xref: /openbmc/linux/fs/ext4/super.c (revision 827beb77)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  linux/fs/ext4/super.c
4  *
5  * Copyright (C) 1992, 1993, 1994, 1995
6  * Remy Card (card@masi.ibp.fr)
7  * Laboratoire MASI - Institut Blaise Pascal
8  * Universite Pierre et Marie Curie (Paris VI)
9  *
10  *  from
11  *
12  *  linux/fs/minix/inode.c
13  *
14  *  Copyright (C) 1991, 1992  Linus Torvalds
15  *
16  *  Big-endian to little-endian byte-swapping/bitmaps by
17  *        David S. Miller (davem@caip.rutgers.edu), 1995
18  */
19 
20 #include <linux/module.h>
21 #include <linux/string.h>
22 #include <linux/fs.h>
23 #include <linux/time.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/dax.h>
42 #include <linux/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
45 #include <linux/unicode.h>
46 #include <linux/part_stat.h>
47 #include <linux/kthread.h>
48 #include <linux/freezer.h>
49 
50 #include "ext4.h"
51 #include "ext4_extents.h"	/* Needed for trace points definition */
52 #include "ext4_jbd2.h"
53 #include "xattr.h"
54 #include "acl.h"
55 #include "mballoc.h"
56 #include "fsmap.h"
57 
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/ext4.h>
60 
61 static struct ext4_lazy_init *ext4_li_info;
62 static DEFINE_MUTEX(ext4_li_mtx);
63 static struct ratelimit_state ext4_mount_msg_ratelimit;
64 
65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
66 			     unsigned long journal_devnum);
67 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
68 static void ext4_update_super(struct super_block *sb);
69 static int ext4_commit_super(struct super_block *sb);
70 static int ext4_mark_recovery_complete(struct super_block *sb,
71 					struct ext4_super_block *es);
72 static int ext4_clear_journal_err(struct super_block *sb,
73 				  struct ext4_super_block *es);
74 static int ext4_sync_fs(struct super_block *sb, int wait);
75 static int ext4_remount(struct super_block *sb, int *flags, char *data);
76 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
77 static int ext4_unfreeze(struct super_block *sb);
78 static int ext4_freeze(struct super_block *sb);
79 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 		       const char *dev_name, void *data);
81 static inline int ext2_feature_set_ok(struct super_block *sb);
82 static inline int ext3_feature_set_ok(struct super_block *sb);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static struct inode *ext4_get_journal_inode(struct super_block *sb,
87 					    unsigned int journal_inum);
88 
89 /*
90  * Lock ordering
91  *
92  * page fault path:
93  * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
94  *   -> page lock -> i_data_sem (rw)
95  *
96  * buffered write path:
97  * sb_start_write -> i_mutex -> mmap_lock
98  * sb_start_write -> i_mutex -> transaction start -> page lock ->
99  *   i_data_sem (rw)
100  *
101  * truncate:
102  * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
103  *   page lock
104  * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
105  *   i_data_sem (rw)
106  *
107  * direct IO:
108  * sb_start_write -> i_mutex -> mmap_lock
109  * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
110  *
111  * writepages:
112  * transaction start -> page lock(s) -> i_data_sem (rw)
113  */
114 
115 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
116 static struct file_system_type ext2_fs_type = {
117 	.owner		= THIS_MODULE,
118 	.name		= "ext2",
119 	.mount		= ext4_mount,
120 	.kill_sb	= kill_block_super,
121 	.fs_flags	= FS_REQUIRES_DEV,
122 };
123 MODULE_ALIAS_FS("ext2");
124 MODULE_ALIAS("ext2");
125 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
126 #else
127 #define IS_EXT2_SB(sb) (0)
128 #endif
129 
130 
131 static struct file_system_type ext3_fs_type = {
132 	.owner		= THIS_MODULE,
133 	.name		= "ext3",
134 	.mount		= ext4_mount,
135 	.kill_sb	= kill_block_super,
136 	.fs_flags	= FS_REQUIRES_DEV,
137 };
138 MODULE_ALIAS_FS("ext3");
139 MODULE_ALIAS("ext3");
140 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
141 
142 
143 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
144 				  bh_end_io_t *end_io)
145 {
146 	/*
147 	 * buffer's verified bit is no longer valid after reading from
148 	 * disk again due to write out error, clear it to make sure we
149 	 * recheck the buffer contents.
150 	 */
151 	clear_buffer_verified(bh);
152 
153 	bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
154 	get_bh(bh);
155 	submit_bh(REQ_OP_READ, op_flags, bh);
156 }
157 
158 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
159 			 bh_end_io_t *end_io)
160 {
161 	BUG_ON(!buffer_locked(bh));
162 
163 	if (ext4_buffer_uptodate(bh)) {
164 		unlock_buffer(bh);
165 		return;
166 	}
167 	__ext4_read_bh(bh, op_flags, end_io);
168 }
169 
170 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
171 {
172 	BUG_ON(!buffer_locked(bh));
173 
174 	if (ext4_buffer_uptodate(bh)) {
175 		unlock_buffer(bh);
176 		return 0;
177 	}
178 
179 	__ext4_read_bh(bh, op_flags, end_io);
180 
181 	wait_on_buffer(bh);
182 	if (buffer_uptodate(bh))
183 		return 0;
184 	return -EIO;
185 }
186 
187 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
188 {
189 	if (trylock_buffer(bh)) {
190 		if (wait)
191 			return ext4_read_bh(bh, op_flags, NULL);
192 		ext4_read_bh_nowait(bh, op_flags, NULL);
193 		return 0;
194 	}
195 	if (wait) {
196 		wait_on_buffer(bh);
197 		if (buffer_uptodate(bh))
198 			return 0;
199 		return -EIO;
200 	}
201 	return 0;
202 }
203 
204 /*
205  * This works like __bread_gfp() except it uses ERR_PTR for error
206  * returns.  Currently with sb_bread it's impossible to distinguish
207  * between ENOMEM and EIO situations (since both result in a NULL
208  * return.
209  */
210 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
211 					       sector_t block, int op_flags,
212 					       gfp_t gfp)
213 {
214 	struct buffer_head *bh;
215 	int ret;
216 
217 	bh = sb_getblk_gfp(sb, block, gfp);
218 	if (bh == NULL)
219 		return ERR_PTR(-ENOMEM);
220 	if (ext4_buffer_uptodate(bh))
221 		return bh;
222 
223 	ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
224 	if (ret) {
225 		put_bh(bh);
226 		return ERR_PTR(ret);
227 	}
228 	return bh;
229 }
230 
231 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
232 				   int op_flags)
233 {
234 	return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
235 }
236 
237 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
238 					    sector_t block)
239 {
240 	return __ext4_sb_bread_gfp(sb, block, 0, 0);
241 }
242 
243 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
244 {
245 	struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
246 
247 	if (likely(bh)) {
248 		ext4_read_bh_lock(bh, REQ_RAHEAD, false);
249 		brelse(bh);
250 	}
251 }
252 
253 static int ext4_verify_csum_type(struct super_block *sb,
254 				 struct ext4_super_block *es)
255 {
256 	if (!ext4_has_feature_metadata_csum(sb))
257 		return 1;
258 
259 	return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
260 }
261 
262 static __le32 ext4_superblock_csum(struct super_block *sb,
263 				   struct ext4_super_block *es)
264 {
265 	struct ext4_sb_info *sbi = EXT4_SB(sb);
266 	int offset = offsetof(struct ext4_super_block, s_checksum);
267 	__u32 csum;
268 
269 	csum = ext4_chksum(sbi, ~0, (char *)es, offset);
270 
271 	return cpu_to_le32(csum);
272 }
273 
274 static int ext4_superblock_csum_verify(struct super_block *sb,
275 				       struct ext4_super_block *es)
276 {
277 	if (!ext4_has_metadata_csum(sb))
278 		return 1;
279 
280 	return es->s_checksum == ext4_superblock_csum(sb, es);
281 }
282 
283 void ext4_superblock_csum_set(struct super_block *sb)
284 {
285 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
286 
287 	if (!ext4_has_metadata_csum(sb))
288 		return;
289 
290 	es->s_checksum = ext4_superblock_csum(sb, es);
291 }
292 
293 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
294 			       struct ext4_group_desc *bg)
295 {
296 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
297 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
298 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
299 }
300 
301 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
302 			       struct ext4_group_desc *bg)
303 {
304 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
305 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
306 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
307 }
308 
309 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
310 			      struct ext4_group_desc *bg)
311 {
312 	return le32_to_cpu(bg->bg_inode_table_lo) |
313 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
314 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
315 }
316 
317 __u32 ext4_free_group_clusters(struct super_block *sb,
318 			       struct ext4_group_desc *bg)
319 {
320 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
321 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
322 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
323 }
324 
325 __u32 ext4_free_inodes_count(struct super_block *sb,
326 			      struct ext4_group_desc *bg)
327 {
328 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
329 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
330 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
331 }
332 
333 __u32 ext4_used_dirs_count(struct super_block *sb,
334 			      struct ext4_group_desc *bg)
335 {
336 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
337 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
338 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
339 }
340 
341 __u32 ext4_itable_unused_count(struct super_block *sb,
342 			      struct ext4_group_desc *bg)
343 {
344 	return le16_to_cpu(bg->bg_itable_unused_lo) |
345 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
346 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
347 }
348 
349 void ext4_block_bitmap_set(struct super_block *sb,
350 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
351 {
352 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
353 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
354 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
355 }
356 
357 void ext4_inode_bitmap_set(struct super_block *sb,
358 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
359 {
360 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
361 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
362 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
363 }
364 
365 void ext4_inode_table_set(struct super_block *sb,
366 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
367 {
368 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
369 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
370 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
371 }
372 
373 void ext4_free_group_clusters_set(struct super_block *sb,
374 				  struct ext4_group_desc *bg, __u32 count)
375 {
376 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
377 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
378 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
379 }
380 
381 void ext4_free_inodes_set(struct super_block *sb,
382 			  struct ext4_group_desc *bg, __u32 count)
383 {
384 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
385 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
386 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
387 }
388 
389 void ext4_used_dirs_set(struct super_block *sb,
390 			  struct ext4_group_desc *bg, __u32 count)
391 {
392 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
393 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
394 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
395 }
396 
397 void ext4_itable_unused_set(struct super_block *sb,
398 			  struct ext4_group_desc *bg, __u32 count)
399 {
400 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
401 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
402 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
403 }
404 
405 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
406 {
407 	now = clamp_val(now, 0, (1ull << 40) - 1);
408 
409 	*lo = cpu_to_le32(lower_32_bits(now));
410 	*hi = upper_32_bits(now);
411 }
412 
413 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
414 {
415 	return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
416 }
417 #define ext4_update_tstamp(es, tstamp) \
418 	__ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
419 			     ktime_get_real_seconds())
420 #define ext4_get_tstamp(es, tstamp) \
421 	__ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
422 
423 /*
424  * The del_gendisk() function uninitializes the disk-specific data
425  * structures, including the bdi structure, without telling anyone
426  * else.  Once this happens, any attempt to call mark_buffer_dirty()
427  * (for example, by ext4_commit_super), will cause a kernel OOPS.
428  * This is a kludge to prevent these oops until we can put in a proper
429  * hook in del_gendisk() to inform the VFS and file system layers.
430  */
431 static int block_device_ejected(struct super_block *sb)
432 {
433 	struct inode *bd_inode = sb->s_bdev->bd_inode;
434 	struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
435 
436 	return bdi->dev == NULL;
437 }
438 
439 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
440 {
441 	struct super_block		*sb = journal->j_private;
442 	struct ext4_sb_info		*sbi = EXT4_SB(sb);
443 	int				error = is_journal_aborted(journal);
444 	struct ext4_journal_cb_entry	*jce;
445 
446 	BUG_ON(txn->t_state == T_FINISHED);
447 
448 	ext4_process_freed_data(sb, txn->t_tid);
449 
450 	spin_lock(&sbi->s_md_lock);
451 	while (!list_empty(&txn->t_private_list)) {
452 		jce = list_entry(txn->t_private_list.next,
453 				 struct ext4_journal_cb_entry, jce_list);
454 		list_del_init(&jce->jce_list);
455 		spin_unlock(&sbi->s_md_lock);
456 		jce->jce_func(sb, jce, error);
457 		spin_lock(&sbi->s_md_lock);
458 	}
459 	spin_unlock(&sbi->s_md_lock);
460 }
461 
462 /*
463  * This writepage callback for write_cache_pages()
464  * takes care of a few cases after page cleaning.
465  *
466  * write_cache_pages() already checks for dirty pages
467  * and calls clear_page_dirty_for_io(), which we want,
468  * to write protect the pages.
469  *
470  * However, we may have to redirty a page (see below.)
471  */
472 static int ext4_journalled_writepage_callback(struct page *page,
473 					      struct writeback_control *wbc,
474 					      void *data)
475 {
476 	transaction_t *transaction = (transaction_t *) data;
477 	struct buffer_head *bh, *head;
478 	struct journal_head *jh;
479 
480 	bh = head = page_buffers(page);
481 	do {
482 		/*
483 		 * We have to redirty a page in these cases:
484 		 * 1) If buffer is dirty, it means the page was dirty because it
485 		 * contains a buffer that needs checkpointing. So the dirty bit
486 		 * needs to be preserved so that checkpointing writes the buffer
487 		 * properly.
488 		 * 2) If buffer is not part of the committing transaction
489 		 * (we may have just accidentally come across this buffer because
490 		 * inode range tracking is not exact) or if the currently running
491 		 * transaction already contains this buffer as well, dirty bit
492 		 * needs to be preserved so that the buffer gets writeprotected
493 		 * properly on running transaction's commit.
494 		 */
495 		jh = bh2jh(bh);
496 		if (buffer_dirty(bh) ||
497 		    (jh && (jh->b_transaction != transaction ||
498 			    jh->b_next_transaction))) {
499 			redirty_page_for_writepage(wbc, page);
500 			goto out;
501 		}
502 	} while ((bh = bh->b_this_page) != head);
503 
504 out:
505 	return AOP_WRITEPAGE_ACTIVATE;
506 }
507 
508 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
509 {
510 	struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
511 	struct writeback_control wbc = {
512 		.sync_mode =  WB_SYNC_ALL,
513 		.nr_to_write = LONG_MAX,
514 		.range_start = jinode->i_dirty_start,
515 		.range_end = jinode->i_dirty_end,
516         };
517 
518 	return write_cache_pages(mapping, &wbc,
519 				 ext4_journalled_writepage_callback,
520 				 jinode->i_transaction);
521 }
522 
523 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
524 {
525 	int ret;
526 
527 	if (ext4_should_journal_data(jinode->i_vfs_inode))
528 		ret = ext4_journalled_submit_inode_data_buffers(jinode);
529 	else
530 		ret = jbd2_journal_submit_inode_data_buffers(jinode);
531 
532 	return ret;
533 }
534 
535 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
536 {
537 	int ret = 0;
538 
539 	if (!ext4_should_journal_data(jinode->i_vfs_inode))
540 		ret = jbd2_journal_finish_inode_data_buffers(jinode);
541 
542 	return ret;
543 }
544 
545 static bool system_going_down(void)
546 {
547 	return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
548 		|| system_state == SYSTEM_RESTART;
549 }
550 
551 struct ext4_err_translation {
552 	int code;
553 	int errno;
554 };
555 
556 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
557 
558 static struct ext4_err_translation err_translation[] = {
559 	EXT4_ERR_TRANSLATE(EIO),
560 	EXT4_ERR_TRANSLATE(ENOMEM),
561 	EXT4_ERR_TRANSLATE(EFSBADCRC),
562 	EXT4_ERR_TRANSLATE(EFSCORRUPTED),
563 	EXT4_ERR_TRANSLATE(ENOSPC),
564 	EXT4_ERR_TRANSLATE(ENOKEY),
565 	EXT4_ERR_TRANSLATE(EROFS),
566 	EXT4_ERR_TRANSLATE(EFBIG),
567 	EXT4_ERR_TRANSLATE(EEXIST),
568 	EXT4_ERR_TRANSLATE(ERANGE),
569 	EXT4_ERR_TRANSLATE(EOVERFLOW),
570 	EXT4_ERR_TRANSLATE(EBUSY),
571 	EXT4_ERR_TRANSLATE(ENOTDIR),
572 	EXT4_ERR_TRANSLATE(ENOTEMPTY),
573 	EXT4_ERR_TRANSLATE(ESHUTDOWN),
574 	EXT4_ERR_TRANSLATE(EFAULT),
575 };
576 
577 static int ext4_errno_to_code(int errno)
578 {
579 	int i;
580 
581 	for (i = 0; i < ARRAY_SIZE(err_translation); i++)
582 		if (err_translation[i].errno == errno)
583 			return err_translation[i].code;
584 	return EXT4_ERR_UNKNOWN;
585 }
586 
587 static void save_error_info(struct super_block *sb, int error,
588 			    __u32 ino, __u64 block,
589 			    const char *func, unsigned int line)
590 {
591 	struct ext4_sb_info *sbi = EXT4_SB(sb);
592 
593 	/* We default to EFSCORRUPTED error... */
594 	if (error == 0)
595 		error = EFSCORRUPTED;
596 
597 	spin_lock(&sbi->s_error_lock);
598 	sbi->s_add_error_count++;
599 	sbi->s_last_error_code = error;
600 	sbi->s_last_error_line = line;
601 	sbi->s_last_error_ino = ino;
602 	sbi->s_last_error_block = block;
603 	sbi->s_last_error_func = func;
604 	sbi->s_last_error_time = ktime_get_real_seconds();
605 	if (!sbi->s_first_error_time) {
606 		sbi->s_first_error_code = error;
607 		sbi->s_first_error_line = line;
608 		sbi->s_first_error_ino = ino;
609 		sbi->s_first_error_block = block;
610 		sbi->s_first_error_func = func;
611 		sbi->s_first_error_time = sbi->s_last_error_time;
612 	}
613 	spin_unlock(&sbi->s_error_lock);
614 }
615 
616 /* Deal with the reporting of failure conditions on a filesystem such as
617  * inconsistencies detected or read IO failures.
618  *
619  * On ext2, we can store the error state of the filesystem in the
620  * superblock.  That is not possible on ext4, because we may have other
621  * write ordering constraints on the superblock which prevent us from
622  * writing it out straight away; and given that the journal is about to
623  * be aborted, we can't rely on the current, or future, transactions to
624  * write out the superblock safely.
625  *
626  * We'll just use the jbd2_journal_abort() error code to record an error in
627  * the journal instead.  On recovery, the journal will complain about
628  * that error until we've noted it down and cleared it.
629  *
630  * If force_ro is set, we unconditionally force the filesystem into an
631  * ABORT|READONLY state, unless the error response on the fs has been set to
632  * panic in which case we take the easy way out and panic immediately. This is
633  * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
634  * at a critical moment in log management.
635  */
636 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
637 			      __u32 ino, __u64 block,
638 			      const char *func, unsigned int line)
639 {
640 	journal_t *journal = EXT4_SB(sb)->s_journal;
641 	bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
642 
643 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
644 	if (test_opt(sb, WARN_ON_ERROR))
645 		WARN_ON_ONCE(1);
646 
647 	if (!continue_fs && !sb_rdonly(sb)) {
648 		ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
649 		if (journal)
650 			jbd2_journal_abort(journal, -EIO);
651 	}
652 
653 	if (!bdev_read_only(sb->s_bdev)) {
654 		save_error_info(sb, error, ino, block, func, line);
655 		/*
656 		 * In case the fs should keep running, we need to writeout
657 		 * superblock through the journal. Due to lock ordering
658 		 * constraints, it may not be safe to do it right here so we
659 		 * defer superblock flushing to a workqueue.
660 		 */
661 		if (continue_fs && journal)
662 			schedule_work(&EXT4_SB(sb)->s_error_work);
663 		else
664 			ext4_commit_super(sb);
665 	}
666 
667 	/*
668 	 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
669 	 * could panic during 'reboot -f' as the underlying device got already
670 	 * disabled.
671 	 */
672 	if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
673 		panic("EXT4-fs (device %s): panic forced after error\n",
674 			sb->s_id);
675 	}
676 
677 	if (sb_rdonly(sb) || continue_fs)
678 		return;
679 
680 	ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
681 	/*
682 	 * Make sure updated value of ->s_mount_flags will be visible before
683 	 * ->s_flags update
684 	 */
685 	smp_wmb();
686 	sb->s_flags |= SB_RDONLY;
687 }
688 
689 static void flush_stashed_error_work(struct work_struct *work)
690 {
691 	struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
692 						s_error_work);
693 	journal_t *journal = sbi->s_journal;
694 	handle_t *handle;
695 
696 	/*
697 	 * If the journal is still running, we have to write out superblock
698 	 * through the journal to avoid collisions of other journalled sb
699 	 * updates.
700 	 *
701 	 * We use directly jbd2 functions here to avoid recursing back into
702 	 * ext4 error handling code during handling of previous errors.
703 	 */
704 	if (!sb_rdonly(sbi->s_sb) && journal) {
705 		struct buffer_head *sbh = sbi->s_sbh;
706 		handle = jbd2_journal_start(journal, 1);
707 		if (IS_ERR(handle))
708 			goto write_directly;
709 		if (jbd2_journal_get_write_access(handle, sbh)) {
710 			jbd2_journal_stop(handle);
711 			goto write_directly;
712 		}
713 		ext4_update_super(sbi->s_sb);
714 		if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
715 			ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
716 				 "superblock detected");
717 			clear_buffer_write_io_error(sbh);
718 			set_buffer_uptodate(sbh);
719 		}
720 
721 		if (jbd2_journal_dirty_metadata(handle, sbh)) {
722 			jbd2_journal_stop(handle);
723 			goto write_directly;
724 		}
725 		jbd2_journal_stop(handle);
726 		ext4_notify_error_sysfs(sbi);
727 		return;
728 	}
729 write_directly:
730 	/*
731 	 * Write through journal failed. Write sb directly to get error info
732 	 * out and hope for the best.
733 	 */
734 	ext4_commit_super(sbi->s_sb);
735 	ext4_notify_error_sysfs(sbi);
736 }
737 
738 #define ext4_error_ratelimit(sb)					\
739 		___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),	\
740 			     "EXT4-fs error")
741 
742 void __ext4_error(struct super_block *sb, const char *function,
743 		  unsigned int line, bool force_ro, int error, __u64 block,
744 		  const char *fmt, ...)
745 {
746 	struct va_format vaf;
747 	va_list args;
748 
749 	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
750 		return;
751 
752 	trace_ext4_error(sb, function, line);
753 	if (ext4_error_ratelimit(sb)) {
754 		va_start(args, fmt);
755 		vaf.fmt = fmt;
756 		vaf.va = &args;
757 		printk(KERN_CRIT
758 		       "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
759 		       sb->s_id, function, line, current->comm, &vaf);
760 		va_end(args);
761 	}
762 	ext4_handle_error(sb, force_ro, error, 0, block, function, line);
763 }
764 
765 void __ext4_error_inode(struct inode *inode, const char *function,
766 			unsigned int line, ext4_fsblk_t block, int error,
767 			const char *fmt, ...)
768 {
769 	va_list args;
770 	struct va_format vaf;
771 
772 	if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
773 		return;
774 
775 	trace_ext4_error(inode->i_sb, function, line);
776 	if (ext4_error_ratelimit(inode->i_sb)) {
777 		va_start(args, fmt);
778 		vaf.fmt = fmt;
779 		vaf.va = &args;
780 		if (block)
781 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
782 			       "inode #%lu: block %llu: comm %s: %pV\n",
783 			       inode->i_sb->s_id, function, line, inode->i_ino,
784 			       block, current->comm, &vaf);
785 		else
786 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
787 			       "inode #%lu: comm %s: %pV\n",
788 			       inode->i_sb->s_id, function, line, inode->i_ino,
789 			       current->comm, &vaf);
790 		va_end(args);
791 	}
792 	ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
793 			  function, line);
794 }
795 
796 void __ext4_error_file(struct file *file, const char *function,
797 		       unsigned int line, ext4_fsblk_t block,
798 		       const char *fmt, ...)
799 {
800 	va_list args;
801 	struct va_format vaf;
802 	struct inode *inode = file_inode(file);
803 	char pathname[80], *path;
804 
805 	if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
806 		return;
807 
808 	trace_ext4_error(inode->i_sb, function, line);
809 	if (ext4_error_ratelimit(inode->i_sb)) {
810 		path = file_path(file, pathname, sizeof(pathname));
811 		if (IS_ERR(path))
812 			path = "(unknown)";
813 		va_start(args, fmt);
814 		vaf.fmt = fmt;
815 		vaf.va = &args;
816 		if (block)
817 			printk(KERN_CRIT
818 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
819 			       "block %llu: comm %s: path %s: %pV\n",
820 			       inode->i_sb->s_id, function, line, inode->i_ino,
821 			       block, current->comm, path, &vaf);
822 		else
823 			printk(KERN_CRIT
824 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
825 			       "comm %s: path %s: %pV\n",
826 			       inode->i_sb->s_id, function, line, inode->i_ino,
827 			       current->comm, path, &vaf);
828 		va_end(args);
829 	}
830 	ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
831 			  function, line);
832 }
833 
834 const char *ext4_decode_error(struct super_block *sb, int errno,
835 			      char nbuf[16])
836 {
837 	char *errstr = NULL;
838 
839 	switch (errno) {
840 	case -EFSCORRUPTED:
841 		errstr = "Corrupt filesystem";
842 		break;
843 	case -EFSBADCRC:
844 		errstr = "Filesystem failed CRC";
845 		break;
846 	case -EIO:
847 		errstr = "IO failure";
848 		break;
849 	case -ENOMEM:
850 		errstr = "Out of memory";
851 		break;
852 	case -EROFS:
853 		if (!sb || (EXT4_SB(sb)->s_journal &&
854 			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
855 			errstr = "Journal has aborted";
856 		else
857 			errstr = "Readonly filesystem";
858 		break;
859 	default:
860 		/* If the caller passed in an extra buffer for unknown
861 		 * errors, textualise them now.  Else we just return
862 		 * NULL. */
863 		if (nbuf) {
864 			/* Check for truncated error codes... */
865 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
866 				errstr = nbuf;
867 		}
868 		break;
869 	}
870 
871 	return errstr;
872 }
873 
874 /* __ext4_std_error decodes expected errors from journaling functions
875  * automatically and invokes the appropriate error response.  */
876 
877 void __ext4_std_error(struct super_block *sb, const char *function,
878 		      unsigned int line, int errno)
879 {
880 	char nbuf[16];
881 	const char *errstr;
882 
883 	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
884 		return;
885 
886 	/* Special case: if the error is EROFS, and we're not already
887 	 * inside a transaction, then there's really no point in logging
888 	 * an error. */
889 	if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
890 		return;
891 
892 	if (ext4_error_ratelimit(sb)) {
893 		errstr = ext4_decode_error(sb, errno, nbuf);
894 		printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
895 		       sb->s_id, function, line, errstr);
896 	}
897 
898 	ext4_handle_error(sb, false, -errno, 0, 0, function, line);
899 }
900 
901 void __ext4_msg(struct super_block *sb,
902 		const char *prefix, const char *fmt, ...)
903 {
904 	struct va_format vaf;
905 	va_list args;
906 
907 	atomic_inc(&EXT4_SB(sb)->s_msg_count);
908 	if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
909 		return;
910 
911 	va_start(args, fmt);
912 	vaf.fmt = fmt;
913 	vaf.va = &args;
914 	printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
915 	va_end(args);
916 }
917 
918 static int ext4_warning_ratelimit(struct super_block *sb)
919 {
920 	atomic_inc(&EXT4_SB(sb)->s_warning_count);
921 	return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
922 			    "EXT4-fs warning");
923 }
924 
925 void __ext4_warning(struct super_block *sb, const char *function,
926 		    unsigned int line, const char *fmt, ...)
927 {
928 	struct va_format vaf;
929 	va_list args;
930 
931 	if (!ext4_warning_ratelimit(sb))
932 		return;
933 
934 	va_start(args, fmt);
935 	vaf.fmt = fmt;
936 	vaf.va = &args;
937 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
938 	       sb->s_id, function, line, &vaf);
939 	va_end(args);
940 }
941 
942 void __ext4_warning_inode(const struct inode *inode, const char *function,
943 			  unsigned int line, const char *fmt, ...)
944 {
945 	struct va_format vaf;
946 	va_list args;
947 
948 	if (!ext4_warning_ratelimit(inode->i_sb))
949 		return;
950 
951 	va_start(args, fmt);
952 	vaf.fmt = fmt;
953 	vaf.va = &args;
954 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
955 	       "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
956 	       function, line, inode->i_ino, current->comm, &vaf);
957 	va_end(args);
958 }
959 
960 void __ext4_grp_locked_error(const char *function, unsigned int line,
961 			     struct super_block *sb, ext4_group_t grp,
962 			     unsigned long ino, ext4_fsblk_t block,
963 			     const char *fmt, ...)
964 __releases(bitlock)
965 __acquires(bitlock)
966 {
967 	struct va_format vaf;
968 	va_list args;
969 
970 	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
971 		return;
972 
973 	trace_ext4_error(sb, function, line);
974 	if (ext4_error_ratelimit(sb)) {
975 		va_start(args, fmt);
976 		vaf.fmt = fmt;
977 		vaf.va = &args;
978 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
979 		       sb->s_id, function, line, grp);
980 		if (ino)
981 			printk(KERN_CONT "inode %lu: ", ino);
982 		if (block)
983 			printk(KERN_CONT "block %llu:",
984 			       (unsigned long long) block);
985 		printk(KERN_CONT "%pV\n", &vaf);
986 		va_end(args);
987 	}
988 
989 	if (test_opt(sb, ERRORS_CONT)) {
990 		if (test_opt(sb, WARN_ON_ERROR))
991 			WARN_ON_ONCE(1);
992 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
993 		if (!bdev_read_only(sb->s_bdev)) {
994 			save_error_info(sb, EFSCORRUPTED, ino, block, function,
995 					line);
996 			schedule_work(&EXT4_SB(sb)->s_error_work);
997 		}
998 		return;
999 	}
1000 	ext4_unlock_group(sb, grp);
1001 	ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1002 	/*
1003 	 * We only get here in the ERRORS_RO case; relocking the group
1004 	 * may be dangerous, but nothing bad will happen since the
1005 	 * filesystem will have already been marked read/only and the
1006 	 * journal has been aborted.  We return 1 as a hint to callers
1007 	 * who might what to use the return value from
1008 	 * ext4_grp_locked_error() to distinguish between the
1009 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1010 	 * aggressively from the ext4 function in question, with a
1011 	 * more appropriate error code.
1012 	 */
1013 	ext4_lock_group(sb, grp);
1014 	return;
1015 }
1016 
1017 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1018 				     ext4_group_t group,
1019 				     unsigned int flags)
1020 {
1021 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1022 	struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1023 	struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1024 	int ret;
1025 
1026 	if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1027 		ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1028 					    &grp->bb_state);
1029 		if (!ret)
1030 			percpu_counter_sub(&sbi->s_freeclusters_counter,
1031 					   grp->bb_free);
1032 	}
1033 
1034 	if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1035 		ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1036 					    &grp->bb_state);
1037 		if (!ret && gdp) {
1038 			int count;
1039 
1040 			count = ext4_free_inodes_count(sb, gdp);
1041 			percpu_counter_sub(&sbi->s_freeinodes_counter,
1042 					   count);
1043 		}
1044 	}
1045 }
1046 
1047 void ext4_update_dynamic_rev(struct super_block *sb)
1048 {
1049 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1050 
1051 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1052 		return;
1053 
1054 	ext4_warning(sb,
1055 		     "updating to rev %d because of new feature flag, "
1056 		     "running e2fsck is recommended",
1057 		     EXT4_DYNAMIC_REV);
1058 
1059 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1060 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1061 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1062 	/* leave es->s_feature_*compat flags alone */
1063 	/* es->s_uuid will be set by e2fsck if empty */
1064 
1065 	/*
1066 	 * The rest of the superblock fields should be zero, and if not it
1067 	 * means they are likely already in use, so leave them alone.  We
1068 	 * can leave it up to e2fsck to clean up any inconsistencies there.
1069 	 */
1070 }
1071 
1072 /*
1073  * Open the external journal device
1074  */
1075 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1076 {
1077 	struct block_device *bdev;
1078 
1079 	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1080 	if (IS_ERR(bdev))
1081 		goto fail;
1082 	return bdev;
1083 
1084 fail:
1085 	ext4_msg(sb, KERN_ERR,
1086 		 "failed to open journal device unknown-block(%u,%u) %ld",
1087 		 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1088 	return NULL;
1089 }
1090 
1091 /*
1092  * Release the journal device
1093  */
1094 static void ext4_blkdev_put(struct block_device *bdev)
1095 {
1096 	blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1097 }
1098 
1099 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1100 {
1101 	struct block_device *bdev;
1102 	bdev = sbi->s_journal_bdev;
1103 	if (bdev) {
1104 		ext4_blkdev_put(bdev);
1105 		sbi->s_journal_bdev = NULL;
1106 	}
1107 }
1108 
1109 static inline struct inode *orphan_list_entry(struct list_head *l)
1110 {
1111 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1112 }
1113 
1114 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1115 {
1116 	struct list_head *l;
1117 
1118 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1119 		 le32_to_cpu(sbi->s_es->s_last_orphan));
1120 
1121 	printk(KERN_ERR "sb_info orphan list:\n");
1122 	list_for_each(l, &sbi->s_orphan) {
1123 		struct inode *inode = orphan_list_entry(l);
1124 		printk(KERN_ERR "  "
1125 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1126 		       inode->i_sb->s_id, inode->i_ino, inode,
1127 		       inode->i_mode, inode->i_nlink,
1128 		       NEXT_ORPHAN(inode));
1129 	}
1130 }
1131 
1132 #ifdef CONFIG_QUOTA
1133 static int ext4_quota_off(struct super_block *sb, int type);
1134 
1135 static inline void ext4_quota_off_umount(struct super_block *sb)
1136 {
1137 	int type;
1138 
1139 	/* Use our quota_off function to clear inode flags etc. */
1140 	for (type = 0; type < EXT4_MAXQUOTAS; type++)
1141 		ext4_quota_off(sb, type);
1142 }
1143 
1144 /*
1145  * This is a helper function which is used in the mount/remount
1146  * codepaths (which holds s_umount) to fetch the quota file name.
1147  */
1148 static inline char *get_qf_name(struct super_block *sb,
1149 				struct ext4_sb_info *sbi,
1150 				int type)
1151 {
1152 	return rcu_dereference_protected(sbi->s_qf_names[type],
1153 					 lockdep_is_held(&sb->s_umount));
1154 }
1155 #else
1156 static inline void ext4_quota_off_umount(struct super_block *sb)
1157 {
1158 }
1159 #endif
1160 
1161 static void ext4_put_super(struct super_block *sb)
1162 {
1163 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1164 	struct ext4_super_block *es = sbi->s_es;
1165 	struct buffer_head **group_desc;
1166 	struct flex_groups **flex_groups;
1167 	int aborted = 0;
1168 	int i, err;
1169 
1170 	ext4_unregister_li_request(sb);
1171 	ext4_quota_off_umount(sb);
1172 
1173 	flush_work(&sbi->s_error_work);
1174 	destroy_workqueue(sbi->rsv_conversion_wq);
1175 	ext4_release_orphan_info(sb);
1176 
1177 	/*
1178 	 * Unregister sysfs before destroying jbd2 journal.
1179 	 * Since we could still access attr_journal_task attribute via sysfs
1180 	 * path which could have sbi->s_journal->j_task as NULL
1181 	 */
1182 	ext4_unregister_sysfs(sb);
1183 
1184 	if (sbi->s_journal) {
1185 		aborted = is_journal_aborted(sbi->s_journal);
1186 		err = jbd2_journal_destroy(sbi->s_journal);
1187 		sbi->s_journal = NULL;
1188 		if ((err < 0) && !aborted) {
1189 			ext4_abort(sb, -err, "Couldn't clean up the journal");
1190 		}
1191 	}
1192 
1193 	ext4_es_unregister_shrinker(sbi);
1194 	del_timer_sync(&sbi->s_err_report);
1195 	ext4_release_system_zone(sb);
1196 	ext4_mb_release(sb);
1197 	ext4_ext_release(sb);
1198 
1199 	if (!sb_rdonly(sb) && !aborted) {
1200 		ext4_clear_feature_journal_needs_recovery(sb);
1201 		ext4_clear_feature_orphan_present(sb);
1202 		es->s_state = cpu_to_le16(sbi->s_mount_state);
1203 	}
1204 	if (!sb_rdonly(sb))
1205 		ext4_commit_super(sb);
1206 
1207 	rcu_read_lock();
1208 	group_desc = rcu_dereference(sbi->s_group_desc);
1209 	for (i = 0; i < sbi->s_gdb_count; i++)
1210 		brelse(group_desc[i]);
1211 	kvfree(group_desc);
1212 	flex_groups = rcu_dereference(sbi->s_flex_groups);
1213 	if (flex_groups) {
1214 		for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1215 			kvfree(flex_groups[i]);
1216 		kvfree(flex_groups);
1217 	}
1218 	rcu_read_unlock();
1219 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
1220 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1221 	percpu_counter_destroy(&sbi->s_dirs_counter);
1222 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1223 	percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1224 	percpu_free_rwsem(&sbi->s_writepages_rwsem);
1225 #ifdef CONFIG_QUOTA
1226 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
1227 		kfree(get_qf_name(sb, sbi, i));
1228 #endif
1229 
1230 	/* Debugging code just in case the in-memory inode orphan list
1231 	 * isn't empty.  The on-disk one can be non-empty if we've
1232 	 * detected an error and taken the fs readonly, but the
1233 	 * in-memory list had better be clean by this point. */
1234 	if (!list_empty(&sbi->s_orphan))
1235 		dump_orphan_list(sb, sbi);
1236 	ASSERT(list_empty(&sbi->s_orphan));
1237 
1238 	sync_blockdev(sb->s_bdev);
1239 	invalidate_bdev(sb->s_bdev);
1240 	if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1241 		/*
1242 		 * Invalidate the journal device's buffers.  We don't want them
1243 		 * floating about in memory - the physical journal device may
1244 		 * hotswapped, and it breaks the `ro-after' testing code.
1245 		 */
1246 		sync_blockdev(sbi->s_journal_bdev);
1247 		invalidate_bdev(sbi->s_journal_bdev);
1248 		ext4_blkdev_remove(sbi);
1249 	}
1250 
1251 	ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1252 	sbi->s_ea_inode_cache = NULL;
1253 
1254 	ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1255 	sbi->s_ea_block_cache = NULL;
1256 
1257 	ext4_stop_mmpd(sbi);
1258 
1259 	brelse(sbi->s_sbh);
1260 	sb->s_fs_info = NULL;
1261 	/*
1262 	 * Now that we are completely done shutting down the
1263 	 * superblock, we need to actually destroy the kobject.
1264 	 */
1265 	kobject_put(&sbi->s_kobj);
1266 	wait_for_completion(&sbi->s_kobj_unregister);
1267 	if (sbi->s_chksum_driver)
1268 		crypto_free_shash(sbi->s_chksum_driver);
1269 	kfree(sbi->s_blockgroup_lock);
1270 	fs_put_dax(sbi->s_daxdev);
1271 	fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1272 #ifdef CONFIG_UNICODE
1273 	utf8_unload(sb->s_encoding);
1274 #endif
1275 	kfree(sbi);
1276 }
1277 
1278 static struct kmem_cache *ext4_inode_cachep;
1279 
1280 /*
1281  * Called inside transaction, so use GFP_NOFS
1282  */
1283 static struct inode *ext4_alloc_inode(struct super_block *sb)
1284 {
1285 	struct ext4_inode_info *ei;
1286 
1287 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1288 	if (!ei)
1289 		return NULL;
1290 
1291 	inode_set_iversion(&ei->vfs_inode, 1);
1292 	spin_lock_init(&ei->i_raw_lock);
1293 	INIT_LIST_HEAD(&ei->i_prealloc_list);
1294 	atomic_set(&ei->i_prealloc_active, 0);
1295 	spin_lock_init(&ei->i_prealloc_lock);
1296 	ext4_es_init_tree(&ei->i_es_tree);
1297 	rwlock_init(&ei->i_es_lock);
1298 	INIT_LIST_HEAD(&ei->i_es_list);
1299 	ei->i_es_all_nr = 0;
1300 	ei->i_es_shk_nr = 0;
1301 	ei->i_es_shrink_lblk = 0;
1302 	ei->i_reserved_data_blocks = 0;
1303 	spin_lock_init(&(ei->i_block_reservation_lock));
1304 	ext4_init_pending_tree(&ei->i_pending_tree);
1305 #ifdef CONFIG_QUOTA
1306 	ei->i_reserved_quota = 0;
1307 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1308 #endif
1309 	ei->jinode = NULL;
1310 	INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1311 	spin_lock_init(&ei->i_completed_io_lock);
1312 	ei->i_sync_tid = 0;
1313 	ei->i_datasync_tid = 0;
1314 	atomic_set(&ei->i_unwritten, 0);
1315 	INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1316 	ext4_fc_init_inode(&ei->vfs_inode);
1317 	mutex_init(&ei->i_fc_lock);
1318 	return &ei->vfs_inode;
1319 }
1320 
1321 static int ext4_drop_inode(struct inode *inode)
1322 {
1323 	int drop = generic_drop_inode(inode);
1324 
1325 	if (!drop)
1326 		drop = fscrypt_drop_inode(inode);
1327 
1328 	trace_ext4_drop_inode(inode, drop);
1329 	return drop;
1330 }
1331 
1332 static void ext4_free_in_core_inode(struct inode *inode)
1333 {
1334 	fscrypt_free_inode(inode);
1335 	if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1336 		pr_warn("%s: inode %ld still in fc list",
1337 			__func__, inode->i_ino);
1338 	}
1339 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1340 }
1341 
1342 static void ext4_destroy_inode(struct inode *inode)
1343 {
1344 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1345 		ext4_msg(inode->i_sb, KERN_ERR,
1346 			 "Inode %lu (%p): orphan list check failed!",
1347 			 inode->i_ino, EXT4_I(inode));
1348 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1349 				EXT4_I(inode), sizeof(struct ext4_inode_info),
1350 				true);
1351 		dump_stack();
1352 	}
1353 
1354 	if (EXT4_I(inode)->i_reserved_data_blocks)
1355 		ext4_msg(inode->i_sb, KERN_ERR,
1356 			 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1357 			 inode->i_ino, EXT4_I(inode),
1358 			 EXT4_I(inode)->i_reserved_data_blocks);
1359 }
1360 
1361 static void init_once(void *foo)
1362 {
1363 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1364 
1365 	INIT_LIST_HEAD(&ei->i_orphan);
1366 	init_rwsem(&ei->xattr_sem);
1367 	init_rwsem(&ei->i_data_sem);
1368 	inode_init_once(&ei->vfs_inode);
1369 	ext4_fc_init_inode(&ei->vfs_inode);
1370 }
1371 
1372 static int __init init_inodecache(void)
1373 {
1374 	ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1375 				sizeof(struct ext4_inode_info), 0,
1376 				(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1377 					SLAB_ACCOUNT),
1378 				offsetof(struct ext4_inode_info, i_data),
1379 				sizeof_field(struct ext4_inode_info, i_data),
1380 				init_once);
1381 	if (ext4_inode_cachep == NULL)
1382 		return -ENOMEM;
1383 	return 0;
1384 }
1385 
1386 static void destroy_inodecache(void)
1387 {
1388 	/*
1389 	 * Make sure all delayed rcu free inodes are flushed before we
1390 	 * destroy cache.
1391 	 */
1392 	rcu_barrier();
1393 	kmem_cache_destroy(ext4_inode_cachep);
1394 }
1395 
1396 void ext4_clear_inode(struct inode *inode)
1397 {
1398 	ext4_fc_del(inode);
1399 	invalidate_inode_buffers(inode);
1400 	clear_inode(inode);
1401 	ext4_discard_preallocations(inode, 0);
1402 	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1403 	dquot_drop(inode);
1404 	if (EXT4_I(inode)->jinode) {
1405 		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1406 					       EXT4_I(inode)->jinode);
1407 		jbd2_free_inode(EXT4_I(inode)->jinode);
1408 		EXT4_I(inode)->jinode = NULL;
1409 	}
1410 	fscrypt_put_encryption_info(inode);
1411 	fsverity_cleanup_inode(inode);
1412 }
1413 
1414 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1415 					u64 ino, u32 generation)
1416 {
1417 	struct inode *inode;
1418 
1419 	/*
1420 	 * Currently we don't know the generation for parent directory, so
1421 	 * a generation of 0 means "accept any"
1422 	 */
1423 	inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1424 	if (IS_ERR(inode))
1425 		return ERR_CAST(inode);
1426 	if (generation && inode->i_generation != generation) {
1427 		iput(inode);
1428 		return ERR_PTR(-ESTALE);
1429 	}
1430 
1431 	return inode;
1432 }
1433 
1434 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1435 					int fh_len, int fh_type)
1436 {
1437 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1438 				    ext4_nfs_get_inode);
1439 }
1440 
1441 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1442 					int fh_len, int fh_type)
1443 {
1444 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1445 				    ext4_nfs_get_inode);
1446 }
1447 
1448 static int ext4_nfs_commit_metadata(struct inode *inode)
1449 {
1450 	struct writeback_control wbc = {
1451 		.sync_mode = WB_SYNC_ALL
1452 	};
1453 
1454 	trace_ext4_nfs_commit_metadata(inode);
1455 	return ext4_write_inode(inode, &wbc);
1456 }
1457 
1458 #ifdef CONFIG_FS_ENCRYPTION
1459 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1460 {
1461 	return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1462 				 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1463 }
1464 
1465 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1466 							void *fs_data)
1467 {
1468 	handle_t *handle = fs_data;
1469 	int res, res2, credits, retries = 0;
1470 
1471 	/*
1472 	 * Encrypting the root directory is not allowed because e2fsck expects
1473 	 * lost+found to exist and be unencrypted, and encrypting the root
1474 	 * directory would imply encrypting the lost+found directory as well as
1475 	 * the filename "lost+found" itself.
1476 	 */
1477 	if (inode->i_ino == EXT4_ROOT_INO)
1478 		return -EPERM;
1479 
1480 	if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1481 		return -EINVAL;
1482 
1483 	if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1484 		return -EOPNOTSUPP;
1485 
1486 	res = ext4_convert_inline_data(inode);
1487 	if (res)
1488 		return res;
1489 
1490 	/*
1491 	 * If a journal handle was specified, then the encryption context is
1492 	 * being set on a new inode via inheritance and is part of a larger
1493 	 * transaction to create the inode.  Otherwise the encryption context is
1494 	 * being set on an existing inode in its own transaction.  Only in the
1495 	 * latter case should the "retry on ENOSPC" logic be used.
1496 	 */
1497 
1498 	if (handle) {
1499 		res = ext4_xattr_set_handle(handle, inode,
1500 					    EXT4_XATTR_INDEX_ENCRYPTION,
1501 					    EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1502 					    ctx, len, 0);
1503 		if (!res) {
1504 			ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1505 			ext4_clear_inode_state(inode,
1506 					EXT4_STATE_MAY_INLINE_DATA);
1507 			/*
1508 			 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1509 			 * S_DAX may be disabled
1510 			 */
1511 			ext4_set_inode_flags(inode, false);
1512 		}
1513 		return res;
1514 	}
1515 
1516 	res = dquot_initialize(inode);
1517 	if (res)
1518 		return res;
1519 retry:
1520 	res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1521 				     &credits);
1522 	if (res)
1523 		return res;
1524 
1525 	handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1526 	if (IS_ERR(handle))
1527 		return PTR_ERR(handle);
1528 
1529 	res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1530 				    EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1531 				    ctx, len, 0);
1532 	if (!res) {
1533 		ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1534 		/*
1535 		 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1536 		 * S_DAX may be disabled
1537 		 */
1538 		ext4_set_inode_flags(inode, false);
1539 		res = ext4_mark_inode_dirty(handle, inode);
1540 		if (res)
1541 			EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1542 	}
1543 	res2 = ext4_journal_stop(handle);
1544 
1545 	if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1546 		goto retry;
1547 	if (!res)
1548 		res = res2;
1549 	return res;
1550 }
1551 
1552 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1553 {
1554 	return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1555 }
1556 
1557 static bool ext4_has_stable_inodes(struct super_block *sb)
1558 {
1559 	return ext4_has_feature_stable_inodes(sb);
1560 }
1561 
1562 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1563 				       int *ino_bits_ret, int *lblk_bits_ret)
1564 {
1565 	*ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1566 	*lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1567 }
1568 
1569 static const struct fscrypt_operations ext4_cryptops = {
1570 	.key_prefix		= "ext4:",
1571 	.get_context		= ext4_get_context,
1572 	.set_context		= ext4_set_context,
1573 	.get_dummy_policy	= ext4_get_dummy_policy,
1574 	.empty_dir		= ext4_empty_dir,
1575 	.has_stable_inodes	= ext4_has_stable_inodes,
1576 	.get_ino_and_lblk_bits	= ext4_get_ino_and_lblk_bits,
1577 };
1578 #endif
1579 
1580 #ifdef CONFIG_QUOTA
1581 static const char * const quotatypes[] = INITQFNAMES;
1582 #define QTYPE2NAME(t) (quotatypes[t])
1583 
1584 static int ext4_write_dquot(struct dquot *dquot);
1585 static int ext4_acquire_dquot(struct dquot *dquot);
1586 static int ext4_release_dquot(struct dquot *dquot);
1587 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1588 static int ext4_write_info(struct super_block *sb, int type);
1589 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1590 			 const struct path *path);
1591 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1592 			       size_t len, loff_t off);
1593 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1594 				const char *data, size_t len, loff_t off);
1595 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1596 			     unsigned int flags);
1597 
1598 static struct dquot **ext4_get_dquots(struct inode *inode)
1599 {
1600 	return EXT4_I(inode)->i_dquot;
1601 }
1602 
1603 static const struct dquot_operations ext4_quota_operations = {
1604 	.get_reserved_space	= ext4_get_reserved_space,
1605 	.write_dquot		= ext4_write_dquot,
1606 	.acquire_dquot		= ext4_acquire_dquot,
1607 	.release_dquot		= ext4_release_dquot,
1608 	.mark_dirty		= ext4_mark_dquot_dirty,
1609 	.write_info		= ext4_write_info,
1610 	.alloc_dquot		= dquot_alloc,
1611 	.destroy_dquot		= dquot_destroy,
1612 	.get_projid		= ext4_get_projid,
1613 	.get_inode_usage	= ext4_get_inode_usage,
1614 	.get_next_id		= dquot_get_next_id,
1615 };
1616 
1617 static const struct quotactl_ops ext4_qctl_operations = {
1618 	.quota_on	= ext4_quota_on,
1619 	.quota_off	= ext4_quota_off,
1620 	.quota_sync	= dquot_quota_sync,
1621 	.get_state	= dquot_get_state,
1622 	.set_info	= dquot_set_dqinfo,
1623 	.get_dqblk	= dquot_get_dqblk,
1624 	.set_dqblk	= dquot_set_dqblk,
1625 	.get_nextdqblk	= dquot_get_next_dqblk,
1626 };
1627 #endif
1628 
1629 static const struct super_operations ext4_sops = {
1630 	.alloc_inode	= ext4_alloc_inode,
1631 	.free_inode	= ext4_free_in_core_inode,
1632 	.destroy_inode	= ext4_destroy_inode,
1633 	.write_inode	= ext4_write_inode,
1634 	.dirty_inode	= ext4_dirty_inode,
1635 	.drop_inode	= ext4_drop_inode,
1636 	.evict_inode	= ext4_evict_inode,
1637 	.put_super	= ext4_put_super,
1638 	.sync_fs	= ext4_sync_fs,
1639 	.freeze_fs	= ext4_freeze,
1640 	.unfreeze_fs	= ext4_unfreeze,
1641 	.statfs		= ext4_statfs,
1642 	.remount_fs	= ext4_remount,
1643 	.show_options	= ext4_show_options,
1644 #ifdef CONFIG_QUOTA
1645 	.quota_read	= ext4_quota_read,
1646 	.quota_write	= ext4_quota_write,
1647 	.get_dquots	= ext4_get_dquots,
1648 #endif
1649 };
1650 
1651 static const struct export_operations ext4_export_ops = {
1652 	.fh_to_dentry = ext4_fh_to_dentry,
1653 	.fh_to_parent = ext4_fh_to_parent,
1654 	.get_parent = ext4_get_parent,
1655 	.commit_metadata = ext4_nfs_commit_metadata,
1656 };
1657 
1658 enum {
1659 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1660 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1661 	Opt_nouid32, Opt_debug, Opt_removed,
1662 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1663 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1664 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1665 	Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1666 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1667 	Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1668 	Opt_inlinecrypt,
1669 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1670 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1671 	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1672 	Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1673 	Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1674 	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1675 	Opt_nowarn_on_error, Opt_mblk_io_submit,
1676 	Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1677 	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1678 	Opt_inode_readahead_blks, Opt_journal_ioprio,
1679 	Opt_dioread_nolock, Opt_dioread_lock,
1680 	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1681 	Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1682 	Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1683 #ifdef CONFIG_EXT4_DEBUG
1684 	Opt_fc_debug_max_replay, Opt_fc_debug_force
1685 #endif
1686 };
1687 
1688 static const match_table_t tokens = {
1689 	{Opt_bsd_df, "bsddf"},
1690 	{Opt_minix_df, "minixdf"},
1691 	{Opt_grpid, "grpid"},
1692 	{Opt_grpid, "bsdgroups"},
1693 	{Opt_nogrpid, "nogrpid"},
1694 	{Opt_nogrpid, "sysvgroups"},
1695 	{Opt_resgid, "resgid=%u"},
1696 	{Opt_resuid, "resuid=%u"},
1697 	{Opt_sb, "sb=%u"},
1698 	{Opt_err_cont, "errors=continue"},
1699 	{Opt_err_panic, "errors=panic"},
1700 	{Opt_err_ro, "errors=remount-ro"},
1701 	{Opt_nouid32, "nouid32"},
1702 	{Opt_debug, "debug"},
1703 	{Opt_removed, "oldalloc"},
1704 	{Opt_removed, "orlov"},
1705 	{Opt_user_xattr, "user_xattr"},
1706 	{Opt_nouser_xattr, "nouser_xattr"},
1707 	{Opt_acl, "acl"},
1708 	{Opt_noacl, "noacl"},
1709 	{Opt_noload, "norecovery"},
1710 	{Opt_noload, "noload"},
1711 	{Opt_removed, "nobh"},
1712 	{Opt_removed, "bh"},
1713 	{Opt_commit, "commit=%u"},
1714 	{Opt_min_batch_time, "min_batch_time=%u"},
1715 	{Opt_max_batch_time, "max_batch_time=%u"},
1716 	{Opt_journal_dev, "journal_dev=%u"},
1717 	{Opt_journal_path, "journal_path=%s"},
1718 	{Opt_journal_checksum, "journal_checksum"},
1719 	{Opt_nojournal_checksum, "nojournal_checksum"},
1720 	{Opt_journal_async_commit, "journal_async_commit"},
1721 	{Opt_abort, "abort"},
1722 	{Opt_data_journal, "data=journal"},
1723 	{Opt_data_ordered, "data=ordered"},
1724 	{Opt_data_writeback, "data=writeback"},
1725 	{Opt_data_err_abort, "data_err=abort"},
1726 	{Opt_data_err_ignore, "data_err=ignore"},
1727 	{Opt_offusrjquota, "usrjquota="},
1728 	{Opt_usrjquota, "usrjquota=%s"},
1729 	{Opt_offgrpjquota, "grpjquota="},
1730 	{Opt_grpjquota, "grpjquota=%s"},
1731 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1732 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1733 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1734 	{Opt_grpquota, "grpquota"},
1735 	{Opt_noquota, "noquota"},
1736 	{Opt_quota, "quota"},
1737 	{Opt_usrquota, "usrquota"},
1738 	{Opt_prjquota, "prjquota"},
1739 	{Opt_barrier, "barrier=%u"},
1740 	{Opt_barrier, "barrier"},
1741 	{Opt_nobarrier, "nobarrier"},
1742 	{Opt_i_version, "i_version"},
1743 	{Opt_dax, "dax"},
1744 	{Opt_dax_always, "dax=always"},
1745 	{Opt_dax_inode, "dax=inode"},
1746 	{Opt_dax_never, "dax=never"},
1747 	{Opt_stripe, "stripe=%u"},
1748 	{Opt_delalloc, "delalloc"},
1749 	{Opt_warn_on_error, "warn_on_error"},
1750 	{Opt_nowarn_on_error, "nowarn_on_error"},
1751 	{Opt_lazytime, "lazytime"},
1752 	{Opt_nolazytime, "nolazytime"},
1753 	{Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1754 	{Opt_nodelalloc, "nodelalloc"},
1755 	{Opt_removed, "mblk_io_submit"},
1756 	{Opt_removed, "nomblk_io_submit"},
1757 	{Opt_block_validity, "block_validity"},
1758 	{Opt_noblock_validity, "noblock_validity"},
1759 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1760 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1761 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1762 	{Opt_auto_da_alloc, "auto_da_alloc"},
1763 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1764 	{Opt_dioread_nolock, "dioread_nolock"},
1765 	{Opt_dioread_lock, "nodioread_nolock"},
1766 	{Opt_dioread_lock, "dioread_lock"},
1767 	{Opt_discard, "discard"},
1768 	{Opt_nodiscard, "nodiscard"},
1769 	{Opt_init_itable, "init_itable=%u"},
1770 	{Opt_init_itable, "init_itable"},
1771 	{Opt_noinit_itable, "noinit_itable"},
1772 #ifdef CONFIG_EXT4_DEBUG
1773 	{Opt_fc_debug_force, "fc_debug_force"},
1774 	{Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1775 #endif
1776 	{Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1777 	{Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1778 	{Opt_test_dummy_encryption, "test_dummy_encryption"},
1779 	{Opt_inlinecrypt, "inlinecrypt"},
1780 	{Opt_nombcache, "nombcache"},
1781 	{Opt_nombcache, "no_mbcache"},	/* for backward compatibility */
1782 	{Opt_removed, "prefetch_block_bitmaps"},
1783 	{Opt_no_prefetch_block_bitmaps, "no_prefetch_block_bitmaps"},
1784 	{Opt_mb_optimize_scan, "mb_optimize_scan=%d"},
1785 	{Opt_removed, "check=none"},	/* mount option from ext2/3 */
1786 	{Opt_removed, "nocheck"},	/* mount option from ext2/3 */
1787 	{Opt_removed, "reservation"},	/* mount option from ext2/3 */
1788 	{Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1789 	{Opt_removed, "journal=%u"},	/* mount option from ext2/3 */
1790 	{Opt_err, NULL},
1791 };
1792 
1793 static ext4_fsblk_t get_sb_block(void **data)
1794 {
1795 	ext4_fsblk_t	sb_block;
1796 	char		*options = (char *) *data;
1797 
1798 	if (!options || strncmp(options, "sb=", 3) != 0)
1799 		return 1;	/* Default location */
1800 
1801 	options += 3;
1802 	/* TODO: use simple_strtoll with >32bit ext4 */
1803 	sb_block = simple_strtoul(options, &options, 0);
1804 	if (*options && *options != ',') {
1805 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1806 		       (char *) *data);
1807 		return 1;
1808 	}
1809 	if (*options == ',')
1810 		options++;
1811 	*data = (void *) options;
1812 
1813 	return sb_block;
1814 }
1815 
1816 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1817 #define DEFAULT_MB_OPTIMIZE_SCAN	(-1)
1818 
1819 static const char deprecated_msg[] =
1820 	"Mount option \"%s\" will be removed by %s\n"
1821 	"Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1822 
1823 #ifdef CONFIG_QUOTA
1824 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1825 {
1826 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1827 	char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1828 	int ret = -1;
1829 
1830 	if (sb_any_quota_loaded(sb) && !old_qname) {
1831 		ext4_msg(sb, KERN_ERR,
1832 			"Cannot change journaled "
1833 			"quota options when quota turned on");
1834 		return -1;
1835 	}
1836 	if (ext4_has_feature_quota(sb)) {
1837 		ext4_msg(sb, KERN_INFO, "Journaled quota options "
1838 			 "ignored when QUOTA feature is enabled");
1839 		return 1;
1840 	}
1841 	qname = match_strdup(args);
1842 	if (!qname) {
1843 		ext4_msg(sb, KERN_ERR,
1844 			"Not enough memory for storing quotafile name");
1845 		return -1;
1846 	}
1847 	if (old_qname) {
1848 		if (strcmp(old_qname, qname) == 0)
1849 			ret = 1;
1850 		else
1851 			ext4_msg(sb, KERN_ERR,
1852 				 "%s quota file already specified",
1853 				 QTYPE2NAME(qtype));
1854 		goto errout;
1855 	}
1856 	if (strchr(qname, '/')) {
1857 		ext4_msg(sb, KERN_ERR,
1858 			"quotafile must be on filesystem root");
1859 		goto errout;
1860 	}
1861 	rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1862 	set_opt(sb, QUOTA);
1863 	return 1;
1864 errout:
1865 	kfree(qname);
1866 	return ret;
1867 }
1868 
1869 static int clear_qf_name(struct super_block *sb, int qtype)
1870 {
1871 
1872 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1873 	char *old_qname = get_qf_name(sb, sbi, qtype);
1874 
1875 	if (sb_any_quota_loaded(sb) && old_qname) {
1876 		ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1877 			" when quota turned on");
1878 		return -1;
1879 	}
1880 	rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1881 	synchronize_rcu();
1882 	kfree(old_qname);
1883 	return 1;
1884 }
1885 #endif
1886 
1887 #define MOPT_SET	0x0001
1888 #define MOPT_CLEAR	0x0002
1889 #define MOPT_NOSUPPORT	0x0004
1890 #define MOPT_EXPLICIT	0x0008
1891 #define MOPT_CLEAR_ERR	0x0010
1892 #define MOPT_GTE0	0x0020
1893 #ifdef CONFIG_QUOTA
1894 #define MOPT_Q		0
1895 #define MOPT_QFMT	0x0040
1896 #else
1897 #define MOPT_Q		MOPT_NOSUPPORT
1898 #define MOPT_QFMT	MOPT_NOSUPPORT
1899 #endif
1900 #define MOPT_DATAJ	0x0080
1901 #define MOPT_NO_EXT2	0x0100
1902 #define MOPT_NO_EXT3	0x0200
1903 #define MOPT_EXT4_ONLY	(MOPT_NO_EXT2 | MOPT_NO_EXT3)
1904 #define MOPT_STRING	0x0400
1905 #define MOPT_SKIP	0x0800
1906 #define	MOPT_2		0x1000
1907 
1908 static const struct mount_opts {
1909 	int	token;
1910 	int	mount_opt;
1911 	int	flags;
1912 } ext4_mount_opts[] = {
1913 	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1914 	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1915 	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1916 	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1917 	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1918 	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1919 	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1920 	 MOPT_EXT4_ONLY | MOPT_SET},
1921 	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1922 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1923 	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1924 	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1925 	{Opt_delalloc, EXT4_MOUNT_DELALLOC,
1926 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1927 	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1928 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1929 	{Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1930 	{Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1931 	{Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1932 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1933 	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1934 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1935 	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1936 				    EXT4_MOUNT_JOURNAL_CHECKSUM),
1937 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1938 	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1939 	{Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1940 	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1941 	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1942 	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1943 	 MOPT_NO_EXT2},
1944 	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1945 	 MOPT_NO_EXT2},
1946 	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1947 	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1948 	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1949 	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1950 	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1951 	{Opt_commit, 0, MOPT_GTE0},
1952 	{Opt_max_batch_time, 0, MOPT_GTE0},
1953 	{Opt_min_batch_time, 0, MOPT_GTE0},
1954 	{Opt_inode_readahead_blks, 0, MOPT_GTE0},
1955 	{Opt_init_itable, 0, MOPT_GTE0},
1956 	{Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1957 	{Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1958 		MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1959 	{Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1960 		MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1961 	{Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1962 		MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1963 	{Opt_stripe, 0, MOPT_GTE0},
1964 	{Opt_resuid, 0, MOPT_GTE0},
1965 	{Opt_resgid, 0, MOPT_GTE0},
1966 	{Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1967 	{Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1968 	{Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1969 	{Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1970 	{Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1971 	{Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1972 	 MOPT_NO_EXT2 | MOPT_DATAJ},
1973 	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1974 	{Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1975 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1976 	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1977 	{Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1978 #else
1979 	{Opt_acl, 0, MOPT_NOSUPPORT},
1980 	{Opt_noacl, 0, MOPT_NOSUPPORT},
1981 #endif
1982 	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1983 	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1984 	{Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1985 	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1986 	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1987 							MOPT_SET | MOPT_Q},
1988 	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1989 							MOPT_SET | MOPT_Q},
1990 	{Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1991 							MOPT_SET | MOPT_Q},
1992 	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1993 		       EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1994 							MOPT_CLEAR | MOPT_Q},
1995 	{Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1996 	{Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1997 	{Opt_offusrjquota, 0, MOPT_Q},
1998 	{Opt_offgrpjquota, 0, MOPT_Q},
1999 	{Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
2000 	{Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
2001 	{Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
2002 	{Opt_max_dir_size_kb, 0, MOPT_GTE0},
2003 	{Opt_test_dummy_encryption, 0, MOPT_STRING},
2004 	{Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
2005 	{Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
2006 	 MOPT_SET},
2007 	{Opt_mb_optimize_scan, EXT4_MOUNT2_MB_OPTIMIZE_SCAN, MOPT_GTE0},
2008 #ifdef CONFIG_EXT4_DEBUG
2009 	{Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2010 	 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
2011 	{Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2012 #endif
2013 	{Opt_err, 0, 0}
2014 };
2015 
2016 #ifdef CONFIG_UNICODE
2017 static const struct ext4_sb_encodings {
2018 	__u16 magic;
2019 	char *name;
2020 	char *version;
2021 } ext4_sb_encoding_map[] = {
2022 	{EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2023 };
2024 
2025 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2026 				 const struct ext4_sb_encodings **encoding,
2027 				 __u16 *flags)
2028 {
2029 	__u16 magic = le16_to_cpu(es->s_encoding);
2030 	int i;
2031 
2032 	for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2033 		if (magic == ext4_sb_encoding_map[i].magic)
2034 			break;
2035 
2036 	if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2037 		return -EINVAL;
2038 
2039 	*encoding = &ext4_sb_encoding_map[i];
2040 	*flags = le16_to_cpu(es->s_encoding_flags);
2041 
2042 	return 0;
2043 }
2044 #endif
2045 
2046 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2047 					  const char *opt,
2048 					  const substring_t *arg,
2049 					  bool is_remount)
2050 {
2051 #ifdef CONFIG_FS_ENCRYPTION
2052 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2053 	int err;
2054 
2055 	/*
2056 	 * This mount option is just for testing, and it's not worthwhile to
2057 	 * implement the extra complexity (e.g. RCU protection) that would be
2058 	 * needed to allow it to be set or changed during remount.  We do allow
2059 	 * it to be specified during remount, but only if there is no change.
2060 	 */
2061 	if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2062 		ext4_msg(sb, KERN_WARNING,
2063 			 "Can't set test_dummy_encryption on remount");
2064 		return -1;
2065 	}
2066 	err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2067 						&sbi->s_dummy_enc_policy);
2068 	if (err) {
2069 		if (err == -EEXIST)
2070 			ext4_msg(sb, KERN_WARNING,
2071 				 "Can't change test_dummy_encryption on remount");
2072 		else if (err == -EINVAL)
2073 			ext4_msg(sb, KERN_WARNING,
2074 				 "Value of option \"%s\" is unrecognized", opt);
2075 		else
2076 			ext4_msg(sb, KERN_WARNING,
2077 				 "Error processing option \"%s\" [%d]",
2078 				 opt, err);
2079 		return -1;
2080 	}
2081 	ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2082 #else
2083 	ext4_msg(sb, KERN_WARNING,
2084 		 "Test dummy encryption mount option ignored");
2085 #endif
2086 	return 1;
2087 }
2088 
2089 struct ext4_parsed_options {
2090 	unsigned long journal_devnum;
2091 	unsigned int journal_ioprio;
2092 	int mb_optimize_scan;
2093 };
2094 
2095 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2096 			    substring_t *args, struct ext4_parsed_options *parsed_opts,
2097 			    int is_remount)
2098 {
2099 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2100 	const struct mount_opts *m;
2101 	kuid_t uid;
2102 	kgid_t gid;
2103 	int arg = 0;
2104 
2105 #ifdef CONFIG_QUOTA
2106 	if (token == Opt_usrjquota)
2107 		return set_qf_name(sb, USRQUOTA, &args[0]);
2108 	else if (token == Opt_grpjquota)
2109 		return set_qf_name(sb, GRPQUOTA, &args[0]);
2110 	else if (token == Opt_offusrjquota)
2111 		return clear_qf_name(sb, USRQUOTA);
2112 	else if (token == Opt_offgrpjquota)
2113 		return clear_qf_name(sb, GRPQUOTA);
2114 #endif
2115 	switch (token) {
2116 	case Opt_noacl:
2117 	case Opt_nouser_xattr:
2118 		ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2119 		break;
2120 	case Opt_sb:
2121 		return 1;	/* handled by get_sb_block() */
2122 	case Opt_removed:
2123 		ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2124 		return 1;
2125 	case Opt_abort:
2126 		ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
2127 		return 1;
2128 	case Opt_i_version:
2129 		sb->s_flags |= SB_I_VERSION;
2130 		return 1;
2131 	case Opt_lazytime:
2132 		sb->s_flags |= SB_LAZYTIME;
2133 		return 1;
2134 	case Opt_nolazytime:
2135 		sb->s_flags &= ~SB_LAZYTIME;
2136 		return 1;
2137 	case Opt_inlinecrypt:
2138 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2139 		sb->s_flags |= SB_INLINECRYPT;
2140 #else
2141 		ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2142 #endif
2143 		return 1;
2144 	}
2145 
2146 	for (m = ext4_mount_opts; m->token != Opt_err; m++)
2147 		if (token == m->token)
2148 			break;
2149 
2150 	if (m->token == Opt_err) {
2151 		ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2152 			 "or missing value", opt);
2153 		return -1;
2154 	}
2155 
2156 	if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2157 		ext4_msg(sb, KERN_ERR,
2158 			 "Mount option \"%s\" incompatible with ext2", opt);
2159 		return -1;
2160 	}
2161 	if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2162 		ext4_msg(sb, KERN_ERR,
2163 			 "Mount option \"%s\" incompatible with ext3", opt);
2164 		return -1;
2165 	}
2166 
2167 	if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2168 		return -1;
2169 	if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2170 		return -1;
2171 	if (m->flags & MOPT_EXPLICIT) {
2172 		if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2173 			set_opt2(sb, EXPLICIT_DELALLOC);
2174 		} else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2175 			set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2176 		} else
2177 			return -1;
2178 	}
2179 	if (m->flags & MOPT_CLEAR_ERR)
2180 		clear_opt(sb, ERRORS_MASK);
2181 	if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2182 		ext4_msg(sb, KERN_ERR, "Cannot change quota "
2183 			 "options when quota turned on");
2184 		return -1;
2185 	}
2186 
2187 	if (m->flags & MOPT_NOSUPPORT) {
2188 		ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2189 	} else if (token == Opt_commit) {
2190 		if (arg == 0)
2191 			arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2192 		else if (arg > INT_MAX / HZ) {
2193 			ext4_msg(sb, KERN_ERR,
2194 				 "Invalid commit interval %d, "
2195 				 "must be smaller than %d",
2196 				 arg, INT_MAX / HZ);
2197 			return -1;
2198 		}
2199 		sbi->s_commit_interval = HZ * arg;
2200 	} else if (token == Opt_debug_want_extra_isize) {
2201 		if ((arg & 1) ||
2202 		    (arg < 4) ||
2203 		    (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2204 			ext4_msg(sb, KERN_ERR,
2205 				 "Invalid want_extra_isize %d", arg);
2206 			return -1;
2207 		}
2208 		sbi->s_want_extra_isize = arg;
2209 	} else if (token == Opt_max_batch_time) {
2210 		sbi->s_max_batch_time = arg;
2211 	} else if (token == Opt_min_batch_time) {
2212 		sbi->s_min_batch_time = arg;
2213 	} else if (token == Opt_inode_readahead_blks) {
2214 		if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2215 			ext4_msg(sb, KERN_ERR,
2216 				 "EXT4-fs: inode_readahead_blks must be "
2217 				 "0 or a power of 2 smaller than 2^31");
2218 			return -1;
2219 		}
2220 		sbi->s_inode_readahead_blks = arg;
2221 	} else if (token == Opt_init_itable) {
2222 		set_opt(sb, INIT_INODE_TABLE);
2223 		if (!args->from)
2224 			arg = EXT4_DEF_LI_WAIT_MULT;
2225 		sbi->s_li_wait_mult = arg;
2226 	} else if (token == Opt_max_dir_size_kb) {
2227 		sbi->s_max_dir_size_kb = arg;
2228 #ifdef CONFIG_EXT4_DEBUG
2229 	} else if (token == Opt_fc_debug_max_replay) {
2230 		sbi->s_fc_debug_max_replay = arg;
2231 #endif
2232 	} else if (token == Opt_stripe) {
2233 		sbi->s_stripe = arg;
2234 	} else if (token == Opt_resuid) {
2235 		uid = make_kuid(current_user_ns(), arg);
2236 		if (!uid_valid(uid)) {
2237 			ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2238 			return -1;
2239 		}
2240 		sbi->s_resuid = uid;
2241 	} else if (token == Opt_resgid) {
2242 		gid = make_kgid(current_user_ns(), arg);
2243 		if (!gid_valid(gid)) {
2244 			ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2245 			return -1;
2246 		}
2247 		sbi->s_resgid = gid;
2248 	} else if (token == Opt_journal_dev) {
2249 		if (is_remount) {
2250 			ext4_msg(sb, KERN_ERR,
2251 				 "Cannot specify journal on remount");
2252 			return -1;
2253 		}
2254 		parsed_opts->journal_devnum = arg;
2255 	} else if (token == Opt_journal_path) {
2256 		char *journal_path;
2257 		struct inode *journal_inode;
2258 		struct path path;
2259 		int error;
2260 
2261 		if (is_remount) {
2262 			ext4_msg(sb, KERN_ERR,
2263 				 "Cannot specify journal on remount");
2264 			return -1;
2265 		}
2266 		journal_path = match_strdup(&args[0]);
2267 		if (!journal_path) {
2268 			ext4_msg(sb, KERN_ERR, "error: could not dup "
2269 				"journal device string");
2270 			return -1;
2271 		}
2272 
2273 		error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2274 		if (error) {
2275 			ext4_msg(sb, KERN_ERR, "error: could not find "
2276 				"journal device path: error %d", error);
2277 			kfree(journal_path);
2278 			return -1;
2279 		}
2280 
2281 		journal_inode = d_inode(path.dentry);
2282 		if (!S_ISBLK(journal_inode->i_mode)) {
2283 			ext4_msg(sb, KERN_ERR, "error: journal path %s "
2284 				"is not a block device", journal_path);
2285 			path_put(&path);
2286 			kfree(journal_path);
2287 			return -1;
2288 		}
2289 
2290 		parsed_opts->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2291 		path_put(&path);
2292 		kfree(journal_path);
2293 	} else if (token == Opt_journal_ioprio) {
2294 		if (arg > 7) {
2295 			ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2296 				 " (must be 0-7)");
2297 			return -1;
2298 		}
2299 		parsed_opts->journal_ioprio =
2300 			IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2301 	} else if (token == Opt_test_dummy_encryption) {
2302 		return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2303 						      is_remount);
2304 	} else if (m->flags & MOPT_DATAJ) {
2305 		if (is_remount) {
2306 			if (!sbi->s_journal)
2307 				ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2308 			else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2309 				ext4_msg(sb, KERN_ERR,
2310 					 "Cannot change data mode on remount");
2311 				return -1;
2312 			}
2313 		} else {
2314 			clear_opt(sb, DATA_FLAGS);
2315 			sbi->s_mount_opt |= m->mount_opt;
2316 		}
2317 #ifdef CONFIG_QUOTA
2318 	} else if (m->flags & MOPT_QFMT) {
2319 		if (sb_any_quota_loaded(sb) &&
2320 		    sbi->s_jquota_fmt != m->mount_opt) {
2321 			ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2322 				 "quota options when quota turned on");
2323 			return -1;
2324 		}
2325 		if (ext4_has_feature_quota(sb)) {
2326 			ext4_msg(sb, KERN_INFO,
2327 				 "Quota format mount options ignored "
2328 				 "when QUOTA feature is enabled");
2329 			return 1;
2330 		}
2331 		sbi->s_jquota_fmt = m->mount_opt;
2332 #endif
2333 	} else if (token == Opt_dax || token == Opt_dax_always ||
2334 		   token == Opt_dax_inode || token == Opt_dax_never) {
2335 #ifdef CONFIG_FS_DAX
2336 		switch (token) {
2337 		case Opt_dax:
2338 		case Opt_dax_always:
2339 			if (is_remount &&
2340 			    (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2341 			     (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2342 			fail_dax_change_remount:
2343 				ext4_msg(sb, KERN_ERR, "can't change "
2344 					 "dax mount option while remounting");
2345 				return -1;
2346 			}
2347 			if (is_remount &&
2348 			    (test_opt(sb, DATA_FLAGS) ==
2349 			     EXT4_MOUNT_JOURNAL_DATA)) {
2350 				    ext4_msg(sb, KERN_ERR, "can't mount with "
2351 					     "both data=journal and dax");
2352 				    return -1;
2353 			}
2354 			ext4_msg(sb, KERN_WARNING,
2355 				"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2356 			sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2357 			sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2358 			break;
2359 		case Opt_dax_never:
2360 			if (is_remount &&
2361 			    (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2362 			     (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2363 				goto fail_dax_change_remount;
2364 			sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2365 			sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2366 			break;
2367 		case Opt_dax_inode:
2368 			if (is_remount &&
2369 			    ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2370 			     (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2371 			     !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2372 				goto fail_dax_change_remount;
2373 			sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2374 			sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2375 			/* Strictly for printing options */
2376 			sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2377 			break;
2378 		}
2379 #else
2380 		ext4_msg(sb, KERN_INFO, "dax option not supported");
2381 		sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2382 		sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2383 		return -1;
2384 #endif
2385 	} else if (token == Opt_data_err_abort) {
2386 		sbi->s_mount_opt |= m->mount_opt;
2387 	} else if (token == Opt_data_err_ignore) {
2388 		sbi->s_mount_opt &= ~m->mount_opt;
2389 	} else if (token == Opt_mb_optimize_scan) {
2390 		if (arg != 0 && arg != 1) {
2391 			ext4_msg(sb, KERN_WARNING,
2392 				 "mb_optimize_scan should be set to 0 or 1.");
2393 			return -1;
2394 		}
2395 		parsed_opts->mb_optimize_scan = arg;
2396 	} else {
2397 		if (!args->from)
2398 			arg = 1;
2399 		if (m->flags & MOPT_CLEAR)
2400 			arg = !arg;
2401 		else if (unlikely(!(m->flags & MOPT_SET))) {
2402 			ext4_msg(sb, KERN_WARNING,
2403 				 "buggy handling of option %s", opt);
2404 			WARN_ON(1);
2405 			return -1;
2406 		}
2407 		if (m->flags & MOPT_2) {
2408 			if (arg != 0)
2409 				sbi->s_mount_opt2 |= m->mount_opt;
2410 			else
2411 				sbi->s_mount_opt2 &= ~m->mount_opt;
2412 		} else {
2413 			if (arg != 0)
2414 				sbi->s_mount_opt |= m->mount_opt;
2415 			else
2416 				sbi->s_mount_opt &= ~m->mount_opt;
2417 		}
2418 	}
2419 	return 1;
2420 }
2421 
2422 static int parse_options(char *options, struct super_block *sb,
2423 			 struct ext4_parsed_options *ret_opts,
2424 			 int is_remount)
2425 {
2426 	struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2427 	char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2428 	substring_t args[MAX_OPT_ARGS];
2429 	int token;
2430 
2431 	if (!options)
2432 		return 1;
2433 
2434 	while ((p = strsep(&options, ",")) != NULL) {
2435 		if (!*p)
2436 			continue;
2437 		/*
2438 		 * Initialize args struct so we know whether arg was
2439 		 * found; some options take optional arguments.
2440 		 */
2441 		args[0].to = args[0].from = NULL;
2442 		token = match_token(p, tokens, args);
2443 		if (handle_mount_opt(sb, p, token, args, ret_opts,
2444 				     is_remount) < 0)
2445 			return 0;
2446 	}
2447 #ifdef CONFIG_QUOTA
2448 	/*
2449 	 * We do the test below only for project quotas. 'usrquota' and
2450 	 * 'grpquota' mount options are allowed even without quota feature
2451 	 * to support legacy quotas in quota files.
2452 	 */
2453 	if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2454 		ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2455 			 "Cannot enable project quota enforcement.");
2456 		return 0;
2457 	}
2458 	usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2459 	grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2460 	if (usr_qf_name || grp_qf_name) {
2461 		if (test_opt(sb, USRQUOTA) && usr_qf_name)
2462 			clear_opt(sb, USRQUOTA);
2463 
2464 		if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2465 			clear_opt(sb, GRPQUOTA);
2466 
2467 		if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2468 			ext4_msg(sb, KERN_ERR, "old and new quota "
2469 					"format mixing");
2470 			return 0;
2471 		}
2472 
2473 		if (!sbi->s_jquota_fmt) {
2474 			ext4_msg(sb, KERN_ERR, "journaled quota format "
2475 					"not specified");
2476 			return 0;
2477 		}
2478 	}
2479 #endif
2480 	if (test_opt(sb, DIOREAD_NOLOCK)) {
2481 		int blocksize =
2482 			BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2483 		if (blocksize < PAGE_SIZE)
2484 			ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2485 				 "experimental mount option 'dioread_nolock' "
2486 				 "for blocksize < PAGE_SIZE");
2487 	}
2488 	return 1;
2489 }
2490 
2491 static inline void ext4_show_quota_options(struct seq_file *seq,
2492 					   struct super_block *sb)
2493 {
2494 #if defined(CONFIG_QUOTA)
2495 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2496 	char *usr_qf_name, *grp_qf_name;
2497 
2498 	if (sbi->s_jquota_fmt) {
2499 		char *fmtname = "";
2500 
2501 		switch (sbi->s_jquota_fmt) {
2502 		case QFMT_VFS_OLD:
2503 			fmtname = "vfsold";
2504 			break;
2505 		case QFMT_VFS_V0:
2506 			fmtname = "vfsv0";
2507 			break;
2508 		case QFMT_VFS_V1:
2509 			fmtname = "vfsv1";
2510 			break;
2511 		}
2512 		seq_printf(seq, ",jqfmt=%s", fmtname);
2513 	}
2514 
2515 	rcu_read_lock();
2516 	usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2517 	grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2518 	if (usr_qf_name)
2519 		seq_show_option(seq, "usrjquota", usr_qf_name);
2520 	if (grp_qf_name)
2521 		seq_show_option(seq, "grpjquota", grp_qf_name);
2522 	rcu_read_unlock();
2523 #endif
2524 }
2525 
2526 static const char *token2str(int token)
2527 {
2528 	const struct match_token *t;
2529 
2530 	for (t = tokens; t->token != Opt_err; t++)
2531 		if (t->token == token && !strchr(t->pattern, '='))
2532 			break;
2533 	return t->pattern;
2534 }
2535 
2536 /*
2537  * Show an option if
2538  *  - it's set to a non-default value OR
2539  *  - if the per-sb default is different from the global default
2540  */
2541 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2542 			      int nodefs)
2543 {
2544 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2545 	struct ext4_super_block *es = sbi->s_es;
2546 	int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2547 	const struct mount_opts *m;
2548 	char sep = nodefs ? '\n' : ',';
2549 
2550 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2551 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2552 
2553 	if (sbi->s_sb_block != 1)
2554 		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2555 
2556 	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2557 		int want_set = m->flags & MOPT_SET;
2558 		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2559 		    (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2560 			continue;
2561 		if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2562 			continue; /* skip if same as the default */
2563 		if ((want_set &&
2564 		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2565 		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2566 			continue; /* select Opt_noFoo vs Opt_Foo */
2567 		SEQ_OPTS_PRINT("%s", token2str(m->token));
2568 	}
2569 
2570 	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2571 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2572 		SEQ_OPTS_PRINT("resuid=%u",
2573 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
2574 	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2575 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2576 		SEQ_OPTS_PRINT("resgid=%u",
2577 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
2578 	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2579 	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2580 		SEQ_OPTS_PUTS("errors=remount-ro");
2581 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2582 		SEQ_OPTS_PUTS("errors=continue");
2583 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2584 		SEQ_OPTS_PUTS("errors=panic");
2585 	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2586 		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2587 	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2588 		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2589 	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2590 		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2591 	if (sb->s_flags & SB_I_VERSION)
2592 		SEQ_OPTS_PUTS("i_version");
2593 	if (nodefs || sbi->s_stripe)
2594 		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2595 	if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2596 			(sbi->s_mount_opt ^ def_mount_opt)) {
2597 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2598 			SEQ_OPTS_PUTS("data=journal");
2599 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2600 			SEQ_OPTS_PUTS("data=ordered");
2601 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2602 			SEQ_OPTS_PUTS("data=writeback");
2603 	}
2604 	if (nodefs ||
2605 	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2606 		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2607 			       sbi->s_inode_readahead_blks);
2608 
2609 	if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2610 		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2611 		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2612 	if (nodefs || sbi->s_max_dir_size_kb)
2613 		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2614 	if (test_opt(sb, DATA_ERR_ABORT))
2615 		SEQ_OPTS_PUTS("data_err=abort");
2616 
2617 	fscrypt_show_test_dummy_encryption(seq, sep, sb);
2618 
2619 	if (sb->s_flags & SB_INLINECRYPT)
2620 		SEQ_OPTS_PUTS("inlinecrypt");
2621 
2622 	if (test_opt(sb, DAX_ALWAYS)) {
2623 		if (IS_EXT2_SB(sb))
2624 			SEQ_OPTS_PUTS("dax");
2625 		else
2626 			SEQ_OPTS_PUTS("dax=always");
2627 	} else if (test_opt2(sb, DAX_NEVER)) {
2628 		SEQ_OPTS_PUTS("dax=never");
2629 	} else if (test_opt2(sb, DAX_INODE)) {
2630 		SEQ_OPTS_PUTS("dax=inode");
2631 	}
2632 	ext4_show_quota_options(seq, sb);
2633 	return 0;
2634 }
2635 
2636 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2637 {
2638 	return _ext4_show_options(seq, root->d_sb, 0);
2639 }
2640 
2641 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2642 {
2643 	struct super_block *sb = seq->private;
2644 	int rc;
2645 
2646 	seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2647 	rc = _ext4_show_options(seq, sb, 1);
2648 	seq_puts(seq, "\n");
2649 	return rc;
2650 }
2651 
2652 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2653 			    int read_only)
2654 {
2655 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2656 	int err = 0;
2657 
2658 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2659 		ext4_msg(sb, KERN_ERR, "revision level too high, "
2660 			 "forcing read-only mode");
2661 		err = -EROFS;
2662 		goto done;
2663 	}
2664 	if (read_only)
2665 		goto done;
2666 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
2667 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2668 			 "running e2fsck is recommended");
2669 	else if (sbi->s_mount_state & EXT4_ERROR_FS)
2670 		ext4_msg(sb, KERN_WARNING,
2671 			 "warning: mounting fs with errors, "
2672 			 "running e2fsck is recommended");
2673 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2674 		 le16_to_cpu(es->s_mnt_count) >=
2675 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2676 		ext4_msg(sb, KERN_WARNING,
2677 			 "warning: maximal mount count reached, "
2678 			 "running e2fsck is recommended");
2679 	else if (le32_to_cpu(es->s_checkinterval) &&
2680 		 (ext4_get_tstamp(es, s_lastcheck) +
2681 		  le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2682 		ext4_msg(sb, KERN_WARNING,
2683 			 "warning: checktime reached, "
2684 			 "running e2fsck is recommended");
2685 	if (!sbi->s_journal)
2686 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2687 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2688 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2689 	le16_add_cpu(&es->s_mnt_count, 1);
2690 	ext4_update_tstamp(es, s_mtime);
2691 	if (sbi->s_journal) {
2692 		ext4_set_feature_journal_needs_recovery(sb);
2693 		if (ext4_has_feature_orphan_file(sb))
2694 			ext4_set_feature_orphan_present(sb);
2695 	}
2696 
2697 	err = ext4_commit_super(sb);
2698 done:
2699 	if (test_opt(sb, DEBUG))
2700 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2701 				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2702 			sb->s_blocksize,
2703 			sbi->s_groups_count,
2704 			EXT4_BLOCKS_PER_GROUP(sb),
2705 			EXT4_INODES_PER_GROUP(sb),
2706 			sbi->s_mount_opt, sbi->s_mount_opt2);
2707 
2708 	cleancache_init_fs(sb);
2709 	return err;
2710 }
2711 
2712 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2713 {
2714 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2715 	struct flex_groups **old_groups, **new_groups;
2716 	int size, i, j;
2717 
2718 	if (!sbi->s_log_groups_per_flex)
2719 		return 0;
2720 
2721 	size = ext4_flex_group(sbi, ngroup - 1) + 1;
2722 	if (size <= sbi->s_flex_groups_allocated)
2723 		return 0;
2724 
2725 	new_groups = kvzalloc(roundup_pow_of_two(size *
2726 			      sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2727 	if (!new_groups) {
2728 		ext4_msg(sb, KERN_ERR,
2729 			 "not enough memory for %d flex group pointers", size);
2730 		return -ENOMEM;
2731 	}
2732 	for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2733 		new_groups[i] = kvzalloc(roundup_pow_of_two(
2734 					 sizeof(struct flex_groups)),
2735 					 GFP_KERNEL);
2736 		if (!new_groups[i]) {
2737 			for (j = sbi->s_flex_groups_allocated; j < i; j++)
2738 				kvfree(new_groups[j]);
2739 			kvfree(new_groups);
2740 			ext4_msg(sb, KERN_ERR,
2741 				 "not enough memory for %d flex groups", size);
2742 			return -ENOMEM;
2743 		}
2744 	}
2745 	rcu_read_lock();
2746 	old_groups = rcu_dereference(sbi->s_flex_groups);
2747 	if (old_groups)
2748 		memcpy(new_groups, old_groups,
2749 		       (sbi->s_flex_groups_allocated *
2750 			sizeof(struct flex_groups *)));
2751 	rcu_read_unlock();
2752 	rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2753 	sbi->s_flex_groups_allocated = size;
2754 	if (old_groups)
2755 		ext4_kvfree_array_rcu(old_groups);
2756 	return 0;
2757 }
2758 
2759 static int ext4_fill_flex_info(struct super_block *sb)
2760 {
2761 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2762 	struct ext4_group_desc *gdp = NULL;
2763 	struct flex_groups *fg;
2764 	ext4_group_t flex_group;
2765 	int i, err;
2766 
2767 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2768 	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2769 		sbi->s_log_groups_per_flex = 0;
2770 		return 1;
2771 	}
2772 
2773 	err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2774 	if (err)
2775 		goto failed;
2776 
2777 	for (i = 0; i < sbi->s_groups_count; i++) {
2778 		gdp = ext4_get_group_desc(sb, i, NULL);
2779 
2780 		flex_group = ext4_flex_group(sbi, i);
2781 		fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2782 		atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2783 		atomic64_add(ext4_free_group_clusters(sb, gdp),
2784 			     &fg->free_clusters);
2785 		atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2786 	}
2787 
2788 	return 1;
2789 failed:
2790 	return 0;
2791 }
2792 
2793 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2794 				   struct ext4_group_desc *gdp)
2795 {
2796 	int offset = offsetof(struct ext4_group_desc, bg_checksum);
2797 	__u16 crc = 0;
2798 	__le32 le_group = cpu_to_le32(block_group);
2799 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2800 
2801 	if (ext4_has_metadata_csum(sbi->s_sb)) {
2802 		/* Use new metadata_csum algorithm */
2803 		__u32 csum32;
2804 		__u16 dummy_csum = 0;
2805 
2806 		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2807 				     sizeof(le_group));
2808 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2809 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2810 				     sizeof(dummy_csum));
2811 		offset += sizeof(dummy_csum);
2812 		if (offset < sbi->s_desc_size)
2813 			csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2814 					     sbi->s_desc_size - offset);
2815 
2816 		crc = csum32 & 0xFFFF;
2817 		goto out;
2818 	}
2819 
2820 	/* old crc16 code */
2821 	if (!ext4_has_feature_gdt_csum(sb))
2822 		return 0;
2823 
2824 	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2825 	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2826 	crc = crc16(crc, (__u8 *)gdp, offset);
2827 	offset += sizeof(gdp->bg_checksum); /* skip checksum */
2828 	/* for checksum of struct ext4_group_desc do the rest...*/
2829 	if (ext4_has_feature_64bit(sb) &&
2830 	    offset < le16_to_cpu(sbi->s_es->s_desc_size))
2831 		crc = crc16(crc, (__u8 *)gdp + offset,
2832 			    le16_to_cpu(sbi->s_es->s_desc_size) -
2833 				offset);
2834 
2835 out:
2836 	return cpu_to_le16(crc);
2837 }
2838 
2839 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2840 				struct ext4_group_desc *gdp)
2841 {
2842 	if (ext4_has_group_desc_csum(sb) &&
2843 	    (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2844 		return 0;
2845 
2846 	return 1;
2847 }
2848 
2849 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2850 			      struct ext4_group_desc *gdp)
2851 {
2852 	if (!ext4_has_group_desc_csum(sb))
2853 		return;
2854 	gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2855 }
2856 
2857 /* Called at mount-time, super-block is locked */
2858 static int ext4_check_descriptors(struct super_block *sb,
2859 				  ext4_fsblk_t sb_block,
2860 				  ext4_group_t *first_not_zeroed)
2861 {
2862 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2863 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2864 	ext4_fsblk_t last_block;
2865 	ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2866 	ext4_fsblk_t block_bitmap;
2867 	ext4_fsblk_t inode_bitmap;
2868 	ext4_fsblk_t inode_table;
2869 	int flexbg_flag = 0;
2870 	ext4_group_t i, grp = sbi->s_groups_count;
2871 
2872 	if (ext4_has_feature_flex_bg(sb))
2873 		flexbg_flag = 1;
2874 
2875 	ext4_debug("Checking group descriptors");
2876 
2877 	for (i = 0; i < sbi->s_groups_count; i++) {
2878 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2879 
2880 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
2881 			last_block = ext4_blocks_count(sbi->s_es) - 1;
2882 		else
2883 			last_block = first_block +
2884 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
2885 
2886 		if ((grp == sbi->s_groups_count) &&
2887 		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2888 			grp = i;
2889 
2890 		block_bitmap = ext4_block_bitmap(sb, gdp);
2891 		if (block_bitmap == sb_block) {
2892 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2893 				 "Block bitmap for group %u overlaps "
2894 				 "superblock", i);
2895 			if (!sb_rdonly(sb))
2896 				return 0;
2897 		}
2898 		if (block_bitmap >= sb_block + 1 &&
2899 		    block_bitmap <= last_bg_block) {
2900 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2901 				 "Block bitmap for group %u overlaps "
2902 				 "block group descriptors", i);
2903 			if (!sb_rdonly(sb))
2904 				return 0;
2905 		}
2906 		if (block_bitmap < first_block || block_bitmap > last_block) {
2907 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2908 			       "Block bitmap for group %u not in group "
2909 			       "(block %llu)!", i, block_bitmap);
2910 			return 0;
2911 		}
2912 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
2913 		if (inode_bitmap == sb_block) {
2914 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2915 				 "Inode bitmap for group %u overlaps "
2916 				 "superblock", i);
2917 			if (!sb_rdonly(sb))
2918 				return 0;
2919 		}
2920 		if (inode_bitmap >= sb_block + 1 &&
2921 		    inode_bitmap <= last_bg_block) {
2922 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2923 				 "Inode bitmap for group %u overlaps "
2924 				 "block group descriptors", i);
2925 			if (!sb_rdonly(sb))
2926 				return 0;
2927 		}
2928 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
2929 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2930 			       "Inode bitmap for group %u not in group "
2931 			       "(block %llu)!", i, inode_bitmap);
2932 			return 0;
2933 		}
2934 		inode_table = ext4_inode_table(sb, gdp);
2935 		if (inode_table == sb_block) {
2936 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2937 				 "Inode table for group %u overlaps "
2938 				 "superblock", i);
2939 			if (!sb_rdonly(sb))
2940 				return 0;
2941 		}
2942 		if (inode_table >= sb_block + 1 &&
2943 		    inode_table <= last_bg_block) {
2944 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2945 				 "Inode table for group %u overlaps "
2946 				 "block group descriptors", i);
2947 			if (!sb_rdonly(sb))
2948 				return 0;
2949 		}
2950 		if (inode_table < first_block ||
2951 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2952 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2953 			       "Inode table for group %u not in group "
2954 			       "(block %llu)!", i, inode_table);
2955 			return 0;
2956 		}
2957 		ext4_lock_group(sb, i);
2958 		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2959 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2960 				 "Checksum for group %u failed (%u!=%u)",
2961 				 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2962 				     gdp)), le16_to_cpu(gdp->bg_checksum));
2963 			if (!sb_rdonly(sb)) {
2964 				ext4_unlock_group(sb, i);
2965 				return 0;
2966 			}
2967 		}
2968 		ext4_unlock_group(sb, i);
2969 		if (!flexbg_flag)
2970 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
2971 	}
2972 	if (NULL != first_not_zeroed)
2973 		*first_not_zeroed = grp;
2974 	return 1;
2975 }
2976 
2977 /*
2978  * Maximal extent format file size.
2979  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2980  * extent format containers, within a sector_t, and within i_blocks
2981  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2982  * so that won't be a limiting factor.
2983  *
2984  * However there is other limiting factor. We do store extents in the form
2985  * of starting block and length, hence the resulting length of the extent
2986  * covering maximum file size must fit into on-disk format containers as
2987  * well. Given that length is always by 1 unit bigger than max unit (because
2988  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2989  *
2990  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2991  */
2992 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2993 {
2994 	loff_t res;
2995 	loff_t upper_limit = MAX_LFS_FILESIZE;
2996 
2997 	BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2998 
2999 	if (!has_huge_files) {
3000 		upper_limit = (1LL << 32) - 1;
3001 
3002 		/* total blocks in file system block size */
3003 		upper_limit >>= (blkbits - 9);
3004 		upper_limit <<= blkbits;
3005 	}
3006 
3007 	/*
3008 	 * 32-bit extent-start container, ee_block. We lower the maxbytes
3009 	 * by one fs block, so ee_len can cover the extent of maximum file
3010 	 * size
3011 	 */
3012 	res = (1LL << 32) - 1;
3013 	res <<= blkbits;
3014 
3015 	/* Sanity check against vm- & vfs- imposed limits */
3016 	if (res > upper_limit)
3017 		res = upper_limit;
3018 
3019 	return res;
3020 }
3021 
3022 /*
3023  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
3024  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3025  * We need to be 1 filesystem block less than the 2^48 sector limit.
3026  */
3027 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3028 {
3029 	unsigned long long upper_limit, res = EXT4_NDIR_BLOCKS;
3030 	int meta_blocks;
3031 
3032 	/*
3033 	 * This is calculated to be the largest file size for a dense, block
3034 	 * mapped file such that the file's total number of 512-byte sectors,
3035 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
3036 	 *
3037 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3038 	 * number of 512-byte sectors of the file.
3039 	 */
3040 	if (!has_huge_files) {
3041 		/*
3042 		 * !has_huge_files or implies that the inode i_block field
3043 		 * represents total file blocks in 2^32 512-byte sectors ==
3044 		 * size of vfs inode i_blocks * 8
3045 		 */
3046 		upper_limit = (1LL << 32) - 1;
3047 
3048 		/* total blocks in file system block size */
3049 		upper_limit >>= (bits - 9);
3050 
3051 	} else {
3052 		/*
3053 		 * We use 48 bit ext4_inode i_blocks
3054 		 * With EXT4_HUGE_FILE_FL set the i_blocks
3055 		 * represent total number of blocks in
3056 		 * file system block size
3057 		 */
3058 		upper_limit = (1LL << 48) - 1;
3059 
3060 	}
3061 
3062 	/* indirect blocks */
3063 	meta_blocks = 1;
3064 	/* double indirect blocks */
3065 	meta_blocks += 1 + (1LL << (bits-2));
3066 	/* tripple indirect blocks */
3067 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3068 
3069 	upper_limit -= meta_blocks;
3070 	upper_limit <<= bits;
3071 
3072 	res += 1LL << (bits-2);
3073 	res += 1LL << (2*(bits-2));
3074 	res += 1LL << (3*(bits-2));
3075 	res <<= bits;
3076 	if (res > upper_limit)
3077 		res = upper_limit;
3078 
3079 	if (res > MAX_LFS_FILESIZE)
3080 		res = MAX_LFS_FILESIZE;
3081 
3082 	return (loff_t)res;
3083 }
3084 
3085 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3086 				   ext4_fsblk_t logical_sb_block, int nr)
3087 {
3088 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3089 	ext4_group_t bg, first_meta_bg;
3090 	int has_super = 0;
3091 
3092 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3093 
3094 	if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3095 		return logical_sb_block + nr + 1;
3096 	bg = sbi->s_desc_per_block * nr;
3097 	if (ext4_bg_has_super(sb, bg))
3098 		has_super = 1;
3099 
3100 	/*
3101 	 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3102 	 * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
3103 	 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3104 	 * compensate.
3105 	 */
3106 	if (sb->s_blocksize == 1024 && nr == 0 &&
3107 	    le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3108 		has_super++;
3109 
3110 	return (has_super + ext4_group_first_block_no(sb, bg));
3111 }
3112 
3113 /**
3114  * ext4_get_stripe_size: Get the stripe size.
3115  * @sbi: In memory super block info
3116  *
3117  * If we have specified it via mount option, then
3118  * use the mount option value. If the value specified at mount time is
3119  * greater than the blocks per group use the super block value.
3120  * If the super block value is greater than blocks per group return 0.
3121  * Allocator needs it be less than blocks per group.
3122  *
3123  */
3124 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3125 {
3126 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3127 	unsigned long stripe_width =
3128 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3129 	int ret;
3130 
3131 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3132 		ret = sbi->s_stripe;
3133 	else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3134 		ret = stripe_width;
3135 	else if (stride && stride <= sbi->s_blocks_per_group)
3136 		ret = stride;
3137 	else
3138 		ret = 0;
3139 
3140 	/*
3141 	 * If the stripe width is 1, this makes no sense and
3142 	 * we set it to 0 to turn off stripe handling code.
3143 	 */
3144 	if (ret <= 1)
3145 		ret = 0;
3146 
3147 	return ret;
3148 }
3149 
3150 /*
3151  * Check whether this filesystem can be mounted based on
3152  * the features present and the RDONLY/RDWR mount requested.
3153  * Returns 1 if this filesystem can be mounted as requested,
3154  * 0 if it cannot be.
3155  */
3156 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3157 {
3158 	if (ext4_has_unknown_ext4_incompat_features(sb)) {
3159 		ext4_msg(sb, KERN_ERR,
3160 			"Couldn't mount because of "
3161 			"unsupported optional features (%x)",
3162 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3163 			~EXT4_FEATURE_INCOMPAT_SUPP));
3164 		return 0;
3165 	}
3166 
3167 #ifndef CONFIG_UNICODE
3168 	if (ext4_has_feature_casefold(sb)) {
3169 		ext4_msg(sb, KERN_ERR,
3170 			 "Filesystem with casefold feature cannot be "
3171 			 "mounted without CONFIG_UNICODE");
3172 		return 0;
3173 	}
3174 #endif
3175 
3176 	if (readonly)
3177 		return 1;
3178 
3179 	if (ext4_has_feature_readonly(sb)) {
3180 		ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3181 		sb->s_flags |= SB_RDONLY;
3182 		return 1;
3183 	}
3184 
3185 	/* Check that feature set is OK for a read-write mount */
3186 	if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3187 		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3188 			 "unsupported optional features (%x)",
3189 			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3190 				~EXT4_FEATURE_RO_COMPAT_SUPP));
3191 		return 0;
3192 	}
3193 	if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3194 		ext4_msg(sb, KERN_ERR,
3195 			 "Can't support bigalloc feature without "
3196 			 "extents feature\n");
3197 		return 0;
3198 	}
3199 
3200 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3201 	if (!readonly && (ext4_has_feature_quota(sb) ||
3202 			  ext4_has_feature_project(sb))) {
3203 		ext4_msg(sb, KERN_ERR,
3204 			 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3205 		return 0;
3206 	}
3207 #endif  /* CONFIG_QUOTA */
3208 	return 1;
3209 }
3210 
3211 /*
3212  * This function is called once a day if we have errors logged
3213  * on the file system
3214  */
3215 static void print_daily_error_info(struct timer_list *t)
3216 {
3217 	struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3218 	struct super_block *sb = sbi->s_sb;
3219 	struct ext4_super_block *es = sbi->s_es;
3220 
3221 	if (es->s_error_count)
3222 		/* fsck newer than v1.41.13 is needed to clean this condition. */
3223 		ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3224 			 le32_to_cpu(es->s_error_count));
3225 	if (es->s_first_error_time) {
3226 		printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3227 		       sb->s_id,
3228 		       ext4_get_tstamp(es, s_first_error_time),
3229 		       (int) sizeof(es->s_first_error_func),
3230 		       es->s_first_error_func,
3231 		       le32_to_cpu(es->s_first_error_line));
3232 		if (es->s_first_error_ino)
3233 			printk(KERN_CONT ": inode %u",
3234 			       le32_to_cpu(es->s_first_error_ino));
3235 		if (es->s_first_error_block)
3236 			printk(KERN_CONT ": block %llu", (unsigned long long)
3237 			       le64_to_cpu(es->s_first_error_block));
3238 		printk(KERN_CONT "\n");
3239 	}
3240 	if (es->s_last_error_time) {
3241 		printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3242 		       sb->s_id,
3243 		       ext4_get_tstamp(es, s_last_error_time),
3244 		       (int) sizeof(es->s_last_error_func),
3245 		       es->s_last_error_func,
3246 		       le32_to_cpu(es->s_last_error_line));
3247 		if (es->s_last_error_ino)
3248 			printk(KERN_CONT ": inode %u",
3249 			       le32_to_cpu(es->s_last_error_ino));
3250 		if (es->s_last_error_block)
3251 			printk(KERN_CONT ": block %llu", (unsigned long long)
3252 			       le64_to_cpu(es->s_last_error_block));
3253 		printk(KERN_CONT "\n");
3254 	}
3255 	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
3256 }
3257 
3258 /* Find next suitable group and run ext4_init_inode_table */
3259 static int ext4_run_li_request(struct ext4_li_request *elr)
3260 {
3261 	struct ext4_group_desc *gdp = NULL;
3262 	struct super_block *sb = elr->lr_super;
3263 	ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3264 	ext4_group_t group = elr->lr_next_group;
3265 	unsigned long timeout = 0;
3266 	unsigned int prefetch_ios = 0;
3267 	int ret = 0;
3268 
3269 	if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3270 		elr->lr_next_group = ext4_mb_prefetch(sb, group,
3271 				EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3272 		if (prefetch_ios)
3273 			ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3274 					      prefetch_ios);
3275 		trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3276 					    prefetch_ios);
3277 		if (group >= elr->lr_next_group) {
3278 			ret = 1;
3279 			if (elr->lr_first_not_zeroed != ngroups &&
3280 			    !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3281 				elr->lr_next_group = elr->lr_first_not_zeroed;
3282 				elr->lr_mode = EXT4_LI_MODE_ITABLE;
3283 				ret = 0;
3284 			}
3285 		}
3286 		return ret;
3287 	}
3288 
3289 	for (; group < ngroups; group++) {
3290 		gdp = ext4_get_group_desc(sb, group, NULL);
3291 		if (!gdp) {
3292 			ret = 1;
3293 			break;
3294 		}
3295 
3296 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3297 			break;
3298 	}
3299 
3300 	if (group >= ngroups)
3301 		ret = 1;
3302 
3303 	if (!ret) {
3304 		timeout = jiffies;
3305 		ret = ext4_init_inode_table(sb, group,
3306 					    elr->lr_timeout ? 0 : 1);
3307 		trace_ext4_lazy_itable_init(sb, group);
3308 		if (elr->lr_timeout == 0) {
3309 			timeout = (jiffies - timeout) *
3310 				EXT4_SB(elr->lr_super)->s_li_wait_mult;
3311 			elr->lr_timeout = timeout;
3312 		}
3313 		elr->lr_next_sched = jiffies + elr->lr_timeout;
3314 		elr->lr_next_group = group + 1;
3315 	}
3316 	return ret;
3317 }
3318 
3319 /*
3320  * Remove lr_request from the list_request and free the
3321  * request structure. Should be called with li_list_mtx held
3322  */
3323 static void ext4_remove_li_request(struct ext4_li_request *elr)
3324 {
3325 	if (!elr)
3326 		return;
3327 
3328 	list_del(&elr->lr_request);
3329 	EXT4_SB(elr->lr_super)->s_li_request = NULL;
3330 	kfree(elr);
3331 }
3332 
3333 static void ext4_unregister_li_request(struct super_block *sb)
3334 {
3335 	mutex_lock(&ext4_li_mtx);
3336 	if (!ext4_li_info) {
3337 		mutex_unlock(&ext4_li_mtx);
3338 		return;
3339 	}
3340 
3341 	mutex_lock(&ext4_li_info->li_list_mtx);
3342 	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3343 	mutex_unlock(&ext4_li_info->li_list_mtx);
3344 	mutex_unlock(&ext4_li_mtx);
3345 }
3346 
3347 static struct task_struct *ext4_lazyinit_task;
3348 
3349 /*
3350  * This is the function where ext4lazyinit thread lives. It walks
3351  * through the request list searching for next scheduled filesystem.
3352  * When such a fs is found, run the lazy initialization request
3353  * (ext4_rn_li_request) and keep track of the time spend in this
3354  * function. Based on that time we compute next schedule time of
3355  * the request. When walking through the list is complete, compute
3356  * next waking time and put itself into sleep.
3357  */
3358 static int ext4_lazyinit_thread(void *arg)
3359 {
3360 	struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3361 	struct list_head *pos, *n;
3362 	struct ext4_li_request *elr;
3363 	unsigned long next_wakeup, cur;
3364 
3365 	BUG_ON(NULL == eli);
3366 
3367 cont_thread:
3368 	while (true) {
3369 		next_wakeup = MAX_JIFFY_OFFSET;
3370 
3371 		mutex_lock(&eli->li_list_mtx);
3372 		if (list_empty(&eli->li_request_list)) {
3373 			mutex_unlock(&eli->li_list_mtx);
3374 			goto exit_thread;
3375 		}
3376 		list_for_each_safe(pos, n, &eli->li_request_list) {
3377 			int err = 0;
3378 			int progress = 0;
3379 			elr = list_entry(pos, struct ext4_li_request,
3380 					 lr_request);
3381 
3382 			if (time_before(jiffies, elr->lr_next_sched)) {
3383 				if (time_before(elr->lr_next_sched, next_wakeup))
3384 					next_wakeup = elr->lr_next_sched;
3385 				continue;
3386 			}
3387 			if (down_read_trylock(&elr->lr_super->s_umount)) {
3388 				if (sb_start_write_trylock(elr->lr_super)) {
3389 					progress = 1;
3390 					/*
3391 					 * We hold sb->s_umount, sb can not
3392 					 * be removed from the list, it is
3393 					 * now safe to drop li_list_mtx
3394 					 */
3395 					mutex_unlock(&eli->li_list_mtx);
3396 					err = ext4_run_li_request(elr);
3397 					sb_end_write(elr->lr_super);
3398 					mutex_lock(&eli->li_list_mtx);
3399 					n = pos->next;
3400 				}
3401 				up_read((&elr->lr_super->s_umount));
3402 			}
3403 			/* error, remove the lazy_init job */
3404 			if (err) {
3405 				ext4_remove_li_request(elr);
3406 				continue;
3407 			}
3408 			if (!progress) {
3409 				elr->lr_next_sched = jiffies +
3410 					(prandom_u32()
3411 					 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3412 			}
3413 			if (time_before(elr->lr_next_sched, next_wakeup))
3414 				next_wakeup = elr->lr_next_sched;
3415 		}
3416 		mutex_unlock(&eli->li_list_mtx);
3417 
3418 		try_to_freeze();
3419 
3420 		cur = jiffies;
3421 		if ((time_after_eq(cur, next_wakeup)) ||
3422 		    (MAX_JIFFY_OFFSET == next_wakeup)) {
3423 			cond_resched();
3424 			continue;
3425 		}
3426 
3427 		schedule_timeout_interruptible(next_wakeup - cur);
3428 
3429 		if (kthread_should_stop()) {
3430 			ext4_clear_request_list();
3431 			goto exit_thread;
3432 		}
3433 	}
3434 
3435 exit_thread:
3436 	/*
3437 	 * It looks like the request list is empty, but we need
3438 	 * to check it under the li_list_mtx lock, to prevent any
3439 	 * additions into it, and of course we should lock ext4_li_mtx
3440 	 * to atomically free the list and ext4_li_info, because at
3441 	 * this point another ext4 filesystem could be registering
3442 	 * new one.
3443 	 */
3444 	mutex_lock(&ext4_li_mtx);
3445 	mutex_lock(&eli->li_list_mtx);
3446 	if (!list_empty(&eli->li_request_list)) {
3447 		mutex_unlock(&eli->li_list_mtx);
3448 		mutex_unlock(&ext4_li_mtx);
3449 		goto cont_thread;
3450 	}
3451 	mutex_unlock(&eli->li_list_mtx);
3452 	kfree(ext4_li_info);
3453 	ext4_li_info = NULL;
3454 	mutex_unlock(&ext4_li_mtx);
3455 
3456 	return 0;
3457 }
3458 
3459 static void ext4_clear_request_list(void)
3460 {
3461 	struct list_head *pos, *n;
3462 	struct ext4_li_request *elr;
3463 
3464 	mutex_lock(&ext4_li_info->li_list_mtx);
3465 	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3466 		elr = list_entry(pos, struct ext4_li_request,
3467 				 lr_request);
3468 		ext4_remove_li_request(elr);
3469 	}
3470 	mutex_unlock(&ext4_li_info->li_list_mtx);
3471 }
3472 
3473 static int ext4_run_lazyinit_thread(void)
3474 {
3475 	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3476 					 ext4_li_info, "ext4lazyinit");
3477 	if (IS_ERR(ext4_lazyinit_task)) {
3478 		int err = PTR_ERR(ext4_lazyinit_task);
3479 		ext4_clear_request_list();
3480 		kfree(ext4_li_info);
3481 		ext4_li_info = NULL;
3482 		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3483 				 "initialization thread\n",
3484 				 err);
3485 		return err;
3486 	}
3487 	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3488 	return 0;
3489 }
3490 
3491 /*
3492  * Check whether it make sense to run itable init. thread or not.
3493  * If there is at least one uninitialized inode table, return
3494  * corresponding group number, else the loop goes through all
3495  * groups and return total number of groups.
3496  */
3497 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3498 {
3499 	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3500 	struct ext4_group_desc *gdp = NULL;
3501 
3502 	if (!ext4_has_group_desc_csum(sb))
3503 		return ngroups;
3504 
3505 	for (group = 0; group < ngroups; group++) {
3506 		gdp = ext4_get_group_desc(sb, group, NULL);
3507 		if (!gdp)
3508 			continue;
3509 
3510 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3511 			break;
3512 	}
3513 
3514 	return group;
3515 }
3516 
3517 static int ext4_li_info_new(void)
3518 {
3519 	struct ext4_lazy_init *eli = NULL;
3520 
3521 	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3522 	if (!eli)
3523 		return -ENOMEM;
3524 
3525 	INIT_LIST_HEAD(&eli->li_request_list);
3526 	mutex_init(&eli->li_list_mtx);
3527 
3528 	eli->li_state |= EXT4_LAZYINIT_QUIT;
3529 
3530 	ext4_li_info = eli;
3531 
3532 	return 0;
3533 }
3534 
3535 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3536 					    ext4_group_t start)
3537 {
3538 	struct ext4_li_request *elr;
3539 
3540 	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3541 	if (!elr)
3542 		return NULL;
3543 
3544 	elr->lr_super = sb;
3545 	elr->lr_first_not_zeroed = start;
3546 	if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3547 		elr->lr_mode = EXT4_LI_MODE_ITABLE;
3548 		elr->lr_next_group = start;
3549 	} else {
3550 		elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3551 	}
3552 
3553 	/*
3554 	 * Randomize first schedule time of the request to
3555 	 * spread the inode table initialization requests
3556 	 * better.
3557 	 */
3558 	elr->lr_next_sched = jiffies + (prandom_u32() %
3559 				(EXT4_DEF_LI_MAX_START_DELAY * HZ));
3560 	return elr;
3561 }
3562 
3563 int ext4_register_li_request(struct super_block *sb,
3564 			     ext4_group_t first_not_zeroed)
3565 {
3566 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3567 	struct ext4_li_request *elr = NULL;
3568 	ext4_group_t ngroups = sbi->s_groups_count;
3569 	int ret = 0;
3570 
3571 	mutex_lock(&ext4_li_mtx);
3572 	if (sbi->s_li_request != NULL) {
3573 		/*
3574 		 * Reset timeout so it can be computed again, because
3575 		 * s_li_wait_mult might have changed.
3576 		 */
3577 		sbi->s_li_request->lr_timeout = 0;
3578 		goto out;
3579 	}
3580 
3581 	if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3582 	    (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3583 	     !test_opt(sb, INIT_INODE_TABLE)))
3584 		goto out;
3585 
3586 	elr = ext4_li_request_new(sb, first_not_zeroed);
3587 	if (!elr) {
3588 		ret = -ENOMEM;
3589 		goto out;
3590 	}
3591 
3592 	if (NULL == ext4_li_info) {
3593 		ret = ext4_li_info_new();
3594 		if (ret)
3595 			goto out;
3596 	}
3597 
3598 	mutex_lock(&ext4_li_info->li_list_mtx);
3599 	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3600 	mutex_unlock(&ext4_li_info->li_list_mtx);
3601 
3602 	sbi->s_li_request = elr;
3603 	/*
3604 	 * set elr to NULL here since it has been inserted to
3605 	 * the request_list and the removal and free of it is
3606 	 * handled by ext4_clear_request_list from now on.
3607 	 */
3608 	elr = NULL;
3609 
3610 	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3611 		ret = ext4_run_lazyinit_thread();
3612 		if (ret)
3613 			goto out;
3614 	}
3615 out:
3616 	mutex_unlock(&ext4_li_mtx);
3617 	if (ret)
3618 		kfree(elr);
3619 	return ret;
3620 }
3621 
3622 /*
3623  * We do not need to lock anything since this is called on
3624  * module unload.
3625  */
3626 static void ext4_destroy_lazyinit_thread(void)
3627 {
3628 	/*
3629 	 * If thread exited earlier
3630 	 * there's nothing to be done.
3631 	 */
3632 	if (!ext4_li_info || !ext4_lazyinit_task)
3633 		return;
3634 
3635 	kthread_stop(ext4_lazyinit_task);
3636 }
3637 
3638 static int set_journal_csum_feature_set(struct super_block *sb)
3639 {
3640 	int ret = 1;
3641 	int compat, incompat;
3642 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3643 
3644 	if (ext4_has_metadata_csum(sb)) {
3645 		/* journal checksum v3 */
3646 		compat = 0;
3647 		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3648 	} else {
3649 		/* journal checksum v1 */
3650 		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3651 		incompat = 0;
3652 	}
3653 
3654 	jbd2_journal_clear_features(sbi->s_journal,
3655 			JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3656 			JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3657 			JBD2_FEATURE_INCOMPAT_CSUM_V2);
3658 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3659 		ret = jbd2_journal_set_features(sbi->s_journal,
3660 				compat, 0,
3661 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3662 				incompat);
3663 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3664 		ret = jbd2_journal_set_features(sbi->s_journal,
3665 				compat, 0,
3666 				incompat);
3667 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3668 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3669 	} else {
3670 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3671 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3672 	}
3673 
3674 	return ret;
3675 }
3676 
3677 /*
3678  * Note: calculating the overhead so we can be compatible with
3679  * historical BSD practice is quite difficult in the face of
3680  * clusters/bigalloc.  This is because multiple metadata blocks from
3681  * different block group can end up in the same allocation cluster.
3682  * Calculating the exact overhead in the face of clustered allocation
3683  * requires either O(all block bitmaps) in memory or O(number of block
3684  * groups**2) in time.  We will still calculate the superblock for
3685  * older file systems --- and if we come across with a bigalloc file
3686  * system with zero in s_overhead_clusters the estimate will be close to
3687  * correct especially for very large cluster sizes --- but for newer
3688  * file systems, it's better to calculate this figure once at mkfs
3689  * time, and store it in the superblock.  If the superblock value is
3690  * present (even for non-bigalloc file systems), we will use it.
3691  */
3692 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3693 			  char *buf)
3694 {
3695 	struct ext4_sb_info	*sbi = EXT4_SB(sb);
3696 	struct ext4_group_desc	*gdp;
3697 	ext4_fsblk_t		first_block, last_block, b;
3698 	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
3699 	int			s, j, count = 0;
3700 
3701 	if (!ext4_has_feature_bigalloc(sb))
3702 		return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3703 			sbi->s_itb_per_group + 2);
3704 
3705 	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3706 		(grp * EXT4_BLOCKS_PER_GROUP(sb));
3707 	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3708 	for (i = 0; i < ngroups; i++) {
3709 		gdp = ext4_get_group_desc(sb, i, NULL);
3710 		b = ext4_block_bitmap(sb, gdp);
3711 		if (b >= first_block && b <= last_block) {
3712 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3713 			count++;
3714 		}
3715 		b = ext4_inode_bitmap(sb, gdp);
3716 		if (b >= first_block && b <= last_block) {
3717 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3718 			count++;
3719 		}
3720 		b = ext4_inode_table(sb, gdp);
3721 		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3722 			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3723 				int c = EXT4_B2C(sbi, b - first_block);
3724 				ext4_set_bit(c, buf);
3725 				count++;
3726 			}
3727 		if (i != grp)
3728 			continue;
3729 		s = 0;
3730 		if (ext4_bg_has_super(sb, grp)) {
3731 			ext4_set_bit(s++, buf);
3732 			count++;
3733 		}
3734 		j = ext4_bg_num_gdb(sb, grp);
3735 		if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3736 			ext4_error(sb, "Invalid number of block group "
3737 				   "descriptor blocks: %d", j);
3738 			j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3739 		}
3740 		count += j;
3741 		for (; j > 0; j--)
3742 			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3743 	}
3744 	if (!count)
3745 		return 0;
3746 	return EXT4_CLUSTERS_PER_GROUP(sb) -
3747 		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3748 }
3749 
3750 /*
3751  * Compute the overhead and stash it in sbi->s_overhead
3752  */
3753 int ext4_calculate_overhead(struct super_block *sb)
3754 {
3755 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3756 	struct ext4_super_block *es = sbi->s_es;
3757 	struct inode *j_inode;
3758 	unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3759 	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3760 	ext4_fsblk_t overhead = 0;
3761 	char *buf = (char *) get_zeroed_page(GFP_NOFS);
3762 
3763 	if (!buf)
3764 		return -ENOMEM;
3765 
3766 	/*
3767 	 * Compute the overhead (FS structures).  This is constant
3768 	 * for a given filesystem unless the number of block groups
3769 	 * changes so we cache the previous value until it does.
3770 	 */
3771 
3772 	/*
3773 	 * All of the blocks before first_data_block are overhead
3774 	 */
3775 	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3776 
3777 	/*
3778 	 * Add the overhead found in each block group
3779 	 */
3780 	for (i = 0; i < ngroups; i++) {
3781 		int blks;
3782 
3783 		blks = count_overhead(sb, i, buf);
3784 		overhead += blks;
3785 		if (blks)
3786 			memset(buf, 0, PAGE_SIZE);
3787 		cond_resched();
3788 	}
3789 
3790 	/*
3791 	 * Add the internal journal blocks whether the journal has been
3792 	 * loaded or not
3793 	 */
3794 	if (sbi->s_journal && !sbi->s_journal_bdev)
3795 		overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
3796 	else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3797 		/* j_inum for internal journal is non-zero */
3798 		j_inode = ext4_get_journal_inode(sb, j_inum);
3799 		if (j_inode) {
3800 			j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3801 			overhead += EXT4_NUM_B2C(sbi, j_blocks);
3802 			iput(j_inode);
3803 		} else {
3804 			ext4_msg(sb, KERN_ERR, "can't get journal size");
3805 		}
3806 	}
3807 	sbi->s_overhead = overhead;
3808 	smp_wmb();
3809 	free_page((unsigned long) buf);
3810 	return 0;
3811 }
3812 
3813 static void ext4_set_resv_clusters(struct super_block *sb)
3814 {
3815 	ext4_fsblk_t resv_clusters;
3816 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3817 
3818 	/*
3819 	 * There's no need to reserve anything when we aren't using extents.
3820 	 * The space estimates are exact, there are no unwritten extents,
3821 	 * hole punching doesn't need new metadata... This is needed especially
3822 	 * to keep ext2/3 backward compatibility.
3823 	 */
3824 	if (!ext4_has_feature_extents(sb))
3825 		return;
3826 	/*
3827 	 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3828 	 * This should cover the situations where we can not afford to run
3829 	 * out of space like for example punch hole, or converting
3830 	 * unwritten extents in delalloc path. In most cases such
3831 	 * allocation would require 1, or 2 blocks, higher numbers are
3832 	 * very rare.
3833 	 */
3834 	resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3835 			 sbi->s_cluster_bits);
3836 
3837 	do_div(resv_clusters, 50);
3838 	resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3839 
3840 	atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3841 }
3842 
3843 static const char *ext4_quota_mode(struct super_block *sb)
3844 {
3845 #ifdef CONFIG_QUOTA
3846 	if (!ext4_quota_capable(sb))
3847 		return "none";
3848 
3849 	if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
3850 		return "journalled";
3851 	else
3852 		return "writeback";
3853 #else
3854 	return "disabled";
3855 #endif
3856 }
3857 
3858 static void ext4_setup_csum_trigger(struct super_block *sb,
3859 				    enum ext4_journal_trigger_type type,
3860 				    void (*trigger)(
3861 					struct jbd2_buffer_trigger_type *type,
3862 					struct buffer_head *bh,
3863 					void *mapped_data,
3864 					size_t size))
3865 {
3866 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3867 
3868 	sbi->s_journal_triggers[type].sb = sb;
3869 	sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
3870 }
3871 
3872 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3873 {
3874 	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3875 	char *orig_data = kstrdup(data, GFP_KERNEL);
3876 	struct buffer_head *bh, **group_desc;
3877 	struct ext4_super_block *es = NULL;
3878 	struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3879 	struct flex_groups **flex_groups;
3880 	ext4_fsblk_t block;
3881 	ext4_fsblk_t sb_block = get_sb_block(&data);
3882 	ext4_fsblk_t logical_sb_block;
3883 	unsigned long offset = 0;
3884 	unsigned long def_mount_opts;
3885 	struct inode *root;
3886 	const char *descr;
3887 	int ret = -ENOMEM;
3888 	int blocksize, clustersize;
3889 	unsigned int db_count;
3890 	unsigned int i;
3891 	int needs_recovery, has_huge_files;
3892 	__u64 blocks_count;
3893 	int err = 0;
3894 	ext4_group_t first_not_zeroed;
3895 	struct ext4_parsed_options parsed_opts;
3896 
3897 	/* Set defaults for the variables that will be set during parsing */
3898 	parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3899 	parsed_opts.journal_devnum = 0;
3900 	parsed_opts.mb_optimize_scan = DEFAULT_MB_OPTIMIZE_SCAN;
3901 
3902 	if ((data && !orig_data) || !sbi)
3903 		goto out_free_base;
3904 
3905 	sbi->s_daxdev = dax_dev;
3906 	sbi->s_blockgroup_lock =
3907 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3908 	if (!sbi->s_blockgroup_lock)
3909 		goto out_free_base;
3910 
3911 	sb->s_fs_info = sbi;
3912 	sbi->s_sb = sb;
3913 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3914 	sbi->s_sb_block = sb_block;
3915 	sbi->s_sectors_written_start =
3916 		part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
3917 
3918 	/* Cleanup superblock name */
3919 	strreplace(sb->s_id, '/', '!');
3920 
3921 	/* -EINVAL is default */
3922 	ret = -EINVAL;
3923 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3924 	if (!blocksize) {
3925 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3926 		goto out_fail;
3927 	}
3928 
3929 	/*
3930 	 * The ext4 superblock will not be buffer aligned for other than 1kB
3931 	 * block sizes.  We need to calculate the offset from buffer start.
3932 	 */
3933 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3934 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3935 		offset = do_div(logical_sb_block, blocksize);
3936 	} else {
3937 		logical_sb_block = sb_block;
3938 	}
3939 
3940 	bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
3941 	if (IS_ERR(bh)) {
3942 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
3943 		ret = PTR_ERR(bh);
3944 		goto out_fail;
3945 	}
3946 	/*
3947 	 * Note: s_es must be initialized as soon as possible because
3948 	 *       some ext4 macro-instructions depend on its value
3949 	 */
3950 	es = (struct ext4_super_block *) (bh->b_data + offset);
3951 	sbi->s_es = es;
3952 	sb->s_magic = le16_to_cpu(es->s_magic);
3953 	if (sb->s_magic != EXT4_SUPER_MAGIC)
3954 		goto cantfind_ext4;
3955 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3956 
3957 	/* Warn if metadata_csum and gdt_csum are both set. */
3958 	if (ext4_has_feature_metadata_csum(sb) &&
3959 	    ext4_has_feature_gdt_csum(sb))
3960 		ext4_warning(sb, "metadata_csum and uninit_bg are "
3961 			     "redundant flags; please run fsck.");
3962 
3963 	/* Check for a known checksum algorithm */
3964 	if (!ext4_verify_csum_type(sb, es)) {
3965 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3966 			 "unknown checksum algorithm.");
3967 		silent = 1;
3968 		goto cantfind_ext4;
3969 	}
3970 	ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
3971 				ext4_orphan_file_block_trigger);
3972 
3973 	/* Load the checksum driver */
3974 	sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3975 	if (IS_ERR(sbi->s_chksum_driver)) {
3976 		ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3977 		ret = PTR_ERR(sbi->s_chksum_driver);
3978 		sbi->s_chksum_driver = NULL;
3979 		goto failed_mount;
3980 	}
3981 
3982 	/* Check superblock checksum */
3983 	if (!ext4_superblock_csum_verify(sb, es)) {
3984 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3985 			 "invalid superblock checksum.  Run e2fsck?");
3986 		silent = 1;
3987 		ret = -EFSBADCRC;
3988 		goto cantfind_ext4;
3989 	}
3990 
3991 	/* Precompute checksum seed for all metadata */
3992 	if (ext4_has_feature_csum_seed(sb))
3993 		sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3994 	else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3995 		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3996 					       sizeof(es->s_uuid));
3997 
3998 	/* Set defaults before we parse the mount options */
3999 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4000 	set_opt(sb, INIT_INODE_TABLE);
4001 	if (def_mount_opts & EXT4_DEFM_DEBUG)
4002 		set_opt(sb, DEBUG);
4003 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4004 		set_opt(sb, GRPID);
4005 	if (def_mount_opts & EXT4_DEFM_UID16)
4006 		set_opt(sb, NO_UID32);
4007 	/* xattr user namespace & acls are now defaulted on */
4008 	set_opt(sb, XATTR_USER);
4009 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4010 	set_opt(sb, POSIX_ACL);
4011 #endif
4012 	if (ext4_has_feature_fast_commit(sb))
4013 		set_opt2(sb, JOURNAL_FAST_COMMIT);
4014 	/* don't forget to enable journal_csum when metadata_csum is enabled. */
4015 	if (ext4_has_metadata_csum(sb))
4016 		set_opt(sb, JOURNAL_CHECKSUM);
4017 
4018 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4019 		set_opt(sb, JOURNAL_DATA);
4020 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4021 		set_opt(sb, ORDERED_DATA);
4022 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4023 		set_opt(sb, WRITEBACK_DATA);
4024 
4025 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4026 		set_opt(sb, ERRORS_PANIC);
4027 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4028 		set_opt(sb, ERRORS_CONT);
4029 	else
4030 		set_opt(sb, ERRORS_RO);
4031 	/* block_validity enabled by default; disable with noblock_validity */
4032 	set_opt(sb, BLOCK_VALIDITY);
4033 	if (def_mount_opts & EXT4_DEFM_DISCARD)
4034 		set_opt(sb, DISCARD);
4035 
4036 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4037 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4038 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4039 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4040 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4041 
4042 	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4043 		set_opt(sb, BARRIER);
4044 
4045 	/*
4046 	 * enable delayed allocation by default
4047 	 * Use -o nodelalloc to turn it off
4048 	 */
4049 	if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4050 	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4051 		set_opt(sb, DELALLOC);
4052 
4053 	/*
4054 	 * set default s_li_wait_mult for lazyinit, for the case there is
4055 	 * no mount option specified.
4056 	 */
4057 	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4058 
4059 	if (le32_to_cpu(es->s_log_block_size) >
4060 	    (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4061 		ext4_msg(sb, KERN_ERR,
4062 			 "Invalid log block size: %u",
4063 			 le32_to_cpu(es->s_log_block_size));
4064 		goto failed_mount;
4065 	}
4066 	if (le32_to_cpu(es->s_log_cluster_size) >
4067 	    (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4068 		ext4_msg(sb, KERN_ERR,
4069 			 "Invalid log cluster size: %u",
4070 			 le32_to_cpu(es->s_log_cluster_size));
4071 		goto failed_mount;
4072 	}
4073 
4074 	blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4075 
4076 	if (blocksize == PAGE_SIZE)
4077 		set_opt(sb, DIOREAD_NOLOCK);
4078 
4079 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4080 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4081 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4082 	} else {
4083 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4084 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4085 		if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4086 			ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4087 				 sbi->s_first_ino);
4088 			goto failed_mount;
4089 		}
4090 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4091 		    (!is_power_of_2(sbi->s_inode_size)) ||
4092 		    (sbi->s_inode_size > blocksize)) {
4093 			ext4_msg(sb, KERN_ERR,
4094 			       "unsupported inode size: %d",
4095 			       sbi->s_inode_size);
4096 			ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4097 			goto failed_mount;
4098 		}
4099 		/*
4100 		 * i_atime_extra is the last extra field available for
4101 		 * [acm]times in struct ext4_inode. Checking for that
4102 		 * field should suffice to ensure we have extra space
4103 		 * for all three.
4104 		 */
4105 		if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4106 			sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4107 			sb->s_time_gran = 1;
4108 			sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4109 		} else {
4110 			sb->s_time_gran = NSEC_PER_SEC;
4111 			sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4112 		}
4113 		sb->s_time_min = EXT4_TIMESTAMP_MIN;
4114 	}
4115 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4116 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4117 			EXT4_GOOD_OLD_INODE_SIZE;
4118 		if (ext4_has_feature_extra_isize(sb)) {
4119 			unsigned v, max = (sbi->s_inode_size -
4120 					   EXT4_GOOD_OLD_INODE_SIZE);
4121 
4122 			v = le16_to_cpu(es->s_want_extra_isize);
4123 			if (v > max) {
4124 				ext4_msg(sb, KERN_ERR,
4125 					 "bad s_want_extra_isize: %d", v);
4126 				goto failed_mount;
4127 			}
4128 			if (sbi->s_want_extra_isize < v)
4129 				sbi->s_want_extra_isize = v;
4130 
4131 			v = le16_to_cpu(es->s_min_extra_isize);
4132 			if (v > max) {
4133 				ext4_msg(sb, KERN_ERR,
4134 					 "bad s_min_extra_isize: %d", v);
4135 				goto failed_mount;
4136 			}
4137 			if (sbi->s_want_extra_isize < v)
4138 				sbi->s_want_extra_isize = v;
4139 		}
4140 	}
4141 
4142 	if (sbi->s_es->s_mount_opts[0]) {
4143 		char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4144 					      sizeof(sbi->s_es->s_mount_opts),
4145 					      GFP_KERNEL);
4146 		if (!s_mount_opts)
4147 			goto failed_mount;
4148 		if (!parse_options(s_mount_opts, sb, &parsed_opts, 0)) {
4149 			ext4_msg(sb, KERN_WARNING,
4150 				 "failed to parse options in superblock: %s",
4151 				 s_mount_opts);
4152 		}
4153 		kfree(s_mount_opts);
4154 	}
4155 	sbi->s_def_mount_opt = sbi->s_mount_opt;
4156 	if (!parse_options((char *) data, sb, &parsed_opts, 0))
4157 		goto failed_mount;
4158 
4159 #ifdef CONFIG_UNICODE
4160 	if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4161 		const struct ext4_sb_encodings *encoding_info;
4162 		struct unicode_map *encoding;
4163 		__u16 encoding_flags;
4164 
4165 		if (ext4_sb_read_encoding(es, &encoding_info,
4166 					  &encoding_flags)) {
4167 			ext4_msg(sb, KERN_ERR,
4168 				 "Encoding requested by superblock is unknown");
4169 			goto failed_mount;
4170 		}
4171 
4172 		encoding = utf8_load(encoding_info->version);
4173 		if (IS_ERR(encoding)) {
4174 			ext4_msg(sb, KERN_ERR,
4175 				 "can't mount with superblock charset: %s-%s "
4176 				 "not supported by the kernel. flags: 0x%x.",
4177 				 encoding_info->name, encoding_info->version,
4178 				 encoding_flags);
4179 			goto failed_mount;
4180 		}
4181 		ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4182 			 "%s-%s with flags 0x%hx", encoding_info->name,
4183 			 encoding_info->version?:"\b", encoding_flags);
4184 
4185 		sb->s_encoding = encoding;
4186 		sb->s_encoding_flags = encoding_flags;
4187 	}
4188 #endif
4189 
4190 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4191 		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4192 		/* can't mount with both data=journal and dioread_nolock. */
4193 		clear_opt(sb, DIOREAD_NOLOCK);
4194 		clear_opt2(sb, JOURNAL_FAST_COMMIT);
4195 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4196 			ext4_msg(sb, KERN_ERR, "can't mount with "
4197 				 "both data=journal and delalloc");
4198 			goto failed_mount;
4199 		}
4200 		if (test_opt(sb, DAX_ALWAYS)) {
4201 			ext4_msg(sb, KERN_ERR, "can't mount with "
4202 				 "both data=journal and dax");
4203 			goto failed_mount;
4204 		}
4205 		if (ext4_has_feature_encrypt(sb)) {
4206 			ext4_msg(sb, KERN_WARNING,
4207 				 "encrypted files will use data=ordered "
4208 				 "instead of data journaling mode");
4209 		}
4210 		if (test_opt(sb, DELALLOC))
4211 			clear_opt(sb, DELALLOC);
4212 	} else {
4213 		sb->s_iflags |= SB_I_CGROUPWB;
4214 	}
4215 
4216 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4217 		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4218 
4219 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4220 	    (ext4_has_compat_features(sb) ||
4221 	     ext4_has_ro_compat_features(sb) ||
4222 	     ext4_has_incompat_features(sb)))
4223 		ext4_msg(sb, KERN_WARNING,
4224 		       "feature flags set on rev 0 fs, "
4225 		       "running e2fsck is recommended");
4226 
4227 	if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4228 		set_opt2(sb, HURD_COMPAT);
4229 		if (ext4_has_feature_64bit(sb)) {
4230 			ext4_msg(sb, KERN_ERR,
4231 				 "The Hurd can't support 64-bit file systems");
4232 			goto failed_mount;
4233 		}
4234 
4235 		/*
4236 		 * ea_inode feature uses l_i_version field which is not
4237 		 * available in HURD_COMPAT mode.
4238 		 */
4239 		if (ext4_has_feature_ea_inode(sb)) {
4240 			ext4_msg(sb, KERN_ERR,
4241 				 "ea_inode feature is not supported for Hurd");
4242 			goto failed_mount;
4243 		}
4244 	}
4245 
4246 	if (IS_EXT2_SB(sb)) {
4247 		if (ext2_feature_set_ok(sb))
4248 			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4249 				 "using the ext4 subsystem");
4250 		else {
4251 			/*
4252 			 * If we're probing be silent, if this looks like
4253 			 * it's actually an ext[34] filesystem.
4254 			 */
4255 			if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4256 				goto failed_mount;
4257 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4258 				 "to feature incompatibilities");
4259 			goto failed_mount;
4260 		}
4261 	}
4262 
4263 	if (IS_EXT3_SB(sb)) {
4264 		if (ext3_feature_set_ok(sb))
4265 			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4266 				 "using the ext4 subsystem");
4267 		else {
4268 			/*
4269 			 * If we're probing be silent, if this looks like
4270 			 * it's actually an ext4 filesystem.
4271 			 */
4272 			if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4273 				goto failed_mount;
4274 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4275 				 "to feature incompatibilities");
4276 			goto failed_mount;
4277 		}
4278 	}
4279 
4280 	/*
4281 	 * Check feature flags regardless of the revision level, since we
4282 	 * previously didn't change the revision level when setting the flags,
4283 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
4284 	 */
4285 	if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4286 		goto failed_mount;
4287 
4288 	if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4289 		ext4_msg(sb, KERN_ERR,
4290 			 "Number of reserved GDT blocks insanely large: %d",
4291 			 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4292 		goto failed_mount;
4293 	}
4294 
4295 	if (dax_supported(dax_dev, sb->s_bdev, blocksize, 0,
4296 			bdev_nr_sectors(sb->s_bdev)))
4297 		set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4298 
4299 	if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4300 		if (ext4_has_feature_inline_data(sb)) {
4301 			ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4302 					" that may contain inline data");
4303 			goto failed_mount;
4304 		}
4305 		if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4306 			ext4_msg(sb, KERN_ERR,
4307 				"DAX unsupported by block device.");
4308 			goto failed_mount;
4309 		}
4310 	}
4311 
4312 	if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4313 		ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4314 			 es->s_encryption_level);
4315 		goto failed_mount;
4316 	}
4317 
4318 	if (sb->s_blocksize != blocksize) {
4319 		/*
4320 		 * bh must be released before kill_bdev(), otherwise
4321 		 * it won't be freed and its page also. kill_bdev()
4322 		 * is called by sb_set_blocksize().
4323 		 */
4324 		brelse(bh);
4325 		/* Validate the filesystem blocksize */
4326 		if (!sb_set_blocksize(sb, blocksize)) {
4327 			ext4_msg(sb, KERN_ERR, "bad block size %d",
4328 					blocksize);
4329 			bh = NULL;
4330 			goto failed_mount;
4331 		}
4332 
4333 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4334 		offset = do_div(logical_sb_block, blocksize);
4335 		bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4336 		if (IS_ERR(bh)) {
4337 			ext4_msg(sb, KERN_ERR,
4338 			       "Can't read superblock on 2nd try");
4339 			ret = PTR_ERR(bh);
4340 			bh = NULL;
4341 			goto failed_mount;
4342 		}
4343 		es = (struct ext4_super_block *)(bh->b_data + offset);
4344 		sbi->s_es = es;
4345 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4346 			ext4_msg(sb, KERN_ERR,
4347 			       "Magic mismatch, very weird!");
4348 			goto failed_mount;
4349 		}
4350 	}
4351 
4352 	has_huge_files = ext4_has_feature_huge_file(sb);
4353 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4354 						      has_huge_files);
4355 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4356 
4357 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4358 	if (ext4_has_feature_64bit(sb)) {
4359 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4360 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4361 		    !is_power_of_2(sbi->s_desc_size)) {
4362 			ext4_msg(sb, KERN_ERR,
4363 			       "unsupported descriptor size %lu",
4364 			       sbi->s_desc_size);
4365 			goto failed_mount;
4366 		}
4367 	} else
4368 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4369 
4370 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4371 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4372 
4373 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4374 	if (sbi->s_inodes_per_block == 0)
4375 		goto cantfind_ext4;
4376 	if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4377 	    sbi->s_inodes_per_group > blocksize * 8) {
4378 		ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4379 			 sbi->s_inodes_per_group);
4380 		goto failed_mount;
4381 	}
4382 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
4383 					sbi->s_inodes_per_block;
4384 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4385 	sbi->s_sbh = bh;
4386 	sbi->s_mount_state = le16_to_cpu(es->s_state);
4387 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4388 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4389 
4390 	for (i = 0; i < 4; i++)
4391 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4392 	sbi->s_def_hash_version = es->s_def_hash_version;
4393 	if (ext4_has_feature_dir_index(sb)) {
4394 		i = le32_to_cpu(es->s_flags);
4395 		if (i & EXT2_FLAGS_UNSIGNED_HASH)
4396 			sbi->s_hash_unsigned = 3;
4397 		else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4398 #ifdef __CHAR_UNSIGNED__
4399 			if (!sb_rdonly(sb))
4400 				es->s_flags |=
4401 					cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4402 			sbi->s_hash_unsigned = 3;
4403 #else
4404 			if (!sb_rdonly(sb))
4405 				es->s_flags |=
4406 					cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4407 #endif
4408 		}
4409 	}
4410 
4411 	/* Handle clustersize */
4412 	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4413 	if (ext4_has_feature_bigalloc(sb)) {
4414 		if (clustersize < blocksize) {
4415 			ext4_msg(sb, KERN_ERR,
4416 				 "cluster size (%d) smaller than "
4417 				 "block size (%d)", clustersize, blocksize);
4418 			goto failed_mount;
4419 		}
4420 		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4421 			le32_to_cpu(es->s_log_block_size);
4422 		sbi->s_clusters_per_group =
4423 			le32_to_cpu(es->s_clusters_per_group);
4424 		if (sbi->s_clusters_per_group > blocksize * 8) {
4425 			ext4_msg(sb, KERN_ERR,
4426 				 "#clusters per group too big: %lu",
4427 				 sbi->s_clusters_per_group);
4428 			goto failed_mount;
4429 		}
4430 		if (sbi->s_blocks_per_group !=
4431 		    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4432 			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4433 				 "clusters per group (%lu) inconsistent",
4434 				 sbi->s_blocks_per_group,
4435 				 sbi->s_clusters_per_group);
4436 			goto failed_mount;
4437 		}
4438 	} else {
4439 		if (clustersize != blocksize) {
4440 			ext4_msg(sb, KERN_ERR,
4441 				 "fragment/cluster size (%d) != "
4442 				 "block size (%d)", clustersize, blocksize);
4443 			goto failed_mount;
4444 		}
4445 		if (sbi->s_blocks_per_group > blocksize * 8) {
4446 			ext4_msg(sb, KERN_ERR,
4447 				 "#blocks per group too big: %lu",
4448 				 sbi->s_blocks_per_group);
4449 			goto failed_mount;
4450 		}
4451 		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4452 		sbi->s_cluster_bits = 0;
4453 	}
4454 	sbi->s_cluster_ratio = clustersize / blocksize;
4455 
4456 	/* Do we have standard group size of clustersize * 8 blocks ? */
4457 	if (sbi->s_blocks_per_group == clustersize << 3)
4458 		set_opt2(sb, STD_GROUP_SIZE);
4459 
4460 	/*
4461 	 * Test whether we have more sectors than will fit in sector_t,
4462 	 * and whether the max offset is addressable by the page cache.
4463 	 */
4464 	err = generic_check_addressable(sb->s_blocksize_bits,
4465 					ext4_blocks_count(es));
4466 	if (err) {
4467 		ext4_msg(sb, KERN_ERR, "filesystem"
4468 			 " too large to mount safely on this system");
4469 		goto failed_mount;
4470 	}
4471 
4472 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4473 		goto cantfind_ext4;
4474 
4475 	/* check blocks count against device size */
4476 	blocks_count = sb_bdev_nr_blocks(sb);
4477 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4478 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4479 		       "exceeds size of device (%llu blocks)",
4480 		       ext4_blocks_count(es), blocks_count);
4481 		goto failed_mount;
4482 	}
4483 
4484 	/*
4485 	 * It makes no sense for the first data block to be beyond the end
4486 	 * of the filesystem.
4487 	 */
4488 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4489 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4490 			 "block %u is beyond end of filesystem (%llu)",
4491 			 le32_to_cpu(es->s_first_data_block),
4492 			 ext4_blocks_count(es));
4493 		goto failed_mount;
4494 	}
4495 	if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4496 	    (sbi->s_cluster_ratio == 1)) {
4497 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4498 			 "block is 0 with a 1k block and cluster size");
4499 		goto failed_mount;
4500 	}
4501 
4502 	blocks_count = (ext4_blocks_count(es) -
4503 			le32_to_cpu(es->s_first_data_block) +
4504 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
4505 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4506 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4507 		ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4508 		       "(block count %llu, first data block %u, "
4509 		       "blocks per group %lu)", blocks_count,
4510 		       ext4_blocks_count(es),
4511 		       le32_to_cpu(es->s_first_data_block),
4512 		       EXT4_BLOCKS_PER_GROUP(sb));
4513 		goto failed_mount;
4514 	}
4515 	sbi->s_groups_count = blocks_count;
4516 	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4517 			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4518 	if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4519 	    le32_to_cpu(es->s_inodes_count)) {
4520 		ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4521 			 le32_to_cpu(es->s_inodes_count),
4522 			 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4523 		ret = -EINVAL;
4524 		goto failed_mount;
4525 	}
4526 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4527 		   EXT4_DESC_PER_BLOCK(sb);
4528 	if (ext4_has_feature_meta_bg(sb)) {
4529 		if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4530 			ext4_msg(sb, KERN_WARNING,
4531 				 "first meta block group too large: %u "
4532 				 "(group descriptor block count %u)",
4533 				 le32_to_cpu(es->s_first_meta_bg), db_count);
4534 			goto failed_mount;
4535 		}
4536 	}
4537 	rcu_assign_pointer(sbi->s_group_desc,
4538 			   kvmalloc_array(db_count,
4539 					  sizeof(struct buffer_head *),
4540 					  GFP_KERNEL));
4541 	if (sbi->s_group_desc == NULL) {
4542 		ext4_msg(sb, KERN_ERR, "not enough memory");
4543 		ret = -ENOMEM;
4544 		goto failed_mount;
4545 	}
4546 
4547 	bgl_lock_init(sbi->s_blockgroup_lock);
4548 
4549 	/* Pre-read the descriptors into the buffer cache */
4550 	for (i = 0; i < db_count; i++) {
4551 		block = descriptor_loc(sb, logical_sb_block, i);
4552 		ext4_sb_breadahead_unmovable(sb, block);
4553 	}
4554 
4555 	for (i = 0; i < db_count; i++) {
4556 		struct buffer_head *bh;
4557 
4558 		block = descriptor_loc(sb, logical_sb_block, i);
4559 		bh = ext4_sb_bread_unmovable(sb, block);
4560 		if (IS_ERR(bh)) {
4561 			ext4_msg(sb, KERN_ERR,
4562 			       "can't read group descriptor %d", i);
4563 			db_count = i;
4564 			ret = PTR_ERR(bh);
4565 			goto failed_mount2;
4566 		}
4567 		rcu_read_lock();
4568 		rcu_dereference(sbi->s_group_desc)[i] = bh;
4569 		rcu_read_unlock();
4570 	}
4571 	sbi->s_gdb_count = db_count;
4572 	if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4573 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4574 		ret = -EFSCORRUPTED;
4575 		goto failed_mount2;
4576 	}
4577 
4578 	timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4579 	spin_lock_init(&sbi->s_error_lock);
4580 	INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
4581 
4582 	/* Register extent status tree shrinker */
4583 	if (ext4_es_register_shrinker(sbi))
4584 		goto failed_mount3;
4585 
4586 	sbi->s_stripe = ext4_get_stripe_size(sbi);
4587 	sbi->s_extent_max_zeroout_kb = 32;
4588 
4589 	/*
4590 	 * set up enough so that it can read an inode
4591 	 */
4592 	sb->s_op = &ext4_sops;
4593 	sb->s_export_op = &ext4_export_ops;
4594 	sb->s_xattr = ext4_xattr_handlers;
4595 #ifdef CONFIG_FS_ENCRYPTION
4596 	sb->s_cop = &ext4_cryptops;
4597 #endif
4598 #ifdef CONFIG_FS_VERITY
4599 	sb->s_vop = &ext4_verityops;
4600 #endif
4601 #ifdef CONFIG_QUOTA
4602 	sb->dq_op = &ext4_quota_operations;
4603 	if (ext4_has_feature_quota(sb))
4604 		sb->s_qcop = &dquot_quotactl_sysfile_ops;
4605 	else
4606 		sb->s_qcop = &ext4_qctl_operations;
4607 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4608 #endif
4609 	memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4610 
4611 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4612 	mutex_init(&sbi->s_orphan_lock);
4613 
4614 	/* Initialize fast commit stuff */
4615 	atomic_set(&sbi->s_fc_subtid, 0);
4616 	atomic_set(&sbi->s_fc_ineligible_updates, 0);
4617 	INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4618 	INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4619 	INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4620 	INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4621 	sbi->s_fc_bytes = 0;
4622 	ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4623 	ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
4624 	spin_lock_init(&sbi->s_fc_lock);
4625 	memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4626 	sbi->s_fc_replay_state.fc_regions = NULL;
4627 	sbi->s_fc_replay_state.fc_regions_size = 0;
4628 	sbi->s_fc_replay_state.fc_regions_used = 0;
4629 	sbi->s_fc_replay_state.fc_regions_valid = 0;
4630 	sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4631 	sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4632 	sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4633 
4634 	sb->s_root = NULL;
4635 
4636 	needs_recovery = (es->s_last_orphan != 0 ||
4637 			  ext4_has_feature_orphan_present(sb) ||
4638 			  ext4_has_feature_journal_needs_recovery(sb));
4639 
4640 	if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4641 		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4642 			goto failed_mount3a;
4643 
4644 	/*
4645 	 * The first inode we look at is the journal inode.  Don't try
4646 	 * root first: it may be modified in the journal!
4647 	 */
4648 	if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4649 		err = ext4_load_journal(sb, es, parsed_opts.journal_devnum);
4650 		if (err)
4651 			goto failed_mount3a;
4652 	} else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4653 		   ext4_has_feature_journal_needs_recovery(sb)) {
4654 		ext4_msg(sb, KERN_ERR, "required journal recovery "
4655 		       "suppressed and not mounted read-only");
4656 		goto failed_mount_wq;
4657 	} else {
4658 		/* Nojournal mode, all journal mount options are illegal */
4659 		if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4660 			ext4_msg(sb, KERN_ERR, "can't mount with "
4661 				 "journal_checksum, fs mounted w/o journal");
4662 			goto failed_mount_wq;
4663 		}
4664 		if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4665 			ext4_msg(sb, KERN_ERR, "can't mount with "
4666 				 "journal_async_commit, fs mounted w/o journal");
4667 			goto failed_mount_wq;
4668 		}
4669 		if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4670 			ext4_msg(sb, KERN_ERR, "can't mount with "
4671 				 "commit=%lu, fs mounted w/o journal",
4672 				 sbi->s_commit_interval / HZ);
4673 			goto failed_mount_wq;
4674 		}
4675 		if (EXT4_MOUNT_DATA_FLAGS &
4676 		    (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4677 			ext4_msg(sb, KERN_ERR, "can't mount with "
4678 				 "data=, fs mounted w/o journal");
4679 			goto failed_mount_wq;
4680 		}
4681 		sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4682 		clear_opt(sb, JOURNAL_CHECKSUM);
4683 		clear_opt(sb, DATA_FLAGS);
4684 		clear_opt2(sb, JOURNAL_FAST_COMMIT);
4685 		sbi->s_journal = NULL;
4686 		needs_recovery = 0;
4687 		goto no_journal;
4688 	}
4689 
4690 	if (ext4_has_feature_64bit(sb) &&
4691 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4692 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
4693 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4694 		goto failed_mount_wq;
4695 	}
4696 
4697 	if (!set_journal_csum_feature_set(sb)) {
4698 		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4699 			 "feature set");
4700 		goto failed_mount_wq;
4701 	}
4702 
4703 	if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4704 		!jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4705 					  JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4706 		ext4_msg(sb, KERN_ERR,
4707 			"Failed to set fast commit journal feature");
4708 		goto failed_mount_wq;
4709 	}
4710 
4711 	/* We have now updated the journal if required, so we can
4712 	 * validate the data journaling mode. */
4713 	switch (test_opt(sb, DATA_FLAGS)) {
4714 	case 0:
4715 		/* No mode set, assume a default based on the journal
4716 		 * capabilities: ORDERED_DATA if the journal can
4717 		 * cope, else JOURNAL_DATA
4718 		 */
4719 		if (jbd2_journal_check_available_features
4720 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4721 			set_opt(sb, ORDERED_DATA);
4722 			sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4723 		} else {
4724 			set_opt(sb, JOURNAL_DATA);
4725 			sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4726 		}
4727 		break;
4728 
4729 	case EXT4_MOUNT_ORDERED_DATA:
4730 	case EXT4_MOUNT_WRITEBACK_DATA:
4731 		if (!jbd2_journal_check_available_features
4732 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4733 			ext4_msg(sb, KERN_ERR, "Journal does not support "
4734 			       "requested data journaling mode");
4735 			goto failed_mount_wq;
4736 		}
4737 		break;
4738 	default:
4739 		break;
4740 	}
4741 
4742 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4743 	    test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4744 		ext4_msg(sb, KERN_ERR, "can't mount with "
4745 			"journal_async_commit in data=ordered mode");
4746 		goto failed_mount_wq;
4747 	}
4748 
4749 	set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio);
4750 
4751 	sbi->s_journal->j_submit_inode_data_buffers =
4752 		ext4_journal_submit_inode_data_buffers;
4753 	sbi->s_journal->j_finish_inode_data_buffers =
4754 		ext4_journal_finish_inode_data_buffers;
4755 
4756 no_journal:
4757 	if (!test_opt(sb, NO_MBCACHE)) {
4758 		sbi->s_ea_block_cache = ext4_xattr_create_cache();
4759 		if (!sbi->s_ea_block_cache) {
4760 			ext4_msg(sb, KERN_ERR,
4761 				 "Failed to create ea_block_cache");
4762 			goto failed_mount_wq;
4763 		}
4764 
4765 		if (ext4_has_feature_ea_inode(sb)) {
4766 			sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4767 			if (!sbi->s_ea_inode_cache) {
4768 				ext4_msg(sb, KERN_ERR,
4769 					 "Failed to create ea_inode_cache");
4770 				goto failed_mount_wq;
4771 			}
4772 		}
4773 	}
4774 
4775 	if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4776 		ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4777 		goto failed_mount_wq;
4778 	}
4779 
4780 	if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4781 	    !ext4_has_feature_encrypt(sb)) {
4782 		ext4_set_feature_encrypt(sb);
4783 		ext4_commit_super(sb);
4784 	}
4785 
4786 	/*
4787 	 * Get the # of file system overhead blocks from the
4788 	 * superblock if present.
4789 	 */
4790 	if (es->s_overhead_clusters)
4791 		sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4792 	else {
4793 		err = ext4_calculate_overhead(sb);
4794 		if (err)
4795 			goto failed_mount_wq;
4796 	}
4797 
4798 	/*
4799 	 * The maximum number of concurrent works can be high and
4800 	 * concurrency isn't really necessary.  Limit it to 1.
4801 	 */
4802 	EXT4_SB(sb)->rsv_conversion_wq =
4803 		alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4804 	if (!EXT4_SB(sb)->rsv_conversion_wq) {
4805 		printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4806 		ret = -ENOMEM;
4807 		goto failed_mount4;
4808 	}
4809 
4810 	/*
4811 	 * The jbd2_journal_load will have done any necessary log recovery,
4812 	 * so we can safely mount the rest of the filesystem now.
4813 	 */
4814 
4815 	root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4816 	if (IS_ERR(root)) {
4817 		ext4_msg(sb, KERN_ERR, "get root inode failed");
4818 		ret = PTR_ERR(root);
4819 		root = NULL;
4820 		goto failed_mount4;
4821 	}
4822 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4823 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4824 		iput(root);
4825 		goto failed_mount4;
4826 	}
4827 
4828 	sb->s_root = d_make_root(root);
4829 	if (!sb->s_root) {
4830 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
4831 		ret = -ENOMEM;
4832 		goto failed_mount4;
4833 	}
4834 
4835 	ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4836 	if (ret == -EROFS) {
4837 		sb->s_flags |= SB_RDONLY;
4838 		ret = 0;
4839 	} else if (ret)
4840 		goto failed_mount4a;
4841 
4842 	ext4_set_resv_clusters(sb);
4843 
4844 	if (test_opt(sb, BLOCK_VALIDITY)) {
4845 		err = ext4_setup_system_zone(sb);
4846 		if (err) {
4847 			ext4_msg(sb, KERN_ERR, "failed to initialize system "
4848 				 "zone (%d)", err);
4849 			goto failed_mount4a;
4850 		}
4851 	}
4852 	ext4_fc_replay_cleanup(sb);
4853 
4854 	ext4_ext_init(sb);
4855 
4856 	/*
4857 	 * Enable optimize_scan if number of groups is > threshold. This can be
4858 	 * turned off by passing "mb_optimize_scan=0". This can also be
4859 	 * turned on forcefully by passing "mb_optimize_scan=1".
4860 	 */
4861 	if (parsed_opts.mb_optimize_scan == 1)
4862 		set_opt2(sb, MB_OPTIMIZE_SCAN);
4863 	else if (parsed_opts.mb_optimize_scan == 0)
4864 		clear_opt2(sb, MB_OPTIMIZE_SCAN);
4865 	else if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
4866 		set_opt2(sb, MB_OPTIMIZE_SCAN);
4867 
4868 	err = ext4_mb_init(sb);
4869 	if (err) {
4870 		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4871 			 err);
4872 		goto failed_mount5;
4873 	}
4874 
4875 	/*
4876 	 * We can only set up the journal commit callback once
4877 	 * mballoc is initialized
4878 	 */
4879 	if (sbi->s_journal)
4880 		sbi->s_journal->j_commit_callback =
4881 			ext4_journal_commit_callback;
4882 
4883 	block = ext4_count_free_clusters(sb);
4884 	ext4_free_blocks_count_set(sbi->s_es,
4885 				   EXT4_C2B(sbi, block));
4886 	err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4887 				  GFP_KERNEL);
4888 	if (!err) {
4889 		unsigned long freei = ext4_count_free_inodes(sb);
4890 		sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4891 		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4892 					  GFP_KERNEL);
4893 	}
4894 	/*
4895 	 * Update the checksum after updating free space/inode
4896 	 * counters.  Otherwise the superblock can have an incorrect
4897 	 * checksum in the buffer cache until it is written out and
4898 	 * e2fsprogs programs trying to open a file system immediately
4899 	 * after it is mounted can fail.
4900 	 */
4901 	ext4_superblock_csum_set(sb);
4902 	if (!err)
4903 		err = percpu_counter_init(&sbi->s_dirs_counter,
4904 					  ext4_count_dirs(sb), GFP_KERNEL);
4905 	if (!err)
4906 		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4907 					  GFP_KERNEL);
4908 	if (!err)
4909 		err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
4910 					  GFP_KERNEL);
4911 	if (!err)
4912 		err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4913 
4914 	if (err) {
4915 		ext4_msg(sb, KERN_ERR, "insufficient memory");
4916 		goto failed_mount6;
4917 	}
4918 
4919 	if (ext4_has_feature_flex_bg(sb))
4920 		if (!ext4_fill_flex_info(sb)) {
4921 			ext4_msg(sb, KERN_ERR,
4922 			       "unable to initialize "
4923 			       "flex_bg meta info!");
4924 			ret = -ENOMEM;
4925 			goto failed_mount6;
4926 		}
4927 
4928 	err = ext4_register_li_request(sb, first_not_zeroed);
4929 	if (err)
4930 		goto failed_mount6;
4931 
4932 	err = ext4_register_sysfs(sb);
4933 	if (err)
4934 		goto failed_mount7;
4935 
4936 	err = ext4_init_orphan_info(sb);
4937 	if (err)
4938 		goto failed_mount8;
4939 #ifdef CONFIG_QUOTA
4940 	/* Enable quota usage during mount. */
4941 	if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4942 		err = ext4_enable_quotas(sb);
4943 		if (err)
4944 			goto failed_mount9;
4945 	}
4946 #endif  /* CONFIG_QUOTA */
4947 
4948 	/*
4949 	 * Save the original bdev mapping's wb_err value which could be
4950 	 * used to detect the metadata async write error.
4951 	 */
4952 	spin_lock_init(&sbi->s_bdev_wb_lock);
4953 	errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
4954 				 &sbi->s_bdev_wb_err);
4955 	sb->s_bdev->bd_super = sb;
4956 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4957 	ext4_orphan_cleanup(sb, es);
4958 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4959 	if (needs_recovery) {
4960 		ext4_msg(sb, KERN_INFO, "recovery complete");
4961 		err = ext4_mark_recovery_complete(sb, es);
4962 		if (err)
4963 			goto failed_mount9;
4964 	}
4965 	if (EXT4_SB(sb)->s_journal) {
4966 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4967 			descr = " journalled data mode";
4968 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4969 			descr = " ordered data mode";
4970 		else
4971 			descr = " writeback data mode";
4972 	} else
4973 		descr = "out journal";
4974 
4975 	if (test_opt(sb, DISCARD)) {
4976 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
4977 		if (!blk_queue_discard(q))
4978 			ext4_msg(sb, KERN_WARNING,
4979 				 "mounting with \"discard\" option, but "
4980 				 "the device does not support discard");
4981 	}
4982 
4983 	if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4984 		ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4985 			 "Opts: %.*s%s%s. Quota mode: %s.", descr,
4986 			 (int) sizeof(sbi->s_es->s_mount_opts),
4987 			 sbi->s_es->s_mount_opts,
4988 			 *sbi->s_es->s_mount_opts ? "; " : "", orig_data,
4989 			 ext4_quota_mode(sb));
4990 
4991 	if (es->s_error_count)
4992 		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4993 
4994 	/* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4995 	ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4996 	ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4997 	ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4998 	atomic_set(&sbi->s_warning_count, 0);
4999 	atomic_set(&sbi->s_msg_count, 0);
5000 
5001 	kfree(orig_data);
5002 	return 0;
5003 
5004 cantfind_ext4:
5005 	if (!silent)
5006 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5007 	goto failed_mount;
5008 
5009 failed_mount9:
5010 	ext4_release_orphan_info(sb);
5011 failed_mount8:
5012 	ext4_unregister_sysfs(sb);
5013 	kobject_put(&sbi->s_kobj);
5014 failed_mount7:
5015 	ext4_unregister_li_request(sb);
5016 failed_mount6:
5017 	ext4_mb_release(sb);
5018 	rcu_read_lock();
5019 	flex_groups = rcu_dereference(sbi->s_flex_groups);
5020 	if (flex_groups) {
5021 		for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5022 			kvfree(flex_groups[i]);
5023 		kvfree(flex_groups);
5024 	}
5025 	rcu_read_unlock();
5026 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
5027 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
5028 	percpu_counter_destroy(&sbi->s_dirs_counter);
5029 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5030 	percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5031 	percpu_free_rwsem(&sbi->s_writepages_rwsem);
5032 failed_mount5:
5033 	ext4_ext_release(sb);
5034 	ext4_release_system_zone(sb);
5035 failed_mount4a:
5036 	dput(sb->s_root);
5037 	sb->s_root = NULL;
5038 failed_mount4:
5039 	ext4_msg(sb, KERN_ERR, "mount failed");
5040 	if (EXT4_SB(sb)->rsv_conversion_wq)
5041 		destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5042 failed_mount_wq:
5043 	ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5044 	sbi->s_ea_inode_cache = NULL;
5045 
5046 	ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5047 	sbi->s_ea_block_cache = NULL;
5048 
5049 	if (sbi->s_journal) {
5050 		/* flush s_error_work before journal destroy. */
5051 		flush_work(&sbi->s_error_work);
5052 		jbd2_journal_destroy(sbi->s_journal);
5053 		sbi->s_journal = NULL;
5054 	}
5055 failed_mount3a:
5056 	ext4_es_unregister_shrinker(sbi);
5057 failed_mount3:
5058 	/* flush s_error_work before sbi destroy */
5059 	flush_work(&sbi->s_error_work);
5060 	del_timer_sync(&sbi->s_err_report);
5061 	ext4_stop_mmpd(sbi);
5062 failed_mount2:
5063 	rcu_read_lock();
5064 	group_desc = rcu_dereference(sbi->s_group_desc);
5065 	for (i = 0; i < db_count; i++)
5066 		brelse(group_desc[i]);
5067 	kvfree(group_desc);
5068 	rcu_read_unlock();
5069 failed_mount:
5070 	if (sbi->s_chksum_driver)
5071 		crypto_free_shash(sbi->s_chksum_driver);
5072 
5073 #ifdef CONFIG_UNICODE
5074 	utf8_unload(sb->s_encoding);
5075 #endif
5076 
5077 #ifdef CONFIG_QUOTA
5078 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
5079 		kfree(get_qf_name(sb, sbi, i));
5080 #endif
5081 	fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5082 	/* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5083 	brelse(bh);
5084 	ext4_blkdev_remove(sbi);
5085 out_fail:
5086 	sb->s_fs_info = NULL;
5087 	kfree(sbi->s_blockgroup_lock);
5088 out_free_base:
5089 	kfree(sbi);
5090 	kfree(orig_data);
5091 	fs_put_dax(dax_dev);
5092 	return err ? err : ret;
5093 }
5094 
5095 /*
5096  * Setup any per-fs journal parameters now.  We'll do this both on
5097  * initial mount, once the journal has been initialised but before we've
5098  * done any recovery; and again on any subsequent remount.
5099  */
5100 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5101 {
5102 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5103 
5104 	journal->j_commit_interval = sbi->s_commit_interval;
5105 	journal->j_min_batch_time = sbi->s_min_batch_time;
5106 	journal->j_max_batch_time = sbi->s_max_batch_time;
5107 	ext4_fc_init(sb, journal);
5108 
5109 	write_lock(&journal->j_state_lock);
5110 	if (test_opt(sb, BARRIER))
5111 		journal->j_flags |= JBD2_BARRIER;
5112 	else
5113 		journal->j_flags &= ~JBD2_BARRIER;
5114 	if (test_opt(sb, DATA_ERR_ABORT))
5115 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5116 	else
5117 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5118 	write_unlock(&journal->j_state_lock);
5119 }
5120 
5121 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5122 					     unsigned int journal_inum)
5123 {
5124 	struct inode *journal_inode;
5125 
5126 	/*
5127 	 * Test for the existence of a valid inode on disk.  Bad things
5128 	 * happen if we iget() an unused inode, as the subsequent iput()
5129 	 * will try to delete it.
5130 	 */
5131 	journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5132 	if (IS_ERR(journal_inode)) {
5133 		ext4_msg(sb, KERN_ERR, "no journal found");
5134 		return NULL;
5135 	}
5136 	if (!journal_inode->i_nlink) {
5137 		make_bad_inode(journal_inode);
5138 		iput(journal_inode);
5139 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5140 		return NULL;
5141 	}
5142 
5143 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5144 		  journal_inode, journal_inode->i_size);
5145 	if (!S_ISREG(journal_inode->i_mode)) {
5146 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
5147 		iput(journal_inode);
5148 		return NULL;
5149 	}
5150 	return journal_inode;
5151 }
5152 
5153 static journal_t *ext4_get_journal(struct super_block *sb,
5154 				   unsigned int journal_inum)
5155 {
5156 	struct inode *journal_inode;
5157 	journal_t *journal;
5158 
5159 	if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5160 		return NULL;
5161 
5162 	journal_inode = ext4_get_journal_inode(sb, journal_inum);
5163 	if (!journal_inode)
5164 		return NULL;
5165 
5166 	journal = jbd2_journal_init_inode(journal_inode);
5167 	if (!journal) {
5168 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5169 		iput(journal_inode);
5170 		return NULL;
5171 	}
5172 	journal->j_private = sb;
5173 	ext4_init_journal_params(sb, journal);
5174 	return journal;
5175 }
5176 
5177 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5178 				       dev_t j_dev)
5179 {
5180 	struct buffer_head *bh;
5181 	journal_t *journal;
5182 	ext4_fsblk_t start;
5183 	ext4_fsblk_t len;
5184 	int hblock, blocksize;
5185 	ext4_fsblk_t sb_block;
5186 	unsigned long offset;
5187 	struct ext4_super_block *es;
5188 	struct block_device *bdev;
5189 
5190 	if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5191 		return NULL;
5192 
5193 	bdev = ext4_blkdev_get(j_dev, sb);
5194 	if (bdev == NULL)
5195 		return NULL;
5196 
5197 	blocksize = sb->s_blocksize;
5198 	hblock = bdev_logical_block_size(bdev);
5199 	if (blocksize < hblock) {
5200 		ext4_msg(sb, KERN_ERR,
5201 			"blocksize too small for journal device");
5202 		goto out_bdev;
5203 	}
5204 
5205 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5206 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5207 	set_blocksize(bdev, blocksize);
5208 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
5209 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5210 		       "external journal");
5211 		goto out_bdev;
5212 	}
5213 
5214 	es = (struct ext4_super_block *) (bh->b_data + offset);
5215 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5216 	    !(le32_to_cpu(es->s_feature_incompat) &
5217 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5218 		ext4_msg(sb, KERN_ERR, "external journal has "
5219 					"bad superblock");
5220 		brelse(bh);
5221 		goto out_bdev;
5222 	}
5223 
5224 	if ((le32_to_cpu(es->s_feature_ro_compat) &
5225 	     EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5226 	    es->s_checksum != ext4_superblock_csum(sb, es)) {
5227 		ext4_msg(sb, KERN_ERR, "external journal has "
5228 				       "corrupt superblock");
5229 		brelse(bh);
5230 		goto out_bdev;
5231 	}
5232 
5233 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5234 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5235 		brelse(bh);
5236 		goto out_bdev;
5237 	}
5238 
5239 	len = ext4_blocks_count(es);
5240 	start = sb_block + 1;
5241 	brelse(bh);	/* we're done with the superblock */
5242 
5243 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5244 					start, len, blocksize);
5245 	if (!journal) {
5246 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
5247 		goto out_bdev;
5248 	}
5249 	journal->j_private = sb;
5250 	if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5251 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5252 		goto out_journal;
5253 	}
5254 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5255 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
5256 					"user (unsupported) - %d",
5257 			be32_to_cpu(journal->j_superblock->s_nr_users));
5258 		goto out_journal;
5259 	}
5260 	EXT4_SB(sb)->s_journal_bdev = bdev;
5261 	ext4_init_journal_params(sb, journal);
5262 	return journal;
5263 
5264 out_journal:
5265 	jbd2_journal_destroy(journal);
5266 out_bdev:
5267 	ext4_blkdev_put(bdev);
5268 	return NULL;
5269 }
5270 
5271 static int ext4_load_journal(struct super_block *sb,
5272 			     struct ext4_super_block *es,
5273 			     unsigned long journal_devnum)
5274 {
5275 	journal_t *journal;
5276 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5277 	dev_t journal_dev;
5278 	int err = 0;
5279 	int really_read_only;
5280 	int journal_dev_ro;
5281 
5282 	if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5283 		return -EFSCORRUPTED;
5284 
5285 	if (journal_devnum &&
5286 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5287 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5288 			"numbers have changed");
5289 		journal_dev = new_decode_dev(journal_devnum);
5290 	} else
5291 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5292 
5293 	if (journal_inum && journal_dev) {
5294 		ext4_msg(sb, KERN_ERR,
5295 			 "filesystem has both journal inode and journal device!");
5296 		return -EINVAL;
5297 	}
5298 
5299 	if (journal_inum) {
5300 		journal = ext4_get_journal(sb, journal_inum);
5301 		if (!journal)
5302 			return -EINVAL;
5303 	} else {
5304 		journal = ext4_get_dev_journal(sb, journal_dev);
5305 		if (!journal)
5306 			return -EINVAL;
5307 	}
5308 
5309 	journal_dev_ro = bdev_read_only(journal->j_dev);
5310 	really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5311 
5312 	if (journal_dev_ro && !sb_rdonly(sb)) {
5313 		ext4_msg(sb, KERN_ERR,
5314 			 "journal device read-only, try mounting with '-o ro'");
5315 		err = -EROFS;
5316 		goto err_out;
5317 	}
5318 
5319 	/*
5320 	 * Are we loading a blank journal or performing recovery after a
5321 	 * crash?  For recovery, we need to check in advance whether we
5322 	 * can get read-write access to the device.
5323 	 */
5324 	if (ext4_has_feature_journal_needs_recovery(sb)) {
5325 		if (sb_rdonly(sb)) {
5326 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
5327 					"required on readonly filesystem");
5328 			if (really_read_only) {
5329 				ext4_msg(sb, KERN_ERR, "write access "
5330 					"unavailable, cannot proceed "
5331 					"(try mounting with noload)");
5332 				err = -EROFS;
5333 				goto err_out;
5334 			}
5335 			ext4_msg(sb, KERN_INFO, "write access will "
5336 			       "be enabled during recovery");
5337 		}
5338 	}
5339 
5340 	if (!(journal->j_flags & JBD2_BARRIER))
5341 		ext4_msg(sb, KERN_INFO, "barriers disabled");
5342 
5343 	if (!ext4_has_feature_journal_needs_recovery(sb))
5344 		err = jbd2_journal_wipe(journal, !really_read_only);
5345 	if (!err) {
5346 		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5347 		if (save)
5348 			memcpy(save, ((char *) es) +
5349 			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5350 		err = jbd2_journal_load(journal);
5351 		if (save)
5352 			memcpy(((char *) es) + EXT4_S_ERR_START,
5353 			       save, EXT4_S_ERR_LEN);
5354 		kfree(save);
5355 	}
5356 
5357 	if (err) {
5358 		ext4_msg(sb, KERN_ERR, "error loading journal");
5359 		goto err_out;
5360 	}
5361 
5362 	EXT4_SB(sb)->s_journal = journal;
5363 	err = ext4_clear_journal_err(sb, es);
5364 	if (err) {
5365 		EXT4_SB(sb)->s_journal = NULL;
5366 		jbd2_journal_destroy(journal);
5367 		return err;
5368 	}
5369 
5370 	if (!really_read_only && journal_devnum &&
5371 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5372 		es->s_journal_dev = cpu_to_le32(journal_devnum);
5373 
5374 		/* Make sure we flush the recovery flag to disk. */
5375 		ext4_commit_super(sb);
5376 	}
5377 
5378 	return 0;
5379 
5380 err_out:
5381 	jbd2_journal_destroy(journal);
5382 	return err;
5383 }
5384 
5385 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5386 static void ext4_update_super(struct super_block *sb)
5387 {
5388 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5389 	struct ext4_super_block *es = sbi->s_es;
5390 	struct buffer_head *sbh = sbi->s_sbh;
5391 
5392 	lock_buffer(sbh);
5393 	/*
5394 	 * If the file system is mounted read-only, don't update the
5395 	 * superblock write time.  This avoids updating the superblock
5396 	 * write time when we are mounting the root file system
5397 	 * read/only but we need to replay the journal; at that point,
5398 	 * for people who are east of GMT and who make their clock
5399 	 * tick in localtime for Windows bug-for-bug compatibility,
5400 	 * the clock is set in the future, and this will cause e2fsck
5401 	 * to complain and force a full file system check.
5402 	 */
5403 	if (!(sb->s_flags & SB_RDONLY))
5404 		ext4_update_tstamp(es, s_wtime);
5405 	es->s_kbytes_written =
5406 		cpu_to_le64(sbi->s_kbytes_written +
5407 		    ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
5408 		      sbi->s_sectors_written_start) >> 1));
5409 	if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
5410 		ext4_free_blocks_count_set(es,
5411 			EXT4_C2B(sbi, percpu_counter_sum_positive(
5412 				&sbi->s_freeclusters_counter)));
5413 	if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
5414 		es->s_free_inodes_count =
5415 			cpu_to_le32(percpu_counter_sum_positive(
5416 				&sbi->s_freeinodes_counter));
5417 	/* Copy error information to the on-disk superblock */
5418 	spin_lock(&sbi->s_error_lock);
5419 	if (sbi->s_add_error_count > 0) {
5420 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5421 		if (!es->s_first_error_time && !es->s_first_error_time_hi) {
5422 			__ext4_update_tstamp(&es->s_first_error_time,
5423 					     &es->s_first_error_time_hi,
5424 					     sbi->s_first_error_time);
5425 			strncpy(es->s_first_error_func, sbi->s_first_error_func,
5426 				sizeof(es->s_first_error_func));
5427 			es->s_first_error_line =
5428 				cpu_to_le32(sbi->s_first_error_line);
5429 			es->s_first_error_ino =
5430 				cpu_to_le32(sbi->s_first_error_ino);
5431 			es->s_first_error_block =
5432 				cpu_to_le64(sbi->s_first_error_block);
5433 			es->s_first_error_errcode =
5434 				ext4_errno_to_code(sbi->s_first_error_code);
5435 		}
5436 		__ext4_update_tstamp(&es->s_last_error_time,
5437 				     &es->s_last_error_time_hi,
5438 				     sbi->s_last_error_time);
5439 		strncpy(es->s_last_error_func, sbi->s_last_error_func,
5440 			sizeof(es->s_last_error_func));
5441 		es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
5442 		es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
5443 		es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
5444 		es->s_last_error_errcode =
5445 				ext4_errno_to_code(sbi->s_last_error_code);
5446 		/*
5447 		 * Start the daily error reporting function if it hasn't been
5448 		 * started already
5449 		 */
5450 		if (!es->s_error_count)
5451 			mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
5452 		le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
5453 		sbi->s_add_error_count = 0;
5454 	}
5455 	spin_unlock(&sbi->s_error_lock);
5456 
5457 	ext4_superblock_csum_set(sb);
5458 	unlock_buffer(sbh);
5459 }
5460 
5461 static int ext4_commit_super(struct super_block *sb)
5462 {
5463 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5464 	int error = 0;
5465 
5466 	if (!sbh)
5467 		return -EINVAL;
5468 	if (block_device_ejected(sb))
5469 		return -ENODEV;
5470 
5471 	ext4_update_super(sb);
5472 
5473 	if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5474 		/*
5475 		 * Oh, dear.  A previous attempt to write the
5476 		 * superblock failed.  This could happen because the
5477 		 * USB device was yanked out.  Or it could happen to
5478 		 * be a transient write error and maybe the block will
5479 		 * be remapped.  Nothing we can do but to retry the
5480 		 * write and hope for the best.
5481 		 */
5482 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
5483 		       "superblock detected");
5484 		clear_buffer_write_io_error(sbh);
5485 		set_buffer_uptodate(sbh);
5486 	}
5487 	BUFFER_TRACE(sbh, "marking dirty");
5488 	mark_buffer_dirty(sbh);
5489 	error = __sync_dirty_buffer(sbh,
5490 		REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5491 	if (buffer_write_io_error(sbh)) {
5492 		ext4_msg(sb, KERN_ERR, "I/O error while writing "
5493 		       "superblock");
5494 		clear_buffer_write_io_error(sbh);
5495 		set_buffer_uptodate(sbh);
5496 	}
5497 	return error;
5498 }
5499 
5500 /*
5501  * Have we just finished recovery?  If so, and if we are mounting (or
5502  * remounting) the filesystem readonly, then we will end up with a
5503  * consistent fs on disk.  Record that fact.
5504  */
5505 static int ext4_mark_recovery_complete(struct super_block *sb,
5506 				       struct ext4_super_block *es)
5507 {
5508 	int err;
5509 	journal_t *journal = EXT4_SB(sb)->s_journal;
5510 
5511 	if (!ext4_has_feature_journal(sb)) {
5512 		if (journal != NULL) {
5513 			ext4_error(sb, "Journal got removed while the fs was "
5514 				   "mounted!");
5515 			return -EFSCORRUPTED;
5516 		}
5517 		return 0;
5518 	}
5519 	jbd2_journal_lock_updates(journal);
5520 	err = jbd2_journal_flush(journal, 0);
5521 	if (err < 0)
5522 		goto out;
5523 
5524 	if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
5525 	    ext4_has_feature_orphan_present(sb))) {
5526 		if (!ext4_orphan_file_empty(sb)) {
5527 			ext4_error(sb, "Orphan file not empty on read-only fs.");
5528 			err = -EFSCORRUPTED;
5529 			goto out;
5530 		}
5531 		ext4_clear_feature_journal_needs_recovery(sb);
5532 		ext4_clear_feature_orphan_present(sb);
5533 		ext4_commit_super(sb);
5534 	}
5535 out:
5536 	jbd2_journal_unlock_updates(journal);
5537 	return err;
5538 }
5539 
5540 /*
5541  * If we are mounting (or read-write remounting) a filesystem whose journal
5542  * has recorded an error from a previous lifetime, move that error to the
5543  * main filesystem now.
5544  */
5545 static int ext4_clear_journal_err(struct super_block *sb,
5546 				   struct ext4_super_block *es)
5547 {
5548 	journal_t *journal;
5549 	int j_errno;
5550 	const char *errstr;
5551 
5552 	if (!ext4_has_feature_journal(sb)) {
5553 		ext4_error(sb, "Journal got removed while the fs was mounted!");
5554 		return -EFSCORRUPTED;
5555 	}
5556 
5557 	journal = EXT4_SB(sb)->s_journal;
5558 
5559 	/*
5560 	 * Now check for any error status which may have been recorded in the
5561 	 * journal by a prior ext4_error() or ext4_abort()
5562 	 */
5563 
5564 	j_errno = jbd2_journal_errno(journal);
5565 	if (j_errno) {
5566 		char nbuf[16];
5567 
5568 		errstr = ext4_decode_error(sb, j_errno, nbuf);
5569 		ext4_warning(sb, "Filesystem error recorded "
5570 			     "from previous mount: %s", errstr);
5571 		ext4_warning(sb, "Marking fs in need of filesystem check.");
5572 
5573 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5574 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5575 		ext4_commit_super(sb);
5576 
5577 		jbd2_journal_clear_err(journal);
5578 		jbd2_journal_update_sb_errno(journal);
5579 	}
5580 	return 0;
5581 }
5582 
5583 /*
5584  * Force the running and committing transactions to commit,
5585  * and wait on the commit.
5586  */
5587 int ext4_force_commit(struct super_block *sb)
5588 {
5589 	journal_t *journal;
5590 
5591 	if (sb_rdonly(sb))
5592 		return 0;
5593 
5594 	journal = EXT4_SB(sb)->s_journal;
5595 	return ext4_journal_force_commit(journal);
5596 }
5597 
5598 static int ext4_sync_fs(struct super_block *sb, int wait)
5599 {
5600 	int ret = 0;
5601 	tid_t target;
5602 	bool needs_barrier = false;
5603 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5604 
5605 	if (unlikely(ext4_forced_shutdown(sbi)))
5606 		return 0;
5607 
5608 	trace_ext4_sync_fs(sb, wait);
5609 	flush_workqueue(sbi->rsv_conversion_wq);
5610 	/*
5611 	 * Writeback quota in non-journalled quota case - journalled quota has
5612 	 * no dirty dquots
5613 	 */
5614 	dquot_writeback_dquots(sb, -1);
5615 	/*
5616 	 * Data writeback is possible w/o journal transaction, so barrier must
5617 	 * being sent at the end of the function. But we can skip it if
5618 	 * transaction_commit will do it for us.
5619 	 */
5620 	if (sbi->s_journal) {
5621 		target = jbd2_get_latest_transaction(sbi->s_journal);
5622 		if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5623 		    !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5624 			needs_barrier = true;
5625 
5626 		if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5627 			if (wait)
5628 				ret = jbd2_log_wait_commit(sbi->s_journal,
5629 							   target);
5630 		}
5631 	} else if (wait && test_opt(sb, BARRIER))
5632 		needs_barrier = true;
5633 	if (needs_barrier) {
5634 		int err;
5635 		err = blkdev_issue_flush(sb->s_bdev);
5636 		if (!ret)
5637 			ret = err;
5638 	}
5639 
5640 	return ret;
5641 }
5642 
5643 /*
5644  * LVM calls this function before a (read-only) snapshot is created.  This
5645  * gives us a chance to flush the journal completely and mark the fs clean.
5646  *
5647  * Note that only this function cannot bring a filesystem to be in a clean
5648  * state independently. It relies on upper layer to stop all data & metadata
5649  * modifications.
5650  */
5651 static int ext4_freeze(struct super_block *sb)
5652 {
5653 	int error = 0;
5654 	journal_t *journal;
5655 
5656 	if (sb_rdonly(sb))
5657 		return 0;
5658 
5659 	journal = EXT4_SB(sb)->s_journal;
5660 
5661 	if (journal) {
5662 		/* Now we set up the journal barrier. */
5663 		jbd2_journal_lock_updates(journal);
5664 
5665 		/*
5666 		 * Don't clear the needs_recovery flag if we failed to
5667 		 * flush the journal.
5668 		 */
5669 		error = jbd2_journal_flush(journal, 0);
5670 		if (error < 0)
5671 			goto out;
5672 
5673 		/* Journal blocked and flushed, clear needs_recovery flag. */
5674 		ext4_clear_feature_journal_needs_recovery(sb);
5675 		if (ext4_orphan_file_empty(sb))
5676 			ext4_clear_feature_orphan_present(sb);
5677 	}
5678 
5679 	error = ext4_commit_super(sb);
5680 out:
5681 	if (journal)
5682 		/* we rely on upper layer to stop further updates */
5683 		jbd2_journal_unlock_updates(journal);
5684 	return error;
5685 }
5686 
5687 /*
5688  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
5689  * flag here, even though the filesystem is not technically dirty yet.
5690  */
5691 static int ext4_unfreeze(struct super_block *sb)
5692 {
5693 	if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5694 		return 0;
5695 
5696 	if (EXT4_SB(sb)->s_journal) {
5697 		/* Reset the needs_recovery flag before the fs is unlocked. */
5698 		ext4_set_feature_journal_needs_recovery(sb);
5699 		if (ext4_has_feature_orphan_file(sb))
5700 			ext4_set_feature_orphan_present(sb);
5701 	}
5702 
5703 	ext4_commit_super(sb);
5704 	return 0;
5705 }
5706 
5707 /*
5708  * Structure to save mount options for ext4_remount's benefit
5709  */
5710 struct ext4_mount_options {
5711 	unsigned long s_mount_opt;
5712 	unsigned long s_mount_opt2;
5713 	kuid_t s_resuid;
5714 	kgid_t s_resgid;
5715 	unsigned long s_commit_interval;
5716 	u32 s_min_batch_time, s_max_batch_time;
5717 #ifdef CONFIG_QUOTA
5718 	int s_jquota_fmt;
5719 	char *s_qf_names[EXT4_MAXQUOTAS];
5720 #endif
5721 };
5722 
5723 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5724 {
5725 	struct ext4_super_block *es;
5726 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5727 	unsigned long old_sb_flags, vfs_flags;
5728 	struct ext4_mount_options old_opts;
5729 	int enable_quota = 0;
5730 	ext4_group_t g;
5731 	int err = 0;
5732 #ifdef CONFIG_QUOTA
5733 	int i, j;
5734 	char *to_free[EXT4_MAXQUOTAS];
5735 #endif
5736 	char *orig_data = kstrdup(data, GFP_KERNEL);
5737 	struct ext4_parsed_options parsed_opts;
5738 
5739 	parsed_opts.journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5740 	parsed_opts.journal_devnum = 0;
5741 
5742 	if (data && !orig_data)
5743 		return -ENOMEM;
5744 
5745 	/* Store the original options */
5746 	old_sb_flags = sb->s_flags;
5747 	old_opts.s_mount_opt = sbi->s_mount_opt;
5748 	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5749 	old_opts.s_resuid = sbi->s_resuid;
5750 	old_opts.s_resgid = sbi->s_resgid;
5751 	old_opts.s_commit_interval = sbi->s_commit_interval;
5752 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
5753 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
5754 #ifdef CONFIG_QUOTA
5755 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5756 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
5757 		if (sbi->s_qf_names[i]) {
5758 			char *qf_name = get_qf_name(sb, sbi, i);
5759 
5760 			old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5761 			if (!old_opts.s_qf_names[i]) {
5762 				for (j = 0; j < i; j++)
5763 					kfree(old_opts.s_qf_names[j]);
5764 				kfree(orig_data);
5765 				return -ENOMEM;
5766 			}
5767 		} else
5768 			old_opts.s_qf_names[i] = NULL;
5769 #endif
5770 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5771 		parsed_opts.journal_ioprio =
5772 			sbi->s_journal->j_task->io_context->ioprio;
5773 
5774 	/*
5775 	 * Some options can be enabled by ext4 and/or by VFS mount flag
5776 	 * either way we need to make sure it matches in both *flags and
5777 	 * s_flags. Copy those selected flags from *flags to s_flags
5778 	 */
5779 	vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5780 	sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5781 
5782 	if (!parse_options(data, sb, &parsed_opts, 1)) {
5783 		err = -EINVAL;
5784 		goto restore_opts;
5785 	}
5786 
5787 	if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5788 	    test_opt(sb, JOURNAL_CHECKSUM)) {
5789 		ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5790 			 "during remount not supported; ignoring");
5791 		sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5792 	}
5793 
5794 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5795 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5796 			ext4_msg(sb, KERN_ERR, "can't mount with "
5797 				 "both data=journal and delalloc");
5798 			err = -EINVAL;
5799 			goto restore_opts;
5800 		}
5801 		if (test_opt(sb, DIOREAD_NOLOCK)) {
5802 			ext4_msg(sb, KERN_ERR, "can't mount with "
5803 				 "both data=journal and dioread_nolock");
5804 			err = -EINVAL;
5805 			goto restore_opts;
5806 		}
5807 	} else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5808 		if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5809 			ext4_msg(sb, KERN_ERR, "can't mount with "
5810 				"journal_async_commit in data=ordered mode");
5811 			err = -EINVAL;
5812 			goto restore_opts;
5813 		}
5814 	}
5815 
5816 	if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5817 		ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5818 		err = -EINVAL;
5819 		goto restore_opts;
5820 	}
5821 
5822 	if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
5823 		ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5824 
5825 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5826 		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5827 
5828 	es = sbi->s_es;
5829 
5830 	if (sbi->s_journal) {
5831 		ext4_init_journal_params(sb, sbi->s_journal);
5832 		set_task_ioprio(sbi->s_journal->j_task, parsed_opts.journal_ioprio);
5833 	}
5834 
5835 	/* Flush outstanding errors before changing fs state */
5836 	flush_work(&sbi->s_error_work);
5837 
5838 	if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5839 		if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
5840 			err = -EROFS;
5841 			goto restore_opts;
5842 		}
5843 
5844 		if (*flags & SB_RDONLY) {
5845 			err = sync_filesystem(sb);
5846 			if (err < 0)
5847 				goto restore_opts;
5848 			err = dquot_suspend(sb, -1);
5849 			if (err < 0)
5850 				goto restore_opts;
5851 
5852 			/*
5853 			 * First of all, the unconditional stuff we have to do
5854 			 * to disable replay of the journal when we next remount
5855 			 */
5856 			sb->s_flags |= SB_RDONLY;
5857 
5858 			/*
5859 			 * OK, test if we are remounting a valid rw partition
5860 			 * readonly, and if so set the rdonly flag and then
5861 			 * mark the partition as valid again.
5862 			 */
5863 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5864 			    (sbi->s_mount_state & EXT4_VALID_FS))
5865 				es->s_state = cpu_to_le16(sbi->s_mount_state);
5866 
5867 			if (sbi->s_journal) {
5868 				/*
5869 				 * We let remount-ro finish even if marking fs
5870 				 * as clean failed...
5871 				 */
5872 				ext4_mark_recovery_complete(sb, es);
5873 			}
5874 		} else {
5875 			/* Make sure we can mount this feature set readwrite */
5876 			if (ext4_has_feature_readonly(sb) ||
5877 			    !ext4_feature_set_ok(sb, 0)) {
5878 				err = -EROFS;
5879 				goto restore_opts;
5880 			}
5881 			/*
5882 			 * Make sure the group descriptor checksums
5883 			 * are sane.  If they aren't, refuse to remount r/w.
5884 			 */
5885 			for (g = 0; g < sbi->s_groups_count; g++) {
5886 				struct ext4_group_desc *gdp =
5887 					ext4_get_group_desc(sb, g, NULL);
5888 
5889 				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5890 					ext4_msg(sb, KERN_ERR,
5891 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
5892 		g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5893 					       le16_to_cpu(gdp->bg_checksum));
5894 					err = -EFSBADCRC;
5895 					goto restore_opts;
5896 				}
5897 			}
5898 
5899 			/*
5900 			 * If we have an unprocessed orphan list hanging
5901 			 * around from a previously readonly bdev mount,
5902 			 * require a full umount/remount for now.
5903 			 */
5904 			if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
5905 				ext4_msg(sb, KERN_WARNING, "Couldn't "
5906 				       "remount RDWR because of unprocessed "
5907 				       "orphan inode list.  Please "
5908 				       "umount/remount instead");
5909 				err = -EINVAL;
5910 				goto restore_opts;
5911 			}
5912 
5913 			/*
5914 			 * Mounting a RDONLY partition read-write, so reread
5915 			 * and store the current valid flag.  (It may have
5916 			 * been changed by e2fsck since we originally mounted
5917 			 * the partition.)
5918 			 */
5919 			if (sbi->s_journal) {
5920 				err = ext4_clear_journal_err(sb, es);
5921 				if (err)
5922 					goto restore_opts;
5923 			}
5924 			sbi->s_mount_state = le16_to_cpu(es->s_state);
5925 
5926 			err = ext4_setup_super(sb, es, 0);
5927 			if (err)
5928 				goto restore_opts;
5929 
5930 			sb->s_flags &= ~SB_RDONLY;
5931 			if (ext4_has_feature_mmp(sb))
5932 				if (ext4_multi_mount_protect(sb,
5933 						le64_to_cpu(es->s_mmp_block))) {
5934 					err = -EROFS;
5935 					goto restore_opts;
5936 				}
5937 			enable_quota = 1;
5938 		}
5939 	}
5940 
5941 	/*
5942 	 * Reinitialize lazy itable initialization thread based on
5943 	 * current settings
5944 	 */
5945 	if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5946 		ext4_unregister_li_request(sb);
5947 	else {
5948 		ext4_group_t first_not_zeroed;
5949 		first_not_zeroed = ext4_has_uninit_itable(sb);
5950 		ext4_register_li_request(sb, first_not_zeroed);
5951 	}
5952 
5953 	/*
5954 	 * Handle creation of system zone data early because it can fail.
5955 	 * Releasing of existing data is done when we are sure remount will
5956 	 * succeed.
5957 	 */
5958 	if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
5959 		err = ext4_setup_system_zone(sb);
5960 		if (err)
5961 			goto restore_opts;
5962 	}
5963 
5964 	if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5965 		err = ext4_commit_super(sb);
5966 		if (err)
5967 			goto restore_opts;
5968 	}
5969 
5970 #ifdef CONFIG_QUOTA
5971 	/* Release old quota file names */
5972 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
5973 		kfree(old_opts.s_qf_names[i]);
5974 	if (enable_quota) {
5975 		if (sb_any_quota_suspended(sb))
5976 			dquot_resume(sb, -1);
5977 		else if (ext4_has_feature_quota(sb)) {
5978 			err = ext4_enable_quotas(sb);
5979 			if (err)
5980 				goto restore_opts;
5981 		}
5982 	}
5983 #endif
5984 	if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5985 		ext4_release_system_zone(sb);
5986 
5987 	if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
5988 		ext4_stop_mmpd(sbi);
5989 
5990 	/*
5991 	 * Some options can be enabled by ext4 and/or by VFS mount flag
5992 	 * either way we need to make sure it matches in both *flags and
5993 	 * s_flags. Copy those selected flags from s_flags to *flags
5994 	 */
5995 	*flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
5996 
5997 	ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.",
5998 		 orig_data, ext4_quota_mode(sb));
5999 	kfree(orig_data);
6000 	return 0;
6001 
6002 restore_opts:
6003 	sb->s_flags = old_sb_flags;
6004 	sbi->s_mount_opt = old_opts.s_mount_opt;
6005 	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6006 	sbi->s_resuid = old_opts.s_resuid;
6007 	sbi->s_resgid = old_opts.s_resgid;
6008 	sbi->s_commit_interval = old_opts.s_commit_interval;
6009 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
6010 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
6011 	if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6012 		ext4_release_system_zone(sb);
6013 #ifdef CONFIG_QUOTA
6014 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6015 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6016 		to_free[i] = get_qf_name(sb, sbi, i);
6017 		rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6018 	}
6019 	synchronize_rcu();
6020 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
6021 		kfree(to_free[i]);
6022 #endif
6023 	if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6024 		ext4_stop_mmpd(sbi);
6025 	kfree(orig_data);
6026 	return err;
6027 }
6028 
6029 #ifdef CONFIG_QUOTA
6030 static int ext4_statfs_project(struct super_block *sb,
6031 			       kprojid_t projid, struct kstatfs *buf)
6032 {
6033 	struct kqid qid;
6034 	struct dquot *dquot;
6035 	u64 limit;
6036 	u64 curblock;
6037 
6038 	qid = make_kqid_projid(projid);
6039 	dquot = dqget(sb, qid);
6040 	if (IS_ERR(dquot))
6041 		return PTR_ERR(dquot);
6042 	spin_lock(&dquot->dq_dqb_lock);
6043 
6044 	limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6045 			     dquot->dq_dqb.dqb_bhardlimit);
6046 	limit >>= sb->s_blocksize_bits;
6047 
6048 	if (limit && buf->f_blocks > limit) {
6049 		curblock = (dquot->dq_dqb.dqb_curspace +
6050 			    dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6051 		buf->f_blocks = limit;
6052 		buf->f_bfree = buf->f_bavail =
6053 			(buf->f_blocks > curblock) ?
6054 			 (buf->f_blocks - curblock) : 0;
6055 	}
6056 
6057 	limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6058 			     dquot->dq_dqb.dqb_ihardlimit);
6059 	if (limit && buf->f_files > limit) {
6060 		buf->f_files = limit;
6061 		buf->f_ffree =
6062 			(buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6063 			 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6064 	}
6065 
6066 	spin_unlock(&dquot->dq_dqb_lock);
6067 	dqput(dquot);
6068 	return 0;
6069 }
6070 #endif
6071 
6072 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6073 {
6074 	struct super_block *sb = dentry->d_sb;
6075 	struct ext4_sb_info *sbi = EXT4_SB(sb);
6076 	struct ext4_super_block *es = sbi->s_es;
6077 	ext4_fsblk_t overhead = 0, resv_blocks;
6078 	s64 bfree;
6079 	resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6080 
6081 	if (!test_opt(sb, MINIX_DF))
6082 		overhead = sbi->s_overhead;
6083 
6084 	buf->f_type = EXT4_SUPER_MAGIC;
6085 	buf->f_bsize = sb->s_blocksize;
6086 	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6087 	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6088 		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6089 	/* prevent underflow in case that few free space is available */
6090 	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6091 	buf->f_bavail = buf->f_bfree -
6092 			(ext4_r_blocks_count(es) + resv_blocks);
6093 	if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6094 		buf->f_bavail = 0;
6095 	buf->f_files = le32_to_cpu(es->s_inodes_count);
6096 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6097 	buf->f_namelen = EXT4_NAME_LEN;
6098 	buf->f_fsid = uuid_to_fsid(es->s_uuid);
6099 
6100 #ifdef CONFIG_QUOTA
6101 	if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6102 	    sb_has_quota_limits_enabled(sb, PRJQUOTA))
6103 		ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6104 #endif
6105 	return 0;
6106 }
6107 
6108 
6109 #ifdef CONFIG_QUOTA
6110 
6111 /*
6112  * Helper functions so that transaction is started before we acquire dqio_sem
6113  * to keep correct lock ordering of transaction > dqio_sem
6114  */
6115 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6116 {
6117 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6118 }
6119 
6120 static int ext4_write_dquot(struct dquot *dquot)
6121 {
6122 	int ret, err;
6123 	handle_t *handle;
6124 	struct inode *inode;
6125 
6126 	inode = dquot_to_inode(dquot);
6127 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6128 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6129 	if (IS_ERR(handle))
6130 		return PTR_ERR(handle);
6131 	ret = dquot_commit(dquot);
6132 	err = ext4_journal_stop(handle);
6133 	if (!ret)
6134 		ret = err;
6135 	return ret;
6136 }
6137 
6138 static int ext4_acquire_dquot(struct dquot *dquot)
6139 {
6140 	int ret, err;
6141 	handle_t *handle;
6142 
6143 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6144 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6145 	if (IS_ERR(handle))
6146 		return PTR_ERR(handle);
6147 	ret = dquot_acquire(dquot);
6148 	err = ext4_journal_stop(handle);
6149 	if (!ret)
6150 		ret = err;
6151 	return ret;
6152 }
6153 
6154 static int ext4_release_dquot(struct dquot *dquot)
6155 {
6156 	int ret, err;
6157 	handle_t *handle;
6158 
6159 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6160 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6161 	if (IS_ERR(handle)) {
6162 		/* Release dquot anyway to avoid endless cycle in dqput() */
6163 		dquot_release(dquot);
6164 		return PTR_ERR(handle);
6165 	}
6166 	ret = dquot_release(dquot);
6167 	err = ext4_journal_stop(handle);
6168 	if (!ret)
6169 		ret = err;
6170 	return ret;
6171 }
6172 
6173 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6174 {
6175 	struct super_block *sb = dquot->dq_sb;
6176 
6177 	if (ext4_is_quota_journalled(sb)) {
6178 		dquot_mark_dquot_dirty(dquot);
6179 		return ext4_write_dquot(dquot);
6180 	} else {
6181 		return dquot_mark_dquot_dirty(dquot);
6182 	}
6183 }
6184 
6185 static int ext4_write_info(struct super_block *sb, int type)
6186 {
6187 	int ret, err;
6188 	handle_t *handle;
6189 
6190 	/* Data block + inode block */
6191 	handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6192 	if (IS_ERR(handle))
6193 		return PTR_ERR(handle);
6194 	ret = dquot_commit_info(sb, type);
6195 	err = ext4_journal_stop(handle);
6196 	if (!ret)
6197 		ret = err;
6198 	return ret;
6199 }
6200 
6201 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6202 {
6203 	struct ext4_inode_info *ei = EXT4_I(inode);
6204 
6205 	/* The first argument of lockdep_set_subclass has to be
6206 	 * *exactly* the same as the argument to init_rwsem() --- in
6207 	 * this case, in init_once() --- or lockdep gets unhappy
6208 	 * because the name of the lock is set using the
6209 	 * stringification of the argument to init_rwsem().
6210 	 */
6211 	(void) ei;	/* shut up clang warning if !CONFIG_LOCKDEP */
6212 	lockdep_set_subclass(&ei->i_data_sem, subclass);
6213 }
6214 
6215 /*
6216  * Standard function to be called on quota_on
6217  */
6218 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6219 			 const struct path *path)
6220 {
6221 	int err;
6222 
6223 	if (!test_opt(sb, QUOTA))
6224 		return -EINVAL;
6225 
6226 	/* Quotafile not on the same filesystem? */
6227 	if (path->dentry->d_sb != sb)
6228 		return -EXDEV;
6229 
6230 	/* Quota already enabled for this file? */
6231 	if (IS_NOQUOTA(d_inode(path->dentry)))
6232 		return -EBUSY;
6233 
6234 	/* Journaling quota? */
6235 	if (EXT4_SB(sb)->s_qf_names[type]) {
6236 		/* Quotafile not in fs root? */
6237 		if (path->dentry->d_parent != sb->s_root)
6238 			ext4_msg(sb, KERN_WARNING,
6239 				"Quota file not on filesystem root. "
6240 				"Journaled quota will not work");
6241 		sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6242 	} else {
6243 		/*
6244 		 * Clear the flag just in case mount options changed since
6245 		 * last time.
6246 		 */
6247 		sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6248 	}
6249 
6250 	/*
6251 	 * When we journal data on quota file, we have to flush journal to see
6252 	 * all updates to the file when we bypass pagecache...
6253 	 */
6254 	if (EXT4_SB(sb)->s_journal &&
6255 	    ext4_should_journal_data(d_inode(path->dentry))) {
6256 		/*
6257 		 * We don't need to lock updates but journal_flush() could
6258 		 * otherwise be livelocked...
6259 		 */
6260 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6261 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
6262 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6263 		if (err)
6264 			return err;
6265 	}
6266 
6267 	lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6268 	err = dquot_quota_on(sb, type, format_id, path);
6269 	if (err) {
6270 		lockdep_set_quota_inode(path->dentry->d_inode,
6271 					     I_DATA_SEM_NORMAL);
6272 	} else {
6273 		struct inode *inode = d_inode(path->dentry);
6274 		handle_t *handle;
6275 
6276 		/*
6277 		 * Set inode flags to prevent userspace from messing with quota
6278 		 * files. If this fails, we return success anyway since quotas
6279 		 * are already enabled and this is not a hard failure.
6280 		 */
6281 		inode_lock(inode);
6282 		handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6283 		if (IS_ERR(handle))
6284 			goto unlock_inode;
6285 		EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6286 		inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6287 				S_NOATIME | S_IMMUTABLE);
6288 		err = ext4_mark_inode_dirty(handle, inode);
6289 		ext4_journal_stop(handle);
6290 	unlock_inode:
6291 		inode_unlock(inode);
6292 	}
6293 	return err;
6294 }
6295 
6296 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6297 			     unsigned int flags)
6298 {
6299 	int err;
6300 	struct inode *qf_inode;
6301 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6302 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6303 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6304 		le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6305 	};
6306 
6307 	BUG_ON(!ext4_has_feature_quota(sb));
6308 
6309 	if (!qf_inums[type])
6310 		return -EPERM;
6311 
6312 	qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6313 	if (IS_ERR(qf_inode)) {
6314 		ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6315 		return PTR_ERR(qf_inode);
6316 	}
6317 
6318 	/* Don't account quota for quota files to avoid recursion */
6319 	qf_inode->i_flags |= S_NOQUOTA;
6320 	lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6321 	err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6322 	if (err)
6323 		lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6324 	iput(qf_inode);
6325 
6326 	return err;
6327 }
6328 
6329 /* Enable usage tracking for all quota types. */
6330 int ext4_enable_quotas(struct super_block *sb)
6331 {
6332 	int type, err = 0;
6333 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6334 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6335 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6336 		le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6337 	};
6338 	bool quota_mopt[EXT4_MAXQUOTAS] = {
6339 		test_opt(sb, USRQUOTA),
6340 		test_opt(sb, GRPQUOTA),
6341 		test_opt(sb, PRJQUOTA),
6342 	};
6343 
6344 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6345 	for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6346 		if (qf_inums[type]) {
6347 			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6348 				DQUOT_USAGE_ENABLED |
6349 				(quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6350 			if (err) {
6351 				ext4_warning(sb,
6352 					"Failed to enable quota tracking "
6353 					"(type=%d, err=%d). Please run "
6354 					"e2fsck to fix.", type, err);
6355 				for (type--; type >= 0; type--)
6356 					dquot_quota_off(sb, type);
6357 
6358 				return err;
6359 			}
6360 		}
6361 	}
6362 	return 0;
6363 }
6364 
6365 static int ext4_quota_off(struct super_block *sb, int type)
6366 {
6367 	struct inode *inode = sb_dqopt(sb)->files[type];
6368 	handle_t *handle;
6369 	int err;
6370 
6371 	/* Force all delayed allocation blocks to be allocated.
6372 	 * Caller already holds s_umount sem */
6373 	if (test_opt(sb, DELALLOC))
6374 		sync_filesystem(sb);
6375 
6376 	if (!inode || !igrab(inode))
6377 		goto out;
6378 
6379 	err = dquot_quota_off(sb, type);
6380 	if (err || ext4_has_feature_quota(sb))
6381 		goto out_put;
6382 
6383 	inode_lock(inode);
6384 	/*
6385 	 * Update modification times of quota files when userspace can
6386 	 * start looking at them. If we fail, we return success anyway since
6387 	 * this is not a hard failure and quotas are already disabled.
6388 	 */
6389 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6390 	if (IS_ERR(handle)) {
6391 		err = PTR_ERR(handle);
6392 		goto out_unlock;
6393 	}
6394 	EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6395 	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6396 	inode->i_mtime = inode->i_ctime = current_time(inode);
6397 	err = ext4_mark_inode_dirty(handle, inode);
6398 	ext4_journal_stop(handle);
6399 out_unlock:
6400 	inode_unlock(inode);
6401 out_put:
6402 	lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6403 	iput(inode);
6404 	return err;
6405 out:
6406 	return dquot_quota_off(sb, type);
6407 }
6408 
6409 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6410  * acquiring the locks... As quota files are never truncated and quota code
6411  * itself serializes the operations (and no one else should touch the files)
6412  * we don't have to be afraid of races */
6413 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6414 			       size_t len, loff_t off)
6415 {
6416 	struct inode *inode = sb_dqopt(sb)->files[type];
6417 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6418 	int offset = off & (sb->s_blocksize - 1);
6419 	int tocopy;
6420 	size_t toread;
6421 	struct buffer_head *bh;
6422 	loff_t i_size = i_size_read(inode);
6423 
6424 	if (off > i_size)
6425 		return 0;
6426 	if (off+len > i_size)
6427 		len = i_size-off;
6428 	toread = len;
6429 	while (toread > 0) {
6430 		tocopy = sb->s_blocksize - offset < toread ?
6431 				sb->s_blocksize - offset : toread;
6432 		bh = ext4_bread(NULL, inode, blk, 0);
6433 		if (IS_ERR(bh))
6434 			return PTR_ERR(bh);
6435 		if (!bh)	/* A hole? */
6436 			memset(data, 0, tocopy);
6437 		else
6438 			memcpy(data, bh->b_data+offset, tocopy);
6439 		brelse(bh);
6440 		offset = 0;
6441 		toread -= tocopy;
6442 		data += tocopy;
6443 		blk++;
6444 	}
6445 	return len;
6446 }
6447 
6448 /* Write to quotafile (we know the transaction is already started and has
6449  * enough credits) */
6450 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6451 				const char *data, size_t len, loff_t off)
6452 {
6453 	struct inode *inode = sb_dqopt(sb)->files[type];
6454 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6455 	int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6456 	int retries = 0;
6457 	struct buffer_head *bh;
6458 	handle_t *handle = journal_current_handle();
6459 
6460 	if (EXT4_SB(sb)->s_journal && !handle) {
6461 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6462 			" cancelled because transaction is not started",
6463 			(unsigned long long)off, (unsigned long long)len);
6464 		return -EIO;
6465 	}
6466 	/*
6467 	 * Since we account only one data block in transaction credits,
6468 	 * then it is impossible to cross a block boundary.
6469 	 */
6470 	if (sb->s_blocksize - offset < len) {
6471 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6472 			" cancelled because not block aligned",
6473 			(unsigned long long)off, (unsigned long long)len);
6474 		return -EIO;
6475 	}
6476 
6477 	do {
6478 		bh = ext4_bread(handle, inode, blk,
6479 				EXT4_GET_BLOCKS_CREATE |
6480 				EXT4_GET_BLOCKS_METADATA_NOFAIL);
6481 	} while (PTR_ERR(bh) == -ENOSPC &&
6482 		 ext4_should_retry_alloc(inode->i_sb, &retries));
6483 	if (IS_ERR(bh))
6484 		return PTR_ERR(bh);
6485 	if (!bh)
6486 		goto out;
6487 	BUFFER_TRACE(bh, "get write access");
6488 	err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
6489 	if (err) {
6490 		brelse(bh);
6491 		return err;
6492 	}
6493 	lock_buffer(bh);
6494 	memcpy(bh->b_data+offset, data, len);
6495 	flush_dcache_page(bh->b_page);
6496 	unlock_buffer(bh);
6497 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
6498 	brelse(bh);
6499 out:
6500 	if (inode->i_size < off + len) {
6501 		i_size_write(inode, off + len);
6502 		EXT4_I(inode)->i_disksize = inode->i_size;
6503 		err2 = ext4_mark_inode_dirty(handle, inode);
6504 		if (unlikely(err2 && !err))
6505 			err = err2;
6506 	}
6507 	return err ? err : len;
6508 }
6509 #endif
6510 
6511 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6512 		       const char *dev_name, void *data)
6513 {
6514 	return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6515 }
6516 
6517 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6518 static inline void register_as_ext2(void)
6519 {
6520 	int err = register_filesystem(&ext2_fs_type);
6521 	if (err)
6522 		printk(KERN_WARNING
6523 		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6524 }
6525 
6526 static inline void unregister_as_ext2(void)
6527 {
6528 	unregister_filesystem(&ext2_fs_type);
6529 }
6530 
6531 static inline int ext2_feature_set_ok(struct super_block *sb)
6532 {
6533 	if (ext4_has_unknown_ext2_incompat_features(sb))
6534 		return 0;
6535 	if (sb_rdonly(sb))
6536 		return 1;
6537 	if (ext4_has_unknown_ext2_ro_compat_features(sb))
6538 		return 0;
6539 	return 1;
6540 }
6541 #else
6542 static inline void register_as_ext2(void) { }
6543 static inline void unregister_as_ext2(void) { }
6544 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6545 #endif
6546 
6547 static inline void register_as_ext3(void)
6548 {
6549 	int err = register_filesystem(&ext3_fs_type);
6550 	if (err)
6551 		printk(KERN_WARNING
6552 		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6553 }
6554 
6555 static inline void unregister_as_ext3(void)
6556 {
6557 	unregister_filesystem(&ext3_fs_type);
6558 }
6559 
6560 static inline int ext3_feature_set_ok(struct super_block *sb)
6561 {
6562 	if (ext4_has_unknown_ext3_incompat_features(sb))
6563 		return 0;
6564 	if (!ext4_has_feature_journal(sb))
6565 		return 0;
6566 	if (sb_rdonly(sb))
6567 		return 1;
6568 	if (ext4_has_unknown_ext3_ro_compat_features(sb))
6569 		return 0;
6570 	return 1;
6571 }
6572 
6573 static struct file_system_type ext4_fs_type = {
6574 	.owner		= THIS_MODULE,
6575 	.name		= "ext4",
6576 	.mount		= ext4_mount,
6577 	.kill_sb	= kill_block_super,
6578 	.fs_flags	= FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
6579 };
6580 MODULE_ALIAS_FS("ext4");
6581 
6582 /* Shared across all ext4 file systems */
6583 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6584 
6585 static int __init ext4_init_fs(void)
6586 {
6587 	int i, err;
6588 
6589 	ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6590 	ext4_li_info = NULL;
6591 
6592 	/* Build-time check for flags consistency */
6593 	ext4_check_flag_values();
6594 
6595 	for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6596 		init_waitqueue_head(&ext4__ioend_wq[i]);
6597 
6598 	err = ext4_init_es();
6599 	if (err)
6600 		return err;
6601 
6602 	err = ext4_init_pending();
6603 	if (err)
6604 		goto out7;
6605 
6606 	err = ext4_init_post_read_processing();
6607 	if (err)
6608 		goto out6;
6609 
6610 	err = ext4_init_pageio();
6611 	if (err)
6612 		goto out5;
6613 
6614 	err = ext4_init_system_zone();
6615 	if (err)
6616 		goto out4;
6617 
6618 	err = ext4_init_sysfs();
6619 	if (err)
6620 		goto out3;
6621 
6622 	err = ext4_init_mballoc();
6623 	if (err)
6624 		goto out2;
6625 	err = init_inodecache();
6626 	if (err)
6627 		goto out1;
6628 
6629 	err = ext4_fc_init_dentry_cache();
6630 	if (err)
6631 		goto out05;
6632 
6633 	register_as_ext3();
6634 	register_as_ext2();
6635 	err = register_filesystem(&ext4_fs_type);
6636 	if (err)
6637 		goto out;
6638 
6639 	return 0;
6640 out:
6641 	unregister_as_ext2();
6642 	unregister_as_ext3();
6643 out05:
6644 	destroy_inodecache();
6645 out1:
6646 	ext4_exit_mballoc();
6647 out2:
6648 	ext4_exit_sysfs();
6649 out3:
6650 	ext4_exit_system_zone();
6651 out4:
6652 	ext4_exit_pageio();
6653 out5:
6654 	ext4_exit_post_read_processing();
6655 out6:
6656 	ext4_exit_pending();
6657 out7:
6658 	ext4_exit_es();
6659 
6660 	return err;
6661 }
6662 
6663 static void __exit ext4_exit_fs(void)
6664 {
6665 	ext4_destroy_lazyinit_thread();
6666 	unregister_as_ext2();
6667 	unregister_as_ext3();
6668 	unregister_filesystem(&ext4_fs_type);
6669 	destroy_inodecache();
6670 	ext4_exit_mballoc();
6671 	ext4_exit_sysfs();
6672 	ext4_exit_system_zone();
6673 	ext4_exit_pageio();
6674 	ext4_exit_post_read_processing();
6675 	ext4_exit_es();
6676 	ext4_exit_pending();
6677 }
6678 
6679 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6680 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6681 MODULE_LICENSE("GPL");
6682 MODULE_SOFTDEP("pre: crc32c");
6683 module_init(ext4_init_fs)
6684 module_exit(ext4_exit_fs)
6685