	sbi = EXT4_SB(sb);
2544 	es = sbi->s_es;
2545 
2546 	if (es->s_error_count)
2547 		/* fsck newer than v1.41.13 is needed to clean this condition. */
2548 		ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2549 			 le32_to_cpu(es->s_error_count));
2550 	if (es->s_first_error_time) {
2551 		printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2552 		       sb->s_id, le32_to_cpu(es->s_first_error_time),
2553 		       (int) sizeof(es->s_first_error_func),
2554 		       es->s_first_error_func,
2555 		       le32_to_cpu(es->s_first_error_line));
2556 		if (es->s_first_error_ino)
2557 			printk(": inode %u",
2558 			       le32_to_cpu(es->s_first_error_ino));
2559 		if (es->s_first_error_block)
2560 			printk(": block %llu", (unsigned long long)
2561 			       le64_to_cpu(es->s_first_error_block));
2562 		printk("\n");
2563 	}
2564 	if (es->s_last_error_time) {
2565 		printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2566 		       sb->s_id, le32_to_cpu(es->s_last_error_time),
2567 		       (int) sizeof(es->s_last_error_func),
2568 		       es->s_last_error_func,
2569 		       le32_to_cpu(es->s_last_error_line));
2570 		if (es->s_last_error_ino)
2571 			printk(": inode %u",
2572 			       le32_to_cpu(es->s_last_error_ino));
2573 		if (es->s_last_error_block)
2574 			printk(": block %llu", (unsigned long long)
2575 			       le64_to_cpu(es->s_last_error_block));
2576 		printk("\n");
2577 	}
2578 	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2579 }
2580 
2581 /* Find next suitable group and run ext4_init_inode_table */
2582 static int ext4_run_li_request(struct ext4_li_request *elr)
2583 {
2584 	struct ext4_group_desc *gdp = NULL;
2585 	ext4_group_t group, ngroups;
2586 	struct super_block *sb;
2587 	unsigned long timeout = 0;
2588 	int ret = 0;
2589 
2590 	sb = elr->lr_super;
2591 	ngroups = EXT4_SB(sb)->s_groups_count;
2592 
2593 	sb_start_write(sb);
2594 	for (group = elr->lr_next_group; group < ngroups; group++) {
2595 		gdp = ext4_get_group_desc(sb, group, NULL);
2596 		if (!gdp) {
2597 			ret = 1;
2598 			break;
2599 		}
2600 
2601 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2602 			break;
2603 	}
2604 
2605 	if (group >= ngroups)
2606 		ret = 1;
2607 
2608 	if (!ret) {
2609 		timeout = jiffies;
2610 		ret = ext4_init_inode_table(sb, group,
2611 					    elr->lr_timeout ? 0 : 1);
2612 		if (elr->lr_timeout == 0) {
2613 			timeout = (jiffies - timeout) *
2614 				  elr->lr_sbi->s_li_wait_mult;
2615 			elr->lr_timeout = timeout;
2616 		}
2617 		elr->lr_next_sched = jiffies + elr->lr_timeout;
2618 		elr->lr_next_group = group + 1;
2619 	}
2620 	sb_end_write(sb);
2621 
2622 	return ret;
2623 }
2624 
2625 /*
2626  * Remove lr_request from the list_request and free the
2627  * request structure. Should be called with li_list_mtx held
2628  */
2629 static void ext4_remove_li_request(struct ext4_li_request *elr)
2630 {
2631 	struct ext4_sb_info *sbi;
2632 
2633 	if (!elr)
2634 		return;
2635 
2636 	sbi = elr->lr_sbi;
2637 
2638 	list_del(&elr->lr_request);
2639 	sbi->s_li_request = NULL;
2640 	kfree(elr);
2641 }
2642 
2643 static void ext4_unregister_li_request(struct super_block *sb)
2644 {
2645 	mutex_lock(&ext4_li_mtx);
2646 	if (!ext4_li_info) {
2647 		mutex_unlock(&ext4_li_mtx);
2648 		return;
2649 	}
2650 
2651 	mutex_lock(&ext4_li_info->li_list_mtx);
2652 	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2653 	mutex_unlock(&ext4_li_info->li_list_mtx);
2654 	mutex_unlock(&ext4_li_mtx);
2655 }
2656 
2657 static struct task_struct *ext4_lazyinit_task;
2658 
2659 /*
2660  * This is the function where ext4lazyinit thread lives. It walks
2661  * through the request list searching for next scheduled filesystem.
2662  * When such a fs is found, run the lazy initialization request
2663  * (ext4_rn_li_request) and keep track of the time spend in this
2664  * function. Based on that time we compute next schedule time of
2665  * the request. When walking through the list is complete, compute
2666  * next waking time and put itself into sleep.
2667  */
2668 static int ext4_lazyinit_thread(void *arg)
2669 {
2670 	struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2671 	struct list_head *pos, *n;
2672 	struct ext4_li_request *elr;
2673 	unsigned long next_wakeup, cur;
2674 
2675 	BUG_ON(NULL == eli);
2676 
2677 cont_thread:
2678 	while (true) {
2679 		next_wakeup = MAX_JIFFY_OFFSET;
2680 
2681 		mutex_lock(&eli->li_list_mtx);
2682 		if (list_empty(&eli->li_request_list)) {
2683 			mutex_unlock(&eli->li_list_mtx);
2684 			goto exit_thread;
2685 		}
2686 
2687 		list_for_each_safe(pos, n, &eli->li_request_list) {
2688 			elr = list_entry(pos, struct ext4_li_request,
2689 					 lr_request);
2690 
2691 			if (time_after_eq(jiffies, elr->lr_next_sched)) {
2692 				if (ext4_run_li_request(elr) != 0) {
2693 					/* error, remove the lazy_init job */
2694 					ext4_remove_li_request(elr);
2695 					continue;
2696 				}
2697 			}
2698 
2699 			if (time_before(elr->lr_next_sched, next_wakeup))
2700 				next_wakeup = elr->lr_next_sched;
2701 		}
2702 		mutex_unlock(&eli->li_list_mtx);
2703 
2704 		try_to_freeze();
2705 
2706 		cur = jiffies;
2707 		if ((time_after_eq(cur, next_wakeup)) ||
2708 		    (MAX_JIFFY_OFFSET == next_wakeup)) {
2709 			cond_resched();
2710 			continue;
2711 		}
2712 
2713 		schedule_timeout_interruptible(next_wakeup - cur);
2714 
2715 		if (kthread_should_stop()) {
2716 			ext4_clear_request_list();
2717 			goto exit_thread;
2718 		}
2719 	}
2720 
2721 exit_thread:
2722 	/*
2723 	 * It looks like the request list is empty, but we need
2724 	 * to check it under the li_list_mtx lock, to prevent any
2725 	 * additions into it, and of course we should lock ext4_li_mtx
2726 	 * to atomically free the list and ext4_li_info, because at
2727 	 * this point another ext4 filesystem could be registering
2728 	 * new one.
2729 	 */
2730 	mutex_lock(&ext4_li_mtx);
2731 	mutex_lock(&eli->li_list_mtx);
2732 	if (!list_empty(&eli->li_request_list)) {
2733 		mutex_unlock(&eli->li_list_mtx);
2734 		mutex_unlock(&ext4_li_mtx);
2735 		goto cont_thread;
2736 	}
2737 	mutex_unlock(&eli->li_list_mtx);
2738 	kfree(ext4_li_info);
2739 	ext4_li_info = NULL;
2740 	mutex_unlock(&ext4_li_mtx);
2741 
2742 	return 0;
2743 }
2744 
2745 static void ext4_clear_request_list(void)
2746 {
2747 	struct list_head *pos, *n;
2748 	struct ext4_li_request *elr;
2749 
2750 	mutex_lock(&ext4_li_info->li_list_mtx);
2751 	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2752 		elr = list_entry(pos, struct ext4_li_request,
2753 				 lr_request);
2754 		ext4_remove_li_request(elr);
2755 	}
2756 	mutex_unlock(&ext4_li_info->li_list_mtx);
2757 }
2758 
2759 static int ext4_run_lazyinit_thread(void)
2760 {
2761 	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2762 					 ext4_li_info, "ext4lazyinit");
2763 	if (IS_ERR(ext4_lazyinit_task)) {
2764 		int err = PTR_ERR(ext4_lazyinit_task);
2765 		ext4_clear_request_list();
2766 		kfree(ext4_li_info);
2767 		ext4_li_info = NULL;
2768 		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2769 				 "initialization thread\n",
2770 				 err);
2771 		return err;
2772 	}
2773 	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2774 	return 0;
2775 }
2776 
2777 /*
2778  * Check whether it make sense to run itable init. thread or not.
2779  * If there is at least one uninitialized inode table, return
2780  * corresponding group number, else the loop goes through all
2781  * groups and return total number of groups.
2782  */
2783 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2784 {
2785 	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2786 	struct ext4_group_desc *gdp = NULL;
2787 
2788 	for (group = 0; group < ngroups; group++) {
2789 		gdp = ext4_get_group_desc(sb, group, NULL);
2790 		if (!gdp)
2791 			continue;
2792 
2793 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2794 			break;
2795 	}
2796 
2797 	return group;
2798 }
2799 
2800 static int ext4_li_info_new(void)
2801 {
2802 	struct ext4_lazy_init *eli = NULL;
2803 
2804 	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2805 	if (!eli)
2806 		return -ENOMEM;
2807 
2808 	INIT_LIST_HEAD(&eli->li_request_list);
2809 	mutex_init(&eli->li_list_mtx);
2810 
2811 	eli->li_state |= EXT4_LAZYINIT_QUIT;
2812 
2813 	ext4_li_info = eli;
2814 
2815 	return 0;
2816 }
2817 
2818 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2819 					    ext4_group_t start)
2820 {
2821 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2822 	struct ext4_li_request *elr;
2823 
2824 	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2825 	if (!elr)
2826 		return NULL;
2827 
2828 	elr->lr_super = sb;
2829 	elr->lr_sbi = sbi;
2830 	elr->lr_next_group = start;
2831 
2832 	/*
2833 	 * Randomize first schedule time of the request to
2834 	 * spread the inode table initialization requests
2835 	 * better.
2836 	 */
2837 	elr->lr_next_sched = jiffies + (prandom_u32() %
2838 				(EXT4_DEF_LI_MAX_START_DELAY * HZ));
2839 	return elr;
2840 }
2841 
2842 int ext4_register_li_request(struct super_block *sb,
2843 			     ext4_group_t first_not_zeroed)
2844 {
2845 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2846 	struct ext4_li_request *elr = NULL;
2847 	ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2848 	int ret = 0;
2849 
2850 	mutex_lock(&ext4_li_mtx);
2851 	if (sbi->s_li_request != NULL) {
2852 		/*
2853 		 * Reset timeout so it can be computed again, because
2854 		 * s_li_wait_mult might have changed.
2855 		 */
2856 		sbi->s_li_request->lr_timeout = 0;
2857 		goto out;
2858 	}
2859 
2860 	if (first_not_zeroed == ngroups ||
2861 	    (sb->s_flags & MS_RDONLY) ||
2862 	    !test_opt(sb, INIT_INODE_TABLE))
2863 		goto out;
2864 
2865 	elr = ext4_li_request_new(sb, first_not_zeroed);
2866 	if (!elr) {
2867 		ret = -ENOMEM;
2868 		goto out;
2869 	}
2870 
2871 	if (NULL == ext4_li_info) {
2872 		ret = ext4_li_info_new();
2873 		if (ret)
2874 			goto out;
2875 	}
2876 
2877 	mutex_lock(&ext4_li_info->li_list_mtx);
2878 	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2879 	mutex_unlock(&ext4_li_info->li_list_mtx);
2880 
2881 	sbi->s_li_request = elr;
2882 	/*
2883 	 * set elr to NULL here since it has been inserted to
2884 	 * the request_list and the removal and free of it is
2885 	 * handled by ext4_clear_request_list from now on.
2886 	 */
2887 	elr = NULL;
2888 
2889 	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2890 		ret = ext4_run_lazyinit_thread();
2891 		if (ret)
2892 			goto out;
2893 	}
2894 out:
2895 	mutex_unlock(&ext4_li_mtx);
2896 	if (ret)
2897 		kfree(elr);
2898 	return ret;
2899 }
2900 
2901 /*
2902  * We do not need to lock anything since this is called on
2903  * module unload.
2904  */
2905 static void ext4_destroy_lazyinit_thread(void)
2906 {
2907 	/*
2908 	 * If thread exited earlier
2909 	 * there's nothing to be done.
2910 	 */
2911 	if (!ext4_li_info || !ext4_lazyinit_task)
2912 		return;
2913 
2914 	kthread_stop(ext4_lazyinit_task);
2915 }
2916 
2917 static int set_journal_csum_feature_set(struct super_block *sb)
2918 {
2919 	int ret = 1;
2920 	int compat, incompat;
2921 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2922 
2923 	if (ext4_has_metadata_csum(sb)) {
2924 		/* journal checksum v3 */
2925 		compat = 0;
2926 		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
2927 	} else {
2928 		/* journal checksum v1 */
2929 		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2930 		incompat = 0;
2931 	}
2932 
2933 	jbd2_journal_clear_features(sbi->s_journal,
2934 			JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2935 			JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2936 			JBD2_FEATURE_INCOMPAT_CSUM_V2);
2937 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2938 		ret = jbd2_journal_set_features(sbi->s_journal,
2939 				compat, 0,
2940 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2941 				incompat);
2942 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2943 		ret = jbd2_journal_set_features(sbi->s_journal,
2944 				compat, 0,
2945 				incompat);
2946 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2947 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2948 	} else {
2949 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2950 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2951 	}
2952 
2953 	return ret;
2954 }
2955 
2956 /*
2957  * Note: calculating the overhead so we can be compatible with
2958  * historical BSD practice is quite difficult in the face of
2959  * clusters/bigalloc.  This is because multiple metadata blocks from
2960  * different block group can end up in the same allocation cluster.
2961  * Calculating the exact overhead in the face of clustered allocation
2962  * requires either O(all block bitmaps) in memory or O(number of block
2963  * groups**2) in time.  We will still calculate the superblock for
2964  * older file systems --- and if we come across with a bigalloc file
2965  * system with zero in s_overhead_clusters the estimate will be close to
2966  * correct especially for very large cluster sizes --- but for newer
2967  * file systems, it's better to calculate this figure once at mkfs
2968  * time, and store it in the superblock.  If the superblock value is
2969  * present (even for non-bigalloc file systems), we will use it.
2970  */
2971 static int count_overhead(struct super_block *sb, ext4_group_t grp,
2972 			  char *buf)
2973 {
2974 	struct ext4_sb_info	*sbi = EXT4_SB(sb);
2975 	struct ext4_group_desc	*gdp;
2976 	ext4_fsblk_t		first_block, last_block, b;
2977 	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
2978 	int			s, j, count = 0;
2979 
2980 	if (!ext4_has_feature_bigalloc(sb))
2981 		return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
2982 			sbi->s_itb_per_group + 2);
2983 
2984 	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
2985 		(grp * EXT4_BLOCKS_PER_GROUP(sb));
2986 	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
2987 	for (i = 0; i < ngroups; i++) {
2988 		gdp = ext4_get_group_desc(sb, i, NULL);
2989 		b = ext4_block_bitmap(sb, gdp);
2990 		if (b >= first_block && b <= last_block) {
2991 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2992 			count++;
2993 		}
2994 		b = ext4_inode_bitmap(sb, gdp);
2995 		if (b >= first_block && b <= last_block) {
2996 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2997 			count++;
2998 		}
2999 		b = ext4_inode_table(sb, gdp);
3000 		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3001 			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3002 				int c = EXT4_B2C(sbi, b - first_block);
3003 				ext4_set_bit(c, buf);
3004 				count++;
3005 			}
3006 		if (i != grp)
3007 			continue;
3008 		s = 0;
3009 		if (ext4_bg_has_super(sb, grp)) {
3010 			ext4_set_bit(s++, buf);
3011 			count++;
3012 		}
3013 		for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3014 			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3015 			count++;
3016 		}
3017 	}
3018 	if (!count)
3019 		return 0;
3020 	return EXT4_CLUSTERS_PER_GROUP(sb) -
3021 		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3022 }
3023 
3024 /*
3025  * Compute the overhead and stash it in sbi->s_overhead
3026  */
3027 int ext4_calculate_overhead(struct super_block *sb)
3028 {
3029 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3030 	struct ext4_super_block *es = sbi->s_es;
3031 	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3032 	ext4_fsblk_t overhead = 0;
3033 	char *buf = (char *) get_zeroed_page(GFP_NOFS);
3034 
3035 	if (!buf)
3036 		return -ENOMEM;
3037 
3038 	/*
3039 	 * Compute the overhead (FS structures).  This is constant
3040 	 * for a given filesystem unless the number of block groups
3041 	 * changes so we cache the previous value until it does.
3042 	 */
3043 
3044 	/*
3045 	 * All of the blocks before first_data_block are overhead
3046 	 */
3047 	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3048 
3049 	/*
3050 	 * Add the overhead found in each block group
3051 	 */
3052 	for (i = 0; i < ngroups; i++) {
3053 		int blks;
3054 
3055 		blks = count_overhead(sb, i, buf);
3056 		overhead += blks;
3057 		if (blks)
3058 			memset(buf, 0, PAGE_SIZE);
3059 		cond_resched();
3060 	}
3061 	/* Add the internal journal blocks as well */
3062 	if (sbi->s_journal && !sbi->journal_bdev)
3063 		overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3064 
3065 	sbi->s_overhead = overhead;
3066 	smp_wmb();
3067 	free_page((unsigned long) buf);
3068 	return 0;
3069 }
3070 
3071 static void ext4_set_resv_clusters(struct super_block *sb)
3072 {
3073 	ext4_fsblk_t resv_clusters;
3074 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3075 
3076 	/*
3077 	 * There's no need to reserve anything when we aren't using extents.
3078 	 * The space estimates are exact, there are no unwritten extents,
3079 	 * hole punching doesn't need new metadata... This is needed especially
3080 	 * to keep ext2/3 backward compatibility.
3081 	 */
3082 	if (!ext4_has_feature_extents(sb))
3083 		return;
3084 	/*
3085 	 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3086 	 * This should cover the situations where we can not afford to run
3087 	 * out of space like for example punch hole, or converting
3088 	 * unwritten extents in delalloc path. In most cases such
3089 	 * allocation would require 1, or 2 blocks, higher numbers are
3090 	 * very rare.
3091 	 */
3092 	resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3093 			 sbi->s_cluster_bits);
3094 
3095 	do_div(resv_clusters, 50);
3096 	resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3097 
3098 	atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3099 }
3100 
3101 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3102 {
3103 	char *orig_data = kstrdup(data, GFP_KERNEL);
3104 	struct buffer_head *bh;
3105 	struct ext4_super_block *es = NULL;
3106 	struct ext4_sb_info *sbi;
3107 	ext4_fsblk_t block;
3108 	ext4_fsblk_t sb_block = get_sb_block(&data);
3109 	ext4_fsblk_t logical_sb_block;
3110 	unsigned long offset = 0;
3111 	unsigned long journal_devnum = 0;
3112 	unsigned long def_mount_opts;
3113 	struct inode *root;
3114 	const char *descr;
3115 	int ret = -ENOMEM;
3116 	int blocksize, clustersize;
3117 	unsigned int db_count;
3118 	unsigned int i;
3119 	int needs_recovery, has_huge_files, has_bigalloc;
3120 	__u64 blocks_count;
3121 	int err = 0;
3122 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3123 	ext4_group_t first_not_zeroed;
3124 
3125 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3126 	if (!sbi)
3127 		goto out_free_orig;
3128 
3129 	sbi->s_blockgroup_lock =
3130 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3131 	if (!sbi->s_blockgroup_lock) {
3132 		kfree(sbi);
3133 		goto out_free_orig;
3134 	}
3135 	sb->s_fs_info = sbi;
3136 	sbi->s_sb = sb;
3137 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3138 	sbi->s_sb_block = sb_block;
3139 	if (sb->s_bdev->bd_part)
3140 		sbi->s_sectors_written_start =
3141 			part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3142 
3143 	/* Cleanup superblock name */
3144 	strreplace(sb->s_id, '/', '!');
3145 
3146 	/* -EINVAL is default */
3147 	ret = -EINVAL;
3148 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3149 	if (!blocksize) {
3150 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3151 		goto out_fail;
3152 	}
3153 
3154 	/*
3155 	 * The ext4 superblock will not be buffer aligned for other than 1kB
3156 	 * block sizes.  We need to calculate the offset from buffer start.
3157 	 */
3158 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3159 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3160 		offset = do_div(logical_sb_block, blocksize);
3161 	} else {
3162 		logical_sb_block = sb_block;
3163 	}
3164 
3165 	if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3166 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
3167 		goto out_fail;
3168 	}
3169 	/*
3170 	 * Note: s_es must be initialized as soon as possible because
3171 	 *       some ext4 macro-instructions depend on its value
3172 	 */
3173 	es = (struct ext4_super_block *) (bh->b_data + offset);
3174 	sbi->s_es = es;
3175 	sb->s_magic = le16_to_cpu(es->s_magic);
3176 	if (sb->s_magic != EXT4_SUPER_MAGIC)
3177 		goto cantfind_ext4;
3178 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3179 
3180 	/* Warn if metadata_csum and gdt_csum are both set. */
3181 	if (ext4_has_feature_metadata_csum(sb) &&
3182 	    ext4_has_feature_gdt_csum(sb))
3183 		ext4_warning(sb, "metadata_csum and uninit_bg are "
3184 			     "redundant flags; please run fsck.");
3185 
3186 	/* Check for a known checksum algorithm */
3187 	if (!ext4_verify_csum_type(sb, es)) {
3188 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3189 			 "unknown checksum algorithm.");
3190 		silent = 1;
3191 		goto cantfind_ext4;
3192 	}
3193 
3194 	/* Load the checksum driver */
3195 	if (ext4_has_feature_metadata_csum(sb)) {
3196 		sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3197 		if (IS_ERR(sbi->s_chksum_driver)) {
3198 			ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3199 			ret = PTR_ERR(sbi->s_chksum_driver);
3200 			sbi->s_chksum_driver = NULL;
3201 			goto failed_mount;
3202 		}
3203 	}
3204 
3205 	/* Check superblock checksum */
3206 	if (!ext4_superblock_csum_verify(sb, es)) {
3207 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3208 			 "invalid superblock checksum.  Run e2fsck?");
3209 		silent = 1;
3210 		ret = -EFSBADCRC;
3211 		goto cantfind_ext4;
3212 	}
3213 
3214 	/* Precompute checksum seed for all metadata */
3215 	if (ext4_has_feature_csum_seed(sb))
3216 		sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3217 	else if (ext4_has_metadata_csum(sb))
3218 		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3219 					       sizeof(es->s_uuid));
3220 
3221 	/* Set defaults before we parse the mount options */
3222 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3223 	set_opt(sb, INIT_INODE_TABLE);
3224 	if (def_mount_opts & EXT4_DEFM_DEBUG)
3225 		set_opt(sb, DEBUG);
3226 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3227 		set_opt(sb, GRPID);
3228 	if (def_mount_opts & EXT4_DEFM_UID16)
3229 		set_opt(sb, NO_UID32);
3230 	/* xattr user namespace & acls are now defaulted on */
3231 	set_opt(sb, XATTR_USER);
3232 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3233 	set_opt(sb, POSIX_ACL);
3234 #endif
3235 	/* don't forget to enable journal_csum when metadata_csum is enabled. */
3236 	if (ext4_has_metadata_csum(sb))
3237 		set_opt(sb, JOURNAL_CHECKSUM);
3238 
3239 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3240 		set_opt(sb, JOURNAL_DATA);
3241 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3242 		set_opt(sb, ORDERED_DATA);
3243 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3244 		set_opt(sb, WRITEBACK_DATA);
3245 
3246 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3247 		set_opt(sb, ERRORS_PANIC);
3248 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3249 		set_opt(sb, ERRORS_CONT);
3250 	else
3251 		set_opt(sb, ERRORS_RO);
3252 	/* block_validity enabled by default; disable with noblock_validity */
3253 	set_opt(sb, BLOCK_VALIDITY);
3254 	if (def_mount_opts & EXT4_DEFM_DISCARD)
3255 		set_opt(sb, DISCARD);
3256 
3257 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3258 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3259 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3260 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3261 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3262 
3263 	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3264 		set_opt(sb, BARRIER);
3265 
3266 	/*
3267 	 * enable delayed allocation by default
3268 	 * Use -o nodelalloc to turn it off
3269 	 */
3270 	if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3271 	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3272 		set_opt(sb, DELALLOC);
3273 
3274 	/*
3275 	 * set default s_li_wait_mult for lazyinit, for the case there is
3276 	 * no mount option specified.
3277 	 */
3278 	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3279 
3280 	if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3281 			   &journal_devnum, &journal_ioprio, 0)) {
3282 		ext4_msg(sb, KERN_WARNING,
3283 			 "failed to parse options in superblock: %s",
3284 			 sbi->s_es->s_mount_opts);
3285 	}
3286 	sbi->s_def_mount_opt = sbi->s_mount_opt;
3287 	if (!parse_options((char *) data, sb, &journal_devnum,
3288 			   &journal_ioprio, 0))
3289 		goto failed_mount;
3290 
3291 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3292 		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3293 			    "with data=journal disables delayed "
3294 			    "allocation and O_DIRECT support!\n");
3295 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3296 			ext4_msg(sb, KERN_ERR, "can't mount with "
3297 				 "both data=journal and delalloc");
3298 			goto failed_mount;
3299 		}
3300 		if (test_opt(sb, DIOREAD_NOLOCK)) {
3301 			ext4_msg(sb, KERN_ERR, "can't mount with "
3302 				 "both data=journal and dioread_nolock");
3303 			goto failed_mount;
3304 		}
3305 		if (test_opt(sb, DAX)) {
3306 			ext4_msg(sb, KERN_ERR, "can't mount with "
3307 				 "both data=journal and dax");
3308 			goto failed_mount;
3309 		}
3310 		if (test_opt(sb, DELALLOC))
3311 			clear_opt(sb, DELALLOC);
3312 	} else {
3313 		sb->s_iflags |= SB_I_CGROUPWB;
3314 	}
3315 
3316 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3317 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3318 
3319 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3320 	    (ext4_has_compat_features(sb) ||
3321 	     ext4_has_ro_compat_features(sb) ||
3322 	     ext4_has_incompat_features(sb)))
3323 		ext4_msg(sb, KERN_WARNING,
3324 		       "feature flags set on rev 0 fs, "
3325 		       "running e2fsck is recommended");
3326 
3327 	if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3328 		set_opt2(sb, HURD_COMPAT);
3329 		if (ext4_has_feature_64bit(sb)) {
3330 			ext4_msg(sb, KERN_ERR,
3331 				 "The Hurd can't support 64-bit file systems");
3332 			goto failed_mount;
3333 		}
3334 	}
3335 
3336 	if (IS_EXT2_SB(sb)) {
3337 		if (ext2_feature_set_ok(sb))
3338 			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3339 				 "using the ext4 subsystem");
3340 		else {
3341 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3342 				 "to feature incompatibilities");
3343 			goto failed_mount;
3344 		}
3345 	}
3346 
3347 	if (IS_EXT3_SB(sb)) {
3348 		if (ext3_feature_set_ok(sb))
3349 			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3350 				 "using the ext4 subsystem");
3351 		else {
3352 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3353 				 "to feature incompatibilities");
3354 			goto failed_mount;
3355 		}
3356 	}
3357 
3358 	/*
3359 	 * Check feature flags regardless of the revision level, since we
3360 	 * previously didn't change the revision level when setting the flags,
3361 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
3362 	 */
3363 	if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3364 		goto failed_mount;
3365 
3366 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3367 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3368 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
3369 		ext4_msg(sb, KERN_ERR,
3370 		       "Unsupported filesystem blocksize %d", blocksize);
3371 		goto failed_mount;
3372 	}
3373 
3374 	if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3375 		if (blocksize != PAGE_SIZE) {
3376 			ext4_msg(sb, KERN_ERR,
3377 					"error: unsupported blocksize for dax");
3378 			goto failed_mount;
3379 		}
3380 		if (!sb->s_bdev->bd_disk->fops->direct_access) {
3381 			ext4_msg(sb, KERN_ERR,
3382 					"error: device does not support dax");
3383 			goto failed_mount;
3384 		}
3385 	}
3386 
3387 	if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3388 		ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3389 			 es->s_encryption_level);
3390 		goto failed_mount;
3391 	}
3392 
3393 	if (sb->s_blocksize != blocksize) {
3394 		/* Validate the filesystem blocksize */
3395 		if (!sb_set_blocksize(sb, blocksize)) {
3396 			ext4_msg(sb, KERN_ERR, "bad block size %d",
3397 					blocksize);
3398 			goto failed_mount;
3399 		}
3400 
3401 		brelse(bh);
3402 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3403 		offset = do_div(logical_sb_block, blocksize);
3404 		bh = sb_bread_unmovable(sb, logical_sb_block);
3405 		if (!bh) {
3406 			ext4_msg(sb, KERN_ERR,
3407 			       "Can't read superblock on 2nd try");
3408 			goto failed_mount;
3409 		}
3410 		es = (struct ext4_super_block *)(bh->b_data + offset);
3411 		sbi->s_es = es;
3412 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3413 			ext4_msg(sb, KERN_ERR,
3414 			       "Magic mismatch, very weird!");
3415 			goto failed_mount;
3416 		}
3417 	}
3418 
3419 	has_huge_files = ext4_has_feature_huge_file(sb);
3420 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3421 						      has_huge_files);
3422 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3423 
3424 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3425 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3426 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3427 	} else {
3428 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3429 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3430 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3431 		    (!is_power_of_2(sbi->s_inode_size)) ||
3432 		    (sbi->s_inode_size > blocksize)) {
3433 			ext4_msg(sb, KERN_ERR,
3434 			       "unsupported inode size: %d",
3435 			       sbi->s_inode_size);
3436 			goto failed_mount;
3437 		}
3438 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3439 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3440 	}
3441 
3442 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3443 	if (ext4_has_feature_64bit(sb)) {
3444 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3445 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3446 		    !is_power_of_2(sbi->s_desc_size)) {
3447 			ext4_msg(sb, KERN_ERR,
3448 			       "unsupported descriptor size %lu",
3449 			       sbi->s_desc_size);
3450 			goto failed_mount;
3451 		}
3452 	} else
3453 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3454 
3455 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3456 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3457 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3458 		goto cantfind_ext4;
3459 
3460 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3461 	if (sbi->s_inodes_per_block == 0)
3462 		goto cantfind_ext4;
3463 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
3464 					sbi->s_inodes_per_block;
3465 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3466 	sbi->s_sbh = bh;
3467 	sbi->s_mount_state = le16_to_cpu(es->s_state);
3468 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3469 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3470 
3471 	for (i = 0; i < 4; i++)
3472 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3473 	sbi->s_def_hash_version = es->s_def_hash_version;
3474 	if (ext4_has_feature_dir_index(sb)) {
3475 		i = le32_to_cpu(es->s_flags);
3476 		if (i & EXT2_FLAGS_UNSIGNED_HASH)
3477 			sbi->s_hash_unsigned = 3;
3478 		else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3479 #ifdef __CHAR_UNSIGNED__
3480 			if (!(sb->s_flags & MS_RDONLY))
3481 				es->s_flags |=
3482 					cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3483 			sbi->s_hash_unsigned = 3;
3484 #else
3485 			if (!(sb->s_flags & MS_RDONLY))
3486 				es->s_flags |=
3487 					cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3488 #endif
3489 		}
3490 	}
3491 
3492 	/* Handle clustersize */
3493 	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3494 	has_bigalloc = ext4_has_feature_bigalloc(sb);
3495 	if (has_bigalloc) {
3496 		if (clustersize < blocksize) {
3497 			ext4_msg(sb, KERN_ERR,
3498 				 "cluster size (%d) smaller than "
3499 				 "block size (%d)", clustersize, blocksize);
3500 			goto failed_mount;
3501 		}
3502 		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3503 			le32_to_cpu(es->s_log_block_size);
3504 		sbi->s_clusters_per_group =
3505 			le32_to_cpu(es->s_clusters_per_group);
3506 		if (sbi->s_clusters_per_group > blocksize * 8) {
3507 			ext4_msg(sb, KERN_ERR,
3508 				 "#clusters per group too big: %lu",
3509 				 sbi->s_clusters_per_group);
3510 			goto failed_mount;
3511 		}
3512 		if (sbi->s_blocks_per_group !=
3513 		    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3514 			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3515 				 "clusters per group (%lu) inconsistent",
3516 				 sbi->s_blocks_per_group,
3517 				 sbi->s_clusters_per_group);
3518 			goto failed_mount;
3519 		}
3520 	} else {
3521 		if (clustersize != blocksize) {
3522 			ext4_warning(sb, "fragment/cluster size (%d) != "
3523 				     "block size (%d)", clustersize,
3524 				     blocksize);
3525 			clustersize = blocksize;
3526 		}
3527 		if (sbi->s_blocks_per_group > blocksize * 8) {
3528 			ext4_msg(sb, KERN_ERR,
3529 				 "#blocks per group too big: %lu",
3530 				 sbi->s_blocks_per_group);
3531 			goto failed_mount;
3532 		}
3533 		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3534 		sbi->s_cluster_bits = 0;
3535 	}
3536 	sbi->s_cluster_ratio = clustersize / blocksize;
3537 
3538 	if (sbi->s_inodes_per_group > blocksize * 8) {
3539 		ext4_msg(sb, KERN_ERR,
3540 		       "#inodes per group too big: %lu",
3541 		       sbi->s_inodes_per_group);
3542 		goto failed_mount;
3543 	}
3544 
3545 	/* Do we have standard group size of clustersize * 8 blocks ? */
3546 	if (sbi->s_blocks_per_group == clustersize << 3)
3547 		set_opt2(sb, STD_GROUP_SIZE);
3548 
3549 	/*
3550 	 * Test whether we have more sectors than will fit in sector_t,
3551 	 * and whether the max offset is addressable by the page cache.
3552 	 */
3553 	err = generic_check_addressable(sb->s_blocksize_bits,
3554 					ext4_blocks_count(es));
3555 	if (err) {
3556 		ext4_msg(sb, KERN_ERR, "filesystem"
3557 			 " too large to mount safely on this system");
3558 		if (sizeof(sector_t) < 8)
3559 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3560 		goto failed_mount;
3561 	}
3562 
3563 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3564 		goto cantfind_ext4;
3565 
3566 	/* check blocks count against device size */
3567 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3568 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3569 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3570 		       "exceeds size of device (%llu blocks)",
3571 		       ext4_blocks_count(es), blocks_count);
3572 		goto failed_mount;
3573 	}
3574 
3575 	/*
3576 	 * It makes no sense for the first data block to be beyond the end
3577 	 * of the filesystem.
3578 	 */
3579 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3580 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3581 			 "block %u is beyond end of filesystem (%llu)",
3582 			 le32_to_cpu(es->s_first_data_block),
3583 			 ext4_blocks_count(es));
3584 		goto failed_mount;
3585 	}
3586 	blocks_count = (ext4_blocks_count(es) -
3587 			le32_to_cpu(es->s_first_data_block) +
3588 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
3589 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3590 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3591 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3592 		       "(block count %llu, first data block %u, "
3593 		       "blocks per group %lu)", sbi->s_groups_count,
3594 		       ext4_blocks_count(es),
3595 		       le32_to_cpu(es->s_first_data_block),
3596 		       EXT4_BLOCKS_PER_GROUP(sb));
3597 		goto failed_mount;
3598 	}
3599 	sbi->s_groups_count = blocks_count;
3600 	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3601 			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3602 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3603 		   EXT4_DESC_PER_BLOCK(sb);
3604 	sbi->s_group_desc = ext4_kvmalloc(db_count *
3605 					  sizeof(struct buffer_head *),
3606 					  GFP_KERNEL);
3607 	if (sbi->s_group_desc == NULL) {
3608 		ext4_msg(sb, KERN_ERR, "not enough memory");
3609 		ret = -ENOMEM;
3610 		goto failed_mount;
3611 	}
3612 
3613 	bgl_lock_init(sbi->s_blockgroup_lock);
3614 
3615 	for (i = 0; i < db_count; i++) {
3616 		block = descriptor_loc(sb, logical_sb_block, i);
3617 		sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3618 		if (!sbi->s_group_desc[i]) {
3619 			ext4_msg(sb, KERN_ERR,
3620 			       "can't read group descriptor %d", i);
3621 			db_count = i;
3622 			goto failed_mount2;
3623 		}
3624 	}
3625 	if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3626 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3627 		ret = -EFSCORRUPTED;
3628 		goto failed_mount2;
3629 	}
3630 
3631 	sbi->s_gdb_count = db_count;
3632 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3633 	spin_lock_init(&sbi->s_next_gen_lock);
3634 
3635 	setup_timer(&sbi->s_err_report, print_daily_error_info,
3636 		(unsigned long) sb);
3637 
3638 	/* Register extent status tree shrinker */
3639 	if (ext4_es_register_shrinker(sbi))
3640 		goto failed_mount3;
3641 
3642 	sbi->s_stripe = ext4_get_stripe_size(sbi);
3643 	sbi->s_extent_max_zeroout_kb = 32;
3644 
3645 	/*
3646 	 * set up enough so that it can read an inode
3647 	 */
3648 	sb->s_op = &ext4_sops;
3649 	sb->s_export_op = &ext4_export_ops;
3650 	sb->s_xattr = ext4_xattr_handlers;
3651 #ifdef CONFIG_QUOTA
3652 	sb->dq_op = &ext4_quota_operations;
3653 	if (ext4_has_feature_quota(sb))
3654 		sb->s_qcop = &dquot_quotactl_sysfile_ops;
3655 	else
3656 		sb->s_qcop = &ext4_qctl_operations;
3657 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3658 #endif
3659 	memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3660 
3661 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3662 	mutex_init(&sbi->s_orphan_lock);
3663 
3664 	sb->s_root = NULL;
3665 
3666 	needs_recovery = (es->s_last_orphan != 0 ||
3667 			  ext4_has_feature_journal_needs_recovery(sb));
3668 
3669 	if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3670 		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3671 			goto failed_mount3a;
3672 
3673 	/*
3674 	 * The first inode we look at is the journal inode.  Don't try
3675 	 * root first: it may be modified in the journal!
3676 	 */
3677 	if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3678 		if (ext4_load_journal(sb, es, journal_devnum))
3679 			goto failed_mount3a;
3680 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3681 		   ext4_has_feature_journal_needs_recovery(sb)) {
3682 		ext4_msg(sb, KERN_ERR, "required journal recovery "
3683 		       "suppressed and not mounted read-only");
3684 		goto failed_mount_wq;
3685 	} else {
3686 		/* Nojournal mode, all journal mount options are illegal */
3687 		if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3688 			ext4_msg(sb, KERN_ERR, "can't mount with "
3689 				 "journal_checksum, fs mounted w/o journal");
3690 			goto failed_mount_wq;
3691 		}
3692 		if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3693 			ext4_msg(sb, KERN_ERR, "can't mount with "
3694 				 "journal_async_commit, fs mounted w/o journal");
3695 			goto failed_mount_wq;
3696 		}
3697 		if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3698 			ext4_msg(sb, KERN_ERR, "can't mount with "
3699 				 "commit=%lu, fs mounted w/o journal",
3700 				 sbi->s_commit_interval / HZ);
3701 			goto failed_mount_wq;
3702 		}
3703 		if (EXT4_MOUNT_DATA_FLAGS &
3704 		    (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3705 			ext4_msg(sb, KERN_ERR, "can't mount with "
3706 				 "data=, fs mounted w/o journal");
3707 			goto failed_mount_wq;
3708 		}
3709 		sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3710 		clear_opt(sb, JOURNAL_CHECKSUM);
3711 		clear_opt(sb, DATA_FLAGS);
3712 		sbi->s_journal = NULL;
3713 		needs_recovery = 0;
3714 		goto no_journal;
3715 	}
3716 
3717 	if (ext4_has_feature_64bit(sb) &&
3718 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3719 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
3720 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3721 		goto failed_mount_wq;
3722 	}
3723 
3724 	if (!set_journal_csum_feature_set(sb)) {
3725 		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3726 			 "feature set");
3727 		goto failed_mount_wq;
3728 	}
3729 
3730 	/* We have now updated the journal if required, so we can
3731 	 * validate the data journaling mode. */
3732 	switch (test_opt(sb, DATA_FLAGS)) {
3733 	case 0:
3734 		/* No mode set, assume a default based on the journal
3735 		 * capabilities: ORDERED_DATA if the journal can
3736 		 * cope, else JOURNAL_DATA
3737 		 */
3738 		if (jbd2_journal_check_available_features
3739 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3740 			set_opt(sb, ORDERED_DATA);
3741 		else
3742 			set_opt(sb, JOURNAL_DATA);
3743 		break;
3744 
3745 	case EXT4_MOUNT_ORDERED_DATA:
3746 	case EXT4_MOUNT_WRITEBACK_DATA:
3747 		if (!jbd2_journal_check_available_features
3748 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3749 			ext4_msg(sb, KERN_ERR, "Journal does not support "
3750 			       "requested data journaling mode");
3751 			goto failed_mount_wq;
3752 		}
3753 	default:
3754 		break;
3755 	}
3756 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3757 
3758 	sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3759 
3760 no_journal:
3761 	if (ext4_mballoc_ready) {
3762 		sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
3763 		if (!sbi->s_mb_cache) {
3764 			ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3765 			goto failed_mount_wq;
3766 		}
3767 	}
3768 
3769 	if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3770 	    (blocksize != PAGE_CACHE_SIZE)) {
3771 		ext4_msg(sb, KERN_ERR,
3772 			 "Unsupported blocksize for fs encryption");
3773 		goto failed_mount_wq;
3774 	}
3775 
3776 	if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3777 	    !ext4_has_feature_encrypt(sb)) {
3778 		ext4_set_feature_encrypt(sb);
3779 		ext4_commit_super(sb, 1);
3780 	}
3781 
3782 	/*
3783 	 * Get the # of file system overhead blocks from the
3784 	 * superblock if present.
3785 	 */
3786 	if (es->s_overhead_clusters)
3787 		sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3788 	else {
3789 		err = ext4_calculate_overhead(sb);
3790 		if (err)
3791 			goto failed_mount_wq;
3792 	}
3793 
3794 	/*
3795 	 * The maximum number of concurrent works can be high and
3796 	 * concurrency isn't really necessary.  Limit it to 1.
3797 	 */
3798 	EXT4_SB(sb)->rsv_conversion_wq =
3799 		alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3800 	if (!EXT4_SB(sb)->rsv_conversion_wq) {
3801 		printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3802 		ret = -ENOMEM;
3803 		goto failed_mount4;
3804 	}
3805 
3806 	/*
3807 	 * The jbd2_journal_load will have done any necessary log recovery,
3808 	 * so we can safely mount the rest of the filesystem now.
3809 	 */
3810 
3811 	root = ext4_iget(sb, EXT4_ROOT_INO);
3812 	if (IS_ERR(root)) {
3813 		ext4_msg(sb, KERN_ERR, "get root inode failed");
3814 		ret = PTR_ERR(root);
3815 		root = NULL;
3816 		goto failed_mount4;
3817 	}
3818 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3819 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3820 		iput(root);
3821 		goto failed_mount4;
3822 	}
3823 	sb->s_root = d_make_root(root);
3824 	if (!sb->s_root) {
3825 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
3826 		ret = -ENOMEM;
3827 		goto failed_mount4;
3828 	}
3829 
3830 	if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3831 		sb->s_flags |= MS_RDONLY;
3832 
3833 	/* determine the minimum size of new large inodes, if present */
3834 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3835 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3836 						     EXT4_GOOD_OLD_INODE_SIZE;
3837 		if (ext4_has_feature_extra_isize(sb)) {
3838 			if (sbi->s_want_extra_isize <
3839 			    le16_to_cpu(es->s_want_extra_isize))
3840 				sbi->s_want_extra_isize =
3841 					le16_to_cpu(es->s_want_extra_isize);
3842 			if (sbi->s_want_extra_isize <
3843 			    le16_to_cpu(es->s_min_extra_isize))
3844 				sbi->s_want_extra_isize =
3845 					le16_to_cpu(es->s_min_extra_isize);
3846 		}
3847 	}
3848 	/* Check if enough inode space is available */
3849 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3850 							sbi->s_inode_size) {
3851 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3852 						       EXT4_GOOD_OLD_INODE_SIZE;
3853 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
3854 			 "available");
3855 	}
3856 
3857 	ext4_set_resv_clusters(sb);
3858 
3859 	err = ext4_setup_system_zone(sb);
3860 	if (err) {
3861 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
3862 			 "zone (%d)", err);
3863 		goto failed_mount4a;
3864 	}
3865 
3866 	ext4_ext_init(sb);
3867 	err = ext4_mb_init(sb);
3868 	if (err) {
3869 		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3870 			 err);
3871 		goto failed_mount5;
3872 	}
3873 
3874 	block = ext4_count_free_clusters(sb);
3875 	ext4_free_blocks_count_set(sbi->s_es,
3876 				   EXT4_C2B(sbi, block));
3877 	err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3878 				  GFP_KERNEL);
3879 	if (!err) {
3880 		unsigned long freei = ext4_count_free_inodes(sb);
3881 		sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
3882 		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3883 					  GFP_KERNEL);
3884 	}
3885 	if (!err)
3886 		err = percpu_counter_init(&sbi->s_dirs_counter,
3887 					  ext4_count_dirs(sb), GFP_KERNEL);
3888 	if (!err)
3889 		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3890 					  GFP_KERNEL);
3891 	if (err) {
3892 		ext4_msg(sb, KERN_ERR, "insufficient memory");
3893 		goto failed_mount6;
3894 	}
3895 
3896 	if (ext4_has_feature_flex_bg(sb))
3897 		if (!ext4_fill_flex_info(sb)) {
3898 			ext4_msg(sb, KERN_ERR,
3899 			       "unable to initialize "
3900 			       "flex_bg meta info!");
3901 			goto failed_mount6;
3902 		}
3903 
3904 	err = ext4_register_li_request(sb, first_not_zeroed);
3905 	if (err)
3906 		goto failed_mount6;
3907 
3908 	err = ext4_register_sysfs(sb);
3909 	if (err)
3910 		goto failed_mount7;
3911 
3912 #ifdef CONFIG_QUOTA
3913 	/* Enable quota usage during mount. */
3914 	if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
3915 		err = ext4_enable_quotas(sb);
3916 		if (err)
3917 			goto failed_mount8;
3918 	}
3919 #endif  /* CONFIG_QUOTA */
3920 
3921 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3922 	ext4_orphan_cleanup(sb, es);
3923 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3924 	if (needs_recovery) {
3925 		ext4_msg(sb, KERN_INFO, "recovery complete");
3926 		ext4_mark_recovery_complete(sb, es);
3927 	}
3928 	if (EXT4_SB(sb)->s_journal) {
3929 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3930 			descr = " journalled data mode";
3931 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3932 			descr = " ordered data mode";
3933 		else
3934 			descr = " writeback data mode";
3935 	} else
3936 		descr = "out journal";
3937 
3938 	if (test_opt(sb, DISCARD)) {
3939 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
3940 		if (!blk_queue_discard(q))
3941 			ext4_msg(sb, KERN_WARNING,
3942 				 "mounting with \"discard\" option, but "
3943 				 "the device does not support discard");
3944 	}
3945 
3946 	if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3947 		ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3948 			 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3949 			 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3950 
3951 	if (es->s_error_count)
3952 		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3953 
3954 	/* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3955 	ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
3956 	ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
3957 	ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
3958 
3959 	kfree(orig_data);
3960 	return 0;
3961 
3962 cantfind_ext4:
3963 	if (!silent)
3964 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3965 	goto failed_mount;
3966 
3967 #ifdef CONFIG_QUOTA
3968 failed_mount8:
3969 	ext4_unregister_sysfs(sb);
3970 #endif
3971 failed_mount7:
3972 	ext4_unregister_li_request(sb);
3973 failed_mount6:
3974 	ext4_mb_release(sb);
3975 	if (sbi->s_flex_groups)
3976 		kvfree(sbi->s_flex_groups);
3977 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
3978 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
3979 	percpu_counter_destroy(&sbi->s_dirs_counter);
3980 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3981 failed_mount5:
3982 	ext4_ext_release(sb);
3983 	ext4_release_system_zone(sb);
3984 failed_mount4a:
3985 	dput(sb->s_root);
3986 	sb->s_root = NULL;
3987 failed_mount4:
3988 	ext4_msg(sb, KERN_ERR, "mount failed");
3989 	if (EXT4_SB(sb)->rsv_conversion_wq)
3990 		destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
3991 failed_mount_wq:
3992 	if (sbi->s_journal) {
3993 		jbd2_journal_destroy(sbi->s_journal);
3994 		sbi->s_journal = NULL;
3995 	}
3996 failed_mount3a:
3997 	ext4_es_unregister_shrinker(sbi);
3998 failed_mount3:
3999 	del_timer_sync(&sbi->s_err_report);
4000 	if (sbi->s_mmp_tsk)
4001 		kthread_stop(sbi->s_mmp_tsk);
4002 failed_mount2:
4003 	for (i = 0; i < db_count; i++)
4004 		brelse(sbi->s_group_desc[i]);
4005 	kvfree(sbi->s_group_desc);
4006 failed_mount:
4007 	if (sbi->s_chksum_driver)
4008 		crypto_free_shash(sbi->s_chksum_driver);
4009 #ifdef CONFIG_QUOTA
4010 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4011 		kfree(sbi->s_qf_names[i]);
4012 #endif
4013 	ext4_blkdev_remove(sbi);
4014 	brelse(bh);
4015 out_fail:
4016 	sb->s_fs_info = NULL;
4017 	kfree(sbi->s_blockgroup_lock);
4018 	kfree(sbi);
4019 out_free_orig:
4020 	kfree(orig_data);
4021 	return err ? err : ret;
4022 }
4023 
4024 /*
4025  * Setup any per-fs journal parameters now.  We'll do this both on
4026  * initial mount, once the journal has been initialised but before we've
4027  * done any recovery; and again on any subsequent remount.
4028  */
4029 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4030 {
4031 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4032 
4033 	journal->j_commit_interval = sbi->s_commit_interval;
4034 	journal->j_min_batch_time = sbi->s_min_batch_time;
4035 	journal->j_max_batch_time = sbi->s_max_batch_time;
4036 
4037 	write_lock(&journal->j_state_lock);
4038 	if (test_opt(sb, BARRIER))
4039 		journal->j_flags |= JBD2_BARRIER;
4040 	else
4041 		journal->j_flags &= ~JBD2_BARRIER;
4042 	if (test_opt(sb, DATA_ERR_ABORT))
4043 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4044 	else
4045 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4046 	write_unlock(&journal->j_state_lock);
4047 }
4048 
4049 static journal_t *ext4_get_journal(struct super_block *sb,
4050 				   unsigned int journal_inum)
4051 {
4052 	struct inode *journal_inode;
4053 	journal_t *journal;
4054 
4055 	BUG_ON(!ext4_has_feature_journal(sb));
4056 
4057 	/* First, test for the existence of a valid inode on disk.  Bad
4058 	 * things happen if we iget() an unused inode, as the subsequent
4059 	 * iput() will try to delete it. */
4060 
4061 	journal_inode = ext4_iget(sb, journal_inum);
4062 	if (IS_ERR(journal_inode)) {
4063 		ext4_msg(sb, KERN_ERR, "no journal found");
4064 		return NULL;
4065 	}
4066 	if (!journal_inode->i_nlink) {
4067 		make_bad_inode(journal_inode);
4068 		iput(journal_inode);
4069 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4070 		return NULL;
4071 	}
4072 
4073 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4074 		  journal_inode, journal_inode->i_size);
4075 	if (!S_ISREG(journal_inode->i_mode)) {
4076 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
4077 		iput(journal_inode);
4078 		return NULL;
4079 	}
4080 
4081 	journal = jbd2_journal_init_inode(journal_inode);
4082 	if (!journal) {
4083 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4084 		iput(journal_inode);
4085 		return NULL;
4086 	}
4087 	journal->j_private = sb;
4088 	ext4_init_journal_params(sb, journal);
4089 	return journal;
4090 }
4091 
4092 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4093 				       dev_t j_dev)
4094 {
4095 	struct buffer_head *bh;
4096 	journal_t *journal;
4097 	ext4_fsblk_t start;
4098 	ext4_fsblk_t len;
4099 	int hblock, blocksize;
4100 	ext4_fsblk_t sb_block;
4101 	unsigned long offset;
4102 	struct ext4_super_block *es;
4103 	struct block_device *bdev;
4104 
4105 	BUG_ON(!ext4_has_feature_journal(sb));
4106 
4107 	bdev = ext4_blkdev_get(j_dev, sb);
4108 	if (bdev == NULL)
4109 		return NULL;
4110 
4111 	blocksize = sb->s_blocksize;
4112 	hblock = bdev_logical_block_size(bdev);
4113 	if (blocksize < hblock) {
4114 		ext4_msg(sb, KERN_ERR,
4115 			"blocksize too small for journal device");
4116 		goto out_bdev;
4117 	}
4118 
4119 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4120 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4121 	set_blocksize(bdev, blocksize);
4122 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
4123 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4124 		       "external journal");
4125 		goto out_bdev;
4126 	}
4127 
4128 	es = (struct ext4_super_block *) (bh->b_data + offset);
4129 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4130 	    !(le32_to_cpu(es->s_feature_incompat) &
4131 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4132 		ext4_msg(sb, KERN_ERR, "external journal has "
4133 					"bad superblock");
4134 		brelse(bh);
4135 		goto out_bdev;
4136 	}
4137 
4138 	if ((le32_to_cpu(es->s_feature_ro_compat) &
4139 	     EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4140 	    es->s_checksum != ext4_superblock_csum(sb, es)) {
4141 		ext4_msg(sb, KERN_ERR, "external journal has "
4142 				       "corrupt superblock");
4143 		brelse(bh);
4144 		goto out_bdev;
4145 	}
4146 
4147 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4148 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4149 		brelse(bh);
4150 		goto out_bdev;
4151 	}
4152 
4153 	len = ext4_blocks_count(es);
4154 	start = sb_block + 1;
4155 	brelse(bh);	/* we're done with the superblock */
4156 
4157 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4158 					start, len, blocksize);
4159 	if (!journal) {
4160 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
4161 		goto out_bdev;
4162 	}
4163 	journal->j_private = sb;
4164 	ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4165 	wait_on_buffer(journal->j_sb_buffer);
4166 	if (!buffer_uptodate(journal->j_sb_buffer)) {
4167 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4168 		goto out_journal;
4169 	}
4170 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4171 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
4172 					"user (unsupported) - %d",
4173 			be32_to_cpu(journal->j_superblock->s_nr_users));
4174 		goto out_journal;
4175 	}
4176 	EXT4_SB(sb)->journal_bdev = bdev;
4177 	ext4_init_journal_params(sb, journal);
4178 	return journal;
4179 
4180 out_journal:
4181 	jbd2_journal_destroy(journal);
4182 out_bdev:
4183 	ext4_blkdev_put(bdev);
4184 	return NULL;
4185 }
4186 
4187 static int ext4_load_journal(struct super_block *sb,
4188 			     struct ext4_super_block *es,
4189 			     unsigned long journal_devnum)
4190 {
4191 	journal_t *journal;
4192 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4193 	dev_t journal_dev;
4194 	int err = 0;
4195 	int really_read_only;
4196 
4197 	BUG_ON(!ext4_has_feature_journal(sb));
4198 
4199 	if (journal_devnum &&
4200 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4201 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4202 			"numbers have changed");
4203 		journal_dev = new_decode_dev(journal_devnum);
4204 	} else
4205 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4206 
4207 	really_read_only = bdev_read_only(sb->s_bdev);
4208 
4209 	/*
4210 	 * Are we loading a blank journal or performing recovery after a
4211 	 * crash?  For recovery, we need to check in advance whether we
4212 	 * can get read-write access to the device.
4213 	 */
4214 	if (ext4_has_feature_journal_needs_recovery(sb)) {
4215 		if (sb->s_flags & MS_RDONLY) {
4216 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
4217 					"required on readonly filesystem");
4218 			if (really_read_only) {
4219 				ext4_msg(sb, KERN_ERR, "write access "
4220 					"unavailable, cannot proceed");
4221 				return -EROFS;
4222 			}
4223 			ext4_msg(sb, KERN_INFO, "write access will "
4224 			       "be enabled during recovery");
4225 		}
4226 	}
4227 
4228 	if (journal_inum && journal_dev) {
4229 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4230 		       "and inode journals!");
4231 		return -EINVAL;
4232 	}
4233 
4234 	if (journal_inum) {
4235 		if (!(journal = ext4_get_journal(sb, journal_inum)))
4236 			return -EINVAL;
4237 	} else {
4238 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4239 			return -EINVAL;
4240 	}
4241 
4242 	if (!(journal->j_flags & JBD2_BARRIER))
4243 		ext4_msg(sb, KERN_INFO, "barriers disabled");
4244 
4245 	if (!ext4_has_feature_journal_needs_recovery(sb))
4246 		err = jbd2_journal_wipe(journal, !really_read_only);
4247 	if (!err) {
4248 		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4249 		if (save)
4250 			memcpy(save, ((char *) es) +
4251 			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4252 		err = jbd2_journal_load(journal);
4253 		if (save)
4254 			memcpy(((char *) es) + EXT4_S_ERR_START,
4255 			       save, EXT4_S_ERR_LEN);
4256 		kfree(save);
4257 	}
4258 
4259 	if (err) {
4260 		ext4_msg(sb, KERN_ERR, "error loading journal");
4261 		jbd2_journal_destroy(journal);
4262 		return err;
4263 	}
4264 
4265 	EXT4_SB(sb)->s_journal = journal;
4266 	ext4_clear_journal_err(sb, es);
4267 
4268 	if (!really_read_only && journal_devnum &&
4269 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4270 		es->s_journal_dev = cpu_to_le32(journal_devnum);
4271 
4272 		/* Make sure we flush the recovery flag to disk. */
4273 		ext4_commit_super(sb, 1);
4274 	}
4275 
4276 	return 0;
4277 }
4278 
4279 static int ext4_commit_super(struct super_block *sb, int sync)
4280 {
4281 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4282 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4283 	int error = 0;
4284 
4285 	if (!sbh || block_device_ejected(sb))
4286 		return error;
4287 	if (buffer_write_io_error(sbh)) {
4288 		/*
4289 		 * Oh, dear.  A previous attempt to write the
4290 		 * superblock failed.  This could happen because the
4291 		 * USB device was yanked out.  Or it could happen to
4292 		 * be a transient write error and maybe the block will
4293 		 * be remapped.  Nothing we can do but to retry the
4294 		 * write and hope for the best.
4295 		 */
4296 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
4297 		       "superblock detected");
4298 		clear_buffer_write_io_error(sbh);
4299 		set_buffer_uptodate(sbh);
4300 	}
4301 	/*
4302 	 * If the file system is mounted read-only, don't update the
4303 	 * superblock write time.  This avoids updating the superblock
4304 	 * write time when we are mounting the root file system
4305 	 * read/only but we need to replay the journal; at that point,
4306 	 * for people who are east of GMT and who make their clock
4307 	 * tick in localtime for Windows bug-for-bug compatibility,
4308 	 * the clock is set in the future, and this will cause e2fsck
4309 	 * to complain and force a full file system check.
4310 	 */
4311 	if (!(sb->s_flags & MS_RDONLY))
4312 		es->s_wtime = cpu_to_le32(get_seconds());
4313 	if (sb->s_bdev->bd_part)
4314 		es->s_kbytes_written =
4315 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4316 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4317 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
4318 	else
4319 		es->s_kbytes_written =
4320 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4321 	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4322 		ext4_free_blocks_count_set(es,
4323 			EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4324 				&EXT4_SB(sb)->s_freeclusters_counter)));
4325 	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4326 		es->s_free_inodes_count =
4327 			cpu_to_le32(percpu_counter_sum_positive(
4328 				&EXT4_SB(sb)->s_freeinodes_counter));
4329 	BUFFER_TRACE(sbh, "marking dirty");
4330 	ext4_superblock_csum_set(sb);
4331 	mark_buffer_dirty(sbh);
4332 	if (sync) {
4333 		error = __sync_dirty_buffer(sbh,
4334 			test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4335 		if (error)
4336 			return error;
4337 
4338 		error = buffer_write_io_error(sbh);
4339 		if (error) {
4340 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
4341 			       "superblock");
4342 			clear_buffer_write_io_error(sbh);
4343 			set_buffer_uptodate(sbh);
4344 		}
4345 	}
4346 	return error;
4347 }
4348 
4349 /*
4350  * Have we just finished recovery?  If so, and if we are mounting (or
4351  * remounting) the filesystem readonly, then we will end up with a
4352  * consistent fs on disk.  Record that fact.
4353  */
4354 static void ext4_mark_recovery_complete(struct super_block *sb,
4355 					struct ext4_super_block *es)
4356 {
4357 	journal_t *journal = EXT4_SB(sb)->s_journal;
4358 
4359 	if (!ext4_has_feature_journal(sb)) {
4360 		BUG_ON(journal != NULL);
4361 		return;
4362 	}
4363 	jbd2_journal_lock_updates(journal);
4364 	if (jbd2_journal_flush(journal) < 0)
4365 		goto out;
4366 
4367 	if (ext4_has_feature_journal_needs_recovery(sb) &&
4368 	    sb->s_flags & MS_RDONLY) {
4369 		ext4_clear_feature_journal_needs_recovery(sb);
4370 		ext4_commit_super(sb, 1);
4371 	}
4372 
4373 out:
4374 	jbd2_journal_unlock_updates(journal);
4375 }
4376 
4377 /*
4378  * If we are mounting (or read-write remounting) a filesystem whose journal
4379  * has recorded an error from a previous lifetime, move that error to the
4380  * main filesystem now.
4381  */
4382 static void ext4_clear_journal_err(struct super_block *sb,
4383 				   struct ext4_super_block *es)
4384 {
4385 	journal_t *journal;
4386 	int j_errno;
4387 	const char *errstr;
4388 
4389 	BUG_ON(!ext4_has_feature_journal(sb));
4390 
4391 	journal = EXT4_SB(sb)->s_journal;
4392 
4393 	/*
4394 	 * Now check for any error status which may have been recorded in the
4395 	 * journal by a prior ext4_error() or ext4_abort()
4396 	 */
4397 
4398 	j_errno = jbd2_journal_errno(journal);
4399 	if (j_errno) {
4400 		char nbuf[16];
4401 
4402 		errstr = ext4_decode_error(sb, j_errno, nbuf);
4403 		ext4_warning(sb, "Filesystem error recorded "
4404 			     "from previous mount: %s", errstr);
4405 		ext4_warning(sb, "Marking fs in need of filesystem check.");
4406 
4407 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4408 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4409 		ext4_commit_super(sb, 1);
4410 
4411 		jbd2_journal_clear_err(journal);
4412 		jbd2_journal_update_sb_errno(journal);
4413 	}
4414 }
4415 
4416 /*
4417  * Force the running and committing transactions to commit,
4418  * and wait on the commit.
4419  */
4420 int ext4_force_commit(struct super_block *sb)
4421 {
4422 	journal_t *journal;
4423 
4424 	if (sb->s_flags & MS_RDONLY)
4425 		return 0;
4426 
4427 	journal = EXT4_SB(sb)->s_journal;
4428 	return ext4_journal_force_commit(journal);
4429 }
4430 
4431 static int ext4_sync_fs(struct super_block *sb, int wait)
4432 {
4433 	int ret = 0;
4434 	tid_t target;
4435 	bool needs_barrier = false;
4436 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4437 
4438 	trace_ext4_sync_fs(sb, wait);
4439 	flush_workqueue(sbi->rsv_conversion_wq);
4440 	/*
4441 	 * Writeback quota in non-journalled quota case - journalled quota has
4442 	 * no dirty dquots
4443 	 */
4444 	dquot_writeback_dquots(sb, -1);
4445 	/*
4446 	 * Data writeback is possible w/o journal transaction, so barrier must
4447 	 * being sent at the end of the function. But we can skip it if
4448 	 * transaction_commit will do it for us.
4449 	 */
4450 	if (sbi->s_journal) {
4451 		target = jbd2_get_latest_transaction(sbi->s_journal);
4452 		if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4453 		    !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4454 			needs_barrier = true;
4455 
4456 		if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4457 			if (wait)
4458 				ret = jbd2_log_wait_commit(sbi->s_journal,
4459 							   target);
4460 		}
4461 	} else if (wait && test_opt(sb, BARRIER))
4462 		needs_barrier = true;
4463 	if (needs_barrier) {
4464 		int err;
4465 		err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4466 		if (!ret)
4467 			ret = err;
4468 	}
4469 
4470 	return ret;
4471 }
4472 
4473 /*
4474  * LVM calls this function before a (read-only) snapshot is created.  This
4475  * gives us a chance to flush the journal completely and mark the fs clean.
4476  *
4477  * Note that only this function cannot bring a filesystem to be in a clean
4478  * state independently. It relies on upper layer to stop all data & metadata
4479  * modifications.
4480  */
4481 static int ext4_freeze(struct super_block *sb)
4482 {
4483 	int error = 0;
4484 	journal_t *journal;
4485 
4486 	if (sb->s_flags & MS_RDONLY)
4487 		return 0;
4488 
4489 	journal = EXT4_SB(sb)->s_journal;
4490 
4491 	if (journal) {
4492 		/* Now we set up the journal barrier. */
4493 		jbd2_journal_lock_updates(journal);
4494 
4495 		/*
4496 		 * Don't clear the needs_recovery flag if we failed to
4497 		 * flush the journal.
4498 		 */
4499 		error = jbd2_journal_flush(journal);
4500 		if (error < 0)
4501 			goto out;
4502 
4503 		/* Journal blocked and flushed, clear needs_recovery flag. */
4504 		ext4_clear_feature_journal_needs_recovery(sb);
4505 	}
4506 
4507 	error = ext4_commit_super(sb, 1);
4508 out:
4509 	if (journal)
4510 		/* we rely on upper layer to stop further updates */
4511 		jbd2_journal_unlock_updates(journal);
4512 	return error;
4513 }
4514 
4515 /*
4516  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4517  * flag here, even though the filesystem is not technically dirty yet.
4518  */
4519 static int ext4_unfreeze(struct super_block *sb)
4520 {
4521 	if (sb->s_flags & MS_RDONLY)
4522 		return 0;
4523 
4524 	if (EXT4_SB(sb)->s_journal) {
4525 		/* Reset the needs_recovery flag before the fs is unlocked. */
4526 		ext4_set_feature_journal_needs_recovery(sb);
4527 	}
4528 
4529 	ext4_commit_super(sb, 1);
4530 	return 0;
4531 }
4532 
4533 /*
4534  * Structure to save mount options for ext4_remount's benefit
4535  */
4536 struct ext4_mount_options {
4537 	unsigned long s_mount_opt;
4538 	unsigned long s_mount_opt2;
4539 	kuid_t s_resuid;
4540 	kgid_t s_resgid;
4541 	unsigned long s_commit_interval;
4542 	u32 s_min_batch_time, s_max_batch_time;
4543 #ifdef CONFIG_QUOTA
4544 	int s_jquota_fmt;
4545 	char *s_qf_names[EXT4_MAXQUOTAS];
4546 #endif
4547 };
4548 
4549 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4550 {
4551 	struct ext4_super_block *es;
4552 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4553 	unsigned long old_sb_flags;
4554 	struct ext4_mount_options old_opts;
4555 	int enable_quota = 0;
4556 	ext4_group_t g;
4557 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4558 	int err = 0;
4559 #ifdef CONFIG_QUOTA
4560 	int i, j;
4561 #endif
4562 	char *orig_data = kstrdup(data, GFP_KERNEL);
4563 
4564 	/* Store the original options */
4565 	old_sb_flags = sb->s_flags;
4566 	old_opts.s_mount_opt = sbi->s_mount_opt;
4567 	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4568 	old_opts.s_resuid = sbi->s_resuid;
4569 	old_opts.s_resgid = sbi->s_resgid;
4570 	old_opts.s_commit_interval = sbi->s_commit_interval;
4571 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
4572 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
4573 #ifdef CONFIG_QUOTA
4574 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4575 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4576 		if (sbi->s_qf_names[i]) {
4577 			old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4578 							 GFP_KERNEL);
4579 			if (!old_opts.s_qf_names[i]) {
4580 				for (j = 0; j < i; j++)
4581 					kfree(old_opts.s_qf_names[j]);
4582 				kfree(orig_data);
4583 				return -ENOMEM;
4584 			}
4585 		} else
4586 			old_opts.s_qf_names[i] = NULL;
4587 #endif
4588 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4589 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4590 
4591 	if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4592 		err = -EINVAL;
4593 		goto restore_opts;
4594 	}
4595 
4596 	if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4597 	    test_opt(sb, JOURNAL_CHECKSUM)) {
4598 		ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4599 			 "during remount not supported; ignoring");
4600 		sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4601 	}
4602 
4603 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4604 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4605 			ext4_msg(sb, KERN_ERR, "can't mount with "
4606 				 "both data=journal and delalloc");
4607 			err = -EINVAL;
4608 			goto restore_opts;
4609 		}
4610 		if (test_opt(sb, DIOREAD_NOLOCK)) {
4611 			ext4_msg(sb, KERN_ERR, "can't mount with "
4612 				 "both data=journal and dioread_nolock");
4613 			err = -EINVAL;
4614 			goto restore_opts;
4615 		}
4616 		if (test_opt(sb, DAX)) {
4617 			ext4_msg(sb, KERN_ERR, "can't mount with "
4618 				 "both data=journal and dax");
4619 			err = -EINVAL;
4620 			goto restore_opts;
4621 		}
4622 	}
4623 
4624 	if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4625 		ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4626 			"dax flag with busy inodes while remounting");
4627 		sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4628 	}
4629 
4630 	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4631 		ext4_abort(sb, "Abort forced by user");
4632 
4633 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4634 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4635 
4636 	es = sbi->s_es;
4637 
4638 	if (sbi->s_journal) {
4639 		ext4_init_journal_params(sb, sbi->s_journal);
4640 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4641 	}
4642 
4643 	if (*flags & MS_LAZYTIME)
4644 		sb->s_flags |= MS_LAZYTIME;
4645 
4646 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4647 		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4648 			err = -EROFS;
4649 			goto restore_opts;
4650 		}
4651 
4652 		if (*flags & MS_RDONLY) {
4653 			err = sync_filesystem(sb);
4654 			if (err < 0)
4655 				goto restore_opts;
4656 			err = dquot_suspend(sb, -1);
4657 			if (err < 0)
4658 				goto restore_opts;
4659 
4660 			/*
4661 			 * First of all, the unconditional stuff we have to do
4662 			 * to disable replay of the journal when we next remount
4663 			 */
4664 			sb->s_flags |= MS_RDONLY;
4665 
4666 			/*
4667 			 * OK, test if we are remounting a valid rw partition
4668 			 * readonly, and if so set the rdonly flag and then
4669 			 * mark the partition as valid again.
4670 			 */
4671 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4672 			    (sbi->s_mount_state & EXT4_VALID_FS))
4673 				es->s_state = cpu_to_le16(sbi->s_mount_state);
4674 
4675 			if (sbi->s_journal)
4676 				ext4_mark_recovery_complete(sb, es);
4677 		} else {
4678 			/* Make sure we can mount this feature set readwrite */
4679 			if (ext4_has_feature_readonly(sb) ||
4680 			    !ext4_feature_set_ok(sb, 0)) {
4681 				err = -EROFS;
4682 				goto restore_opts;
4683 			}
4684 			/*
4685 			 * Make sure the group descriptor checksums
4686 			 * are sane.  If they aren't, refuse to remount r/w.
4687 			 */
4688 			for (g = 0; g < sbi->s_groups_count; g++) {
4689 				struct ext4_group_desc *gdp =
4690 					ext4_get_group_desc(sb, g, NULL);
4691 
4692 				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4693 					ext4_msg(sb, KERN_ERR,
4694 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
4695 		g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4696 					       le16_to_cpu(gdp->bg_checksum));
4697 					err = -EFSBADCRC;
4698 					goto restore_opts;
4699 				}
4700 			}
4701 
4702 			/*
4703 			 * If we have an unprocessed orphan list hanging
4704 			 * around from a previously readonly bdev mount,
4705 			 * require a full umount/remount for now.
4706 			 */
4707 			if (es->s_last_orphan) {
4708 				ext4_msg(sb, KERN_WARNING, "Couldn't "
4709 				       "remount RDWR because of unprocessed "
4710 				       "orphan inode list.  Please "
4711 				       "umount/remount instead");
4712 				err = -EINVAL;
4713 				goto restore_opts;
4714 			}
4715 
4716 			/*
4717 			 * Mounting a RDONLY partition read-write, so reread
4718 			 * and store the current valid flag.  (It may have
4719 			 * been changed by e2fsck since we originally mounted
4720 			 * the partition.)
4721 			 */
4722 			if (sbi->s_journal)
4723 				ext4_clear_journal_err(sb, es);
4724 			sbi->s_mount_state = le16_to_cpu(es->s_state);
4725 			if (!ext4_setup_super(sb, es, 0))
4726 				sb->s_flags &= ~MS_RDONLY;
4727 			if (ext4_has_feature_mmp(sb))
4728 				if (ext4_multi_mount_protect(sb,
4729 						le64_to_cpu(es->s_mmp_block))) {
4730 					err = -EROFS;
4731 					goto restore_opts;
4732 				}
4733 			enable_quota = 1;
4734 		}
4735 	}
4736 
4737 	/*
4738 	 * Reinitialize lazy itable initialization thread based on
4739 	 * current settings
4740 	 */
4741 	if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4742 		ext4_unregister_li_request(sb);
4743 	else {
4744 		ext4_group_t first_not_zeroed;
4745 		first_not_zeroed = ext4_has_uninit_itable(sb);
4746 		ext4_register_li_request(sb, first_not_zeroed);
4747 	}
4748 
4749 	ext4_setup_system_zone(sb);
4750 	if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4751 		ext4_commit_super(sb, 1);
4752 
4753 #ifdef CONFIG_QUOTA
4754 	/* Release old quota file names */
4755 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4756 		kfree(old_opts.s_qf_names[i]);
4757 	if (enable_quota) {
4758 		if (sb_any_quota_suspended(sb))
4759 			dquot_resume(sb, -1);
4760 		else if (ext4_has_feature_quota(sb)) {
4761 			err = ext4_enable_quotas(sb);
4762 			if (err)
4763 				goto restore_opts;
4764 		}
4765 	}
4766 #endif
4767 
4768 	*flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4769 	ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4770 	kfree(orig_data);
4771 	return 0;
4772 
4773 restore_opts:
4774 	sb->s_flags = old_sb_flags;
4775 	sbi->s_mount_opt = old_opts.s_mount_opt;
4776 	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4777 	sbi->s_resuid = old_opts.s_resuid;
4778 	sbi->s_resgid = old_opts.s_resgid;
4779 	sbi->s_commit_interval = old_opts.s_commit_interval;
4780 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
4781 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
4782 #ifdef CONFIG_QUOTA
4783 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4784 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
4785 		kfree(sbi->s_qf_names[i]);
4786 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4787 	}
4788 #endif
4789 	kfree(orig_data);
4790 	return err;
4791 }
4792 
4793 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4794 {
4795 	struct super_block *sb = dentry->d_sb;
4796 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4797 	struct ext4_super_block *es = sbi->s_es;
4798 	ext4_fsblk_t overhead = 0, resv_blocks;
4799 	u64 fsid;
4800 	s64 bfree;
4801 	resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4802 
4803 	if (!test_opt(sb, MINIX_DF))
4804 		overhead = sbi->s_overhead;
4805 
4806 	buf->f_type = EXT4_SUPER_MAGIC;
4807 	buf->f_bsize = sb->s_blocksize;
4808 	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4809 	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4810 		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4811 	/* prevent underflow in case that few free space is available */
4812 	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4813 	buf->f_bavail = buf->f_bfree -
4814 			(ext4_r_blocks_count(es) + resv_blocks);
4815 	if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4816 		buf->f_bavail = 0;
4817 	buf->f_files = le32_to_cpu(es->s_inodes_count);
4818 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4819 	buf->f_namelen = EXT4_NAME_LEN;
4820 	fsid = le64_to_cpup((void *)es->s_uuid) ^
4821 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4822 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4823 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4824 
4825 	return 0;
4826 }
4827 
4828 /* Helper function for writing quotas on sync - we need to start transaction
4829  * before quota file is locked for write. Otherwise the are possible deadlocks:
4830  * Process 1                         Process 2
4831  * ext4_create()                     quota_sync()
4832  *   jbd2_journal_start()                  write_dquot()
4833  *   dquot_initialize()                         down(dqio_mutex)
4834  *     down(dqio_mutex)                    jbd2_journal_start()
4835  *
4836  */
4837 
4838 #ifdef CONFIG_QUOTA
4839 
4840 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4841 {
4842 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4843 }
4844 
4845 static int ext4_write_dquot(struct dquot *dquot)
4846 {
4847 	int ret, err;
4848 	handle_t *handle;
4849 	struct inode *inode;
4850 
4851 	inode = dquot_to_inode(dquot);
4852 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4853 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4854 	if (IS_ERR(handle))
4855 		return PTR_ERR(handle);
4856 	ret = dquot_commit(dquot);
4857 	err = ext4_journal_stop(handle);
4858 	if (!ret)
4859 		ret = err;
4860 	return ret;
4861 }
4862 
4863 static int ext4_acquire_dquot(struct dquot *dquot)
4864 {
4865 	int ret, err;
4866 	handle_t *handle;
4867 
4868 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4869 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4870 	if (IS_ERR(handle))
4871 		return PTR_ERR(handle);
4872 	ret = dquot_acquire(dquot);
4873 	err = ext4_journal_stop(handle);
4874 	if (!ret)
4875 		ret = err;
4876 	return ret;
4877 }
4878 
4879 static int ext4_release_dquot(struct dquot *dquot)
4880 {
4881 	int ret, err;
4882 	handle_t *handle;
4883 
4884 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4885 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4886 	if (IS_ERR(handle)) {
4887 		/* Release dquot anyway to avoid endless cycle in dqput() */
4888 		dquot_release(dquot);
4889 		return PTR_ERR(handle);
4890 	}
4891 	ret = dquot_release(dquot);
4892 	err = ext4_journal_stop(handle);
4893 	if (!ret)
4894 		ret = err;
4895 	return ret;
4896 }
4897 
4898 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4899 {
4900 	struct super_block *sb = dquot->dq_sb;
4901 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4902 
4903 	/* Are we journaling quotas? */
4904 	if (ext4_has_feature_quota(sb) ||
4905 	    sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4906 		dquot_mark_dquot_dirty(dquot);
4907 		return ext4_write_dquot(dquot);
4908 	} else {
4909 		return dquot_mark_dquot_dirty(dquot);
4910 	}
4911 }
4912 
4913 static int ext4_write_info(struct super_block *sb, int type)
4914 {
4915 	int ret, err;
4916 	handle_t *handle;
4917 
4918 	/* Data block + inode block */
4919 	handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
4920 	if (IS_ERR(handle))
4921 		return PTR_ERR(handle);
4922 	ret = dquot_commit_info(sb, type);
4923 	err = ext4_journal_stop(handle);
4924 	if (!ret)
4925 		ret = err;
4926 	return ret;
4927 }
4928 
4929 /*
4930  * Turn on quotas during mount time - we need to find
4931  * the quota file and such...
4932  */
4933 static int ext4_quota_on_mount(struct super_block *sb, int type)
4934 {
4935 	return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4936 					EXT4_SB(sb)->s_jquota_fmt, type);
4937 }
4938 
4939 /*
4940  * Standard function to be called on quota_on
4941  */
4942 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4943 			 struct path *path)
4944 {
4945 	int err;
4946 
4947 	if (!test_opt(sb, QUOTA))
4948 		return -EINVAL;
4949 
4950 	/* Quotafile not on the same filesystem? */
4951 	if (path->dentry->d_sb != sb)
4952 		return -EXDEV;
4953 	/* Journaling quota? */
4954 	if (EXT4_SB(sb)->s_qf_names[type]) {
4955 		/* Quotafile not in fs root? */
4956 		if (path->dentry->d_parent != sb->s_root)
4957 			ext4_msg(sb, KERN_WARNING,
4958 				"Quota file not on filesystem root. "
4959 				"Journaled quota will not work");
4960 	}
4961 
4962 	/*
4963 	 * When we journal data on quota file, we have to flush journal to see
4964 	 * all updates to the file when we bypass pagecache...
4965 	 */
4966 	if (EXT4_SB(sb)->s_journal &&
4967 	    ext4_should_journal_data(d_inode(path->dentry))) {
4968 		/*
4969 		 * We don't need to lock updates but journal_flush() could
4970 		 * otherwise be livelocked...
4971 		 */
4972 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4973 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4974 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4975 		if (err)
4976 			return err;
4977 	}
4978 
4979 	return dquot_quota_on(sb, type, format_id, path);
4980 }
4981 
4982 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4983 			     unsigned int flags)
4984 {
4985 	int err;
4986 	struct inode *qf_inode;
4987 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
4988 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4989 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4990 	};
4991 
4992 	BUG_ON(!ext4_has_feature_quota(sb));
4993 
4994 	if (!qf_inums[type])
4995 		return -EPERM;
4996 
4997 	qf_inode = ext4_iget(sb, qf_inums[type]);
4998 	if (IS_ERR(qf_inode)) {
4999 		ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5000 		return PTR_ERR(qf_inode);
5001 	}
5002 
5003 	/* Don't account quota for quota files to avoid recursion */
5004 	qf_inode->i_flags |= S_NOQUOTA;
5005 	err = dquot_enable(qf_inode, type, format_id, flags);
5006 	iput(qf_inode);
5007 
5008 	return err;
5009 }
5010 
5011 /* Enable usage tracking for all quota types. */
5012 static int ext4_enable_quotas(struct super_block *sb)
5013 {
5014 	int type, err = 0;
5015 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5016 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5017 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5018 	};
5019 
5020 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5021 	for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5022 		if (qf_inums[type]) {
5023 			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5024 						DQUOT_USAGE_ENABLED);
5025 			if (err) {
5026 				ext4_warning(sb,
5027 					"Failed to enable quota tracking "
5028 					"(type=%d, err=%d). Please run "
5029 					"e2fsck to fix.", type, err);
5030 				return err;
5031 			}
5032 		}
5033 	}
5034 	return 0;
5035 }
5036 
5037 static int ext4_quota_off(struct super_block *sb, int type)
5038 {
5039 	struct inode *inode = sb_dqopt(sb)->files[type];
5040 	handle_t *handle;
5041 
5042 	/* Force all delayed allocation blocks to be allocated.
5043 	 * Caller already holds s_umount sem */
5044 	if (test_opt(sb, DELALLOC))
5045 		sync_filesystem(sb);
5046 
5047 	if (!inode)
5048 		goto out;
5049 
5050 	/* Update modification times of quota files when userspace can
5051 	 * start looking at them */
5052 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5053 	if (IS_ERR(handle))
5054 		goto out;
5055 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5056 	ext4_mark_inode_dirty(handle, inode);
5057 	ext4_journal_stop(handle);
5058 
5059 out:
5060 	return dquot_quota_off(sb, type);
5061 }
5062 
5063 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5064  * acquiring the locks... As quota files are never truncated and quota code
5065  * itself serializes the operations (and no one else should touch the files)
5066  * we don't have to be afraid of races */
5067 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5068 			       size_t len, loff_t off)
5069 {
5070 	struct inode *inode = sb_dqopt(sb)->files[type];
5071 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5072 	int offset = off & (sb->s_blocksize - 1);
5073 	int tocopy;
5074 	size_t toread;
5075 	struct buffer_head *bh;
5076 	loff_t i_size = i_size_read(inode);
5077 
5078 	if (off > i_size)
5079 		return 0;
5080 	if (off+len > i_size)
5081 		len = i_size-off;
5082 	toread = len;
5083 	while (toread > 0) {
5084 		tocopy = sb->s_blocksize - offset < toread ?
5085 				sb->s_blocksize - offset : toread;
5086 		bh = ext4_bread(NULL, inode, blk, 0);
5087 		if (IS_ERR(bh))
5088 			return PTR_ERR(bh);
5089 		if (!bh)	/* A hole? */
5090 			memset(data, 0, tocopy);
5091 		else
5092 			memcpy(data, bh->b_data+offset, tocopy);
5093 		brelse(bh);
5094 		offset = 0;
5095 		toread -= tocopy;
5096 		data += tocopy;
5097 		blk++;
5098 	}
5099 	return len;
5100 }
5101 
5102 /* Write to quotafile (we know the transaction is already started and has
5103  * enough credits) */
5104 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5105 				const char *data, size_t len, loff_t off)
5106 {
5107 	struct inode *inode = sb_dqopt(sb)->files[type];
5108 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5109 	int err, offset = off & (sb->s_blocksize - 1);
5110 	int retries = 0;
5111 	struct buffer_head *bh;
5112 	handle_t *handle = journal_current_handle();
5113 
5114 	if (EXT4_SB(sb)->s_journal && !handle) {
5115 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5116 			" cancelled because transaction is not started",
5117 			(unsigned long long)off, (unsigned long long)len);
5118 		return -EIO;
5119 	}
5120 	/*
5121 	 * Since we account only one data block in transaction credits,
5122 	 * then it is impossible to cross a block boundary.
5123 	 */
5124 	if (sb->s_blocksize - offset < len) {
5125 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5126 			" cancelled because not block aligned",
5127 			(unsigned long long)off, (unsigned long long)len);
5128 		return -EIO;
5129 	}
5130 
5131 	do {
5132 		bh = ext4_bread(handle, inode, blk,
5133 				EXT4_GET_BLOCKS_CREATE |
5134 				EXT4_GET_BLOCKS_METADATA_NOFAIL);
5135 	} while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5136 		 ext4_should_retry_alloc(inode->i_sb, &retries));
5137 	if (IS_ERR(bh))
5138 		return PTR_ERR(bh);
5139 	if (!bh)
5140 		goto out;
5141 	BUFFER_TRACE(bh, "get write access");
5142 	err = ext4_journal_get_write_access(handle, bh);
5143 	if (err) {
5144 		brelse(bh);
5145 		return err;
5146 	}
5147 	lock_buffer(bh);
5148 	memcpy(bh->b_data+offset, data, len);
5149 	flush_dcache_page(bh->b_page);
5150 	unlock_buffer(bh);
5151 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
5152 	brelse(bh);
5153 out:
5154 	if (inode->i_size < off + len) {
5155 		i_size_write(inode, off + len);
5156 		EXT4_I(inode)->i_disksize = inode->i_size;
5157 		ext4_mark_inode_dirty(handle, inode);
5158 	}
5159 	return len;
5160 }
5161 
5162 #endif
5163 
5164 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5165 		       const char *dev_name, void *data)
5166 {
5167 	return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5168 }
5169 
5170 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5171 static inline void register_as_ext2(void)
5172 {
5173 	int err = register_filesystem(&ext2_fs_type);
5174 	if (err)
5175 		printk(KERN_WARNING
5176 		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5177 }
5178 
5179 static inline void unregister_as_ext2(void)
5180 {
5181 	unregister_filesystem(&ext2_fs_type);
5182 }
5183 
5184 static inline int ext2_feature_set_ok(struct super_block *sb)
5185 {
5186 	if (ext4_has_unknown_ext2_incompat_features(sb))
5187 		return 0;
5188 	if (sb->s_flags & MS_RDONLY)
5189 		return 1;
5190 	if (ext4_has_unknown_ext2_ro_compat_features(sb))
5191 		return 0;
5192 	return 1;
5193 }
5194 #else
5195 static inline void register_as_ext2(void) { }
5196 static inline void unregister_as_ext2(void) { }
5197 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5198 #endif
5199 
5200 static inline void register_as_ext3(void)
5201 {
5202 	int err = register_filesystem(&ext3_fs_type);
5203 	if (err)
5204 		printk(KERN_WARNING
5205 		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5206 }
5207 
5208 static inline void unregister_as_ext3(void)
5209 {
5210 	unregister_filesystem(&ext3_fs_type);
5211 }
5212 
5213 static inline int ext3_feature_set_ok(struct super_block *sb)
5214 {
5215 	if (ext4_has_unknown_ext3_incompat_features(sb))
5216 		return 0;
5217 	if (!ext4_has_feature_journal(sb))
5218 		return 0;
5219 	if (sb->s_flags & MS_RDONLY)
5220 		return 1;
5221 	if (ext4_has_unknown_ext3_ro_compat_features(sb))
5222 		return 0;
5223 	return 1;
5224 }
5225 
5226 static struct file_system_type ext4_fs_type = {
5227 	.owner		= THIS_MODULE,
5228 	.name		= "ext4",
5229 	.mount		= ext4_mount,
5230 	.kill_sb	= kill_block_super,
5231 	.fs_flags	= FS_REQUIRES_DEV,
5232 };
5233 MODULE_ALIAS_FS("ext4");
5234 
5235 /* Shared across all ext4 file systems */
5236 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5237 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5238 
5239 static int __init ext4_init_fs(void)
5240 {
5241 	int i, err;
5242 
5243 	ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5244 	ext4_li_info = NULL;
5245 	mutex_init(&ext4_li_mtx);
5246 
5247 	/* Build-time check for flags consistency */
5248 	ext4_check_flag_values();
5249 
5250 	for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5251 		mutex_init(&ext4__aio_mutex[i]);
5252 		init_waitqueue_head(&ext4__ioend_wq[i]);
5253 	}
5254 
5255 	err = ext4_init_es();
5256 	if (err)
5257 		return err;
5258 
5259 	err = ext4_init_pageio();
5260 	if (err)
5261 		goto out5;
5262 
5263 	err = ext4_init_system_zone();
5264 	if (err)
5265 		goto out4;
5266 
5267 	err = ext4_init_sysfs();
5268 	if (err)
5269 		goto out3;
5270 
5271 	err = ext4_init_mballoc();
5272 	if (err)
5273 		goto out2;
5274 	else
5275 		ext4_mballoc_ready = 1;
5276 	err = init_inodecache();
5277 	if (err)
5278 		goto out1;
5279 	register_as_ext3();
5280 	register_as_ext2();
5281 	err = register_filesystem(&ext4_fs_type);
5282 	if (err)
5283 		goto out;
5284 
5285 	return 0;
5286 out:
5287 	unregister_as_ext2();
5288 	unregister_as_ext3();
5289 	destroy_inodecache();
5290 out1:
5291 	ext4_mballoc_ready = 0;
5292 	ext4_exit_mballoc();
5293 out2:
5294 	ext4_exit_sysfs();
5295 out3:
5296 	ext4_exit_system_zone();
5297 out4:
5298 	ext4_exit_pageio();
5299 out5:
5300 	ext4_exit_es();
5301 
5302 	return err;
5303 }
5304 
5305 static void __exit ext4_exit_fs(void)
5306 {
5307 	ext4_exit_crypto();
5308 	ext4_destroy_lazyinit_thread();
5309 	unregister_as_ext2();
5310 	unregister_as_ext3();
5311 	unregister_filesystem(&ext4_fs_type);
5312 	destroy_inodecache();
5313 	ext4_exit_mballoc();
5314 	ext4_exit_sysfs();
5315 	ext4_exit_system_zone();
5316 	ext4_exit_pageio();
5317 	ext4_exit_es();
5318 }
5319 
5320 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5321 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5322 MODULE_LICENSE("GPL");
5323 module_init(ext4_init_fs)
5324 module_exit(ext4_exit_fs)
5325