xref: /openbmc/linux/fs/ext4/super.c (revision e2b911c53584a92266943f3b7f2cdbc19c1a4e80)
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 		panic("EXT4-fs (device %s): panic forced after error\n",
393 			sb->s_id);
394 }
395 
396 #define ext4_error_ratelimit(sb)					\
397 		___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),	\
398 			     "EXT4-fs error")
399 
400 void __ext4_error(struct super_block *sb, const char *function,
401 		  unsigned int line, const char *fmt, ...)
402 {
403 	struct va_format vaf;
404 	va_list args;
405 
406 	if (ext4_error_ratelimit(sb)) {
407 		va_start(args, fmt);
408 		vaf.fmt = fmt;
409 		vaf.va = &args;
410 		printk(KERN_CRIT
411 		       "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
412 		       sb->s_id, function, line, current->comm, &vaf);
413 		va_end(args);
414 	}
415 	save_error_info(sb, function, line);
416 	ext4_handle_error(sb);
417 }
418 
419 void __ext4_error_inode(struct inode *inode, const char *function,
420 			unsigned int line, ext4_fsblk_t block,
421 			const char *fmt, ...)
422 {
423 	va_list args;
424 	struct va_format vaf;
425 	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
426 
427 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
428 	es->s_last_error_block = cpu_to_le64(block);
429 	if (ext4_error_ratelimit(inode->i_sb)) {
430 		va_start(args, fmt);
431 		vaf.fmt = fmt;
432 		vaf.va = &args;
433 		if (block)
434 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
435 			       "inode #%lu: block %llu: comm %s: %pV\n",
436 			       inode->i_sb->s_id, function, line, inode->i_ino,
437 			       block, current->comm, &vaf);
438 		else
439 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
440 			       "inode #%lu: comm %s: %pV\n",
441 			       inode->i_sb->s_id, function, line, inode->i_ino,
442 			       current->comm, &vaf);
443 		va_end(args);
444 	}
445 	save_error_info(inode->i_sb, function, line);
446 	ext4_handle_error(inode->i_sb);
447 }
448 
449 void __ext4_error_file(struct file *file, const char *function,
450 		       unsigned int line, ext4_fsblk_t block,
451 		       const char *fmt, ...)
452 {
453 	va_list args;
454 	struct va_format vaf;
455 	struct ext4_super_block *es;
456 	struct inode *inode = file_inode(file);
457 	char pathname[80], *path;
458 
459 	es = EXT4_SB(inode->i_sb)->s_es;
460 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
461 	if (ext4_error_ratelimit(inode->i_sb)) {
462 		path = file_path(file, pathname, sizeof(pathname));
463 		if (IS_ERR(path))
464 			path = "(unknown)";
465 		va_start(args, fmt);
466 		vaf.fmt = fmt;
467 		vaf.va = &args;
468 		if (block)
469 			printk(KERN_CRIT
470 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
471 			       "block %llu: comm %s: path %s: %pV\n",
472 			       inode->i_sb->s_id, function, line, inode->i_ino,
473 			       block, current->comm, path, &vaf);
474 		else
475 			printk(KERN_CRIT
476 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
477 			       "comm %s: path %s: %pV\n",
478 			       inode->i_sb->s_id, function, line, inode->i_ino,
479 			       current->comm, path, &vaf);
480 		va_end(args);
481 	}
482 	save_error_info(inode->i_sb, function, line);
483 	ext4_handle_error(inode->i_sb);
484 }
485 
486 const char *ext4_decode_error(struct super_block *sb, int errno,
487 			      char nbuf[16])
488 {
489 	char *errstr = NULL;
490 
491 	switch (errno) {
492 	case -EFSCORRUPTED:
493 		errstr = "Corrupt filesystem";
494 		break;
495 	case -EFSBADCRC:
496 		errstr = "Filesystem failed CRC";
497 		break;
498 	case -EIO:
499 		errstr = "IO failure";
500 		break;
501 	case -ENOMEM:
502 		errstr = "Out of memory";
503 		break;
504 	case -EROFS:
505 		if (!sb || (EXT4_SB(sb)->s_journal &&
506 			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
507 			errstr = "Journal has aborted";
508 		else
509 			errstr = "Readonly filesystem";
510 		break;
511 	default:
512 		/* If the caller passed in an extra buffer for unknown
513 		 * errors, textualise them now.  Else we just return
514 		 * NULL. */
515 		if (nbuf) {
516 			/* Check for truncated error codes... */
517 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
518 				errstr = nbuf;
519 		}
520 		break;
521 	}
522 
523 	return errstr;
524 }
525 
526 /* __ext4_std_error decodes expected errors from journaling functions
527  * automatically and invokes the appropriate error response.  */
528 
529 void __ext4_std_error(struct super_block *sb, const char *function,
530 		      unsigned int line, int errno)
531 {
532 	char nbuf[16];
533 	const char *errstr;
534 
535 	/* Special case: if the error is EROFS, and we're not already
536 	 * inside a transaction, then there's really no point in logging
537 	 * an error. */
538 	if (errno == -EROFS && journal_current_handle() == NULL &&
539 	    (sb->s_flags & MS_RDONLY))
540 		return;
541 
542 	if (ext4_error_ratelimit(sb)) {
543 		errstr = ext4_decode_error(sb, errno, nbuf);
544 		printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
545 		       sb->s_id, function, line, errstr);
546 	}
547 
548 	save_error_info(sb, function, line);
549 	ext4_handle_error(sb);
550 }
551 
552 /*
553  * ext4_abort is a much stronger failure handler than ext4_error.  The
554  * abort function may be used to deal with unrecoverable failures such
555  * as journal IO errors or ENOMEM at a critical moment in log management.
556  *
557  * We unconditionally force the filesystem into an ABORT|READONLY state,
558  * unless the error response on the fs has been set to panic in which
559  * case we take the easy way out and panic immediately.
560  */
561 
562 void __ext4_abort(struct super_block *sb, const char *function,
563 		unsigned int line, const char *fmt, ...)
564 {
565 	va_list args;
566 
567 	save_error_info(sb, function, line);
568 	va_start(args, fmt);
569 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
570 	       function, line);
571 	vprintk(fmt, args);
572 	printk("\n");
573 	va_end(args);
574 
575 	if ((sb->s_flags & MS_RDONLY) == 0) {
576 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
577 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
578 		/*
579 		 * Make sure updated value of ->s_mount_flags will be visible
580 		 * before ->s_flags update
581 		 */
582 		smp_wmb();
583 		sb->s_flags |= MS_RDONLY;
584 		if (EXT4_SB(sb)->s_journal)
585 			jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
586 		save_error_info(sb, function, line);
587 	}
588 	if (test_opt(sb, ERRORS_PANIC))
589 		panic("EXT4-fs panic from previous error\n");
590 }
591 
592 void __ext4_msg(struct super_block *sb,
593 		const char *prefix, const char *fmt, ...)
594 {
595 	struct va_format vaf;
596 	va_list args;
597 
598 	if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
599 		return;
600 
601 	va_start(args, fmt);
602 	vaf.fmt = fmt;
603 	vaf.va = &args;
604 	printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
605 	va_end(args);
606 }
607 
608 #define ext4_warning_ratelimit(sb)					\
609 		___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),	\
610 			     "EXT4-fs warning")
611 
612 void __ext4_warning(struct super_block *sb, const char *function,
613 		    unsigned int line, const char *fmt, ...)
614 {
615 	struct va_format vaf;
616 	va_list args;
617 
618 	if (!ext4_warning_ratelimit(sb))
619 		return;
620 
621 	va_start(args, fmt);
622 	vaf.fmt = fmt;
623 	vaf.va = &args;
624 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
625 	       sb->s_id, function, line, &vaf);
626 	va_end(args);
627 }
628 
629 void __ext4_warning_inode(const struct inode *inode, const char *function,
630 			  unsigned int line, const char *fmt, ...)
631 {
632 	struct va_format vaf;
633 	va_list args;
634 
635 	if (!ext4_warning_ratelimit(inode->i_sb))
636 		return;
637 
638 	va_start(args, fmt);
639 	vaf.fmt = fmt;
640 	vaf.va = &args;
641 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
642 	       "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
643 	       function, line, inode->i_ino, current->comm, &vaf);
644 	va_end(args);
645 }
646 
647 void __ext4_grp_locked_error(const char *function, unsigned int line,
648 			     struct super_block *sb, ext4_group_t grp,
649 			     unsigned long ino, ext4_fsblk_t block,
650 			     const char *fmt, ...)
651 __releases(bitlock)
652 __acquires(bitlock)
653 {
654 	struct va_format vaf;
655 	va_list args;
656 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
657 
658 	es->s_last_error_ino = cpu_to_le32(ino);
659 	es->s_last_error_block = cpu_to_le64(block);
660 	__save_error_info(sb, function, line);
661 
662 	if (ext4_error_ratelimit(sb)) {
663 		va_start(args, fmt);
664 		vaf.fmt = fmt;
665 		vaf.va = &args;
666 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
667 		       sb->s_id, function, line, grp);
668 		if (ino)
669 			printk(KERN_CONT "inode %lu: ", ino);
670 		if (block)
671 			printk(KERN_CONT "block %llu:",
672 			       (unsigned long long) block);
673 		printk(KERN_CONT "%pV\n", &vaf);
674 		va_end(args);
675 	}
676 
677 	if (test_opt(sb, ERRORS_CONT)) {
678 		ext4_commit_super(sb, 0);
679 		return;
680 	}
681 
682 	ext4_unlock_group(sb, grp);
683 	ext4_handle_error(sb);
684 	/*
685 	 * We only get here in the ERRORS_RO case; relocking the group
686 	 * may be dangerous, but nothing bad will happen since the
687 	 * filesystem will have already been marked read/only and the
688 	 * journal has been aborted.  We return 1 as a hint to callers
689 	 * who might what to use the return value from
690 	 * ext4_grp_locked_error() to distinguish between the
691 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
692 	 * aggressively from the ext4 function in question, with a
693 	 * more appropriate error code.
694 	 */
695 	ext4_lock_group(sb, grp);
696 	return;
697 }
698 
699 void ext4_update_dynamic_rev(struct super_block *sb)
700 {
701 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
702 
703 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
704 		return;
705 
706 	ext4_warning(sb,
707 		     "updating to rev %d because of new feature flag, "
708 		     "running e2fsck is recommended",
709 		     EXT4_DYNAMIC_REV);
710 
711 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
712 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
713 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
714 	/* leave es->s_feature_*compat flags alone */
715 	/* es->s_uuid will be set by e2fsck if empty */
716 
717 	/*
718 	 * The rest of the superblock fields should be zero, and if not it
719 	 * means they are likely already in use, so leave them alone.  We
720 	 * can leave it up to e2fsck to clean up any inconsistencies there.
721 	 */
722 }
723 
724 /*
725  * Open the external journal device
726  */
727 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
728 {
729 	struct block_device *bdev;
730 	char b[BDEVNAME_SIZE];
731 
732 	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
733 	if (IS_ERR(bdev))
734 		goto fail;
735 	return bdev;
736 
737 fail:
738 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
739 			__bdevname(dev, b), PTR_ERR(bdev));
740 	return NULL;
741 }
742 
743 /*
744  * Release the journal device
745  */
746 static void ext4_blkdev_put(struct block_device *bdev)
747 {
748 	blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
749 }
750 
751 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
752 {
753 	struct block_device *bdev;
754 	bdev = sbi->journal_bdev;
755 	if (bdev) {
756 		ext4_blkdev_put(bdev);
757 		sbi->journal_bdev = NULL;
758 	}
759 }
760 
761 static inline struct inode *orphan_list_entry(struct list_head *l)
762 {
763 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
764 }
765 
766 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
767 {
768 	struct list_head *l;
769 
770 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
771 		 le32_to_cpu(sbi->s_es->s_last_orphan));
772 
773 	printk(KERN_ERR "sb_info orphan list:\n");
774 	list_for_each(l, &sbi->s_orphan) {
775 		struct inode *inode = orphan_list_entry(l);
776 		printk(KERN_ERR "  "
777 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
778 		       inode->i_sb->s_id, inode->i_ino, inode,
779 		       inode->i_mode, inode->i_nlink,
780 		       NEXT_ORPHAN(inode));
781 	}
782 }
783 
784 static void ext4_put_super(struct super_block *sb)
785 {
786 	struct ext4_sb_info *sbi = EXT4_SB(sb);
787 	struct ext4_super_block *es = sbi->s_es;
788 	int i, err;
789 
790 	ext4_unregister_li_request(sb);
791 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
792 
793 	flush_workqueue(sbi->rsv_conversion_wq);
794 	destroy_workqueue(sbi->rsv_conversion_wq);
795 
796 	if (sbi->s_journal) {
797 		err = jbd2_journal_destroy(sbi->s_journal);
798 		sbi->s_journal = NULL;
799 		if (err < 0)
800 			ext4_abort(sb, "Couldn't clean up the journal");
801 	}
802 
803 	ext4_unregister_sysfs(sb);
804 	ext4_es_unregister_shrinker(sbi);
805 	del_timer_sync(&sbi->s_err_report);
806 	ext4_release_system_zone(sb);
807 	ext4_mb_release(sb);
808 	ext4_ext_release(sb);
809 	ext4_xattr_put_super(sb);
810 
811 	if (!(sb->s_flags & MS_RDONLY)) {
812 		ext4_clear_feature_journal_needs_recovery(sb);
813 		es->s_state = cpu_to_le16(sbi->s_mount_state);
814 	}
815 	if (!(sb->s_flags & MS_RDONLY))
816 		ext4_commit_super(sb, 1);
817 
818 	for (i = 0; i < sbi->s_gdb_count; i++)
819 		brelse(sbi->s_group_desc[i]);
820 	kvfree(sbi->s_group_desc);
821 	kvfree(sbi->s_flex_groups);
822 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
823 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
824 	percpu_counter_destroy(&sbi->s_dirs_counter);
825 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
826 	brelse(sbi->s_sbh);
827 #ifdef CONFIG_QUOTA
828 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
829 		kfree(sbi->s_qf_names[i]);
830 #endif
831 
832 	/* Debugging code just in case the in-memory inode orphan list
833 	 * isn't empty.  The on-disk one can be non-empty if we've
834 	 * detected an error and taken the fs readonly, but the
835 	 * in-memory list had better be clean by this point. */
836 	if (!list_empty(&sbi->s_orphan))
837 		dump_orphan_list(sb, sbi);
838 	J_ASSERT(list_empty(&sbi->s_orphan));
839 
840 	sync_blockdev(sb->s_bdev);
841 	invalidate_bdev(sb->s_bdev);
842 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
843 		/*
844 		 * Invalidate the journal device's buffers.  We don't want them
845 		 * floating about in memory - the physical journal device may
846 		 * hotswapped, and it breaks the `ro-after' testing code.
847 		 */
848 		sync_blockdev(sbi->journal_bdev);
849 		invalidate_bdev(sbi->journal_bdev);
850 		ext4_blkdev_remove(sbi);
851 	}
852 	if (sbi->s_mb_cache) {
853 		ext4_xattr_destroy_cache(sbi->s_mb_cache);
854 		sbi->s_mb_cache = NULL;
855 	}
856 	if (sbi->s_mmp_tsk)
857 		kthread_stop(sbi->s_mmp_tsk);
858 	sb->s_fs_info = NULL;
859 	/*
860 	 * Now that we are completely done shutting down the
861 	 * superblock, we need to actually destroy the kobject.
862 	 */
863 	kobject_put(&sbi->s_kobj);
864 	wait_for_completion(&sbi->s_kobj_unregister);
865 	if (sbi->s_chksum_driver)
866 		crypto_free_shash(sbi->s_chksum_driver);
867 	kfree(sbi->s_blockgroup_lock);
868 	kfree(sbi);
869 }
870 
871 static struct kmem_cache *ext4_inode_cachep;
872 
873 /*
874  * Called inside transaction, so use GFP_NOFS
875  */
876 static struct inode *ext4_alloc_inode(struct super_block *sb)
877 {
878 	struct ext4_inode_info *ei;
879 
880 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
881 	if (!ei)
882 		return NULL;
883 
884 	ei->vfs_inode.i_version = 1;
885 	spin_lock_init(&ei->i_raw_lock);
886 	INIT_LIST_HEAD(&ei->i_prealloc_list);
887 	spin_lock_init(&ei->i_prealloc_lock);
888 	ext4_es_init_tree(&ei->i_es_tree);
889 	rwlock_init(&ei->i_es_lock);
890 	INIT_LIST_HEAD(&ei->i_es_list);
891 	ei->i_es_all_nr = 0;
892 	ei->i_es_shk_nr = 0;
893 	ei->i_es_shrink_lblk = 0;
894 	ei->i_reserved_data_blocks = 0;
895 	ei->i_reserved_meta_blocks = 0;
896 	ei->i_allocated_meta_blocks = 0;
897 	ei->i_da_metadata_calc_len = 0;
898 	ei->i_da_metadata_calc_last_lblock = 0;
899 	spin_lock_init(&(ei->i_block_reservation_lock));
900 #ifdef CONFIG_QUOTA
901 	ei->i_reserved_quota = 0;
902 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
903 #endif
904 	ei->jinode = NULL;
905 	INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
906 	spin_lock_init(&ei->i_completed_io_lock);
907 	ei->i_sync_tid = 0;
908 	ei->i_datasync_tid = 0;
909 	atomic_set(&ei->i_ioend_count, 0);
910 	atomic_set(&ei->i_unwritten, 0);
911 	INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
912 #ifdef CONFIG_EXT4_FS_ENCRYPTION
913 	ei->i_crypt_info = NULL;
914 #endif
915 	return &ei->vfs_inode;
916 }
917 
918 static int ext4_drop_inode(struct inode *inode)
919 {
920 	int drop = generic_drop_inode(inode);
921 
922 	trace_ext4_drop_inode(inode, drop);
923 	return drop;
924 }
925 
926 static void ext4_i_callback(struct rcu_head *head)
927 {
928 	struct inode *inode = container_of(head, struct inode, i_rcu);
929 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
930 }
931 
932 static void ext4_destroy_inode(struct inode *inode)
933 {
934 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
935 		ext4_msg(inode->i_sb, KERN_ERR,
936 			 "Inode %lu (%p): orphan list check failed!",
937 			 inode->i_ino, EXT4_I(inode));
938 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
939 				EXT4_I(inode), sizeof(struct ext4_inode_info),
940 				true);
941 		dump_stack();
942 	}
943 	call_rcu(&inode->i_rcu, ext4_i_callback);
944 }
945 
946 static void init_once(void *foo)
947 {
948 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
949 
950 	INIT_LIST_HEAD(&ei->i_orphan);
951 	init_rwsem(&ei->xattr_sem);
952 	init_rwsem(&ei->i_data_sem);
953 	inode_init_once(&ei->vfs_inode);
954 }
955 
956 static int __init init_inodecache(void)
957 {
958 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
959 					     sizeof(struct ext4_inode_info),
960 					     0, (SLAB_RECLAIM_ACCOUNT|
961 						SLAB_MEM_SPREAD),
962 					     init_once);
963 	if (ext4_inode_cachep == NULL)
964 		return -ENOMEM;
965 	return 0;
966 }
967 
968 static void destroy_inodecache(void)
969 {
970 	/*
971 	 * Make sure all delayed rcu free inodes are flushed before we
972 	 * destroy cache.
973 	 */
974 	rcu_barrier();
975 	kmem_cache_destroy(ext4_inode_cachep);
976 }
977 
978 void ext4_clear_inode(struct inode *inode)
979 {
980 	invalidate_inode_buffers(inode);
981 	clear_inode(inode);
982 	dquot_drop(inode);
983 	ext4_discard_preallocations(inode);
984 	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
985 	if (EXT4_I(inode)->jinode) {
986 		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
987 					       EXT4_I(inode)->jinode);
988 		jbd2_free_inode(EXT4_I(inode)->jinode);
989 		EXT4_I(inode)->jinode = NULL;
990 	}
991 #ifdef CONFIG_EXT4_FS_ENCRYPTION
992 	if (EXT4_I(inode)->i_crypt_info)
993 		ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
994 #endif
995 }
996 
997 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
998 					u64 ino, u32 generation)
999 {
1000 	struct inode *inode;
1001 
1002 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1003 		return ERR_PTR(-ESTALE);
1004 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1005 		return ERR_PTR(-ESTALE);
1006 
1007 	/* iget isn't really right if the inode is currently unallocated!!
1008 	 *
1009 	 * ext4_read_inode will return a bad_inode if the inode had been
1010 	 * deleted, so we should be safe.
1011 	 *
1012 	 * Currently we don't know the generation for parent directory, so
1013 	 * a generation of 0 means "accept any"
1014 	 */
1015 	inode = ext4_iget_normal(sb, ino);
1016 	if (IS_ERR(inode))
1017 		return ERR_CAST(inode);
1018 	if (generation && inode->i_generation != generation) {
1019 		iput(inode);
1020 		return ERR_PTR(-ESTALE);
1021 	}
1022 
1023 	return inode;
1024 }
1025 
1026 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1027 					int fh_len, int fh_type)
1028 {
1029 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1030 				    ext4_nfs_get_inode);
1031 }
1032 
1033 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1034 					int fh_len, int fh_type)
1035 {
1036 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1037 				    ext4_nfs_get_inode);
1038 }
1039 
1040 /*
1041  * Try to release metadata pages (indirect blocks, directories) which are
1042  * mapped via the block device.  Since these pages could have journal heads
1043  * which would prevent try_to_free_buffers() from freeing them, we must use
1044  * jbd2 layer's try_to_free_buffers() function to release them.
1045  */
1046 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1047 				 gfp_t wait)
1048 {
1049 	journal_t *journal = EXT4_SB(sb)->s_journal;
1050 
1051 	WARN_ON(PageChecked(page));
1052 	if (!page_has_buffers(page))
1053 		return 0;
1054 	if (journal)
1055 		return jbd2_journal_try_to_free_buffers(journal, page,
1056 							wait & ~__GFP_WAIT);
1057 	return try_to_free_buffers(page);
1058 }
1059 
1060 #ifdef CONFIG_QUOTA
1061 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1062 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1063 
1064 static int ext4_write_dquot(struct dquot *dquot);
1065 static int ext4_acquire_dquot(struct dquot *dquot);
1066 static int ext4_release_dquot(struct dquot *dquot);
1067 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1068 static int ext4_write_info(struct super_block *sb, int type);
1069 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1070 			 struct path *path);
1071 static int ext4_quota_off(struct super_block *sb, int type);
1072 static int ext4_quota_on_mount(struct super_block *sb, int type);
1073 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1074 			       size_t len, loff_t off);
1075 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1076 				const char *data, size_t len, loff_t off);
1077 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1078 			     unsigned int flags);
1079 static int ext4_enable_quotas(struct super_block *sb);
1080 
1081 static struct dquot **ext4_get_dquots(struct inode *inode)
1082 {
1083 	return EXT4_I(inode)->i_dquot;
1084 }
1085 
1086 static const struct dquot_operations ext4_quota_operations = {
1087 	.get_reserved_space = ext4_get_reserved_space,
1088 	.write_dquot	= ext4_write_dquot,
1089 	.acquire_dquot	= ext4_acquire_dquot,
1090 	.release_dquot	= ext4_release_dquot,
1091 	.mark_dirty	= ext4_mark_dquot_dirty,
1092 	.write_info	= ext4_write_info,
1093 	.alloc_dquot	= dquot_alloc,
1094 	.destroy_dquot	= dquot_destroy,
1095 };
1096 
1097 static const struct quotactl_ops ext4_qctl_operations = {
1098 	.quota_on	= ext4_quota_on,
1099 	.quota_off	= ext4_quota_off,
1100 	.quota_sync	= dquot_quota_sync,
1101 	.get_state	= dquot_get_state,
1102 	.set_info	= dquot_set_dqinfo,
1103 	.get_dqblk	= dquot_get_dqblk,
1104 	.set_dqblk	= dquot_set_dqblk
1105 };
1106 #endif
1107 
1108 static const struct super_operations ext4_sops = {
1109 	.alloc_inode	= ext4_alloc_inode,
1110 	.destroy_inode	= ext4_destroy_inode,
1111 	.write_inode	= ext4_write_inode,
1112 	.dirty_inode	= ext4_dirty_inode,
1113 	.drop_inode	= ext4_drop_inode,
1114 	.evict_inode	= ext4_evict_inode,
1115 	.put_super	= ext4_put_super,
1116 	.sync_fs	= ext4_sync_fs,
1117 	.freeze_fs	= ext4_freeze,
1118 	.unfreeze_fs	= ext4_unfreeze,
1119 	.statfs		= ext4_statfs,
1120 	.remount_fs	= ext4_remount,
1121 	.show_options	= ext4_show_options,
1122 #ifdef CONFIG_QUOTA
1123 	.quota_read	= ext4_quota_read,
1124 	.quota_write	= ext4_quota_write,
1125 	.get_dquots	= ext4_get_dquots,
1126 #endif
1127 	.bdev_try_to_free_page = bdev_try_to_free_page,
1128 };
1129 
1130 static const struct export_operations ext4_export_ops = {
1131 	.fh_to_dentry = ext4_fh_to_dentry,
1132 	.fh_to_parent = ext4_fh_to_parent,
1133 	.get_parent = ext4_get_parent,
1134 };
1135 
1136 enum {
1137 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1138 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1139 	Opt_nouid32, Opt_debug, Opt_removed,
1140 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1141 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1142 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1143 	Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1144 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1145 	Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1146 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1147 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1148 	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1149 	Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1150 	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1151 	Opt_lazytime, Opt_nolazytime,
1152 	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1153 	Opt_inode_readahead_blks, Opt_journal_ioprio,
1154 	Opt_dioread_nolock, Opt_dioread_lock,
1155 	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1156 	Opt_max_dir_size_kb, Opt_nojournal_checksum,
1157 };
1158 
1159 static const match_table_t tokens = {
1160 	{Opt_bsd_df, "bsddf"},
1161 	{Opt_minix_df, "minixdf"},
1162 	{Opt_grpid, "grpid"},
1163 	{Opt_grpid, "bsdgroups"},
1164 	{Opt_nogrpid, "nogrpid"},
1165 	{Opt_nogrpid, "sysvgroups"},
1166 	{Opt_resgid, "resgid=%u"},
1167 	{Opt_resuid, "resuid=%u"},
1168 	{Opt_sb, "sb=%u"},
1169 	{Opt_err_cont, "errors=continue"},
1170 	{Opt_err_panic, "errors=panic"},
1171 	{Opt_err_ro, "errors=remount-ro"},
1172 	{Opt_nouid32, "nouid32"},
1173 	{Opt_debug, "debug"},
1174 	{Opt_removed, "oldalloc"},
1175 	{Opt_removed, "orlov"},
1176 	{Opt_user_xattr, "user_xattr"},
1177 	{Opt_nouser_xattr, "nouser_xattr"},
1178 	{Opt_acl, "acl"},
1179 	{Opt_noacl, "noacl"},
1180 	{Opt_noload, "norecovery"},
1181 	{Opt_noload, "noload"},
1182 	{Opt_removed, "nobh"},
1183 	{Opt_removed, "bh"},
1184 	{Opt_commit, "commit=%u"},
1185 	{Opt_min_batch_time, "min_batch_time=%u"},
1186 	{Opt_max_batch_time, "max_batch_time=%u"},
1187 	{Opt_journal_dev, "journal_dev=%u"},
1188 	{Opt_journal_path, "journal_path=%s"},
1189 	{Opt_journal_checksum, "journal_checksum"},
1190 	{Opt_nojournal_checksum, "nojournal_checksum"},
1191 	{Opt_journal_async_commit, "journal_async_commit"},
1192 	{Opt_abort, "abort"},
1193 	{Opt_data_journal, "data=journal"},
1194 	{Opt_data_ordered, "data=ordered"},
1195 	{Opt_data_writeback, "data=writeback"},
1196 	{Opt_data_err_abort, "data_err=abort"},
1197 	{Opt_data_err_ignore, "data_err=ignore"},
1198 	{Opt_offusrjquota, "usrjquota="},
1199 	{Opt_usrjquota, "usrjquota=%s"},
1200 	{Opt_offgrpjquota, "grpjquota="},
1201 	{Opt_grpjquota, "grpjquota=%s"},
1202 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1203 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1204 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1205 	{Opt_grpquota, "grpquota"},
1206 	{Opt_noquota, "noquota"},
1207 	{Opt_quota, "quota"},
1208 	{Opt_usrquota, "usrquota"},
1209 	{Opt_barrier, "barrier=%u"},
1210 	{Opt_barrier, "barrier"},
1211 	{Opt_nobarrier, "nobarrier"},
1212 	{Opt_i_version, "i_version"},
1213 	{Opt_dax, "dax"},
1214 	{Opt_stripe, "stripe=%u"},
1215 	{Opt_delalloc, "delalloc"},
1216 	{Opt_lazytime, "lazytime"},
1217 	{Opt_nolazytime, "nolazytime"},
1218 	{Opt_nodelalloc, "nodelalloc"},
1219 	{Opt_removed, "mblk_io_submit"},
1220 	{Opt_removed, "nomblk_io_submit"},
1221 	{Opt_block_validity, "block_validity"},
1222 	{Opt_noblock_validity, "noblock_validity"},
1223 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1224 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1225 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1226 	{Opt_auto_da_alloc, "auto_da_alloc"},
1227 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1228 	{Opt_dioread_nolock, "dioread_nolock"},
1229 	{Opt_dioread_lock, "dioread_lock"},
1230 	{Opt_discard, "discard"},
1231 	{Opt_nodiscard, "nodiscard"},
1232 	{Opt_init_itable, "init_itable=%u"},
1233 	{Opt_init_itable, "init_itable"},
1234 	{Opt_noinit_itable, "noinit_itable"},
1235 	{Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1236 	{Opt_test_dummy_encryption, "test_dummy_encryption"},
1237 	{Opt_removed, "check=none"},	/* mount option from ext2/3 */
1238 	{Opt_removed, "nocheck"},	/* mount option from ext2/3 */
1239 	{Opt_removed, "reservation"},	/* mount option from ext2/3 */
1240 	{Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1241 	{Opt_removed, "journal=%u"},	/* mount option from ext2/3 */
1242 	{Opt_err, NULL},
1243 };
1244 
1245 static ext4_fsblk_t get_sb_block(void **data)
1246 {
1247 	ext4_fsblk_t	sb_block;
1248 	char		*options = (char *) *data;
1249 
1250 	if (!options || strncmp(options, "sb=", 3) != 0)
1251 		return 1;	/* Default location */
1252 
1253 	options += 3;
1254 	/* TODO: use simple_strtoll with >32bit ext4 */
1255 	sb_block = simple_strtoul(options, &options, 0);
1256 	if (*options && *options != ',') {
1257 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1258 		       (char *) *data);
1259 		return 1;
1260 	}
1261 	if (*options == ',')
1262 		options++;
1263 	*data = (void *) options;
1264 
1265 	return sb_block;
1266 }
1267 
1268 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1269 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1270 	"Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1271 
1272 #ifdef CONFIG_QUOTA
1273 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1274 {
1275 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1276 	char *qname;
1277 	int ret = -1;
1278 
1279 	if (sb_any_quota_loaded(sb) &&
1280 		!sbi->s_qf_names[qtype]) {
1281 		ext4_msg(sb, KERN_ERR,
1282 			"Cannot change journaled "
1283 			"quota options when quota turned on");
1284 		return -1;
1285 	}
1286 	if (ext4_has_feature_quota(sb)) {
1287 		ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1288 			 "when QUOTA feature is enabled");
1289 		return -1;
1290 	}
1291 	qname = match_strdup(args);
1292 	if (!qname) {
1293 		ext4_msg(sb, KERN_ERR,
1294 			"Not enough memory for storing quotafile name");
1295 		return -1;
1296 	}
1297 	if (sbi->s_qf_names[qtype]) {
1298 		if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1299 			ret = 1;
1300 		else
1301 			ext4_msg(sb, KERN_ERR,
1302 				 "%s quota file already specified",
1303 				 QTYPE2NAME(qtype));
1304 		goto errout;
1305 	}
1306 	if (strchr(qname, '/')) {
1307 		ext4_msg(sb, KERN_ERR,
1308 			"quotafile must be on filesystem root");
1309 		goto errout;
1310 	}
1311 	sbi->s_qf_names[qtype] = qname;
1312 	set_opt(sb, QUOTA);
1313 	return 1;
1314 errout:
1315 	kfree(qname);
1316 	return ret;
1317 }
1318 
1319 static int clear_qf_name(struct super_block *sb, int qtype)
1320 {
1321 
1322 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1323 
1324 	if (sb_any_quota_loaded(sb) &&
1325 		sbi->s_qf_names[qtype]) {
1326 		ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1327 			" when quota turned on");
1328 		return -1;
1329 	}
1330 	kfree(sbi->s_qf_names[qtype]);
1331 	sbi->s_qf_names[qtype] = NULL;
1332 	return 1;
1333 }
1334 #endif
1335 
1336 #define MOPT_SET	0x0001
1337 #define MOPT_CLEAR	0x0002
1338 #define MOPT_NOSUPPORT	0x0004
1339 #define MOPT_EXPLICIT	0x0008
1340 #define MOPT_CLEAR_ERR	0x0010
1341 #define MOPT_GTE0	0x0020
1342 #ifdef CONFIG_QUOTA
1343 #define MOPT_Q		0
1344 #define MOPT_QFMT	0x0040
1345 #else
1346 #define MOPT_Q		MOPT_NOSUPPORT
1347 #define MOPT_QFMT	MOPT_NOSUPPORT
1348 #endif
1349 #define MOPT_DATAJ	0x0080
1350 #define MOPT_NO_EXT2	0x0100
1351 #define MOPT_NO_EXT3	0x0200
1352 #define MOPT_EXT4_ONLY	(MOPT_NO_EXT2 | MOPT_NO_EXT3)
1353 #define MOPT_STRING	0x0400
1354 
1355 static const struct mount_opts {
1356 	int	token;
1357 	int	mount_opt;
1358 	int	flags;
1359 } ext4_mount_opts[] = {
1360 	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1361 	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1362 	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1363 	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1364 	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1365 	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1366 	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1367 	 MOPT_EXT4_ONLY | MOPT_SET},
1368 	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1369 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1370 	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1371 	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1372 	{Opt_delalloc, EXT4_MOUNT_DELALLOC,
1373 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1374 	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1375 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1376 	{Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1377 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1378 	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1379 	 MOPT_EXT4_ONLY | MOPT_SET},
1380 	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1381 				    EXT4_MOUNT_JOURNAL_CHECKSUM),
1382 	 MOPT_EXT4_ONLY | MOPT_SET},
1383 	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1384 	{Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1385 	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1386 	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1387 	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1388 	 MOPT_NO_EXT2 | MOPT_SET},
1389 	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1390 	 MOPT_NO_EXT2 | MOPT_CLEAR},
1391 	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1392 	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1393 	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1394 	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1395 	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1396 	{Opt_commit, 0, MOPT_GTE0},
1397 	{Opt_max_batch_time, 0, MOPT_GTE0},
1398 	{Opt_min_batch_time, 0, MOPT_GTE0},
1399 	{Opt_inode_readahead_blks, 0, MOPT_GTE0},
1400 	{Opt_init_itable, 0, MOPT_GTE0},
1401 	{Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1402 	{Opt_stripe, 0, MOPT_GTE0},
1403 	{Opt_resuid, 0, MOPT_GTE0},
1404 	{Opt_resgid, 0, MOPT_GTE0},
1405 	{Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1406 	{Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1407 	{Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1408 	{Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1409 	{Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1410 	{Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1411 	 MOPT_NO_EXT2 | MOPT_DATAJ},
1412 	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1413 	{Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1414 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1415 	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1416 	{Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1417 #else
1418 	{Opt_acl, 0, MOPT_NOSUPPORT},
1419 	{Opt_noacl, 0, MOPT_NOSUPPORT},
1420 #endif
1421 	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1422 	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1423 	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1424 	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1425 							MOPT_SET | MOPT_Q},
1426 	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1427 							MOPT_SET | MOPT_Q},
1428 	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1429 		       EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1430 	{Opt_usrjquota, 0, MOPT_Q},
1431 	{Opt_grpjquota, 0, MOPT_Q},
1432 	{Opt_offusrjquota, 0, MOPT_Q},
1433 	{Opt_offgrpjquota, 0, MOPT_Q},
1434 	{Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1435 	{Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1436 	{Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1437 	{Opt_max_dir_size_kb, 0, MOPT_GTE0},
1438 	{Opt_test_dummy_encryption, 0, MOPT_GTE0},
1439 	{Opt_err, 0, 0}
1440 };
1441 
1442 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1443 			    substring_t *args, unsigned long *journal_devnum,
1444 			    unsigned int *journal_ioprio, int is_remount)
1445 {
1446 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1447 	const struct mount_opts *m;
1448 	kuid_t uid;
1449 	kgid_t gid;
1450 	int arg = 0;
1451 
1452 #ifdef CONFIG_QUOTA
1453 	if (token == Opt_usrjquota)
1454 		return set_qf_name(sb, USRQUOTA, &args[0]);
1455 	else if (token == Opt_grpjquota)
1456 		return set_qf_name(sb, GRPQUOTA, &args[0]);
1457 	else if (token == Opt_offusrjquota)
1458 		return clear_qf_name(sb, USRQUOTA);
1459 	else if (token == Opt_offgrpjquota)
1460 		return clear_qf_name(sb, GRPQUOTA);
1461 #endif
1462 	switch (token) {
1463 	case Opt_noacl:
1464 	case Opt_nouser_xattr:
1465 		ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1466 		break;
1467 	case Opt_sb:
1468 		return 1;	/* handled by get_sb_block() */
1469 	case Opt_removed:
1470 		ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1471 		return 1;
1472 	case Opt_abort:
1473 		sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1474 		return 1;
1475 	case Opt_i_version:
1476 		sb->s_flags |= MS_I_VERSION;
1477 		return 1;
1478 	case Opt_lazytime:
1479 		sb->s_flags |= MS_LAZYTIME;
1480 		return 1;
1481 	case Opt_nolazytime:
1482 		sb->s_flags &= ~MS_LAZYTIME;
1483 		return 1;
1484 	}
1485 
1486 	for (m = ext4_mount_opts; m->token != Opt_err; m++)
1487 		if (token == m->token)
1488 			break;
1489 
1490 	if (m->token == Opt_err) {
1491 		ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1492 			 "or missing value", opt);
1493 		return -1;
1494 	}
1495 
1496 	if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1497 		ext4_msg(sb, KERN_ERR,
1498 			 "Mount option \"%s\" incompatible with ext2", opt);
1499 		return -1;
1500 	}
1501 	if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1502 		ext4_msg(sb, KERN_ERR,
1503 			 "Mount option \"%s\" incompatible with ext3", opt);
1504 		return -1;
1505 	}
1506 
1507 	if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1508 		return -1;
1509 	if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1510 		return -1;
1511 	if (m->flags & MOPT_EXPLICIT)
1512 		set_opt2(sb, EXPLICIT_DELALLOC);
1513 	if (m->flags & MOPT_CLEAR_ERR)
1514 		clear_opt(sb, ERRORS_MASK);
1515 	if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1516 		ext4_msg(sb, KERN_ERR, "Cannot change quota "
1517 			 "options when quota turned on");
1518 		return -1;
1519 	}
1520 
1521 	if (m->flags & MOPT_NOSUPPORT) {
1522 		ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1523 	} else if (token == Opt_commit) {
1524 		if (arg == 0)
1525 			arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1526 		sbi->s_commit_interval = HZ * arg;
1527 	} else if (token == Opt_max_batch_time) {
1528 		sbi->s_max_batch_time = arg;
1529 	} else if (token == Opt_min_batch_time) {
1530 		sbi->s_min_batch_time = arg;
1531 	} else if (token == Opt_inode_readahead_blks) {
1532 		if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1533 			ext4_msg(sb, KERN_ERR,
1534 				 "EXT4-fs: inode_readahead_blks must be "
1535 				 "0 or a power of 2 smaller than 2^31");
1536 			return -1;
1537 		}
1538 		sbi->s_inode_readahead_blks = arg;
1539 	} else if (token == Opt_init_itable) {
1540 		set_opt(sb, INIT_INODE_TABLE);
1541 		if (!args->from)
1542 			arg = EXT4_DEF_LI_WAIT_MULT;
1543 		sbi->s_li_wait_mult = arg;
1544 	} else if (token == Opt_max_dir_size_kb) {
1545 		sbi->s_max_dir_size_kb = arg;
1546 	} else if (token == Opt_stripe) {
1547 		sbi->s_stripe = arg;
1548 	} else if (token == Opt_resuid) {
1549 		uid = make_kuid(current_user_ns(), arg);
1550 		if (!uid_valid(uid)) {
1551 			ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1552 			return -1;
1553 		}
1554 		sbi->s_resuid = uid;
1555 	} else if (token == Opt_resgid) {
1556 		gid = make_kgid(current_user_ns(), arg);
1557 		if (!gid_valid(gid)) {
1558 			ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1559 			return -1;
1560 		}
1561 		sbi->s_resgid = gid;
1562 	} else if (token == Opt_journal_dev) {
1563 		if (is_remount) {
1564 			ext4_msg(sb, KERN_ERR,
1565 				 "Cannot specify journal on remount");
1566 			return -1;
1567 		}
1568 		*journal_devnum = arg;
1569 	} else if (token == Opt_journal_path) {
1570 		char *journal_path;
1571 		struct inode *journal_inode;
1572 		struct path path;
1573 		int error;
1574 
1575 		if (is_remount) {
1576 			ext4_msg(sb, KERN_ERR,
1577 				 "Cannot specify journal on remount");
1578 			return -1;
1579 		}
1580 		journal_path = match_strdup(&args[0]);
1581 		if (!journal_path) {
1582 			ext4_msg(sb, KERN_ERR, "error: could not dup "
1583 				"journal device string");
1584 			return -1;
1585 		}
1586 
1587 		error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1588 		if (error) {
1589 			ext4_msg(sb, KERN_ERR, "error: could not find "
1590 				"journal device path: error %d", error);
1591 			kfree(journal_path);
1592 			return -1;
1593 		}
1594 
1595 		journal_inode = d_inode(path.dentry);
1596 		if (!S_ISBLK(journal_inode->i_mode)) {
1597 			ext4_msg(sb, KERN_ERR, "error: journal path %s "
1598 				"is not a block device", journal_path);
1599 			path_put(&path);
1600 			kfree(journal_path);
1601 			return -1;
1602 		}
1603 
1604 		*journal_devnum = new_encode_dev(journal_inode->i_rdev);
1605 		path_put(&path);
1606 		kfree(journal_path);
1607 	} else if (token == Opt_journal_ioprio) {
1608 		if (arg > 7) {
1609 			ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1610 				 " (must be 0-7)");
1611 			return -1;
1612 		}
1613 		*journal_ioprio =
1614 			IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1615 	} else if (token == Opt_test_dummy_encryption) {
1616 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1617 		sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1618 		ext4_msg(sb, KERN_WARNING,
1619 			 "Test dummy encryption mode enabled");
1620 #else
1621 		ext4_msg(sb, KERN_WARNING,
1622 			 "Test dummy encryption mount option ignored");
1623 #endif
1624 	} else if (m->flags & MOPT_DATAJ) {
1625 		if (is_remount) {
1626 			if (!sbi->s_journal)
1627 				ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1628 			else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1629 				ext4_msg(sb, KERN_ERR,
1630 					 "Cannot change data mode on remount");
1631 				return -1;
1632 			}
1633 		} else {
1634 			clear_opt(sb, DATA_FLAGS);
1635 			sbi->s_mount_opt |= m->mount_opt;
1636 		}
1637 #ifdef CONFIG_QUOTA
1638 	} else if (m->flags & MOPT_QFMT) {
1639 		if (sb_any_quota_loaded(sb) &&
1640 		    sbi->s_jquota_fmt != m->mount_opt) {
1641 			ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1642 				 "quota options when quota turned on");
1643 			return -1;
1644 		}
1645 		if (ext4_has_feature_quota(sb)) {
1646 			ext4_msg(sb, KERN_ERR,
1647 				 "Cannot set journaled quota options "
1648 				 "when QUOTA feature is enabled");
1649 			return -1;
1650 		}
1651 		sbi->s_jquota_fmt = m->mount_opt;
1652 #endif
1653 #ifndef CONFIG_FS_DAX
1654 	} else if (token == Opt_dax) {
1655 		ext4_msg(sb, KERN_INFO, "dax option not supported");
1656 		return -1;
1657 #endif
1658 	} else {
1659 		if (!args->from)
1660 			arg = 1;
1661 		if (m->flags & MOPT_CLEAR)
1662 			arg = !arg;
1663 		else if (unlikely(!(m->flags & MOPT_SET))) {
1664 			ext4_msg(sb, KERN_WARNING,
1665 				 "buggy handling of option %s", opt);
1666 			WARN_ON(1);
1667 			return -1;
1668 		}
1669 		if (arg != 0)
1670 			sbi->s_mount_opt |= m->mount_opt;
1671 		else
1672 			sbi->s_mount_opt &= ~m->mount_opt;
1673 	}
1674 	return 1;
1675 }
1676 
1677 static int parse_options(char *options, struct super_block *sb,
1678 			 unsigned long *journal_devnum,
1679 			 unsigned int *journal_ioprio,
1680 			 int is_remount)
1681 {
1682 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1683 	char *p;
1684 	substring_t args[MAX_OPT_ARGS];
1685 	int token;
1686 
1687 	if (!options)
1688 		return 1;
1689 
1690 	while ((p = strsep(&options, ",")) != NULL) {
1691 		if (!*p)
1692 			continue;
1693 		/*
1694 		 * Initialize args struct so we know whether arg was
1695 		 * found; some options take optional arguments.
1696 		 */
1697 		args[0].to = args[0].from = NULL;
1698 		token = match_token(p, tokens, args);
1699 		if (handle_mount_opt(sb, p, token, args, journal_devnum,
1700 				     journal_ioprio, is_remount) < 0)
1701 			return 0;
1702 	}
1703 #ifdef CONFIG_QUOTA
1704 	if (ext4_has_feature_quota(sb) &&
1705 	    (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1706 		ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1707 			 "feature is enabled");
1708 		return 0;
1709 	}
1710 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1711 		if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1712 			clear_opt(sb, USRQUOTA);
1713 
1714 		if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1715 			clear_opt(sb, GRPQUOTA);
1716 
1717 		if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1718 			ext4_msg(sb, KERN_ERR, "old and new quota "
1719 					"format mixing");
1720 			return 0;
1721 		}
1722 
1723 		if (!sbi->s_jquota_fmt) {
1724 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1725 					"not specified");
1726 			return 0;
1727 		}
1728 	}
1729 #endif
1730 	if (test_opt(sb, DIOREAD_NOLOCK)) {
1731 		int blocksize =
1732 			BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1733 
1734 		if (blocksize < PAGE_CACHE_SIZE) {
1735 			ext4_msg(sb, KERN_ERR, "can't mount with "
1736 				 "dioread_nolock if block size != PAGE_SIZE");
1737 			return 0;
1738 		}
1739 	}
1740 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1741 	    test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1742 		ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1743 			 "in data=ordered mode");
1744 		return 0;
1745 	}
1746 	return 1;
1747 }
1748 
1749 static inline void ext4_show_quota_options(struct seq_file *seq,
1750 					   struct super_block *sb)
1751 {
1752 #if defined(CONFIG_QUOTA)
1753 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1754 
1755 	if (sbi->s_jquota_fmt) {
1756 		char *fmtname = "";
1757 
1758 		switch (sbi->s_jquota_fmt) {
1759 		case QFMT_VFS_OLD:
1760 			fmtname = "vfsold";
1761 			break;
1762 		case QFMT_VFS_V0:
1763 			fmtname = "vfsv0";
1764 			break;
1765 		case QFMT_VFS_V1:
1766 			fmtname = "vfsv1";
1767 			break;
1768 		}
1769 		seq_printf(seq, ",jqfmt=%s", fmtname);
1770 	}
1771 
1772 	if (sbi->s_qf_names[USRQUOTA])
1773 		seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1774 
1775 	if (sbi->s_qf_names[GRPQUOTA])
1776 		seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1777 #endif
1778 }
1779 
1780 static const char *token2str(int token)
1781 {
1782 	const struct match_token *t;
1783 
1784 	for (t = tokens; t->token != Opt_err; t++)
1785 		if (t->token == token && !strchr(t->pattern, '='))
1786 			break;
1787 	return t->pattern;
1788 }
1789 
1790 /*
1791  * Show an option if
1792  *  - it's set to a non-default value OR
1793  *  - if the per-sb default is different from the global default
1794  */
1795 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1796 			      int nodefs)
1797 {
1798 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1799 	struct ext4_super_block *es = sbi->s_es;
1800 	int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1801 	const struct mount_opts *m;
1802 	char sep = nodefs ? '\n' : ',';
1803 
1804 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1805 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1806 
1807 	if (sbi->s_sb_block != 1)
1808 		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1809 
1810 	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1811 		int want_set = m->flags & MOPT_SET;
1812 		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1813 		    (m->flags & MOPT_CLEAR_ERR))
1814 			continue;
1815 		if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1816 			continue; /* skip if same as the default */
1817 		if ((want_set &&
1818 		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1819 		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1820 			continue; /* select Opt_noFoo vs Opt_Foo */
1821 		SEQ_OPTS_PRINT("%s", token2str(m->token));
1822 	}
1823 
1824 	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1825 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1826 		SEQ_OPTS_PRINT("resuid=%u",
1827 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
1828 	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1829 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1830 		SEQ_OPTS_PRINT("resgid=%u",
1831 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
1832 	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1833 	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1834 		SEQ_OPTS_PUTS("errors=remount-ro");
1835 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1836 		SEQ_OPTS_PUTS("errors=continue");
1837 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1838 		SEQ_OPTS_PUTS("errors=panic");
1839 	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1840 		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1841 	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1842 		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1843 	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1844 		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1845 	if (sb->s_flags & MS_I_VERSION)
1846 		SEQ_OPTS_PUTS("i_version");
1847 	if (nodefs || sbi->s_stripe)
1848 		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1849 	if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1850 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1851 			SEQ_OPTS_PUTS("data=journal");
1852 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1853 			SEQ_OPTS_PUTS("data=ordered");
1854 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1855 			SEQ_OPTS_PUTS("data=writeback");
1856 	}
1857 	if (nodefs ||
1858 	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1859 		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1860 			       sbi->s_inode_readahead_blks);
1861 
1862 	if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1863 		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1864 		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1865 	if (nodefs || sbi->s_max_dir_size_kb)
1866 		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1867 
1868 	ext4_show_quota_options(seq, sb);
1869 	return 0;
1870 }
1871 
1872 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1873 {
1874 	return _ext4_show_options(seq, root->d_sb, 0);
1875 }
1876 
1877 int ext4_seq_options_show(struct seq_file *seq, void *offset)
1878 {
1879 	struct super_block *sb = seq->private;
1880 	int rc;
1881 
1882 	seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1883 	rc = _ext4_show_options(seq, sb, 1);
1884 	seq_puts(seq, "\n");
1885 	return rc;
1886 }
1887 
1888 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1889 			    int read_only)
1890 {
1891 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1892 	int res = 0;
1893 
1894 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1895 		ext4_msg(sb, KERN_ERR, "revision level too high, "
1896 			 "forcing read-only mode");
1897 		res = MS_RDONLY;
1898 	}
1899 	if (read_only)
1900 		goto done;
1901 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1902 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1903 			 "running e2fsck is recommended");
1904 	else if (sbi->s_mount_state & EXT4_ERROR_FS)
1905 		ext4_msg(sb, KERN_WARNING,
1906 			 "warning: mounting fs with errors, "
1907 			 "running e2fsck is recommended");
1908 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1909 		 le16_to_cpu(es->s_mnt_count) >=
1910 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1911 		ext4_msg(sb, KERN_WARNING,
1912 			 "warning: maximal mount count reached, "
1913 			 "running e2fsck is recommended");
1914 	else if (le32_to_cpu(es->s_checkinterval) &&
1915 		(le32_to_cpu(es->s_lastcheck) +
1916 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1917 		ext4_msg(sb, KERN_WARNING,
1918 			 "warning: checktime reached, "
1919 			 "running e2fsck is recommended");
1920 	if (!sbi->s_journal)
1921 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1922 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1923 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1924 	le16_add_cpu(&es->s_mnt_count, 1);
1925 	es->s_mtime = cpu_to_le32(get_seconds());
1926 	ext4_update_dynamic_rev(sb);
1927 	if (sbi->s_journal)
1928 		ext4_set_feature_journal_needs_recovery(sb);
1929 
1930 	ext4_commit_super(sb, 1);
1931 done:
1932 	if (test_opt(sb, DEBUG))
1933 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1934 				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1935 			sb->s_blocksize,
1936 			sbi->s_groups_count,
1937 			EXT4_BLOCKS_PER_GROUP(sb),
1938 			EXT4_INODES_PER_GROUP(sb),
1939 			sbi->s_mount_opt, sbi->s_mount_opt2);
1940 
1941 	cleancache_init_fs(sb);
1942 	return res;
1943 }
1944 
1945 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1946 {
1947 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1948 	struct flex_groups *new_groups;
1949 	int size;
1950 
1951 	if (!sbi->s_log_groups_per_flex)
1952 		return 0;
1953 
1954 	size = ext4_flex_group(sbi, ngroup - 1) + 1;
1955 	if (size <= sbi->s_flex_groups_allocated)
1956 		return 0;
1957 
1958 	size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1959 	new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1960 	if (!new_groups) {
1961 		ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1962 			 size / (int) sizeof(struct flex_groups));
1963 		return -ENOMEM;
1964 	}
1965 
1966 	if (sbi->s_flex_groups) {
1967 		memcpy(new_groups, sbi->s_flex_groups,
1968 		       (sbi->s_flex_groups_allocated *
1969 			sizeof(struct flex_groups)));
1970 		kvfree(sbi->s_flex_groups);
1971 	}
1972 	sbi->s_flex_groups = new_groups;
1973 	sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1974 	return 0;
1975 }
1976 
1977 static int ext4_fill_flex_info(struct super_block *sb)
1978 {
1979 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1980 	struct ext4_group_desc *gdp = NULL;
1981 	ext4_group_t flex_group;
1982 	int i, err;
1983 
1984 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1985 	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1986 		sbi->s_log_groups_per_flex = 0;
1987 		return 1;
1988 	}
1989 
1990 	err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1991 	if (err)
1992 		goto failed;
1993 
1994 	for (i = 0; i < sbi->s_groups_count; i++) {
1995 		gdp = ext4_get_group_desc(sb, i, NULL);
1996 
1997 		flex_group = ext4_flex_group(sbi, i);
1998 		atomic_add(ext4_free_inodes_count(sb, gdp),
1999 			   &sbi->s_flex_groups[flex_group].free_inodes);
2000 		atomic64_add(ext4_free_group_clusters(sb, gdp),
2001 			     &sbi->s_flex_groups[flex_group].free_clusters);
2002 		atomic_add(ext4_used_dirs_count(sb, gdp),
2003 			   &sbi->s_flex_groups[flex_group].used_dirs);
2004 	}
2005 
2006 	return 1;
2007 failed:
2008 	return 0;
2009 }
2010 
2011 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2012 				   struct ext4_group_desc *gdp)
2013 {
2014 	int offset;
2015 	__u16 crc = 0;
2016 	__le32 le_group = cpu_to_le32(block_group);
2017 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2018 
2019 	if (ext4_has_metadata_csum(sbi->s_sb)) {
2020 		/* Use new metadata_csum algorithm */
2021 		__le16 save_csum;
2022 		__u32 csum32;
2023 
2024 		save_csum = gdp->bg_checksum;
2025 		gdp->bg_checksum = 0;
2026 		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2027 				     sizeof(le_group));
2028 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2029 				     sbi->s_desc_size);
2030 		gdp->bg_checksum = save_csum;
2031 
2032 		crc = csum32 & 0xFFFF;
2033 		goto out;
2034 	}
2035 
2036 	/* old crc16 code */
2037 	if (!ext4_has_feature_gdt_csum(sb))
2038 		return 0;
2039 
2040 	offset = offsetof(struct ext4_group_desc, bg_checksum);
2041 
2042 	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2043 	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2044 	crc = crc16(crc, (__u8 *)gdp, offset);
2045 	offset += sizeof(gdp->bg_checksum); /* skip checksum */
2046 	/* for checksum of struct ext4_group_desc do the rest...*/
2047 	if (ext4_has_feature_64bit(sb) &&
2048 	    offset < le16_to_cpu(sbi->s_es->s_desc_size))
2049 		crc = crc16(crc, (__u8 *)gdp + offset,
2050 			    le16_to_cpu(sbi->s_es->s_desc_size) -
2051 				offset);
2052 
2053 out:
2054 	return cpu_to_le16(crc);
2055 }
2056 
2057 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2058 				struct ext4_group_desc *gdp)
2059 {
2060 	if (ext4_has_group_desc_csum(sb) &&
2061 	    (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2062 		return 0;
2063 
2064 	return 1;
2065 }
2066 
2067 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2068 			      struct ext4_group_desc *gdp)
2069 {
2070 	if (!ext4_has_group_desc_csum(sb))
2071 		return;
2072 	gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2073 }
2074 
2075 /* Called at mount-time, super-block is locked */
2076 static int ext4_check_descriptors(struct super_block *sb,
2077 				  ext4_group_t *first_not_zeroed)
2078 {
2079 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2080 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2081 	ext4_fsblk_t last_block;
2082 	ext4_fsblk_t block_bitmap;
2083 	ext4_fsblk_t inode_bitmap;
2084 	ext4_fsblk_t inode_table;
2085 	int flexbg_flag = 0;
2086 	ext4_group_t i, grp = sbi->s_groups_count;
2087 
2088 	if (ext4_has_feature_flex_bg(sb))
2089 		flexbg_flag = 1;
2090 
2091 	ext4_debug("Checking group descriptors");
2092 
2093 	for (i = 0; i < sbi->s_groups_count; i++) {
2094 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2095 
2096 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
2097 			last_block = ext4_blocks_count(sbi->s_es) - 1;
2098 		else
2099 			last_block = first_block +
2100 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
2101 
2102 		if ((grp == sbi->s_groups_count) &&
2103 		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2104 			grp = i;
2105 
2106 		block_bitmap = ext4_block_bitmap(sb, gdp);
2107 		if (block_bitmap < first_block || block_bitmap > last_block) {
2108 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2109 			       "Block bitmap for group %u not in group "
2110 			       "(block %llu)!", i, block_bitmap);
2111 			return 0;
2112 		}
2113 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
2114 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
2115 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2116 			       "Inode bitmap for group %u not in group "
2117 			       "(block %llu)!", i, inode_bitmap);
2118 			return 0;
2119 		}
2120 		inode_table = ext4_inode_table(sb, gdp);
2121 		if (inode_table < first_block ||
2122 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2123 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2124 			       "Inode table for group %u not in group "
2125 			       "(block %llu)!", i, inode_table);
2126 			return 0;
2127 		}
2128 		ext4_lock_group(sb, i);
2129 		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2130 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2131 				 "Checksum for group %u failed (%u!=%u)",
2132 				 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2133 				     gdp)), le16_to_cpu(gdp->bg_checksum));
2134 			if (!(sb->s_flags & MS_RDONLY)) {
2135 				ext4_unlock_group(sb, i);
2136 				return 0;
2137 			}
2138 		}
2139 		ext4_unlock_group(sb, i);
2140 		if (!flexbg_flag)
2141 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
2142 	}
2143 	if (NULL != first_not_zeroed)
2144 		*first_not_zeroed = grp;
2145 	return 1;
2146 }
2147 
2148 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2149  * the superblock) which were deleted from all directories, but held open by
2150  * a process at the time of a crash.  We walk the list and try to delete these
2151  * inodes at recovery time (only with a read-write filesystem).
2152  *
2153  * In order to keep the orphan inode chain consistent during traversal (in
2154  * case of crash during recovery), we link each inode into the superblock
2155  * orphan list_head and handle it the same way as an inode deletion during
2156  * normal operation (which journals the operations for us).
2157  *
2158  * We only do an iget() and an iput() on each inode, which is very safe if we
2159  * accidentally point at an in-use or already deleted inode.  The worst that
2160  * can happen in this case is that we get a "bit already cleared" message from
2161  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2162  * e2fsck was run on this filesystem, and it must have already done the orphan
2163  * inode cleanup for us, so we can safely abort without any further action.
2164  */
2165 static void ext4_orphan_cleanup(struct super_block *sb,
2166 				struct ext4_super_block *es)
2167 {
2168 	unsigned int s_flags = sb->s_flags;
2169 	int nr_orphans = 0, nr_truncates = 0;
2170 #ifdef CONFIG_QUOTA
2171 	int i;
2172 #endif
2173 	if (!es->s_last_orphan) {
2174 		jbd_debug(4, "no orphan inodes to clean up\n");
2175 		return;
2176 	}
2177 
2178 	if (bdev_read_only(sb->s_bdev)) {
2179 		ext4_msg(sb, KERN_ERR, "write access "
2180 			"unavailable, skipping orphan cleanup");
2181 		return;
2182 	}
2183 
2184 	/* Check if feature set would not allow a r/w mount */
2185 	if (!ext4_feature_set_ok(sb, 0)) {
2186 		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2187 			 "unknown ROCOMPAT features");
2188 		return;
2189 	}
2190 
2191 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2192 		/* don't clear list on RO mount w/ errors */
2193 		if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2194 			ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2195 				  "clearing orphan list.\n");
2196 			es->s_last_orphan = 0;
2197 		}
2198 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2199 		return;
2200 	}
2201 
2202 	if (s_flags & MS_RDONLY) {
2203 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2204 		sb->s_flags &= ~MS_RDONLY;
2205 	}
2206 #ifdef CONFIG_QUOTA
2207 	/* Needed for iput() to work correctly and not trash data */
2208 	sb->s_flags |= MS_ACTIVE;
2209 	/* Turn on quotas so that they are updated correctly */
2210 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2211 		if (EXT4_SB(sb)->s_qf_names[i]) {
2212 			int ret = ext4_quota_on_mount(sb, i);
2213 			if (ret < 0)
2214 				ext4_msg(sb, KERN_ERR,
2215 					"Cannot turn on journaled "
2216 					"quota: error %d", ret);
2217 		}
2218 	}
2219 #endif
2220 
2221 	while (es->s_last_orphan) {
2222 		struct inode *inode;
2223 
2224 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2225 		if (IS_ERR(inode)) {
2226 			es->s_last_orphan = 0;
2227 			break;
2228 		}
2229 
2230 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2231 		dquot_initialize(inode);
2232 		if (inode->i_nlink) {
2233 			if (test_opt(sb, DEBUG))
2234 				ext4_msg(sb, KERN_DEBUG,
2235 					"%s: truncating inode %lu to %lld bytes",
2236 					__func__, inode->i_ino, inode->i_size);
2237 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2238 				  inode->i_ino, inode->i_size);
2239 			mutex_lock(&inode->i_mutex);
2240 			truncate_inode_pages(inode->i_mapping, inode->i_size);
2241 			ext4_truncate(inode);
2242 			mutex_unlock(&inode->i_mutex);
2243 			nr_truncates++;
2244 		} else {
2245 			if (test_opt(sb, DEBUG))
2246 				ext4_msg(sb, KERN_DEBUG,
2247 					"%s: deleting unreferenced inode %lu",
2248 					__func__, inode->i_ino);
2249 			jbd_debug(2, "deleting unreferenced inode %lu\n",
2250 				  inode->i_ino);
2251 			nr_orphans++;
2252 		}
2253 		iput(inode);  /* The delete magic happens here! */
2254 	}
2255 
2256 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2257 
2258 	if (nr_orphans)
2259 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2260 		       PLURAL(nr_orphans));
2261 	if (nr_truncates)
2262 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2263 		       PLURAL(nr_truncates));
2264 #ifdef CONFIG_QUOTA
2265 	/* Turn quotas off */
2266 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2267 		if (sb_dqopt(sb)->files[i])
2268 			dquot_quota_off(sb, i);
2269 	}
2270 #endif
2271 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2272 }
2273 
2274 /*
2275  * Maximal extent format file size.
2276  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2277  * extent format containers, within a sector_t, and within i_blocks
2278  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2279  * so that won't be a limiting factor.
2280  *
2281  * However there is other limiting factor. We do store extents in the form
2282  * of starting block and length, hence the resulting length of the extent
2283  * covering maximum file size must fit into on-disk format containers as
2284  * well. Given that length is always by 1 unit bigger than max unit (because
2285  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2286  *
2287  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2288  */
2289 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2290 {
2291 	loff_t res;
2292 	loff_t upper_limit = MAX_LFS_FILESIZE;
2293 
2294 	/* small i_blocks in vfs inode? */
2295 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2296 		/*
2297 		 * CONFIG_LBDAF is not enabled implies the inode
2298 		 * i_block represent total blocks in 512 bytes
2299 		 * 32 == size of vfs inode i_blocks * 8
2300 		 */
2301 		upper_limit = (1LL << 32) - 1;
2302 
2303 		/* total blocks in file system block size */
2304 		upper_limit >>= (blkbits - 9);
2305 		upper_limit <<= blkbits;
2306 	}
2307 
2308 	/*
2309 	 * 32-bit extent-start container, ee_block. We lower the maxbytes
2310 	 * by one fs block, so ee_len can cover the extent of maximum file
2311 	 * size
2312 	 */
2313 	res = (1LL << 32) - 1;
2314 	res <<= blkbits;
2315 
2316 	/* Sanity check against vm- & vfs- imposed limits */
2317 	if (res > upper_limit)
2318 		res = upper_limit;
2319 
2320 	return res;
2321 }
2322 
2323 /*
2324  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2325  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2326  * We need to be 1 filesystem block less than the 2^48 sector limit.
2327  */
2328 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2329 {
2330 	loff_t res = EXT4_NDIR_BLOCKS;
2331 	int meta_blocks;
2332 	loff_t upper_limit;
2333 	/* This is calculated to be the largest file size for a dense, block
2334 	 * mapped file such that the file's total number of 512-byte sectors,
2335 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
2336 	 *
2337 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2338 	 * number of 512-byte sectors of the file.
2339 	 */
2340 
2341 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2342 		/*
2343 		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2344 		 * the inode i_block field represents total file blocks in
2345 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2346 		 */
2347 		upper_limit = (1LL << 32) - 1;
2348 
2349 		/* total blocks in file system block size */
2350 		upper_limit >>= (bits - 9);
2351 
2352 	} else {
2353 		/*
2354 		 * We use 48 bit ext4_inode i_blocks
2355 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2356 		 * represent total number of blocks in
2357 		 * file system block size
2358 		 */
2359 		upper_limit = (1LL << 48) - 1;
2360 
2361 	}
2362 
2363 	/* indirect blocks */
2364 	meta_blocks = 1;
2365 	/* double indirect blocks */
2366 	meta_blocks += 1 + (1LL << (bits-2));
2367 	/* tripple indirect blocks */
2368 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2369 
2370 	upper_limit -= meta_blocks;
2371 	upper_limit <<= bits;
2372 
2373 	res += 1LL << (bits-2);
2374 	res += 1LL << (2*(bits-2));
2375 	res += 1LL << (3*(bits-2));
2376 	res <<= bits;
2377 	if (res > upper_limit)
2378 		res = upper_limit;
2379 
2380 	if (res > MAX_LFS_FILESIZE)
2381 		res = MAX_LFS_FILESIZE;
2382 
2383 	return res;
2384 }
2385 
2386 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2387 				   ext4_fsblk_t logical_sb_block, int nr)
2388 {
2389 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2390 	ext4_group_t bg, first_meta_bg;
2391 	int has_super = 0;
2392 
2393 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2394 
2395 	if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2396 		return logical_sb_block + nr + 1;
2397 	bg = sbi->s_desc_per_block * nr;
2398 	if (ext4_bg_has_super(sb, bg))
2399 		has_super = 1;
2400 
2401 	/*
2402 	 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2403 	 * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
2404 	 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2405 	 * compensate.
2406 	 */
2407 	if (sb->s_blocksize == 1024 && nr == 0 &&
2408 	    le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2409 		has_super++;
2410 
2411 	return (has_super + ext4_group_first_block_no(sb, bg));
2412 }
2413 
2414 /**
2415  * ext4_get_stripe_size: Get the stripe size.
2416  * @sbi: In memory super block info
2417  *
2418  * If we have specified it via mount option, then
2419  * use the mount option value. If the value specified at mount time is
2420  * greater than the blocks per group use the super block value.
2421  * If the super block value is greater than blocks per group return 0.
2422  * Allocator needs it be less than blocks per group.
2423  *
2424  */
2425 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2426 {
2427 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2428 	unsigned long stripe_width =
2429 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2430 	int ret;
2431 
2432 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2433 		ret = sbi->s_stripe;
2434 	else if (stripe_width <= sbi->s_blocks_per_group)
2435 		ret = stripe_width;
2436 	else if (stride <= sbi->s_blocks_per_group)
2437 		ret = stride;
2438 	else
2439 		ret = 0;
2440 
2441 	/*
2442 	 * If the stripe width is 1, this makes no sense and
2443 	 * we set it to 0 to turn off stripe handling code.
2444 	 */
2445 	if (ret <= 1)
2446 		ret = 0;
2447 
2448 	return ret;
2449 }
2450 
2451 /*
2452  * Check whether this filesystem can be mounted based on
2453  * the features present and the RDONLY/RDWR mount requested.
2454  * Returns 1 if this filesystem can be mounted as requested,
2455  * 0 if it cannot be.
2456  */
2457 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2458 {
2459 	if (ext4_has_unknown_ext4_incompat_features(sb)) {
2460 		ext4_msg(sb, KERN_ERR,
2461 			"Couldn't mount because of "
2462 			"unsupported optional features (%x)",
2463 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2464 			~EXT4_FEATURE_INCOMPAT_SUPP));
2465 		return 0;
2466 	}
2467 
2468 	if (readonly)
2469 		return 1;
2470 
2471 	if (ext4_has_feature_readonly(sb)) {
2472 		ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2473 		sb->s_flags |= MS_RDONLY;
2474 		return 1;
2475 	}
2476 
2477 	/* Check that feature set is OK for a read-write mount */
2478 	if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2479 		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2480 			 "unsupported optional features (%x)",
2481 			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2482 				~EXT4_FEATURE_RO_COMPAT_SUPP));
2483 		return 0;
2484 	}
2485 	/*
2486 	 * Large file size enabled file system can only be mounted
2487 	 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2488 	 */
2489 	if (ext4_has_feature_huge_file(sb)) {
2490 		if (sizeof(blkcnt_t) < sizeof(u64)) {
2491 			ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2492 				 "cannot be mounted RDWR without "
2493 				 "CONFIG_LBDAF");
2494 			return 0;
2495 		}
2496 	}
2497 	if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2498 		ext4_msg(sb, KERN_ERR,
2499 			 "Can't support bigalloc feature without "
2500 			 "extents feature\n");
2501 		return 0;
2502 	}
2503 
2504 #ifndef CONFIG_QUOTA
2505 	if (ext4_has_feature_quota(sb) && !readonly) {
2506 		ext4_msg(sb, KERN_ERR,
2507 			 "Filesystem with quota feature cannot be mounted RDWR "
2508 			 "without CONFIG_QUOTA");
2509 		return 0;
2510 	}
2511 #endif  /* CONFIG_QUOTA */
2512 	return 1;
2513 }
2514 
2515 /*
2516  * This function is called once a day if we have errors logged
2517  * on the file system
2518  */
2519 static void print_daily_error_info(unsigned long arg)
2520 {
2521 	struct super_block *sb = (struct super_block *) arg;
2522 	struct ext4_sb_info *sbi;
2523 	struct ext4_super_block *es;
2524 
2525 	sbi = EXT4_SB(sb);
2526 	es = sbi->s_es;
2527 
2528 	if (es->s_error_count)
2529 		/* fsck newer than v1.41.13 is needed to clean this condition. */
2530 		ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2531 			 le32_to_cpu(es->s_error_count));
2532 	if (es->s_first_error_time) {
2533 		printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2534 		       sb->s_id, le32_to_cpu(es->s_first_error_time),
2535 		       (int) sizeof(es->s_first_error_func),
2536 		       es->s_first_error_func,
2537 		       le32_to_cpu(es->s_first_error_line));
2538 		if (es->s_first_error_ino)
2539 			printk(": inode %u",
2540 			       le32_to_cpu(es->s_first_error_ino));
2541 		if (es->s_first_error_block)
2542 			printk(": block %llu", (unsigned long long)
2543 			       le64_to_cpu(es->s_first_error_block));
2544 		printk("\n");
2545 	}
2546 	if (es->s_last_error_time) {
2547 		printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2548 		       sb->s_id, le32_to_cpu(es->s_last_error_time),
2549 		       (int) sizeof(es->s_last_error_func),
2550 		       es->s_last_error_func,
2551 		       le32_to_cpu(es->s_last_error_line));
2552 		if (es->s_last_error_ino)
2553 			printk(": inode %u",
2554 			       le32_to_cpu(es->s_last_error_ino));
2555 		if (es->s_last_error_block)
2556 			printk(": block %llu", (unsigned long long)
2557 			       le64_to_cpu(es->s_last_error_block));
2558 		printk("\n");
2559 	}
2560 	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2561 }
2562 
2563 /* Find next suitable group and run ext4_init_inode_table */
2564 static int ext4_run_li_request(struct ext4_li_request *elr)
2565 {
2566 	struct ext4_group_desc *gdp = NULL;
2567 	ext4_group_t group, ngroups;
2568 	struct super_block *sb;
2569 	unsigned long timeout = 0;
2570 	int ret = 0;
2571 
2572 	sb = elr->lr_super;
2573 	ngroups = EXT4_SB(sb)->s_groups_count;
2574 
2575 	sb_start_write(sb);
2576 	for (group = elr->lr_next_group; group < ngroups; group++) {
2577 		gdp = ext4_get_group_desc(sb, group, NULL);
2578 		if (!gdp) {
2579 			ret = 1;
2580 			break;
2581 		}
2582 
2583 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2584 			break;
2585 	}
2586 
2587 	if (group >= ngroups)
2588 		ret = 1;
2589 
2590 	if (!ret) {
2591 		timeout = jiffies;
2592 		ret = ext4_init_inode_table(sb, group,
2593 					    elr->lr_timeout ? 0 : 1);
2594 		if (elr->lr_timeout == 0) {
2595 			timeout = (jiffies - timeout) *
2596 				  elr->lr_sbi->s_li_wait_mult;
2597 			elr->lr_timeout = timeout;
2598 		}
2599 		elr->lr_next_sched = jiffies + elr->lr_timeout;
2600 		elr->lr_next_group = group + 1;
2601 	}
2602 	sb_end_write(sb);
2603 
2604 	return ret;
2605 }
2606 
2607 /*
2608  * Remove lr_request from the list_request and free the
2609  * request structure. Should be called with li_list_mtx held
2610  */
2611 static void ext4_remove_li_request(struct ext4_li_request *elr)
2612 {
2613 	struct ext4_sb_info *sbi;
2614 
2615 	if (!elr)
2616 		return;
2617 
2618 	sbi = elr->lr_sbi;
2619 
2620 	list_del(&elr->lr_request);
2621 	sbi->s_li_request = NULL;
2622 	kfree(elr);
2623 }
2624 
2625 static void ext4_unregister_li_request(struct super_block *sb)
2626 {
2627 	mutex_lock(&ext4_li_mtx);
2628 	if (!ext4_li_info) {
2629 		mutex_unlock(&ext4_li_mtx);
2630 		return;
2631 	}
2632 
2633 	mutex_lock(&ext4_li_info->li_list_mtx);
2634 	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2635 	mutex_unlock(&ext4_li_info->li_list_mtx);
2636 	mutex_unlock(&ext4_li_mtx);
2637 }
2638 
2639 static struct task_struct *ext4_lazyinit_task;
2640 
2641 /*
2642  * This is the function where ext4lazyinit thread lives. It walks
2643  * through the request list searching for next scheduled filesystem.
2644  * When such a fs is found, run the lazy initialization request
2645  * (ext4_rn_li_request) and keep track of the time spend in this
2646  * function. Based on that time we compute next schedule time of
2647  * the request. When walking through the list is complete, compute
2648  * next waking time and put itself into sleep.
2649  */
2650 static int ext4_lazyinit_thread(void *arg)
2651 {
2652 	struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2653 	struct list_head *pos, *n;
2654 	struct ext4_li_request *elr;
2655 	unsigned long next_wakeup, cur;
2656 
2657 	BUG_ON(NULL == eli);
2658 
2659 cont_thread:
2660 	while (true) {
2661 		next_wakeup = MAX_JIFFY_OFFSET;
2662 
2663 		mutex_lock(&eli->li_list_mtx);
2664 		if (list_empty(&eli->li_request_list)) {
2665 			mutex_unlock(&eli->li_list_mtx);
2666 			goto exit_thread;
2667 		}
2668 
2669 		list_for_each_safe(pos, n, &eli->li_request_list) {
2670 			elr = list_entry(pos, struct ext4_li_request,
2671 					 lr_request);
2672 
2673 			if (time_after_eq(jiffies, elr->lr_next_sched)) {
2674 				if (ext4_run_li_request(elr) != 0) {
2675 					/* error, remove the lazy_init job */
2676 					ext4_remove_li_request(elr);
2677 					continue;
2678 				}
2679 			}
2680 
2681 			if (time_before(elr->lr_next_sched, next_wakeup))
2682 				next_wakeup = elr->lr_next_sched;
2683 		}
2684 		mutex_unlock(&eli->li_list_mtx);
2685 
2686 		try_to_freeze();
2687 
2688 		cur = jiffies;
2689 		if ((time_after_eq(cur, next_wakeup)) ||
2690 		    (MAX_JIFFY_OFFSET == next_wakeup)) {
2691 			cond_resched();
2692 			continue;
2693 		}
2694 
2695 		schedule_timeout_interruptible(next_wakeup - cur);
2696 
2697 		if (kthread_should_stop()) {
2698 			ext4_clear_request_list();
2699 			goto exit_thread;
2700 		}
2701 	}
2702 
2703 exit_thread:
2704 	/*
2705 	 * It looks like the request list is empty, but we need
2706 	 * to check it under the li_list_mtx lock, to prevent any
2707 	 * additions into it, and of course we should lock ext4_li_mtx
2708 	 * to atomically free the list and ext4_li_info, because at
2709 	 * this point another ext4 filesystem could be registering
2710 	 * new one.
2711 	 */
2712 	mutex_lock(&ext4_li_mtx);
2713 	mutex_lock(&eli->li_list_mtx);
2714 	if (!list_empty(&eli->li_request_list)) {
2715 		mutex_unlock(&eli->li_list_mtx);
2716 		mutex_unlock(&ext4_li_mtx);
2717 		goto cont_thread;
2718 	}
2719 	mutex_unlock(&eli->li_list_mtx);
2720 	kfree(ext4_li_info);
2721 	ext4_li_info = NULL;
2722 	mutex_unlock(&ext4_li_mtx);
2723 
2724 	return 0;
2725 }
2726 
2727 static void ext4_clear_request_list(void)
2728 {
2729 	struct list_head *pos, *n;
2730 	struct ext4_li_request *elr;
2731 
2732 	mutex_lock(&ext4_li_info->li_list_mtx);
2733 	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2734 		elr = list_entry(pos, struct ext4_li_request,
2735 				 lr_request);
2736 		ext4_remove_li_request(elr);
2737 	}
2738 	mutex_unlock(&ext4_li_info->li_list_mtx);
2739 }
2740 
2741 static int ext4_run_lazyinit_thread(void)
2742 {
2743 	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2744 					 ext4_li_info, "ext4lazyinit");
2745 	if (IS_ERR(ext4_lazyinit_task)) {
2746 		int err = PTR_ERR(ext4_lazyinit_task);
2747 		ext4_clear_request_list();
2748 		kfree(ext4_li_info);
2749 		ext4_li_info = NULL;
2750 		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2751 				 "initialization thread\n",
2752 				 err);
2753 		return err;
2754 	}
2755 	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2756 	return 0;
2757 }
2758 
2759 /*
2760  * Check whether it make sense to run itable init. thread or not.
2761  * If there is at least one uninitialized inode table, return
2762  * corresponding group number, else the loop goes through all
2763  * groups and return total number of groups.
2764  */
2765 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2766 {
2767 	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2768 	struct ext4_group_desc *gdp = NULL;
2769 
2770 	for (group = 0; group < ngroups; group++) {
2771 		gdp = ext4_get_group_desc(sb, group, NULL);
2772 		if (!gdp)
2773 			continue;
2774 
2775 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2776 			break;
2777 	}
2778 
2779 	return group;
2780 }
2781 
2782 static int ext4_li_info_new(void)
2783 {
2784 	struct ext4_lazy_init *eli = NULL;
2785 
2786 	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2787 	if (!eli)
2788 		return -ENOMEM;
2789 
2790 	INIT_LIST_HEAD(&eli->li_request_list);
2791 	mutex_init(&eli->li_list_mtx);
2792 
2793 	eli->li_state |= EXT4_LAZYINIT_QUIT;
2794 
2795 	ext4_li_info = eli;
2796 
2797 	return 0;
2798 }
2799 
2800 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2801 					    ext4_group_t start)
2802 {
2803 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2804 	struct ext4_li_request *elr;
2805 
2806 	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2807 	if (!elr)
2808 		return NULL;
2809 
2810 	elr->lr_super = sb;
2811 	elr->lr_sbi = sbi;
2812 	elr->lr_next_group = start;
2813 
2814 	/*
2815 	 * Randomize first schedule time of the request to
2816 	 * spread the inode table initialization requests
2817 	 * better.
2818 	 */
2819 	elr->lr_next_sched = jiffies + (prandom_u32() %
2820 				(EXT4_DEF_LI_MAX_START_DELAY * HZ));
2821 	return elr;
2822 }
2823 
2824 int ext4_register_li_request(struct super_block *sb,
2825 			     ext4_group_t first_not_zeroed)
2826 {
2827 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2828 	struct ext4_li_request *elr = NULL;
2829 	ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2830 	int ret = 0;
2831 
2832 	mutex_lock(&ext4_li_mtx);
2833 	if (sbi->s_li_request != NULL) {
2834 		/*
2835 		 * Reset timeout so it can be computed again, because
2836 		 * s_li_wait_mult might have changed.
2837 		 */
2838 		sbi->s_li_request->lr_timeout = 0;
2839 		goto out;
2840 	}
2841 
2842 	if (first_not_zeroed == ngroups ||
2843 	    (sb->s_flags & MS_RDONLY) ||
2844 	    !test_opt(sb, INIT_INODE_TABLE))
2845 		goto out;
2846 
2847 	elr = ext4_li_request_new(sb, first_not_zeroed);
2848 	if (!elr) {
2849 		ret = -ENOMEM;
2850 		goto out;
2851 	}
2852 
2853 	if (NULL == ext4_li_info) {
2854 		ret = ext4_li_info_new();
2855 		if (ret)
2856 			goto out;
2857 	}
2858 
2859 	mutex_lock(&ext4_li_info->li_list_mtx);
2860 	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2861 	mutex_unlock(&ext4_li_info->li_list_mtx);
2862 
2863 	sbi->s_li_request = elr;
2864 	/*
2865 	 * set elr to NULL here since it has been inserted to
2866 	 * the request_list and the removal and free of it is
2867 	 * handled by ext4_clear_request_list from now on.
2868 	 */
2869 	elr = NULL;
2870 
2871 	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2872 		ret = ext4_run_lazyinit_thread();
2873 		if (ret)
2874 			goto out;
2875 	}
2876 out:
2877 	mutex_unlock(&ext4_li_mtx);
2878 	if (ret)
2879 		kfree(elr);
2880 	return ret;
2881 }
2882 
2883 /*
2884  * We do not need to lock anything since this is called on
2885  * module unload.
2886  */
2887 static void ext4_destroy_lazyinit_thread(void)
2888 {
2889 	/*
2890 	 * If thread exited earlier
2891 	 * there's nothing to be done.
2892 	 */
2893 	if (!ext4_li_info || !ext4_lazyinit_task)
2894 		return;
2895 
2896 	kthread_stop(ext4_lazyinit_task);
2897 }
2898 
2899 static int set_journal_csum_feature_set(struct super_block *sb)
2900 {
2901 	int ret = 1;
2902 	int compat, incompat;
2903 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2904 
2905 	if (ext4_has_metadata_csum(sb)) {
2906 		/* journal checksum v3 */
2907 		compat = 0;
2908 		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
2909 	} else {
2910 		/* journal checksum v1 */
2911 		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2912 		incompat = 0;
2913 	}
2914 
2915 	jbd2_journal_clear_features(sbi->s_journal,
2916 			JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2917 			JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2918 			JBD2_FEATURE_INCOMPAT_CSUM_V2);
2919 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2920 		ret = jbd2_journal_set_features(sbi->s_journal,
2921 				compat, 0,
2922 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2923 				incompat);
2924 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2925 		ret = jbd2_journal_set_features(sbi->s_journal,
2926 				compat, 0,
2927 				incompat);
2928 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2929 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2930 	} else {
2931 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2932 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2933 	}
2934 
2935 	return ret;
2936 }
2937 
2938 /*
2939  * Note: calculating the overhead so we can be compatible with
2940  * historical BSD practice is quite difficult in the face of
2941  * clusters/bigalloc.  This is because multiple metadata blocks from
2942  * different block group can end up in the same allocation cluster.
2943  * Calculating the exact overhead in the face of clustered allocation
2944  * requires either O(all block bitmaps) in memory or O(number of block
2945  * groups**2) in time.  We will still calculate the superblock for
2946  * older file systems --- and if we come across with a bigalloc file
2947  * system with zero in s_overhead_clusters the estimate will be close to
2948  * correct especially for very large cluster sizes --- but for newer
2949  * file systems, it's better to calculate this figure once at mkfs
2950  * time, and store it in the superblock.  If the superblock value is
2951  * present (even for non-bigalloc file systems), we will use it.
2952  */
2953 static int count_overhead(struct super_block *sb, ext4_group_t grp,
2954 			  char *buf)
2955 {
2956 	struct ext4_sb_info	*sbi = EXT4_SB(sb);
2957 	struct ext4_group_desc	*gdp;
2958 	ext4_fsblk_t		first_block, last_block, b;
2959 	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
2960 	int			s, j, count = 0;
2961 
2962 	if (!ext4_has_feature_bigalloc(sb))
2963 		return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
2964 			sbi->s_itb_per_group + 2);
2965 
2966 	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
2967 		(grp * EXT4_BLOCKS_PER_GROUP(sb));
2968 	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
2969 	for (i = 0; i < ngroups; i++) {
2970 		gdp = ext4_get_group_desc(sb, i, NULL);
2971 		b = ext4_block_bitmap(sb, gdp);
2972 		if (b >= first_block && b <= last_block) {
2973 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2974 			count++;
2975 		}
2976 		b = ext4_inode_bitmap(sb, gdp);
2977 		if (b >= first_block && b <= last_block) {
2978 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
2979 			count++;
2980 		}
2981 		b = ext4_inode_table(sb, gdp);
2982 		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
2983 			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
2984 				int c = EXT4_B2C(sbi, b - first_block);
2985 				ext4_set_bit(c, buf);
2986 				count++;
2987 			}
2988 		if (i != grp)
2989 			continue;
2990 		s = 0;
2991 		if (ext4_bg_has_super(sb, grp)) {
2992 			ext4_set_bit(s++, buf);
2993 			count++;
2994 		}
2995 		for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
2996 			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
2997 			count++;
2998 		}
2999 	}
3000 	if (!count)
3001 		return 0;
3002 	return EXT4_CLUSTERS_PER_GROUP(sb) -
3003 		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3004 }
3005 
3006 /*
3007  * Compute the overhead and stash it in sbi->s_overhead
3008  */
3009 int ext4_calculate_overhead(struct super_block *sb)
3010 {
3011 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3012 	struct ext4_super_block *es = sbi->s_es;
3013 	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3014 	ext4_fsblk_t overhead = 0;
3015 	char *buf = (char *) get_zeroed_page(GFP_NOFS);
3016 
3017 	if (!buf)
3018 		return -ENOMEM;
3019 
3020 	/*
3021 	 * Compute the overhead (FS structures).  This is constant
3022 	 * for a given filesystem unless the number of block groups
3023 	 * changes so we cache the previous value until it does.
3024 	 */
3025 
3026 	/*
3027 	 * All of the blocks before first_data_block are overhead
3028 	 */
3029 	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3030 
3031 	/*
3032 	 * Add the overhead found in each block group
3033 	 */
3034 	for (i = 0; i < ngroups; i++) {
3035 		int blks;
3036 
3037 		blks = count_overhead(sb, i, buf);
3038 		overhead += blks;
3039 		if (blks)
3040 			memset(buf, 0, PAGE_SIZE);
3041 		cond_resched();
3042 	}
3043 	/* Add the internal journal blocks as well */
3044 	if (sbi->s_journal && !sbi->journal_bdev)
3045 		overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3046 
3047 	sbi->s_overhead = overhead;
3048 	smp_wmb();
3049 	free_page((unsigned long) buf);
3050 	return 0;
3051 }
3052 
3053 static void ext4_set_resv_clusters(struct super_block *sb)
3054 {
3055 	ext4_fsblk_t resv_clusters;
3056 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3057 
3058 	/*
3059 	 * There's no need to reserve anything when we aren't using extents.
3060 	 * The space estimates are exact, there are no unwritten extents,
3061 	 * hole punching doesn't need new metadata... This is needed especially
3062 	 * to keep ext2/3 backward compatibility.
3063 	 */
3064 	if (!ext4_has_feature_extents(sb))
3065 		return;
3066 	/*
3067 	 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3068 	 * This should cover the situations where we can not afford to run
3069 	 * out of space like for example punch hole, or converting
3070 	 * unwritten extents in delalloc path. In most cases such
3071 	 * allocation would require 1, or 2 blocks, higher numbers are
3072 	 * very rare.
3073 	 */
3074 	resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3075 			 sbi->s_cluster_bits);
3076 
3077 	do_div(resv_clusters, 50);
3078 	resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3079 
3080 	atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3081 }
3082 
3083 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3084 {
3085 	char *orig_data = kstrdup(data, GFP_KERNEL);
3086 	struct buffer_head *bh;
3087 	struct ext4_super_block *es = NULL;
3088 	struct ext4_sb_info *sbi;
3089 	ext4_fsblk_t block;
3090 	ext4_fsblk_t sb_block = get_sb_block(&data);
3091 	ext4_fsblk_t logical_sb_block;
3092 	unsigned long offset = 0;
3093 	unsigned long journal_devnum = 0;
3094 	unsigned long def_mount_opts;
3095 	struct inode *root;
3096 	const char *descr;
3097 	int ret = -ENOMEM;
3098 	int blocksize, clustersize;
3099 	unsigned int db_count;
3100 	unsigned int i;
3101 	int needs_recovery, has_huge_files, has_bigalloc;
3102 	__u64 blocks_count;
3103 	int err = 0;
3104 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3105 	ext4_group_t first_not_zeroed;
3106 
3107 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3108 	if (!sbi)
3109 		goto out_free_orig;
3110 
3111 	sbi->s_blockgroup_lock =
3112 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3113 	if (!sbi->s_blockgroup_lock) {
3114 		kfree(sbi);
3115 		goto out_free_orig;
3116 	}
3117 	sb->s_fs_info = sbi;
3118 	sbi->s_sb = sb;
3119 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3120 	sbi->s_sb_block = sb_block;
3121 	if (sb->s_bdev->bd_part)
3122 		sbi->s_sectors_written_start =
3123 			part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3124 
3125 	/* Cleanup superblock name */
3126 	strreplace(sb->s_id, '/', '!');
3127 
3128 	/* -EINVAL is default */
3129 	ret = -EINVAL;
3130 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3131 	if (!blocksize) {
3132 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3133 		goto out_fail;
3134 	}
3135 
3136 	/*
3137 	 * The ext4 superblock will not be buffer aligned for other than 1kB
3138 	 * block sizes.  We need to calculate the offset from buffer start.
3139 	 */
3140 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3141 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3142 		offset = do_div(logical_sb_block, blocksize);
3143 	} else {
3144 		logical_sb_block = sb_block;
3145 	}
3146 
3147 	if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3148 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
3149 		goto out_fail;
3150 	}
3151 	/*
3152 	 * Note: s_es must be initialized as soon as possible because
3153 	 *       some ext4 macro-instructions depend on its value
3154 	 */
3155 	es = (struct ext4_super_block *) (bh->b_data + offset);
3156 	sbi->s_es = es;
3157 	sb->s_magic = le16_to_cpu(es->s_magic);
3158 	if (sb->s_magic != EXT4_SUPER_MAGIC)
3159 		goto cantfind_ext4;
3160 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3161 
3162 	/* Warn if metadata_csum and gdt_csum are both set. */
3163 	if (ext4_has_feature_metadata_csum(sb) &&
3164 	    ext4_has_feature_gdt_csum(sb))
3165 		ext4_warning(sb, "metadata_csum and uninit_bg are "
3166 			     "redundant flags; please run fsck.");
3167 
3168 	/* Check for a known checksum algorithm */
3169 	if (!ext4_verify_csum_type(sb, es)) {
3170 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3171 			 "unknown checksum algorithm.");
3172 		silent = 1;
3173 		goto cantfind_ext4;
3174 	}
3175 
3176 	/* Load the checksum driver */
3177 	if (ext4_has_feature_metadata_csum(sb)) {
3178 		sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3179 		if (IS_ERR(sbi->s_chksum_driver)) {
3180 			ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3181 			ret = PTR_ERR(sbi->s_chksum_driver);
3182 			sbi->s_chksum_driver = NULL;
3183 			goto failed_mount;
3184 		}
3185 	}
3186 
3187 	/* Check superblock checksum */
3188 	if (!ext4_superblock_csum_verify(sb, es)) {
3189 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3190 			 "invalid superblock checksum.  Run e2fsck?");
3191 		silent = 1;
3192 		ret = -EFSBADCRC;
3193 		goto cantfind_ext4;
3194 	}
3195 
3196 	/* Precompute checksum seed for all metadata */
3197 	if (ext4_has_feature_csum_seed(sb))
3198 		sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3199 	else if (ext4_has_metadata_csum(sb))
3200 		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3201 					       sizeof(es->s_uuid));
3202 
3203 	/* Set defaults before we parse the mount options */
3204 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3205 	set_opt(sb, INIT_INODE_TABLE);
3206 	if (def_mount_opts & EXT4_DEFM_DEBUG)
3207 		set_opt(sb, DEBUG);
3208 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3209 		set_opt(sb, GRPID);
3210 	if (def_mount_opts & EXT4_DEFM_UID16)
3211 		set_opt(sb, NO_UID32);
3212 	/* xattr user namespace & acls are now defaulted on */
3213 	set_opt(sb, XATTR_USER);
3214 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3215 	set_opt(sb, POSIX_ACL);
3216 #endif
3217 	/* don't forget to enable journal_csum when metadata_csum is enabled. */
3218 	if (ext4_has_metadata_csum(sb))
3219 		set_opt(sb, JOURNAL_CHECKSUM);
3220 
3221 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3222 		set_opt(sb, JOURNAL_DATA);
3223 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3224 		set_opt(sb, ORDERED_DATA);
3225 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3226 		set_opt(sb, WRITEBACK_DATA);
3227 
3228 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3229 		set_opt(sb, ERRORS_PANIC);
3230 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3231 		set_opt(sb, ERRORS_CONT);
3232 	else
3233 		set_opt(sb, ERRORS_RO);
3234 	/* block_validity enabled by default; disable with noblock_validity */
3235 	set_opt(sb, BLOCK_VALIDITY);
3236 	if (def_mount_opts & EXT4_DEFM_DISCARD)
3237 		set_opt(sb, DISCARD);
3238 
3239 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3240 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3241 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3242 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3243 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3244 
3245 	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3246 		set_opt(sb, BARRIER);
3247 
3248 	/*
3249 	 * enable delayed allocation by default
3250 	 * Use -o nodelalloc to turn it off
3251 	 */
3252 	if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3253 	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3254 		set_opt(sb, DELALLOC);
3255 
3256 	/*
3257 	 * set default s_li_wait_mult for lazyinit, for the case there is
3258 	 * no mount option specified.
3259 	 */
3260 	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3261 
3262 	if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3263 			   &journal_devnum, &journal_ioprio, 0)) {
3264 		ext4_msg(sb, KERN_WARNING,
3265 			 "failed to parse options in superblock: %s",
3266 			 sbi->s_es->s_mount_opts);
3267 	}
3268 	sbi->s_def_mount_opt = sbi->s_mount_opt;
3269 	if (!parse_options((char *) data, sb, &journal_devnum,
3270 			   &journal_ioprio, 0))
3271 		goto failed_mount;
3272 
3273 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3274 		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3275 			    "with data=journal disables delayed "
3276 			    "allocation and O_DIRECT support!\n");
3277 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3278 			ext4_msg(sb, KERN_ERR, "can't mount with "
3279 				 "both data=journal and delalloc");
3280 			goto failed_mount;
3281 		}
3282 		if (test_opt(sb, DIOREAD_NOLOCK)) {
3283 			ext4_msg(sb, KERN_ERR, "can't mount with "
3284 				 "both data=journal and dioread_nolock");
3285 			goto failed_mount;
3286 		}
3287 		if (test_opt(sb, DAX)) {
3288 			ext4_msg(sb, KERN_ERR, "can't mount with "
3289 				 "both data=journal and dax");
3290 			goto failed_mount;
3291 		}
3292 		if (test_opt(sb, DELALLOC))
3293 			clear_opt(sb, DELALLOC);
3294 	} else {
3295 		sb->s_iflags |= SB_I_CGROUPWB;
3296 	}
3297 
3298 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3299 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3300 
3301 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3302 	    (ext4_has_compat_features(sb) ||
3303 	     ext4_has_ro_compat_features(sb) ||
3304 	     ext4_has_incompat_features(sb)))
3305 		ext4_msg(sb, KERN_WARNING,
3306 		       "feature flags set on rev 0 fs, "
3307 		       "running e2fsck is recommended");
3308 
3309 	if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3310 		set_opt2(sb, HURD_COMPAT);
3311 		if (ext4_has_feature_64bit(sb)) {
3312 			ext4_msg(sb, KERN_ERR,
3313 				 "The Hurd can't support 64-bit file systems");
3314 			goto failed_mount;
3315 		}
3316 	}
3317 
3318 	if (IS_EXT2_SB(sb)) {
3319 		if (ext2_feature_set_ok(sb))
3320 			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3321 				 "using the ext4 subsystem");
3322 		else {
3323 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3324 				 "to feature incompatibilities");
3325 			goto failed_mount;
3326 		}
3327 	}
3328 
3329 	if (IS_EXT3_SB(sb)) {
3330 		if (ext3_feature_set_ok(sb))
3331 			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3332 				 "using the ext4 subsystem");
3333 		else {
3334 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3335 				 "to feature incompatibilities");
3336 			goto failed_mount;
3337 		}
3338 	}
3339 
3340 	/*
3341 	 * Check feature flags regardless of the revision level, since we
3342 	 * previously didn't change the revision level when setting the flags,
3343 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
3344 	 */
3345 	if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3346 		goto failed_mount;
3347 
3348 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3349 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3350 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
3351 		ext4_msg(sb, KERN_ERR,
3352 		       "Unsupported filesystem blocksize %d", blocksize);
3353 		goto failed_mount;
3354 	}
3355 
3356 	if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3357 		if (blocksize != PAGE_SIZE) {
3358 			ext4_msg(sb, KERN_ERR,
3359 					"error: unsupported blocksize for dax");
3360 			goto failed_mount;
3361 		}
3362 		if (!sb->s_bdev->bd_disk->fops->direct_access) {
3363 			ext4_msg(sb, KERN_ERR,
3364 					"error: device does not support dax");
3365 			goto failed_mount;
3366 		}
3367 	}
3368 
3369 	if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3370 		ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3371 			 es->s_encryption_level);
3372 		goto failed_mount;
3373 	}
3374 
3375 	if (sb->s_blocksize != blocksize) {
3376 		/* Validate the filesystem blocksize */
3377 		if (!sb_set_blocksize(sb, blocksize)) {
3378 			ext4_msg(sb, KERN_ERR, "bad block size %d",
3379 					blocksize);
3380 			goto failed_mount;
3381 		}
3382 
3383 		brelse(bh);
3384 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3385 		offset = do_div(logical_sb_block, blocksize);
3386 		bh = sb_bread_unmovable(sb, logical_sb_block);
3387 		if (!bh) {
3388 			ext4_msg(sb, KERN_ERR,
3389 			       "Can't read superblock on 2nd try");
3390 			goto failed_mount;
3391 		}
3392 		es = (struct ext4_super_block *)(bh->b_data + offset);
3393 		sbi->s_es = es;
3394 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3395 			ext4_msg(sb, KERN_ERR,
3396 			       "Magic mismatch, very weird!");
3397 			goto failed_mount;
3398 		}
3399 	}
3400 
3401 	has_huge_files = ext4_has_feature_huge_file(sb);
3402 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3403 						      has_huge_files);
3404 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3405 
3406 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3407 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3408 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3409 	} else {
3410 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3411 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3412 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3413 		    (!is_power_of_2(sbi->s_inode_size)) ||
3414 		    (sbi->s_inode_size > blocksize)) {
3415 			ext4_msg(sb, KERN_ERR,
3416 			       "unsupported inode size: %d",
3417 			       sbi->s_inode_size);
3418 			goto failed_mount;
3419 		}
3420 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3421 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3422 	}
3423 
3424 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3425 	if (ext4_has_feature_64bit(sb)) {
3426 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3427 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3428 		    !is_power_of_2(sbi->s_desc_size)) {
3429 			ext4_msg(sb, KERN_ERR,
3430 			       "unsupported descriptor size %lu",
3431 			       sbi->s_desc_size);
3432 			goto failed_mount;
3433 		}
3434 	} else
3435 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3436 
3437 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3438 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3439 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3440 		goto cantfind_ext4;
3441 
3442 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3443 	if (sbi->s_inodes_per_block == 0)
3444 		goto cantfind_ext4;
3445 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
3446 					sbi->s_inodes_per_block;
3447 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3448 	sbi->s_sbh = bh;
3449 	sbi->s_mount_state = le16_to_cpu(es->s_state);
3450 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3451 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3452 
3453 	for (i = 0; i < 4; i++)
3454 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3455 	sbi->s_def_hash_version = es->s_def_hash_version;
3456 	if (ext4_has_feature_dir_index(sb)) {
3457 		i = le32_to_cpu(es->s_flags);
3458 		if (i & EXT2_FLAGS_UNSIGNED_HASH)
3459 			sbi->s_hash_unsigned = 3;
3460 		else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3461 #ifdef __CHAR_UNSIGNED__
3462 			if (!(sb->s_flags & MS_RDONLY))
3463 				es->s_flags |=
3464 					cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3465 			sbi->s_hash_unsigned = 3;
3466 #else
3467 			if (!(sb->s_flags & MS_RDONLY))
3468 				es->s_flags |=
3469 					cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3470 #endif
3471 		}
3472 	}
3473 
3474 	/* Handle clustersize */
3475 	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3476 	has_bigalloc = ext4_has_feature_bigalloc(sb);
3477 	if (has_bigalloc) {
3478 		if (clustersize < blocksize) {
3479 			ext4_msg(sb, KERN_ERR,
3480 				 "cluster size (%d) smaller than "
3481 				 "block size (%d)", clustersize, blocksize);
3482 			goto failed_mount;
3483 		}
3484 		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3485 			le32_to_cpu(es->s_log_block_size);
3486 		sbi->s_clusters_per_group =
3487 			le32_to_cpu(es->s_clusters_per_group);
3488 		if (sbi->s_clusters_per_group > blocksize * 8) {
3489 			ext4_msg(sb, KERN_ERR,
3490 				 "#clusters per group too big: %lu",
3491 				 sbi->s_clusters_per_group);
3492 			goto failed_mount;
3493 		}
3494 		if (sbi->s_blocks_per_group !=
3495 		    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3496 			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3497 				 "clusters per group (%lu) inconsistent",
3498 				 sbi->s_blocks_per_group,
3499 				 sbi->s_clusters_per_group);
3500 			goto failed_mount;
3501 		}
3502 	} else {
3503 		if (clustersize != blocksize) {
3504 			ext4_warning(sb, "fragment/cluster size (%d) != "
3505 				     "block size (%d)", clustersize,
3506 				     blocksize);
3507 			clustersize = blocksize;
3508 		}
3509 		if (sbi->s_blocks_per_group > blocksize * 8) {
3510 			ext4_msg(sb, KERN_ERR,
3511 				 "#blocks per group too big: %lu",
3512 				 sbi->s_blocks_per_group);
3513 			goto failed_mount;
3514 		}
3515 		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3516 		sbi->s_cluster_bits = 0;
3517 	}
3518 	sbi->s_cluster_ratio = clustersize / blocksize;
3519 
3520 	if (sbi->s_inodes_per_group > blocksize * 8) {
3521 		ext4_msg(sb, KERN_ERR,
3522 		       "#inodes per group too big: %lu",
3523 		       sbi->s_inodes_per_group);
3524 		goto failed_mount;
3525 	}
3526 
3527 	/* Do we have standard group size of clustersize * 8 blocks ? */
3528 	if (sbi->s_blocks_per_group == clustersize << 3)
3529 		set_opt2(sb, STD_GROUP_SIZE);
3530 
3531 	/*
3532 	 * Test whether we have more sectors than will fit in sector_t,
3533 	 * and whether the max offset is addressable by the page cache.
3534 	 */
3535 	err = generic_check_addressable(sb->s_blocksize_bits,
3536 					ext4_blocks_count(es));
3537 	if (err) {
3538 		ext4_msg(sb, KERN_ERR, "filesystem"
3539 			 " too large to mount safely on this system");
3540 		if (sizeof(sector_t) < 8)
3541 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3542 		goto failed_mount;
3543 	}
3544 
3545 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3546 		goto cantfind_ext4;
3547 
3548 	/* check blocks count against device size */
3549 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3550 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3551 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3552 		       "exceeds size of device (%llu blocks)",
3553 		       ext4_blocks_count(es), blocks_count);
3554 		goto failed_mount;
3555 	}
3556 
3557 	/*
3558 	 * It makes no sense for the first data block to be beyond the end
3559 	 * of the filesystem.
3560 	 */
3561 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3562 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3563 			 "block %u is beyond end of filesystem (%llu)",
3564 			 le32_to_cpu(es->s_first_data_block),
3565 			 ext4_blocks_count(es));
3566 		goto failed_mount;
3567 	}
3568 	blocks_count = (ext4_blocks_count(es) -
3569 			le32_to_cpu(es->s_first_data_block) +
3570 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
3571 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3572 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3573 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3574 		       "(block count %llu, first data block %u, "
3575 		       "blocks per group %lu)", sbi->s_groups_count,
3576 		       ext4_blocks_count(es),
3577 		       le32_to_cpu(es->s_first_data_block),
3578 		       EXT4_BLOCKS_PER_GROUP(sb));
3579 		goto failed_mount;
3580 	}
3581 	sbi->s_groups_count = blocks_count;
3582 	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3583 			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3584 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3585 		   EXT4_DESC_PER_BLOCK(sb);
3586 	sbi->s_group_desc = ext4_kvmalloc(db_count *
3587 					  sizeof(struct buffer_head *),
3588 					  GFP_KERNEL);
3589 	if (sbi->s_group_desc == NULL) {
3590 		ext4_msg(sb, KERN_ERR, "not enough memory");
3591 		ret = -ENOMEM;
3592 		goto failed_mount;
3593 	}
3594 
3595 	bgl_lock_init(sbi->s_blockgroup_lock);
3596 
3597 	for (i = 0; i < db_count; i++) {
3598 		block = descriptor_loc(sb, logical_sb_block, i);
3599 		sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3600 		if (!sbi->s_group_desc[i]) {
3601 			ext4_msg(sb, KERN_ERR,
3602 			       "can't read group descriptor %d", i);
3603 			db_count = i;
3604 			goto failed_mount2;
3605 		}
3606 	}
3607 	if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3608 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3609 		ret = -EFSCORRUPTED;
3610 		goto failed_mount2;
3611 	}
3612 
3613 	sbi->s_gdb_count = db_count;
3614 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3615 	spin_lock_init(&sbi->s_next_gen_lock);
3616 
3617 	setup_timer(&sbi->s_err_report, print_daily_error_info,
3618 		(unsigned long) sb);
3619 
3620 	/* Register extent status tree shrinker */
3621 	if (ext4_es_register_shrinker(sbi))
3622 		goto failed_mount3;
3623 
3624 	sbi->s_stripe = ext4_get_stripe_size(sbi);
3625 	sbi->s_extent_max_zeroout_kb = 32;
3626 
3627 	/*
3628 	 * set up enough so that it can read an inode
3629 	 */
3630 	sb->s_op = &ext4_sops;
3631 	sb->s_export_op = &ext4_export_ops;
3632 	sb->s_xattr = ext4_xattr_handlers;
3633 #ifdef CONFIG_QUOTA
3634 	sb->dq_op = &ext4_quota_operations;
3635 	if (ext4_has_feature_quota(sb))
3636 		sb->s_qcop = &dquot_quotactl_sysfile_ops;
3637 	else
3638 		sb->s_qcop = &ext4_qctl_operations;
3639 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3640 #endif
3641 	memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3642 
3643 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3644 	mutex_init(&sbi->s_orphan_lock);
3645 
3646 	sb->s_root = NULL;
3647 
3648 	needs_recovery = (es->s_last_orphan != 0 ||
3649 			  ext4_has_feature_journal_needs_recovery(sb));
3650 
3651 	if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3652 		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3653 			goto failed_mount3a;
3654 
3655 	/*
3656 	 * The first inode we look at is the journal inode.  Don't try
3657 	 * root first: it may be modified in the journal!
3658 	 */
3659 	if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3660 		if (ext4_load_journal(sb, es, journal_devnum))
3661 			goto failed_mount3a;
3662 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3663 		   ext4_has_feature_journal_needs_recovery(sb)) {
3664 		ext4_msg(sb, KERN_ERR, "required journal recovery "
3665 		       "suppressed and not mounted read-only");
3666 		goto failed_mount_wq;
3667 	} else {
3668 		clear_opt(sb, DATA_FLAGS);
3669 		sbi->s_journal = NULL;
3670 		needs_recovery = 0;
3671 		goto no_journal;
3672 	}
3673 
3674 	if (ext4_has_feature_64bit(sb) &&
3675 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3676 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
3677 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3678 		goto failed_mount_wq;
3679 	}
3680 
3681 	if (!set_journal_csum_feature_set(sb)) {
3682 		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3683 			 "feature set");
3684 		goto failed_mount_wq;
3685 	}
3686 
3687 	/* We have now updated the journal if required, so we can
3688 	 * validate the data journaling mode. */
3689 	switch (test_opt(sb, DATA_FLAGS)) {
3690 	case 0:
3691 		/* No mode set, assume a default based on the journal
3692 		 * capabilities: ORDERED_DATA if the journal can
3693 		 * cope, else JOURNAL_DATA
3694 		 */
3695 		if (jbd2_journal_check_available_features
3696 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3697 			set_opt(sb, ORDERED_DATA);
3698 		else
3699 			set_opt(sb, JOURNAL_DATA);
3700 		break;
3701 
3702 	case EXT4_MOUNT_ORDERED_DATA:
3703 	case EXT4_MOUNT_WRITEBACK_DATA:
3704 		if (!jbd2_journal_check_available_features
3705 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3706 			ext4_msg(sb, KERN_ERR, "Journal does not support "
3707 			       "requested data journaling mode");
3708 			goto failed_mount_wq;
3709 		}
3710 	default:
3711 		break;
3712 	}
3713 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3714 
3715 	sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3716 
3717 no_journal:
3718 	if (ext4_mballoc_ready) {
3719 		sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
3720 		if (!sbi->s_mb_cache) {
3721 			ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3722 			goto failed_mount_wq;
3723 		}
3724 	}
3725 
3726 	if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3727 	    (blocksize != PAGE_CACHE_SIZE)) {
3728 		ext4_msg(sb, KERN_ERR,
3729 			 "Unsupported blocksize for fs encryption");
3730 		goto failed_mount_wq;
3731 	}
3732 
3733 	if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3734 	    !ext4_has_feature_encrypt(sb)) {
3735 		ext4_set_feature_encrypt(sb);
3736 		ext4_commit_super(sb, 1);
3737 	}
3738 
3739 	/*
3740 	 * Get the # of file system overhead blocks from the
3741 	 * superblock if present.
3742 	 */
3743 	if (es->s_overhead_clusters)
3744 		sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3745 	else {
3746 		err = ext4_calculate_overhead(sb);
3747 		if (err)
3748 			goto failed_mount_wq;
3749 	}
3750 
3751 	/*
3752 	 * The maximum number of concurrent works can be high and
3753 	 * concurrency isn't really necessary.  Limit it to 1.
3754 	 */
3755 	EXT4_SB(sb)->rsv_conversion_wq =
3756 		alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3757 	if (!EXT4_SB(sb)->rsv_conversion_wq) {
3758 		printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3759 		ret = -ENOMEM;
3760 		goto failed_mount4;
3761 	}
3762 
3763 	/*
3764 	 * The jbd2_journal_load will have done any necessary log recovery,
3765 	 * so we can safely mount the rest of the filesystem now.
3766 	 */
3767 
3768 	root = ext4_iget(sb, EXT4_ROOT_INO);
3769 	if (IS_ERR(root)) {
3770 		ext4_msg(sb, KERN_ERR, "get root inode failed");
3771 		ret = PTR_ERR(root);
3772 		root = NULL;
3773 		goto failed_mount4;
3774 	}
3775 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3776 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3777 		iput(root);
3778 		goto failed_mount4;
3779 	}
3780 	sb->s_root = d_make_root(root);
3781 	if (!sb->s_root) {
3782 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
3783 		ret = -ENOMEM;
3784 		goto failed_mount4;
3785 	}
3786 
3787 	if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3788 		sb->s_flags |= MS_RDONLY;
3789 
3790 	/* determine the minimum size of new large inodes, if present */
3791 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3792 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3793 						     EXT4_GOOD_OLD_INODE_SIZE;
3794 		if (ext4_has_feature_extra_isize(sb)) {
3795 			if (sbi->s_want_extra_isize <
3796 			    le16_to_cpu(es->s_want_extra_isize))
3797 				sbi->s_want_extra_isize =
3798 					le16_to_cpu(es->s_want_extra_isize);
3799 			if (sbi->s_want_extra_isize <
3800 			    le16_to_cpu(es->s_min_extra_isize))
3801 				sbi->s_want_extra_isize =
3802 					le16_to_cpu(es->s_min_extra_isize);
3803 		}
3804 	}
3805 	/* Check if enough inode space is available */
3806 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3807 							sbi->s_inode_size) {
3808 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3809 						       EXT4_GOOD_OLD_INODE_SIZE;
3810 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
3811 			 "available");
3812 	}
3813 
3814 	ext4_set_resv_clusters(sb);
3815 
3816 	err = ext4_setup_system_zone(sb);
3817 	if (err) {
3818 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
3819 			 "zone (%d)", err);
3820 		goto failed_mount4a;
3821 	}
3822 
3823 	ext4_ext_init(sb);
3824 	err = ext4_mb_init(sb);
3825 	if (err) {
3826 		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3827 			 err);
3828 		goto failed_mount5;
3829 	}
3830 
3831 	block = ext4_count_free_clusters(sb);
3832 	ext4_free_blocks_count_set(sbi->s_es,
3833 				   EXT4_C2B(sbi, block));
3834 	err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3835 				  GFP_KERNEL);
3836 	if (!err) {
3837 		unsigned long freei = ext4_count_free_inodes(sb);
3838 		sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
3839 		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3840 					  GFP_KERNEL);
3841 	}
3842 	if (!err)
3843 		err = percpu_counter_init(&sbi->s_dirs_counter,
3844 					  ext4_count_dirs(sb), GFP_KERNEL);
3845 	if (!err)
3846 		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3847 					  GFP_KERNEL);
3848 	if (err) {
3849 		ext4_msg(sb, KERN_ERR, "insufficient memory");
3850 		goto failed_mount6;
3851 	}
3852 
3853 	if (ext4_has_feature_flex_bg(sb))
3854 		if (!ext4_fill_flex_info(sb)) {
3855 			ext4_msg(sb, KERN_ERR,
3856 			       "unable to initialize "
3857 			       "flex_bg meta info!");
3858 			goto failed_mount6;
3859 		}
3860 
3861 	err = ext4_register_li_request(sb, first_not_zeroed);
3862 	if (err)
3863 		goto failed_mount6;
3864 
3865 	err = ext4_register_sysfs(sb);
3866 	if (err)
3867 		goto failed_mount7;
3868 
3869 #ifdef CONFIG_QUOTA
3870 	/* Enable quota usage during mount. */
3871 	if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
3872 		err = ext4_enable_quotas(sb);
3873 		if (err)
3874 			goto failed_mount8;
3875 	}
3876 #endif  /* CONFIG_QUOTA */
3877 
3878 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3879 	ext4_orphan_cleanup(sb, es);
3880 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3881 	if (needs_recovery) {
3882 		ext4_msg(sb, KERN_INFO, "recovery complete");
3883 		ext4_mark_recovery_complete(sb, es);
3884 	}
3885 	if (EXT4_SB(sb)->s_journal) {
3886 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3887 			descr = " journalled data mode";
3888 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3889 			descr = " ordered data mode";
3890 		else
3891 			descr = " writeback data mode";
3892 	} else
3893 		descr = "out journal";
3894 
3895 	if (test_opt(sb, DISCARD)) {
3896 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
3897 		if (!blk_queue_discard(q))
3898 			ext4_msg(sb, KERN_WARNING,
3899 				 "mounting with \"discard\" option, but "
3900 				 "the device does not support discard");
3901 	}
3902 
3903 	if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3904 		ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3905 			 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3906 			 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3907 
3908 	if (es->s_error_count)
3909 		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3910 
3911 	/* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3912 	ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
3913 	ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
3914 	ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
3915 
3916 	kfree(orig_data);
3917 	return 0;
3918 
3919 cantfind_ext4:
3920 	if (!silent)
3921 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3922 	goto failed_mount;
3923 
3924 #ifdef CONFIG_QUOTA
3925 failed_mount8:
3926 	ext4_unregister_sysfs(sb);
3927 #endif
3928 failed_mount7:
3929 	ext4_unregister_li_request(sb);
3930 failed_mount6:
3931 	ext4_mb_release(sb);
3932 	if (sbi->s_flex_groups)
3933 		kvfree(sbi->s_flex_groups);
3934 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
3935 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
3936 	percpu_counter_destroy(&sbi->s_dirs_counter);
3937 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3938 failed_mount5:
3939 	ext4_ext_release(sb);
3940 	ext4_release_system_zone(sb);
3941 failed_mount4a:
3942 	dput(sb->s_root);
3943 	sb->s_root = NULL;
3944 failed_mount4:
3945 	ext4_msg(sb, KERN_ERR, "mount failed");
3946 	if (EXT4_SB(sb)->rsv_conversion_wq)
3947 		destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
3948 failed_mount_wq:
3949 	if (sbi->s_journal) {
3950 		jbd2_journal_destroy(sbi->s_journal);
3951 		sbi->s_journal = NULL;
3952 	}
3953 failed_mount3a:
3954 	ext4_es_unregister_shrinker(sbi);
3955 failed_mount3:
3956 	del_timer_sync(&sbi->s_err_report);
3957 	if (sbi->s_mmp_tsk)
3958 		kthread_stop(sbi->s_mmp_tsk);
3959 failed_mount2:
3960 	for (i = 0; i < db_count; i++)
3961 		brelse(sbi->s_group_desc[i]);
3962 	kvfree(sbi->s_group_desc);
3963 failed_mount:
3964 	if (sbi->s_chksum_driver)
3965 		crypto_free_shash(sbi->s_chksum_driver);
3966 #ifdef CONFIG_QUOTA
3967 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
3968 		kfree(sbi->s_qf_names[i]);
3969 #endif
3970 	ext4_blkdev_remove(sbi);
3971 	brelse(bh);
3972 out_fail:
3973 	sb->s_fs_info = NULL;
3974 	kfree(sbi->s_blockgroup_lock);
3975 	kfree(sbi);
3976 out_free_orig:
3977 	kfree(orig_data);
3978 	return err ? err : ret;
3979 }
3980 
3981 /*
3982  * Setup any per-fs journal parameters now.  We'll do this both on
3983  * initial mount, once the journal has been initialised but before we've
3984  * done any recovery; and again on any subsequent remount.
3985  */
3986 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3987 {
3988 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3989 
3990 	journal->j_commit_interval = sbi->s_commit_interval;
3991 	journal->j_min_batch_time = sbi->s_min_batch_time;
3992 	journal->j_max_batch_time = sbi->s_max_batch_time;
3993 
3994 	write_lock(&journal->j_state_lock);
3995 	if (test_opt(sb, BARRIER))
3996 		journal->j_flags |= JBD2_BARRIER;
3997 	else
3998 		journal->j_flags &= ~JBD2_BARRIER;
3999 	if (test_opt(sb, DATA_ERR_ABORT))
4000 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4001 	else
4002 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4003 	write_unlock(&journal->j_state_lock);
4004 }
4005 
4006 static journal_t *ext4_get_journal(struct super_block *sb,
4007 				   unsigned int journal_inum)
4008 {
4009 	struct inode *journal_inode;
4010 	journal_t *journal;
4011 
4012 	BUG_ON(!ext4_has_feature_journal(sb));
4013 
4014 	/* First, test for the existence of a valid inode on disk.  Bad
4015 	 * things happen if we iget() an unused inode, as the subsequent
4016 	 * iput() will try to delete it. */
4017 
4018 	journal_inode = ext4_iget(sb, journal_inum);
4019 	if (IS_ERR(journal_inode)) {
4020 		ext4_msg(sb, KERN_ERR, "no journal found");
4021 		return NULL;
4022 	}
4023 	if (!journal_inode->i_nlink) {
4024 		make_bad_inode(journal_inode);
4025 		iput(journal_inode);
4026 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4027 		return NULL;
4028 	}
4029 
4030 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4031 		  journal_inode, journal_inode->i_size);
4032 	if (!S_ISREG(journal_inode->i_mode)) {
4033 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
4034 		iput(journal_inode);
4035 		return NULL;
4036 	}
4037 
4038 	journal = jbd2_journal_init_inode(journal_inode);
4039 	if (!journal) {
4040 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4041 		iput(journal_inode);
4042 		return NULL;
4043 	}
4044 	journal->j_private = sb;
4045 	ext4_init_journal_params(sb, journal);
4046 	return journal;
4047 }
4048 
4049 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4050 				       dev_t j_dev)
4051 {
4052 	struct buffer_head *bh;
4053 	journal_t *journal;
4054 	ext4_fsblk_t start;
4055 	ext4_fsblk_t len;
4056 	int hblock, blocksize;
4057 	ext4_fsblk_t sb_block;
4058 	unsigned long offset;
4059 	struct ext4_super_block *es;
4060 	struct block_device *bdev;
4061 
4062 	BUG_ON(!ext4_has_feature_journal(sb));
4063 
4064 	bdev = ext4_blkdev_get(j_dev, sb);
4065 	if (bdev == NULL)
4066 		return NULL;
4067 
4068 	blocksize = sb->s_blocksize;
4069 	hblock = bdev_logical_block_size(bdev);
4070 	if (blocksize < hblock) {
4071 		ext4_msg(sb, KERN_ERR,
4072 			"blocksize too small for journal device");
4073 		goto out_bdev;
4074 	}
4075 
4076 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4077 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4078 	set_blocksize(bdev, blocksize);
4079 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
4080 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4081 		       "external journal");
4082 		goto out_bdev;
4083 	}
4084 
4085 	es = (struct ext4_super_block *) (bh->b_data + offset);
4086 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4087 	    !(le32_to_cpu(es->s_feature_incompat) &
4088 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4089 		ext4_msg(sb, KERN_ERR, "external journal has "
4090 					"bad superblock");
4091 		brelse(bh);
4092 		goto out_bdev;
4093 	}
4094 
4095 	if ((le32_to_cpu(es->s_feature_ro_compat) &
4096 	     EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4097 	    es->s_checksum != ext4_superblock_csum(sb, es)) {
4098 		ext4_msg(sb, KERN_ERR, "external journal has "
4099 				       "corrupt superblock");
4100 		brelse(bh);
4101 		goto out_bdev;
4102 	}
4103 
4104 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4105 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4106 		brelse(bh);
4107 		goto out_bdev;
4108 	}
4109 
4110 	len = ext4_blocks_count(es);
4111 	start = sb_block + 1;
4112 	brelse(bh);	/* we're done with the superblock */
4113 
4114 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4115 					start, len, blocksize);
4116 	if (!journal) {
4117 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
4118 		goto out_bdev;
4119 	}
4120 	journal->j_private = sb;
4121 	ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4122 	wait_on_buffer(journal->j_sb_buffer);
4123 	if (!buffer_uptodate(journal->j_sb_buffer)) {
4124 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4125 		goto out_journal;
4126 	}
4127 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4128 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
4129 					"user (unsupported) - %d",
4130 			be32_to_cpu(journal->j_superblock->s_nr_users));
4131 		goto out_journal;
4132 	}
4133 	EXT4_SB(sb)->journal_bdev = bdev;
4134 	ext4_init_journal_params(sb, journal);
4135 	return journal;
4136 
4137 out_journal:
4138 	jbd2_journal_destroy(journal);
4139 out_bdev:
4140 	ext4_blkdev_put(bdev);
4141 	return NULL;
4142 }
4143 
4144 static int ext4_load_journal(struct super_block *sb,
4145 			     struct ext4_super_block *es,
4146 			     unsigned long journal_devnum)
4147 {
4148 	journal_t *journal;
4149 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4150 	dev_t journal_dev;
4151 	int err = 0;
4152 	int really_read_only;
4153 
4154 	BUG_ON(!ext4_has_feature_journal(sb));
4155 
4156 	if (journal_devnum &&
4157 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4158 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4159 			"numbers have changed");
4160 		journal_dev = new_decode_dev(journal_devnum);
4161 	} else
4162 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4163 
4164 	really_read_only = bdev_read_only(sb->s_bdev);
4165 
4166 	/*
4167 	 * Are we loading a blank journal or performing recovery after a
4168 	 * crash?  For recovery, we need to check in advance whether we
4169 	 * can get read-write access to the device.
4170 	 */
4171 	if (ext4_has_feature_journal_needs_recovery(sb)) {
4172 		if (sb->s_flags & MS_RDONLY) {
4173 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
4174 					"required on readonly filesystem");
4175 			if (really_read_only) {
4176 				ext4_msg(sb, KERN_ERR, "write access "
4177 					"unavailable, cannot proceed");
4178 				return -EROFS;
4179 			}
4180 			ext4_msg(sb, KERN_INFO, "write access will "
4181 			       "be enabled during recovery");
4182 		}
4183 	}
4184 
4185 	if (journal_inum && journal_dev) {
4186 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4187 		       "and inode journals!");
4188 		return -EINVAL;
4189 	}
4190 
4191 	if (journal_inum) {
4192 		if (!(journal = ext4_get_journal(sb, journal_inum)))
4193 			return -EINVAL;
4194 	} else {
4195 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4196 			return -EINVAL;
4197 	}
4198 
4199 	if (!(journal->j_flags & JBD2_BARRIER))
4200 		ext4_msg(sb, KERN_INFO, "barriers disabled");
4201 
4202 	if (!ext4_has_feature_journal_needs_recovery(sb))
4203 		err = jbd2_journal_wipe(journal, !really_read_only);
4204 	if (!err) {
4205 		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4206 		if (save)
4207 			memcpy(save, ((char *) es) +
4208 			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4209 		err = jbd2_journal_load(journal);
4210 		if (save)
4211 			memcpy(((char *) es) + EXT4_S_ERR_START,
4212 			       save, EXT4_S_ERR_LEN);
4213 		kfree(save);
4214 	}
4215 
4216 	if (err) {
4217 		ext4_msg(sb, KERN_ERR, "error loading journal");
4218 		jbd2_journal_destroy(journal);
4219 		return err;
4220 	}
4221 
4222 	EXT4_SB(sb)->s_journal = journal;
4223 	ext4_clear_journal_err(sb, es);
4224 
4225 	if (!really_read_only && journal_devnum &&
4226 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4227 		es->s_journal_dev = cpu_to_le32(journal_devnum);
4228 
4229 		/* Make sure we flush the recovery flag to disk. */
4230 		ext4_commit_super(sb, 1);
4231 	}
4232 
4233 	return 0;
4234 }
4235 
4236 static int ext4_commit_super(struct super_block *sb, int sync)
4237 {
4238 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4239 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4240 	int error = 0;
4241 
4242 	if (!sbh || block_device_ejected(sb))
4243 		return error;
4244 	if (buffer_write_io_error(sbh)) {
4245 		/*
4246 		 * Oh, dear.  A previous attempt to write the
4247 		 * superblock failed.  This could happen because the
4248 		 * USB device was yanked out.  Or it could happen to
4249 		 * be a transient write error and maybe the block will
4250 		 * be remapped.  Nothing we can do but to retry the
4251 		 * write and hope for the best.
4252 		 */
4253 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
4254 		       "superblock detected");
4255 		clear_buffer_write_io_error(sbh);
4256 		set_buffer_uptodate(sbh);
4257 	}
4258 	/*
4259 	 * If the file system is mounted read-only, don't update the
4260 	 * superblock write time.  This avoids updating the superblock
4261 	 * write time when we are mounting the root file system
4262 	 * read/only but we need to replay the journal; at that point,
4263 	 * for people who are east of GMT and who make their clock
4264 	 * tick in localtime for Windows bug-for-bug compatibility,
4265 	 * the clock is set in the future, and this will cause e2fsck
4266 	 * to complain and force a full file system check.
4267 	 */
4268 	if (!(sb->s_flags & MS_RDONLY))
4269 		es->s_wtime = cpu_to_le32(get_seconds());
4270 	if (sb->s_bdev->bd_part)
4271 		es->s_kbytes_written =
4272 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4273 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4274 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
4275 	else
4276 		es->s_kbytes_written =
4277 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4278 	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4279 		ext4_free_blocks_count_set(es,
4280 			EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4281 				&EXT4_SB(sb)->s_freeclusters_counter)));
4282 	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4283 		es->s_free_inodes_count =
4284 			cpu_to_le32(percpu_counter_sum_positive(
4285 				&EXT4_SB(sb)->s_freeinodes_counter));
4286 	BUFFER_TRACE(sbh, "marking dirty");
4287 	ext4_superblock_csum_set(sb);
4288 	mark_buffer_dirty(sbh);
4289 	if (sync) {
4290 		error = __sync_dirty_buffer(sbh,
4291 			test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4292 		if (error)
4293 			return error;
4294 
4295 		error = buffer_write_io_error(sbh);
4296 		if (error) {
4297 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
4298 			       "superblock");
4299 			clear_buffer_write_io_error(sbh);
4300 			set_buffer_uptodate(sbh);
4301 		}
4302 	}
4303 	return error;
4304 }
4305 
4306 /*
4307  * Have we just finished recovery?  If so, and if we are mounting (or
4308  * remounting) the filesystem readonly, then we will end up with a
4309  * consistent fs on disk.  Record that fact.
4310  */
4311 static void ext4_mark_recovery_complete(struct super_block *sb,
4312 					struct ext4_super_block *es)
4313 {
4314 	journal_t *journal = EXT4_SB(sb)->s_journal;
4315 
4316 	if (!ext4_has_feature_journal(sb)) {
4317 		BUG_ON(journal != NULL);
4318 		return;
4319 	}
4320 	jbd2_journal_lock_updates(journal);
4321 	if (jbd2_journal_flush(journal) < 0)
4322 		goto out;
4323 
4324 	if (ext4_has_feature_journal_needs_recovery(sb) &&
4325 	    sb->s_flags & MS_RDONLY) {
4326 		ext4_clear_feature_journal_needs_recovery(sb);
4327 		ext4_commit_super(sb, 1);
4328 	}
4329 
4330 out:
4331 	jbd2_journal_unlock_updates(journal);
4332 }
4333 
4334 /*
4335  * If we are mounting (or read-write remounting) a filesystem whose journal
4336  * has recorded an error from a previous lifetime, move that error to the
4337  * main filesystem now.
4338  */
4339 static void ext4_clear_journal_err(struct super_block *sb,
4340 				   struct ext4_super_block *es)
4341 {
4342 	journal_t *journal;
4343 	int j_errno;
4344 	const char *errstr;
4345 
4346 	BUG_ON(!ext4_has_feature_journal(sb));
4347 
4348 	journal = EXT4_SB(sb)->s_journal;
4349 
4350 	/*
4351 	 * Now check for any error status which may have been recorded in the
4352 	 * journal by a prior ext4_error() or ext4_abort()
4353 	 */
4354 
4355 	j_errno = jbd2_journal_errno(journal);
4356 	if (j_errno) {
4357 		char nbuf[16];
4358 
4359 		errstr = ext4_decode_error(sb, j_errno, nbuf);
4360 		ext4_warning(sb, "Filesystem error recorded "
4361 			     "from previous mount: %s", errstr);
4362 		ext4_warning(sb, "Marking fs in need of filesystem check.");
4363 
4364 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4365 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4366 		ext4_commit_super(sb, 1);
4367 
4368 		jbd2_journal_clear_err(journal);
4369 		jbd2_journal_update_sb_errno(journal);
4370 	}
4371 }
4372 
4373 /*
4374  * Force the running and committing transactions to commit,
4375  * and wait on the commit.
4376  */
4377 int ext4_force_commit(struct super_block *sb)
4378 {
4379 	journal_t *journal;
4380 
4381 	if (sb->s_flags & MS_RDONLY)
4382 		return 0;
4383 
4384 	journal = EXT4_SB(sb)->s_journal;
4385 	return ext4_journal_force_commit(journal);
4386 }
4387 
4388 static int ext4_sync_fs(struct super_block *sb, int wait)
4389 {
4390 	int ret = 0;
4391 	tid_t target;
4392 	bool needs_barrier = false;
4393 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4394 
4395 	trace_ext4_sync_fs(sb, wait);
4396 	flush_workqueue(sbi->rsv_conversion_wq);
4397 	/*
4398 	 * Writeback quota in non-journalled quota case - journalled quota has
4399 	 * no dirty dquots
4400 	 */
4401 	dquot_writeback_dquots(sb, -1);
4402 	/*
4403 	 * Data writeback is possible w/o journal transaction, so barrier must
4404 	 * being sent at the end of the function. But we can skip it if
4405 	 * transaction_commit will do it for us.
4406 	 */
4407 	if (sbi->s_journal) {
4408 		target = jbd2_get_latest_transaction(sbi->s_journal);
4409 		if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4410 		    !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4411 			needs_barrier = true;
4412 
4413 		if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4414 			if (wait)
4415 				ret = jbd2_log_wait_commit(sbi->s_journal,
4416 							   target);
4417 		}
4418 	} else if (wait && test_opt(sb, BARRIER))
4419 		needs_barrier = true;
4420 	if (needs_barrier) {
4421 		int err;
4422 		err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4423 		if (!ret)
4424 			ret = err;
4425 	}
4426 
4427 	return ret;
4428 }
4429 
4430 /*
4431  * LVM calls this function before a (read-only) snapshot is created.  This
4432  * gives us a chance to flush the journal completely and mark the fs clean.
4433  *
4434  * Note that only this function cannot bring a filesystem to be in a clean
4435  * state independently. It relies on upper layer to stop all data & metadata
4436  * modifications.
4437  */
4438 static int ext4_freeze(struct super_block *sb)
4439 {
4440 	int error = 0;
4441 	journal_t *journal;
4442 
4443 	if (sb->s_flags & MS_RDONLY)
4444 		return 0;
4445 
4446 	journal = EXT4_SB(sb)->s_journal;
4447 
4448 	if (journal) {
4449 		/* Now we set up the journal barrier. */
4450 		jbd2_journal_lock_updates(journal);
4451 
4452 		/*
4453 		 * Don't clear the needs_recovery flag if we failed to
4454 		 * flush the journal.
4455 		 */
4456 		error = jbd2_journal_flush(journal);
4457 		if (error < 0)
4458 			goto out;
4459 
4460 		/* Journal blocked and flushed, clear needs_recovery flag. */
4461 		ext4_clear_feature_journal_needs_recovery(sb);
4462 	}
4463 
4464 	error = ext4_commit_super(sb, 1);
4465 out:
4466 	if (journal)
4467 		/* we rely on upper layer to stop further updates */
4468 		jbd2_journal_unlock_updates(journal);
4469 	return error;
4470 }
4471 
4472 /*
4473  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4474  * flag here, even though the filesystem is not technically dirty yet.
4475  */
4476 static int ext4_unfreeze(struct super_block *sb)
4477 {
4478 	if (sb->s_flags & MS_RDONLY)
4479 		return 0;
4480 
4481 	if (EXT4_SB(sb)->s_journal) {
4482 		/* Reset the needs_recovery flag before the fs is unlocked. */
4483 		ext4_set_feature_journal_needs_recovery(sb);
4484 	}
4485 
4486 	ext4_commit_super(sb, 1);
4487 	return 0;
4488 }
4489 
4490 /*
4491  * Structure to save mount options for ext4_remount's benefit
4492  */
4493 struct ext4_mount_options {
4494 	unsigned long s_mount_opt;
4495 	unsigned long s_mount_opt2;
4496 	kuid_t s_resuid;
4497 	kgid_t s_resgid;
4498 	unsigned long s_commit_interval;
4499 	u32 s_min_batch_time, s_max_batch_time;
4500 #ifdef CONFIG_QUOTA
4501 	int s_jquota_fmt;
4502 	char *s_qf_names[EXT4_MAXQUOTAS];
4503 #endif
4504 };
4505 
4506 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4507 {
4508 	struct ext4_super_block *es;
4509 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4510 	unsigned long old_sb_flags;
4511 	struct ext4_mount_options old_opts;
4512 	int enable_quota = 0;
4513 	ext4_group_t g;
4514 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4515 	int err = 0;
4516 #ifdef CONFIG_QUOTA
4517 	int i, j;
4518 #endif
4519 	char *orig_data = kstrdup(data, GFP_KERNEL);
4520 
4521 	/* Store the original options */
4522 	old_sb_flags = sb->s_flags;
4523 	old_opts.s_mount_opt = sbi->s_mount_opt;
4524 	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4525 	old_opts.s_resuid = sbi->s_resuid;
4526 	old_opts.s_resgid = sbi->s_resgid;
4527 	old_opts.s_commit_interval = sbi->s_commit_interval;
4528 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
4529 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
4530 #ifdef CONFIG_QUOTA
4531 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4532 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4533 		if (sbi->s_qf_names[i]) {
4534 			old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4535 							 GFP_KERNEL);
4536 			if (!old_opts.s_qf_names[i]) {
4537 				for (j = 0; j < i; j++)
4538 					kfree(old_opts.s_qf_names[j]);
4539 				kfree(orig_data);
4540 				return -ENOMEM;
4541 			}
4542 		} else
4543 			old_opts.s_qf_names[i] = NULL;
4544 #endif
4545 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4546 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4547 
4548 	if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4549 		err = -EINVAL;
4550 		goto restore_opts;
4551 	}
4552 
4553 	if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4554 	    test_opt(sb, JOURNAL_CHECKSUM)) {
4555 		ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4556 			 "during remount not supported; ignoring");
4557 		sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4558 	}
4559 
4560 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4561 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4562 			ext4_msg(sb, KERN_ERR, "can't mount with "
4563 				 "both data=journal and delalloc");
4564 			err = -EINVAL;
4565 			goto restore_opts;
4566 		}
4567 		if (test_opt(sb, DIOREAD_NOLOCK)) {
4568 			ext4_msg(sb, KERN_ERR, "can't mount with "
4569 				 "both data=journal and dioread_nolock");
4570 			err = -EINVAL;
4571 			goto restore_opts;
4572 		}
4573 		if (test_opt(sb, DAX)) {
4574 			ext4_msg(sb, KERN_ERR, "can't mount with "
4575 				 "both data=journal and dax");
4576 			err = -EINVAL;
4577 			goto restore_opts;
4578 		}
4579 	}
4580 
4581 	if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4582 		ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4583 			"dax flag with busy inodes while remounting");
4584 		sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4585 	}
4586 
4587 	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4588 		ext4_abort(sb, "Abort forced by user");
4589 
4590 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4591 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4592 
4593 	es = sbi->s_es;
4594 
4595 	if (sbi->s_journal) {
4596 		ext4_init_journal_params(sb, sbi->s_journal);
4597 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4598 	}
4599 
4600 	if (*flags & MS_LAZYTIME)
4601 		sb->s_flags |= MS_LAZYTIME;
4602 
4603 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4604 		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4605 			err = -EROFS;
4606 			goto restore_opts;
4607 		}
4608 
4609 		if (*flags & MS_RDONLY) {
4610 			err = sync_filesystem(sb);
4611 			if (err < 0)
4612 				goto restore_opts;
4613 			err = dquot_suspend(sb, -1);
4614 			if (err < 0)
4615 				goto restore_opts;
4616 
4617 			/*
4618 			 * First of all, the unconditional stuff we have to do
4619 			 * to disable replay of the journal when we next remount
4620 			 */
4621 			sb->s_flags |= MS_RDONLY;
4622 
4623 			/*
4624 			 * OK, test if we are remounting a valid rw partition
4625 			 * readonly, and if so set the rdonly flag and then
4626 			 * mark the partition as valid again.
4627 			 */
4628 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4629 			    (sbi->s_mount_state & EXT4_VALID_FS))
4630 				es->s_state = cpu_to_le16(sbi->s_mount_state);
4631 
4632 			if (sbi->s_journal)
4633 				ext4_mark_recovery_complete(sb, es);
4634 		} else {
4635 			/* Make sure we can mount this feature set readwrite */
4636 			if (ext4_has_feature_readonly(sb) ||
4637 			    !ext4_feature_set_ok(sb, 0)) {
4638 				err = -EROFS;
4639 				goto restore_opts;
4640 			}
4641 			/*
4642 			 * Make sure the group descriptor checksums
4643 			 * are sane.  If they aren't, refuse to remount r/w.
4644 			 */
4645 			for (g = 0; g < sbi->s_groups_count; g++) {
4646 				struct ext4_group_desc *gdp =
4647 					ext4_get_group_desc(sb, g, NULL);
4648 
4649 				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4650 					ext4_msg(sb, KERN_ERR,
4651 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
4652 		g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4653 					       le16_to_cpu(gdp->bg_checksum));
4654 					err = -EFSBADCRC;
4655 					goto restore_opts;
4656 				}
4657 			}
4658 
4659 			/*
4660 			 * If we have an unprocessed orphan list hanging
4661 			 * around from a previously readonly bdev mount,
4662 			 * require a full umount/remount for now.
4663 			 */
4664 			if (es->s_last_orphan) {
4665 				ext4_msg(sb, KERN_WARNING, "Couldn't "
4666 				       "remount RDWR because of unprocessed "
4667 				       "orphan inode list.  Please "
4668 				       "umount/remount instead");
4669 				err = -EINVAL;
4670 				goto restore_opts;
4671 			}
4672 
4673 			/*
4674 			 * Mounting a RDONLY partition read-write, so reread
4675 			 * and store the current valid flag.  (It may have
4676 			 * been changed by e2fsck since we originally mounted
4677 			 * the partition.)
4678 			 */
4679 			if (sbi->s_journal)
4680 				ext4_clear_journal_err(sb, es);
4681 			sbi->s_mount_state = le16_to_cpu(es->s_state);
4682 			if (!ext4_setup_super(sb, es, 0))
4683 				sb->s_flags &= ~MS_RDONLY;
4684 			if (ext4_has_feature_mmp(sb))
4685 				if (ext4_multi_mount_protect(sb,
4686 						le64_to_cpu(es->s_mmp_block))) {
4687 					err = -EROFS;
4688 					goto restore_opts;
4689 				}
4690 			enable_quota = 1;
4691 		}
4692 	}
4693 
4694 	/*
4695 	 * Reinitialize lazy itable initialization thread based on
4696 	 * current settings
4697 	 */
4698 	if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4699 		ext4_unregister_li_request(sb);
4700 	else {
4701 		ext4_group_t first_not_zeroed;
4702 		first_not_zeroed = ext4_has_uninit_itable(sb);
4703 		ext4_register_li_request(sb, first_not_zeroed);
4704 	}
4705 
4706 	ext4_setup_system_zone(sb);
4707 	if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4708 		ext4_commit_super(sb, 1);
4709 
4710 #ifdef CONFIG_QUOTA
4711 	/* Release old quota file names */
4712 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4713 		kfree(old_opts.s_qf_names[i]);
4714 	if (enable_quota) {
4715 		if (sb_any_quota_suspended(sb))
4716 			dquot_resume(sb, -1);
4717 		else if (ext4_has_feature_quota(sb)) {
4718 			err = ext4_enable_quotas(sb);
4719 			if (err)
4720 				goto restore_opts;
4721 		}
4722 	}
4723 #endif
4724 
4725 	*flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4726 	ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4727 	kfree(orig_data);
4728 	return 0;
4729 
4730 restore_opts:
4731 	sb->s_flags = old_sb_flags;
4732 	sbi->s_mount_opt = old_opts.s_mount_opt;
4733 	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4734 	sbi->s_resuid = old_opts.s_resuid;
4735 	sbi->s_resgid = old_opts.s_resgid;
4736 	sbi->s_commit_interval = old_opts.s_commit_interval;
4737 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
4738 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
4739 #ifdef CONFIG_QUOTA
4740 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4741 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
4742 		kfree(sbi->s_qf_names[i]);
4743 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4744 	}
4745 #endif
4746 	kfree(orig_data);
4747 	return err;
4748 }
4749 
4750 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4751 {
4752 	struct super_block *sb = dentry->d_sb;
4753 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4754 	struct ext4_super_block *es = sbi->s_es;
4755 	ext4_fsblk_t overhead = 0, resv_blocks;
4756 	u64 fsid;
4757 	s64 bfree;
4758 	resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4759 
4760 	if (!test_opt(sb, MINIX_DF))
4761 		overhead = sbi->s_overhead;
4762 
4763 	buf->f_type = EXT4_SUPER_MAGIC;
4764 	buf->f_bsize = sb->s_blocksize;
4765 	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4766 	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4767 		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4768 	/* prevent underflow in case that few free space is available */
4769 	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4770 	buf->f_bavail = buf->f_bfree -
4771 			(ext4_r_blocks_count(es) + resv_blocks);
4772 	if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4773 		buf->f_bavail = 0;
4774 	buf->f_files = le32_to_cpu(es->s_inodes_count);
4775 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4776 	buf->f_namelen = EXT4_NAME_LEN;
4777 	fsid = le64_to_cpup((void *)es->s_uuid) ^
4778 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4779 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4780 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4781 
4782 	return 0;
4783 }
4784 
4785 /* Helper function for writing quotas on sync - we need to start transaction
4786  * before quota file is locked for write. Otherwise the are possible deadlocks:
4787  * Process 1                         Process 2
4788  * ext4_create()                     quota_sync()
4789  *   jbd2_journal_start()                  write_dquot()
4790  *   dquot_initialize()                         down(dqio_mutex)
4791  *     down(dqio_mutex)                    jbd2_journal_start()
4792  *
4793  */
4794 
4795 #ifdef CONFIG_QUOTA
4796 
4797 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4798 {
4799 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4800 }
4801 
4802 static int ext4_write_dquot(struct dquot *dquot)
4803 {
4804 	int ret, err;
4805 	handle_t *handle;
4806 	struct inode *inode;
4807 
4808 	inode = dquot_to_inode(dquot);
4809 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4810 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4811 	if (IS_ERR(handle))
4812 		return PTR_ERR(handle);
4813 	ret = dquot_commit(dquot);
4814 	err = ext4_journal_stop(handle);
4815 	if (!ret)
4816 		ret = err;
4817 	return ret;
4818 }
4819 
4820 static int ext4_acquire_dquot(struct dquot *dquot)
4821 {
4822 	int ret, err;
4823 	handle_t *handle;
4824 
4825 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4826 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4827 	if (IS_ERR(handle))
4828 		return PTR_ERR(handle);
4829 	ret = dquot_acquire(dquot);
4830 	err = ext4_journal_stop(handle);
4831 	if (!ret)
4832 		ret = err;
4833 	return ret;
4834 }
4835 
4836 static int ext4_release_dquot(struct dquot *dquot)
4837 {
4838 	int ret, err;
4839 	handle_t *handle;
4840 
4841 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4842 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4843 	if (IS_ERR(handle)) {
4844 		/* Release dquot anyway to avoid endless cycle in dqput() */
4845 		dquot_release(dquot);
4846 		return PTR_ERR(handle);
4847 	}
4848 	ret = dquot_release(dquot);
4849 	err = ext4_journal_stop(handle);
4850 	if (!ret)
4851 		ret = err;
4852 	return ret;
4853 }
4854 
4855 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4856 {
4857 	struct super_block *sb = dquot->dq_sb;
4858 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4859 
4860 	/* Are we journaling quotas? */
4861 	if (ext4_has_feature_quota(sb) ||
4862 	    sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4863 		dquot_mark_dquot_dirty(dquot);
4864 		return ext4_write_dquot(dquot);
4865 	} else {
4866 		return dquot_mark_dquot_dirty(dquot);
4867 	}
4868 }
4869 
4870 static int ext4_write_info(struct super_block *sb, int type)
4871 {
4872 	int ret, err;
4873 	handle_t *handle;
4874 
4875 	/* Data block + inode block */
4876 	handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
4877 	if (IS_ERR(handle))
4878 		return PTR_ERR(handle);
4879 	ret = dquot_commit_info(sb, type);
4880 	err = ext4_journal_stop(handle);
4881 	if (!ret)
4882 		ret = err;
4883 	return ret;
4884 }
4885 
4886 /*
4887  * Turn on quotas during mount time - we need to find
4888  * the quota file and such...
4889  */
4890 static int ext4_quota_on_mount(struct super_block *sb, int type)
4891 {
4892 	return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4893 					EXT4_SB(sb)->s_jquota_fmt, type);
4894 }
4895 
4896 /*
4897  * Standard function to be called on quota_on
4898  */
4899 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4900 			 struct path *path)
4901 {
4902 	int err;
4903 
4904 	if (!test_opt(sb, QUOTA))
4905 		return -EINVAL;
4906 
4907 	/* Quotafile not on the same filesystem? */
4908 	if (path->dentry->d_sb != sb)
4909 		return -EXDEV;
4910 	/* Journaling quota? */
4911 	if (EXT4_SB(sb)->s_qf_names[type]) {
4912 		/* Quotafile not in fs root? */
4913 		if (path->dentry->d_parent != sb->s_root)
4914 			ext4_msg(sb, KERN_WARNING,
4915 				"Quota file not on filesystem root. "
4916 				"Journaled quota will not work");
4917 	}
4918 
4919 	/*
4920 	 * When we journal data on quota file, we have to flush journal to see
4921 	 * all updates to the file when we bypass pagecache...
4922 	 */
4923 	if (EXT4_SB(sb)->s_journal &&
4924 	    ext4_should_journal_data(d_inode(path->dentry))) {
4925 		/*
4926 		 * We don't need to lock updates but journal_flush() could
4927 		 * otherwise be livelocked...
4928 		 */
4929 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4930 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4931 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4932 		if (err)
4933 			return err;
4934 	}
4935 
4936 	return dquot_quota_on(sb, type, format_id, path);
4937 }
4938 
4939 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4940 			     unsigned int flags)
4941 {
4942 	int err;
4943 	struct inode *qf_inode;
4944 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
4945 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4946 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4947 	};
4948 
4949 	BUG_ON(!ext4_has_feature_quota(sb));
4950 
4951 	if (!qf_inums[type])
4952 		return -EPERM;
4953 
4954 	qf_inode = ext4_iget(sb, qf_inums[type]);
4955 	if (IS_ERR(qf_inode)) {
4956 		ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4957 		return PTR_ERR(qf_inode);
4958 	}
4959 
4960 	/* Don't account quota for quota files to avoid recursion */
4961 	qf_inode->i_flags |= S_NOQUOTA;
4962 	err = dquot_enable(qf_inode, type, format_id, flags);
4963 	iput(qf_inode);
4964 
4965 	return err;
4966 }
4967 
4968 /* Enable usage tracking for all quota types. */
4969 static int ext4_enable_quotas(struct super_block *sb)
4970 {
4971 	int type, err = 0;
4972 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
4973 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4974 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4975 	};
4976 
4977 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4978 	for (type = 0; type < EXT4_MAXQUOTAS; type++) {
4979 		if (qf_inums[type]) {
4980 			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4981 						DQUOT_USAGE_ENABLED);
4982 			if (err) {
4983 				ext4_warning(sb,
4984 					"Failed to enable quota tracking "
4985 					"(type=%d, err=%d). Please run "
4986 					"e2fsck to fix.", type, err);
4987 				return err;
4988 			}
4989 		}
4990 	}
4991 	return 0;
4992 }
4993 
4994 static int ext4_quota_off(struct super_block *sb, int type)
4995 {
4996 	struct inode *inode = sb_dqopt(sb)->files[type];
4997 	handle_t *handle;
4998 
4999 	/* Force all delayed allocation blocks to be allocated.
5000 	 * Caller already holds s_umount sem */
5001 	if (test_opt(sb, DELALLOC))
5002 		sync_filesystem(sb);
5003 
5004 	if (!inode)
5005 		goto out;
5006 
5007 	/* Update modification times of quota files when userspace can
5008 	 * start looking at them */
5009 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5010 	if (IS_ERR(handle))
5011 		goto out;
5012 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5013 	ext4_mark_inode_dirty(handle, inode);
5014 	ext4_journal_stop(handle);
5015 
5016 out:
5017 	return dquot_quota_off(sb, type);
5018 }
5019 
5020 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5021  * acquiring the locks... As quota files are never truncated and quota code
5022  * itself serializes the operations (and no one else should touch the files)
5023  * we don't have to be afraid of races */
5024 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5025 			       size_t len, loff_t off)
5026 {
5027 	struct inode *inode = sb_dqopt(sb)->files[type];
5028 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5029 	int offset = off & (sb->s_blocksize - 1);
5030 	int tocopy;
5031 	size_t toread;
5032 	struct buffer_head *bh;
5033 	loff_t i_size = i_size_read(inode);
5034 
5035 	if (off > i_size)
5036 		return 0;
5037 	if (off+len > i_size)
5038 		len = i_size-off;
5039 	toread = len;
5040 	while (toread > 0) {
5041 		tocopy = sb->s_blocksize - offset < toread ?
5042 				sb->s_blocksize - offset : toread;
5043 		bh = ext4_bread(NULL, inode, blk, 0);
5044 		if (IS_ERR(bh))
5045 			return PTR_ERR(bh);
5046 		if (!bh)	/* A hole? */
5047 			memset(data, 0, tocopy);
5048 		else
5049 			memcpy(data, bh->b_data+offset, tocopy);
5050 		brelse(bh);
5051 		offset = 0;
5052 		toread -= tocopy;
5053 		data += tocopy;
5054 		blk++;
5055 	}
5056 	return len;
5057 }
5058 
5059 /* Write to quotafile (we know the transaction is already started and has
5060  * enough credits) */
5061 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5062 				const char *data, size_t len, loff_t off)
5063 {
5064 	struct inode *inode = sb_dqopt(sb)->files[type];
5065 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5066 	int err, offset = off & (sb->s_blocksize - 1);
5067 	int retries = 0;
5068 	struct buffer_head *bh;
5069 	handle_t *handle = journal_current_handle();
5070 
5071 	if (EXT4_SB(sb)->s_journal && !handle) {
5072 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5073 			" cancelled because transaction is not started",
5074 			(unsigned long long)off, (unsigned long long)len);
5075 		return -EIO;
5076 	}
5077 	/*
5078 	 * Since we account only one data block in transaction credits,
5079 	 * then it is impossible to cross a block boundary.
5080 	 */
5081 	if (sb->s_blocksize - offset < len) {
5082 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5083 			" cancelled because not block aligned",
5084 			(unsigned long long)off, (unsigned long long)len);
5085 		return -EIO;
5086 	}
5087 
5088 	do {
5089 		bh = ext4_bread(handle, inode, blk,
5090 				EXT4_GET_BLOCKS_CREATE |
5091 				EXT4_GET_BLOCKS_METADATA_NOFAIL);
5092 	} while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5093 		 ext4_should_retry_alloc(inode->i_sb, &retries));
5094 	if (IS_ERR(bh))
5095 		return PTR_ERR(bh);
5096 	if (!bh)
5097 		goto out;
5098 	BUFFER_TRACE(bh, "get write access");
5099 	err = ext4_journal_get_write_access(handle, bh);
5100 	if (err) {
5101 		brelse(bh);
5102 		return err;
5103 	}
5104 	lock_buffer(bh);
5105 	memcpy(bh->b_data+offset, data, len);
5106 	flush_dcache_page(bh->b_page);
5107 	unlock_buffer(bh);
5108 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
5109 	brelse(bh);
5110 out:
5111 	if (inode->i_size < off + len) {
5112 		i_size_write(inode, off + len);
5113 		EXT4_I(inode)->i_disksize = inode->i_size;
5114 		ext4_mark_inode_dirty(handle, inode);
5115 	}
5116 	return len;
5117 }
5118 
5119 #endif
5120 
5121 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5122 		       const char *dev_name, void *data)
5123 {
5124 	return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5125 }
5126 
5127 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5128 static inline void register_as_ext2(void)
5129 {
5130 	int err = register_filesystem(&ext2_fs_type);
5131 	if (err)
5132 		printk(KERN_WARNING
5133 		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5134 }
5135 
5136 static inline void unregister_as_ext2(void)
5137 {
5138 	unregister_filesystem(&ext2_fs_type);
5139 }
5140 
5141 static inline int ext2_feature_set_ok(struct super_block *sb)
5142 {
5143 	if (ext4_has_unknown_ext2_incompat_features(sb))
5144 		return 0;
5145 	if (sb->s_flags & MS_RDONLY)
5146 		return 1;
5147 	if (ext4_has_unknown_ext2_ro_compat_features(sb))
5148 		return 0;
5149 	return 1;
5150 }
5151 #else
5152 static inline void register_as_ext2(void) { }
5153 static inline void unregister_as_ext2(void) { }
5154 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5155 #endif
5156 
5157 static inline void register_as_ext3(void)
5158 {
5159 	int err = register_filesystem(&ext3_fs_type);
5160 	if (err)
5161 		printk(KERN_WARNING
5162 		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5163 }
5164 
5165 static inline void unregister_as_ext3(void)
5166 {
5167 	unregister_filesystem(&ext3_fs_type);
5168 }
5169 
5170 static inline int ext3_feature_set_ok(struct super_block *sb)
5171 {
5172 	if (ext4_has_unknown_ext3_incompat_features(sb))
5173 		return 0;
5174 	if (!ext4_has_feature_journal(sb))
5175 		return 0;
5176 	if (sb->s_flags & MS_RDONLY)
5177 		return 1;
5178 	if (ext4_has_unknown_ext3_ro_compat_features(sb))
5179 		return 0;
5180 	return 1;
5181 }
5182 
5183 static struct file_system_type ext4_fs_type = {
5184 	.owner		= THIS_MODULE,
5185 	.name		= "ext4",
5186 	.mount		= ext4_mount,
5187 	.kill_sb	= kill_block_super,
5188 	.fs_flags	= FS_REQUIRES_DEV,
5189 };
5190 MODULE_ALIAS_FS("ext4");
5191 
5192 /* Shared across all ext4 file systems */
5193 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5194 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5195 
5196 static int __init ext4_init_fs(void)
5197 {
5198 	int i, err;
5199 
5200 	ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5201 	ext4_li_info = NULL;
5202 	mutex_init(&ext4_li_mtx);
5203 
5204 	/* Build-time check for flags consistency */
5205 	ext4_check_flag_values();
5206 
5207 	for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5208 		mutex_init(&ext4__aio_mutex[i]);
5209 		init_waitqueue_head(&ext4__ioend_wq[i]);
5210 	}
5211 
5212 	err = ext4_init_es();
5213 	if (err)
5214 		return err;
5215 
5216 	err = ext4_init_pageio();
5217 	if (err)
5218 		goto out5;
5219 
5220 	err = ext4_init_system_zone();
5221 	if (err)
5222 		goto out4;
5223 
5224 	err = ext4_init_sysfs();
5225 	if (err)
5226 		goto out3;
5227 
5228 	err = ext4_init_mballoc();
5229 	if (err)
5230 		goto out2;
5231 	else
5232 		ext4_mballoc_ready = 1;
5233 	err = init_inodecache();
5234 	if (err)
5235 		goto out1;
5236 	register_as_ext3();
5237 	register_as_ext2();
5238 	err = register_filesystem(&ext4_fs_type);
5239 	if (err)
5240 		goto out;
5241 
5242 	return 0;
5243 out:
5244 	unregister_as_ext2();
5245 	unregister_as_ext3();
5246 	destroy_inodecache();
5247 out1:
5248 	ext4_mballoc_ready = 0;
5249 	ext4_exit_mballoc();
5250 out2:
5251 	ext4_exit_sysfs();
5252 out3:
5253 	ext4_exit_system_zone();
5254 out4:
5255 	ext4_exit_pageio();
5256 out5:
5257 	ext4_exit_es();
5258 
5259 	return err;
5260 }
5261 
5262 static void __exit ext4_exit_fs(void)
5263 {
5264 	ext4_exit_crypto();
5265 	ext4_destroy_lazyinit_thread();
5266 	unregister_as_ext2();
5267 	unregister_as_ext3();
5268 	unregister_filesystem(&ext4_fs_type);
5269 	destroy_inodecache();
5270 	ext4_exit_mballoc();
5271 	ext4_exit_sysfs();
5272 	ext4_exit_system_zone();
5273 	ext4_exit_pageio();
5274 	ext4_exit_es();
5275 }
5276 
5277 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5278 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5279 MODULE_LICENSE("GPL");
5280 module_init(ext4_init_fs)
5281 module_exit(ext4_exit_fs)
5282