xref: /openbmc/linux/fs/ext4/super.c (revision d7a3d85e)
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/parser.h>
28 #include <linux/buffer_head.h>
29 #include <linux/exportfs.h>
30 #include <linux/vfs.h>
31 #include <linux/random.h>
32 #include <linux/mount.h>
33 #include <linux/namei.h>
34 #include <linux/quotaops.h>
35 #include <linux/seq_file.h>
36 #include <linux/proc_fs.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 proc_dir_entry *ext4_proc_root;
57 static struct kset *ext4_kset;
58 static struct ext4_lazy_init *ext4_li_info;
59 static struct mutex ext4_li_mtx;
60 static struct ext4_features *ext4_feat;
61 static int ext4_mballoc_ready;
62 
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 			     unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 					struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 				   struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 		       const char *dev_name, void *data);
78 static inline int ext2_feature_set_ok(struct super_block *sb);
79 static inline int ext3_feature_set_ok(struct super_block *sb);
80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
84 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
85 
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type = {
88 	.owner		= THIS_MODULE,
89 	.name		= "ext2",
90 	.mount		= ext4_mount,
91 	.kill_sb	= kill_block_super,
92 	.fs_flags	= FS_REQUIRES_DEV,
93 };
94 MODULE_ALIAS_FS("ext2");
95 MODULE_ALIAS("ext2");
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #else
98 #define IS_EXT2_SB(sb) (0)
99 #endif
100 
101 
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104 	.owner		= THIS_MODULE,
105 	.name		= "ext3",
106 	.mount		= ext4_mount,
107 	.kill_sb	= kill_block_super,
108 	.fs_flags	= FS_REQUIRES_DEV,
109 };
110 MODULE_ALIAS_FS("ext3");
111 MODULE_ALIAS("ext3");
112 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
113 #else
114 #define IS_EXT3_SB(sb) (0)
115 #endif
116 
117 static int ext4_verify_csum_type(struct super_block *sb,
118 				 struct ext4_super_block *es)
119 {
120 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 					EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
122 		return 1;
123 
124 	return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
125 }
126 
127 static __le32 ext4_superblock_csum(struct super_block *sb,
128 				   struct ext4_super_block *es)
129 {
130 	struct ext4_sb_info *sbi = EXT4_SB(sb);
131 	int offset = offsetof(struct ext4_super_block, s_checksum);
132 	__u32 csum;
133 
134 	csum = ext4_chksum(sbi, ~0, (char *)es, offset);
135 
136 	return cpu_to_le32(csum);
137 }
138 
139 static int ext4_superblock_csum_verify(struct super_block *sb,
140 				       struct ext4_super_block *es)
141 {
142 	if (!ext4_has_metadata_csum(sb))
143 		return 1;
144 
145 	return es->s_checksum == ext4_superblock_csum(sb, es);
146 }
147 
148 void ext4_superblock_csum_set(struct super_block *sb)
149 {
150 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
151 
152 	if (!ext4_has_metadata_csum(sb))
153 		return;
154 
155 	es->s_checksum = ext4_superblock_csum(sb, es);
156 }
157 
158 void *ext4_kvmalloc(size_t size, gfp_t flags)
159 {
160 	void *ret;
161 
162 	ret = kmalloc(size, flags | __GFP_NOWARN);
163 	if (!ret)
164 		ret = __vmalloc(size, flags, PAGE_KERNEL);
165 	return ret;
166 }
167 
168 void *ext4_kvzalloc(size_t size, gfp_t flags)
169 {
170 	void *ret;
171 
172 	ret = kzalloc(size, flags | __GFP_NOWARN);
173 	if (!ret)
174 		ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
175 	return ret;
176 }
177 
178 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
179 			       struct ext4_group_desc *bg)
180 {
181 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
182 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
183 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
184 }
185 
186 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
187 			       struct ext4_group_desc *bg)
188 {
189 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
190 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
192 }
193 
194 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
195 			      struct ext4_group_desc *bg)
196 {
197 	return le32_to_cpu(bg->bg_inode_table_lo) |
198 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
200 }
201 
202 __u32 ext4_free_group_clusters(struct super_block *sb,
203 			       struct ext4_group_desc *bg)
204 {
205 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
206 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
208 }
209 
210 __u32 ext4_free_inodes_count(struct super_block *sb,
211 			      struct ext4_group_desc *bg)
212 {
213 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
214 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
216 }
217 
218 __u32 ext4_used_dirs_count(struct super_block *sb,
219 			      struct ext4_group_desc *bg)
220 {
221 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
222 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
224 }
225 
226 __u32 ext4_itable_unused_count(struct super_block *sb,
227 			      struct ext4_group_desc *bg)
228 {
229 	return le16_to_cpu(bg->bg_itable_unused_lo) |
230 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
231 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
232 }
233 
234 void ext4_block_bitmap_set(struct super_block *sb,
235 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
236 {
237 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
238 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
240 }
241 
242 void ext4_inode_bitmap_set(struct super_block *sb,
243 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
244 {
245 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
246 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
248 }
249 
250 void ext4_inode_table_set(struct super_block *sb,
251 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
252 {
253 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
254 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
256 }
257 
258 void ext4_free_group_clusters_set(struct super_block *sb,
259 				  struct ext4_group_desc *bg, __u32 count)
260 {
261 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
262 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
264 }
265 
266 void ext4_free_inodes_set(struct super_block *sb,
267 			  struct ext4_group_desc *bg, __u32 count)
268 {
269 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
270 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
272 }
273 
274 void ext4_used_dirs_set(struct super_block *sb,
275 			  struct ext4_group_desc *bg, __u32 count)
276 {
277 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
278 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
280 }
281 
282 void ext4_itable_unused_set(struct super_block *sb,
283 			  struct ext4_group_desc *bg, __u32 count)
284 {
285 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
286 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
288 }
289 
290 
291 static void __save_error_info(struct super_block *sb, const char *func,
292 			    unsigned int line)
293 {
294 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
295 
296 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
297 	if (bdev_read_only(sb->s_bdev))
298 		return;
299 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
300 	es->s_last_error_time = cpu_to_le32(get_seconds());
301 	strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
302 	es->s_last_error_line = cpu_to_le32(line);
303 	if (!es->s_first_error_time) {
304 		es->s_first_error_time = es->s_last_error_time;
305 		strncpy(es->s_first_error_func, func,
306 			sizeof(es->s_first_error_func));
307 		es->s_first_error_line = cpu_to_le32(line);
308 		es->s_first_error_ino = es->s_last_error_ino;
309 		es->s_first_error_block = es->s_last_error_block;
310 	}
311 	/*
312 	 * Start the daily error reporting function if it hasn't been
313 	 * started already
314 	 */
315 	if (!es->s_error_count)
316 		mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
317 	le32_add_cpu(&es->s_error_count, 1);
318 }
319 
320 static void save_error_info(struct super_block *sb, const char *func,
321 			    unsigned int line)
322 {
323 	__save_error_info(sb, func, line);
324 	ext4_commit_super(sb, 1);
325 }
326 
327 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
328 {
329 	struct super_block		*sb = journal->j_private;
330 	struct ext4_sb_info		*sbi = EXT4_SB(sb);
331 	int				error = is_journal_aborted(journal);
332 	struct ext4_journal_cb_entry	*jce;
333 
334 	BUG_ON(txn->t_state == T_FINISHED);
335 	spin_lock(&sbi->s_md_lock);
336 	while (!list_empty(&txn->t_private_list)) {
337 		jce = list_entry(txn->t_private_list.next,
338 				 struct ext4_journal_cb_entry, jce_list);
339 		list_del_init(&jce->jce_list);
340 		spin_unlock(&sbi->s_md_lock);
341 		jce->jce_func(sb, jce, error);
342 		spin_lock(&sbi->s_md_lock);
343 	}
344 	spin_unlock(&sbi->s_md_lock);
345 }
346 
347 /* Deal with the reporting of failure conditions on a filesystem such as
348  * inconsistencies detected or read IO failures.
349  *
350  * On ext2, we can store the error state of the filesystem in the
351  * superblock.  That is not possible on ext4, because we may have other
352  * write ordering constraints on the superblock which prevent us from
353  * writing it out straight away; and given that the journal is about to
354  * be aborted, we can't rely on the current, or future, transactions to
355  * write out the superblock safely.
356  *
357  * We'll just use the jbd2_journal_abort() error code to record an error in
358  * the journal instead.  On recovery, the journal will complain about
359  * that error until we've noted it down and cleared it.
360  */
361 
362 static void ext4_handle_error(struct super_block *sb)
363 {
364 	if (sb->s_flags & MS_RDONLY)
365 		return;
366 
367 	if (!test_opt(sb, ERRORS_CONT)) {
368 		journal_t *journal = EXT4_SB(sb)->s_journal;
369 
370 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
371 		if (journal)
372 			jbd2_journal_abort(journal, -EIO);
373 	}
374 	if (test_opt(sb, ERRORS_RO)) {
375 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
376 		/*
377 		 * Make sure updated value of ->s_mount_flags will be visible
378 		 * before ->s_flags update
379 		 */
380 		smp_wmb();
381 		sb->s_flags |= MS_RDONLY;
382 	}
383 	if (test_opt(sb, ERRORS_PANIC))
384 		panic("EXT4-fs (device %s): panic forced after error\n",
385 			sb->s_id);
386 }
387 
388 #define ext4_error_ratelimit(sb)					\
389 		___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),	\
390 			     "EXT4-fs error")
391 
392 void __ext4_error(struct super_block *sb, const char *function,
393 		  unsigned int line, const char *fmt, ...)
394 {
395 	struct va_format vaf;
396 	va_list args;
397 
398 	if (ext4_error_ratelimit(sb)) {
399 		va_start(args, fmt);
400 		vaf.fmt = fmt;
401 		vaf.va = &args;
402 		printk(KERN_CRIT
403 		       "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
404 		       sb->s_id, function, line, current->comm, &vaf);
405 		va_end(args);
406 	}
407 	save_error_info(sb, function, line);
408 	ext4_handle_error(sb);
409 }
410 
411 void __ext4_error_inode(struct inode *inode, const char *function,
412 			unsigned int line, ext4_fsblk_t block,
413 			const char *fmt, ...)
414 {
415 	va_list args;
416 	struct va_format vaf;
417 	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
418 
419 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
420 	es->s_last_error_block = cpu_to_le64(block);
421 	if (ext4_error_ratelimit(inode->i_sb)) {
422 		va_start(args, fmt);
423 		vaf.fmt = fmt;
424 		vaf.va = &args;
425 		if (block)
426 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
427 			       "inode #%lu: block %llu: comm %s: %pV\n",
428 			       inode->i_sb->s_id, function, line, inode->i_ino,
429 			       block, current->comm, &vaf);
430 		else
431 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
432 			       "inode #%lu: comm %s: %pV\n",
433 			       inode->i_sb->s_id, function, line, inode->i_ino,
434 			       current->comm, &vaf);
435 		va_end(args);
436 	}
437 	save_error_info(inode->i_sb, function, line);
438 	ext4_handle_error(inode->i_sb);
439 }
440 
441 void __ext4_error_file(struct file *file, const char *function,
442 		       unsigned int line, ext4_fsblk_t block,
443 		       const char *fmt, ...)
444 {
445 	va_list args;
446 	struct va_format vaf;
447 	struct ext4_super_block *es;
448 	struct inode *inode = file_inode(file);
449 	char pathname[80], *path;
450 
451 	es = EXT4_SB(inode->i_sb)->s_es;
452 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
453 	if (ext4_error_ratelimit(inode->i_sb)) {
454 		path = d_path(&(file->f_path), pathname, sizeof(pathname));
455 		if (IS_ERR(path))
456 			path = "(unknown)";
457 		va_start(args, fmt);
458 		vaf.fmt = fmt;
459 		vaf.va = &args;
460 		if (block)
461 			printk(KERN_CRIT
462 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
463 			       "block %llu: comm %s: path %s: %pV\n",
464 			       inode->i_sb->s_id, function, line, inode->i_ino,
465 			       block, current->comm, path, &vaf);
466 		else
467 			printk(KERN_CRIT
468 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
469 			       "comm %s: path %s: %pV\n",
470 			       inode->i_sb->s_id, function, line, inode->i_ino,
471 			       current->comm, path, &vaf);
472 		va_end(args);
473 	}
474 	save_error_info(inode->i_sb, function, line);
475 	ext4_handle_error(inode->i_sb);
476 }
477 
478 const char *ext4_decode_error(struct super_block *sb, int errno,
479 			      char nbuf[16])
480 {
481 	char *errstr = NULL;
482 
483 	switch (errno) {
484 	case -EIO:
485 		errstr = "IO failure";
486 		break;
487 	case -ENOMEM:
488 		errstr = "Out of memory";
489 		break;
490 	case -EROFS:
491 		if (!sb || (EXT4_SB(sb)->s_journal &&
492 			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
493 			errstr = "Journal has aborted";
494 		else
495 			errstr = "Readonly filesystem";
496 		break;
497 	default:
498 		/* If the caller passed in an extra buffer for unknown
499 		 * errors, textualise them now.  Else we just return
500 		 * NULL. */
501 		if (nbuf) {
502 			/* Check for truncated error codes... */
503 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
504 				errstr = nbuf;
505 		}
506 		break;
507 	}
508 
509 	return errstr;
510 }
511 
512 /* __ext4_std_error decodes expected errors from journaling functions
513  * automatically and invokes the appropriate error response.  */
514 
515 void __ext4_std_error(struct super_block *sb, const char *function,
516 		      unsigned int line, int errno)
517 {
518 	char nbuf[16];
519 	const char *errstr;
520 
521 	/* Special case: if the error is EROFS, and we're not already
522 	 * inside a transaction, then there's really no point in logging
523 	 * an error. */
524 	if (errno == -EROFS && journal_current_handle() == NULL &&
525 	    (sb->s_flags & MS_RDONLY))
526 		return;
527 
528 	if (ext4_error_ratelimit(sb)) {
529 		errstr = ext4_decode_error(sb, errno, nbuf);
530 		printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
531 		       sb->s_id, function, line, errstr);
532 	}
533 
534 	save_error_info(sb, function, line);
535 	ext4_handle_error(sb);
536 }
537 
538 /*
539  * ext4_abort is a much stronger failure handler than ext4_error.  The
540  * abort function may be used to deal with unrecoverable failures such
541  * as journal IO errors or ENOMEM at a critical moment in log management.
542  *
543  * We unconditionally force the filesystem into an ABORT|READONLY state,
544  * unless the error response on the fs has been set to panic in which
545  * case we take the easy way out and panic immediately.
546  */
547 
548 void __ext4_abort(struct super_block *sb, const char *function,
549 		unsigned int line, const char *fmt, ...)
550 {
551 	va_list args;
552 
553 	save_error_info(sb, function, line);
554 	va_start(args, fmt);
555 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
556 	       function, line);
557 	vprintk(fmt, args);
558 	printk("\n");
559 	va_end(args);
560 
561 	if ((sb->s_flags & MS_RDONLY) == 0) {
562 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
563 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
564 		/*
565 		 * Make sure updated value of ->s_mount_flags will be visible
566 		 * before ->s_flags update
567 		 */
568 		smp_wmb();
569 		sb->s_flags |= MS_RDONLY;
570 		if (EXT4_SB(sb)->s_journal)
571 			jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
572 		save_error_info(sb, function, line);
573 	}
574 	if (test_opt(sb, ERRORS_PANIC))
575 		panic("EXT4-fs panic from previous error\n");
576 }
577 
578 void __ext4_msg(struct super_block *sb,
579 		const char *prefix, const char *fmt, ...)
580 {
581 	struct va_format vaf;
582 	va_list args;
583 
584 	if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
585 		return;
586 
587 	va_start(args, fmt);
588 	vaf.fmt = fmt;
589 	vaf.va = &args;
590 	printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
591 	va_end(args);
592 }
593 
594 void __ext4_warning(struct super_block *sb, const char *function,
595 		    unsigned int line, const char *fmt, ...)
596 {
597 	struct va_format vaf;
598 	va_list args;
599 
600 	if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
601 			  "EXT4-fs warning"))
602 		return;
603 
604 	va_start(args, fmt);
605 	vaf.fmt = fmt;
606 	vaf.va = &args;
607 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
608 	       sb->s_id, function, line, &vaf);
609 	va_end(args);
610 }
611 
612 void __ext4_grp_locked_error(const char *function, unsigned int line,
613 			     struct super_block *sb, ext4_group_t grp,
614 			     unsigned long ino, ext4_fsblk_t block,
615 			     const char *fmt, ...)
616 __releases(bitlock)
617 __acquires(bitlock)
618 {
619 	struct va_format vaf;
620 	va_list args;
621 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
622 
623 	es->s_last_error_ino = cpu_to_le32(ino);
624 	es->s_last_error_block = cpu_to_le64(block);
625 	__save_error_info(sb, function, line);
626 
627 	if (ext4_error_ratelimit(sb)) {
628 		va_start(args, fmt);
629 		vaf.fmt = fmt;
630 		vaf.va = &args;
631 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
632 		       sb->s_id, function, line, grp);
633 		if (ino)
634 			printk(KERN_CONT "inode %lu: ", ino);
635 		if (block)
636 			printk(KERN_CONT "block %llu:",
637 			       (unsigned long long) block);
638 		printk(KERN_CONT "%pV\n", &vaf);
639 		va_end(args);
640 	}
641 
642 	if (test_opt(sb, ERRORS_CONT)) {
643 		ext4_commit_super(sb, 0);
644 		return;
645 	}
646 
647 	ext4_unlock_group(sb, grp);
648 	ext4_handle_error(sb);
649 	/*
650 	 * We only get here in the ERRORS_RO case; relocking the group
651 	 * may be dangerous, but nothing bad will happen since the
652 	 * filesystem will have already been marked read/only and the
653 	 * journal has been aborted.  We return 1 as a hint to callers
654 	 * who might what to use the return value from
655 	 * ext4_grp_locked_error() to distinguish between the
656 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
657 	 * aggressively from the ext4 function in question, with a
658 	 * more appropriate error code.
659 	 */
660 	ext4_lock_group(sb, grp);
661 	return;
662 }
663 
664 void ext4_update_dynamic_rev(struct super_block *sb)
665 {
666 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
667 
668 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
669 		return;
670 
671 	ext4_warning(sb,
672 		     "updating to rev %d because of new feature flag, "
673 		     "running e2fsck is recommended",
674 		     EXT4_DYNAMIC_REV);
675 
676 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
677 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
678 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
679 	/* leave es->s_feature_*compat flags alone */
680 	/* es->s_uuid will be set by e2fsck if empty */
681 
682 	/*
683 	 * The rest of the superblock fields should be zero, and if not it
684 	 * means they are likely already in use, so leave them alone.  We
685 	 * can leave it up to e2fsck to clean up any inconsistencies there.
686 	 */
687 }
688 
689 /*
690  * Open the external journal device
691  */
692 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
693 {
694 	struct block_device *bdev;
695 	char b[BDEVNAME_SIZE];
696 
697 	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
698 	if (IS_ERR(bdev))
699 		goto fail;
700 	return bdev;
701 
702 fail:
703 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
704 			__bdevname(dev, b), PTR_ERR(bdev));
705 	return NULL;
706 }
707 
708 /*
709  * Release the journal device
710  */
711 static void ext4_blkdev_put(struct block_device *bdev)
712 {
713 	blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
714 }
715 
716 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
717 {
718 	struct block_device *bdev;
719 	bdev = sbi->journal_bdev;
720 	if (bdev) {
721 		ext4_blkdev_put(bdev);
722 		sbi->journal_bdev = NULL;
723 	}
724 }
725 
726 static inline struct inode *orphan_list_entry(struct list_head *l)
727 {
728 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
729 }
730 
731 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
732 {
733 	struct list_head *l;
734 
735 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
736 		 le32_to_cpu(sbi->s_es->s_last_orphan));
737 
738 	printk(KERN_ERR "sb_info orphan list:\n");
739 	list_for_each(l, &sbi->s_orphan) {
740 		struct inode *inode = orphan_list_entry(l);
741 		printk(KERN_ERR "  "
742 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
743 		       inode->i_sb->s_id, inode->i_ino, inode,
744 		       inode->i_mode, inode->i_nlink,
745 		       NEXT_ORPHAN(inode));
746 	}
747 }
748 
749 static void ext4_put_super(struct super_block *sb)
750 {
751 	struct ext4_sb_info *sbi = EXT4_SB(sb);
752 	struct ext4_super_block *es = sbi->s_es;
753 	int i, err;
754 
755 	ext4_unregister_li_request(sb);
756 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
757 
758 	flush_workqueue(sbi->rsv_conversion_wq);
759 	destroy_workqueue(sbi->rsv_conversion_wq);
760 
761 	if (sbi->s_journal) {
762 		err = jbd2_journal_destroy(sbi->s_journal);
763 		sbi->s_journal = NULL;
764 		if (err < 0)
765 			ext4_abort(sb, "Couldn't clean up the journal");
766 	}
767 
768 	ext4_es_unregister_shrinker(sbi);
769 	del_timer_sync(&sbi->s_err_report);
770 	ext4_release_system_zone(sb);
771 	ext4_mb_release(sb);
772 	ext4_ext_release(sb);
773 	ext4_xattr_put_super(sb);
774 
775 	if (!(sb->s_flags & MS_RDONLY)) {
776 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
777 		es->s_state = cpu_to_le16(sbi->s_mount_state);
778 	}
779 	if (!(sb->s_flags & MS_RDONLY))
780 		ext4_commit_super(sb, 1);
781 
782 	if (sbi->s_proc) {
783 		remove_proc_entry("options", sbi->s_proc);
784 		remove_proc_entry(sb->s_id, ext4_proc_root);
785 	}
786 	kobject_del(&sbi->s_kobj);
787 
788 	for (i = 0; i < sbi->s_gdb_count; i++)
789 		brelse(sbi->s_group_desc[i]);
790 	kvfree(sbi->s_group_desc);
791 	kvfree(sbi->s_flex_groups);
792 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
793 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
794 	percpu_counter_destroy(&sbi->s_dirs_counter);
795 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
796 	brelse(sbi->s_sbh);
797 #ifdef CONFIG_QUOTA
798 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
799 		kfree(sbi->s_qf_names[i]);
800 #endif
801 
802 	/* Debugging code just in case the in-memory inode orphan list
803 	 * isn't empty.  The on-disk one can be non-empty if we've
804 	 * detected an error and taken the fs readonly, but the
805 	 * in-memory list had better be clean by this point. */
806 	if (!list_empty(&sbi->s_orphan))
807 		dump_orphan_list(sb, sbi);
808 	J_ASSERT(list_empty(&sbi->s_orphan));
809 
810 	invalidate_bdev(sb->s_bdev);
811 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
812 		/*
813 		 * Invalidate the journal device's buffers.  We don't want them
814 		 * floating about in memory - the physical journal device may
815 		 * hotswapped, and it breaks the `ro-after' testing code.
816 		 */
817 		sync_blockdev(sbi->journal_bdev);
818 		invalidate_bdev(sbi->journal_bdev);
819 		ext4_blkdev_remove(sbi);
820 	}
821 	if (sbi->s_mb_cache) {
822 		ext4_xattr_destroy_cache(sbi->s_mb_cache);
823 		sbi->s_mb_cache = NULL;
824 	}
825 	if (sbi->s_mmp_tsk)
826 		kthread_stop(sbi->s_mmp_tsk);
827 	sb->s_fs_info = NULL;
828 	/*
829 	 * Now that we are completely done shutting down the
830 	 * superblock, we need to actually destroy the kobject.
831 	 */
832 	kobject_put(&sbi->s_kobj);
833 	wait_for_completion(&sbi->s_kobj_unregister);
834 	if (sbi->s_chksum_driver)
835 		crypto_free_shash(sbi->s_chksum_driver);
836 	kfree(sbi->s_blockgroup_lock);
837 	kfree(sbi);
838 }
839 
840 static struct kmem_cache *ext4_inode_cachep;
841 
842 /*
843  * Called inside transaction, so use GFP_NOFS
844  */
845 static struct inode *ext4_alloc_inode(struct super_block *sb)
846 {
847 	struct ext4_inode_info *ei;
848 
849 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
850 	if (!ei)
851 		return NULL;
852 
853 	ei->vfs_inode.i_version = 1;
854 	spin_lock_init(&ei->i_raw_lock);
855 	INIT_LIST_HEAD(&ei->i_prealloc_list);
856 	spin_lock_init(&ei->i_prealloc_lock);
857 	ext4_es_init_tree(&ei->i_es_tree);
858 	rwlock_init(&ei->i_es_lock);
859 	INIT_LIST_HEAD(&ei->i_es_list);
860 	ei->i_es_all_nr = 0;
861 	ei->i_es_shk_nr = 0;
862 	ei->i_es_shrink_lblk = 0;
863 	ei->i_reserved_data_blocks = 0;
864 	ei->i_reserved_meta_blocks = 0;
865 	ei->i_allocated_meta_blocks = 0;
866 	ei->i_da_metadata_calc_len = 0;
867 	ei->i_da_metadata_calc_last_lblock = 0;
868 	spin_lock_init(&(ei->i_block_reservation_lock));
869 #ifdef CONFIG_QUOTA
870 	ei->i_reserved_quota = 0;
871 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
872 #endif
873 	ei->jinode = NULL;
874 	INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
875 	spin_lock_init(&ei->i_completed_io_lock);
876 	ei->i_sync_tid = 0;
877 	ei->i_datasync_tid = 0;
878 	atomic_set(&ei->i_ioend_count, 0);
879 	atomic_set(&ei->i_unwritten, 0);
880 	INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
881 #ifdef CONFIG_EXT4_FS_ENCRYPTION
882 	ei->i_encryption_key.mode = EXT4_ENCRYPTION_MODE_INVALID;
883 #endif
884 
885 	return &ei->vfs_inode;
886 }
887 
888 static int ext4_drop_inode(struct inode *inode)
889 {
890 	int drop = generic_drop_inode(inode);
891 
892 	trace_ext4_drop_inode(inode, drop);
893 	return drop;
894 }
895 
896 static void ext4_i_callback(struct rcu_head *head)
897 {
898 	struct inode *inode = container_of(head, struct inode, i_rcu);
899 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
900 }
901 
902 static void ext4_destroy_inode(struct inode *inode)
903 {
904 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
905 		ext4_msg(inode->i_sb, KERN_ERR,
906 			 "Inode %lu (%p): orphan list check failed!",
907 			 inode->i_ino, EXT4_I(inode));
908 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
909 				EXT4_I(inode), sizeof(struct ext4_inode_info),
910 				true);
911 		dump_stack();
912 	}
913 	call_rcu(&inode->i_rcu, ext4_i_callback);
914 }
915 
916 static void init_once(void *foo)
917 {
918 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
919 
920 	INIT_LIST_HEAD(&ei->i_orphan);
921 	init_rwsem(&ei->xattr_sem);
922 	init_rwsem(&ei->i_data_sem);
923 	inode_init_once(&ei->vfs_inode);
924 }
925 
926 static int __init init_inodecache(void)
927 {
928 	ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
929 					     sizeof(struct ext4_inode_info),
930 					     0, (SLAB_RECLAIM_ACCOUNT|
931 						SLAB_MEM_SPREAD),
932 					     init_once);
933 	if (ext4_inode_cachep == NULL)
934 		return -ENOMEM;
935 	return 0;
936 }
937 
938 static void destroy_inodecache(void)
939 {
940 	/*
941 	 * Make sure all delayed rcu free inodes are flushed before we
942 	 * destroy cache.
943 	 */
944 	rcu_barrier();
945 	kmem_cache_destroy(ext4_inode_cachep);
946 }
947 
948 void ext4_clear_inode(struct inode *inode)
949 {
950 	invalidate_inode_buffers(inode);
951 	clear_inode(inode);
952 	dquot_drop(inode);
953 	ext4_discard_preallocations(inode);
954 	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
955 	if (EXT4_I(inode)->jinode) {
956 		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
957 					       EXT4_I(inode)->jinode);
958 		jbd2_free_inode(EXT4_I(inode)->jinode);
959 		EXT4_I(inode)->jinode = NULL;
960 	}
961 }
962 
963 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
964 					u64 ino, u32 generation)
965 {
966 	struct inode *inode;
967 
968 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
969 		return ERR_PTR(-ESTALE);
970 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
971 		return ERR_PTR(-ESTALE);
972 
973 	/* iget isn't really right if the inode is currently unallocated!!
974 	 *
975 	 * ext4_read_inode will return a bad_inode if the inode had been
976 	 * deleted, so we should be safe.
977 	 *
978 	 * Currently we don't know the generation for parent directory, so
979 	 * a generation of 0 means "accept any"
980 	 */
981 	inode = ext4_iget_normal(sb, ino);
982 	if (IS_ERR(inode))
983 		return ERR_CAST(inode);
984 	if (generation && inode->i_generation != generation) {
985 		iput(inode);
986 		return ERR_PTR(-ESTALE);
987 	}
988 
989 	return inode;
990 }
991 
992 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
993 					int fh_len, int fh_type)
994 {
995 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
996 				    ext4_nfs_get_inode);
997 }
998 
999 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1000 					int fh_len, int fh_type)
1001 {
1002 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1003 				    ext4_nfs_get_inode);
1004 }
1005 
1006 /*
1007  * Try to release metadata pages (indirect blocks, directories) which are
1008  * mapped via the block device.  Since these pages could have journal heads
1009  * which would prevent try_to_free_buffers() from freeing them, we must use
1010  * jbd2 layer's try_to_free_buffers() function to release them.
1011  */
1012 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1013 				 gfp_t wait)
1014 {
1015 	journal_t *journal = EXT4_SB(sb)->s_journal;
1016 
1017 	WARN_ON(PageChecked(page));
1018 	if (!page_has_buffers(page))
1019 		return 0;
1020 	if (journal)
1021 		return jbd2_journal_try_to_free_buffers(journal, page,
1022 							wait & ~__GFP_WAIT);
1023 	return try_to_free_buffers(page);
1024 }
1025 
1026 #ifdef CONFIG_QUOTA
1027 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1028 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1029 
1030 static int ext4_write_dquot(struct dquot *dquot);
1031 static int ext4_acquire_dquot(struct dquot *dquot);
1032 static int ext4_release_dquot(struct dquot *dquot);
1033 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1034 static int ext4_write_info(struct super_block *sb, int type);
1035 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1036 			 struct path *path);
1037 static int ext4_quota_off(struct super_block *sb, int type);
1038 static int ext4_quota_on_mount(struct super_block *sb, int type);
1039 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1040 			       size_t len, loff_t off);
1041 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1042 				const char *data, size_t len, loff_t off);
1043 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1044 			     unsigned int flags);
1045 static int ext4_enable_quotas(struct super_block *sb);
1046 
1047 static struct dquot **ext4_get_dquots(struct inode *inode)
1048 {
1049 	return EXT4_I(inode)->i_dquot;
1050 }
1051 
1052 static const struct dquot_operations ext4_quota_operations = {
1053 	.get_reserved_space = ext4_get_reserved_space,
1054 	.write_dquot	= ext4_write_dquot,
1055 	.acquire_dquot	= ext4_acquire_dquot,
1056 	.release_dquot	= ext4_release_dquot,
1057 	.mark_dirty	= ext4_mark_dquot_dirty,
1058 	.write_info	= ext4_write_info,
1059 	.alloc_dquot	= dquot_alloc,
1060 	.destroy_dquot	= dquot_destroy,
1061 };
1062 
1063 static const struct quotactl_ops ext4_qctl_operations = {
1064 	.quota_on	= ext4_quota_on,
1065 	.quota_off	= ext4_quota_off,
1066 	.quota_sync	= dquot_quota_sync,
1067 	.get_state	= dquot_get_state,
1068 	.set_info	= dquot_set_dqinfo,
1069 	.get_dqblk	= dquot_get_dqblk,
1070 	.set_dqblk	= dquot_set_dqblk
1071 };
1072 #endif
1073 
1074 static const struct super_operations ext4_sops = {
1075 	.alloc_inode	= ext4_alloc_inode,
1076 	.destroy_inode	= ext4_destroy_inode,
1077 	.write_inode	= ext4_write_inode,
1078 	.dirty_inode	= ext4_dirty_inode,
1079 	.drop_inode	= ext4_drop_inode,
1080 	.evict_inode	= ext4_evict_inode,
1081 	.put_super	= ext4_put_super,
1082 	.sync_fs	= ext4_sync_fs,
1083 	.freeze_fs	= ext4_freeze,
1084 	.unfreeze_fs	= ext4_unfreeze,
1085 	.statfs		= ext4_statfs,
1086 	.remount_fs	= ext4_remount,
1087 	.show_options	= ext4_show_options,
1088 #ifdef CONFIG_QUOTA
1089 	.quota_read	= ext4_quota_read,
1090 	.quota_write	= ext4_quota_write,
1091 	.get_dquots	= ext4_get_dquots,
1092 #endif
1093 	.bdev_try_to_free_page = bdev_try_to_free_page,
1094 };
1095 
1096 static const struct export_operations ext4_export_ops = {
1097 	.fh_to_dentry = ext4_fh_to_dentry,
1098 	.fh_to_parent = ext4_fh_to_parent,
1099 	.get_parent = ext4_get_parent,
1100 };
1101 
1102 enum {
1103 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1104 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1105 	Opt_nouid32, Opt_debug, Opt_removed,
1106 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1107 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1108 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1109 	Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1110 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1111 	Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1112 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1113 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1114 	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1115 	Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1116 	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1117 	Opt_lazytime, Opt_nolazytime,
1118 	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1119 	Opt_inode_readahead_blks, Opt_journal_ioprio,
1120 	Opt_dioread_nolock, Opt_dioread_lock,
1121 	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1122 	Opt_max_dir_size_kb, Opt_nojournal_checksum,
1123 };
1124 
1125 static const match_table_t tokens = {
1126 	{Opt_bsd_df, "bsddf"},
1127 	{Opt_minix_df, "minixdf"},
1128 	{Opt_grpid, "grpid"},
1129 	{Opt_grpid, "bsdgroups"},
1130 	{Opt_nogrpid, "nogrpid"},
1131 	{Opt_nogrpid, "sysvgroups"},
1132 	{Opt_resgid, "resgid=%u"},
1133 	{Opt_resuid, "resuid=%u"},
1134 	{Opt_sb, "sb=%u"},
1135 	{Opt_err_cont, "errors=continue"},
1136 	{Opt_err_panic, "errors=panic"},
1137 	{Opt_err_ro, "errors=remount-ro"},
1138 	{Opt_nouid32, "nouid32"},
1139 	{Opt_debug, "debug"},
1140 	{Opt_removed, "oldalloc"},
1141 	{Opt_removed, "orlov"},
1142 	{Opt_user_xattr, "user_xattr"},
1143 	{Opt_nouser_xattr, "nouser_xattr"},
1144 	{Opt_acl, "acl"},
1145 	{Opt_noacl, "noacl"},
1146 	{Opt_noload, "norecovery"},
1147 	{Opt_noload, "noload"},
1148 	{Opt_removed, "nobh"},
1149 	{Opt_removed, "bh"},
1150 	{Opt_commit, "commit=%u"},
1151 	{Opt_min_batch_time, "min_batch_time=%u"},
1152 	{Opt_max_batch_time, "max_batch_time=%u"},
1153 	{Opt_journal_dev, "journal_dev=%u"},
1154 	{Opt_journal_path, "journal_path=%s"},
1155 	{Opt_journal_checksum, "journal_checksum"},
1156 	{Opt_nojournal_checksum, "nojournal_checksum"},
1157 	{Opt_journal_async_commit, "journal_async_commit"},
1158 	{Opt_abort, "abort"},
1159 	{Opt_data_journal, "data=journal"},
1160 	{Opt_data_ordered, "data=ordered"},
1161 	{Opt_data_writeback, "data=writeback"},
1162 	{Opt_data_err_abort, "data_err=abort"},
1163 	{Opt_data_err_ignore, "data_err=ignore"},
1164 	{Opt_offusrjquota, "usrjquota="},
1165 	{Opt_usrjquota, "usrjquota=%s"},
1166 	{Opt_offgrpjquota, "grpjquota="},
1167 	{Opt_grpjquota, "grpjquota=%s"},
1168 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1169 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1170 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1171 	{Opt_grpquota, "grpquota"},
1172 	{Opt_noquota, "noquota"},
1173 	{Opt_quota, "quota"},
1174 	{Opt_usrquota, "usrquota"},
1175 	{Opt_barrier, "barrier=%u"},
1176 	{Opt_barrier, "barrier"},
1177 	{Opt_nobarrier, "nobarrier"},
1178 	{Opt_i_version, "i_version"},
1179 	{Opt_dax, "dax"},
1180 	{Opt_stripe, "stripe=%u"},
1181 	{Opt_delalloc, "delalloc"},
1182 	{Opt_lazytime, "lazytime"},
1183 	{Opt_nolazytime, "nolazytime"},
1184 	{Opt_nodelalloc, "nodelalloc"},
1185 	{Opt_removed, "mblk_io_submit"},
1186 	{Opt_removed, "nomblk_io_submit"},
1187 	{Opt_block_validity, "block_validity"},
1188 	{Opt_noblock_validity, "noblock_validity"},
1189 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1190 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1191 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1192 	{Opt_auto_da_alloc, "auto_da_alloc"},
1193 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1194 	{Opt_dioread_nolock, "dioread_nolock"},
1195 	{Opt_dioread_lock, "dioread_lock"},
1196 	{Opt_discard, "discard"},
1197 	{Opt_nodiscard, "nodiscard"},
1198 	{Opt_init_itable, "init_itable=%u"},
1199 	{Opt_init_itable, "init_itable"},
1200 	{Opt_noinit_itable, "noinit_itable"},
1201 	{Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1202 	{Opt_test_dummy_encryption, "test_dummy_encryption"},
1203 	{Opt_removed, "check=none"},	/* mount option from ext2/3 */
1204 	{Opt_removed, "nocheck"},	/* mount option from ext2/3 */
1205 	{Opt_removed, "reservation"},	/* mount option from ext2/3 */
1206 	{Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1207 	{Opt_removed, "journal=%u"},	/* mount option from ext2/3 */
1208 	{Opt_err, NULL},
1209 };
1210 
1211 static ext4_fsblk_t get_sb_block(void **data)
1212 {
1213 	ext4_fsblk_t	sb_block;
1214 	char		*options = (char *) *data;
1215 
1216 	if (!options || strncmp(options, "sb=", 3) != 0)
1217 		return 1;	/* Default location */
1218 
1219 	options += 3;
1220 	/* TODO: use simple_strtoll with >32bit ext4 */
1221 	sb_block = simple_strtoul(options, &options, 0);
1222 	if (*options && *options != ',') {
1223 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1224 		       (char *) *data);
1225 		return 1;
1226 	}
1227 	if (*options == ',')
1228 		options++;
1229 	*data = (void *) options;
1230 
1231 	return sb_block;
1232 }
1233 
1234 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1235 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1236 	"Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1237 
1238 #ifdef CONFIG_QUOTA
1239 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1240 {
1241 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1242 	char *qname;
1243 	int ret = -1;
1244 
1245 	if (sb_any_quota_loaded(sb) &&
1246 		!sbi->s_qf_names[qtype]) {
1247 		ext4_msg(sb, KERN_ERR,
1248 			"Cannot change journaled "
1249 			"quota options when quota turned on");
1250 		return -1;
1251 	}
1252 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1253 		ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1254 			 "when QUOTA feature is enabled");
1255 		return -1;
1256 	}
1257 	qname = match_strdup(args);
1258 	if (!qname) {
1259 		ext4_msg(sb, KERN_ERR,
1260 			"Not enough memory for storing quotafile name");
1261 		return -1;
1262 	}
1263 	if (sbi->s_qf_names[qtype]) {
1264 		if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1265 			ret = 1;
1266 		else
1267 			ext4_msg(sb, KERN_ERR,
1268 				 "%s quota file already specified",
1269 				 QTYPE2NAME(qtype));
1270 		goto errout;
1271 	}
1272 	if (strchr(qname, '/')) {
1273 		ext4_msg(sb, KERN_ERR,
1274 			"quotafile must be on filesystem root");
1275 		goto errout;
1276 	}
1277 	sbi->s_qf_names[qtype] = qname;
1278 	set_opt(sb, QUOTA);
1279 	return 1;
1280 errout:
1281 	kfree(qname);
1282 	return ret;
1283 }
1284 
1285 static int clear_qf_name(struct super_block *sb, int qtype)
1286 {
1287 
1288 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1289 
1290 	if (sb_any_quota_loaded(sb) &&
1291 		sbi->s_qf_names[qtype]) {
1292 		ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1293 			" when quota turned on");
1294 		return -1;
1295 	}
1296 	kfree(sbi->s_qf_names[qtype]);
1297 	sbi->s_qf_names[qtype] = NULL;
1298 	return 1;
1299 }
1300 #endif
1301 
1302 #define MOPT_SET	0x0001
1303 #define MOPT_CLEAR	0x0002
1304 #define MOPT_NOSUPPORT	0x0004
1305 #define MOPT_EXPLICIT	0x0008
1306 #define MOPT_CLEAR_ERR	0x0010
1307 #define MOPT_GTE0	0x0020
1308 #ifdef CONFIG_QUOTA
1309 #define MOPT_Q		0
1310 #define MOPT_QFMT	0x0040
1311 #else
1312 #define MOPT_Q		MOPT_NOSUPPORT
1313 #define MOPT_QFMT	MOPT_NOSUPPORT
1314 #endif
1315 #define MOPT_DATAJ	0x0080
1316 #define MOPT_NO_EXT2	0x0100
1317 #define MOPT_NO_EXT3	0x0200
1318 #define MOPT_EXT4_ONLY	(MOPT_NO_EXT2 | MOPT_NO_EXT3)
1319 #define MOPT_STRING	0x0400
1320 
1321 static const struct mount_opts {
1322 	int	token;
1323 	int	mount_opt;
1324 	int	flags;
1325 } ext4_mount_opts[] = {
1326 	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1327 	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1328 	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1329 	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1330 	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1331 	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1332 	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1333 	 MOPT_EXT4_ONLY | MOPT_SET},
1334 	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1335 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1336 	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1337 	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1338 	{Opt_delalloc, EXT4_MOUNT_DELALLOC,
1339 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1340 	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1341 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1342 	{Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1343 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1344 	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1345 	 MOPT_EXT4_ONLY | MOPT_SET},
1346 	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1347 				    EXT4_MOUNT_JOURNAL_CHECKSUM),
1348 	 MOPT_EXT4_ONLY | MOPT_SET},
1349 	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1350 	{Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1351 	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1352 	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1353 	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1354 	 MOPT_NO_EXT2 | MOPT_SET},
1355 	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1356 	 MOPT_NO_EXT2 | MOPT_CLEAR},
1357 	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1358 	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1359 	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1360 	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1361 	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1362 	{Opt_commit, 0, MOPT_GTE0},
1363 	{Opt_max_batch_time, 0, MOPT_GTE0},
1364 	{Opt_min_batch_time, 0, MOPT_GTE0},
1365 	{Opt_inode_readahead_blks, 0, MOPT_GTE0},
1366 	{Opt_init_itable, 0, MOPT_GTE0},
1367 	{Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1368 	{Opt_stripe, 0, MOPT_GTE0},
1369 	{Opt_resuid, 0, MOPT_GTE0},
1370 	{Opt_resgid, 0, MOPT_GTE0},
1371 	{Opt_journal_dev, 0, MOPT_GTE0},
1372 	{Opt_journal_path, 0, MOPT_STRING},
1373 	{Opt_journal_ioprio, 0, MOPT_GTE0},
1374 	{Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1375 	{Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1376 	{Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1377 	 MOPT_NO_EXT2 | MOPT_DATAJ},
1378 	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1379 	{Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1380 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1381 	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1382 	{Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1383 #else
1384 	{Opt_acl, 0, MOPT_NOSUPPORT},
1385 	{Opt_noacl, 0, MOPT_NOSUPPORT},
1386 #endif
1387 	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1388 	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1389 	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1390 	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1391 							MOPT_SET | MOPT_Q},
1392 	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1393 							MOPT_SET | MOPT_Q},
1394 	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1395 		       EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1396 	{Opt_usrjquota, 0, MOPT_Q},
1397 	{Opt_grpjquota, 0, MOPT_Q},
1398 	{Opt_offusrjquota, 0, MOPT_Q},
1399 	{Opt_offgrpjquota, 0, MOPT_Q},
1400 	{Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1401 	{Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1402 	{Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1403 	{Opt_max_dir_size_kb, 0, MOPT_GTE0},
1404 	{Opt_test_dummy_encryption, 0, MOPT_GTE0},
1405 	{Opt_err, 0, 0}
1406 };
1407 
1408 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1409 			    substring_t *args, unsigned long *journal_devnum,
1410 			    unsigned int *journal_ioprio, int is_remount)
1411 {
1412 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1413 	const struct mount_opts *m;
1414 	kuid_t uid;
1415 	kgid_t gid;
1416 	int arg = 0;
1417 
1418 #ifdef CONFIG_QUOTA
1419 	if (token == Opt_usrjquota)
1420 		return set_qf_name(sb, USRQUOTA, &args[0]);
1421 	else if (token == Opt_grpjquota)
1422 		return set_qf_name(sb, GRPQUOTA, &args[0]);
1423 	else if (token == Opt_offusrjquota)
1424 		return clear_qf_name(sb, USRQUOTA);
1425 	else if (token == Opt_offgrpjquota)
1426 		return clear_qf_name(sb, GRPQUOTA);
1427 #endif
1428 	switch (token) {
1429 	case Opt_noacl:
1430 	case Opt_nouser_xattr:
1431 		ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1432 		break;
1433 	case Opt_sb:
1434 		return 1;	/* handled by get_sb_block() */
1435 	case Opt_removed:
1436 		ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1437 		return 1;
1438 	case Opt_abort:
1439 		sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1440 		return 1;
1441 	case Opt_i_version:
1442 		sb->s_flags |= MS_I_VERSION;
1443 		return 1;
1444 	case Opt_lazytime:
1445 		sb->s_flags |= MS_LAZYTIME;
1446 		return 1;
1447 	case Opt_nolazytime:
1448 		sb->s_flags &= ~MS_LAZYTIME;
1449 		return 1;
1450 	}
1451 
1452 	for (m = ext4_mount_opts; m->token != Opt_err; m++)
1453 		if (token == m->token)
1454 			break;
1455 
1456 	if (m->token == Opt_err) {
1457 		ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1458 			 "or missing value", opt);
1459 		return -1;
1460 	}
1461 
1462 	if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1463 		ext4_msg(sb, KERN_ERR,
1464 			 "Mount option \"%s\" incompatible with ext2", opt);
1465 		return -1;
1466 	}
1467 	if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1468 		ext4_msg(sb, KERN_ERR,
1469 			 "Mount option \"%s\" incompatible with ext3", opt);
1470 		return -1;
1471 	}
1472 
1473 	if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1474 		return -1;
1475 	if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1476 		return -1;
1477 	if (m->flags & MOPT_EXPLICIT)
1478 		set_opt2(sb, EXPLICIT_DELALLOC);
1479 	if (m->flags & MOPT_CLEAR_ERR)
1480 		clear_opt(sb, ERRORS_MASK);
1481 	if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1482 		ext4_msg(sb, KERN_ERR, "Cannot change quota "
1483 			 "options when quota turned on");
1484 		return -1;
1485 	}
1486 
1487 	if (m->flags & MOPT_NOSUPPORT) {
1488 		ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1489 	} else if (token == Opt_commit) {
1490 		if (arg == 0)
1491 			arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1492 		sbi->s_commit_interval = HZ * arg;
1493 	} else if (token == Opt_max_batch_time) {
1494 		sbi->s_max_batch_time = arg;
1495 	} else if (token == Opt_min_batch_time) {
1496 		sbi->s_min_batch_time = arg;
1497 	} else if (token == Opt_inode_readahead_blks) {
1498 		if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1499 			ext4_msg(sb, KERN_ERR,
1500 				 "EXT4-fs: inode_readahead_blks must be "
1501 				 "0 or a power of 2 smaller than 2^31");
1502 			return -1;
1503 		}
1504 		sbi->s_inode_readahead_blks = arg;
1505 	} else if (token == Opt_init_itable) {
1506 		set_opt(sb, INIT_INODE_TABLE);
1507 		if (!args->from)
1508 			arg = EXT4_DEF_LI_WAIT_MULT;
1509 		sbi->s_li_wait_mult = arg;
1510 	} else if (token == Opt_max_dir_size_kb) {
1511 		sbi->s_max_dir_size_kb = arg;
1512 	} else if (token == Opt_stripe) {
1513 		sbi->s_stripe = arg;
1514 	} else if (token == Opt_resuid) {
1515 		uid = make_kuid(current_user_ns(), arg);
1516 		if (!uid_valid(uid)) {
1517 			ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1518 			return -1;
1519 		}
1520 		sbi->s_resuid = uid;
1521 	} else if (token == Opt_resgid) {
1522 		gid = make_kgid(current_user_ns(), arg);
1523 		if (!gid_valid(gid)) {
1524 			ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1525 			return -1;
1526 		}
1527 		sbi->s_resgid = gid;
1528 	} else if (token == Opt_journal_dev) {
1529 		if (is_remount) {
1530 			ext4_msg(sb, KERN_ERR,
1531 				 "Cannot specify journal on remount");
1532 			return -1;
1533 		}
1534 		*journal_devnum = arg;
1535 	} else if (token == Opt_journal_path) {
1536 		char *journal_path;
1537 		struct inode *journal_inode;
1538 		struct path path;
1539 		int error;
1540 
1541 		if (is_remount) {
1542 			ext4_msg(sb, KERN_ERR,
1543 				 "Cannot specify journal on remount");
1544 			return -1;
1545 		}
1546 		journal_path = match_strdup(&args[0]);
1547 		if (!journal_path) {
1548 			ext4_msg(sb, KERN_ERR, "error: could not dup "
1549 				"journal device string");
1550 			return -1;
1551 		}
1552 
1553 		error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1554 		if (error) {
1555 			ext4_msg(sb, KERN_ERR, "error: could not find "
1556 				"journal device path: error %d", error);
1557 			kfree(journal_path);
1558 			return -1;
1559 		}
1560 
1561 		journal_inode = d_inode(path.dentry);
1562 		if (!S_ISBLK(journal_inode->i_mode)) {
1563 			ext4_msg(sb, KERN_ERR, "error: journal path %s "
1564 				"is not a block device", journal_path);
1565 			path_put(&path);
1566 			kfree(journal_path);
1567 			return -1;
1568 		}
1569 
1570 		*journal_devnum = new_encode_dev(journal_inode->i_rdev);
1571 		path_put(&path);
1572 		kfree(journal_path);
1573 	} else if (token == Opt_journal_ioprio) {
1574 		if (arg > 7) {
1575 			ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1576 				 " (must be 0-7)");
1577 			return -1;
1578 		}
1579 		*journal_ioprio =
1580 			IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1581 	} else if (token == Opt_test_dummy_encryption) {
1582 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1583 		sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1584 		ext4_msg(sb, KERN_WARNING,
1585 			 "Test dummy encryption mode enabled");
1586 #else
1587 		ext4_msg(sb, KERN_WARNING,
1588 			 "Test dummy encryption mount option ignored");
1589 #endif
1590 	} else if (m->flags & MOPT_DATAJ) {
1591 		if (is_remount) {
1592 			if (!sbi->s_journal)
1593 				ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1594 			else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1595 				ext4_msg(sb, KERN_ERR,
1596 					 "Cannot change data mode on remount");
1597 				return -1;
1598 			}
1599 		} else {
1600 			clear_opt(sb, DATA_FLAGS);
1601 			sbi->s_mount_opt |= m->mount_opt;
1602 		}
1603 #ifdef CONFIG_QUOTA
1604 	} else if (m->flags & MOPT_QFMT) {
1605 		if (sb_any_quota_loaded(sb) &&
1606 		    sbi->s_jquota_fmt != m->mount_opt) {
1607 			ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1608 				 "quota options when quota turned on");
1609 			return -1;
1610 		}
1611 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1612 					       EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1613 			ext4_msg(sb, KERN_ERR,
1614 				 "Cannot set journaled quota options "
1615 				 "when QUOTA feature is enabled");
1616 			return -1;
1617 		}
1618 		sbi->s_jquota_fmt = m->mount_opt;
1619 #endif
1620 #ifndef CONFIG_FS_DAX
1621 	} else if (token == Opt_dax) {
1622 		ext4_msg(sb, KERN_INFO, "dax option not supported");
1623 		return -1;
1624 #endif
1625 	} else {
1626 		if (!args->from)
1627 			arg = 1;
1628 		if (m->flags & MOPT_CLEAR)
1629 			arg = !arg;
1630 		else if (unlikely(!(m->flags & MOPT_SET))) {
1631 			ext4_msg(sb, KERN_WARNING,
1632 				 "buggy handling of option %s", opt);
1633 			WARN_ON(1);
1634 			return -1;
1635 		}
1636 		if (arg != 0)
1637 			sbi->s_mount_opt |= m->mount_opt;
1638 		else
1639 			sbi->s_mount_opt &= ~m->mount_opt;
1640 	}
1641 	return 1;
1642 }
1643 
1644 static int parse_options(char *options, struct super_block *sb,
1645 			 unsigned long *journal_devnum,
1646 			 unsigned int *journal_ioprio,
1647 			 int is_remount)
1648 {
1649 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1650 	char *p;
1651 	substring_t args[MAX_OPT_ARGS];
1652 	int token;
1653 
1654 	if (!options)
1655 		return 1;
1656 
1657 	while ((p = strsep(&options, ",")) != NULL) {
1658 		if (!*p)
1659 			continue;
1660 		/*
1661 		 * Initialize args struct so we know whether arg was
1662 		 * found; some options take optional arguments.
1663 		 */
1664 		args[0].to = args[0].from = NULL;
1665 		token = match_token(p, tokens, args);
1666 		if (handle_mount_opt(sb, p, token, args, journal_devnum,
1667 				     journal_ioprio, is_remount) < 0)
1668 			return 0;
1669 	}
1670 #ifdef CONFIG_QUOTA
1671 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1672 	    (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1673 		ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1674 			 "feature is enabled");
1675 		return 0;
1676 	}
1677 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1678 		if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1679 			clear_opt(sb, USRQUOTA);
1680 
1681 		if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1682 			clear_opt(sb, GRPQUOTA);
1683 
1684 		if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1685 			ext4_msg(sb, KERN_ERR, "old and new quota "
1686 					"format mixing");
1687 			return 0;
1688 		}
1689 
1690 		if (!sbi->s_jquota_fmt) {
1691 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1692 					"not specified");
1693 			return 0;
1694 		}
1695 	}
1696 #endif
1697 	if (test_opt(sb, DIOREAD_NOLOCK)) {
1698 		int blocksize =
1699 			BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1700 
1701 		if (blocksize < PAGE_CACHE_SIZE) {
1702 			ext4_msg(sb, KERN_ERR, "can't mount with "
1703 				 "dioread_nolock if block size != PAGE_SIZE");
1704 			return 0;
1705 		}
1706 	}
1707 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1708 	    test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1709 		ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1710 			 "in data=ordered mode");
1711 		return 0;
1712 	}
1713 	return 1;
1714 }
1715 
1716 static inline void ext4_show_quota_options(struct seq_file *seq,
1717 					   struct super_block *sb)
1718 {
1719 #if defined(CONFIG_QUOTA)
1720 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1721 
1722 	if (sbi->s_jquota_fmt) {
1723 		char *fmtname = "";
1724 
1725 		switch (sbi->s_jquota_fmt) {
1726 		case QFMT_VFS_OLD:
1727 			fmtname = "vfsold";
1728 			break;
1729 		case QFMT_VFS_V0:
1730 			fmtname = "vfsv0";
1731 			break;
1732 		case QFMT_VFS_V1:
1733 			fmtname = "vfsv1";
1734 			break;
1735 		}
1736 		seq_printf(seq, ",jqfmt=%s", fmtname);
1737 	}
1738 
1739 	if (sbi->s_qf_names[USRQUOTA])
1740 		seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1741 
1742 	if (sbi->s_qf_names[GRPQUOTA])
1743 		seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1744 #endif
1745 }
1746 
1747 static const char *token2str(int token)
1748 {
1749 	const struct match_token *t;
1750 
1751 	for (t = tokens; t->token != Opt_err; t++)
1752 		if (t->token == token && !strchr(t->pattern, '='))
1753 			break;
1754 	return t->pattern;
1755 }
1756 
1757 /*
1758  * Show an option if
1759  *  - it's set to a non-default value OR
1760  *  - if the per-sb default is different from the global default
1761  */
1762 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1763 			      int nodefs)
1764 {
1765 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1766 	struct ext4_super_block *es = sbi->s_es;
1767 	int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1768 	const struct mount_opts *m;
1769 	char sep = nodefs ? '\n' : ',';
1770 
1771 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1772 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1773 
1774 	if (sbi->s_sb_block != 1)
1775 		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1776 
1777 	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1778 		int want_set = m->flags & MOPT_SET;
1779 		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1780 		    (m->flags & MOPT_CLEAR_ERR))
1781 			continue;
1782 		if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1783 			continue; /* skip if same as the default */
1784 		if ((want_set &&
1785 		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1786 		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1787 			continue; /* select Opt_noFoo vs Opt_Foo */
1788 		SEQ_OPTS_PRINT("%s", token2str(m->token));
1789 	}
1790 
1791 	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1792 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1793 		SEQ_OPTS_PRINT("resuid=%u",
1794 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
1795 	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1796 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1797 		SEQ_OPTS_PRINT("resgid=%u",
1798 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
1799 	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1800 	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1801 		SEQ_OPTS_PUTS("errors=remount-ro");
1802 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1803 		SEQ_OPTS_PUTS("errors=continue");
1804 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1805 		SEQ_OPTS_PUTS("errors=panic");
1806 	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1807 		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1808 	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1809 		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1810 	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1811 		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1812 	if (sb->s_flags & MS_I_VERSION)
1813 		SEQ_OPTS_PUTS("i_version");
1814 	if (nodefs || sbi->s_stripe)
1815 		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1816 	if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1817 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1818 			SEQ_OPTS_PUTS("data=journal");
1819 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1820 			SEQ_OPTS_PUTS("data=ordered");
1821 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1822 			SEQ_OPTS_PUTS("data=writeback");
1823 	}
1824 	if (nodefs ||
1825 	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1826 		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1827 			       sbi->s_inode_readahead_blks);
1828 
1829 	if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1830 		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1831 		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1832 	if (nodefs || sbi->s_max_dir_size_kb)
1833 		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1834 
1835 	ext4_show_quota_options(seq, sb);
1836 	return 0;
1837 }
1838 
1839 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1840 {
1841 	return _ext4_show_options(seq, root->d_sb, 0);
1842 }
1843 
1844 static int options_seq_show(struct seq_file *seq, void *offset)
1845 {
1846 	struct super_block *sb = seq->private;
1847 	int rc;
1848 
1849 	seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1850 	rc = _ext4_show_options(seq, sb, 1);
1851 	seq_puts(seq, "\n");
1852 	return rc;
1853 }
1854 
1855 static int options_open_fs(struct inode *inode, struct file *file)
1856 {
1857 	return single_open(file, options_seq_show, PDE_DATA(inode));
1858 }
1859 
1860 static const struct file_operations ext4_seq_options_fops = {
1861 	.owner = THIS_MODULE,
1862 	.open = options_open_fs,
1863 	.read = seq_read,
1864 	.llseek = seq_lseek,
1865 	.release = single_release,
1866 };
1867 
1868 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1869 			    int read_only)
1870 {
1871 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1872 	int res = 0;
1873 
1874 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1875 		ext4_msg(sb, KERN_ERR, "revision level too high, "
1876 			 "forcing read-only mode");
1877 		res = MS_RDONLY;
1878 	}
1879 	if (read_only)
1880 		goto done;
1881 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
1882 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1883 			 "running e2fsck is recommended");
1884 	else if (sbi->s_mount_state & EXT4_ERROR_FS)
1885 		ext4_msg(sb, KERN_WARNING,
1886 			 "warning: mounting fs with errors, "
1887 			 "running e2fsck is recommended");
1888 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1889 		 le16_to_cpu(es->s_mnt_count) >=
1890 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1891 		ext4_msg(sb, KERN_WARNING,
1892 			 "warning: maximal mount count reached, "
1893 			 "running e2fsck is recommended");
1894 	else if (le32_to_cpu(es->s_checkinterval) &&
1895 		(le32_to_cpu(es->s_lastcheck) +
1896 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1897 		ext4_msg(sb, KERN_WARNING,
1898 			 "warning: checktime reached, "
1899 			 "running e2fsck is recommended");
1900 	if (!sbi->s_journal)
1901 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1902 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1903 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1904 	le16_add_cpu(&es->s_mnt_count, 1);
1905 	es->s_mtime = cpu_to_le32(get_seconds());
1906 	ext4_update_dynamic_rev(sb);
1907 	if (sbi->s_journal)
1908 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1909 
1910 	ext4_commit_super(sb, 1);
1911 done:
1912 	if (test_opt(sb, DEBUG))
1913 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1914 				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1915 			sb->s_blocksize,
1916 			sbi->s_groups_count,
1917 			EXT4_BLOCKS_PER_GROUP(sb),
1918 			EXT4_INODES_PER_GROUP(sb),
1919 			sbi->s_mount_opt, sbi->s_mount_opt2);
1920 
1921 	cleancache_init_fs(sb);
1922 	return res;
1923 }
1924 
1925 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1926 {
1927 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1928 	struct flex_groups *new_groups;
1929 	int size;
1930 
1931 	if (!sbi->s_log_groups_per_flex)
1932 		return 0;
1933 
1934 	size = ext4_flex_group(sbi, ngroup - 1) + 1;
1935 	if (size <= sbi->s_flex_groups_allocated)
1936 		return 0;
1937 
1938 	size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1939 	new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1940 	if (!new_groups) {
1941 		ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1942 			 size / (int) sizeof(struct flex_groups));
1943 		return -ENOMEM;
1944 	}
1945 
1946 	if (sbi->s_flex_groups) {
1947 		memcpy(new_groups, sbi->s_flex_groups,
1948 		       (sbi->s_flex_groups_allocated *
1949 			sizeof(struct flex_groups)));
1950 		kvfree(sbi->s_flex_groups);
1951 	}
1952 	sbi->s_flex_groups = new_groups;
1953 	sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1954 	return 0;
1955 }
1956 
1957 static int ext4_fill_flex_info(struct super_block *sb)
1958 {
1959 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1960 	struct ext4_group_desc *gdp = NULL;
1961 	ext4_group_t flex_group;
1962 	int i, err;
1963 
1964 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1965 	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1966 		sbi->s_log_groups_per_flex = 0;
1967 		return 1;
1968 	}
1969 
1970 	err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1971 	if (err)
1972 		goto failed;
1973 
1974 	for (i = 0; i < sbi->s_groups_count; i++) {
1975 		gdp = ext4_get_group_desc(sb, i, NULL);
1976 
1977 		flex_group = ext4_flex_group(sbi, i);
1978 		atomic_add(ext4_free_inodes_count(sb, gdp),
1979 			   &sbi->s_flex_groups[flex_group].free_inodes);
1980 		atomic64_add(ext4_free_group_clusters(sb, gdp),
1981 			     &sbi->s_flex_groups[flex_group].free_clusters);
1982 		atomic_add(ext4_used_dirs_count(sb, gdp),
1983 			   &sbi->s_flex_groups[flex_group].used_dirs);
1984 	}
1985 
1986 	return 1;
1987 failed:
1988 	return 0;
1989 }
1990 
1991 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1992 				   struct ext4_group_desc *gdp)
1993 {
1994 	int offset;
1995 	__u16 crc = 0;
1996 	__le32 le_group = cpu_to_le32(block_group);
1997 
1998 	if (ext4_has_metadata_csum(sbi->s_sb)) {
1999 		/* Use new metadata_csum algorithm */
2000 		__le16 save_csum;
2001 		__u32 csum32;
2002 
2003 		save_csum = gdp->bg_checksum;
2004 		gdp->bg_checksum = 0;
2005 		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2006 				     sizeof(le_group));
2007 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2008 				     sbi->s_desc_size);
2009 		gdp->bg_checksum = save_csum;
2010 
2011 		crc = csum32 & 0xFFFF;
2012 		goto out;
2013 	}
2014 
2015 	/* old crc16 code */
2016 	if (!(sbi->s_es->s_feature_ro_compat &
2017 	      cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2018 		return 0;
2019 
2020 	offset = offsetof(struct ext4_group_desc, bg_checksum);
2021 
2022 	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2023 	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2024 	crc = crc16(crc, (__u8 *)gdp, offset);
2025 	offset += sizeof(gdp->bg_checksum); /* skip checksum */
2026 	/* for checksum of struct ext4_group_desc do the rest...*/
2027 	if ((sbi->s_es->s_feature_incompat &
2028 	     cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2029 	    offset < le16_to_cpu(sbi->s_es->s_desc_size))
2030 		crc = crc16(crc, (__u8 *)gdp + offset,
2031 			    le16_to_cpu(sbi->s_es->s_desc_size) -
2032 				offset);
2033 
2034 out:
2035 	return cpu_to_le16(crc);
2036 }
2037 
2038 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2039 				struct ext4_group_desc *gdp)
2040 {
2041 	if (ext4_has_group_desc_csum(sb) &&
2042 	    (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2043 						      block_group, gdp)))
2044 		return 0;
2045 
2046 	return 1;
2047 }
2048 
2049 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2050 			      struct ext4_group_desc *gdp)
2051 {
2052 	if (!ext4_has_group_desc_csum(sb))
2053 		return;
2054 	gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2055 }
2056 
2057 /* Called at mount-time, super-block is locked */
2058 static int ext4_check_descriptors(struct super_block *sb,
2059 				  ext4_group_t *first_not_zeroed)
2060 {
2061 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2062 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2063 	ext4_fsblk_t last_block;
2064 	ext4_fsblk_t block_bitmap;
2065 	ext4_fsblk_t inode_bitmap;
2066 	ext4_fsblk_t inode_table;
2067 	int flexbg_flag = 0;
2068 	ext4_group_t i, grp = sbi->s_groups_count;
2069 
2070 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2071 		flexbg_flag = 1;
2072 
2073 	ext4_debug("Checking group descriptors");
2074 
2075 	for (i = 0; i < sbi->s_groups_count; i++) {
2076 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2077 
2078 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
2079 			last_block = ext4_blocks_count(sbi->s_es) - 1;
2080 		else
2081 			last_block = first_block +
2082 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
2083 
2084 		if ((grp == sbi->s_groups_count) &&
2085 		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2086 			grp = i;
2087 
2088 		block_bitmap = ext4_block_bitmap(sb, gdp);
2089 		if (block_bitmap < first_block || block_bitmap > last_block) {
2090 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2091 			       "Block bitmap for group %u not in group "
2092 			       "(block %llu)!", i, block_bitmap);
2093 			return 0;
2094 		}
2095 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
2096 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
2097 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2098 			       "Inode bitmap for group %u not in group "
2099 			       "(block %llu)!", i, inode_bitmap);
2100 			return 0;
2101 		}
2102 		inode_table = ext4_inode_table(sb, gdp);
2103 		if (inode_table < first_block ||
2104 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2105 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2106 			       "Inode table for group %u not in group "
2107 			       "(block %llu)!", i, inode_table);
2108 			return 0;
2109 		}
2110 		ext4_lock_group(sb, i);
2111 		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2112 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2113 				 "Checksum for group %u failed (%u!=%u)",
2114 				 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2115 				     gdp)), le16_to_cpu(gdp->bg_checksum));
2116 			if (!(sb->s_flags & MS_RDONLY)) {
2117 				ext4_unlock_group(sb, i);
2118 				return 0;
2119 			}
2120 		}
2121 		ext4_unlock_group(sb, i);
2122 		if (!flexbg_flag)
2123 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
2124 	}
2125 	if (NULL != first_not_zeroed)
2126 		*first_not_zeroed = grp;
2127 	return 1;
2128 }
2129 
2130 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2131  * the superblock) which were deleted from all directories, but held open by
2132  * a process at the time of a crash.  We walk the list and try to delete these
2133  * inodes at recovery time (only with a read-write filesystem).
2134  *
2135  * In order to keep the orphan inode chain consistent during traversal (in
2136  * case of crash during recovery), we link each inode into the superblock
2137  * orphan list_head and handle it the same way as an inode deletion during
2138  * normal operation (which journals the operations for us).
2139  *
2140  * We only do an iget() and an iput() on each inode, which is very safe if we
2141  * accidentally point at an in-use or already deleted inode.  The worst that
2142  * can happen in this case is that we get a "bit already cleared" message from
2143  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2144  * e2fsck was run on this filesystem, and it must have already done the orphan
2145  * inode cleanup for us, so we can safely abort without any further action.
2146  */
2147 static void ext4_orphan_cleanup(struct super_block *sb,
2148 				struct ext4_super_block *es)
2149 {
2150 	unsigned int s_flags = sb->s_flags;
2151 	int nr_orphans = 0, nr_truncates = 0;
2152 #ifdef CONFIG_QUOTA
2153 	int i;
2154 #endif
2155 	if (!es->s_last_orphan) {
2156 		jbd_debug(4, "no orphan inodes to clean up\n");
2157 		return;
2158 	}
2159 
2160 	if (bdev_read_only(sb->s_bdev)) {
2161 		ext4_msg(sb, KERN_ERR, "write access "
2162 			"unavailable, skipping orphan cleanup");
2163 		return;
2164 	}
2165 
2166 	/* Check if feature set would not allow a r/w mount */
2167 	if (!ext4_feature_set_ok(sb, 0)) {
2168 		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2169 			 "unknown ROCOMPAT features");
2170 		return;
2171 	}
2172 
2173 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2174 		/* don't clear list on RO mount w/ errors */
2175 		if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2176 			ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2177 				  "clearing orphan list.\n");
2178 			es->s_last_orphan = 0;
2179 		}
2180 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2181 		return;
2182 	}
2183 
2184 	if (s_flags & MS_RDONLY) {
2185 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2186 		sb->s_flags &= ~MS_RDONLY;
2187 	}
2188 #ifdef CONFIG_QUOTA
2189 	/* Needed for iput() to work correctly and not trash data */
2190 	sb->s_flags |= MS_ACTIVE;
2191 	/* Turn on quotas so that they are updated correctly */
2192 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2193 		if (EXT4_SB(sb)->s_qf_names[i]) {
2194 			int ret = ext4_quota_on_mount(sb, i);
2195 			if (ret < 0)
2196 				ext4_msg(sb, KERN_ERR,
2197 					"Cannot turn on journaled "
2198 					"quota: error %d", ret);
2199 		}
2200 	}
2201 #endif
2202 
2203 	while (es->s_last_orphan) {
2204 		struct inode *inode;
2205 
2206 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2207 		if (IS_ERR(inode)) {
2208 			es->s_last_orphan = 0;
2209 			break;
2210 		}
2211 
2212 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2213 		dquot_initialize(inode);
2214 		if (inode->i_nlink) {
2215 			if (test_opt(sb, DEBUG))
2216 				ext4_msg(sb, KERN_DEBUG,
2217 					"%s: truncating inode %lu to %lld bytes",
2218 					__func__, inode->i_ino, inode->i_size);
2219 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2220 				  inode->i_ino, inode->i_size);
2221 			mutex_lock(&inode->i_mutex);
2222 			truncate_inode_pages(inode->i_mapping, inode->i_size);
2223 			ext4_truncate(inode);
2224 			mutex_unlock(&inode->i_mutex);
2225 			nr_truncates++;
2226 		} else {
2227 			if (test_opt(sb, DEBUG))
2228 				ext4_msg(sb, KERN_DEBUG,
2229 					"%s: deleting unreferenced inode %lu",
2230 					__func__, inode->i_ino);
2231 			jbd_debug(2, "deleting unreferenced inode %lu\n",
2232 				  inode->i_ino);
2233 			nr_orphans++;
2234 		}
2235 		iput(inode);  /* The delete magic happens here! */
2236 	}
2237 
2238 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2239 
2240 	if (nr_orphans)
2241 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2242 		       PLURAL(nr_orphans));
2243 	if (nr_truncates)
2244 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2245 		       PLURAL(nr_truncates));
2246 #ifdef CONFIG_QUOTA
2247 	/* Turn quotas off */
2248 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2249 		if (sb_dqopt(sb)->files[i])
2250 			dquot_quota_off(sb, i);
2251 	}
2252 #endif
2253 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2254 }
2255 
2256 /*
2257  * Maximal extent format file size.
2258  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2259  * extent format containers, within a sector_t, and within i_blocks
2260  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2261  * so that won't be a limiting factor.
2262  *
2263  * However there is other limiting factor. We do store extents in the form
2264  * of starting block and length, hence the resulting length of the extent
2265  * covering maximum file size must fit into on-disk format containers as
2266  * well. Given that length is always by 1 unit bigger than max unit (because
2267  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2268  *
2269  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2270  */
2271 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2272 {
2273 	loff_t res;
2274 	loff_t upper_limit = MAX_LFS_FILESIZE;
2275 
2276 	/* small i_blocks in vfs inode? */
2277 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2278 		/*
2279 		 * CONFIG_LBDAF is not enabled implies the inode
2280 		 * i_block represent total blocks in 512 bytes
2281 		 * 32 == size of vfs inode i_blocks * 8
2282 		 */
2283 		upper_limit = (1LL << 32) - 1;
2284 
2285 		/* total blocks in file system block size */
2286 		upper_limit >>= (blkbits - 9);
2287 		upper_limit <<= blkbits;
2288 	}
2289 
2290 	/*
2291 	 * 32-bit extent-start container, ee_block. We lower the maxbytes
2292 	 * by one fs block, so ee_len can cover the extent of maximum file
2293 	 * size
2294 	 */
2295 	res = (1LL << 32) - 1;
2296 	res <<= blkbits;
2297 
2298 	/* Sanity check against vm- & vfs- imposed limits */
2299 	if (res > upper_limit)
2300 		res = upper_limit;
2301 
2302 	return res;
2303 }
2304 
2305 /*
2306  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2307  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2308  * We need to be 1 filesystem block less than the 2^48 sector limit.
2309  */
2310 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2311 {
2312 	loff_t res = EXT4_NDIR_BLOCKS;
2313 	int meta_blocks;
2314 	loff_t upper_limit;
2315 	/* This is calculated to be the largest file size for a dense, block
2316 	 * mapped file such that the file's total number of 512-byte sectors,
2317 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
2318 	 *
2319 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2320 	 * number of 512-byte sectors of the file.
2321 	 */
2322 
2323 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2324 		/*
2325 		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2326 		 * the inode i_block field represents total file blocks in
2327 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2328 		 */
2329 		upper_limit = (1LL << 32) - 1;
2330 
2331 		/* total blocks in file system block size */
2332 		upper_limit >>= (bits - 9);
2333 
2334 	} else {
2335 		/*
2336 		 * We use 48 bit ext4_inode i_blocks
2337 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2338 		 * represent total number of blocks in
2339 		 * file system block size
2340 		 */
2341 		upper_limit = (1LL << 48) - 1;
2342 
2343 	}
2344 
2345 	/* indirect blocks */
2346 	meta_blocks = 1;
2347 	/* double indirect blocks */
2348 	meta_blocks += 1 + (1LL << (bits-2));
2349 	/* tripple indirect blocks */
2350 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2351 
2352 	upper_limit -= meta_blocks;
2353 	upper_limit <<= bits;
2354 
2355 	res += 1LL << (bits-2);
2356 	res += 1LL << (2*(bits-2));
2357 	res += 1LL << (3*(bits-2));
2358 	res <<= bits;
2359 	if (res > upper_limit)
2360 		res = upper_limit;
2361 
2362 	if (res > MAX_LFS_FILESIZE)
2363 		res = MAX_LFS_FILESIZE;
2364 
2365 	return res;
2366 }
2367 
2368 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2369 				   ext4_fsblk_t logical_sb_block, int nr)
2370 {
2371 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2372 	ext4_group_t bg, first_meta_bg;
2373 	int has_super = 0;
2374 
2375 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2376 
2377 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2378 	    nr < first_meta_bg)
2379 		return logical_sb_block + nr + 1;
2380 	bg = sbi->s_desc_per_block * nr;
2381 	if (ext4_bg_has_super(sb, bg))
2382 		has_super = 1;
2383 
2384 	/*
2385 	 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2386 	 * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
2387 	 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2388 	 * compensate.
2389 	 */
2390 	if (sb->s_blocksize == 1024 && nr == 0 &&
2391 	    le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2392 		has_super++;
2393 
2394 	return (has_super + ext4_group_first_block_no(sb, bg));
2395 }
2396 
2397 /**
2398  * ext4_get_stripe_size: Get the stripe size.
2399  * @sbi: In memory super block info
2400  *
2401  * If we have specified it via mount option, then
2402  * use the mount option value. If the value specified at mount time is
2403  * greater than the blocks per group use the super block value.
2404  * If the super block value is greater than blocks per group return 0.
2405  * Allocator needs it be less than blocks per group.
2406  *
2407  */
2408 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2409 {
2410 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2411 	unsigned long stripe_width =
2412 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2413 	int ret;
2414 
2415 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2416 		ret = sbi->s_stripe;
2417 	else if (stripe_width <= sbi->s_blocks_per_group)
2418 		ret = stripe_width;
2419 	else if (stride <= sbi->s_blocks_per_group)
2420 		ret = stride;
2421 	else
2422 		ret = 0;
2423 
2424 	/*
2425 	 * If the stripe width is 1, this makes no sense and
2426 	 * we set it to 0 to turn off stripe handling code.
2427 	 */
2428 	if (ret <= 1)
2429 		ret = 0;
2430 
2431 	return ret;
2432 }
2433 
2434 /* sysfs supprt */
2435 
2436 struct ext4_attr {
2437 	struct attribute attr;
2438 	ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2439 	ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2440 			 const char *, size_t);
2441 	union {
2442 		int offset;
2443 		int deprecated_val;
2444 	} u;
2445 };
2446 
2447 static int parse_strtoull(const char *buf,
2448 		unsigned long long max, unsigned long long *value)
2449 {
2450 	int ret;
2451 
2452 	ret = kstrtoull(skip_spaces(buf), 0, value);
2453 	if (!ret && *value > max)
2454 		ret = -EINVAL;
2455 	return ret;
2456 }
2457 
2458 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2459 					      struct ext4_sb_info *sbi,
2460 					      char *buf)
2461 {
2462 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2463 		(s64) EXT4_C2B(sbi,
2464 			percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2465 }
2466 
2467 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2468 					 struct ext4_sb_info *sbi, char *buf)
2469 {
2470 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2471 
2472 	if (!sb->s_bdev->bd_part)
2473 		return snprintf(buf, PAGE_SIZE, "0\n");
2474 	return snprintf(buf, PAGE_SIZE, "%lu\n",
2475 			(part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2476 			 sbi->s_sectors_written_start) >> 1);
2477 }
2478 
2479 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2480 					  struct ext4_sb_info *sbi, char *buf)
2481 {
2482 	struct super_block *sb = sbi->s_buddy_cache->i_sb;
2483 
2484 	if (!sb->s_bdev->bd_part)
2485 		return snprintf(buf, PAGE_SIZE, "0\n");
2486 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2487 			(unsigned long long)(sbi->s_kbytes_written +
2488 			((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2489 			  EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2490 }
2491 
2492 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2493 					  struct ext4_sb_info *sbi,
2494 					  const char *buf, size_t count)
2495 {
2496 	unsigned long t;
2497 	int ret;
2498 
2499 	ret = kstrtoul(skip_spaces(buf), 0, &t);
2500 	if (ret)
2501 		return ret;
2502 
2503 	if (t && (!is_power_of_2(t) || t > 0x40000000))
2504 		return -EINVAL;
2505 
2506 	sbi->s_inode_readahead_blks = t;
2507 	return count;
2508 }
2509 
2510 static ssize_t sbi_ui_show(struct ext4_attr *a,
2511 			   struct ext4_sb_info *sbi, char *buf)
2512 {
2513 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2514 
2515 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2516 }
2517 
2518 static ssize_t sbi_ui_store(struct ext4_attr *a,
2519 			    struct ext4_sb_info *sbi,
2520 			    const char *buf, size_t count)
2521 {
2522 	unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2523 	unsigned long t;
2524 	int ret;
2525 
2526 	ret = kstrtoul(skip_spaces(buf), 0, &t);
2527 	if (ret)
2528 		return ret;
2529 	*ui = t;
2530 	return count;
2531 }
2532 
2533 static ssize_t es_ui_show(struct ext4_attr *a,
2534 			   struct ext4_sb_info *sbi, char *buf)
2535 {
2536 
2537 	unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2538 			   a->u.offset);
2539 
2540 	return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2541 }
2542 
2543 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2544 				  struct ext4_sb_info *sbi, char *buf)
2545 {
2546 	return snprintf(buf, PAGE_SIZE, "%llu\n",
2547 		(unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2548 }
2549 
2550 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2551 				   struct ext4_sb_info *sbi,
2552 				   const char *buf, size_t count)
2553 {
2554 	unsigned long long val;
2555 	int ret;
2556 
2557 	if (parse_strtoull(buf, -1ULL, &val))
2558 		return -EINVAL;
2559 	ret = ext4_reserve_clusters(sbi, val);
2560 
2561 	return ret ? ret : count;
2562 }
2563 
2564 static ssize_t trigger_test_error(struct ext4_attr *a,
2565 				  struct ext4_sb_info *sbi,
2566 				  const char *buf, size_t count)
2567 {
2568 	int len = count;
2569 
2570 	if (!capable(CAP_SYS_ADMIN))
2571 		return -EPERM;
2572 
2573 	if (len && buf[len-1] == '\n')
2574 		len--;
2575 
2576 	if (len)
2577 		ext4_error(sbi->s_sb, "%.*s", len, buf);
2578 	return count;
2579 }
2580 
2581 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2582 				   struct ext4_sb_info *sbi, char *buf)
2583 {
2584 	return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2585 }
2586 
2587 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2588 static struct ext4_attr ext4_attr_##_name = {			\
2589 	.attr = {.name = __stringify(_name), .mode = _mode },	\
2590 	.show	= _show,					\
2591 	.store	= _store,					\
2592 	.u = {							\
2593 		.offset = offsetof(struct ext4_sb_info, _elname),\
2594 	},							\
2595 }
2596 
2597 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname)		\
2598 static struct ext4_attr ext4_attr_##_name = {				\
2599 	.attr = {.name = __stringify(_name), .mode = _mode },		\
2600 	.show	= _show,						\
2601 	.store	= _store,						\
2602 	.u = {								\
2603 		.offset = offsetof(struct ext4_super_block, _elname),	\
2604 	},								\
2605 }
2606 
2607 #define EXT4_ATTR(name, mode, show, store) \
2608 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2609 
2610 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2611 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2612 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2613 
2614 #define EXT4_RO_ATTR_ES_UI(name, elname)	\
2615 	EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2616 #define EXT4_RW_ATTR_SBI_UI(name, elname)	\
2617 	EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2618 
2619 #define ATTR_LIST(name) &ext4_attr_##name.attr
2620 #define EXT4_DEPRECATED_ATTR(_name, _val)	\
2621 static struct ext4_attr ext4_attr_##_name = {			\
2622 	.attr = {.name = __stringify(_name), .mode = 0444 },	\
2623 	.show	= sbi_deprecated_show,				\
2624 	.u = {							\
2625 		.deprecated_val = _val,				\
2626 	},							\
2627 }
2628 
2629 EXT4_RO_ATTR(delayed_allocation_blocks);
2630 EXT4_RO_ATTR(session_write_kbytes);
2631 EXT4_RO_ATTR(lifetime_write_kbytes);
2632 EXT4_RW_ATTR(reserved_clusters);
2633 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2634 		 inode_readahead_blks_store, s_inode_readahead_blks);
2635 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2636 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2637 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2638 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2639 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2640 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2641 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2642 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2643 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2644 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2645 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2646 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2647 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2648 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2649 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2650 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2651 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2652 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2653 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2654 
2655 static struct attribute *ext4_attrs[] = {
2656 	ATTR_LIST(delayed_allocation_blocks),
2657 	ATTR_LIST(session_write_kbytes),
2658 	ATTR_LIST(lifetime_write_kbytes),
2659 	ATTR_LIST(reserved_clusters),
2660 	ATTR_LIST(inode_readahead_blks),
2661 	ATTR_LIST(inode_goal),
2662 	ATTR_LIST(mb_stats),
2663 	ATTR_LIST(mb_max_to_scan),
2664 	ATTR_LIST(mb_min_to_scan),
2665 	ATTR_LIST(mb_order2_req),
2666 	ATTR_LIST(mb_stream_req),
2667 	ATTR_LIST(mb_group_prealloc),
2668 	ATTR_LIST(max_writeback_mb_bump),
2669 	ATTR_LIST(extent_max_zeroout_kb),
2670 	ATTR_LIST(trigger_fs_error),
2671 	ATTR_LIST(err_ratelimit_interval_ms),
2672 	ATTR_LIST(err_ratelimit_burst),
2673 	ATTR_LIST(warning_ratelimit_interval_ms),
2674 	ATTR_LIST(warning_ratelimit_burst),
2675 	ATTR_LIST(msg_ratelimit_interval_ms),
2676 	ATTR_LIST(msg_ratelimit_burst),
2677 	ATTR_LIST(errors_count),
2678 	ATTR_LIST(first_error_time),
2679 	ATTR_LIST(last_error_time),
2680 	NULL,
2681 };
2682 
2683 /* Features this copy of ext4 supports */
2684 EXT4_INFO_ATTR(lazy_itable_init);
2685 EXT4_INFO_ATTR(batched_discard);
2686 EXT4_INFO_ATTR(meta_bg_resize);
2687 EXT4_INFO_ATTR(encryption);
2688 
2689 static struct attribute *ext4_feat_attrs[] = {
2690 	ATTR_LIST(lazy_itable_init),
2691 	ATTR_LIST(batched_discard),
2692 	ATTR_LIST(meta_bg_resize),
2693 	ATTR_LIST(encryption),
2694 	NULL,
2695 };
2696 
2697 static ssize_t ext4_attr_show(struct kobject *kobj,
2698 			      struct attribute *attr, char *buf)
2699 {
2700 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2701 						s_kobj);
2702 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2703 
2704 	return a->show ? a->show(a, sbi, buf) : 0;
2705 }
2706 
2707 static ssize_t ext4_attr_store(struct kobject *kobj,
2708 			       struct attribute *attr,
2709 			       const char *buf, size_t len)
2710 {
2711 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2712 						s_kobj);
2713 	struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2714 
2715 	return a->store ? a->store(a, sbi, buf, len) : 0;
2716 }
2717 
2718 static void ext4_sb_release(struct kobject *kobj)
2719 {
2720 	struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2721 						s_kobj);
2722 	complete(&sbi->s_kobj_unregister);
2723 }
2724 
2725 static const struct sysfs_ops ext4_attr_ops = {
2726 	.show	= ext4_attr_show,
2727 	.store	= ext4_attr_store,
2728 };
2729 
2730 static struct kobj_type ext4_ktype = {
2731 	.default_attrs	= ext4_attrs,
2732 	.sysfs_ops	= &ext4_attr_ops,
2733 	.release	= ext4_sb_release,
2734 };
2735 
2736 static void ext4_feat_release(struct kobject *kobj)
2737 {
2738 	complete(&ext4_feat->f_kobj_unregister);
2739 }
2740 
2741 static ssize_t ext4_feat_show(struct kobject *kobj,
2742 			      struct attribute *attr, char *buf)
2743 {
2744 	return snprintf(buf, PAGE_SIZE, "supported\n");
2745 }
2746 
2747 /*
2748  * We can not use ext4_attr_show/store because it relies on the kobject
2749  * being embedded in the ext4_sb_info structure which is definitely not
2750  * true in this case.
2751  */
2752 static const struct sysfs_ops ext4_feat_ops = {
2753 	.show	= ext4_feat_show,
2754 	.store	= NULL,
2755 };
2756 
2757 static struct kobj_type ext4_feat_ktype = {
2758 	.default_attrs	= ext4_feat_attrs,
2759 	.sysfs_ops	= &ext4_feat_ops,
2760 	.release	= ext4_feat_release,
2761 };
2762 
2763 /*
2764  * Check whether this filesystem can be mounted based on
2765  * the features present and the RDONLY/RDWR mount requested.
2766  * Returns 1 if this filesystem can be mounted as requested,
2767  * 0 if it cannot be.
2768  */
2769 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2770 {
2771 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2772 		ext4_msg(sb, KERN_ERR,
2773 			"Couldn't mount because of "
2774 			"unsupported optional features (%x)",
2775 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2776 			~EXT4_FEATURE_INCOMPAT_SUPP));
2777 		return 0;
2778 	}
2779 
2780 	if (readonly)
2781 		return 1;
2782 
2783 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2784 		ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2785 		sb->s_flags |= MS_RDONLY;
2786 		return 1;
2787 	}
2788 
2789 	/* Check that feature set is OK for a read-write mount */
2790 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2791 		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2792 			 "unsupported optional features (%x)",
2793 			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2794 				~EXT4_FEATURE_RO_COMPAT_SUPP));
2795 		return 0;
2796 	}
2797 	/*
2798 	 * Large file size enabled file system can only be mounted
2799 	 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2800 	 */
2801 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2802 		if (sizeof(blkcnt_t) < sizeof(u64)) {
2803 			ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2804 				 "cannot be mounted RDWR without "
2805 				 "CONFIG_LBDAF");
2806 			return 0;
2807 		}
2808 	}
2809 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2810 	    !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2811 		ext4_msg(sb, KERN_ERR,
2812 			 "Can't support bigalloc feature without "
2813 			 "extents feature\n");
2814 		return 0;
2815 	}
2816 
2817 #ifndef CONFIG_QUOTA
2818 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2819 	    !readonly) {
2820 		ext4_msg(sb, KERN_ERR,
2821 			 "Filesystem with quota feature cannot be mounted RDWR "
2822 			 "without CONFIG_QUOTA");
2823 		return 0;
2824 	}
2825 #endif  /* CONFIG_QUOTA */
2826 	return 1;
2827 }
2828 
2829 /*
2830  * This function is called once a day if we have errors logged
2831  * on the file system
2832  */
2833 static void print_daily_error_info(unsigned long arg)
2834 {
2835 	struct super_block *sb = (struct super_block *) arg;
2836 	struct ext4_sb_info *sbi;
2837 	struct ext4_super_block *es;
2838 
2839 	sbi = EXT4_SB(sb);
2840 	es = sbi->s_es;
2841 
2842 	if (es->s_error_count)
2843 		/* fsck newer than v1.41.13 is needed to clean this condition. */
2844 		ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2845 			 le32_to_cpu(es->s_error_count));
2846 	if (es->s_first_error_time) {
2847 		printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2848 		       sb->s_id, le32_to_cpu(es->s_first_error_time),
2849 		       (int) sizeof(es->s_first_error_func),
2850 		       es->s_first_error_func,
2851 		       le32_to_cpu(es->s_first_error_line));
2852 		if (es->s_first_error_ino)
2853 			printk(": inode %u",
2854 			       le32_to_cpu(es->s_first_error_ino));
2855 		if (es->s_first_error_block)
2856 			printk(": block %llu", (unsigned long long)
2857 			       le64_to_cpu(es->s_first_error_block));
2858 		printk("\n");
2859 	}
2860 	if (es->s_last_error_time) {
2861 		printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2862 		       sb->s_id, le32_to_cpu(es->s_last_error_time),
2863 		       (int) sizeof(es->s_last_error_func),
2864 		       es->s_last_error_func,
2865 		       le32_to_cpu(es->s_last_error_line));
2866 		if (es->s_last_error_ino)
2867 			printk(": inode %u",
2868 			       le32_to_cpu(es->s_last_error_ino));
2869 		if (es->s_last_error_block)
2870 			printk(": block %llu", (unsigned long long)
2871 			       le64_to_cpu(es->s_last_error_block));
2872 		printk("\n");
2873 	}
2874 	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2875 }
2876 
2877 /* Find next suitable group and run ext4_init_inode_table */
2878 static int ext4_run_li_request(struct ext4_li_request *elr)
2879 {
2880 	struct ext4_group_desc *gdp = NULL;
2881 	ext4_group_t group, ngroups;
2882 	struct super_block *sb;
2883 	unsigned long timeout = 0;
2884 	int ret = 0;
2885 
2886 	sb = elr->lr_super;
2887 	ngroups = EXT4_SB(sb)->s_groups_count;
2888 
2889 	sb_start_write(sb);
2890 	for (group = elr->lr_next_group; group < ngroups; group++) {
2891 		gdp = ext4_get_group_desc(sb, group, NULL);
2892 		if (!gdp) {
2893 			ret = 1;
2894 			break;
2895 		}
2896 
2897 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2898 			break;
2899 	}
2900 
2901 	if (group >= ngroups)
2902 		ret = 1;
2903 
2904 	if (!ret) {
2905 		timeout = jiffies;
2906 		ret = ext4_init_inode_table(sb, group,
2907 					    elr->lr_timeout ? 0 : 1);
2908 		if (elr->lr_timeout == 0) {
2909 			timeout = (jiffies - timeout) *
2910 				  elr->lr_sbi->s_li_wait_mult;
2911 			elr->lr_timeout = timeout;
2912 		}
2913 		elr->lr_next_sched = jiffies + elr->lr_timeout;
2914 		elr->lr_next_group = group + 1;
2915 	}
2916 	sb_end_write(sb);
2917 
2918 	return ret;
2919 }
2920 
2921 /*
2922  * Remove lr_request from the list_request and free the
2923  * request structure. Should be called with li_list_mtx held
2924  */
2925 static void ext4_remove_li_request(struct ext4_li_request *elr)
2926 {
2927 	struct ext4_sb_info *sbi;
2928 
2929 	if (!elr)
2930 		return;
2931 
2932 	sbi = elr->lr_sbi;
2933 
2934 	list_del(&elr->lr_request);
2935 	sbi->s_li_request = NULL;
2936 	kfree(elr);
2937 }
2938 
2939 static void ext4_unregister_li_request(struct super_block *sb)
2940 {
2941 	mutex_lock(&ext4_li_mtx);
2942 	if (!ext4_li_info) {
2943 		mutex_unlock(&ext4_li_mtx);
2944 		return;
2945 	}
2946 
2947 	mutex_lock(&ext4_li_info->li_list_mtx);
2948 	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2949 	mutex_unlock(&ext4_li_info->li_list_mtx);
2950 	mutex_unlock(&ext4_li_mtx);
2951 }
2952 
2953 static struct task_struct *ext4_lazyinit_task;
2954 
2955 /*
2956  * This is the function where ext4lazyinit thread lives. It walks
2957  * through the request list searching for next scheduled filesystem.
2958  * When such a fs is found, run the lazy initialization request
2959  * (ext4_rn_li_request) and keep track of the time spend in this
2960  * function. Based on that time we compute next schedule time of
2961  * the request. When walking through the list is complete, compute
2962  * next waking time and put itself into sleep.
2963  */
2964 static int ext4_lazyinit_thread(void *arg)
2965 {
2966 	struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2967 	struct list_head *pos, *n;
2968 	struct ext4_li_request *elr;
2969 	unsigned long next_wakeup, cur;
2970 
2971 	BUG_ON(NULL == eli);
2972 
2973 cont_thread:
2974 	while (true) {
2975 		next_wakeup = MAX_JIFFY_OFFSET;
2976 
2977 		mutex_lock(&eli->li_list_mtx);
2978 		if (list_empty(&eli->li_request_list)) {
2979 			mutex_unlock(&eli->li_list_mtx);
2980 			goto exit_thread;
2981 		}
2982 
2983 		list_for_each_safe(pos, n, &eli->li_request_list) {
2984 			elr = list_entry(pos, struct ext4_li_request,
2985 					 lr_request);
2986 
2987 			if (time_after_eq(jiffies, elr->lr_next_sched)) {
2988 				if (ext4_run_li_request(elr) != 0) {
2989 					/* error, remove the lazy_init job */
2990 					ext4_remove_li_request(elr);
2991 					continue;
2992 				}
2993 			}
2994 
2995 			if (time_before(elr->lr_next_sched, next_wakeup))
2996 				next_wakeup = elr->lr_next_sched;
2997 		}
2998 		mutex_unlock(&eli->li_list_mtx);
2999 
3000 		try_to_freeze();
3001 
3002 		cur = jiffies;
3003 		if ((time_after_eq(cur, next_wakeup)) ||
3004 		    (MAX_JIFFY_OFFSET == next_wakeup)) {
3005 			cond_resched();
3006 			continue;
3007 		}
3008 
3009 		schedule_timeout_interruptible(next_wakeup - cur);
3010 
3011 		if (kthread_should_stop()) {
3012 			ext4_clear_request_list();
3013 			goto exit_thread;
3014 		}
3015 	}
3016 
3017 exit_thread:
3018 	/*
3019 	 * It looks like the request list is empty, but we need
3020 	 * to check it under the li_list_mtx lock, to prevent any
3021 	 * additions into it, and of course we should lock ext4_li_mtx
3022 	 * to atomically free the list and ext4_li_info, because at
3023 	 * this point another ext4 filesystem could be registering
3024 	 * new one.
3025 	 */
3026 	mutex_lock(&ext4_li_mtx);
3027 	mutex_lock(&eli->li_list_mtx);
3028 	if (!list_empty(&eli->li_request_list)) {
3029 		mutex_unlock(&eli->li_list_mtx);
3030 		mutex_unlock(&ext4_li_mtx);
3031 		goto cont_thread;
3032 	}
3033 	mutex_unlock(&eli->li_list_mtx);
3034 	kfree(ext4_li_info);
3035 	ext4_li_info = NULL;
3036 	mutex_unlock(&ext4_li_mtx);
3037 
3038 	return 0;
3039 }
3040 
3041 static void ext4_clear_request_list(void)
3042 {
3043 	struct list_head *pos, *n;
3044 	struct ext4_li_request *elr;
3045 
3046 	mutex_lock(&ext4_li_info->li_list_mtx);
3047 	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3048 		elr = list_entry(pos, struct ext4_li_request,
3049 				 lr_request);
3050 		ext4_remove_li_request(elr);
3051 	}
3052 	mutex_unlock(&ext4_li_info->li_list_mtx);
3053 }
3054 
3055 static int ext4_run_lazyinit_thread(void)
3056 {
3057 	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3058 					 ext4_li_info, "ext4lazyinit");
3059 	if (IS_ERR(ext4_lazyinit_task)) {
3060 		int err = PTR_ERR(ext4_lazyinit_task);
3061 		ext4_clear_request_list();
3062 		kfree(ext4_li_info);
3063 		ext4_li_info = NULL;
3064 		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3065 				 "initialization thread\n",
3066 				 err);
3067 		return err;
3068 	}
3069 	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3070 	return 0;
3071 }
3072 
3073 /*
3074  * Check whether it make sense to run itable init. thread or not.
3075  * If there is at least one uninitialized inode table, return
3076  * corresponding group number, else the loop goes through all
3077  * groups and return total number of groups.
3078  */
3079 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3080 {
3081 	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3082 	struct ext4_group_desc *gdp = NULL;
3083 
3084 	for (group = 0; group < ngroups; group++) {
3085 		gdp = ext4_get_group_desc(sb, group, NULL);
3086 		if (!gdp)
3087 			continue;
3088 
3089 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3090 			break;
3091 	}
3092 
3093 	return group;
3094 }
3095 
3096 static int ext4_li_info_new(void)
3097 {
3098 	struct ext4_lazy_init *eli = NULL;
3099 
3100 	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3101 	if (!eli)
3102 		return -ENOMEM;
3103 
3104 	INIT_LIST_HEAD(&eli->li_request_list);
3105 	mutex_init(&eli->li_list_mtx);
3106 
3107 	eli->li_state |= EXT4_LAZYINIT_QUIT;
3108 
3109 	ext4_li_info = eli;
3110 
3111 	return 0;
3112 }
3113 
3114 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3115 					    ext4_group_t start)
3116 {
3117 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3118 	struct ext4_li_request *elr;
3119 
3120 	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3121 	if (!elr)
3122 		return NULL;
3123 
3124 	elr->lr_super = sb;
3125 	elr->lr_sbi = sbi;
3126 	elr->lr_next_group = start;
3127 
3128 	/*
3129 	 * Randomize first schedule time of the request to
3130 	 * spread the inode table initialization requests
3131 	 * better.
3132 	 */
3133 	elr->lr_next_sched = jiffies + (prandom_u32() %
3134 				(EXT4_DEF_LI_MAX_START_DELAY * HZ));
3135 	return elr;
3136 }
3137 
3138 int ext4_register_li_request(struct super_block *sb,
3139 			     ext4_group_t first_not_zeroed)
3140 {
3141 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3142 	struct ext4_li_request *elr = NULL;
3143 	ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3144 	int ret = 0;
3145 
3146 	mutex_lock(&ext4_li_mtx);
3147 	if (sbi->s_li_request != NULL) {
3148 		/*
3149 		 * Reset timeout so it can be computed again, because
3150 		 * s_li_wait_mult might have changed.
3151 		 */
3152 		sbi->s_li_request->lr_timeout = 0;
3153 		goto out;
3154 	}
3155 
3156 	if (first_not_zeroed == ngroups ||
3157 	    (sb->s_flags & MS_RDONLY) ||
3158 	    !test_opt(sb, INIT_INODE_TABLE))
3159 		goto out;
3160 
3161 	elr = ext4_li_request_new(sb, first_not_zeroed);
3162 	if (!elr) {
3163 		ret = -ENOMEM;
3164 		goto out;
3165 	}
3166 
3167 	if (NULL == ext4_li_info) {
3168 		ret = ext4_li_info_new();
3169 		if (ret)
3170 			goto out;
3171 	}
3172 
3173 	mutex_lock(&ext4_li_info->li_list_mtx);
3174 	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3175 	mutex_unlock(&ext4_li_info->li_list_mtx);
3176 
3177 	sbi->s_li_request = elr;
3178 	/*
3179 	 * set elr to NULL here since it has been inserted to
3180 	 * the request_list and the removal and free of it is
3181 	 * handled by ext4_clear_request_list from now on.
3182 	 */
3183 	elr = NULL;
3184 
3185 	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3186 		ret = ext4_run_lazyinit_thread();
3187 		if (ret)
3188 			goto out;
3189 	}
3190 out:
3191 	mutex_unlock(&ext4_li_mtx);
3192 	if (ret)
3193 		kfree(elr);
3194 	return ret;
3195 }
3196 
3197 /*
3198  * We do not need to lock anything since this is called on
3199  * module unload.
3200  */
3201 static void ext4_destroy_lazyinit_thread(void)
3202 {
3203 	/*
3204 	 * If thread exited earlier
3205 	 * there's nothing to be done.
3206 	 */
3207 	if (!ext4_li_info || !ext4_lazyinit_task)
3208 		return;
3209 
3210 	kthread_stop(ext4_lazyinit_task);
3211 }
3212 
3213 static int set_journal_csum_feature_set(struct super_block *sb)
3214 {
3215 	int ret = 1;
3216 	int compat, incompat;
3217 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3218 
3219 	if (ext4_has_metadata_csum(sb)) {
3220 		/* journal checksum v3 */
3221 		compat = 0;
3222 		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3223 	} else {
3224 		/* journal checksum v1 */
3225 		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3226 		incompat = 0;
3227 	}
3228 
3229 	jbd2_journal_clear_features(sbi->s_journal,
3230 			JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3231 			JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3232 			JBD2_FEATURE_INCOMPAT_CSUM_V2);
3233 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3234 		ret = jbd2_journal_set_features(sbi->s_journal,
3235 				compat, 0,
3236 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3237 				incompat);
3238 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3239 		ret = jbd2_journal_set_features(sbi->s_journal,
3240 				compat, 0,
3241 				incompat);
3242 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3243 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3244 	} else {
3245 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3246 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3247 	}
3248 
3249 	return ret;
3250 }
3251 
3252 /*
3253  * Note: calculating the overhead so we can be compatible with
3254  * historical BSD practice is quite difficult in the face of
3255  * clusters/bigalloc.  This is because multiple metadata blocks from
3256  * different block group can end up in the same allocation cluster.
3257  * Calculating the exact overhead in the face of clustered allocation
3258  * requires either O(all block bitmaps) in memory or O(number of block
3259  * groups**2) in time.  We will still calculate the superblock for
3260  * older file systems --- and if we come across with a bigalloc file
3261  * system with zero in s_overhead_clusters the estimate will be close to
3262  * correct especially for very large cluster sizes --- but for newer
3263  * file systems, it's better to calculate this figure once at mkfs
3264  * time, and store it in the superblock.  If the superblock value is
3265  * present (even for non-bigalloc file systems), we will use it.
3266  */
3267 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3268 			  char *buf)
3269 {
3270 	struct ext4_sb_info	*sbi = EXT4_SB(sb);
3271 	struct ext4_group_desc	*gdp;
3272 	ext4_fsblk_t		first_block, last_block, b;
3273 	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
3274 	int			s, j, count = 0;
3275 
3276 	if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3277 		return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3278 			sbi->s_itb_per_group + 2);
3279 
3280 	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3281 		(grp * EXT4_BLOCKS_PER_GROUP(sb));
3282 	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3283 	for (i = 0; i < ngroups; i++) {
3284 		gdp = ext4_get_group_desc(sb, i, NULL);
3285 		b = ext4_block_bitmap(sb, gdp);
3286 		if (b >= first_block && b <= last_block) {
3287 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3288 			count++;
3289 		}
3290 		b = ext4_inode_bitmap(sb, gdp);
3291 		if (b >= first_block && b <= last_block) {
3292 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3293 			count++;
3294 		}
3295 		b = ext4_inode_table(sb, gdp);
3296 		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3297 			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3298 				int c = EXT4_B2C(sbi, b - first_block);
3299 				ext4_set_bit(c, buf);
3300 				count++;
3301 			}
3302 		if (i != grp)
3303 			continue;
3304 		s = 0;
3305 		if (ext4_bg_has_super(sb, grp)) {
3306 			ext4_set_bit(s++, buf);
3307 			count++;
3308 		}
3309 		for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3310 			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3311 			count++;
3312 		}
3313 	}
3314 	if (!count)
3315 		return 0;
3316 	return EXT4_CLUSTERS_PER_GROUP(sb) -
3317 		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3318 }
3319 
3320 /*
3321  * Compute the overhead and stash it in sbi->s_overhead
3322  */
3323 int ext4_calculate_overhead(struct super_block *sb)
3324 {
3325 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3326 	struct ext4_super_block *es = sbi->s_es;
3327 	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3328 	ext4_fsblk_t overhead = 0;
3329 	char *buf = (char *) get_zeroed_page(GFP_NOFS);
3330 
3331 	if (!buf)
3332 		return -ENOMEM;
3333 
3334 	/*
3335 	 * Compute the overhead (FS structures).  This is constant
3336 	 * for a given filesystem unless the number of block groups
3337 	 * changes so we cache the previous value until it does.
3338 	 */
3339 
3340 	/*
3341 	 * All of the blocks before first_data_block are overhead
3342 	 */
3343 	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3344 
3345 	/*
3346 	 * Add the overhead found in each block group
3347 	 */
3348 	for (i = 0; i < ngroups; i++) {
3349 		int blks;
3350 
3351 		blks = count_overhead(sb, i, buf);
3352 		overhead += blks;
3353 		if (blks)
3354 			memset(buf, 0, PAGE_SIZE);
3355 		cond_resched();
3356 	}
3357 	/* Add the internal journal blocks as well */
3358 	if (sbi->s_journal && !sbi->journal_bdev)
3359 		overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3360 
3361 	sbi->s_overhead = overhead;
3362 	smp_wmb();
3363 	free_page((unsigned long) buf);
3364 	return 0;
3365 }
3366 
3367 
3368 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3369 {
3370 	ext4_fsblk_t resv_clusters;
3371 
3372 	/*
3373 	 * There's no need to reserve anything when we aren't using extents.
3374 	 * The space estimates are exact, there are no unwritten extents,
3375 	 * hole punching doesn't need new metadata... This is needed especially
3376 	 * to keep ext2/3 backward compatibility.
3377 	 */
3378 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3379 		return 0;
3380 	/*
3381 	 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3382 	 * This should cover the situations where we can not afford to run
3383 	 * out of space like for example punch hole, or converting
3384 	 * unwritten extents in delalloc path. In most cases such
3385 	 * allocation would require 1, or 2 blocks, higher numbers are
3386 	 * very rare.
3387 	 */
3388 	resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3389 			EXT4_SB(sb)->s_cluster_bits;
3390 
3391 	do_div(resv_clusters, 50);
3392 	resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3393 
3394 	return resv_clusters;
3395 }
3396 
3397 
3398 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3399 {
3400 	ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3401 				sbi->s_cluster_bits;
3402 
3403 	if (count >= clusters)
3404 		return -EINVAL;
3405 
3406 	atomic64_set(&sbi->s_resv_clusters, count);
3407 	return 0;
3408 }
3409 
3410 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3411 {
3412 	char *orig_data = kstrdup(data, GFP_KERNEL);
3413 	struct buffer_head *bh;
3414 	struct ext4_super_block *es = NULL;
3415 	struct ext4_sb_info *sbi;
3416 	ext4_fsblk_t block;
3417 	ext4_fsblk_t sb_block = get_sb_block(&data);
3418 	ext4_fsblk_t logical_sb_block;
3419 	unsigned long offset = 0;
3420 	unsigned long journal_devnum = 0;
3421 	unsigned long def_mount_opts;
3422 	struct inode *root;
3423 	char *cp;
3424 	const char *descr;
3425 	int ret = -ENOMEM;
3426 	int blocksize, clustersize;
3427 	unsigned int db_count;
3428 	unsigned int i;
3429 	int needs_recovery, has_huge_files, has_bigalloc;
3430 	__u64 blocks_count;
3431 	int err = 0;
3432 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3433 	ext4_group_t first_not_zeroed;
3434 
3435 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3436 	if (!sbi)
3437 		goto out_free_orig;
3438 
3439 	sbi->s_blockgroup_lock =
3440 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3441 	if (!sbi->s_blockgroup_lock) {
3442 		kfree(sbi);
3443 		goto out_free_orig;
3444 	}
3445 	sb->s_fs_info = sbi;
3446 	sbi->s_sb = sb;
3447 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3448 	sbi->s_sb_block = sb_block;
3449 	if (sb->s_bdev->bd_part)
3450 		sbi->s_sectors_written_start =
3451 			part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3452 #ifdef CONFIG_EXT4_FS_ENCRYPTION
3453 	/* Modes of operations for file and directory encryption. */
3454 	sbi->s_file_encryption_mode = EXT4_ENCRYPTION_MODE_AES_256_XTS;
3455 	sbi->s_dir_encryption_mode = EXT4_ENCRYPTION_MODE_INVALID;
3456 #endif
3457 
3458 	/* Cleanup superblock name */
3459 	for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3460 		*cp = '!';
3461 
3462 	/* -EINVAL is default */
3463 	ret = -EINVAL;
3464 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3465 	if (!blocksize) {
3466 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3467 		goto out_fail;
3468 	}
3469 
3470 	/*
3471 	 * The ext4 superblock will not be buffer aligned for other than 1kB
3472 	 * block sizes.  We need to calculate the offset from buffer start.
3473 	 */
3474 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3475 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3476 		offset = do_div(logical_sb_block, blocksize);
3477 	} else {
3478 		logical_sb_block = sb_block;
3479 	}
3480 
3481 	if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3482 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
3483 		goto out_fail;
3484 	}
3485 	/*
3486 	 * Note: s_es must be initialized as soon as possible because
3487 	 *       some ext4 macro-instructions depend on its value
3488 	 */
3489 	es = (struct ext4_super_block *) (bh->b_data + offset);
3490 	sbi->s_es = es;
3491 	sb->s_magic = le16_to_cpu(es->s_magic);
3492 	if (sb->s_magic != EXT4_SUPER_MAGIC)
3493 		goto cantfind_ext4;
3494 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3495 
3496 	/* Warn if metadata_csum and gdt_csum are both set. */
3497 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3498 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3499 	    EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3500 		ext4_warning(sb, "metadata_csum and uninit_bg are "
3501 			     "redundant flags; please run fsck.");
3502 
3503 	/* Check for a known checksum algorithm */
3504 	if (!ext4_verify_csum_type(sb, es)) {
3505 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3506 			 "unknown checksum algorithm.");
3507 		silent = 1;
3508 		goto cantfind_ext4;
3509 	}
3510 
3511 	/* Load the checksum driver */
3512 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3513 				       EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3514 		sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3515 		if (IS_ERR(sbi->s_chksum_driver)) {
3516 			ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3517 			ret = PTR_ERR(sbi->s_chksum_driver);
3518 			sbi->s_chksum_driver = NULL;
3519 			goto failed_mount;
3520 		}
3521 	}
3522 
3523 	/* Check superblock checksum */
3524 	if (!ext4_superblock_csum_verify(sb, es)) {
3525 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3526 			 "invalid superblock checksum.  Run e2fsck?");
3527 		silent = 1;
3528 		goto cantfind_ext4;
3529 	}
3530 
3531 	/* Precompute checksum seed for all metadata */
3532 	if (ext4_has_metadata_csum(sb))
3533 		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3534 					       sizeof(es->s_uuid));
3535 
3536 	/* Set defaults before we parse the mount options */
3537 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3538 	set_opt(sb, INIT_INODE_TABLE);
3539 	if (def_mount_opts & EXT4_DEFM_DEBUG)
3540 		set_opt(sb, DEBUG);
3541 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3542 		set_opt(sb, GRPID);
3543 	if (def_mount_opts & EXT4_DEFM_UID16)
3544 		set_opt(sb, NO_UID32);
3545 	/* xattr user namespace & acls are now defaulted on */
3546 	set_opt(sb, XATTR_USER);
3547 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3548 	set_opt(sb, POSIX_ACL);
3549 #endif
3550 	/* don't forget to enable journal_csum when metadata_csum is enabled. */
3551 	if (ext4_has_metadata_csum(sb))
3552 		set_opt(sb, JOURNAL_CHECKSUM);
3553 
3554 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3555 		set_opt(sb, JOURNAL_DATA);
3556 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3557 		set_opt(sb, ORDERED_DATA);
3558 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3559 		set_opt(sb, WRITEBACK_DATA);
3560 
3561 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3562 		set_opt(sb, ERRORS_PANIC);
3563 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3564 		set_opt(sb, ERRORS_CONT);
3565 	else
3566 		set_opt(sb, ERRORS_RO);
3567 	/* block_validity enabled by default; disable with noblock_validity */
3568 	set_opt(sb, BLOCK_VALIDITY);
3569 	if (def_mount_opts & EXT4_DEFM_DISCARD)
3570 		set_opt(sb, DISCARD);
3571 
3572 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3573 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3574 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3575 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3576 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3577 
3578 	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3579 		set_opt(sb, BARRIER);
3580 
3581 	/*
3582 	 * enable delayed allocation by default
3583 	 * Use -o nodelalloc to turn it off
3584 	 */
3585 	if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3586 	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3587 		set_opt(sb, DELALLOC);
3588 
3589 	/*
3590 	 * set default s_li_wait_mult for lazyinit, for the case there is
3591 	 * no mount option specified.
3592 	 */
3593 	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3594 
3595 	if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3596 			   &journal_devnum, &journal_ioprio, 0)) {
3597 		ext4_msg(sb, KERN_WARNING,
3598 			 "failed to parse options in superblock: %s",
3599 			 sbi->s_es->s_mount_opts);
3600 	}
3601 	sbi->s_def_mount_opt = sbi->s_mount_opt;
3602 	if (!parse_options((char *) data, sb, &journal_devnum,
3603 			   &journal_ioprio, 0))
3604 		goto failed_mount;
3605 
3606 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3607 		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3608 			    "with data=journal disables delayed "
3609 			    "allocation and O_DIRECT support!\n");
3610 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3611 			ext4_msg(sb, KERN_ERR, "can't mount with "
3612 				 "both data=journal and delalloc");
3613 			goto failed_mount;
3614 		}
3615 		if (test_opt(sb, DIOREAD_NOLOCK)) {
3616 			ext4_msg(sb, KERN_ERR, "can't mount with "
3617 				 "both data=journal and dioread_nolock");
3618 			goto failed_mount;
3619 		}
3620 		if (test_opt(sb, DAX)) {
3621 			ext4_msg(sb, KERN_ERR, "can't mount with "
3622 				 "both data=journal and dax");
3623 			goto failed_mount;
3624 		}
3625 		if (test_opt(sb, DELALLOC))
3626 			clear_opt(sb, DELALLOC);
3627 	}
3628 
3629 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3630 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3631 
3632 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3633 	    (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3634 	     EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3635 	     EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3636 		ext4_msg(sb, KERN_WARNING,
3637 		       "feature flags set on rev 0 fs, "
3638 		       "running e2fsck is recommended");
3639 
3640 	if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3641 		set_opt2(sb, HURD_COMPAT);
3642 		if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3643 					      EXT4_FEATURE_INCOMPAT_64BIT)) {
3644 			ext4_msg(sb, KERN_ERR,
3645 				 "The Hurd can't support 64-bit file systems");
3646 			goto failed_mount;
3647 		}
3648 	}
3649 
3650 	if (IS_EXT2_SB(sb)) {
3651 		if (ext2_feature_set_ok(sb))
3652 			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3653 				 "using the ext4 subsystem");
3654 		else {
3655 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3656 				 "to feature incompatibilities");
3657 			goto failed_mount;
3658 		}
3659 	}
3660 
3661 	if (IS_EXT3_SB(sb)) {
3662 		if (ext3_feature_set_ok(sb))
3663 			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3664 				 "using the ext4 subsystem");
3665 		else {
3666 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3667 				 "to feature incompatibilities");
3668 			goto failed_mount;
3669 		}
3670 	}
3671 
3672 	/*
3673 	 * Check feature flags regardless of the revision level, since we
3674 	 * previously didn't change the revision level when setting the flags,
3675 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
3676 	 */
3677 	if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3678 		goto failed_mount;
3679 
3680 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3681 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3682 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
3683 		ext4_msg(sb, KERN_ERR,
3684 		       "Unsupported filesystem blocksize %d", blocksize);
3685 		goto failed_mount;
3686 	}
3687 
3688 	if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3689 		if (blocksize != PAGE_SIZE) {
3690 			ext4_msg(sb, KERN_ERR,
3691 					"error: unsupported blocksize for dax");
3692 			goto failed_mount;
3693 		}
3694 		if (!sb->s_bdev->bd_disk->fops->direct_access) {
3695 			ext4_msg(sb, KERN_ERR,
3696 					"error: device does not support dax");
3697 			goto failed_mount;
3698 		}
3699 	}
3700 
3701 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3702 	    es->s_encryption_level) {
3703 		ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3704 			 es->s_encryption_level);
3705 		goto failed_mount;
3706 	}
3707 
3708 	if (sb->s_blocksize != blocksize) {
3709 		/* Validate the filesystem blocksize */
3710 		if (!sb_set_blocksize(sb, blocksize)) {
3711 			ext4_msg(sb, KERN_ERR, "bad block size %d",
3712 					blocksize);
3713 			goto failed_mount;
3714 		}
3715 
3716 		brelse(bh);
3717 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3718 		offset = do_div(logical_sb_block, blocksize);
3719 		bh = sb_bread_unmovable(sb, logical_sb_block);
3720 		if (!bh) {
3721 			ext4_msg(sb, KERN_ERR,
3722 			       "Can't read superblock on 2nd try");
3723 			goto failed_mount;
3724 		}
3725 		es = (struct ext4_super_block *)(bh->b_data + offset);
3726 		sbi->s_es = es;
3727 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3728 			ext4_msg(sb, KERN_ERR,
3729 			       "Magic mismatch, very weird!");
3730 			goto failed_mount;
3731 		}
3732 	}
3733 
3734 	has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3735 				EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3736 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3737 						      has_huge_files);
3738 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3739 
3740 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3741 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3742 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3743 	} else {
3744 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3745 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3746 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3747 		    (!is_power_of_2(sbi->s_inode_size)) ||
3748 		    (sbi->s_inode_size > blocksize)) {
3749 			ext4_msg(sb, KERN_ERR,
3750 			       "unsupported inode size: %d",
3751 			       sbi->s_inode_size);
3752 			goto failed_mount;
3753 		}
3754 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3755 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3756 	}
3757 
3758 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3759 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3760 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3761 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3762 		    !is_power_of_2(sbi->s_desc_size)) {
3763 			ext4_msg(sb, KERN_ERR,
3764 			       "unsupported descriptor size %lu",
3765 			       sbi->s_desc_size);
3766 			goto failed_mount;
3767 		}
3768 	} else
3769 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3770 
3771 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3772 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3773 	if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3774 		goto cantfind_ext4;
3775 
3776 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3777 	if (sbi->s_inodes_per_block == 0)
3778 		goto cantfind_ext4;
3779 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
3780 					sbi->s_inodes_per_block;
3781 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3782 	sbi->s_sbh = bh;
3783 	sbi->s_mount_state = le16_to_cpu(es->s_state);
3784 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3785 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3786 
3787 	for (i = 0; i < 4; i++)
3788 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3789 	sbi->s_def_hash_version = es->s_def_hash_version;
3790 	if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3791 		i = le32_to_cpu(es->s_flags);
3792 		if (i & EXT2_FLAGS_UNSIGNED_HASH)
3793 			sbi->s_hash_unsigned = 3;
3794 		else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3795 #ifdef __CHAR_UNSIGNED__
3796 			if (!(sb->s_flags & MS_RDONLY))
3797 				es->s_flags |=
3798 					cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3799 			sbi->s_hash_unsigned = 3;
3800 #else
3801 			if (!(sb->s_flags & MS_RDONLY))
3802 				es->s_flags |=
3803 					cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3804 #endif
3805 		}
3806 	}
3807 
3808 	/* Handle clustersize */
3809 	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3810 	has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3811 				EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3812 	if (has_bigalloc) {
3813 		if (clustersize < blocksize) {
3814 			ext4_msg(sb, KERN_ERR,
3815 				 "cluster size (%d) smaller than "
3816 				 "block size (%d)", clustersize, blocksize);
3817 			goto failed_mount;
3818 		}
3819 		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3820 			le32_to_cpu(es->s_log_block_size);
3821 		sbi->s_clusters_per_group =
3822 			le32_to_cpu(es->s_clusters_per_group);
3823 		if (sbi->s_clusters_per_group > blocksize * 8) {
3824 			ext4_msg(sb, KERN_ERR,
3825 				 "#clusters per group too big: %lu",
3826 				 sbi->s_clusters_per_group);
3827 			goto failed_mount;
3828 		}
3829 		if (sbi->s_blocks_per_group !=
3830 		    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3831 			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3832 				 "clusters per group (%lu) inconsistent",
3833 				 sbi->s_blocks_per_group,
3834 				 sbi->s_clusters_per_group);
3835 			goto failed_mount;
3836 		}
3837 	} else {
3838 		if (clustersize != blocksize) {
3839 			ext4_warning(sb, "fragment/cluster size (%d) != "
3840 				     "block size (%d)", clustersize,
3841 				     blocksize);
3842 			clustersize = blocksize;
3843 		}
3844 		if (sbi->s_blocks_per_group > blocksize * 8) {
3845 			ext4_msg(sb, KERN_ERR,
3846 				 "#blocks per group too big: %lu",
3847 				 sbi->s_blocks_per_group);
3848 			goto failed_mount;
3849 		}
3850 		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3851 		sbi->s_cluster_bits = 0;
3852 	}
3853 	sbi->s_cluster_ratio = clustersize / blocksize;
3854 
3855 	if (sbi->s_inodes_per_group > blocksize * 8) {
3856 		ext4_msg(sb, KERN_ERR,
3857 		       "#inodes per group too big: %lu",
3858 		       sbi->s_inodes_per_group);
3859 		goto failed_mount;
3860 	}
3861 
3862 	/* Do we have standard group size of clustersize * 8 blocks ? */
3863 	if (sbi->s_blocks_per_group == clustersize << 3)
3864 		set_opt2(sb, STD_GROUP_SIZE);
3865 
3866 	/*
3867 	 * Test whether we have more sectors than will fit in sector_t,
3868 	 * and whether the max offset is addressable by the page cache.
3869 	 */
3870 	err = generic_check_addressable(sb->s_blocksize_bits,
3871 					ext4_blocks_count(es));
3872 	if (err) {
3873 		ext4_msg(sb, KERN_ERR, "filesystem"
3874 			 " too large to mount safely on this system");
3875 		if (sizeof(sector_t) < 8)
3876 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3877 		goto failed_mount;
3878 	}
3879 
3880 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3881 		goto cantfind_ext4;
3882 
3883 	/* check blocks count against device size */
3884 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3885 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3886 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3887 		       "exceeds size of device (%llu blocks)",
3888 		       ext4_blocks_count(es), blocks_count);
3889 		goto failed_mount;
3890 	}
3891 
3892 	/*
3893 	 * It makes no sense for the first data block to be beyond the end
3894 	 * of the filesystem.
3895 	 */
3896 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3897 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3898 			 "block %u is beyond end of filesystem (%llu)",
3899 			 le32_to_cpu(es->s_first_data_block),
3900 			 ext4_blocks_count(es));
3901 		goto failed_mount;
3902 	}
3903 	blocks_count = (ext4_blocks_count(es) -
3904 			le32_to_cpu(es->s_first_data_block) +
3905 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
3906 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3907 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3908 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3909 		       "(block count %llu, first data block %u, "
3910 		       "blocks per group %lu)", sbi->s_groups_count,
3911 		       ext4_blocks_count(es),
3912 		       le32_to_cpu(es->s_first_data_block),
3913 		       EXT4_BLOCKS_PER_GROUP(sb));
3914 		goto failed_mount;
3915 	}
3916 	sbi->s_groups_count = blocks_count;
3917 	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3918 			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3919 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3920 		   EXT4_DESC_PER_BLOCK(sb);
3921 	sbi->s_group_desc = ext4_kvmalloc(db_count *
3922 					  sizeof(struct buffer_head *),
3923 					  GFP_KERNEL);
3924 	if (sbi->s_group_desc == NULL) {
3925 		ext4_msg(sb, KERN_ERR, "not enough memory");
3926 		ret = -ENOMEM;
3927 		goto failed_mount;
3928 	}
3929 
3930 	if (ext4_proc_root)
3931 		sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3932 
3933 	if (sbi->s_proc)
3934 		proc_create_data("options", S_IRUGO, sbi->s_proc,
3935 				 &ext4_seq_options_fops, sb);
3936 
3937 	bgl_lock_init(sbi->s_blockgroup_lock);
3938 
3939 	for (i = 0; i < db_count; i++) {
3940 		block = descriptor_loc(sb, logical_sb_block, i);
3941 		sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3942 		if (!sbi->s_group_desc[i]) {
3943 			ext4_msg(sb, KERN_ERR,
3944 			       "can't read group descriptor %d", i);
3945 			db_count = i;
3946 			goto failed_mount2;
3947 		}
3948 	}
3949 	if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3950 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3951 		goto failed_mount2;
3952 	}
3953 
3954 	sbi->s_gdb_count = db_count;
3955 	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3956 	spin_lock_init(&sbi->s_next_gen_lock);
3957 
3958 	setup_timer(&sbi->s_err_report, print_daily_error_info,
3959 		(unsigned long) sb);
3960 
3961 	/* Register extent status tree shrinker */
3962 	if (ext4_es_register_shrinker(sbi))
3963 		goto failed_mount3;
3964 
3965 	sbi->s_stripe = ext4_get_stripe_size(sbi);
3966 	sbi->s_extent_max_zeroout_kb = 32;
3967 
3968 	/*
3969 	 * set up enough so that it can read an inode
3970 	 */
3971 	sb->s_op = &ext4_sops;
3972 	sb->s_export_op = &ext4_export_ops;
3973 	sb->s_xattr = ext4_xattr_handlers;
3974 #ifdef CONFIG_QUOTA
3975 	sb->dq_op = &ext4_quota_operations;
3976 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3977 		sb->s_qcop = &dquot_quotactl_sysfile_ops;
3978 	else
3979 		sb->s_qcop = &ext4_qctl_operations;
3980 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3981 #endif
3982 	memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3983 
3984 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3985 	mutex_init(&sbi->s_orphan_lock);
3986 
3987 	sb->s_root = NULL;
3988 
3989 	needs_recovery = (es->s_last_orphan != 0 ||
3990 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
3991 				    EXT4_FEATURE_INCOMPAT_RECOVER));
3992 
3993 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3994 	    !(sb->s_flags & MS_RDONLY))
3995 		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3996 			goto failed_mount3a;
3997 
3998 	/*
3999 	 * The first inode we look at is the journal inode.  Don't try
4000 	 * root first: it may be modified in the journal!
4001 	 */
4002 	if (!test_opt(sb, NOLOAD) &&
4003 	    EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4004 		if (ext4_load_journal(sb, es, journal_devnum))
4005 			goto failed_mount3a;
4006 	} else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4007 	      EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4008 		ext4_msg(sb, KERN_ERR, "required journal recovery "
4009 		       "suppressed and not mounted read-only");
4010 		goto failed_mount_wq;
4011 	} else {
4012 		clear_opt(sb, DATA_FLAGS);
4013 		sbi->s_journal = NULL;
4014 		needs_recovery = 0;
4015 		goto no_journal;
4016 	}
4017 
4018 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
4019 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4020 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
4021 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4022 		goto failed_mount_wq;
4023 	}
4024 
4025 	if (!set_journal_csum_feature_set(sb)) {
4026 		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4027 			 "feature set");
4028 		goto failed_mount_wq;
4029 	}
4030 
4031 	/* We have now updated the journal if required, so we can
4032 	 * validate the data journaling mode. */
4033 	switch (test_opt(sb, DATA_FLAGS)) {
4034 	case 0:
4035 		/* No mode set, assume a default based on the journal
4036 		 * capabilities: ORDERED_DATA if the journal can
4037 		 * cope, else JOURNAL_DATA
4038 		 */
4039 		if (jbd2_journal_check_available_features
4040 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4041 			set_opt(sb, ORDERED_DATA);
4042 		else
4043 			set_opt(sb, JOURNAL_DATA);
4044 		break;
4045 
4046 	case EXT4_MOUNT_ORDERED_DATA:
4047 	case EXT4_MOUNT_WRITEBACK_DATA:
4048 		if (!jbd2_journal_check_available_features
4049 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4050 			ext4_msg(sb, KERN_ERR, "Journal does not support "
4051 			       "requested data journaling mode");
4052 			goto failed_mount_wq;
4053 		}
4054 	default:
4055 		break;
4056 	}
4057 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4058 
4059 	sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4060 
4061 no_journal:
4062 	if (ext4_mballoc_ready) {
4063 		sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4064 		if (!sbi->s_mb_cache) {
4065 			ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4066 			goto failed_mount_wq;
4067 		}
4068 	}
4069 
4070 	if (unlikely(sbi->s_mount_flags & EXT4_MF_TEST_DUMMY_ENCRYPTION) &&
4071 	    !(sb->s_flags & MS_RDONLY) &&
4072 	    !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
4073 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
4074 		ext4_commit_super(sb, 1);
4075 	}
4076 
4077 	/*
4078 	 * Get the # of file system overhead blocks from the
4079 	 * superblock if present.
4080 	 */
4081 	if (es->s_overhead_clusters)
4082 		sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4083 	else {
4084 		err = ext4_calculate_overhead(sb);
4085 		if (err)
4086 			goto failed_mount_wq;
4087 	}
4088 
4089 	/*
4090 	 * The maximum number of concurrent works can be high and
4091 	 * concurrency isn't really necessary.  Limit it to 1.
4092 	 */
4093 	EXT4_SB(sb)->rsv_conversion_wq =
4094 		alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4095 	if (!EXT4_SB(sb)->rsv_conversion_wq) {
4096 		printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4097 		ret = -ENOMEM;
4098 		goto failed_mount4;
4099 	}
4100 
4101 	/*
4102 	 * The jbd2_journal_load will have done any necessary log recovery,
4103 	 * so we can safely mount the rest of the filesystem now.
4104 	 */
4105 
4106 	root = ext4_iget(sb, EXT4_ROOT_INO);
4107 	if (IS_ERR(root)) {
4108 		ext4_msg(sb, KERN_ERR, "get root inode failed");
4109 		ret = PTR_ERR(root);
4110 		root = NULL;
4111 		goto failed_mount4;
4112 	}
4113 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4114 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4115 		iput(root);
4116 		goto failed_mount4;
4117 	}
4118 	sb->s_root = d_make_root(root);
4119 	if (!sb->s_root) {
4120 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
4121 		ret = -ENOMEM;
4122 		goto failed_mount4;
4123 	}
4124 
4125 	if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4126 		sb->s_flags |= MS_RDONLY;
4127 
4128 	/* determine the minimum size of new large inodes, if present */
4129 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4130 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4131 						     EXT4_GOOD_OLD_INODE_SIZE;
4132 		if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4133 				       EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4134 			if (sbi->s_want_extra_isize <
4135 			    le16_to_cpu(es->s_want_extra_isize))
4136 				sbi->s_want_extra_isize =
4137 					le16_to_cpu(es->s_want_extra_isize);
4138 			if (sbi->s_want_extra_isize <
4139 			    le16_to_cpu(es->s_min_extra_isize))
4140 				sbi->s_want_extra_isize =
4141 					le16_to_cpu(es->s_min_extra_isize);
4142 		}
4143 	}
4144 	/* Check if enough inode space is available */
4145 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4146 							sbi->s_inode_size) {
4147 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4148 						       EXT4_GOOD_OLD_INODE_SIZE;
4149 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
4150 			 "available");
4151 	}
4152 
4153 	err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4154 	if (err) {
4155 		ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4156 			 "reserved pool", ext4_calculate_resv_clusters(sb));
4157 		goto failed_mount4a;
4158 	}
4159 
4160 	err = ext4_setup_system_zone(sb);
4161 	if (err) {
4162 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
4163 			 "zone (%d)", err);
4164 		goto failed_mount4a;
4165 	}
4166 
4167 	ext4_ext_init(sb);
4168 	err = ext4_mb_init(sb);
4169 	if (err) {
4170 		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4171 			 err);
4172 		goto failed_mount5;
4173 	}
4174 
4175 	block = ext4_count_free_clusters(sb);
4176 	ext4_free_blocks_count_set(sbi->s_es,
4177 				   EXT4_C2B(sbi, block));
4178 	err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4179 				  GFP_KERNEL);
4180 	if (!err) {
4181 		unsigned long freei = ext4_count_free_inodes(sb);
4182 		sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4183 		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4184 					  GFP_KERNEL);
4185 	}
4186 	if (!err)
4187 		err = percpu_counter_init(&sbi->s_dirs_counter,
4188 					  ext4_count_dirs(sb), GFP_KERNEL);
4189 	if (!err)
4190 		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4191 					  GFP_KERNEL);
4192 	if (err) {
4193 		ext4_msg(sb, KERN_ERR, "insufficient memory");
4194 		goto failed_mount6;
4195 	}
4196 
4197 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4198 		if (!ext4_fill_flex_info(sb)) {
4199 			ext4_msg(sb, KERN_ERR,
4200 			       "unable to initialize "
4201 			       "flex_bg meta info!");
4202 			goto failed_mount6;
4203 		}
4204 
4205 	err = ext4_register_li_request(sb, first_not_zeroed);
4206 	if (err)
4207 		goto failed_mount6;
4208 
4209 	sbi->s_kobj.kset = ext4_kset;
4210 	init_completion(&sbi->s_kobj_unregister);
4211 	err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4212 				   "%s", sb->s_id);
4213 	if (err)
4214 		goto failed_mount7;
4215 
4216 #ifdef CONFIG_QUOTA
4217 	/* Enable quota usage during mount. */
4218 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4219 	    !(sb->s_flags & MS_RDONLY)) {
4220 		err = ext4_enable_quotas(sb);
4221 		if (err)
4222 			goto failed_mount8;
4223 	}
4224 #endif  /* CONFIG_QUOTA */
4225 
4226 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4227 	ext4_orphan_cleanup(sb, es);
4228 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4229 	if (needs_recovery) {
4230 		ext4_msg(sb, KERN_INFO, "recovery complete");
4231 		ext4_mark_recovery_complete(sb, es);
4232 	}
4233 	if (EXT4_SB(sb)->s_journal) {
4234 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4235 			descr = " journalled data mode";
4236 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4237 			descr = " ordered data mode";
4238 		else
4239 			descr = " writeback data mode";
4240 	} else
4241 		descr = "out journal";
4242 
4243 	if (test_opt(sb, DISCARD)) {
4244 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
4245 		if (!blk_queue_discard(q))
4246 			ext4_msg(sb, KERN_WARNING,
4247 				 "mounting with \"discard\" option, but "
4248 				 "the device does not support discard");
4249 	}
4250 
4251 	ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4252 		 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4253 		 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4254 
4255 	if (es->s_error_count)
4256 		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4257 
4258 	/* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4259 	ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4260 	ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4261 	ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4262 
4263 	kfree(orig_data);
4264 	return 0;
4265 
4266 cantfind_ext4:
4267 	if (!silent)
4268 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4269 	goto failed_mount;
4270 
4271 #ifdef CONFIG_QUOTA
4272 failed_mount8:
4273 	kobject_del(&sbi->s_kobj);
4274 #endif
4275 failed_mount7:
4276 	ext4_unregister_li_request(sb);
4277 failed_mount6:
4278 	ext4_mb_release(sb);
4279 	if (sbi->s_flex_groups)
4280 		kvfree(sbi->s_flex_groups);
4281 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
4282 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
4283 	percpu_counter_destroy(&sbi->s_dirs_counter);
4284 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4285 failed_mount5:
4286 	ext4_ext_release(sb);
4287 	ext4_release_system_zone(sb);
4288 failed_mount4a:
4289 	dput(sb->s_root);
4290 	sb->s_root = NULL;
4291 failed_mount4:
4292 	ext4_msg(sb, KERN_ERR, "mount failed");
4293 	if (EXT4_SB(sb)->rsv_conversion_wq)
4294 		destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4295 failed_mount_wq:
4296 	if (sbi->s_journal) {
4297 		jbd2_journal_destroy(sbi->s_journal);
4298 		sbi->s_journal = NULL;
4299 	}
4300 failed_mount3a:
4301 	ext4_es_unregister_shrinker(sbi);
4302 failed_mount3:
4303 	del_timer_sync(&sbi->s_err_report);
4304 	if (sbi->s_mmp_tsk)
4305 		kthread_stop(sbi->s_mmp_tsk);
4306 failed_mount2:
4307 	for (i = 0; i < db_count; i++)
4308 		brelse(sbi->s_group_desc[i]);
4309 	kvfree(sbi->s_group_desc);
4310 failed_mount:
4311 	if (sbi->s_chksum_driver)
4312 		crypto_free_shash(sbi->s_chksum_driver);
4313 	if (sbi->s_proc) {
4314 		remove_proc_entry("options", sbi->s_proc);
4315 		remove_proc_entry(sb->s_id, ext4_proc_root);
4316 	}
4317 #ifdef CONFIG_QUOTA
4318 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4319 		kfree(sbi->s_qf_names[i]);
4320 #endif
4321 	ext4_blkdev_remove(sbi);
4322 	brelse(bh);
4323 out_fail:
4324 	sb->s_fs_info = NULL;
4325 	kfree(sbi->s_blockgroup_lock);
4326 	kfree(sbi);
4327 out_free_orig:
4328 	kfree(orig_data);
4329 	return err ? err : ret;
4330 }
4331 
4332 /*
4333  * Setup any per-fs journal parameters now.  We'll do this both on
4334  * initial mount, once the journal has been initialised but before we've
4335  * done any recovery; and again on any subsequent remount.
4336  */
4337 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4338 {
4339 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4340 
4341 	journal->j_commit_interval = sbi->s_commit_interval;
4342 	journal->j_min_batch_time = sbi->s_min_batch_time;
4343 	journal->j_max_batch_time = sbi->s_max_batch_time;
4344 
4345 	write_lock(&journal->j_state_lock);
4346 	if (test_opt(sb, BARRIER))
4347 		journal->j_flags |= JBD2_BARRIER;
4348 	else
4349 		journal->j_flags &= ~JBD2_BARRIER;
4350 	if (test_opt(sb, DATA_ERR_ABORT))
4351 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4352 	else
4353 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4354 	write_unlock(&journal->j_state_lock);
4355 }
4356 
4357 static journal_t *ext4_get_journal(struct super_block *sb,
4358 				   unsigned int journal_inum)
4359 {
4360 	struct inode *journal_inode;
4361 	journal_t *journal;
4362 
4363 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4364 
4365 	/* First, test for the existence of a valid inode on disk.  Bad
4366 	 * things happen if we iget() an unused inode, as the subsequent
4367 	 * iput() will try to delete it. */
4368 
4369 	journal_inode = ext4_iget(sb, journal_inum);
4370 	if (IS_ERR(journal_inode)) {
4371 		ext4_msg(sb, KERN_ERR, "no journal found");
4372 		return NULL;
4373 	}
4374 	if (!journal_inode->i_nlink) {
4375 		make_bad_inode(journal_inode);
4376 		iput(journal_inode);
4377 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4378 		return NULL;
4379 	}
4380 
4381 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4382 		  journal_inode, journal_inode->i_size);
4383 	if (!S_ISREG(journal_inode->i_mode)) {
4384 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
4385 		iput(journal_inode);
4386 		return NULL;
4387 	}
4388 
4389 	journal = jbd2_journal_init_inode(journal_inode);
4390 	if (!journal) {
4391 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4392 		iput(journal_inode);
4393 		return NULL;
4394 	}
4395 	journal->j_private = sb;
4396 	ext4_init_journal_params(sb, journal);
4397 	return journal;
4398 }
4399 
4400 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4401 				       dev_t j_dev)
4402 {
4403 	struct buffer_head *bh;
4404 	journal_t *journal;
4405 	ext4_fsblk_t start;
4406 	ext4_fsblk_t len;
4407 	int hblock, blocksize;
4408 	ext4_fsblk_t sb_block;
4409 	unsigned long offset;
4410 	struct ext4_super_block *es;
4411 	struct block_device *bdev;
4412 
4413 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4414 
4415 	bdev = ext4_blkdev_get(j_dev, sb);
4416 	if (bdev == NULL)
4417 		return NULL;
4418 
4419 	blocksize = sb->s_blocksize;
4420 	hblock = bdev_logical_block_size(bdev);
4421 	if (blocksize < hblock) {
4422 		ext4_msg(sb, KERN_ERR,
4423 			"blocksize too small for journal device");
4424 		goto out_bdev;
4425 	}
4426 
4427 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4428 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4429 	set_blocksize(bdev, blocksize);
4430 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
4431 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4432 		       "external journal");
4433 		goto out_bdev;
4434 	}
4435 
4436 	es = (struct ext4_super_block *) (bh->b_data + offset);
4437 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4438 	    !(le32_to_cpu(es->s_feature_incompat) &
4439 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4440 		ext4_msg(sb, KERN_ERR, "external journal has "
4441 					"bad superblock");
4442 		brelse(bh);
4443 		goto out_bdev;
4444 	}
4445 
4446 	if ((le32_to_cpu(es->s_feature_ro_compat) &
4447 	     EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4448 	    es->s_checksum != ext4_superblock_csum(sb, es)) {
4449 		ext4_msg(sb, KERN_ERR, "external journal has "
4450 				       "corrupt superblock");
4451 		brelse(bh);
4452 		goto out_bdev;
4453 	}
4454 
4455 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4456 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4457 		brelse(bh);
4458 		goto out_bdev;
4459 	}
4460 
4461 	len = ext4_blocks_count(es);
4462 	start = sb_block + 1;
4463 	brelse(bh);	/* we're done with the superblock */
4464 
4465 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4466 					start, len, blocksize);
4467 	if (!journal) {
4468 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
4469 		goto out_bdev;
4470 	}
4471 	journal->j_private = sb;
4472 	ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4473 	wait_on_buffer(journal->j_sb_buffer);
4474 	if (!buffer_uptodate(journal->j_sb_buffer)) {
4475 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4476 		goto out_journal;
4477 	}
4478 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4479 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
4480 					"user (unsupported) - %d",
4481 			be32_to_cpu(journal->j_superblock->s_nr_users));
4482 		goto out_journal;
4483 	}
4484 	EXT4_SB(sb)->journal_bdev = bdev;
4485 	ext4_init_journal_params(sb, journal);
4486 	return journal;
4487 
4488 out_journal:
4489 	jbd2_journal_destroy(journal);
4490 out_bdev:
4491 	ext4_blkdev_put(bdev);
4492 	return NULL;
4493 }
4494 
4495 static int ext4_load_journal(struct super_block *sb,
4496 			     struct ext4_super_block *es,
4497 			     unsigned long journal_devnum)
4498 {
4499 	journal_t *journal;
4500 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4501 	dev_t journal_dev;
4502 	int err = 0;
4503 	int really_read_only;
4504 
4505 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4506 
4507 	if (journal_devnum &&
4508 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4509 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4510 			"numbers have changed");
4511 		journal_dev = new_decode_dev(journal_devnum);
4512 	} else
4513 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4514 
4515 	really_read_only = bdev_read_only(sb->s_bdev);
4516 
4517 	/*
4518 	 * Are we loading a blank journal or performing recovery after a
4519 	 * crash?  For recovery, we need to check in advance whether we
4520 	 * can get read-write access to the device.
4521 	 */
4522 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4523 		if (sb->s_flags & MS_RDONLY) {
4524 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
4525 					"required on readonly filesystem");
4526 			if (really_read_only) {
4527 				ext4_msg(sb, KERN_ERR, "write access "
4528 					"unavailable, cannot proceed");
4529 				return -EROFS;
4530 			}
4531 			ext4_msg(sb, KERN_INFO, "write access will "
4532 			       "be enabled during recovery");
4533 		}
4534 	}
4535 
4536 	if (journal_inum && journal_dev) {
4537 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4538 		       "and inode journals!");
4539 		return -EINVAL;
4540 	}
4541 
4542 	if (journal_inum) {
4543 		if (!(journal = ext4_get_journal(sb, journal_inum)))
4544 			return -EINVAL;
4545 	} else {
4546 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4547 			return -EINVAL;
4548 	}
4549 
4550 	if (!(journal->j_flags & JBD2_BARRIER))
4551 		ext4_msg(sb, KERN_INFO, "barriers disabled");
4552 
4553 	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4554 		err = jbd2_journal_wipe(journal, !really_read_only);
4555 	if (!err) {
4556 		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4557 		if (save)
4558 			memcpy(save, ((char *) es) +
4559 			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4560 		err = jbd2_journal_load(journal);
4561 		if (save)
4562 			memcpy(((char *) es) + EXT4_S_ERR_START,
4563 			       save, EXT4_S_ERR_LEN);
4564 		kfree(save);
4565 	}
4566 
4567 	if (err) {
4568 		ext4_msg(sb, KERN_ERR, "error loading journal");
4569 		jbd2_journal_destroy(journal);
4570 		return err;
4571 	}
4572 
4573 	EXT4_SB(sb)->s_journal = journal;
4574 	ext4_clear_journal_err(sb, es);
4575 
4576 	if (!really_read_only && journal_devnum &&
4577 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4578 		es->s_journal_dev = cpu_to_le32(journal_devnum);
4579 
4580 		/* Make sure we flush the recovery flag to disk. */
4581 		ext4_commit_super(sb, 1);
4582 	}
4583 
4584 	return 0;
4585 }
4586 
4587 static int ext4_commit_super(struct super_block *sb, int sync)
4588 {
4589 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4590 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4591 	int error = 0;
4592 
4593 	if (!sbh)
4594 		return error;
4595 	if (buffer_write_io_error(sbh)) {
4596 		/*
4597 		 * Oh, dear.  A previous attempt to write the
4598 		 * superblock failed.  This could happen because the
4599 		 * USB device was yanked out.  Or it could happen to
4600 		 * be a transient write error and maybe the block will
4601 		 * be remapped.  Nothing we can do but to retry the
4602 		 * write and hope for the best.
4603 		 */
4604 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
4605 		       "superblock detected");
4606 		clear_buffer_write_io_error(sbh);
4607 		set_buffer_uptodate(sbh);
4608 	}
4609 	/*
4610 	 * If the file system is mounted read-only, don't update the
4611 	 * superblock write time.  This avoids updating the superblock
4612 	 * write time when we are mounting the root file system
4613 	 * read/only but we need to replay the journal; at that point,
4614 	 * for people who are east of GMT and who make their clock
4615 	 * tick in localtime for Windows bug-for-bug compatibility,
4616 	 * the clock is set in the future, and this will cause e2fsck
4617 	 * to complain and force a full file system check.
4618 	 */
4619 	if (!(sb->s_flags & MS_RDONLY))
4620 		es->s_wtime = cpu_to_le32(get_seconds());
4621 	if (sb->s_bdev->bd_part)
4622 		es->s_kbytes_written =
4623 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4624 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4625 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
4626 	else
4627 		es->s_kbytes_written =
4628 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4629 	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4630 		ext4_free_blocks_count_set(es,
4631 			EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4632 				&EXT4_SB(sb)->s_freeclusters_counter)));
4633 	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4634 		es->s_free_inodes_count =
4635 			cpu_to_le32(percpu_counter_sum_positive(
4636 				&EXT4_SB(sb)->s_freeinodes_counter));
4637 	BUFFER_TRACE(sbh, "marking dirty");
4638 	ext4_superblock_csum_set(sb);
4639 	mark_buffer_dirty(sbh);
4640 	if (sync) {
4641 		error = sync_dirty_buffer(sbh);
4642 		if (error)
4643 			return error;
4644 
4645 		error = buffer_write_io_error(sbh);
4646 		if (error) {
4647 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
4648 			       "superblock");
4649 			clear_buffer_write_io_error(sbh);
4650 			set_buffer_uptodate(sbh);
4651 		}
4652 	}
4653 	return error;
4654 }
4655 
4656 /*
4657  * Have we just finished recovery?  If so, and if we are mounting (or
4658  * remounting) the filesystem readonly, then we will end up with a
4659  * consistent fs on disk.  Record that fact.
4660  */
4661 static void ext4_mark_recovery_complete(struct super_block *sb,
4662 					struct ext4_super_block *es)
4663 {
4664 	journal_t *journal = EXT4_SB(sb)->s_journal;
4665 
4666 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4667 		BUG_ON(journal != NULL);
4668 		return;
4669 	}
4670 	jbd2_journal_lock_updates(journal);
4671 	if (jbd2_journal_flush(journal) < 0)
4672 		goto out;
4673 
4674 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4675 	    sb->s_flags & MS_RDONLY) {
4676 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4677 		ext4_commit_super(sb, 1);
4678 	}
4679 
4680 out:
4681 	jbd2_journal_unlock_updates(journal);
4682 }
4683 
4684 /*
4685  * If we are mounting (or read-write remounting) a filesystem whose journal
4686  * has recorded an error from a previous lifetime, move that error to the
4687  * main filesystem now.
4688  */
4689 static void ext4_clear_journal_err(struct super_block *sb,
4690 				   struct ext4_super_block *es)
4691 {
4692 	journal_t *journal;
4693 	int j_errno;
4694 	const char *errstr;
4695 
4696 	BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4697 
4698 	journal = EXT4_SB(sb)->s_journal;
4699 
4700 	/*
4701 	 * Now check for any error status which may have been recorded in the
4702 	 * journal by a prior ext4_error() or ext4_abort()
4703 	 */
4704 
4705 	j_errno = jbd2_journal_errno(journal);
4706 	if (j_errno) {
4707 		char nbuf[16];
4708 
4709 		errstr = ext4_decode_error(sb, j_errno, nbuf);
4710 		ext4_warning(sb, "Filesystem error recorded "
4711 			     "from previous mount: %s", errstr);
4712 		ext4_warning(sb, "Marking fs in need of filesystem check.");
4713 
4714 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4715 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4716 		ext4_commit_super(sb, 1);
4717 
4718 		jbd2_journal_clear_err(journal);
4719 		jbd2_journal_update_sb_errno(journal);
4720 	}
4721 }
4722 
4723 /*
4724  * Force the running and committing transactions to commit,
4725  * and wait on the commit.
4726  */
4727 int ext4_force_commit(struct super_block *sb)
4728 {
4729 	journal_t *journal;
4730 
4731 	if (sb->s_flags & MS_RDONLY)
4732 		return 0;
4733 
4734 	journal = EXT4_SB(sb)->s_journal;
4735 	return ext4_journal_force_commit(journal);
4736 }
4737 
4738 static int ext4_sync_fs(struct super_block *sb, int wait)
4739 {
4740 	int ret = 0;
4741 	tid_t target;
4742 	bool needs_barrier = false;
4743 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4744 
4745 	trace_ext4_sync_fs(sb, wait);
4746 	flush_workqueue(sbi->rsv_conversion_wq);
4747 	/*
4748 	 * Writeback quota in non-journalled quota case - journalled quota has
4749 	 * no dirty dquots
4750 	 */
4751 	dquot_writeback_dquots(sb, -1);
4752 	/*
4753 	 * Data writeback is possible w/o journal transaction, so barrier must
4754 	 * being sent at the end of the function. But we can skip it if
4755 	 * transaction_commit will do it for us.
4756 	 */
4757 	if (sbi->s_journal) {
4758 		target = jbd2_get_latest_transaction(sbi->s_journal);
4759 		if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4760 		    !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4761 			needs_barrier = true;
4762 
4763 		if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4764 			if (wait)
4765 				ret = jbd2_log_wait_commit(sbi->s_journal,
4766 							   target);
4767 		}
4768 	} else if (wait && test_opt(sb, BARRIER))
4769 		needs_barrier = true;
4770 	if (needs_barrier) {
4771 		int err;
4772 		err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4773 		if (!ret)
4774 			ret = err;
4775 	}
4776 
4777 	return ret;
4778 }
4779 
4780 /*
4781  * LVM calls this function before a (read-only) snapshot is created.  This
4782  * gives us a chance to flush the journal completely and mark the fs clean.
4783  *
4784  * Note that only this function cannot bring a filesystem to be in a clean
4785  * state independently. It relies on upper layer to stop all data & metadata
4786  * modifications.
4787  */
4788 static int ext4_freeze(struct super_block *sb)
4789 {
4790 	int error = 0;
4791 	journal_t *journal;
4792 
4793 	if (sb->s_flags & MS_RDONLY)
4794 		return 0;
4795 
4796 	journal = EXT4_SB(sb)->s_journal;
4797 
4798 	if (journal) {
4799 		/* Now we set up the journal barrier. */
4800 		jbd2_journal_lock_updates(journal);
4801 
4802 		/*
4803 		 * Don't clear the needs_recovery flag if we failed to
4804 		 * flush the journal.
4805 		 */
4806 		error = jbd2_journal_flush(journal);
4807 		if (error < 0)
4808 			goto out;
4809 	}
4810 
4811 	/* Journal blocked and flushed, clear needs_recovery flag. */
4812 	EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4813 	error = ext4_commit_super(sb, 1);
4814 out:
4815 	if (journal)
4816 		/* we rely on upper layer to stop further updates */
4817 		jbd2_journal_unlock_updates(journal);
4818 	return error;
4819 }
4820 
4821 /*
4822  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4823  * flag here, even though the filesystem is not technically dirty yet.
4824  */
4825 static int ext4_unfreeze(struct super_block *sb)
4826 {
4827 	if (sb->s_flags & MS_RDONLY)
4828 		return 0;
4829 
4830 	/* Reset the needs_recovery flag before the fs is unlocked. */
4831 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4832 	ext4_commit_super(sb, 1);
4833 	return 0;
4834 }
4835 
4836 /*
4837  * Structure to save mount options for ext4_remount's benefit
4838  */
4839 struct ext4_mount_options {
4840 	unsigned long s_mount_opt;
4841 	unsigned long s_mount_opt2;
4842 	kuid_t s_resuid;
4843 	kgid_t s_resgid;
4844 	unsigned long s_commit_interval;
4845 	u32 s_min_batch_time, s_max_batch_time;
4846 #ifdef CONFIG_QUOTA
4847 	int s_jquota_fmt;
4848 	char *s_qf_names[EXT4_MAXQUOTAS];
4849 #endif
4850 };
4851 
4852 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4853 {
4854 	struct ext4_super_block *es;
4855 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4856 	unsigned long old_sb_flags;
4857 	struct ext4_mount_options old_opts;
4858 	int enable_quota = 0;
4859 	ext4_group_t g;
4860 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4861 	int err = 0;
4862 #ifdef CONFIG_QUOTA
4863 	int i, j;
4864 #endif
4865 	char *orig_data = kstrdup(data, GFP_KERNEL);
4866 
4867 	/* Store the original options */
4868 	old_sb_flags = sb->s_flags;
4869 	old_opts.s_mount_opt = sbi->s_mount_opt;
4870 	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4871 	old_opts.s_resuid = sbi->s_resuid;
4872 	old_opts.s_resgid = sbi->s_resgid;
4873 	old_opts.s_commit_interval = sbi->s_commit_interval;
4874 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
4875 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
4876 #ifdef CONFIG_QUOTA
4877 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4878 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4879 		if (sbi->s_qf_names[i]) {
4880 			old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4881 							 GFP_KERNEL);
4882 			if (!old_opts.s_qf_names[i]) {
4883 				for (j = 0; j < i; j++)
4884 					kfree(old_opts.s_qf_names[j]);
4885 				kfree(orig_data);
4886 				return -ENOMEM;
4887 			}
4888 		} else
4889 			old_opts.s_qf_names[i] = NULL;
4890 #endif
4891 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4892 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4893 
4894 	if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4895 		err = -EINVAL;
4896 		goto restore_opts;
4897 	}
4898 
4899 	if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4900 	    test_opt(sb, JOURNAL_CHECKSUM)) {
4901 		ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4902 			 "during remount not supported; ignoring");
4903 		sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4904 	}
4905 
4906 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4907 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4908 			ext4_msg(sb, KERN_ERR, "can't mount with "
4909 				 "both data=journal and delalloc");
4910 			err = -EINVAL;
4911 			goto restore_opts;
4912 		}
4913 		if (test_opt(sb, DIOREAD_NOLOCK)) {
4914 			ext4_msg(sb, KERN_ERR, "can't mount with "
4915 				 "both data=journal and dioread_nolock");
4916 			err = -EINVAL;
4917 			goto restore_opts;
4918 		}
4919 		if (test_opt(sb, DAX)) {
4920 			ext4_msg(sb, KERN_ERR, "can't mount with "
4921 				 "both data=journal and dax");
4922 			err = -EINVAL;
4923 			goto restore_opts;
4924 		}
4925 	}
4926 
4927 	if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4928 		ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4929 			"dax flag with busy inodes while remounting");
4930 		sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4931 	}
4932 
4933 	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4934 		ext4_abort(sb, "Abort forced by user");
4935 
4936 	sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4937 		(test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4938 
4939 	es = sbi->s_es;
4940 
4941 	if (sbi->s_journal) {
4942 		ext4_init_journal_params(sb, sbi->s_journal);
4943 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4944 	}
4945 
4946 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4947 		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4948 			err = -EROFS;
4949 			goto restore_opts;
4950 		}
4951 
4952 		if (*flags & MS_RDONLY) {
4953 			err = sync_filesystem(sb);
4954 			if (err < 0)
4955 				goto restore_opts;
4956 			err = dquot_suspend(sb, -1);
4957 			if (err < 0)
4958 				goto restore_opts;
4959 
4960 			/*
4961 			 * First of all, the unconditional stuff we have to do
4962 			 * to disable replay of the journal when we next remount
4963 			 */
4964 			sb->s_flags |= MS_RDONLY;
4965 
4966 			/*
4967 			 * OK, test if we are remounting a valid rw partition
4968 			 * readonly, and if so set the rdonly flag and then
4969 			 * mark the partition as valid again.
4970 			 */
4971 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4972 			    (sbi->s_mount_state & EXT4_VALID_FS))
4973 				es->s_state = cpu_to_le16(sbi->s_mount_state);
4974 
4975 			if (sbi->s_journal)
4976 				ext4_mark_recovery_complete(sb, es);
4977 		} else {
4978 			/* Make sure we can mount this feature set readwrite */
4979 			if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4980 					EXT4_FEATURE_RO_COMPAT_READONLY) ||
4981 			    !ext4_feature_set_ok(sb, 0)) {
4982 				err = -EROFS;
4983 				goto restore_opts;
4984 			}
4985 			/*
4986 			 * Make sure the group descriptor checksums
4987 			 * are sane.  If they aren't, refuse to remount r/w.
4988 			 */
4989 			for (g = 0; g < sbi->s_groups_count; g++) {
4990 				struct ext4_group_desc *gdp =
4991 					ext4_get_group_desc(sb, g, NULL);
4992 
4993 				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4994 					ext4_msg(sb, KERN_ERR,
4995 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
4996 		g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4997 					       le16_to_cpu(gdp->bg_checksum));
4998 					err = -EINVAL;
4999 					goto restore_opts;
5000 				}
5001 			}
5002 
5003 			/*
5004 			 * If we have an unprocessed orphan list hanging
5005 			 * around from a previously readonly bdev mount,
5006 			 * require a full umount/remount for now.
5007 			 */
5008 			if (es->s_last_orphan) {
5009 				ext4_msg(sb, KERN_WARNING, "Couldn't "
5010 				       "remount RDWR because of unprocessed "
5011 				       "orphan inode list.  Please "
5012 				       "umount/remount instead");
5013 				err = -EINVAL;
5014 				goto restore_opts;
5015 			}
5016 
5017 			/*
5018 			 * Mounting a RDONLY partition read-write, so reread
5019 			 * and store the current valid flag.  (It may have
5020 			 * been changed by e2fsck since we originally mounted
5021 			 * the partition.)
5022 			 */
5023 			if (sbi->s_journal)
5024 				ext4_clear_journal_err(sb, es);
5025 			sbi->s_mount_state = le16_to_cpu(es->s_state);
5026 			if (!ext4_setup_super(sb, es, 0))
5027 				sb->s_flags &= ~MS_RDONLY;
5028 			if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5029 						     EXT4_FEATURE_INCOMPAT_MMP))
5030 				if (ext4_multi_mount_protect(sb,
5031 						le64_to_cpu(es->s_mmp_block))) {
5032 					err = -EROFS;
5033 					goto restore_opts;
5034 				}
5035 			enable_quota = 1;
5036 		}
5037 	}
5038 
5039 	/*
5040 	 * Reinitialize lazy itable initialization thread based on
5041 	 * current settings
5042 	 */
5043 	if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5044 		ext4_unregister_li_request(sb);
5045 	else {
5046 		ext4_group_t first_not_zeroed;
5047 		first_not_zeroed = ext4_has_uninit_itable(sb);
5048 		ext4_register_li_request(sb, first_not_zeroed);
5049 	}
5050 
5051 	ext4_setup_system_zone(sb);
5052 	if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5053 		ext4_commit_super(sb, 1);
5054 
5055 #ifdef CONFIG_QUOTA
5056 	/* Release old quota file names */
5057 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
5058 		kfree(old_opts.s_qf_names[i]);
5059 	if (enable_quota) {
5060 		if (sb_any_quota_suspended(sb))
5061 			dquot_resume(sb, -1);
5062 		else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5063 					EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5064 			err = ext4_enable_quotas(sb);
5065 			if (err)
5066 				goto restore_opts;
5067 		}
5068 	}
5069 #endif
5070 
5071 	*flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5072 	ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5073 	kfree(orig_data);
5074 	return 0;
5075 
5076 restore_opts:
5077 	sb->s_flags = old_sb_flags;
5078 	sbi->s_mount_opt = old_opts.s_mount_opt;
5079 	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5080 	sbi->s_resuid = old_opts.s_resuid;
5081 	sbi->s_resgid = old_opts.s_resgid;
5082 	sbi->s_commit_interval = old_opts.s_commit_interval;
5083 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
5084 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
5085 #ifdef CONFIG_QUOTA
5086 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5087 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5088 		kfree(sbi->s_qf_names[i]);
5089 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5090 	}
5091 #endif
5092 	kfree(orig_data);
5093 	return err;
5094 }
5095 
5096 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5097 {
5098 	struct super_block *sb = dentry->d_sb;
5099 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5100 	struct ext4_super_block *es = sbi->s_es;
5101 	ext4_fsblk_t overhead = 0, resv_blocks;
5102 	u64 fsid;
5103 	s64 bfree;
5104 	resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5105 
5106 	if (!test_opt(sb, MINIX_DF))
5107 		overhead = sbi->s_overhead;
5108 
5109 	buf->f_type = EXT4_SUPER_MAGIC;
5110 	buf->f_bsize = sb->s_blocksize;
5111 	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5112 	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5113 		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5114 	/* prevent underflow in case that few free space is available */
5115 	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5116 	buf->f_bavail = buf->f_bfree -
5117 			(ext4_r_blocks_count(es) + resv_blocks);
5118 	if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5119 		buf->f_bavail = 0;
5120 	buf->f_files = le32_to_cpu(es->s_inodes_count);
5121 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5122 	buf->f_namelen = EXT4_NAME_LEN;
5123 	fsid = le64_to_cpup((void *)es->s_uuid) ^
5124 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5125 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5126 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5127 
5128 	return 0;
5129 }
5130 
5131 /* Helper function for writing quotas on sync - we need to start transaction
5132  * before quota file is locked for write. Otherwise the are possible deadlocks:
5133  * Process 1                         Process 2
5134  * ext4_create()                     quota_sync()
5135  *   jbd2_journal_start()                  write_dquot()
5136  *   dquot_initialize()                         down(dqio_mutex)
5137  *     down(dqio_mutex)                    jbd2_journal_start()
5138  *
5139  */
5140 
5141 #ifdef CONFIG_QUOTA
5142 
5143 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5144 {
5145 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5146 }
5147 
5148 static int ext4_write_dquot(struct dquot *dquot)
5149 {
5150 	int ret, err;
5151 	handle_t *handle;
5152 	struct inode *inode;
5153 
5154 	inode = dquot_to_inode(dquot);
5155 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5156 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5157 	if (IS_ERR(handle))
5158 		return PTR_ERR(handle);
5159 	ret = dquot_commit(dquot);
5160 	err = ext4_journal_stop(handle);
5161 	if (!ret)
5162 		ret = err;
5163 	return ret;
5164 }
5165 
5166 static int ext4_acquire_dquot(struct dquot *dquot)
5167 {
5168 	int ret, err;
5169 	handle_t *handle;
5170 
5171 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5172 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5173 	if (IS_ERR(handle))
5174 		return PTR_ERR(handle);
5175 	ret = dquot_acquire(dquot);
5176 	err = ext4_journal_stop(handle);
5177 	if (!ret)
5178 		ret = err;
5179 	return ret;
5180 }
5181 
5182 static int ext4_release_dquot(struct dquot *dquot)
5183 {
5184 	int ret, err;
5185 	handle_t *handle;
5186 
5187 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5188 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5189 	if (IS_ERR(handle)) {
5190 		/* Release dquot anyway to avoid endless cycle in dqput() */
5191 		dquot_release(dquot);
5192 		return PTR_ERR(handle);
5193 	}
5194 	ret = dquot_release(dquot);
5195 	err = ext4_journal_stop(handle);
5196 	if (!ret)
5197 		ret = err;
5198 	return ret;
5199 }
5200 
5201 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5202 {
5203 	struct super_block *sb = dquot->dq_sb;
5204 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5205 
5206 	/* Are we journaling quotas? */
5207 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5208 	    sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5209 		dquot_mark_dquot_dirty(dquot);
5210 		return ext4_write_dquot(dquot);
5211 	} else {
5212 		return dquot_mark_dquot_dirty(dquot);
5213 	}
5214 }
5215 
5216 static int ext4_write_info(struct super_block *sb, int type)
5217 {
5218 	int ret, err;
5219 	handle_t *handle;
5220 
5221 	/* Data block + inode block */
5222 	handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5223 	if (IS_ERR(handle))
5224 		return PTR_ERR(handle);
5225 	ret = dquot_commit_info(sb, type);
5226 	err = ext4_journal_stop(handle);
5227 	if (!ret)
5228 		ret = err;
5229 	return ret;
5230 }
5231 
5232 /*
5233  * Turn on quotas during mount time - we need to find
5234  * the quota file and such...
5235  */
5236 static int ext4_quota_on_mount(struct super_block *sb, int type)
5237 {
5238 	return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5239 					EXT4_SB(sb)->s_jquota_fmt, type);
5240 }
5241 
5242 /*
5243  * Standard function to be called on quota_on
5244  */
5245 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5246 			 struct path *path)
5247 {
5248 	int err;
5249 
5250 	if (!test_opt(sb, QUOTA))
5251 		return -EINVAL;
5252 
5253 	/* Quotafile not on the same filesystem? */
5254 	if (path->dentry->d_sb != sb)
5255 		return -EXDEV;
5256 	/* Journaling quota? */
5257 	if (EXT4_SB(sb)->s_qf_names[type]) {
5258 		/* Quotafile not in fs root? */
5259 		if (path->dentry->d_parent != sb->s_root)
5260 			ext4_msg(sb, KERN_WARNING,
5261 				"Quota file not on filesystem root. "
5262 				"Journaled quota will not work");
5263 	}
5264 
5265 	/*
5266 	 * When we journal data on quota file, we have to flush journal to see
5267 	 * all updates to the file when we bypass pagecache...
5268 	 */
5269 	if (EXT4_SB(sb)->s_journal &&
5270 	    ext4_should_journal_data(d_inode(path->dentry))) {
5271 		/*
5272 		 * We don't need to lock updates but journal_flush() could
5273 		 * otherwise be livelocked...
5274 		 */
5275 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5276 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5277 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5278 		if (err)
5279 			return err;
5280 	}
5281 
5282 	return dquot_quota_on(sb, type, format_id, path);
5283 }
5284 
5285 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5286 			     unsigned int flags)
5287 {
5288 	int err;
5289 	struct inode *qf_inode;
5290 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5291 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5292 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5293 	};
5294 
5295 	BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5296 
5297 	if (!qf_inums[type])
5298 		return -EPERM;
5299 
5300 	qf_inode = ext4_iget(sb, qf_inums[type]);
5301 	if (IS_ERR(qf_inode)) {
5302 		ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5303 		return PTR_ERR(qf_inode);
5304 	}
5305 
5306 	/* Don't account quota for quota files to avoid recursion */
5307 	qf_inode->i_flags |= S_NOQUOTA;
5308 	err = dquot_enable(qf_inode, type, format_id, flags);
5309 	iput(qf_inode);
5310 
5311 	return err;
5312 }
5313 
5314 /* Enable usage tracking for all quota types. */
5315 static int ext4_enable_quotas(struct super_block *sb)
5316 {
5317 	int type, err = 0;
5318 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5319 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5320 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5321 	};
5322 
5323 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5324 	for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5325 		if (qf_inums[type]) {
5326 			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5327 						DQUOT_USAGE_ENABLED);
5328 			if (err) {
5329 				ext4_warning(sb,
5330 					"Failed to enable quota tracking "
5331 					"(type=%d, err=%d). Please run "
5332 					"e2fsck to fix.", type, err);
5333 				return err;
5334 			}
5335 		}
5336 	}
5337 	return 0;
5338 }
5339 
5340 static int ext4_quota_off(struct super_block *sb, int type)
5341 {
5342 	struct inode *inode = sb_dqopt(sb)->files[type];
5343 	handle_t *handle;
5344 
5345 	/* Force all delayed allocation blocks to be allocated.
5346 	 * Caller already holds s_umount sem */
5347 	if (test_opt(sb, DELALLOC))
5348 		sync_filesystem(sb);
5349 
5350 	if (!inode)
5351 		goto out;
5352 
5353 	/* Update modification times of quota files when userspace can
5354 	 * start looking at them */
5355 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5356 	if (IS_ERR(handle))
5357 		goto out;
5358 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5359 	ext4_mark_inode_dirty(handle, inode);
5360 	ext4_journal_stop(handle);
5361 
5362 out:
5363 	return dquot_quota_off(sb, type);
5364 }
5365 
5366 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5367  * acquiring the locks... As quota files are never truncated and quota code
5368  * itself serializes the operations (and no one else should touch the files)
5369  * we don't have to be afraid of races */
5370 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5371 			       size_t len, loff_t off)
5372 {
5373 	struct inode *inode = sb_dqopt(sb)->files[type];
5374 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5375 	int offset = off & (sb->s_blocksize - 1);
5376 	int tocopy;
5377 	size_t toread;
5378 	struct buffer_head *bh;
5379 	loff_t i_size = i_size_read(inode);
5380 
5381 	if (off > i_size)
5382 		return 0;
5383 	if (off+len > i_size)
5384 		len = i_size-off;
5385 	toread = len;
5386 	while (toread > 0) {
5387 		tocopy = sb->s_blocksize - offset < toread ?
5388 				sb->s_blocksize - offset : toread;
5389 		bh = ext4_bread(NULL, inode, blk, 0);
5390 		if (IS_ERR(bh))
5391 			return PTR_ERR(bh);
5392 		if (!bh)	/* A hole? */
5393 			memset(data, 0, tocopy);
5394 		else
5395 			memcpy(data, bh->b_data+offset, tocopy);
5396 		brelse(bh);
5397 		offset = 0;
5398 		toread -= tocopy;
5399 		data += tocopy;
5400 		blk++;
5401 	}
5402 	return len;
5403 }
5404 
5405 /* Write to quotafile (we know the transaction is already started and has
5406  * enough credits) */
5407 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5408 				const char *data, size_t len, loff_t off)
5409 {
5410 	struct inode *inode = sb_dqopt(sb)->files[type];
5411 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5412 	int err, offset = off & (sb->s_blocksize - 1);
5413 	struct buffer_head *bh;
5414 	handle_t *handle = journal_current_handle();
5415 
5416 	if (EXT4_SB(sb)->s_journal && !handle) {
5417 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5418 			" cancelled because transaction is not started",
5419 			(unsigned long long)off, (unsigned long long)len);
5420 		return -EIO;
5421 	}
5422 	/*
5423 	 * Since we account only one data block in transaction credits,
5424 	 * then it is impossible to cross a block boundary.
5425 	 */
5426 	if (sb->s_blocksize - offset < len) {
5427 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5428 			" cancelled because not block aligned",
5429 			(unsigned long long)off, (unsigned long long)len);
5430 		return -EIO;
5431 	}
5432 
5433 	bh = ext4_bread(handle, inode, blk, 1);
5434 	if (IS_ERR(bh))
5435 		return PTR_ERR(bh);
5436 	if (!bh)
5437 		goto out;
5438 	BUFFER_TRACE(bh, "get write access");
5439 	err = ext4_journal_get_write_access(handle, bh);
5440 	if (err) {
5441 		brelse(bh);
5442 		return err;
5443 	}
5444 	lock_buffer(bh);
5445 	memcpy(bh->b_data+offset, data, len);
5446 	flush_dcache_page(bh->b_page);
5447 	unlock_buffer(bh);
5448 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
5449 	brelse(bh);
5450 out:
5451 	if (inode->i_size < off + len) {
5452 		i_size_write(inode, off + len);
5453 		EXT4_I(inode)->i_disksize = inode->i_size;
5454 		ext4_mark_inode_dirty(handle, inode);
5455 	}
5456 	return len;
5457 }
5458 
5459 #endif
5460 
5461 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5462 		       const char *dev_name, void *data)
5463 {
5464 	return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5465 }
5466 
5467 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5468 static inline void register_as_ext2(void)
5469 {
5470 	int err = register_filesystem(&ext2_fs_type);
5471 	if (err)
5472 		printk(KERN_WARNING
5473 		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5474 }
5475 
5476 static inline void unregister_as_ext2(void)
5477 {
5478 	unregister_filesystem(&ext2_fs_type);
5479 }
5480 
5481 static inline int ext2_feature_set_ok(struct super_block *sb)
5482 {
5483 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5484 		return 0;
5485 	if (sb->s_flags & MS_RDONLY)
5486 		return 1;
5487 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5488 		return 0;
5489 	return 1;
5490 }
5491 #else
5492 static inline void register_as_ext2(void) { }
5493 static inline void unregister_as_ext2(void) { }
5494 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5495 #endif
5496 
5497 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5498 static inline void register_as_ext3(void)
5499 {
5500 	int err = register_filesystem(&ext3_fs_type);
5501 	if (err)
5502 		printk(KERN_WARNING
5503 		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5504 }
5505 
5506 static inline void unregister_as_ext3(void)
5507 {
5508 	unregister_filesystem(&ext3_fs_type);
5509 }
5510 
5511 static inline int ext3_feature_set_ok(struct super_block *sb)
5512 {
5513 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5514 		return 0;
5515 	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5516 		return 0;
5517 	if (sb->s_flags & MS_RDONLY)
5518 		return 1;
5519 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5520 		return 0;
5521 	return 1;
5522 }
5523 #else
5524 static inline void register_as_ext3(void) { }
5525 static inline void unregister_as_ext3(void) { }
5526 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5527 #endif
5528 
5529 static struct file_system_type ext4_fs_type = {
5530 	.owner		= THIS_MODULE,
5531 	.name		= "ext4",
5532 	.mount		= ext4_mount,
5533 	.kill_sb	= kill_block_super,
5534 	.fs_flags	= FS_REQUIRES_DEV,
5535 };
5536 MODULE_ALIAS_FS("ext4");
5537 
5538 static int __init ext4_init_feat_adverts(void)
5539 {
5540 	struct ext4_features *ef;
5541 	int ret = -ENOMEM;
5542 
5543 	ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5544 	if (!ef)
5545 		goto out;
5546 
5547 	ef->f_kobj.kset = ext4_kset;
5548 	init_completion(&ef->f_kobj_unregister);
5549 	ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5550 				   "features");
5551 	if (ret) {
5552 		kfree(ef);
5553 		goto out;
5554 	}
5555 
5556 	ext4_feat = ef;
5557 	ret = 0;
5558 out:
5559 	return ret;
5560 }
5561 
5562 static void ext4_exit_feat_adverts(void)
5563 {
5564 	kobject_put(&ext4_feat->f_kobj);
5565 	wait_for_completion(&ext4_feat->f_kobj_unregister);
5566 	kfree(ext4_feat);
5567 }
5568 
5569 /* Shared across all ext4 file systems */
5570 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5571 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5572 
5573 static int __init ext4_init_fs(void)
5574 {
5575 	int i, err;
5576 
5577 	ext4_li_info = NULL;
5578 	mutex_init(&ext4_li_mtx);
5579 
5580 	/* Build-time check for flags consistency */
5581 	ext4_check_flag_values();
5582 
5583 	for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5584 		mutex_init(&ext4__aio_mutex[i]);
5585 		init_waitqueue_head(&ext4__ioend_wq[i]);
5586 	}
5587 
5588 	err = ext4_init_es();
5589 	if (err)
5590 		return err;
5591 
5592 	err = ext4_init_pageio();
5593 	if (err)
5594 		goto out7;
5595 
5596 	err = ext4_init_system_zone();
5597 	if (err)
5598 		goto out6;
5599 	ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5600 	if (!ext4_kset) {
5601 		err = -ENOMEM;
5602 		goto out5;
5603 	}
5604 	ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5605 
5606 	err = ext4_init_feat_adverts();
5607 	if (err)
5608 		goto out4;
5609 
5610 	err = ext4_init_mballoc();
5611 	if (err)
5612 		goto out2;
5613 	else
5614 		ext4_mballoc_ready = 1;
5615 	err = init_inodecache();
5616 	if (err)
5617 		goto out1;
5618 	register_as_ext3();
5619 	register_as_ext2();
5620 	err = register_filesystem(&ext4_fs_type);
5621 	if (err)
5622 		goto out;
5623 
5624 	return 0;
5625 out:
5626 	unregister_as_ext2();
5627 	unregister_as_ext3();
5628 	destroy_inodecache();
5629 out1:
5630 	ext4_mballoc_ready = 0;
5631 	ext4_exit_mballoc();
5632 out2:
5633 	ext4_exit_feat_adverts();
5634 out4:
5635 	if (ext4_proc_root)
5636 		remove_proc_entry("fs/ext4", NULL);
5637 	kset_unregister(ext4_kset);
5638 out5:
5639 	ext4_exit_system_zone();
5640 out6:
5641 	ext4_exit_pageio();
5642 out7:
5643 	ext4_exit_es();
5644 
5645 	return err;
5646 }
5647 
5648 static void __exit ext4_exit_fs(void)
5649 {
5650 	ext4_destroy_lazyinit_thread();
5651 	unregister_as_ext2();
5652 	unregister_as_ext3();
5653 	unregister_filesystem(&ext4_fs_type);
5654 	destroy_inodecache();
5655 	ext4_exit_mballoc();
5656 	ext4_exit_feat_adverts();
5657 	remove_proc_entry("fs/ext4", NULL);
5658 	kset_unregister(ext4_kset);
5659 	ext4_exit_system_zone();
5660 	ext4_exit_pageio();
5661 	ext4_exit_es();
5662 }
5663 
5664 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5665 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5666 MODULE_LICENSE("GPL");
5667 module_init(ext4_init_fs)
5668 module_exit(ext4_exit_fs)
5669