xref: /openbmc/linux/fs/ext4/super.c (revision a9a08845)
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 
46 #include <linux/kthread.h>
47 #include <linux/freezer.h>
48 
49 #include "ext4.h"
50 #include "ext4_extents.h"	/* Needed for trace points definition */
51 #include "ext4_jbd2.h"
52 #include "xattr.h"
53 #include "acl.h"
54 #include "mballoc.h"
55 #include "fsmap.h"
56 
57 #define CREATE_TRACE_POINTS
58 #include <trace/events/ext4.h>
59 
60 static struct ext4_lazy_init *ext4_li_info;
61 static struct mutex ext4_li_mtx;
62 static struct ratelimit_state ext4_mount_msg_ratelimit;
63 
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 			     unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69 					struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71 				   struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 		       const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static struct inode *ext4_get_journal_inode(struct super_block *sb,
86 					    unsigned int journal_inum);
87 
88 /*
89  * Lock ordering
90  *
91  * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
92  * i_mmap_rwsem (inode->i_mmap_rwsem)!
93  *
94  * page fault path:
95  * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
96  *   page lock -> i_data_sem (rw)
97  *
98  * buffered write path:
99  * sb_start_write -> i_mutex -> mmap_sem
100  * sb_start_write -> i_mutex -> transaction start -> page lock ->
101  *   i_data_sem (rw)
102  *
103  * truncate:
104  * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
105  *   i_mmap_rwsem (w) -> page lock
106  * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
107  *   transaction start -> i_data_sem (rw)
108  *
109  * direct IO:
110  * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
111  * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
112  *   transaction start -> i_data_sem (rw)
113  *
114  * writepages:
115  * transaction start -> page lock(s) -> i_data_sem (rw)
116  */
117 
118 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
119 static struct file_system_type ext2_fs_type = {
120 	.owner		= THIS_MODULE,
121 	.name		= "ext2",
122 	.mount		= ext4_mount,
123 	.kill_sb	= kill_block_super,
124 	.fs_flags	= FS_REQUIRES_DEV,
125 };
126 MODULE_ALIAS_FS("ext2");
127 MODULE_ALIAS("ext2");
128 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
129 #else
130 #define IS_EXT2_SB(sb) (0)
131 #endif
132 
133 
134 static struct file_system_type ext3_fs_type = {
135 	.owner		= THIS_MODULE,
136 	.name		= "ext3",
137 	.mount		= ext4_mount,
138 	.kill_sb	= kill_block_super,
139 	.fs_flags	= FS_REQUIRES_DEV,
140 };
141 MODULE_ALIAS_FS("ext3");
142 MODULE_ALIAS("ext3");
143 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
144 
145 static int ext4_verify_csum_type(struct super_block *sb,
146 				 struct ext4_super_block *es)
147 {
148 	if (!ext4_has_feature_metadata_csum(sb))
149 		return 1;
150 
151 	return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
152 }
153 
154 static __le32 ext4_superblock_csum(struct super_block *sb,
155 				   struct ext4_super_block *es)
156 {
157 	struct ext4_sb_info *sbi = EXT4_SB(sb);
158 	int offset = offsetof(struct ext4_super_block, s_checksum);
159 	__u32 csum;
160 
161 	csum = ext4_chksum(sbi, ~0, (char *)es, offset);
162 
163 	return cpu_to_le32(csum);
164 }
165 
166 static int ext4_superblock_csum_verify(struct super_block *sb,
167 				       struct ext4_super_block *es)
168 {
169 	if (!ext4_has_metadata_csum(sb))
170 		return 1;
171 
172 	return es->s_checksum == ext4_superblock_csum(sb, es);
173 }
174 
175 void ext4_superblock_csum_set(struct super_block *sb)
176 {
177 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
178 
179 	if (!ext4_has_metadata_csum(sb))
180 		return;
181 
182 	es->s_checksum = ext4_superblock_csum(sb, es);
183 }
184 
185 void *ext4_kvmalloc(size_t size, gfp_t flags)
186 {
187 	void *ret;
188 
189 	ret = kmalloc(size, flags | __GFP_NOWARN);
190 	if (!ret)
191 		ret = __vmalloc(size, flags, PAGE_KERNEL);
192 	return ret;
193 }
194 
195 void *ext4_kvzalloc(size_t size, gfp_t flags)
196 {
197 	void *ret;
198 
199 	ret = kzalloc(size, flags | __GFP_NOWARN);
200 	if (!ret)
201 		ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
202 	return ret;
203 }
204 
205 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
206 			       struct ext4_group_desc *bg)
207 {
208 	return le32_to_cpu(bg->bg_block_bitmap_lo) |
209 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
210 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
211 }
212 
213 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
214 			       struct ext4_group_desc *bg)
215 {
216 	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
217 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
218 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
219 }
220 
221 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
222 			      struct ext4_group_desc *bg)
223 {
224 	return le32_to_cpu(bg->bg_inode_table_lo) |
225 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
226 		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
227 }
228 
229 __u32 ext4_free_group_clusters(struct super_block *sb,
230 			       struct ext4_group_desc *bg)
231 {
232 	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
233 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
234 		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
235 }
236 
237 __u32 ext4_free_inodes_count(struct super_block *sb,
238 			      struct ext4_group_desc *bg)
239 {
240 	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
241 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
242 		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
243 }
244 
245 __u32 ext4_used_dirs_count(struct super_block *sb,
246 			      struct ext4_group_desc *bg)
247 {
248 	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
249 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
250 		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
251 }
252 
253 __u32 ext4_itable_unused_count(struct super_block *sb,
254 			      struct ext4_group_desc *bg)
255 {
256 	return le16_to_cpu(bg->bg_itable_unused_lo) |
257 		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
258 		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
259 }
260 
261 void ext4_block_bitmap_set(struct super_block *sb,
262 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
263 {
264 	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
265 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
266 		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
267 }
268 
269 void ext4_inode_bitmap_set(struct super_block *sb,
270 			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
271 {
272 	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
273 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
274 		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
275 }
276 
277 void ext4_inode_table_set(struct super_block *sb,
278 			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
279 {
280 	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
281 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
282 		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
283 }
284 
285 void ext4_free_group_clusters_set(struct super_block *sb,
286 				  struct ext4_group_desc *bg, __u32 count)
287 {
288 	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
289 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
290 		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
291 }
292 
293 void ext4_free_inodes_set(struct super_block *sb,
294 			  struct ext4_group_desc *bg, __u32 count)
295 {
296 	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
297 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
298 		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
299 }
300 
301 void ext4_used_dirs_set(struct super_block *sb,
302 			  struct ext4_group_desc *bg, __u32 count)
303 {
304 	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
305 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
306 		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
307 }
308 
309 void ext4_itable_unused_set(struct super_block *sb,
310 			  struct ext4_group_desc *bg, __u32 count)
311 {
312 	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
313 	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
314 		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
315 }
316 
317 
318 static void __save_error_info(struct super_block *sb, const char *func,
319 			    unsigned int line)
320 {
321 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
322 
323 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
324 	if (bdev_read_only(sb->s_bdev))
325 		return;
326 	es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
327 	es->s_last_error_time = cpu_to_le32(get_seconds());
328 	strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
329 	es->s_last_error_line = cpu_to_le32(line);
330 	if (!es->s_first_error_time) {
331 		es->s_first_error_time = es->s_last_error_time;
332 		strncpy(es->s_first_error_func, func,
333 			sizeof(es->s_first_error_func));
334 		es->s_first_error_line = cpu_to_le32(line);
335 		es->s_first_error_ino = es->s_last_error_ino;
336 		es->s_first_error_block = es->s_last_error_block;
337 	}
338 	/*
339 	 * Start the daily error reporting function if it hasn't been
340 	 * started already
341 	 */
342 	if (!es->s_error_count)
343 		mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
344 	le32_add_cpu(&es->s_error_count, 1);
345 }
346 
347 static void save_error_info(struct super_block *sb, const char *func,
348 			    unsigned int line)
349 {
350 	__save_error_info(sb, func, line);
351 	ext4_commit_super(sb, 1);
352 }
353 
354 /*
355  * The del_gendisk() function uninitializes the disk-specific data
356  * structures, including the bdi structure, without telling anyone
357  * else.  Once this happens, any attempt to call mark_buffer_dirty()
358  * (for example, by ext4_commit_super), will cause a kernel OOPS.
359  * This is a kludge to prevent these oops until we can put in a proper
360  * hook in del_gendisk() to inform the VFS and file system layers.
361  */
362 static int block_device_ejected(struct super_block *sb)
363 {
364 	struct inode *bd_inode = sb->s_bdev->bd_inode;
365 	struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
366 
367 	return bdi->dev == NULL;
368 }
369 
370 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
371 {
372 	struct super_block		*sb = journal->j_private;
373 	struct ext4_sb_info		*sbi = EXT4_SB(sb);
374 	int				error = is_journal_aborted(journal);
375 	struct ext4_journal_cb_entry	*jce;
376 
377 	BUG_ON(txn->t_state == T_FINISHED);
378 
379 	ext4_process_freed_data(sb, txn->t_tid);
380 
381 	spin_lock(&sbi->s_md_lock);
382 	while (!list_empty(&txn->t_private_list)) {
383 		jce = list_entry(txn->t_private_list.next,
384 				 struct ext4_journal_cb_entry, jce_list);
385 		list_del_init(&jce->jce_list);
386 		spin_unlock(&sbi->s_md_lock);
387 		jce->jce_func(sb, jce, error);
388 		spin_lock(&sbi->s_md_lock);
389 	}
390 	spin_unlock(&sbi->s_md_lock);
391 }
392 
393 /* Deal with the reporting of failure conditions on a filesystem such as
394  * inconsistencies detected or read IO failures.
395  *
396  * On ext2, we can store the error state of the filesystem in the
397  * superblock.  That is not possible on ext4, because we may have other
398  * write ordering constraints on the superblock which prevent us from
399  * writing it out straight away; and given that the journal is about to
400  * be aborted, we can't rely on the current, or future, transactions to
401  * write out the superblock safely.
402  *
403  * We'll just use the jbd2_journal_abort() error code to record an error in
404  * the journal instead.  On recovery, the journal will complain about
405  * that error until we've noted it down and cleared it.
406  */
407 
408 static void ext4_handle_error(struct super_block *sb)
409 {
410 	if (sb_rdonly(sb))
411 		return;
412 
413 	if (!test_opt(sb, ERRORS_CONT)) {
414 		journal_t *journal = EXT4_SB(sb)->s_journal;
415 
416 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
417 		if (journal)
418 			jbd2_journal_abort(journal, -EIO);
419 	}
420 	if (test_opt(sb, ERRORS_RO)) {
421 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
422 		/*
423 		 * Make sure updated value of ->s_mount_flags will be visible
424 		 * before ->s_flags update
425 		 */
426 		smp_wmb();
427 		sb->s_flags |= SB_RDONLY;
428 	}
429 	if (test_opt(sb, ERRORS_PANIC)) {
430 		if (EXT4_SB(sb)->s_journal &&
431 		  !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
432 			return;
433 		panic("EXT4-fs (device %s): panic forced after error\n",
434 			sb->s_id);
435 	}
436 }
437 
438 #define ext4_error_ratelimit(sb)					\
439 		___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),	\
440 			     "EXT4-fs error")
441 
442 void __ext4_error(struct super_block *sb, const char *function,
443 		  unsigned int line, const char *fmt, ...)
444 {
445 	struct va_format vaf;
446 	va_list args;
447 
448 	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
449 		return;
450 
451 	if (ext4_error_ratelimit(sb)) {
452 		va_start(args, fmt);
453 		vaf.fmt = fmt;
454 		vaf.va = &args;
455 		printk(KERN_CRIT
456 		       "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
457 		       sb->s_id, function, line, current->comm, &vaf);
458 		va_end(args);
459 	}
460 	save_error_info(sb, function, line);
461 	ext4_handle_error(sb);
462 }
463 
464 void __ext4_error_inode(struct inode *inode, const char *function,
465 			unsigned int line, ext4_fsblk_t block,
466 			const char *fmt, ...)
467 {
468 	va_list args;
469 	struct va_format vaf;
470 	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
471 
472 	if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
473 		return;
474 
475 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
476 	es->s_last_error_block = cpu_to_le64(block);
477 	if (ext4_error_ratelimit(inode->i_sb)) {
478 		va_start(args, fmt);
479 		vaf.fmt = fmt;
480 		vaf.va = &args;
481 		if (block)
482 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
483 			       "inode #%lu: block %llu: comm %s: %pV\n",
484 			       inode->i_sb->s_id, function, line, inode->i_ino,
485 			       block, current->comm, &vaf);
486 		else
487 			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
488 			       "inode #%lu: comm %s: %pV\n",
489 			       inode->i_sb->s_id, function, line, inode->i_ino,
490 			       current->comm, &vaf);
491 		va_end(args);
492 	}
493 	save_error_info(inode->i_sb, function, line);
494 	ext4_handle_error(inode->i_sb);
495 }
496 
497 void __ext4_error_file(struct file *file, const char *function,
498 		       unsigned int line, ext4_fsblk_t block,
499 		       const char *fmt, ...)
500 {
501 	va_list args;
502 	struct va_format vaf;
503 	struct ext4_super_block *es;
504 	struct inode *inode = file_inode(file);
505 	char pathname[80], *path;
506 
507 	if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
508 		return;
509 
510 	es = EXT4_SB(inode->i_sb)->s_es;
511 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
512 	if (ext4_error_ratelimit(inode->i_sb)) {
513 		path = file_path(file, pathname, sizeof(pathname));
514 		if (IS_ERR(path))
515 			path = "(unknown)";
516 		va_start(args, fmt);
517 		vaf.fmt = fmt;
518 		vaf.va = &args;
519 		if (block)
520 			printk(KERN_CRIT
521 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
522 			       "block %llu: comm %s: path %s: %pV\n",
523 			       inode->i_sb->s_id, function, line, inode->i_ino,
524 			       block, current->comm, path, &vaf);
525 		else
526 			printk(KERN_CRIT
527 			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
528 			       "comm %s: path %s: %pV\n",
529 			       inode->i_sb->s_id, function, line, inode->i_ino,
530 			       current->comm, path, &vaf);
531 		va_end(args);
532 	}
533 	save_error_info(inode->i_sb, function, line);
534 	ext4_handle_error(inode->i_sb);
535 }
536 
537 const char *ext4_decode_error(struct super_block *sb, int errno,
538 			      char nbuf[16])
539 {
540 	char *errstr = NULL;
541 
542 	switch (errno) {
543 	case -EFSCORRUPTED:
544 		errstr = "Corrupt filesystem";
545 		break;
546 	case -EFSBADCRC:
547 		errstr = "Filesystem failed CRC";
548 		break;
549 	case -EIO:
550 		errstr = "IO failure";
551 		break;
552 	case -ENOMEM:
553 		errstr = "Out of memory";
554 		break;
555 	case -EROFS:
556 		if (!sb || (EXT4_SB(sb)->s_journal &&
557 			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
558 			errstr = "Journal has aborted";
559 		else
560 			errstr = "Readonly filesystem";
561 		break;
562 	default:
563 		/* If the caller passed in an extra buffer for unknown
564 		 * errors, textualise them now.  Else we just return
565 		 * NULL. */
566 		if (nbuf) {
567 			/* Check for truncated error codes... */
568 			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
569 				errstr = nbuf;
570 		}
571 		break;
572 	}
573 
574 	return errstr;
575 }
576 
577 /* __ext4_std_error decodes expected errors from journaling functions
578  * automatically and invokes the appropriate error response.  */
579 
580 void __ext4_std_error(struct super_block *sb, const char *function,
581 		      unsigned int line, int errno)
582 {
583 	char nbuf[16];
584 	const char *errstr;
585 
586 	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
587 		return;
588 
589 	/* Special case: if the error is EROFS, and we're not already
590 	 * inside a transaction, then there's really no point in logging
591 	 * an error. */
592 	if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
593 		return;
594 
595 	if (ext4_error_ratelimit(sb)) {
596 		errstr = ext4_decode_error(sb, errno, nbuf);
597 		printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
598 		       sb->s_id, function, line, errstr);
599 	}
600 
601 	save_error_info(sb, function, line);
602 	ext4_handle_error(sb);
603 }
604 
605 /*
606  * ext4_abort is a much stronger failure handler than ext4_error.  The
607  * abort function may be used to deal with unrecoverable failures such
608  * as journal IO errors or ENOMEM at a critical moment in log management.
609  *
610  * We unconditionally force the filesystem into an ABORT|READONLY state,
611  * unless the error response on the fs has been set to panic in which
612  * case we take the easy way out and panic immediately.
613  */
614 
615 void __ext4_abort(struct super_block *sb, const char *function,
616 		unsigned int line, const char *fmt, ...)
617 {
618 	struct va_format vaf;
619 	va_list args;
620 
621 	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
622 		return;
623 
624 	save_error_info(sb, function, line);
625 	va_start(args, fmt);
626 	vaf.fmt = fmt;
627 	vaf.va = &args;
628 	printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
629 	       sb->s_id, function, line, &vaf);
630 	va_end(args);
631 
632 	if (sb_rdonly(sb) == 0) {
633 		ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
634 		EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
635 		/*
636 		 * Make sure updated value of ->s_mount_flags will be visible
637 		 * before ->s_flags update
638 		 */
639 		smp_wmb();
640 		sb->s_flags |= SB_RDONLY;
641 		if (EXT4_SB(sb)->s_journal)
642 			jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
643 		save_error_info(sb, function, line);
644 	}
645 	if (test_opt(sb, ERRORS_PANIC)) {
646 		if (EXT4_SB(sb)->s_journal &&
647 		  !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
648 			return;
649 		panic("EXT4-fs panic from previous error\n");
650 	}
651 }
652 
653 void __ext4_msg(struct super_block *sb,
654 		const char *prefix, const char *fmt, ...)
655 {
656 	struct va_format vaf;
657 	va_list args;
658 
659 	if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
660 		return;
661 
662 	va_start(args, fmt);
663 	vaf.fmt = fmt;
664 	vaf.va = &args;
665 	printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
666 	va_end(args);
667 }
668 
669 #define ext4_warning_ratelimit(sb)					\
670 		___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),	\
671 			     "EXT4-fs warning")
672 
673 void __ext4_warning(struct super_block *sb, const char *function,
674 		    unsigned int line, const char *fmt, ...)
675 {
676 	struct va_format vaf;
677 	va_list args;
678 
679 	if (!ext4_warning_ratelimit(sb))
680 		return;
681 
682 	va_start(args, fmt);
683 	vaf.fmt = fmt;
684 	vaf.va = &args;
685 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
686 	       sb->s_id, function, line, &vaf);
687 	va_end(args);
688 }
689 
690 void __ext4_warning_inode(const struct inode *inode, const char *function,
691 			  unsigned int line, const char *fmt, ...)
692 {
693 	struct va_format vaf;
694 	va_list args;
695 
696 	if (!ext4_warning_ratelimit(inode->i_sb))
697 		return;
698 
699 	va_start(args, fmt);
700 	vaf.fmt = fmt;
701 	vaf.va = &args;
702 	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
703 	       "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
704 	       function, line, inode->i_ino, current->comm, &vaf);
705 	va_end(args);
706 }
707 
708 void __ext4_grp_locked_error(const char *function, unsigned int line,
709 			     struct super_block *sb, ext4_group_t grp,
710 			     unsigned long ino, ext4_fsblk_t block,
711 			     const char *fmt, ...)
712 __releases(bitlock)
713 __acquires(bitlock)
714 {
715 	struct va_format vaf;
716 	va_list args;
717 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
718 
719 	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
720 		return;
721 
722 	es->s_last_error_ino = cpu_to_le32(ino);
723 	es->s_last_error_block = cpu_to_le64(block);
724 	__save_error_info(sb, function, line);
725 
726 	if (ext4_error_ratelimit(sb)) {
727 		va_start(args, fmt);
728 		vaf.fmt = fmt;
729 		vaf.va = &args;
730 		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
731 		       sb->s_id, function, line, grp);
732 		if (ino)
733 			printk(KERN_CONT "inode %lu: ", ino);
734 		if (block)
735 			printk(KERN_CONT "block %llu:",
736 			       (unsigned long long) block);
737 		printk(KERN_CONT "%pV\n", &vaf);
738 		va_end(args);
739 	}
740 
741 	if (test_opt(sb, ERRORS_CONT)) {
742 		ext4_commit_super(sb, 0);
743 		return;
744 	}
745 
746 	ext4_unlock_group(sb, grp);
747 	ext4_commit_super(sb, 1);
748 	ext4_handle_error(sb);
749 	/*
750 	 * We only get here in the ERRORS_RO case; relocking the group
751 	 * may be dangerous, but nothing bad will happen since the
752 	 * filesystem will have already been marked read/only and the
753 	 * journal has been aborted.  We return 1 as a hint to callers
754 	 * who might what to use the return value from
755 	 * ext4_grp_locked_error() to distinguish between the
756 	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
757 	 * aggressively from the ext4 function in question, with a
758 	 * more appropriate error code.
759 	 */
760 	ext4_lock_group(sb, grp);
761 	return;
762 }
763 
764 void ext4_update_dynamic_rev(struct super_block *sb)
765 {
766 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
767 
768 	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
769 		return;
770 
771 	ext4_warning(sb,
772 		     "updating to rev %d because of new feature flag, "
773 		     "running e2fsck is recommended",
774 		     EXT4_DYNAMIC_REV);
775 
776 	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
777 	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
778 	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
779 	/* leave es->s_feature_*compat flags alone */
780 	/* es->s_uuid will be set by e2fsck if empty */
781 
782 	/*
783 	 * The rest of the superblock fields should be zero, and if not it
784 	 * means they are likely already in use, so leave them alone.  We
785 	 * can leave it up to e2fsck to clean up any inconsistencies there.
786 	 */
787 }
788 
789 /*
790  * Open the external journal device
791  */
792 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
793 {
794 	struct block_device *bdev;
795 	char b[BDEVNAME_SIZE];
796 
797 	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
798 	if (IS_ERR(bdev))
799 		goto fail;
800 	return bdev;
801 
802 fail:
803 	ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
804 			__bdevname(dev, b), PTR_ERR(bdev));
805 	return NULL;
806 }
807 
808 /*
809  * Release the journal device
810  */
811 static void ext4_blkdev_put(struct block_device *bdev)
812 {
813 	blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
814 }
815 
816 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
817 {
818 	struct block_device *bdev;
819 	bdev = sbi->journal_bdev;
820 	if (bdev) {
821 		ext4_blkdev_put(bdev);
822 		sbi->journal_bdev = NULL;
823 	}
824 }
825 
826 static inline struct inode *orphan_list_entry(struct list_head *l)
827 {
828 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
829 }
830 
831 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
832 {
833 	struct list_head *l;
834 
835 	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
836 		 le32_to_cpu(sbi->s_es->s_last_orphan));
837 
838 	printk(KERN_ERR "sb_info orphan list:\n");
839 	list_for_each(l, &sbi->s_orphan) {
840 		struct inode *inode = orphan_list_entry(l);
841 		printk(KERN_ERR "  "
842 		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
843 		       inode->i_sb->s_id, inode->i_ino, inode,
844 		       inode->i_mode, inode->i_nlink,
845 		       NEXT_ORPHAN(inode));
846 	}
847 }
848 
849 #ifdef CONFIG_QUOTA
850 static int ext4_quota_off(struct super_block *sb, int type);
851 
852 static inline void ext4_quota_off_umount(struct super_block *sb)
853 {
854 	int type;
855 
856 	/* Use our quota_off function to clear inode flags etc. */
857 	for (type = 0; type < EXT4_MAXQUOTAS; type++)
858 		ext4_quota_off(sb, type);
859 }
860 #else
861 static inline void ext4_quota_off_umount(struct super_block *sb)
862 {
863 }
864 #endif
865 
866 static void ext4_put_super(struct super_block *sb)
867 {
868 	struct ext4_sb_info *sbi = EXT4_SB(sb);
869 	struct ext4_super_block *es = sbi->s_es;
870 	int aborted = 0;
871 	int i, err;
872 
873 	ext4_unregister_li_request(sb);
874 	ext4_quota_off_umount(sb);
875 
876 	destroy_workqueue(sbi->rsv_conversion_wq);
877 
878 	if (sbi->s_journal) {
879 		aborted = is_journal_aborted(sbi->s_journal);
880 		err = jbd2_journal_destroy(sbi->s_journal);
881 		sbi->s_journal = NULL;
882 		if ((err < 0) && !aborted)
883 			ext4_abort(sb, "Couldn't clean up the journal");
884 	}
885 
886 	ext4_unregister_sysfs(sb);
887 	ext4_es_unregister_shrinker(sbi);
888 	del_timer_sync(&sbi->s_err_report);
889 	ext4_release_system_zone(sb);
890 	ext4_mb_release(sb);
891 	ext4_ext_release(sb);
892 
893 	if (!sb_rdonly(sb) && !aborted) {
894 		ext4_clear_feature_journal_needs_recovery(sb);
895 		es->s_state = cpu_to_le16(sbi->s_mount_state);
896 	}
897 	if (!sb_rdonly(sb))
898 		ext4_commit_super(sb, 1);
899 
900 	for (i = 0; i < sbi->s_gdb_count; i++)
901 		brelse(sbi->s_group_desc[i]);
902 	kvfree(sbi->s_group_desc);
903 	kvfree(sbi->s_flex_groups);
904 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
905 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
906 	percpu_counter_destroy(&sbi->s_dirs_counter);
907 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
908 	percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
909 #ifdef CONFIG_QUOTA
910 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
911 		kfree(sbi->s_qf_names[i]);
912 #endif
913 
914 	/* Debugging code just in case the in-memory inode orphan list
915 	 * isn't empty.  The on-disk one can be non-empty if we've
916 	 * detected an error and taken the fs readonly, but the
917 	 * in-memory list had better be clean by this point. */
918 	if (!list_empty(&sbi->s_orphan))
919 		dump_orphan_list(sb, sbi);
920 	J_ASSERT(list_empty(&sbi->s_orphan));
921 
922 	sync_blockdev(sb->s_bdev);
923 	invalidate_bdev(sb->s_bdev);
924 	if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
925 		/*
926 		 * Invalidate the journal device's buffers.  We don't want them
927 		 * floating about in memory - the physical journal device may
928 		 * hotswapped, and it breaks the `ro-after' testing code.
929 		 */
930 		sync_blockdev(sbi->journal_bdev);
931 		invalidate_bdev(sbi->journal_bdev);
932 		ext4_blkdev_remove(sbi);
933 	}
934 	if (sbi->s_ea_inode_cache) {
935 		ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
936 		sbi->s_ea_inode_cache = NULL;
937 	}
938 	if (sbi->s_ea_block_cache) {
939 		ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
940 		sbi->s_ea_block_cache = NULL;
941 	}
942 	if (sbi->s_mmp_tsk)
943 		kthread_stop(sbi->s_mmp_tsk);
944 	brelse(sbi->s_sbh);
945 	sb->s_fs_info = NULL;
946 	/*
947 	 * Now that we are completely done shutting down the
948 	 * superblock, we need to actually destroy the kobject.
949 	 */
950 	kobject_put(&sbi->s_kobj);
951 	wait_for_completion(&sbi->s_kobj_unregister);
952 	if (sbi->s_chksum_driver)
953 		crypto_free_shash(sbi->s_chksum_driver);
954 	kfree(sbi->s_blockgroup_lock);
955 	fs_put_dax(sbi->s_daxdev);
956 	kfree(sbi);
957 }
958 
959 static struct kmem_cache *ext4_inode_cachep;
960 
961 /*
962  * Called inside transaction, so use GFP_NOFS
963  */
964 static struct inode *ext4_alloc_inode(struct super_block *sb)
965 {
966 	struct ext4_inode_info *ei;
967 
968 	ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
969 	if (!ei)
970 		return NULL;
971 
972 	inode_set_iversion(&ei->vfs_inode, 1);
973 	spin_lock_init(&ei->i_raw_lock);
974 	INIT_LIST_HEAD(&ei->i_prealloc_list);
975 	spin_lock_init(&ei->i_prealloc_lock);
976 	ext4_es_init_tree(&ei->i_es_tree);
977 	rwlock_init(&ei->i_es_lock);
978 	INIT_LIST_HEAD(&ei->i_es_list);
979 	ei->i_es_all_nr = 0;
980 	ei->i_es_shk_nr = 0;
981 	ei->i_es_shrink_lblk = 0;
982 	ei->i_reserved_data_blocks = 0;
983 	ei->i_da_metadata_calc_len = 0;
984 	ei->i_da_metadata_calc_last_lblock = 0;
985 	spin_lock_init(&(ei->i_block_reservation_lock));
986 #ifdef CONFIG_QUOTA
987 	ei->i_reserved_quota = 0;
988 	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
989 #endif
990 	ei->jinode = NULL;
991 	INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
992 	spin_lock_init(&ei->i_completed_io_lock);
993 	ei->i_sync_tid = 0;
994 	ei->i_datasync_tid = 0;
995 	atomic_set(&ei->i_unwritten, 0);
996 	INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
997 	return &ei->vfs_inode;
998 }
999 
1000 static int ext4_drop_inode(struct inode *inode)
1001 {
1002 	int drop = generic_drop_inode(inode);
1003 
1004 	trace_ext4_drop_inode(inode, drop);
1005 	return drop;
1006 }
1007 
1008 static void ext4_i_callback(struct rcu_head *head)
1009 {
1010 	struct inode *inode = container_of(head, struct inode, i_rcu);
1011 	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1012 }
1013 
1014 static void ext4_destroy_inode(struct inode *inode)
1015 {
1016 	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1017 		ext4_msg(inode->i_sb, KERN_ERR,
1018 			 "Inode %lu (%p): orphan list check failed!",
1019 			 inode->i_ino, EXT4_I(inode));
1020 		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1021 				EXT4_I(inode), sizeof(struct ext4_inode_info),
1022 				true);
1023 		dump_stack();
1024 	}
1025 	call_rcu(&inode->i_rcu, ext4_i_callback);
1026 }
1027 
1028 static void init_once(void *foo)
1029 {
1030 	struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1031 
1032 	INIT_LIST_HEAD(&ei->i_orphan);
1033 	init_rwsem(&ei->xattr_sem);
1034 	init_rwsem(&ei->i_data_sem);
1035 	init_rwsem(&ei->i_mmap_sem);
1036 	inode_init_once(&ei->vfs_inode);
1037 }
1038 
1039 static int __init init_inodecache(void)
1040 {
1041 	ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1042 				sizeof(struct ext4_inode_info), 0,
1043 				(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1044 					SLAB_ACCOUNT),
1045 				offsetof(struct ext4_inode_info, i_data),
1046 				sizeof_field(struct ext4_inode_info, i_data),
1047 				init_once);
1048 	if (ext4_inode_cachep == NULL)
1049 		return -ENOMEM;
1050 	return 0;
1051 }
1052 
1053 static void destroy_inodecache(void)
1054 {
1055 	/*
1056 	 * Make sure all delayed rcu free inodes are flushed before we
1057 	 * destroy cache.
1058 	 */
1059 	rcu_barrier();
1060 	kmem_cache_destroy(ext4_inode_cachep);
1061 }
1062 
1063 void ext4_clear_inode(struct inode *inode)
1064 {
1065 	invalidate_inode_buffers(inode);
1066 	clear_inode(inode);
1067 	dquot_drop(inode);
1068 	ext4_discard_preallocations(inode);
1069 	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1070 	if (EXT4_I(inode)->jinode) {
1071 		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1072 					       EXT4_I(inode)->jinode);
1073 		jbd2_free_inode(EXT4_I(inode)->jinode);
1074 		EXT4_I(inode)->jinode = NULL;
1075 	}
1076 	fscrypt_put_encryption_info(inode);
1077 }
1078 
1079 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1080 					u64 ino, u32 generation)
1081 {
1082 	struct inode *inode;
1083 
1084 	if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1085 		return ERR_PTR(-ESTALE);
1086 	if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1087 		return ERR_PTR(-ESTALE);
1088 
1089 	/* iget isn't really right if the inode is currently unallocated!!
1090 	 *
1091 	 * ext4_read_inode will return a bad_inode if the inode had been
1092 	 * deleted, so we should be safe.
1093 	 *
1094 	 * Currently we don't know the generation for parent directory, so
1095 	 * a generation of 0 means "accept any"
1096 	 */
1097 	inode = ext4_iget_normal(sb, ino);
1098 	if (IS_ERR(inode))
1099 		return ERR_CAST(inode);
1100 	if (generation && inode->i_generation != generation) {
1101 		iput(inode);
1102 		return ERR_PTR(-ESTALE);
1103 	}
1104 
1105 	return inode;
1106 }
1107 
1108 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1109 					int fh_len, int fh_type)
1110 {
1111 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1112 				    ext4_nfs_get_inode);
1113 }
1114 
1115 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1116 					int fh_len, int fh_type)
1117 {
1118 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1119 				    ext4_nfs_get_inode);
1120 }
1121 
1122 /*
1123  * Try to release metadata pages (indirect blocks, directories) which are
1124  * mapped via the block device.  Since these pages could have journal heads
1125  * which would prevent try_to_free_buffers() from freeing them, we must use
1126  * jbd2 layer's try_to_free_buffers() function to release them.
1127  */
1128 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1129 				 gfp_t wait)
1130 {
1131 	journal_t *journal = EXT4_SB(sb)->s_journal;
1132 
1133 	WARN_ON(PageChecked(page));
1134 	if (!page_has_buffers(page))
1135 		return 0;
1136 	if (journal)
1137 		return jbd2_journal_try_to_free_buffers(journal, page,
1138 						wait & ~__GFP_DIRECT_RECLAIM);
1139 	return try_to_free_buffers(page);
1140 }
1141 
1142 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1143 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1144 {
1145 	return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1146 				 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1147 }
1148 
1149 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1150 							void *fs_data)
1151 {
1152 	handle_t *handle = fs_data;
1153 	int res, res2, credits, retries = 0;
1154 
1155 	/*
1156 	 * Encrypting the root directory is not allowed because e2fsck expects
1157 	 * lost+found to exist and be unencrypted, and encrypting the root
1158 	 * directory would imply encrypting the lost+found directory as well as
1159 	 * the filename "lost+found" itself.
1160 	 */
1161 	if (inode->i_ino == EXT4_ROOT_INO)
1162 		return -EPERM;
1163 
1164 	if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1165 		return -EINVAL;
1166 
1167 	res = ext4_convert_inline_data(inode);
1168 	if (res)
1169 		return res;
1170 
1171 	/*
1172 	 * If a journal handle was specified, then the encryption context is
1173 	 * being set on a new inode via inheritance and is part of a larger
1174 	 * transaction to create the inode.  Otherwise the encryption context is
1175 	 * being set on an existing inode in its own transaction.  Only in the
1176 	 * latter case should the "retry on ENOSPC" logic be used.
1177 	 */
1178 
1179 	if (handle) {
1180 		res = ext4_xattr_set_handle(handle, inode,
1181 					    EXT4_XATTR_INDEX_ENCRYPTION,
1182 					    EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1183 					    ctx, len, 0);
1184 		if (!res) {
1185 			ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1186 			ext4_clear_inode_state(inode,
1187 					EXT4_STATE_MAY_INLINE_DATA);
1188 			/*
1189 			 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1190 			 * S_DAX may be disabled
1191 			 */
1192 			ext4_set_inode_flags(inode);
1193 		}
1194 		return res;
1195 	}
1196 
1197 	res = dquot_initialize(inode);
1198 	if (res)
1199 		return res;
1200 retry:
1201 	res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1202 				     &credits);
1203 	if (res)
1204 		return res;
1205 
1206 	handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1207 	if (IS_ERR(handle))
1208 		return PTR_ERR(handle);
1209 
1210 	res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1211 				    EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1212 				    ctx, len, 0);
1213 	if (!res) {
1214 		ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1215 		/*
1216 		 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1217 		 * S_DAX may be disabled
1218 		 */
1219 		ext4_set_inode_flags(inode);
1220 		res = ext4_mark_inode_dirty(handle, inode);
1221 		if (res)
1222 			EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1223 	}
1224 	res2 = ext4_journal_stop(handle);
1225 
1226 	if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1227 		goto retry;
1228 	if (!res)
1229 		res = res2;
1230 	return res;
1231 }
1232 
1233 static bool ext4_dummy_context(struct inode *inode)
1234 {
1235 	return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1236 }
1237 
1238 static unsigned ext4_max_namelen(struct inode *inode)
1239 {
1240 	return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1241 		EXT4_NAME_LEN;
1242 }
1243 
1244 static const struct fscrypt_operations ext4_cryptops = {
1245 	.key_prefix		= "ext4:",
1246 	.get_context		= ext4_get_context,
1247 	.set_context		= ext4_set_context,
1248 	.dummy_context		= ext4_dummy_context,
1249 	.empty_dir		= ext4_empty_dir,
1250 	.max_namelen		= ext4_max_namelen,
1251 };
1252 #endif
1253 
1254 #ifdef CONFIG_QUOTA
1255 static const char * const quotatypes[] = INITQFNAMES;
1256 #define QTYPE2NAME(t) (quotatypes[t])
1257 
1258 static int ext4_write_dquot(struct dquot *dquot);
1259 static int ext4_acquire_dquot(struct dquot *dquot);
1260 static int ext4_release_dquot(struct dquot *dquot);
1261 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1262 static int ext4_write_info(struct super_block *sb, int type);
1263 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1264 			 const struct path *path);
1265 static int ext4_quota_on_mount(struct super_block *sb, int type);
1266 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1267 			       size_t len, loff_t off);
1268 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1269 				const char *data, size_t len, loff_t off);
1270 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1271 			     unsigned int flags);
1272 static int ext4_enable_quotas(struct super_block *sb);
1273 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1274 
1275 static struct dquot **ext4_get_dquots(struct inode *inode)
1276 {
1277 	return EXT4_I(inode)->i_dquot;
1278 }
1279 
1280 static const struct dquot_operations ext4_quota_operations = {
1281 	.get_reserved_space	= ext4_get_reserved_space,
1282 	.write_dquot		= ext4_write_dquot,
1283 	.acquire_dquot		= ext4_acquire_dquot,
1284 	.release_dquot		= ext4_release_dquot,
1285 	.mark_dirty		= ext4_mark_dquot_dirty,
1286 	.write_info		= ext4_write_info,
1287 	.alloc_dquot		= dquot_alloc,
1288 	.destroy_dquot		= dquot_destroy,
1289 	.get_projid		= ext4_get_projid,
1290 	.get_inode_usage	= ext4_get_inode_usage,
1291 	.get_next_id		= ext4_get_next_id,
1292 };
1293 
1294 static const struct quotactl_ops ext4_qctl_operations = {
1295 	.quota_on	= ext4_quota_on,
1296 	.quota_off	= ext4_quota_off,
1297 	.quota_sync	= dquot_quota_sync,
1298 	.get_state	= dquot_get_state,
1299 	.set_info	= dquot_set_dqinfo,
1300 	.get_dqblk	= dquot_get_dqblk,
1301 	.set_dqblk	= dquot_set_dqblk,
1302 	.get_nextdqblk	= dquot_get_next_dqblk,
1303 };
1304 #endif
1305 
1306 static const struct super_operations ext4_sops = {
1307 	.alloc_inode	= ext4_alloc_inode,
1308 	.destroy_inode	= ext4_destroy_inode,
1309 	.write_inode	= ext4_write_inode,
1310 	.dirty_inode	= ext4_dirty_inode,
1311 	.drop_inode	= ext4_drop_inode,
1312 	.evict_inode	= ext4_evict_inode,
1313 	.put_super	= ext4_put_super,
1314 	.sync_fs	= ext4_sync_fs,
1315 	.freeze_fs	= ext4_freeze,
1316 	.unfreeze_fs	= ext4_unfreeze,
1317 	.statfs		= ext4_statfs,
1318 	.remount_fs	= ext4_remount,
1319 	.show_options	= ext4_show_options,
1320 #ifdef CONFIG_QUOTA
1321 	.quota_read	= ext4_quota_read,
1322 	.quota_write	= ext4_quota_write,
1323 	.get_dquots	= ext4_get_dquots,
1324 #endif
1325 	.bdev_try_to_free_page = bdev_try_to_free_page,
1326 };
1327 
1328 static const struct export_operations ext4_export_ops = {
1329 	.fh_to_dentry = ext4_fh_to_dentry,
1330 	.fh_to_parent = ext4_fh_to_parent,
1331 	.get_parent = ext4_get_parent,
1332 };
1333 
1334 enum {
1335 	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1336 	Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1337 	Opt_nouid32, Opt_debug, Opt_removed,
1338 	Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1339 	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1340 	Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1341 	Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1342 	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1343 	Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1344 	Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1345 	Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1346 	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1347 	Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1348 	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1349 	Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1350 	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1351 	Opt_inode_readahead_blks, Opt_journal_ioprio,
1352 	Opt_dioread_nolock, Opt_dioread_lock,
1353 	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1354 	Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1355 };
1356 
1357 static const match_table_t tokens = {
1358 	{Opt_bsd_df, "bsddf"},
1359 	{Opt_minix_df, "minixdf"},
1360 	{Opt_grpid, "grpid"},
1361 	{Opt_grpid, "bsdgroups"},
1362 	{Opt_nogrpid, "nogrpid"},
1363 	{Opt_nogrpid, "sysvgroups"},
1364 	{Opt_resgid, "resgid=%u"},
1365 	{Opt_resuid, "resuid=%u"},
1366 	{Opt_sb, "sb=%u"},
1367 	{Opt_err_cont, "errors=continue"},
1368 	{Opt_err_panic, "errors=panic"},
1369 	{Opt_err_ro, "errors=remount-ro"},
1370 	{Opt_nouid32, "nouid32"},
1371 	{Opt_debug, "debug"},
1372 	{Opt_removed, "oldalloc"},
1373 	{Opt_removed, "orlov"},
1374 	{Opt_user_xattr, "user_xattr"},
1375 	{Opt_nouser_xattr, "nouser_xattr"},
1376 	{Opt_acl, "acl"},
1377 	{Opt_noacl, "noacl"},
1378 	{Opt_noload, "norecovery"},
1379 	{Opt_noload, "noload"},
1380 	{Opt_removed, "nobh"},
1381 	{Opt_removed, "bh"},
1382 	{Opt_commit, "commit=%u"},
1383 	{Opt_min_batch_time, "min_batch_time=%u"},
1384 	{Opt_max_batch_time, "max_batch_time=%u"},
1385 	{Opt_journal_dev, "journal_dev=%u"},
1386 	{Opt_journal_path, "journal_path=%s"},
1387 	{Opt_journal_checksum, "journal_checksum"},
1388 	{Opt_nojournal_checksum, "nojournal_checksum"},
1389 	{Opt_journal_async_commit, "journal_async_commit"},
1390 	{Opt_abort, "abort"},
1391 	{Opt_data_journal, "data=journal"},
1392 	{Opt_data_ordered, "data=ordered"},
1393 	{Opt_data_writeback, "data=writeback"},
1394 	{Opt_data_err_abort, "data_err=abort"},
1395 	{Opt_data_err_ignore, "data_err=ignore"},
1396 	{Opt_offusrjquota, "usrjquota="},
1397 	{Opt_usrjquota, "usrjquota=%s"},
1398 	{Opt_offgrpjquota, "grpjquota="},
1399 	{Opt_grpjquota, "grpjquota=%s"},
1400 	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1401 	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1402 	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1403 	{Opt_grpquota, "grpquota"},
1404 	{Opt_noquota, "noquota"},
1405 	{Opt_quota, "quota"},
1406 	{Opt_usrquota, "usrquota"},
1407 	{Opt_prjquota, "prjquota"},
1408 	{Opt_barrier, "barrier=%u"},
1409 	{Opt_barrier, "barrier"},
1410 	{Opt_nobarrier, "nobarrier"},
1411 	{Opt_i_version, "i_version"},
1412 	{Opt_dax, "dax"},
1413 	{Opt_stripe, "stripe=%u"},
1414 	{Opt_delalloc, "delalloc"},
1415 	{Opt_lazytime, "lazytime"},
1416 	{Opt_nolazytime, "nolazytime"},
1417 	{Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1418 	{Opt_nodelalloc, "nodelalloc"},
1419 	{Opt_removed, "mblk_io_submit"},
1420 	{Opt_removed, "nomblk_io_submit"},
1421 	{Opt_block_validity, "block_validity"},
1422 	{Opt_noblock_validity, "noblock_validity"},
1423 	{Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1424 	{Opt_journal_ioprio, "journal_ioprio=%u"},
1425 	{Opt_auto_da_alloc, "auto_da_alloc=%u"},
1426 	{Opt_auto_da_alloc, "auto_da_alloc"},
1427 	{Opt_noauto_da_alloc, "noauto_da_alloc"},
1428 	{Opt_dioread_nolock, "dioread_nolock"},
1429 	{Opt_dioread_lock, "dioread_lock"},
1430 	{Opt_discard, "discard"},
1431 	{Opt_nodiscard, "nodiscard"},
1432 	{Opt_init_itable, "init_itable=%u"},
1433 	{Opt_init_itable, "init_itable"},
1434 	{Opt_noinit_itable, "noinit_itable"},
1435 	{Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1436 	{Opt_test_dummy_encryption, "test_dummy_encryption"},
1437 	{Opt_nombcache, "nombcache"},
1438 	{Opt_nombcache, "no_mbcache"},	/* for backward compatibility */
1439 	{Opt_removed, "check=none"},	/* mount option from ext2/3 */
1440 	{Opt_removed, "nocheck"},	/* mount option from ext2/3 */
1441 	{Opt_removed, "reservation"},	/* mount option from ext2/3 */
1442 	{Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1443 	{Opt_removed, "journal=%u"},	/* mount option from ext2/3 */
1444 	{Opt_err, NULL},
1445 };
1446 
1447 static ext4_fsblk_t get_sb_block(void **data)
1448 {
1449 	ext4_fsblk_t	sb_block;
1450 	char		*options = (char *) *data;
1451 
1452 	if (!options || strncmp(options, "sb=", 3) != 0)
1453 		return 1;	/* Default location */
1454 
1455 	options += 3;
1456 	/* TODO: use simple_strtoll with >32bit ext4 */
1457 	sb_block = simple_strtoul(options, &options, 0);
1458 	if (*options && *options != ',') {
1459 		printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1460 		       (char *) *data);
1461 		return 1;
1462 	}
1463 	if (*options == ',')
1464 		options++;
1465 	*data = (void *) options;
1466 
1467 	return sb_block;
1468 }
1469 
1470 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1471 static const char deprecated_msg[] =
1472 	"Mount option \"%s\" will be removed by %s\n"
1473 	"Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1474 
1475 #ifdef CONFIG_QUOTA
1476 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1477 {
1478 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1479 	char *qname;
1480 	int ret = -1;
1481 
1482 	if (sb_any_quota_loaded(sb) &&
1483 		!sbi->s_qf_names[qtype]) {
1484 		ext4_msg(sb, KERN_ERR,
1485 			"Cannot change journaled "
1486 			"quota options when quota turned on");
1487 		return -1;
1488 	}
1489 	if (ext4_has_feature_quota(sb)) {
1490 		ext4_msg(sb, KERN_INFO, "Journaled quota options "
1491 			 "ignored when QUOTA feature is enabled");
1492 		return 1;
1493 	}
1494 	qname = match_strdup(args);
1495 	if (!qname) {
1496 		ext4_msg(sb, KERN_ERR,
1497 			"Not enough memory for storing quotafile name");
1498 		return -1;
1499 	}
1500 	if (sbi->s_qf_names[qtype]) {
1501 		if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1502 			ret = 1;
1503 		else
1504 			ext4_msg(sb, KERN_ERR,
1505 				 "%s quota file already specified",
1506 				 QTYPE2NAME(qtype));
1507 		goto errout;
1508 	}
1509 	if (strchr(qname, '/')) {
1510 		ext4_msg(sb, KERN_ERR,
1511 			"quotafile must be on filesystem root");
1512 		goto errout;
1513 	}
1514 	sbi->s_qf_names[qtype] = qname;
1515 	set_opt(sb, QUOTA);
1516 	return 1;
1517 errout:
1518 	kfree(qname);
1519 	return ret;
1520 }
1521 
1522 static int clear_qf_name(struct super_block *sb, int qtype)
1523 {
1524 
1525 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1526 
1527 	if (sb_any_quota_loaded(sb) &&
1528 		sbi->s_qf_names[qtype]) {
1529 		ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1530 			" when quota turned on");
1531 		return -1;
1532 	}
1533 	kfree(sbi->s_qf_names[qtype]);
1534 	sbi->s_qf_names[qtype] = NULL;
1535 	return 1;
1536 }
1537 #endif
1538 
1539 #define MOPT_SET	0x0001
1540 #define MOPT_CLEAR	0x0002
1541 #define MOPT_NOSUPPORT	0x0004
1542 #define MOPT_EXPLICIT	0x0008
1543 #define MOPT_CLEAR_ERR	0x0010
1544 #define MOPT_GTE0	0x0020
1545 #ifdef CONFIG_QUOTA
1546 #define MOPT_Q		0
1547 #define MOPT_QFMT	0x0040
1548 #else
1549 #define MOPT_Q		MOPT_NOSUPPORT
1550 #define MOPT_QFMT	MOPT_NOSUPPORT
1551 #endif
1552 #define MOPT_DATAJ	0x0080
1553 #define MOPT_NO_EXT2	0x0100
1554 #define MOPT_NO_EXT3	0x0200
1555 #define MOPT_EXT4_ONLY	(MOPT_NO_EXT2 | MOPT_NO_EXT3)
1556 #define MOPT_STRING	0x0400
1557 
1558 static const struct mount_opts {
1559 	int	token;
1560 	int	mount_opt;
1561 	int	flags;
1562 } ext4_mount_opts[] = {
1563 	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1564 	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1565 	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1566 	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1567 	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1568 	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1569 	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1570 	 MOPT_EXT4_ONLY | MOPT_SET},
1571 	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1572 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1573 	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1574 	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1575 	{Opt_delalloc, EXT4_MOUNT_DELALLOC,
1576 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1577 	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1578 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1579 	{Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1580 	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1581 	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1582 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1583 	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1584 				    EXT4_MOUNT_JOURNAL_CHECKSUM),
1585 	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1586 	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1587 	{Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1588 	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1589 	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1590 	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1591 	 MOPT_NO_EXT2},
1592 	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1593 	 MOPT_NO_EXT2},
1594 	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1595 	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1596 	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1597 	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1598 	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1599 	{Opt_commit, 0, MOPT_GTE0},
1600 	{Opt_max_batch_time, 0, MOPT_GTE0},
1601 	{Opt_min_batch_time, 0, MOPT_GTE0},
1602 	{Opt_inode_readahead_blks, 0, MOPT_GTE0},
1603 	{Opt_init_itable, 0, MOPT_GTE0},
1604 	{Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1605 	{Opt_stripe, 0, MOPT_GTE0},
1606 	{Opt_resuid, 0, MOPT_GTE0},
1607 	{Opt_resgid, 0, MOPT_GTE0},
1608 	{Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1609 	{Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1610 	{Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1611 	{Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1612 	{Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1613 	{Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1614 	 MOPT_NO_EXT2 | MOPT_DATAJ},
1615 	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1616 	{Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1617 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1618 	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1619 	{Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1620 #else
1621 	{Opt_acl, 0, MOPT_NOSUPPORT},
1622 	{Opt_noacl, 0, MOPT_NOSUPPORT},
1623 #endif
1624 	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1625 	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1626 	{Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1627 	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1628 	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1629 							MOPT_SET | MOPT_Q},
1630 	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1631 							MOPT_SET | MOPT_Q},
1632 	{Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1633 							MOPT_SET | MOPT_Q},
1634 	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1635 		       EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1636 							MOPT_CLEAR | MOPT_Q},
1637 	{Opt_usrjquota, 0, MOPT_Q},
1638 	{Opt_grpjquota, 0, MOPT_Q},
1639 	{Opt_offusrjquota, 0, MOPT_Q},
1640 	{Opt_offgrpjquota, 0, MOPT_Q},
1641 	{Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1642 	{Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1643 	{Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1644 	{Opt_max_dir_size_kb, 0, MOPT_GTE0},
1645 	{Opt_test_dummy_encryption, 0, MOPT_GTE0},
1646 	{Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1647 	{Opt_err, 0, 0}
1648 };
1649 
1650 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1651 			    substring_t *args, unsigned long *journal_devnum,
1652 			    unsigned int *journal_ioprio, int is_remount)
1653 {
1654 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1655 	const struct mount_opts *m;
1656 	kuid_t uid;
1657 	kgid_t gid;
1658 	int arg = 0;
1659 
1660 #ifdef CONFIG_QUOTA
1661 	if (token == Opt_usrjquota)
1662 		return set_qf_name(sb, USRQUOTA, &args[0]);
1663 	else if (token == Opt_grpjquota)
1664 		return set_qf_name(sb, GRPQUOTA, &args[0]);
1665 	else if (token == Opt_offusrjquota)
1666 		return clear_qf_name(sb, USRQUOTA);
1667 	else if (token == Opt_offgrpjquota)
1668 		return clear_qf_name(sb, GRPQUOTA);
1669 #endif
1670 	switch (token) {
1671 	case Opt_noacl:
1672 	case Opt_nouser_xattr:
1673 		ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1674 		break;
1675 	case Opt_sb:
1676 		return 1;	/* handled by get_sb_block() */
1677 	case Opt_removed:
1678 		ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1679 		return 1;
1680 	case Opt_abort:
1681 		sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1682 		return 1;
1683 	case Opt_i_version:
1684 		sb->s_flags |= SB_I_VERSION;
1685 		return 1;
1686 	case Opt_lazytime:
1687 		sb->s_flags |= SB_LAZYTIME;
1688 		return 1;
1689 	case Opt_nolazytime:
1690 		sb->s_flags &= ~SB_LAZYTIME;
1691 		return 1;
1692 	}
1693 
1694 	for (m = ext4_mount_opts; m->token != Opt_err; m++)
1695 		if (token == m->token)
1696 			break;
1697 
1698 	if (m->token == Opt_err) {
1699 		ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1700 			 "or missing value", opt);
1701 		return -1;
1702 	}
1703 
1704 	if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1705 		ext4_msg(sb, KERN_ERR,
1706 			 "Mount option \"%s\" incompatible with ext2", opt);
1707 		return -1;
1708 	}
1709 	if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1710 		ext4_msg(sb, KERN_ERR,
1711 			 "Mount option \"%s\" incompatible with ext3", opt);
1712 		return -1;
1713 	}
1714 
1715 	if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1716 		return -1;
1717 	if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1718 		return -1;
1719 	if (m->flags & MOPT_EXPLICIT) {
1720 		if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1721 			set_opt2(sb, EXPLICIT_DELALLOC);
1722 		} else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1723 			set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1724 		} else
1725 			return -1;
1726 	}
1727 	if (m->flags & MOPT_CLEAR_ERR)
1728 		clear_opt(sb, ERRORS_MASK);
1729 	if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1730 		ext4_msg(sb, KERN_ERR, "Cannot change quota "
1731 			 "options when quota turned on");
1732 		return -1;
1733 	}
1734 
1735 	if (m->flags & MOPT_NOSUPPORT) {
1736 		ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1737 	} else if (token == Opt_commit) {
1738 		if (arg == 0)
1739 			arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1740 		sbi->s_commit_interval = HZ * arg;
1741 	} else if (token == Opt_debug_want_extra_isize) {
1742 		sbi->s_want_extra_isize = arg;
1743 	} else if (token == Opt_max_batch_time) {
1744 		sbi->s_max_batch_time = arg;
1745 	} else if (token == Opt_min_batch_time) {
1746 		sbi->s_min_batch_time = arg;
1747 	} else if (token == Opt_inode_readahead_blks) {
1748 		if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1749 			ext4_msg(sb, KERN_ERR,
1750 				 "EXT4-fs: inode_readahead_blks must be "
1751 				 "0 or a power of 2 smaller than 2^31");
1752 			return -1;
1753 		}
1754 		sbi->s_inode_readahead_blks = arg;
1755 	} else if (token == Opt_init_itable) {
1756 		set_opt(sb, INIT_INODE_TABLE);
1757 		if (!args->from)
1758 			arg = EXT4_DEF_LI_WAIT_MULT;
1759 		sbi->s_li_wait_mult = arg;
1760 	} else if (token == Opt_max_dir_size_kb) {
1761 		sbi->s_max_dir_size_kb = arg;
1762 	} else if (token == Opt_stripe) {
1763 		sbi->s_stripe = arg;
1764 	} else if (token == Opt_resuid) {
1765 		uid = make_kuid(current_user_ns(), arg);
1766 		if (!uid_valid(uid)) {
1767 			ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1768 			return -1;
1769 		}
1770 		sbi->s_resuid = uid;
1771 	} else if (token == Opt_resgid) {
1772 		gid = make_kgid(current_user_ns(), arg);
1773 		if (!gid_valid(gid)) {
1774 			ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1775 			return -1;
1776 		}
1777 		sbi->s_resgid = gid;
1778 	} else if (token == Opt_journal_dev) {
1779 		if (is_remount) {
1780 			ext4_msg(sb, KERN_ERR,
1781 				 "Cannot specify journal on remount");
1782 			return -1;
1783 		}
1784 		*journal_devnum = arg;
1785 	} else if (token == Opt_journal_path) {
1786 		char *journal_path;
1787 		struct inode *journal_inode;
1788 		struct path path;
1789 		int error;
1790 
1791 		if (is_remount) {
1792 			ext4_msg(sb, KERN_ERR,
1793 				 "Cannot specify journal on remount");
1794 			return -1;
1795 		}
1796 		journal_path = match_strdup(&args[0]);
1797 		if (!journal_path) {
1798 			ext4_msg(sb, KERN_ERR, "error: could not dup "
1799 				"journal device string");
1800 			return -1;
1801 		}
1802 
1803 		error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1804 		if (error) {
1805 			ext4_msg(sb, KERN_ERR, "error: could not find "
1806 				"journal device path: error %d", error);
1807 			kfree(journal_path);
1808 			return -1;
1809 		}
1810 
1811 		journal_inode = d_inode(path.dentry);
1812 		if (!S_ISBLK(journal_inode->i_mode)) {
1813 			ext4_msg(sb, KERN_ERR, "error: journal path %s "
1814 				"is not a block device", journal_path);
1815 			path_put(&path);
1816 			kfree(journal_path);
1817 			return -1;
1818 		}
1819 
1820 		*journal_devnum = new_encode_dev(journal_inode->i_rdev);
1821 		path_put(&path);
1822 		kfree(journal_path);
1823 	} else if (token == Opt_journal_ioprio) {
1824 		if (arg > 7) {
1825 			ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1826 				 " (must be 0-7)");
1827 			return -1;
1828 		}
1829 		*journal_ioprio =
1830 			IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1831 	} else if (token == Opt_test_dummy_encryption) {
1832 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1833 		sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1834 		ext4_msg(sb, KERN_WARNING,
1835 			 "Test dummy encryption mode enabled");
1836 #else
1837 		ext4_msg(sb, KERN_WARNING,
1838 			 "Test dummy encryption mount option ignored");
1839 #endif
1840 	} else if (m->flags & MOPT_DATAJ) {
1841 		if (is_remount) {
1842 			if (!sbi->s_journal)
1843 				ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1844 			else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1845 				ext4_msg(sb, KERN_ERR,
1846 					 "Cannot change data mode on remount");
1847 				return -1;
1848 			}
1849 		} else {
1850 			clear_opt(sb, DATA_FLAGS);
1851 			sbi->s_mount_opt |= m->mount_opt;
1852 		}
1853 #ifdef CONFIG_QUOTA
1854 	} else if (m->flags & MOPT_QFMT) {
1855 		if (sb_any_quota_loaded(sb) &&
1856 		    sbi->s_jquota_fmt != m->mount_opt) {
1857 			ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1858 				 "quota options when quota turned on");
1859 			return -1;
1860 		}
1861 		if (ext4_has_feature_quota(sb)) {
1862 			ext4_msg(sb, KERN_INFO,
1863 				 "Quota format mount options ignored "
1864 				 "when QUOTA feature is enabled");
1865 			return 1;
1866 		}
1867 		sbi->s_jquota_fmt = m->mount_opt;
1868 #endif
1869 	} else if (token == Opt_dax) {
1870 #ifdef CONFIG_FS_DAX
1871 		ext4_msg(sb, KERN_WARNING,
1872 		"DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1873 			sbi->s_mount_opt |= m->mount_opt;
1874 #else
1875 		ext4_msg(sb, KERN_INFO, "dax option not supported");
1876 		return -1;
1877 #endif
1878 	} else if (token == Opt_data_err_abort) {
1879 		sbi->s_mount_opt |= m->mount_opt;
1880 	} else if (token == Opt_data_err_ignore) {
1881 		sbi->s_mount_opt &= ~m->mount_opt;
1882 	} else {
1883 		if (!args->from)
1884 			arg = 1;
1885 		if (m->flags & MOPT_CLEAR)
1886 			arg = !arg;
1887 		else if (unlikely(!(m->flags & MOPT_SET))) {
1888 			ext4_msg(sb, KERN_WARNING,
1889 				 "buggy handling of option %s", opt);
1890 			WARN_ON(1);
1891 			return -1;
1892 		}
1893 		if (arg != 0)
1894 			sbi->s_mount_opt |= m->mount_opt;
1895 		else
1896 			sbi->s_mount_opt &= ~m->mount_opt;
1897 	}
1898 	return 1;
1899 }
1900 
1901 static int parse_options(char *options, struct super_block *sb,
1902 			 unsigned long *journal_devnum,
1903 			 unsigned int *journal_ioprio,
1904 			 int is_remount)
1905 {
1906 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1907 	char *p;
1908 	substring_t args[MAX_OPT_ARGS];
1909 	int token;
1910 
1911 	if (!options)
1912 		return 1;
1913 
1914 	while ((p = strsep(&options, ",")) != NULL) {
1915 		if (!*p)
1916 			continue;
1917 		/*
1918 		 * Initialize args struct so we know whether arg was
1919 		 * found; some options take optional arguments.
1920 		 */
1921 		args[0].to = args[0].from = NULL;
1922 		token = match_token(p, tokens, args);
1923 		if (handle_mount_opt(sb, p, token, args, journal_devnum,
1924 				     journal_ioprio, is_remount) < 0)
1925 			return 0;
1926 	}
1927 #ifdef CONFIG_QUOTA
1928 	/*
1929 	 * We do the test below only for project quotas. 'usrquota' and
1930 	 * 'grpquota' mount options are allowed even without quota feature
1931 	 * to support legacy quotas in quota files.
1932 	 */
1933 	if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1934 		ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1935 			 "Cannot enable project quota enforcement.");
1936 		return 0;
1937 	}
1938 	if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1939 		if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1940 			clear_opt(sb, USRQUOTA);
1941 
1942 		if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1943 			clear_opt(sb, GRPQUOTA);
1944 
1945 		if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1946 			ext4_msg(sb, KERN_ERR, "old and new quota "
1947 					"format mixing");
1948 			return 0;
1949 		}
1950 
1951 		if (!sbi->s_jquota_fmt) {
1952 			ext4_msg(sb, KERN_ERR, "journaled quota format "
1953 					"not specified");
1954 			return 0;
1955 		}
1956 	}
1957 #endif
1958 	if (test_opt(sb, DIOREAD_NOLOCK)) {
1959 		int blocksize =
1960 			BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1961 
1962 		if (blocksize < PAGE_SIZE) {
1963 			ext4_msg(sb, KERN_ERR, "can't mount with "
1964 				 "dioread_nolock if block size != PAGE_SIZE");
1965 			return 0;
1966 		}
1967 	}
1968 	return 1;
1969 }
1970 
1971 static inline void ext4_show_quota_options(struct seq_file *seq,
1972 					   struct super_block *sb)
1973 {
1974 #if defined(CONFIG_QUOTA)
1975 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1976 
1977 	if (sbi->s_jquota_fmt) {
1978 		char *fmtname = "";
1979 
1980 		switch (sbi->s_jquota_fmt) {
1981 		case QFMT_VFS_OLD:
1982 			fmtname = "vfsold";
1983 			break;
1984 		case QFMT_VFS_V0:
1985 			fmtname = "vfsv0";
1986 			break;
1987 		case QFMT_VFS_V1:
1988 			fmtname = "vfsv1";
1989 			break;
1990 		}
1991 		seq_printf(seq, ",jqfmt=%s", fmtname);
1992 	}
1993 
1994 	if (sbi->s_qf_names[USRQUOTA])
1995 		seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1996 
1997 	if (sbi->s_qf_names[GRPQUOTA])
1998 		seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1999 #endif
2000 }
2001 
2002 static const char *token2str(int token)
2003 {
2004 	const struct match_token *t;
2005 
2006 	for (t = tokens; t->token != Opt_err; t++)
2007 		if (t->token == token && !strchr(t->pattern, '='))
2008 			break;
2009 	return t->pattern;
2010 }
2011 
2012 /*
2013  * Show an option if
2014  *  - it's set to a non-default value OR
2015  *  - if the per-sb default is different from the global default
2016  */
2017 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2018 			      int nodefs)
2019 {
2020 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2021 	struct ext4_super_block *es = sbi->s_es;
2022 	int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
2023 	const struct mount_opts *m;
2024 	char sep = nodefs ? '\n' : ',';
2025 
2026 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2027 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2028 
2029 	if (sbi->s_sb_block != 1)
2030 		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2031 
2032 	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2033 		int want_set = m->flags & MOPT_SET;
2034 		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2035 		    (m->flags & MOPT_CLEAR_ERR))
2036 			continue;
2037 		if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2038 			continue; /* skip if same as the default */
2039 		if ((want_set &&
2040 		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2041 		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2042 			continue; /* select Opt_noFoo vs Opt_Foo */
2043 		SEQ_OPTS_PRINT("%s", token2str(m->token));
2044 	}
2045 
2046 	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2047 	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2048 		SEQ_OPTS_PRINT("resuid=%u",
2049 				from_kuid_munged(&init_user_ns, sbi->s_resuid));
2050 	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2051 	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2052 		SEQ_OPTS_PRINT("resgid=%u",
2053 				from_kgid_munged(&init_user_ns, sbi->s_resgid));
2054 	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2055 	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2056 		SEQ_OPTS_PUTS("errors=remount-ro");
2057 	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2058 		SEQ_OPTS_PUTS("errors=continue");
2059 	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2060 		SEQ_OPTS_PUTS("errors=panic");
2061 	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2062 		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2063 	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2064 		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2065 	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2066 		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2067 	if (sb->s_flags & SB_I_VERSION)
2068 		SEQ_OPTS_PUTS("i_version");
2069 	if (nodefs || sbi->s_stripe)
2070 		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2071 	if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2072 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2073 			SEQ_OPTS_PUTS("data=journal");
2074 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2075 			SEQ_OPTS_PUTS("data=ordered");
2076 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2077 			SEQ_OPTS_PUTS("data=writeback");
2078 	}
2079 	if (nodefs ||
2080 	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2081 		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2082 			       sbi->s_inode_readahead_blks);
2083 
2084 	if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2085 		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2086 		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2087 	if (nodefs || sbi->s_max_dir_size_kb)
2088 		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2089 	if (test_opt(sb, DATA_ERR_ABORT))
2090 		SEQ_OPTS_PUTS("data_err=abort");
2091 
2092 	ext4_show_quota_options(seq, sb);
2093 	return 0;
2094 }
2095 
2096 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2097 {
2098 	return _ext4_show_options(seq, root->d_sb, 0);
2099 }
2100 
2101 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2102 {
2103 	struct super_block *sb = seq->private;
2104 	int rc;
2105 
2106 	seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2107 	rc = _ext4_show_options(seq, sb, 1);
2108 	seq_puts(seq, "\n");
2109 	return rc;
2110 }
2111 
2112 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2113 			    int read_only)
2114 {
2115 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2116 	int res = 0;
2117 
2118 	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2119 		ext4_msg(sb, KERN_ERR, "revision level too high, "
2120 			 "forcing read-only mode");
2121 		res = SB_RDONLY;
2122 	}
2123 	if (read_only)
2124 		goto done;
2125 	if (!(sbi->s_mount_state & EXT4_VALID_FS))
2126 		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2127 			 "running e2fsck is recommended");
2128 	else if (sbi->s_mount_state & EXT4_ERROR_FS)
2129 		ext4_msg(sb, KERN_WARNING,
2130 			 "warning: mounting fs with errors, "
2131 			 "running e2fsck is recommended");
2132 	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2133 		 le16_to_cpu(es->s_mnt_count) >=
2134 		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2135 		ext4_msg(sb, KERN_WARNING,
2136 			 "warning: maximal mount count reached, "
2137 			 "running e2fsck is recommended");
2138 	else if (le32_to_cpu(es->s_checkinterval) &&
2139 		(le32_to_cpu(es->s_lastcheck) +
2140 			le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2141 		ext4_msg(sb, KERN_WARNING,
2142 			 "warning: checktime reached, "
2143 			 "running e2fsck is recommended");
2144 	if (!sbi->s_journal)
2145 		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2146 	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2147 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2148 	le16_add_cpu(&es->s_mnt_count, 1);
2149 	es->s_mtime = cpu_to_le32(get_seconds());
2150 	ext4_update_dynamic_rev(sb);
2151 	if (sbi->s_journal)
2152 		ext4_set_feature_journal_needs_recovery(sb);
2153 
2154 	ext4_commit_super(sb, 1);
2155 done:
2156 	if (test_opt(sb, DEBUG))
2157 		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2158 				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2159 			sb->s_blocksize,
2160 			sbi->s_groups_count,
2161 			EXT4_BLOCKS_PER_GROUP(sb),
2162 			EXT4_INODES_PER_GROUP(sb),
2163 			sbi->s_mount_opt, sbi->s_mount_opt2);
2164 
2165 	cleancache_init_fs(sb);
2166 	return res;
2167 }
2168 
2169 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2170 {
2171 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2172 	struct flex_groups *new_groups;
2173 	int size;
2174 
2175 	if (!sbi->s_log_groups_per_flex)
2176 		return 0;
2177 
2178 	size = ext4_flex_group(sbi, ngroup - 1) + 1;
2179 	if (size <= sbi->s_flex_groups_allocated)
2180 		return 0;
2181 
2182 	size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2183 	new_groups = kvzalloc(size, GFP_KERNEL);
2184 	if (!new_groups) {
2185 		ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2186 			 size / (int) sizeof(struct flex_groups));
2187 		return -ENOMEM;
2188 	}
2189 
2190 	if (sbi->s_flex_groups) {
2191 		memcpy(new_groups, sbi->s_flex_groups,
2192 		       (sbi->s_flex_groups_allocated *
2193 			sizeof(struct flex_groups)));
2194 		kvfree(sbi->s_flex_groups);
2195 	}
2196 	sbi->s_flex_groups = new_groups;
2197 	sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2198 	return 0;
2199 }
2200 
2201 static int ext4_fill_flex_info(struct super_block *sb)
2202 {
2203 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2204 	struct ext4_group_desc *gdp = NULL;
2205 	ext4_group_t flex_group;
2206 	int i, err;
2207 
2208 	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2209 	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2210 		sbi->s_log_groups_per_flex = 0;
2211 		return 1;
2212 	}
2213 
2214 	err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2215 	if (err)
2216 		goto failed;
2217 
2218 	for (i = 0; i < sbi->s_groups_count; i++) {
2219 		gdp = ext4_get_group_desc(sb, i, NULL);
2220 
2221 		flex_group = ext4_flex_group(sbi, i);
2222 		atomic_add(ext4_free_inodes_count(sb, gdp),
2223 			   &sbi->s_flex_groups[flex_group].free_inodes);
2224 		atomic64_add(ext4_free_group_clusters(sb, gdp),
2225 			     &sbi->s_flex_groups[flex_group].free_clusters);
2226 		atomic_add(ext4_used_dirs_count(sb, gdp),
2227 			   &sbi->s_flex_groups[flex_group].used_dirs);
2228 	}
2229 
2230 	return 1;
2231 failed:
2232 	return 0;
2233 }
2234 
2235 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2236 				   struct ext4_group_desc *gdp)
2237 {
2238 	int offset = offsetof(struct ext4_group_desc, bg_checksum);
2239 	__u16 crc = 0;
2240 	__le32 le_group = cpu_to_le32(block_group);
2241 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2242 
2243 	if (ext4_has_metadata_csum(sbi->s_sb)) {
2244 		/* Use new metadata_csum algorithm */
2245 		__u32 csum32;
2246 		__u16 dummy_csum = 0;
2247 
2248 		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2249 				     sizeof(le_group));
2250 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2251 		csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2252 				     sizeof(dummy_csum));
2253 		offset += sizeof(dummy_csum);
2254 		if (offset < sbi->s_desc_size)
2255 			csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2256 					     sbi->s_desc_size - offset);
2257 
2258 		crc = csum32 & 0xFFFF;
2259 		goto out;
2260 	}
2261 
2262 	/* old crc16 code */
2263 	if (!ext4_has_feature_gdt_csum(sb))
2264 		return 0;
2265 
2266 	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2267 	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2268 	crc = crc16(crc, (__u8 *)gdp, offset);
2269 	offset += sizeof(gdp->bg_checksum); /* skip checksum */
2270 	/* for checksum of struct ext4_group_desc do the rest...*/
2271 	if (ext4_has_feature_64bit(sb) &&
2272 	    offset < le16_to_cpu(sbi->s_es->s_desc_size))
2273 		crc = crc16(crc, (__u8 *)gdp + offset,
2274 			    le16_to_cpu(sbi->s_es->s_desc_size) -
2275 				offset);
2276 
2277 out:
2278 	return cpu_to_le16(crc);
2279 }
2280 
2281 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2282 				struct ext4_group_desc *gdp)
2283 {
2284 	if (ext4_has_group_desc_csum(sb) &&
2285 	    (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2286 		return 0;
2287 
2288 	return 1;
2289 }
2290 
2291 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2292 			      struct ext4_group_desc *gdp)
2293 {
2294 	if (!ext4_has_group_desc_csum(sb))
2295 		return;
2296 	gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2297 }
2298 
2299 /* Called at mount-time, super-block is locked */
2300 static int ext4_check_descriptors(struct super_block *sb,
2301 				  ext4_fsblk_t sb_block,
2302 				  ext4_group_t *first_not_zeroed)
2303 {
2304 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2305 	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2306 	ext4_fsblk_t last_block;
2307 	ext4_fsblk_t block_bitmap;
2308 	ext4_fsblk_t inode_bitmap;
2309 	ext4_fsblk_t inode_table;
2310 	int flexbg_flag = 0;
2311 	ext4_group_t i, grp = sbi->s_groups_count;
2312 
2313 	if (ext4_has_feature_flex_bg(sb))
2314 		flexbg_flag = 1;
2315 
2316 	ext4_debug("Checking group descriptors");
2317 
2318 	for (i = 0; i < sbi->s_groups_count; i++) {
2319 		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2320 
2321 		if (i == sbi->s_groups_count - 1 || flexbg_flag)
2322 			last_block = ext4_blocks_count(sbi->s_es) - 1;
2323 		else
2324 			last_block = first_block +
2325 				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
2326 
2327 		if ((grp == sbi->s_groups_count) &&
2328 		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2329 			grp = i;
2330 
2331 		block_bitmap = ext4_block_bitmap(sb, gdp);
2332 		if (block_bitmap == sb_block) {
2333 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2334 				 "Block bitmap for group %u overlaps "
2335 				 "superblock", i);
2336 		}
2337 		if (block_bitmap < first_block || block_bitmap > last_block) {
2338 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2339 			       "Block bitmap for group %u not in group "
2340 			       "(block %llu)!", i, block_bitmap);
2341 			return 0;
2342 		}
2343 		inode_bitmap = ext4_inode_bitmap(sb, gdp);
2344 		if (inode_bitmap == sb_block) {
2345 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2346 				 "Inode bitmap for group %u overlaps "
2347 				 "superblock", i);
2348 		}
2349 		if (inode_bitmap < first_block || inode_bitmap > last_block) {
2350 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2351 			       "Inode bitmap for group %u not in group "
2352 			       "(block %llu)!", i, inode_bitmap);
2353 			return 0;
2354 		}
2355 		inode_table = ext4_inode_table(sb, gdp);
2356 		if (inode_table == sb_block) {
2357 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2358 				 "Inode table for group %u overlaps "
2359 				 "superblock", i);
2360 		}
2361 		if (inode_table < first_block ||
2362 		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
2363 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2364 			       "Inode table for group %u not in group "
2365 			       "(block %llu)!", i, inode_table);
2366 			return 0;
2367 		}
2368 		ext4_lock_group(sb, i);
2369 		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2370 			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2371 				 "Checksum for group %u failed (%u!=%u)",
2372 				 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2373 				     gdp)), le16_to_cpu(gdp->bg_checksum));
2374 			if (!sb_rdonly(sb)) {
2375 				ext4_unlock_group(sb, i);
2376 				return 0;
2377 			}
2378 		}
2379 		ext4_unlock_group(sb, i);
2380 		if (!flexbg_flag)
2381 			first_block += EXT4_BLOCKS_PER_GROUP(sb);
2382 	}
2383 	if (NULL != first_not_zeroed)
2384 		*first_not_zeroed = grp;
2385 	return 1;
2386 }
2387 
2388 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2389  * the superblock) which were deleted from all directories, but held open by
2390  * a process at the time of a crash.  We walk the list and try to delete these
2391  * inodes at recovery time (only with a read-write filesystem).
2392  *
2393  * In order to keep the orphan inode chain consistent during traversal (in
2394  * case of crash during recovery), we link each inode into the superblock
2395  * orphan list_head and handle it the same way as an inode deletion during
2396  * normal operation (which journals the operations for us).
2397  *
2398  * We only do an iget() and an iput() on each inode, which is very safe if we
2399  * accidentally point at an in-use or already deleted inode.  The worst that
2400  * can happen in this case is that we get a "bit already cleared" message from
2401  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2402  * e2fsck was run on this filesystem, and it must have already done the orphan
2403  * inode cleanup for us, so we can safely abort without any further action.
2404  */
2405 static void ext4_orphan_cleanup(struct super_block *sb,
2406 				struct ext4_super_block *es)
2407 {
2408 	unsigned int s_flags = sb->s_flags;
2409 	int ret, nr_orphans = 0, nr_truncates = 0;
2410 #ifdef CONFIG_QUOTA
2411 	int quota_update = 0;
2412 	int i;
2413 #endif
2414 	if (!es->s_last_orphan) {
2415 		jbd_debug(4, "no orphan inodes to clean up\n");
2416 		return;
2417 	}
2418 
2419 	if (bdev_read_only(sb->s_bdev)) {
2420 		ext4_msg(sb, KERN_ERR, "write access "
2421 			"unavailable, skipping orphan cleanup");
2422 		return;
2423 	}
2424 
2425 	/* Check if feature set would not allow a r/w mount */
2426 	if (!ext4_feature_set_ok(sb, 0)) {
2427 		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2428 			 "unknown ROCOMPAT features");
2429 		return;
2430 	}
2431 
2432 	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2433 		/* don't clear list on RO mount w/ errors */
2434 		if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2435 			ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2436 				  "clearing orphan list.\n");
2437 			es->s_last_orphan = 0;
2438 		}
2439 		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2440 		return;
2441 	}
2442 
2443 	if (s_flags & SB_RDONLY) {
2444 		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2445 		sb->s_flags &= ~SB_RDONLY;
2446 	}
2447 #ifdef CONFIG_QUOTA
2448 	/* Needed for iput() to work correctly and not trash data */
2449 	sb->s_flags |= SB_ACTIVE;
2450 
2451 	/*
2452 	 * Turn on quotas which were not enabled for read-only mounts if
2453 	 * filesystem has quota feature, so that they are updated correctly.
2454 	 */
2455 	if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2456 		int ret = ext4_enable_quotas(sb);
2457 
2458 		if (!ret)
2459 			quota_update = 1;
2460 		else
2461 			ext4_msg(sb, KERN_ERR,
2462 				"Cannot turn on quotas: error %d", ret);
2463 	}
2464 
2465 	/* Turn on journaled quotas used for old sytle */
2466 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2467 		if (EXT4_SB(sb)->s_qf_names[i]) {
2468 			int ret = ext4_quota_on_mount(sb, i);
2469 
2470 			if (!ret)
2471 				quota_update = 1;
2472 			else
2473 				ext4_msg(sb, KERN_ERR,
2474 					"Cannot turn on journaled "
2475 					"quota: type %d: error %d", i, ret);
2476 		}
2477 	}
2478 #endif
2479 
2480 	while (es->s_last_orphan) {
2481 		struct inode *inode;
2482 
2483 		/*
2484 		 * We may have encountered an error during cleanup; if
2485 		 * so, skip the rest.
2486 		 */
2487 		if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2488 			jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2489 			es->s_last_orphan = 0;
2490 			break;
2491 		}
2492 
2493 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2494 		if (IS_ERR(inode)) {
2495 			es->s_last_orphan = 0;
2496 			break;
2497 		}
2498 
2499 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2500 		dquot_initialize(inode);
2501 		if (inode->i_nlink) {
2502 			if (test_opt(sb, DEBUG))
2503 				ext4_msg(sb, KERN_DEBUG,
2504 					"%s: truncating inode %lu to %lld bytes",
2505 					__func__, inode->i_ino, inode->i_size);
2506 			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2507 				  inode->i_ino, inode->i_size);
2508 			inode_lock(inode);
2509 			truncate_inode_pages(inode->i_mapping, inode->i_size);
2510 			ret = ext4_truncate(inode);
2511 			if (ret)
2512 				ext4_std_error(inode->i_sb, ret);
2513 			inode_unlock(inode);
2514 			nr_truncates++;
2515 		} else {
2516 			if (test_opt(sb, DEBUG))
2517 				ext4_msg(sb, KERN_DEBUG,
2518 					"%s: deleting unreferenced inode %lu",
2519 					__func__, inode->i_ino);
2520 			jbd_debug(2, "deleting unreferenced inode %lu\n",
2521 				  inode->i_ino);
2522 			nr_orphans++;
2523 		}
2524 		iput(inode);  /* The delete magic happens here! */
2525 	}
2526 
2527 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2528 
2529 	if (nr_orphans)
2530 		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2531 		       PLURAL(nr_orphans));
2532 	if (nr_truncates)
2533 		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2534 		       PLURAL(nr_truncates));
2535 #ifdef CONFIG_QUOTA
2536 	/* Turn off quotas if they were enabled for orphan cleanup */
2537 	if (quota_update) {
2538 		for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2539 			if (sb_dqopt(sb)->files[i])
2540 				dquot_quota_off(sb, i);
2541 		}
2542 	}
2543 #endif
2544 	sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2545 }
2546 
2547 /*
2548  * Maximal extent format file size.
2549  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2550  * extent format containers, within a sector_t, and within i_blocks
2551  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2552  * so that won't be a limiting factor.
2553  *
2554  * However there is other limiting factor. We do store extents in the form
2555  * of starting block and length, hence the resulting length of the extent
2556  * covering maximum file size must fit into on-disk format containers as
2557  * well. Given that length is always by 1 unit bigger than max unit (because
2558  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2559  *
2560  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2561  */
2562 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2563 {
2564 	loff_t res;
2565 	loff_t upper_limit = MAX_LFS_FILESIZE;
2566 
2567 	/* small i_blocks in vfs inode? */
2568 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2569 		/*
2570 		 * CONFIG_LBDAF is not enabled implies the inode
2571 		 * i_block represent total blocks in 512 bytes
2572 		 * 32 == size of vfs inode i_blocks * 8
2573 		 */
2574 		upper_limit = (1LL << 32) - 1;
2575 
2576 		/* total blocks in file system block size */
2577 		upper_limit >>= (blkbits - 9);
2578 		upper_limit <<= blkbits;
2579 	}
2580 
2581 	/*
2582 	 * 32-bit extent-start container, ee_block. We lower the maxbytes
2583 	 * by one fs block, so ee_len can cover the extent of maximum file
2584 	 * size
2585 	 */
2586 	res = (1LL << 32) - 1;
2587 	res <<= blkbits;
2588 
2589 	/* Sanity check against vm- & vfs- imposed limits */
2590 	if (res > upper_limit)
2591 		res = upper_limit;
2592 
2593 	return res;
2594 }
2595 
2596 /*
2597  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2598  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2599  * We need to be 1 filesystem block less than the 2^48 sector limit.
2600  */
2601 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2602 {
2603 	loff_t res = EXT4_NDIR_BLOCKS;
2604 	int meta_blocks;
2605 	loff_t upper_limit;
2606 	/* This is calculated to be the largest file size for a dense, block
2607 	 * mapped file such that the file's total number of 512-byte sectors,
2608 	 * including data and all indirect blocks, does not exceed (2^48 - 1).
2609 	 *
2610 	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2611 	 * number of 512-byte sectors of the file.
2612 	 */
2613 
2614 	if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2615 		/*
2616 		 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2617 		 * the inode i_block field represents total file blocks in
2618 		 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2619 		 */
2620 		upper_limit = (1LL << 32) - 1;
2621 
2622 		/* total blocks in file system block size */
2623 		upper_limit >>= (bits - 9);
2624 
2625 	} else {
2626 		/*
2627 		 * We use 48 bit ext4_inode i_blocks
2628 		 * With EXT4_HUGE_FILE_FL set the i_blocks
2629 		 * represent total number of blocks in
2630 		 * file system block size
2631 		 */
2632 		upper_limit = (1LL << 48) - 1;
2633 
2634 	}
2635 
2636 	/* indirect blocks */
2637 	meta_blocks = 1;
2638 	/* double indirect blocks */
2639 	meta_blocks += 1 + (1LL << (bits-2));
2640 	/* tripple indirect blocks */
2641 	meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2642 
2643 	upper_limit -= meta_blocks;
2644 	upper_limit <<= bits;
2645 
2646 	res += 1LL << (bits-2);
2647 	res += 1LL << (2*(bits-2));
2648 	res += 1LL << (3*(bits-2));
2649 	res <<= bits;
2650 	if (res > upper_limit)
2651 		res = upper_limit;
2652 
2653 	if (res > MAX_LFS_FILESIZE)
2654 		res = MAX_LFS_FILESIZE;
2655 
2656 	return res;
2657 }
2658 
2659 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2660 				   ext4_fsblk_t logical_sb_block, int nr)
2661 {
2662 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2663 	ext4_group_t bg, first_meta_bg;
2664 	int has_super = 0;
2665 
2666 	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2667 
2668 	if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2669 		return logical_sb_block + nr + 1;
2670 	bg = sbi->s_desc_per_block * nr;
2671 	if (ext4_bg_has_super(sb, bg))
2672 		has_super = 1;
2673 
2674 	/*
2675 	 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2676 	 * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
2677 	 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2678 	 * compensate.
2679 	 */
2680 	if (sb->s_blocksize == 1024 && nr == 0 &&
2681 	    le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
2682 		has_super++;
2683 
2684 	return (has_super + ext4_group_first_block_no(sb, bg));
2685 }
2686 
2687 /**
2688  * ext4_get_stripe_size: Get the stripe size.
2689  * @sbi: In memory super block info
2690  *
2691  * If we have specified it via mount option, then
2692  * use the mount option value. If the value specified at mount time is
2693  * greater than the blocks per group use the super block value.
2694  * If the super block value is greater than blocks per group return 0.
2695  * Allocator needs it be less than blocks per group.
2696  *
2697  */
2698 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2699 {
2700 	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2701 	unsigned long stripe_width =
2702 			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2703 	int ret;
2704 
2705 	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2706 		ret = sbi->s_stripe;
2707 	else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2708 		ret = stripe_width;
2709 	else if (stride && stride <= sbi->s_blocks_per_group)
2710 		ret = stride;
2711 	else
2712 		ret = 0;
2713 
2714 	/*
2715 	 * If the stripe width is 1, this makes no sense and
2716 	 * we set it to 0 to turn off stripe handling code.
2717 	 */
2718 	if (ret <= 1)
2719 		ret = 0;
2720 
2721 	return ret;
2722 }
2723 
2724 /*
2725  * Check whether this filesystem can be mounted based on
2726  * the features present and the RDONLY/RDWR mount requested.
2727  * Returns 1 if this filesystem can be mounted as requested,
2728  * 0 if it cannot be.
2729  */
2730 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2731 {
2732 	if (ext4_has_unknown_ext4_incompat_features(sb)) {
2733 		ext4_msg(sb, KERN_ERR,
2734 			"Couldn't mount because of "
2735 			"unsupported optional features (%x)",
2736 			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2737 			~EXT4_FEATURE_INCOMPAT_SUPP));
2738 		return 0;
2739 	}
2740 
2741 	if (readonly)
2742 		return 1;
2743 
2744 	if (ext4_has_feature_readonly(sb)) {
2745 		ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2746 		sb->s_flags |= SB_RDONLY;
2747 		return 1;
2748 	}
2749 
2750 	/* Check that feature set is OK for a read-write mount */
2751 	if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2752 		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2753 			 "unsupported optional features (%x)",
2754 			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2755 				~EXT4_FEATURE_RO_COMPAT_SUPP));
2756 		return 0;
2757 	}
2758 	/*
2759 	 * Large file size enabled file system can only be mounted
2760 	 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2761 	 */
2762 	if (ext4_has_feature_huge_file(sb)) {
2763 		if (sizeof(blkcnt_t) < sizeof(u64)) {
2764 			ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2765 				 "cannot be mounted RDWR without "
2766 				 "CONFIG_LBDAF");
2767 			return 0;
2768 		}
2769 	}
2770 	if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2771 		ext4_msg(sb, KERN_ERR,
2772 			 "Can't support bigalloc feature without "
2773 			 "extents feature\n");
2774 		return 0;
2775 	}
2776 
2777 #ifndef CONFIG_QUOTA
2778 	if (ext4_has_feature_quota(sb) && !readonly) {
2779 		ext4_msg(sb, KERN_ERR,
2780 			 "Filesystem with quota feature cannot be mounted RDWR "
2781 			 "without CONFIG_QUOTA");
2782 		return 0;
2783 	}
2784 	if (ext4_has_feature_project(sb) && !readonly) {
2785 		ext4_msg(sb, KERN_ERR,
2786 			 "Filesystem with project quota feature cannot be mounted RDWR "
2787 			 "without CONFIG_QUOTA");
2788 		return 0;
2789 	}
2790 #endif  /* CONFIG_QUOTA */
2791 	return 1;
2792 }
2793 
2794 /*
2795  * This function is called once a day if we have errors logged
2796  * on the file system
2797  */
2798 static void print_daily_error_info(struct timer_list *t)
2799 {
2800 	struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2801 	struct super_block *sb = sbi->s_sb;
2802 	struct ext4_super_block *es = sbi->s_es;
2803 
2804 	if (es->s_error_count)
2805 		/* fsck newer than v1.41.13 is needed to clean this condition. */
2806 		ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2807 			 le32_to_cpu(es->s_error_count));
2808 	if (es->s_first_error_time) {
2809 		printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2810 		       sb->s_id, le32_to_cpu(es->s_first_error_time),
2811 		       (int) sizeof(es->s_first_error_func),
2812 		       es->s_first_error_func,
2813 		       le32_to_cpu(es->s_first_error_line));
2814 		if (es->s_first_error_ino)
2815 			printk(KERN_CONT ": inode %u",
2816 			       le32_to_cpu(es->s_first_error_ino));
2817 		if (es->s_first_error_block)
2818 			printk(KERN_CONT ": block %llu", (unsigned long long)
2819 			       le64_to_cpu(es->s_first_error_block));
2820 		printk(KERN_CONT "\n");
2821 	}
2822 	if (es->s_last_error_time) {
2823 		printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2824 		       sb->s_id, le32_to_cpu(es->s_last_error_time),
2825 		       (int) sizeof(es->s_last_error_func),
2826 		       es->s_last_error_func,
2827 		       le32_to_cpu(es->s_last_error_line));
2828 		if (es->s_last_error_ino)
2829 			printk(KERN_CONT ": inode %u",
2830 			       le32_to_cpu(es->s_last_error_ino));
2831 		if (es->s_last_error_block)
2832 			printk(KERN_CONT ": block %llu", (unsigned long long)
2833 			       le64_to_cpu(es->s_last_error_block));
2834 		printk(KERN_CONT "\n");
2835 	}
2836 	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2837 }
2838 
2839 /* Find next suitable group and run ext4_init_inode_table */
2840 static int ext4_run_li_request(struct ext4_li_request *elr)
2841 {
2842 	struct ext4_group_desc *gdp = NULL;
2843 	ext4_group_t group, ngroups;
2844 	struct super_block *sb;
2845 	unsigned long timeout = 0;
2846 	int ret = 0;
2847 
2848 	sb = elr->lr_super;
2849 	ngroups = EXT4_SB(sb)->s_groups_count;
2850 
2851 	for (group = elr->lr_next_group; group < ngroups; group++) {
2852 		gdp = ext4_get_group_desc(sb, group, NULL);
2853 		if (!gdp) {
2854 			ret = 1;
2855 			break;
2856 		}
2857 
2858 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2859 			break;
2860 	}
2861 
2862 	if (group >= ngroups)
2863 		ret = 1;
2864 
2865 	if (!ret) {
2866 		timeout = jiffies;
2867 		ret = ext4_init_inode_table(sb, group,
2868 					    elr->lr_timeout ? 0 : 1);
2869 		if (elr->lr_timeout == 0) {
2870 			timeout = (jiffies - timeout) *
2871 				  elr->lr_sbi->s_li_wait_mult;
2872 			elr->lr_timeout = timeout;
2873 		}
2874 		elr->lr_next_sched = jiffies + elr->lr_timeout;
2875 		elr->lr_next_group = group + 1;
2876 	}
2877 	return ret;
2878 }
2879 
2880 /*
2881  * Remove lr_request from the list_request and free the
2882  * request structure. Should be called with li_list_mtx held
2883  */
2884 static void ext4_remove_li_request(struct ext4_li_request *elr)
2885 {
2886 	struct ext4_sb_info *sbi;
2887 
2888 	if (!elr)
2889 		return;
2890 
2891 	sbi = elr->lr_sbi;
2892 
2893 	list_del(&elr->lr_request);
2894 	sbi->s_li_request = NULL;
2895 	kfree(elr);
2896 }
2897 
2898 static void ext4_unregister_li_request(struct super_block *sb)
2899 {
2900 	mutex_lock(&ext4_li_mtx);
2901 	if (!ext4_li_info) {
2902 		mutex_unlock(&ext4_li_mtx);
2903 		return;
2904 	}
2905 
2906 	mutex_lock(&ext4_li_info->li_list_mtx);
2907 	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2908 	mutex_unlock(&ext4_li_info->li_list_mtx);
2909 	mutex_unlock(&ext4_li_mtx);
2910 }
2911 
2912 static struct task_struct *ext4_lazyinit_task;
2913 
2914 /*
2915  * This is the function where ext4lazyinit thread lives. It walks
2916  * through the request list searching for next scheduled filesystem.
2917  * When such a fs is found, run the lazy initialization request
2918  * (ext4_rn_li_request) and keep track of the time spend in this
2919  * function. Based on that time we compute next schedule time of
2920  * the request. When walking through the list is complete, compute
2921  * next waking time and put itself into sleep.
2922  */
2923 static int ext4_lazyinit_thread(void *arg)
2924 {
2925 	struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2926 	struct list_head *pos, *n;
2927 	struct ext4_li_request *elr;
2928 	unsigned long next_wakeup, cur;
2929 
2930 	BUG_ON(NULL == eli);
2931 
2932 cont_thread:
2933 	while (true) {
2934 		next_wakeup = MAX_JIFFY_OFFSET;
2935 
2936 		mutex_lock(&eli->li_list_mtx);
2937 		if (list_empty(&eli->li_request_list)) {
2938 			mutex_unlock(&eli->li_list_mtx);
2939 			goto exit_thread;
2940 		}
2941 		list_for_each_safe(pos, n, &eli->li_request_list) {
2942 			int err = 0;
2943 			int progress = 0;
2944 			elr = list_entry(pos, struct ext4_li_request,
2945 					 lr_request);
2946 
2947 			if (time_before(jiffies, elr->lr_next_sched)) {
2948 				if (time_before(elr->lr_next_sched, next_wakeup))
2949 					next_wakeup = elr->lr_next_sched;
2950 				continue;
2951 			}
2952 			if (down_read_trylock(&elr->lr_super->s_umount)) {
2953 				if (sb_start_write_trylock(elr->lr_super)) {
2954 					progress = 1;
2955 					/*
2956 					 * We hold sb->s_umount, sb can not
2957 					 * be removed from the list, it is
2958 					 * now safe to drop li_list_mtx
2959 					 */
2960 					mutex_unlock(&eli->li_list_mtx);
2961 					err = ext4_run_li_request(elr);
2962 					sb_end_write(elr->lr_super);
2963 					mutex_lock(&eli->li_list_mtx);
2964 					n = pos->next;
2965 				}
2966 				up_read((&elr->lr_super->s_umount));
2967 			}
2968 			/* error, remove the lazy_init job */
2969 			if (err) {
2970 				ext4_remove_li_request(elr);
2971 				continue;
2972 			}
2973 			if (!progress) {
2974 				elr->lr_next_sched = jiffies +
2975 					(prandom_u32()
2976 					 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2977 			}
2978 			if (time_before(elr->lr_next_sched, next_wakeup))
2979 				next_wakeup = elr->lr_next_sched;
2980 		}
2981 		mutex_unlock(&eli->li_list_mtx);
2982 
2983 		try_to_freeze();
2984 
2985 		cur = jiffies;
2986 		if ((time_after_eq(cur, next_wakeup)) ||
2987 		    (MAX_JIFFY_OFFSET == next_wakeup)) {
2988 			cond_resched();
2989 			continue;
2990 		}
2991 
2992 		schedule_timeout_interruptible(next_wakeup - cur);
2993 
2994 		if (kthread_should_stop()) {
2995 			ext4_clear_request_list();
2996 			goto exit_thread;
2997 		}
2998 	}
2999 
3000 exit_thread:
3001 	/*
3002 	 * It looks like the request list is empty, but we need
3003 	 * to check it under the li_list_mtx lock, to prevent any
3004 	 * additions into it, and of course we should lock ext4_li_mtx
3005 	 * to atomically free the list and ext4_li_info, because at
3006 	 * this point another ext4 filesystem could be registering
3007 	 * new one.
3008 	 */
3009 	mutex_lock(&ext4_li_mtx);
3010 	mutex_lock(&eli->li_list_mtx);
3011 	if (!list_empty(&eli->li_request_list)) {
3012 		mutex_unlock(&eli->li_list_mtx);
3013 		mutex_unlock(&ext4_li_mtx);
3014 		goto cont_thread;
3015 	}
3016 	mutex_unlock(&eli->li_list_mtx);
3017 	kfree(ext4_li_info);
3018 	ext4_li_info = NULL;
3019 	mutex_unlock(&ext4_li_mtx);
3020 
3021 	return 0;
3022 }
3023 
3024 static void ext4_clear_request_list(void)
3025 {
3026 	struct list_head *pos, *n;
3027 	struct ext4_li_request *elr;
3028 
3029 	mutex_lock(&ext4_li_info->li_list_mtx);
3030 	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3031 		elr = list_entry(pos, struct ext4_li_request,
3032 				 lr_request);
3033 		ext4_remove_li_request(elr);
3034 	}
3035 	mutex_unlock(&ext4_li_info->li_list_mtx);
3036 }
3037 
3038 static int ext4_run_lazyinit_thread(void)
3039 {
3040 	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3041 					 ext4_li_info, "ext4lazyinit");
3042 	if (IS_ERR(ext4_lazyinit_task)) {
3043 		int err = PTR_ERR(ext4_lazyinit_task);
3044 		ext4_clear_request_list();
3045 		kfree(ext4_li_info);
3046 		ext4_li_info = NULL;
3047 		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3048 				 "initialization thread\n",
3049 				 err);
3050 		return err;
3051 	}
3052 	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3053 	return 0;
3054 }
3055 
3056 /*
3057  * Check whether it make sense to run itable init. thread or not.
3058  * If there is at least one uninitialized inode table, return
3059  * corresponding group number, else the loop goes through all
3060  * groups and return total number of groups.
3061  */
3062 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3063 {
3064 	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3065 	struct ext4_group_desc *gdp = NULL;
3066 
3067 	for (group = 0; group < ngroups; group++) {
3068 		gdp = ext4_get_group_desc(sb, group, NULL);
3069 		if (!gdp)
3070 			continue;
3071 
3072 		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3073 			break;
3074 	}
3075 
3076 	return group;
3077 }
3078 
3079 static int ext4_li_info_new(void)
3080 {
3081 	struct ext4_lazy_init *eli = NULL;
3082 
3083 	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3084 	if (!eli)
3085 		return -ENOMEM;
3086 
3087 	INIT_LIST_HEAD(&eli->li_request_list);
3088 	mutex_init(&eli->li_list_mtx);
3089 
3090 	eli->li_state |= EXT4_LAZYINIT_QUIT;
3091 
3092 	ext4_li_info = eli;
3093 
3094 	return 0;
3095 }
3096 
3097 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3098 					    ext4_group_t start)
3099 {
3100 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3101 	struct ext4_li_request *elr;
3102 
3103 	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3104 	if (!elr)
3105 		return NULL;
3106 
3107 	elr->lr_super = sb;
3108 	elr->lr_sbi = sbi;
3109 	elr->lr_next_group = start;
3110 
3111 	/*
3112 	 * Randomize first schedule time of the request to
3113 	 * spread the inode table initialization requests
3114 	 * better.
3115 	 */
3116 	elr->lr_next_sched = jiffies + (prandom_u32() %
3117 				(EXT4_DEF_LI_MAX_START_DELAY * HZ));
3118 	return elr;
3119 }
3120 
3121 int ext4_register_li_request(struct super_block *sb,
3122 			     ext4_group_t first_not_zeroed)
3123 {
3124 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3125 	struct ext4_li_request *elr = NULL;
3126 	ext4_group_t ngroups = sbi->s_groups_count;
3127 	int ret = 0;
3128 
3129 	mutex_lock(&ext4_li_mtx);
3130 	if (sbi->s_li_request != NULL) {
3131 		/*
3132 		 * Reset timeout so it can be computed again, because
3133 		 * s_li_wait_mult might have changed.
3134 		 */
3135 		sbi->s_li_request->lr_timeout = 0;
3136 		goto out;
3137 	}
3138 
3139 	if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3140 	    !test_opt(sb, INIT_INODE_TABLE))
3141 		goto out;
3142 
3143 	elr = ext4_li_request_new(sb, first_not_zeroed);
3144 	if (!elr) {
3145 		ret = -ENOMEM;
3146 		goto out;
3147 	}
3148 
3149 	if (NULL == ext4_li_info) {
3150 		ret = ext4_li_info_new();
3151 		if (ret)
3152 			goto out;
3153 	}
3154 
3155 	mutex_lock(&ext4_li_info->li_list_mtx);
3156 	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3157 	mutex_unlock(&ext4_li_info->li_list_mtx);
3158 
3159 	sbi->s_li_request = elr;
3160 	/*
3161 	 * set elr to NULL here since it has been inserted to
3162 	 * the request_list and the removal and free of it is
3163 	 * handled by ext4_clear_request_list from now on.
3164 	 */
3165 	elr = NULL;
3166 
3167 	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3168 		ret = ext4_run_lazyinit_thread();
3169 		if (ret)
3170 			goto out;
3171 	}
3172 out:
3173 	mutex_unlock(&ext4_li_mtx);
3174 	if (ret)
3175 		kfree(elr);
3176 	return ret;
3177 }
3178 
3179 /*
3180  * We do not need to lock anything since this is called on
3181  * module unload.
3182  */
3183 static void ext4_destroy_lazyinit_thread(void)
3184 {
3185 	/*
3186 	 * If thread exited earlier
3187 	 * there's nothing to be done.
3188 	 */
3189 	if (!ext4_li_info || !ext4_lazyinit_task)
3190 		return;
3191 
3192 	kthread_stop(ext4_lazyinit_task);
3193 }
3194 
3195 static int set_journal_csum_feature_set(struct super_block *sb)
3196 {
3197 	int ret = 1;
3198 	int compat, incompat;
3199 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3200 
3201 	if (ext4_has_metadata_csum(sb)) {
3202 		/* journal checksum v3 */
3203 		compat = 0;
3204 		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3205 	} else {
3206 		/* journal checksum v1 */
3207 		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3208 		incompat = 0;
3209 	}
3210 
3211 	jbd2_journal_clear_features(sbi->s_journal,
3212 			JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3213 			JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3214 			JBD2_FEATURE_INCOMPAT_CSUM_V2);
3215 	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3216 		ret = jbd2_journal_set_features(sbi->s_journal,
3217 				compat, 0,
3218 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3219 				incompat);
3220 	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3221 		ret = jbd2_journal_set_features(sbi->s_journal,
3222 				compat, 0,
3223 				incompat);
3224 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3225 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3226 	} else {
3227 		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3228 				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3229 	}
3230 
3231 	return ret;
3232 }
3233 
3234 /*
3235  * Note: calculating the overhead so we can be compatible with
3236  * historical BSD practice is quite difficult in the face of
3237  * clusters/bigalloc.  This is because multiple metadata blocks from
3238  * different block group can end up in the same allocation cluster.
3239  * Calculating the exact overhead in the face of clustered allocation
3240  * requires either O(all block bitmaps) in memory or O(number of block
3241  * groups**2) in time.  We will still calculate the superblock for
3242  * older file systems --- and if we come across with a bigalloc file
3243  * system with zero in s_overhead_clusters the estimate will be close to
3244  * correct especially for very large cluster sizes --- but for newer
3245  * file systems, it's better to calculate this figure once at mkfs
3246  * time, and store it in the superblock.  If the superblock value is
3247  * present (even for non-bigalloc file systems), we will use it.
3248  */
3249 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3250 			  char *buf)
3251 {
3252 	struct ext4_sb_info	*sbi = EXT4_SB(sb);
3253 	struct ext4_group_desc	*gdp;
3254 	ext4_fsblk_t		first_block, last_block, b;
3255 	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
3256 	int			s, j, count = 0;
3257 
3258 	if (!ext4_has_feature_bigalloc(sb))
3259 		return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3260 			sbi->s_itb_per_group + 2);
3261 
3262 	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3263 		(grp * EXT4_BLOCKS_PER_GROUP(sb));
3264 	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3265 	for (i = 0; i < ngroups; i++) {
3266 		gdp = ext4_get_group_desc(sb, i, NULL);
3267 		b = ext4_block_bitmap(sb, gdp);
3268 		if (b >= first_block && b <= last_block) {
3269 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3270 			count++;
3271 		}
3272 		b = ext4_inode_bitmap(sb, gdp);
3273 		if (b >= first_block && b <= last_block) {
3274 			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3275 			count++;
3276 		}
3277 		b = ext4_inode_table(sb, gdp);
3278 		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3279 			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3280 				int c = EXT4_B2C(sbi, b - first_block);
3281 				ext4_set_bit(c, buf);
3282 				count++;
3283 			}
3284 		if (i != grp)
3285 			continue;
3286 		s = 0;
3287 		if (ext4_bg_has_super(sb, grp)) {
3288 			ext4_set_bit(s++, buf);
3289 			count++;
3290 		}
3291 		j = ext4_bg_num_gdb(sb, grp);
3292 		if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3293 			ext4_error(sb, "Invalid number of block group "
3294 				   "descriptor blocks: %d", j);
3295 			j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3296 		}
3297 		count += j;
3298 		for (; j > 0; j--)
3299 			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3300 	}
3301 	if (!count)
3302 		return 0;
3303 	return EXT4_CLUSTERS_PER_GROUP(sb) -
3304 		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3305 }
3306 
3307 /*
3308  * Compute the overhead and stash it in sbi->s_overhead
3309  */
3310 int ext4_calculate_overhead(struct super_block *sb)
3311 {
3312 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3313 	struct ext4_super_block *es = sbi->s_es;
3314 	struct inode *j_inode;
3315 	unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3316 	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3317 	ext4_fsblk_t overhead = 0;
3318 	char *buf = (char *) get_zeroed_page(GFP_NOFS);
3319 
3320 	if (!buf)
3321 		return -ENOMEM;
3322 
3323 	/*
3324 	 * Compute the overhead (FS structures).  This is constant
3325 	 * for a given filesystem unless the number of block groups
3326 	 * changes so we cache the previous value until it does.
3327 	 */
3328 
3329 	/*
3330 	 * All of the blocks before first_data_block are overhead
3331 	 */
3332 	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3333 
3334 	/*
3335 	 * Add the overhead found in each block group
3336 	 */
3337 	for (i = 0; i < ngroups; i++) {
3338 		int blks;
3339 
3340 		blks = count_overhead(sb, i, buf);
3341 		overhead += blks;
3342 		if (blks)
3343 			memset(buf, 0, PAGE_SIZE);
3344 		cond_resched();
3345 	}
3346 
3347 	/*
3348 	 * Add the internal journal blocks whether the journal has been
3349 	 * loaded or not
3350 	 */
3351 	if (sbi->s_journal && !sbi->journal_bdev)
3352 		overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3353 	else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3354 		j_inode = ext4_get_journal_inode(sb, j_inum);
3355 		if (j_inode) {
3356 			j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3357 			overhead += EXT4_NUM_B2C(sbi, j_blocks);
3358 			iput(j_inode);
3359 		} else {
3360 			ext4_msg(sb, KERN_ERR, "can't get journal size");
3361 		}
3362 	}
3363 	sbi->s_overhead = overhead;
3364 	smp_wmb();
3365 	free_page((unsigned long) buf);
3366 	return 0;
3367 }
3368 
3369 static void ext4_set_resv_clusters(struct super_block *sb)
3370 {
3371 	ext4_fsblk_t resv_clusters;
3372 	struct ext4_sb_info *sbi = EXT4_SB(sb);
3373 
3374 	/*
3375 	 * There's no need to reserve anything when we aren't using extents.
3376 	 * The space estimates are exact, there are no unwritten extents,
3377 	 * hole punching doesn't need new metadata... This is needed especially
3378 	 * to keep ext2/3 backward compatibility.
3379 	 */
3380 	if (!ext4_has_feature_extents(sb))
3381 		return;
3382 	/*
3383 	 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3384 	 * This should cover the situations where we can not afford to run
3385 	 * out of space like for example punch hole, or converting
3386 	 * unwritten extents in delalloc path. In most cases such
3387 	 * allocation would require 1, or 2 blocks, higher numbers are
3388 	 * very rare.
3389 	 */
3390 	resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3391 			 sbi->s_cluster_bits);
3392 
3393 	do_div(resv_clusters, 50);
3394 	resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3395 
3396 	atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3397 }
3398 
3399 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3400 {
3401 	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3402 	char *orig_data = kstrdup(data, GFP_KERNEL);
3403 	struct buffer_head *bh;
3404 	struct ext4_super_block *es = NULL;
3405 	struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3406 	ext4_fsblk_t block;
3407 	ext4_fsblk_t sb_block = get_sb_block(&data);
3408 	ext4_fsblk_t logical_sb_block;
3409 	unsigned long offset = 0;
3410 	unsigned long journal_devnum = 0;
3411 	unsigned long def_mount_opts;
3412 	struct inode *root;
3413 	const char *descr;
3414 	int ret = -ENOMEM;
3415 	int blocksize, clustersize;
3416 	unsigned int db_count;
3417 	unsigned int i;
3418 	int needs_recovery, has_huge_files, has_bigalloc;
3419 	__u64 blocks_count;
3420 	int err = 0;
3421 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3422 	ext4_group_t first_not_zeroed;
3423 
3424 	if ((data && !orig_data) || !sbi)
3425 		goto out_free_base;
3426 
3427 	sbi->s_daxdev = dax_dev;
3428 	sbi->s_blockgroup_lock =
3429 		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3430 	if (!sbi->s_blockgroup_lock)
3431 		goto out_free_base;
3432 
3433 	sb->s_fs_info = sbi;
3434 	sbi->s_sb = sb;
3435 	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3436 	sbi->s_sb_block = sb_block;
3437 	if (sb->s_bdev->bd_part)
3438 		sbi->s_sectors_written_start =
3439 			part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3440 
3441 	/* Cleanup superblock name */
3442 	strreplace(sb->s_id, '/', '!');
3443 
3444 	/* -EINVAL is default */
3445 	ret = -EINVAL;
3446 	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3447 	if (!blocksize) {
3448 		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3449 		goto out_fail;
3450 	}
3451 
3452 	/*
3453 	 * The ext4 superblock will not be buffer aligned for other than 1kB
3454 	 * block sizes.  We need to calculate the offset from buffer start.
3455 	 */
3456 	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3457 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3458 		offset = do_div(logical_sb_block, blocksize);
3459 	} else {
3460 		logical_sb_block = sb_block;
3461 	}
3462 
3463 	if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3464 		ext4_msg(sb, KERN_ERR, "unable to read superblock");
3465 		goto out_fail;
3466 	}
3467 	/*
3468 	 * Note: s_es must be initialized as soon as possible because
3469 	 *       some ext4 macro-instructions depend on its value
3470 	 */
3471 	es = (struct ext4_super_block *) (bh->b_data + offset);
3472 	sbi->s_es = es;
3473 	sb->s_magic = le16_to_cpu(es->s_magic);
3474 	if (sb->s_magic != EXT4_SUPER_MAGIC)
3475 		goto cantfind_ext4;
3476 	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3477 
3478 	/* Warn if metadata_csum and gdt_csum are both set. */
3479 	if (ext4_has_feature_metadata_csum(sb) &&
3480 	    ext4_has_feature_gdt_csum(sb))
3481 		ext4_warning(sb, "metadata_csum and uninit_bg are "
3482 			     "redundant flags; please run fsck.");
3483 
3484 	/* Check for a known checksum algorithm */
3485 	if (!ext4_verify_csum_type(sb, es)) {
3486 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3487 			 "unknown checksum algorithm.");
3488 		silent = 1;
3489 		goto cantfind_ext4;
3490 	}
3491 
3492 	/* Load the checksum driver */
3493 	if (ext4_has_feature_metadata_csum(sb) ||
3494 	    ext4_has_feature_ea_inode(sb)) {
3495 		sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3496 		if (IS_ERR(sbi->s_chksum_driver)) {
3497 			ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3498 			ret = PTR_ERR(sbi->s_chksum_driver);
3499 			sbi->s_chksum_driver = NULL;
3500 			goto failed_mount;
3501 		}
3502 	}
3503 
3504 	/* Check superblock checksum */
3505 	if (!ext4_superblock_csum_verify(sb, es)) {
3506 		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3507 			 "invalid superblock checksum.  Run e2fsck?");
3508 		silent = 1;
3509 		ret = -EFSBADCRC;
3510 		goto cantfind_ext4;
3511 	}
3512 
3513 	/* Precompute checksum seed for all metadata */
3514 	if (ext4_has_feature_csum_seed(sb))
3515 		sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3516 	else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3517 		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3518 					       sizeof(es->s_uuid));
3519 
3520 	/* Set defaults before we parse the mount options */
3521 	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3522 	set_opt(sb, INIT_INODE_TABLE);
3523 	if (def_mount_opts & EXT4_DEFM_DEBUG)
3524 		set_opt(sb, DEBUG);
3525 	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3526 		set_opt(sb, GRPID);
3527 	if (def_mount_opts & EXT4_DEFM_UID16)
3528 		set_opt(sb, NO_UID32);
3529 	/* xattr user namespace & acls are now defaulted on */
3530 	set_opt(sb, XATTR_USER);
3531 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3532 	set_opt(sb, POSIX_ACL);
3533 #endif
3534 	/* don't forget to enable journal_csum when metadata_csum is enabled. */
3535 	if (ext4_has_metadata_csum(sb))
3536 		set_opt(sb, JOURNAL_CHECKSUM);
3537 
3538 	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3539 		set_opt(sb, JOURNAL_DATA);
3540 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3541 		set_opt(sb, ORDERED_DATA);
3542 	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3543 		set_opt(sb, WRITEBACK_DATA);
3544 
3545 	if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3546 		set_opt(sb, ERRORS_PANIC);
3547 	else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3548 		set_opt(sb, ERRORS_CONT);
3549 	else
3550 		set_opt(sb, ERRORS_RO);
3551 	/* block_validity enabled by default; disable with noblock_validity */
3552 	set_opt(sb, BLOCK_VALIDITY);
3553 	if (def_mount_opts & EXT4_DEFM_DISCARD)
3554 		set_opt(sb, DISCARD);
3555 
3556 	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3557 	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3558 	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3559 	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3560 	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3561 
3562 	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3563 		set_opt(sb, BARRIER);
3564 
3565 	/*
3566 	 * enable delayed allocation by default
3567 	 * Use -o nodelalloc to turn it off
3568 	 */
3569 	if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3570 	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3571 		set_opt(sb, DELALLOC);
3572 
3573 	/*
3574 	 * set default s_li_wait_mult for lazyinit, for the case there is
3575 	 * no mount option specified.
3576 	 */
3577 	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3578 
3579 	if (sbi->s_es->s_mount_opts[0]) {
3580 		char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3581 					      sizeof(sbi->s_es->s_mount_opts),
3582 					      GFP_KERNEL);
3583 		if (!s_mount_opts)
3584 			goto failed_mount;
3585 		if (!parse_options(s_mount_opts, sb, &journal_devnum,
3586 				   &journal_ioprio, 0)) {
3587 			ext4_msg(sb, KERN_WARNING,
3588 				 "failed to parse options in superblock: %s",
3589 				 s_mount_opts);
3590 		}
3591 		kfree(s_mount_opts);
3592 	}
3593 	sbi->s_def_mount_opt = sbi->s_mount_opt;
3594 	if (!parse_options((char *) data, sb, &journal_devnum,
3595 			   &journal_ioprio, 0))
3596 		goto failed_mount;
3597 
3598 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3599 		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3600 			    "with data=journal disables delayed "
3601 			    "allocation and O_DIRECT support!\n");
3602 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3603 			ext4_msg(sb, KERN_ERR, "can't mount with "
3604 				 "both data=journal and delalloc");
3605 			goto failed_mount;
3606 		}
3607 		if (test_opt(sb, DIOREAD_NOLOCK)) {
3608 			ext4_msg(sb, KERN_ERR, "can't mount with "
3609 				 "both data=journal and dioread_nolock");
3610 			goto failed_mount;
3611 		}
3612 		if (test_opt(sb, DAX)) {
3613 			ext4_msg(sb, KERN_ERR, "can't mount with "
3614 				 "both data=journal and dax");
3615 			goto failed_mount;
3616 		}
3617 		if (ext4_has_feature_encrypt(sb)) {
3618 			ext4_msg(sb, KERN_WARNING,
3619 				 "encrypted files will use data=ordered "
3620 				 "instead of data journaling mode");
3621 		}
3622 		if (test_opt(sb, DELALLOC))
3623 			clear_opt(sb, DELALLOC);
3624 	} else {
3625 		sb->s_iflags |= SB_I_CGROUPWB;
3626 	}
3627 
3628 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3629 		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
3630 
3631 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3632 	    (ext4_has_compat_features(sb) ||
3633 	     ext4_has_ro_compat_features(sb) ||
3634 	     ext4_has_incompat_features(sb)))
3635 		ext4_msg(sb, KERN_WARNING,
3636 		       "feature flags set on rev 0 fs, "
3637 		       "running e2fsck is recommended");
3638 
3639 	if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3640 		set_opt2(sb, HURD_COMPAT);
3641 		if (ext4_has_feature_64bit(sb)) {
3642 			ext4_msg(sb, KERN_ERR,
3643 				 "The Hurd can't support 64-bit file systems");
3644 			goto failed_mount;
3645 		}
3646 
3647 		/*
3648 		 * ea_inode feature uses l_i_version field which is not
3649 		 * available in HURD_COMPAT mode.
3650 		 */
3651 		if (ext4_has_feature_ea_inode(sb)) {
3652 			ext4_msg(sb, KERN_ERR,
3653 				 "ea_inode feature is not supported for Hurd");
3654 			goto failed_mount;
3655 		}
3656 	}
3657 
3658 	if (IS_EXT2_SB(sb)) {
3659 		if (ext2_feature_set_ok(sb))
3660 			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3661 				 "using the ext4 subsystem");
3662 		else {
3663 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3664 				 "to feature incompatibilities");
3665 			goto failed_mount;
3666 		}
3667 	}
3668 
3669 	if (IS_EXT3_SB(sb)) {
3670 		if (ext3_feature_set_ok(sb))
3671 			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3672 				 "using the ext4 subsystem");
3673 		else {
3674 			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3675 				 "to feature incompatibilities");
3676 			goto failed_mount;
3677 		}
3678 	}
3679 
3680 	/*
3681 	 * Check feature flags regardless of the revision level, since we
3682 	 * previously didn't change the revision level when setting the flags,
3683 	 * so there is a chance incompat flags are set on a rev 0 filesystem.
3684 	 */
3685 	if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
3686 		goto failed_mount;
3687 
3688 	blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3689 	if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3690 	    blocksize > EXT4_MAX_BLOCK_SIZE) {
3691 		ext4_msg(sb, KERN_ERR,
3692 		       "Unsupported filesystem blocksize %d (%d log_block_size)",
3693 			 blocksize, le32_to_cpu(es->s_log_block_size));
3694 		goto failed_mount;
3695 	}
3696 	if (le32_to_cpu(es->s_log_block_size) >
3697 	    (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3698 		ext4_msg(sb, KERN_ERR,
3699 			 "Invalid log block size: %u",
3700 			 le32_to_cpu(es->s_log_block_size));
3701 		goto failed_mount;
3702 	}
3703 
3704 	if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3705 		ext4_msg(sb, KERN_ERR,
3706 			 "Number of reserved GDT blocks insanely large: %d",
3707 			 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3708 		goto failed_mount;
3709 	}
3710 
3711 	if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3712 		if (ext4_has_feature_inline_data(sb)) {
3713 			ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3714 					" that may contain inline data");
3715 			sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3716 		}
3717 		err = bdev_dax_supported(sb, blocksize);
3718 		if (err) {
3719 			ext4_msg(sb, KERN_ERR,
3720 				"DAX unsupported by block device. Turning off DAX.");
3721 			sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3722 		}
3723 	}
3724 
3725 	if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3726 		ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3727 			 es->s_encryption_level);
3728 		goto failed_mount;
3729 	}
3730 
3731 	if (sb->s_blocksize != blocksize) {
3732 		/* Validate the filesystem blocksize */
3733 		if (!sb_set_blocksize(sb, blocksize)) {
3734 			ext4_msg(sb, KERN_ERR, "bad block size %d",
3735 					blocksize);
3736 			goto failed_mount;
3737 		}
3738 
3739 		brelse(bh);
3740 		logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3741 		offset = do_div(logical_sb_block, blocksize);
3742 		bh = sb_bread_unmovable(sb, logical_sb_block);
3743 		if (!bh) {
3744 			ext4_msg(sb, KERN_ERR,
3745 			       "Can't read superblock on 2nd try");
3746 			goto failed_mount;
3747 		}
3748 		es = (struct ext4_super_block *)(bh->b_data + offset);
3749 		sbi->s_es = es;
3750 		if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3751 			ext4_msg(sb, KERN_ERR,
3752 			       "Magic mismatch, very weird!");
3753 			goto failed_mount;
3754 		}
3755 	}
3756 
3757 	has_huge_files = ext4_has_feature_huge_file(sb);
3758 	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3759 						      has_huge_files);
3760 	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3761 
3762 	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3763 		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3764 		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3765 	} else {
3766 		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3767 		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3768 		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3769 		    (!is_power_of_2(sbi->s_inode_size)) ||
3770 		    (sbi->s_inode_size > blocksize)) {
3771 			ext4_msg(sb, KERN_ERR,
3772 			       "unsupported inode size: %d",
3773 			       sbi->s_inode_size);
3774 			goto failed_mount;
3775 		}
3776 		if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3777 			sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3778 	}
3779 
3780 	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3781 	if (ext4_has_feature_64bit(sb)) {
3782 		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3783 		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3784 		    !is_power_of_2(sbi->s_desc_size)) {
3785 			ext4_msg(sb, KERN_ERR,
3786 			       "unsupported descriptor size %lu",
3787 			       sbi->s_desc_size);
3788 			goto failed_mount;
3789 		}
3790 	} else
3791 		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3792 
3793 	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3794 	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3795 
3796 	sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3797 	if (sbi->s_inodes_per_block == 0)
3798 		goto cantfind_ext4;
3799 	if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3800 	    sbi->s_inodes_per_group > blocksize * 8) {
3801 		ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3802 			 sbi->s_blocks_per_group);
3803 		goto failed_mount;
3804 	}
3805 	sbi->s_itb_per_group = sbi->s_inodes_per_group /
3806 					sbi->s_inodes_per_block;
3807 	sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3808 	sbi->s_sbh = bh;
3809 	sbi->s_mount_state = le16_to_cpu(es->s_state);
3810 	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3811 	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3812 
3813 	for (i = 0; i < 4; i++)
3814 		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3815 	sbi->s_def_hash_version = es->s_def_hash_version;
3816 	if (ext4_has_feature_dir_index(sb)) {
3817 		i = le32_to_cpu(es->s_flags);
3818 		if (i & EXT2_FLAGS_UNSIGNED_HASH)
3819 			sbi->s_hash_unsigned = 3;
3820 		else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3821 #ifdef __CHAR_UNSIGNED__
3822 			if (!sb_rdonly(sb))
3823 				es->s_flags |=
3824 					cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3825 			sbi->s_hash_unsigned = 3;
3826 #else
3827 			if (!sb_rdonly(sb))
3828 				es->s_flags |=
3829 					cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3830 #endif
3831 		}
3832 	}
3833 
3834 	/* Handle clustersize */
3835 	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3836 	has_bigalloc = ext4_has_feature_bigalloc(sb);
3837 	if (has_bigalloc) {
3838 		if (clustersize < blocksize) {
3839 			ext4_msg(sb, KERN_ERR,
3840 				 "cluster size (%d) smaller than "
3841 				 "block size (%d)", clustersize, blocksize);
3842 			goto failed_mount;
3843 		}
3844 		if (le32_to_cpu(es->s_log_cluster_size) >
3845 		    (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3846 			ext4_msg(sb, KERN_ERR,
3847 				 "Invalid log cluster size: %u",
3848 				 le32_to_cpu(es->s_log_cluster_size));
3849 			goto failed_mount;
3850 		}
3851 		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3852 			le32_to_cpu(es->s_log_block_size);
3853 		sbi->s_clusters_per_group =
3854 			le32_to_cpu(es->s_clusters_per_group);
3855 		if (sbi->s_clusters_per_group > blocksize * 8) {
3856 			ext4_msg(sb, KERN_ERR,
3857 				 "#clusters per group too big: %lu",
3858 				 sbi->s_clusters_per_group);
3859 			goto failed_mount;
3860 		}
3861 		if (sbi->s_blocks_per_group !=
3862 		    (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3863 			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3864 				 "clusters per group (%lu) inconsistent",
3865 				 sbi->s_blocks_per_group,
3866 				 sbi->s_clusters_per_group);
3867 			goto failed_mount;
3868 		}
3869 	} else {
3870 		if (clustersize != blocksize) {
3871 			ext4_warning(sb, "fragment/cluster size (%d) != "
3872 				     "block size (%d)", clustersize,
3873 				     blocksize);
3874 			clustersize = blocksize;
3875 		}
3876 		if (sbi->s_blocks_per_group > blocksize * 8) {
3877 			ext4_msg(sb, KERN_ERR,
3878 				 "#blocks per group too big: %lu",
3879 				 sbi->s_blocks_per_group);
3880 			goto failed_mount;
3881 		}
3882 		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3883 		sbi->s_cluster_bits = 0;
3884 	}
3885 	sbi->s_cluster_ratio = clustersize / blocksize;
3886 
3887 	/* Do we have standard group size of clustersize * 8 blocks ? */
3888 	if (sbi->s_blocks_per_group == clustersize << 3)
3889 		set_opt2(sb, STD_GROUP_SIZE);
3890 
3891 	/*
3892 	 * Test whether we have more sectors than will fit in sector_t,
3893 	 * and whether the max offset is addressable by the page cache.
3894 	 */
3895 	err = generic_check_addressable(sb->s_blocksize_bits,
3896 					ext4_blocks_count(es));
3897 	if (err) {
3898 		ext4_msg(sb, KERN_ERR, "filesystem"
3899 			 " too large to mount safely on this system");
3900 		if (sizeof(sector_t) < 8)
3901 			ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3902 		goto failed_mount;
3903 	}
3904 
3905 	if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3906 		goto cantfind_ext4;
3907 
3908 	/* check blocks count against device size */
3909 	blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3910 	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3911 		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3912 		       "exceeds size of device (%llu blocks)",
3913 		       ext4_blocks_count(es), blocks_count);
3914 		goto failed_mount;
3915 	}
3916 
3917 	/*
3918 	 * It makes no sense for the first data block to be beyond the end
3919 	 * of the filesystem.
3920 	 */
3921 	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3922 		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3923 			 "block %u is beyond end of filesystem (%llu)",
3924 			 le32_to_cpu(es->s_first_data_block),
3925 			 ext4_blocks_count(es));
3926 		goto failed_mount;
3927 	}
3928 	blocks_count = (ext4_blocks_count(es) -
3929 			le32_to_cpu(es->s_first_data_block) +
3930 			EXT4_BLOCKS_PER_GROUP(sb) - 1);
3931 	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3932 	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3933 		ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3934 		       "(block count %llu, first data block %u, "
3935 		       "blocks per group %lu)", sbi->s_groups_count,
3936 		       ext4_blocks_count(es),
3937 		       le32_to_cpu(es->s_first_data_block),
3938 		       EXT4_BLOCKS_PER_GROUP(sb));
3939 		goto failed_mount;
3940 	}
3941 	sbi->s_groups_count = blocks_count;
3942 	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3943 			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3944 	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3945 		   EXT4_DESC_PER_BLOCK(sb);
3946 	if (ext4_has_feature_meta_bg(sb)) {
3947 		if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3948 			ext4_msg(sb, KERN_WARNING,
3949 				 "first meta block group too large: %u "
3950 				 "(group descriptor block count %u)",
3951 				 le32_to_cpu(es->s_first_meta_bg), db_count);
3952 			goto failed_mount;
3953 		}
3954 	}
3955 	sbi->s_group_desc = kvmalloc(db_count *
3956 					  sizeof(struct buffer_head *),
3957 					  GFP_KERNEL);
3958 	if (sbi->s_group_desc == NULL) {
3959 		ext4_msg(sb, KERN_ERR, "not enough memory");
3960 		ret = -ENOMEM;
3961 		goto failed_mount;
3962 	}
3963 
3964 	bgl_lock_init(sbi->s_blockgroup_lock);
3965 
3966 	/* Pre-read the descriptors into the buffer cache */
3967 	for (i = 0; i < db_count; i++) {
3968 		block = descriptor_loc(sb, logical_sb_block, i);
3969 		sb_breadahead(sb, block);
3970 	}
3971 
3972 	for (i = 0; i < db_count; i++) {
3973 		block = descriptor_loc(sb, logical_sb_block, i);
3974 		sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3975 		if (!sbi->s_group_desc[i]) {
3976 			ext4_msg(sb, KERN_ERR,
3977 			       "can't read group descriptor %d", i);
3978 			db_count = i;
3979 			goto failed_mount2;
3980 		}
3981 	}
3982 	if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3983 		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3984 		ret = -EFSCORRUPTED;
3985 		goto failed_mount2;
3986 	}
3987 
3988 	sbi->s_gdb_count = db_count;
3989 
3990 	timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
3991 
3992 	/* Register extent status tree shrinker */
3993 	if (ext4_es_register_shrinker(sbi))
3994 		goto failed_mount3;
3995 
3996 	sbi->s_stripe = ext4_get_stripe_size(sbi);
3997 	sbi->s_extent_max_zeroout_kb = 32;
3998 
3999 	/*
4000 	 * set up enough so that it can read an inode
4001 	 */
4002 	sb->s_op = &ext4_sops;
4003 	sb->s_export_op = &ext4_export_ops;
4004 	sb->s_xattr = ext4_xattr_handlers;
4005 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4006 	sb->s_cop = &ext4_cryptops;
4007 #endif
4008 #ifdef CONFIG_QUOTA
4009 	sb->dq_op = &ext4_quota_operations;
4010 	if (ext4_has_feature_quota(sb))
4011 		sb->s_qcop = &dquot_quotactl_sysfile_ops;
4012 	else
4013 		sb->s_qcop = &ext4_qctl_operations;
4014 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4015 #endif
4016 	memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4017 
4018 	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4019 	mutex_init(&sbi->s_orphan_lock);
4020 
4021 	sb->s_root = NULL;
4022 
4023 	needs_recovery = (es->s_last_orphan != 0 ||
4024 			  ext4_has_feature_journal_needs_recovery(sb));
4025 
4026 	if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4027 		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4028 			goto failed_mount3a;
4029 
4030 	/*
4031 	 * The first inode we look at is the journal inode.  Don't try
4032 	 * root first: it may be modified in the journal!
4033 	 */
4034 	if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4035 		err = ext4_load_journal(sb, es, journal_devnum);
4036 		if (err)
4037 			goto failed_mount3a;
4038 	} else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4039 		   ext4_has_feature_journal_needs_recovery(sb)) {
4040 		ext4_msg(sb, KERN_ERR, "required journal recovery "
4041 		       "suppressed and not mounted read-only");
4042 		goto failed_mount_wq;
4043 	} else {
4044 		/* Nojournal mode, all journal mount options are illegal */
4045 		if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4046 			ext4_msg(sb, KERN_ERR, "can't mount with "
4047 				 "journal_checksum, fs mounted w/o journal");
4048 			goto failed_mount_wq;
4049 		}
4050 		if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4051 			ext4_msg(sb, KERN_ERR, "can't mount with "
4052 				 "journal_async_commit, fs mounted w/o journal");
4053 			goto failed_mount_wq;
4054 		}
4055 		if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4056 			ext4_msg(sb, KERN_ERR, "can't mount with "
4057 				 "commit=%lu, fs mounted w/o journal",
4058 				 sbi->s_commit_interval / HZ);
4059 			goto failed_mount_wq;
4060 		}
4061 		if (EXT4_MOUNT_DATA_FLAGS &
4062 		    (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4063 			ext4_msg(sb, KERN_ERR, "can't mount with "
4064 				 "data=, fs mounted w/o journal");
4065 			goto failed_mount_wq;
4066 		}
4067 		sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4068 		clear_opt(sb, JOURNAL_CHECKSUM);
4069 		clear_opt(sb, DATA_FLAGS);
4070 		sbi->s_journal = NULL;
4071 		needs_recovery = 0;
4072 		goto no_journal;
4073 	}
4074 
4075 	if (ext4_has_feature_64bit(sb) &&
4076 	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4077 				       JBD2_FEATURE_INCOMPAT_64BIT)) {
4078 		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4079 		goto failed_mount_wq;
4080 	}
4081 
4082 	if (!set_journal_csum_feature_set(sb)) {
4083 		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4084 			 "feature set");
4085 		goto failed_mount_wq;
4086 	}
4087 
4088 	/* We have now updated the journal if required, so we can
4089 	 * validate the data journaling mode. */
4090 	switch (test_opt(sb, DATA_FLAGS)) {
4091 	case 0:
4092 		/* No mode set, assume a default based on the journal
4093 		 * capabilities: ORDERED_DATA if the journal can
4094 		 * cope, else JOURNAL_DATA
4095 		 */
4096 		if (jbd2_journal_check_available_features
4097 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4098 			set_opt(sb, ORDERED_DATA);
4099 		else
4100 			set_opt(sb, JOURNAL_DATA);
4101 		break;
4102 
4103 	case EXT4_MOUNT_ORDERED_DATA:
4104 	case EXT4_MOUNT_WRITEBACK_DATA:
4105 		if (!jbd2_journal_check_available_features
4106 		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4107 			ext4_msg(sb, KERN_ERR, "Journal does not support "
4108 			       "requested data journaling mode");
4109 			goto failed_mount_wq;
4110 		}
4111 	default:
4112 		break;
4113 	}
4114 
4115 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4116 	    test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4117 		ext4_msg(sb, KERN_ERR, "can't mount with "
4118 			"journal_async_commit in data=ordered mode");
4119 		goto failed_mount_wq;
4120 	}
4121 
4122 	set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4123 
4124 	sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4125 
4126 no_journal:
4127 	if (!test_opt(sb, NO_MBCACHE)) {
4128 		sbi->s_ea_block_cache = ext4_xattr_create_cache();
4129 		if (!sbi->s_ea_block_cache) {
4130 			ext4_msg(sb, KERN_ERR,
4131 				 "Failed to create ea_block_cache");
4132 			goto failed_mount_wq;
4133 		}
4134 
4135 		if (ext4_has_feature_ea_inode(sb)) {
4136 			sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4137 			if (!sbi->s_ea_inode_cache) {
4138 				ext4_msg(sb, KERN_ERR,
4139 					 "Failed to create ea_inode_cache");
4140 				goto failed_mount_wq;
4141 			}
4142 		}
4143 	}
4144 
4145 	if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4146 	    (blocksize != PAGE_SIZE)) {
4147 		ext4_msg(sb, KERN_ERR,
4148 			 "Unsupported blocksize for fs encryption");
4149 		goto failed_mount_wq;
4150 	}
4151 
4152 	if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4153 	    !ext4_has_feature_encrypt(sb)) {
4154 		ext4_set_feature_encrypt(sb);
4155 		ext4_commit_super(sb, 1);
4156 	}
4157 
4158 	/*
4159 	 * Get the # of file system overhead blocks from the
4160 	 * superblock if present.
4161 	 */
4162 	if (es->s_overhead_clusters)
4163 		sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4164 	else {
4165 		err = ext4_calculate_overhead(sb);
4166 		if (err)
4167 			goto failed_mount_wq;
4168 	}
4169 
4170 	/*
4171 	 * The maximum number of concurrent works can be high and
4172 	 * concurrency isn't really necessary.  Limit it to 1.
4173 	 */
4174 	EXT4_SB(sb)->rsv_conversion_wq =
4175 		alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4176 	if (!EXT4_SB(sb)->rsv_conversion_wq) {
4177 		printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4178 		ret = -ENOMEM;
4179 		goto failed_mount4;
4180 	}
4181 
4182 	/*
4183 	 * The jbd2_journal_load will have done any necessary log recovery,
4184 	 * so we can safely mount the rest of the filesystem now.
4185 	 */
4186 
4187 	root = ext4_iget(sb, EXT4_ROOT_INO);
4188 	if (IS_ERR(root)) {
4189 		ext4_msg(sb, KERN_ERR, "get root inode failed");
4190 		ret = PTR_ERR(root);
4191 		root = NULL;
4192 		goto failed_mount4;
4193 	}
4194 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4195 		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4196 		iput(root);
4197 		goto failed_mount4;
4198 	}
4199 	sb->s_root = d_make_root(root);
4200 	if (!sb->s_root) {
4201 		ext4_msg(sb, KERN_ERR, "get root dentry failed");
4202 		ret = -ENOMEM;
4203 		goto failed_mount4;
4204 	}
4205 
4206 	if (ext4_setup_super(sb, es, sb_rdonly(sb)))
4207 		sb->s_flags |= SB_RDONLY;
4208 
4209 	/* determine the minimum size of new large inodes, if present */
4210 	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4211 	    sbi->s_want_extra_isize == 0) {
4212 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4213 						     EXT4_GOOD_OLD_INODE_SIZE;
4214 		if (ext4_has_feature_extra_isize(sb)) {
4215 			if (sbi->s_want_extra_isize <
4216 			    le16_to_cpu(es->s_want_extra_isize))
4217 				sbi->s_want_extra_isize =
4218 					le16_to_cpu(es->s_want_extra_isize);
4219 			if (sbi->s_want_extra_isize <
4220 			    le16_to_cpu(es->s_min_extra_isize))
4221 				sbi->s_want_extra_isize =
4222 					le16_to_cpu(es->s_min_extra_isize);
4223 		}
4224 	}
4225 	/* Check if enough inode space is available */
4226 	if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4227 							sbi->s_inode_size) {
4228 		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4229 						       EXT4_GOOD_OLD_INODE_SIZE;
4230 		ext4_msg(sb, KERN_INFO, "required extra inode space not"
4231 			 "available");
4232 	}
4233 
4234 	ext4_set_resv_clusters(sb);
4235 
4236 	err = ext4_setup_system_zone(sb);
4237 	if (err) {
4238 		ext4_msg(sb, KERN_ERR, "failed to initialize system "
4239 			 "zone (%d)", err);
4240 		goto failed_mount4a;
4241 	}
4242 
4243 	ext4_ext_init(sb);
4244 	err = ext4_mb_init(sb);
4245 	if (err) {
4246 		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4247 			 err);
4248 		goto failed_mount5;
4249 	}
4250 
4251 	block = ext4_count_free_clusters(sb);
4252 	ext4_free_blocks_count_set(sbi->s_es,
4253 				   EXT4_C2B(sbi, block));
4254 	err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4255 				  GFP_KERNEL);
4256 	if (!err) {
4257 		unsigned long freei = ext4_count_free_inodes(sb);
4258 		sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4259 		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4260 					  GFP_KERNEL);
4261 	}
4262 	if (!err)
4263 		err = percpu_counter_init(&sbi->s_dirs_counter,
4264 					  ext4_count_dirs(sb), GFP_KERNEL);
4265 	if (!err)
4266 		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4267 					  GFP_KERNEL);
4268 	if (!err)
4269 		err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4270 
4271 	if (err) {
4272 		ext4_msg(sb, KERN_ERR, "insufficient memory");
4273 		goto failed_mount6;
4274 	}
4275 
4276 	if (ext4_has_feature_flex_bg(sb))
4277 		if (!ext4_fill_flex_info(sb)) {
4278 			ext4_msg(sb, KERN_ERR,
4279 			       "unable to initialize "
4280 			       "flex_bg meta info!");
4281 			goto failed_mount6;
4282 		}
4283 
4284 	err = ext4_register_li_request(sb, first_not_zeroed);
4285 	if (err)
4286 		goto failed_mount6;
4287 
4288 	err = ext4_register_sysfs(sb);
4289 	if (err)
4290 		goto failed_mount7;
4291 
4292 #ifdef CONFIG_QUOTA
4293 	/* Enable quota usage during mount. */
4294 	if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4295 		err = ext4_enable_quotas(sb);
4296 		if (err)
4297 			goto failed_mount8;
4298 	}
4299 #endif  /* CONFIG_QUOTA */
4300 
4301 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4302 	ext4_orphan_cleanup(sb, es);
4303 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4304 	if (needs_recovery) {
4305 		ext4_msg(sb, KERN_INFO, "recovery complete");
4306 		ext4_mark_recovery_complete(sb, es);
4307 	}
4308 	if (EXT4_SB(sb)->s_journal) {
4309 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4310 			descr = " journalled data mode";
4311 		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4312 			descr = " ordered data mode";
4313 		else
4314 			descr = " writeback data mode";
4315 	} else
4316 		descr = "out journal";
4317 
4318 	if (test_opt(sb, DISCARD)) {
4319 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
4320 		if (!blk_queue_discard(q))
4321 			ext4_msg(sb, KERN_WARNING,
4322 				 "mounting with \"discard\" option, but "
4323 				 "the device does not support discard");
4324 	}
4325 
4326 	if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4327 		ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4328 			 "Opts: %.*s%s%s", descr,
4329 			 (int) sizeof(sbi->s_es->s_mount_opts),
4330 			 sbi->s_es->s_mount_opts,
4331 			 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4332 
4333 	if (es->s_error_count)
4334 		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4335 
4336 	/* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4337 	ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4338 	ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4339 	ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4340 
4341 	kfree(orig_data);
4342 	return 0;
4343 
4344 cantfind_ext4:
4345 	if (!silent)
4346 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4347 	goto failed_mount;
4348 
4349 #ifdef CONFIG_QUOTA
4350 failed_mount8:
4351 	ext4_unregister_sysfs(sb);
4352 #endif
4353 failed_mount7:
4354 	ext4_unregister_li_request(sb);
4355 failed_mount6:
4356 	ext4_mb_release(sb);
4357 	if (sbi->s_flex_groups)
4358 		kvfree(sbi->s_flex_groups);
4359 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
4360 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
4361 	percpu_counter_destroy(&sbi->s_dirs_counter);
4362 	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4363 failed_mount5:
4364 	ext4_ext_release(sb);
4365 	ext4_release_system_zone(sb);
4366 failed_mount4a:
4367 	dput(sb->s_root);
4368 	sb->s_root = NULL;
4369 failed_mount4:
4370 	ext4_msg(sb, KERN_ERR, "mount failed");
4371 	if (EXT4_SB(sb)->rsv_conversion_wq)
4372 		destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4373 failed_mount_wq:
4374 	if (sbi->s_ea_inode_cache) {
4375 		ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4376 		sbi->s_ea_inode_cache = NULL;
4377 	}
4378 	if (sbi->s_ea_block_cache) {
4379 		ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4380 		sbi->s_ea_block_cache = NULL;
4381 	}
4382 	if (sbi->s_journal) {
4383 		jbd2_journal_destroy(sbi->s_journal);
4384 		sbi->s_journal = NULL;
4385 	}
4386 failed_mount3a:
4387 	ext4_es_unregister_shrinker(sbi);
4388 failed_mount3:
4389 	del_timer_sync(&sbi->s_err_report);
4390 	if (sbi->s_mmp_tsk)
4391 		kthread_stop(sbi->s_mmp_tsk);
4392 failed_mount2:
4393 	for (i = 0; i < db_count; i++)
4394 		brelse(sbi->s_group_desc[i]);
4395 	kvfree(sbi->s_group_desc);
4396 failed_mount:
4397 	if (sbi->s_chksum_driver)
4398 		crypto_free_shash(sbi->s_chksum_driver);
4399 #ifdef CONFIG_QUOTA
4400 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4401 		kfree(sbi->s_qf_names[i]);
4402 #endif
4403 	ext4_blkdev_remove(sbi);
4404 	brelse(bh);
4405 out_fail:
4406 	sb->s_fs_info = NULL;
4407 	kfree(sbi->s_blockgroup_lock);
4408 out_free_base:
4409 	kfree(sbi);
4410 	kfree(orig_data);
4411 	fs_put_dax(dax_dev);
4412 	return err ? err : ret;
4413 }
4414 
4415 /*
4416  * Setup any per-fs journal parameters now.  We'll do this both on
4417  * initial mount, once the journal has been initialised but before we've
4418  * done any recovery; and again on any subsequent remount.
4419  */
4420 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4421 {
4422 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4423 
4424 	journal->j_commit_interval = sbi->s_commit_interval;
4425 	journal->j_min_batch_time = sbi->s_min_batch_time;
4426 	journal->j_max_batch_time = sbi->s_max_batch_time;
4427 
4428 	write_lock(&journal->j_state_lock);
4429 	if (test_opt(sb, BARRIER))
4430 		journal->j_flags |= JBD2_BARRIER;
4431 	else
4432 		journal->j_flags &= ~JBD2_BARRIER;
4433 	if (test_opt(sb, DATA_ERR_ABORT))
4434 		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4435 	else
4436 		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4437 	write_unlock(&journal->j_state_lock);
4438 }
4439 
4440 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4441 					     unsigned int journal_inum)
4442 {
4443 	struct inode *journal_inode;
4444 
4445 	/*
4446 	 * Test for the existence of a valid inode on disk.  Bad things
4447 	 * happen if we iget() an unused inode, as the subsequent iput()
4448 	 * will try to delete it.
4449 	 */
4450 	journal_inode = ext4_iget(sb, journal_inum);
4451 	if (IS_ERR(journal_inode)) {
4452 		ext4_msg(sb, KERN_ERR, "no journal found");
4453 		return NULL;
4454 	}
4455 	if (!journal_inode->i_nlink) {
4456 		make_bad_inode(journal_inode);
4457 		iput(journal_inode);
4458 		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4459 		return NULL;
4460 	}
4461 
4462 	jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4463 		  journal_inode, journal_inode->i_size);
4464 	if (!S_ISREG(journal_inode->i_mode)) {
4465 		ext4_msg(sb, KERN_ERR, "invalid journal inode");
4466 		iput(journal_inode);
4467 		return NULL;
4468 	}
4469 	return journal_inode;
4470 }
4471 
4472 static journal_t *ext4_get_journal(struct super_block *sb,
4473 				   unsigned int journal_inum)
4474 {
4475 	struct inode *journal_inode;
4476 	journal_t *journal;
4477 
4478 	BUG_ON(!ext4_has_feature_journal(sb));
4479 
4480 	journal_inode = ext4_get_journal_inode(sb, journal_inum);
4481 	if (!journal_inode)
4482 		return NULL;
4483 
4484 	journal = jbd2_journal_init_inode(journal_inode);
4485 	if (!journal) {
4486 		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4487 		iput(journal_inode);
4488 		return NULL;
4489 	}
4490 	journal->j_private = sb;
4491 	ext4_init_journal_params(sb, journal);
4492 	return journal;
4493 }
4494 
4495 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4496 				       dev_t j_dev)
4497 {
4498 	struct buffer_head *bh;
4499 	journal_t *journal;
4500 	ext4_fsblk_t start;
4501 	ext4_fsblk_t len;
4502 	int hblock, blocksize;
4503 	ext4_fsblk_t sb_block;
4504 	unsigned long offset;
4505 	struct ext4_super_block *es;
4506 	struct block_device *bdev;
4507 
4508 	BUG_ON(!ext4_has_feature_journal(sb));
4509 
4510 	bdev = ext4_blkdev_get(j_dev, sb);
4511 	if (bdev == NULL)
4512 		return NULL;
4513 
4514 	blocksize = sb->s_blocksize;
4515 	hblock = bdev_logical_block_size(bdev);
4516 	if (blocksize < hblock) {
4517 		ext4_msg(sb, KERN_ERR,
4518 			"blocksize too small for journal device");
4519 		goto out_bdev;
4520 	}
4521 
4522 	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4523 	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4524 	set_blocksize(bdev, blocksize);
4525 	if (!(bh = __bread(bdev, sb_block, blocksize))) {
4526 		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4527 		       "external journal");
4528 		goto out_bdev;
4529 	}
4530 
4531 	es = (struct ext4_super_block *) (bh->b_data + offset);
4532 	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4533 	    !(le32_to_cpu(es->s_feature_incompat) &
4534 	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4535 		ext4_msg(sb, KERN_ERR, "external journal has "
4536 					"bad superblock");
4537 		brelse(bh);
4538 		goto out_bdev;
4539 	}
4540 
4541 	if ((le32_to_cpu(es->s_feature_ro_compat) &
4542 	     EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4543 	    es->s_checksum != ext4_superblock_csum(sb, es)) {
4544 		ext4_msg(sb, KERN_ERR, "external journal has "
4545 				       "corrupt superblock");
4546 		brelse(bh);
4547 		goto out_bdev;
4548 	}
4549 
4550 	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4551 		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4552 		brelse(bh);
4553 		goto out_bdev;
4554 	}
4555 
4556 	len = ext4_blocks_count(es);
4557 	start = sb_block + 1;
4558 	brelse(bh);	/* we're done with the superblock */
4559 
4560 	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4561 					start, len, blocksize);
4562 	if (!journal) {
4563 		ext4_msg(sb, KERN_ERR, "failed to create device journal");
4564 		goto out_bdev;
4565 	}
4566 	journal->j_private = sb;
4567 	ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4568 	wait_on_buffer(journal->j_sb_buffer);
4569 	if (!buffer_uptodate(journal->j_sb_buffer)) {
4570 		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4571 		goto out_journal;
4572 	}
4573 	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4574 		ext4_msg(sb, KERN_ERR, "External journal has more than one "
4575 					"user (unsupported) - %d",
4576 			be32_to_cpu(journal->j_superblock->s_nr_users));
4577 		goto out_journal;
4578 	}
4579 	EXT4_SB(sb)->journal_bdev = bdev;
4580 	ext4_init_journal_params(sb, journal);
4581 	return journal;
4582 
4583 out_journal:
4584 	jbd2_journal_destroy(journal);
4585 out_bdev:
4586 	ext4_blkdev_put(bdev);
4587 	return NULL;
4588 }
4589 
4590 static int ext4_load_journal(struct super_block *sb,
4591 			     struct ext4_super_block *es,
4592 			     unsigned long journal_devnum)
4593 {
4594 	journal_t *journal;
4595 	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4596 	dev_t journal_dev;
4597 	int err = 0;
4598 	int really_read_only;
4599 
4600 	BUG_ON(!ext4_has_feature_journal(sb));
4601 
4602 	if (journal_devnum &&
4603 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4604 		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4605 			"numbers have changed");
4606 		journal_dev = new_decode_dev(journal_devnum);
4607 	} else
4608 		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4609 
4610 	really_read_only = bdev_read_only(sb->s_bdev);
4611 
4612 	/*
4613 	 * Are we loading a blank journal or performing recovery after a
4614 	 * crash?  For recovery, we need to check in advance whether we
4615 	 * can get read-write access to the device.
4616 	 */
4617 	if (ext4_has_feature_journal_needs_recovery(sb)) {
4618 		if (sb_rdonly(sb)) {
4619 			ext4_msg(sb, KERN_INFO, "INFO: recovery "
4620 					"required on readonly filesystem");
4621 			if (really_read_only) {
4622 				ext4_msg(sb, KERN_ERR, "write access "
4623 					"unavailable, cannot proceed "
4624 					"(try mounting with noload)");
4625 				return -EROFS;
4626 			}
4627 			ext4_msg(sb, KERN_INFO, "write access will "
4628 			       "be enabled during recovery");
4629 		}
4630 	}
4631 
4632 	if (journal_inum && journal_dev) {
4633 		ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4634 		       "and inode journals!");
4635 		return -EINVAL;
4636 	}
4637 
4638 	if (journal_inum) {
4639 		if (!(journal = ext4_get_journal(sb, journal_inum)))
4640 			return -EINVAL;
4641 	} else {
4642 		if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4643 			return -EINVAL;
4644 	}
4645 
4646 	if (!(journal->j_flags & JBD2_BARRIER))
4647 		ext4_msg(sb, KERN_INFO, "barriers disabled");
4648 
4649 	if (!ext4_has_feature_journal_needs_recovery(sb))
4650 		err = jbd2_journal_wipe(journal, !really_read_only);
4651 	if (!err) {
4652 		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4653 		if (save)
4654 			memcpy(save, ((char *) es) +
4655 			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4656 		err = jbd2_journal_load(journal);
4657 		if (save)
4658 			memcpy(((char *) es) + EXT4_S_ERR_START,
4659 			       save, EXT4_S_ERR_LEN);
4660 		kfree(save);
4661 	}
4662 
4663 	if (err) {
4664 		ext4_msg(sb, KERN_ERR, "error loading journal");
4665 		jbd2_journal_destroy(journal);
4666 		return err;
4667 	}
4668 
4669 	EXT4_SB(sb)->s_journal = journal;
4670 	ext4_clear_journal_err(sb, es);
4671 
4672 	if (!really_read_only && journal_devnum &&
4673 	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4674 		es->s_journal_dev = cpu_to_le32(journal_devnum);
4675 
4676 		/* Make sure we flush the recovery flag to disk. */
4677 		ext4_commit_super(sb, 1);
4678 	}
4679 
4680 	return 0;
4681 }
4682 
4683 static int ext4_commit_super(struct super_block *sb, int sync)
4684 {
4685 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4686 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4687 	int error = 0;
4688 
4689 	if (!sbh || block_device_ejected(sb))
4690 		return error;
4691 	/*
4692 	 * If the file system is mounted read-only, don't update the
4693 	 * superblock write time.  This avoids updating the superblock
4694 	 * write time when we are mounting the root file system
4695 	 * read/only but we need to replay the journal; at that point,
4696 	 * for people who are east of GMT and who make their clock
4697 	 * tick in localtime for Windows bug-for-bug compatibility,
4698 	 * the clock is set in the future, and this will cause e2fsck
4699 	 * to complain and force a full file system check.
4700 	 */
4701 	if (!(sb->s_flags & SB_RDONLY))
4702 		es->s_wtime = cpu_to_le32(get_seconds());
4703 	if (sb->s_bdev->bd_part)
4704 		es->s_kbytes_written =
4705 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4706 			    ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4707 			      EXT4_SB(sb)->s_sectors_written_start) >> 1));
4708 	else
4709 		es->s_kbytes_written =
4710 			cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4711 	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4712 		ext4_free_blocks_count_set(es,
4713 			EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4714 				&EXT4_SB(sb)->s_freeclusters_counter)));
4715 	if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4716 		es->s_free_inodes_count =
4717 			cpu_to_le32(percpu_counter_sum_positive(
4718 				&EXT4_SB(sb)->s_freeinodes_counter));
4719 	BUFFER_TRACE(sbh, "marking dirty");
4720 	ext4_superblock_csum_set(sb);
4721 	if (sync)
4722 		lock_buffer(sbh);
4723 	if (buffer_write_io_error(sbh)) {
4724 		/*
4725 		 * Oh, dear.  A previous attempt to write the
4726 		 * superblock failed.  This could happen because the
4727 		 * USB device was yanked out.  Or it could happen to
4728 		 * be a transient write error and maybe the block will
4729 		 * be remapped.  Nothing we can do but to retry the
4730 		 * write and hope for the best.
4731 		 */
4732 		ext4_msg(sb, KERN_ERR, "previous I/O error to "
4733 		       "superblock detected");
4734 		clear_buffer_write_io_error(sbh);
4735 		set_buffer_uptodate(sbh);
4736 	}
4737 	mark_buffer_dirty(sbh);
4738 	if (sync) {
4739 		unlock_buffer(sbh);
4740 		error = __sync_dirty_buffer(sbh,
4741 			REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
4742 		if (error)
4743 			return error;
4744 
4745 		error = buffer_write_io_error(sbh);
4746 		if (error) {
4747 			ext4_msg(sb, KERN_ERR, "I/O error while writing "
4748 			       "superblock");
4749 			clear_buffer_write_io_error(sbh);
4750 			set_buffer_uptodate(sbh);
4751 		}
4752 	}
4753 	return error;
4754 }
4755 
4756 /*
4757  * Have we just finished recovery?  If so, and if we are mounting (or
4758  * remounting) the filesystem readonly, then we will end up with a
4759  * consistent fs on disk.  Record that fact.
4760  */
4761 static void ext4_mark_recovery_complete(struct super_block *sb,
4762 					struct ext4_super_block *es)
4763 {
4764 	journal_t *journal = EXT4_SB(sb)->s_journal;
4765 
4766 	if (!ext4_has_feature_journal(sb)) {
4767 		BUG_ON(journal != NULL);
4768 		return;
4769 	}
4770 	jbd2_journal_lock_updates(journal);
4771 	if (jbd2_journal_flush(journal) < 0)
4772 		goto out;
4773 
4774 	if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
4775 		ext4_clear_feature_journal_needs_recovery(sb);
4776 		ext4_commit_super(sb, 1);
4777 	}
4778 
4779 out:
4780 	jbd2_journal_unlock_updates(journal);
4781 }
4782 
4783 /*
4784  * If we are mounting (or read-write remounting) a filesystem whose journal
4785  * has recorded an error from a previous lifetime, move that error to the
4786  * main filesystem now.
4787  */
4788 static void ext4_clear_journal_err(struct super_block *sb,
4789 				   struct ext4_super_block *es)
4790 {
4791 	journal_t *journal;
4792 	int j_errno;
4793 	const char *errstr;
4794 
4795 	BUG_ON(!ext4_has_feature_journal(sb));
4796 
4797 	journal = EXT4_SB(sb)->s_journal;
4798 
4799 	/*
4800 	 * Now check for any error status which may have been recorded in the
4801 	 * journal by a prior ext4_error() or ext4_abort()
4802 	 */
4803 
4804 	j_errno = jbd2_journal_errno(journal);
4805 	if (j_errno) {
4806 		char nbuf[16];
4807 
4808 		errstr = ext4_decode_error(sb, j_errno, nbuf);
4809 		ext4_warning(sb, "Filesystem error recorded "
4810 			     "from previous mount: %s", errstr);
4811 		ext4_warning(sb, "Marking fs in need of filesystem check.");
4812 
4813 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4814 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4815 		ext4_commit_super(sb, 1);
4816 
4817 		jbd2_journal_clear_err(journal);
4818 		jbd2_journal_update_sb_errno(journal);
4819 	}
4820 }
4821 
4822 /*
4823  * Force the running and committing transactions to commit,
4824  * and wait on the commit.
4825  */
4826 int ext4_force_commit(struct super_block *sb)
4827 {
4828 	journal_t *journal;
4829 
4830 	if (sb_rdonly(sb))
4831 		return 0;
4832 
4833 	journal = EXT4_SB(sb)->s_journal;
4834 	return ext4_journal_force_commit(journal);
4835 }
4836 
4837 static int ext4_sync_fs(struct super_block *sb, int wait)
4838 {
4839 	int ret = 0;
4840 	tid_t target;
4841 	bool needs_barrier = false;
4842 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4843 
4844 	if (unlikely(ext4_forced_shutdown(sbi)))
4845 		return 0;
4846 
4847 	trace_ext4_sync_fs(sb, wait);
4848 	flush_workqueue(sbi->rsv_conversion_wq);
4849 	/*
4850 	 * Writeback quota in non-journalled quota case - journalled quota has
4851 	 * no dirty dquots
4852 	 */
4853 	dquot_writeback_dquots(sb, -1);
4854 	/*
4855 	 * Data writeback is possible w/o journal transaction, so barrier must
4856 	 * being sent at the end of the function. But we can skip it if
4857 	 * transaction_commit will do it for us.
4858 	 */
4859 	if (sbi->s_journal) {
4860 		target = jbd2_get_latest_transaction(sbi->s_journal);
4861 		if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4862 		    !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4863 			needs_barrier = true;
4864 
4865 		if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4866 			if (wait)
4867 				ret = jbd2_log_wait_commit(sbi->s_journal,
4868 							   target);
4869 		}
4870 	} else if (wait && test_opt(sb, BARRIER))
4871 		needs_barrier = true;
4872 	if (needs_barrier) {
4873 		int err;
4874 		err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4875 		if (!ret)
4876 			ret = err;
4877 	}
4878 
4879 	return ret;
4880 }
4881 
4882 /*
4883  * LVM calls this function before a (read-only) snapshot is created.  This
4884  * gives us a chance to flush the journal completely and mark the fs clean.
4885  *
4886  * Note that only this function cannot bring a filesystem to be in a clean
4887  * state independently. It relies on upper layer to stop all data & metadata
4888  * modifications.
4889  */
4890 static int ext4_freeze(struct super_block *sb)
4891 {
4892 	int error = 0;
4893 	journal_t *journal;
4894 
4895 	if (sb_rdonly(sb))
4896 		return 0;
4897 
4898 	journal = EXT4_SB(sb)->s_journal;
4899 
4900 	if (journal) {
4901 		/* Now we set up the journal barrier. */
4902 		jbd2_journal_lock_updates(journal);
4903 
4904 		/*
4905 		 * Don't clear the needs_recovery flag if we failed to
4906 		 * flush the journal.
4907 		 */
4908 		error = jbd2_journal_flush(journal);
4909 		if (error < 0)
4910 			goto out;
4911 
4912 		/* Journal blocked and flushed, clear needs_recovery flag. */
4913 		ext4_clear_feature_journal_needs_recovery(sb);
4914 	}
4915 
4916 	error = ext4_commit_super(sb, 1);
4917 out:
4918 	if (journal)
4919 		/* we rely on upper layer to stop further updates */
4920 		jbd2_journal_unlock_updates(journal);
4921 	return error;
4922 }
4923 
4924 /*
4925  * Called by LVM after the snapshot is done.  We need to reset the RECOVER
4926  * flag here, even though the filesystem is not technically dirty yet.
4927  */
4928 static int ext4_unfreeze(struct super_block *sb)
4929 {
4930 	if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
4931 		return 0;
4932 
4933 	if (EXT4_SB(sb)->s_journal) {
4934 		/* Reset the needs_recovery flag before the fs is unlocked. */
4935 		ext4_set_feature_journal_needs_recovery(sb);
4936 	}
4937 
4938 	ext4_commit_super(sb, 1);
4939 	return 0;
4940 }
4941 
4942 /*
4943  * Structure to save mount options for ext4_remount's benefit
4944  */
4945 struct ext4_mount_options {
4946 	unsigned long s_mount_opt;
4947 	unsigned long s_mount_opt2;
4948 	kuid_t s_resuid;
4949 	kgid_t s_resgid;
4950 	unsigned long s_commit_interval;
4951 	u32 s_min_batch_time, s_max_batch_time;
4952 #ifdef CONFIG_QUOTA
4953 	int s_jquota_fmt;
4954 	char *s_qf_names[EXT4_MAXQUOTAS];
4955 #endif
4956 };
4957 
4958 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4959 {
4960 	struct ext4_super_block *es;
4961 	struct ext4_sb_info *sbi = EXT4_SB(sb);
4962 	unsigned long old_sb_flags;
4963 	struct ext4_mount_options old_opts;
4964 	int enable_quota = 0;
4965 	ext4_group_t g;
4966 	unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4967 	int err = 0;
4968 #ifdef CONFIG_QUOTA
4969 	int i, j;
4970 #endif
4971 	char *orig_data = kstrdup(data, GFP_KERNEL);
4972 
4973 	/* Store the original options */
4974 	old_sb_flags = sb->s_flags;
4975 	old_opts.s_mount_opt = sbi->s_mount_opt;
4976 	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4977 	old_opts.s_resuid = sbi->s_resuid;
4978 	old_opts.s_resgid = sbi->s_resgid;
4979 	old_opts.s_commit_interval = sbi->s_commit_interval;
4980 	old_opts.s_min_batch_time = sbi->s_min_batch_time;
4981 	old_opts.s_max_batch_time = sbi->s_max_batch_time;
4982 #ifdef CONFIG_QUOTA
4983 	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4984 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
4985 		if (sbi->s_qf_names[i]) {
4986 			old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4987 							 GFP_KERNEL);
4988 			if (!old_opts.s_qf_names[i]) {
4989 				for (j = 0; j < i; j++)
4990 					kfree(old_opts.s_qf_names[j]);
4991 				kfree(orig_data);
4992 				return -ENOMEM;
4993 			}
4994 		} else
4995 			old_opts.s_qf_names[i] = NULL;
4996 #endif
4997 	if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4998 		journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4999 
5000 	if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5001 		err = -EINVAL;
5002 		goto restore_opts;
5003 	}
5004 
5005 	if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5006 	    test_opt(sb, JOURNAL_CHECKSUM)) {
5007 		ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5008 			 "during remount not supported; ignoring");
5009 		sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5010 	}
5011 
5012 	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5013 		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5014 			ext4_msg(sb, KERN_ERR, "can't mount with "
5015 				 "both data=journal and delalloc");
5016 			err = -EINVAL;
5017 			goto restore_opts;
5018 		}
5019 		if (test_opt(sb, DIOREAD_NOLOCK)) {
5020 			ext4_msg(sb, KERN_ERR, "can't mount with "
5021 				 "both data=journal and dioread_nolock");
5022 			err = -EINVAL;
5023 			goto restore_opts;
5024 		}
5025 		if (test_opt(sb, DAX)) {
5026 			ext4_msg(sb, KERN_ERR, "can't mount with "
5027 				 "both data=journal and dax");
5028 			err = -EINVAL;
5029 			goto restore_opts;
5030 		}
5031 	} else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5032 		if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5033 			ext4_msg(sb, KERN_ERR, "can't mount with "
5034 				"journal_async_commit in data=ordered mode");
5035 			err = -EINVAL;
5036 			goto restore_opts;
5037 		}
5038 	}
5039 
5040 	if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5041 		ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5042 		err = -EINVAL;
5043 		goto restore_opts;
5044 	}
5045 
5046 	if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5047 		ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5048 			"dax flag with busy inodes while remounting");
5049 		sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5050 	}
5051 
5052 	if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5053 		ext4_abort(sb, "Abort forced by user");
5054 
5055 	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5056 		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5057 
5058 	es = sbi->s_es;
5059 
5060 	if (sbi->s_journal) {
5061 		ext4_init_journal_params(sb, sbi->s_journal);
5062 		set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5063 	}
5064 
5065 	if (*flags & SB_LAZYTIME)
5066 		sb->s_flags |= SB_LAZYTIME;
5067 
5068 	if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5069 		if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5070 			err = -EROFS;
5071 			goto restore_opts;
5072 		}
5073 
5074 		if (*flags & SB_RDONLY) {
5075 			err = sync_filesystem(sb);
5076 			if (err < 0)
5077 				goto restore_opts;
5078 			err = dquot_suspend(sb, -1);
5079 			if (err < 0)
5080 				goto restore_opts;
5081 
5082 			/*
5083 			 * First of all, the unconditional stuff we have to do
5084 			 * to disable replay of the journal when we next remount
5085 			 */
5086 			sb->s_flags |= SB_RDONLY;
5087 
5088 			/*
5089 			 * OK, test if we are remounting a valid rw partition
5090 			 * readonly, and if so set the rdonly flag and then
5091 			 * mark the partition as valid again.
5092 			 */
5093 			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5094 			    (sbi->s_mount_state & EXT4_VALID_FS))
5095 				es->s_state = cpu_to_le16(sbi->s_mount_state);
5096 
5097 			if (sbi->s_journal)
5098 				ext4_mark_recovery_complete(sb, es);
5099 		} else {
5100 			/* Make sure we can mount this feature set readwrite */
5101 			if (ext4_has_feature_readonly(sb) ||
5102 			    !ext4_feature_set_ok(sb, 0)) {
5103 				err = -EROFS;
5104 				goto restore_opts;
5105 			}
5106 			/*
5107 			 * Make sure the group descriptor checksums
5108 			 * are sane.  If they aren't, refuse to remount r/w.
5109 			 */
5110 			for (g = 0; g < sbi->s_groups_count; g++) {
5111 				struct ext4_group_desc *gdp =
5112 					ext4_get_group_desc(sb, g, NULL);
5113 
5114 				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5115 					ext4_msg(sb, KERN_ERR,
5116 	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
5117 		g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5118 					       le16_to_cpu(gdp->bg_checksum));
5119 					err = -EFSBADCRC;
5120 					goto restore_opts;
5121 				}
5122 			}
5123 
5124 			/*
5125 			 * If we have an unprocessed orphan list hanging
5126 			 * around from a previously readonly bdev mount,
5127 			 * require a full umount/remount for now.
5128 			 */
5129 			if (es->s_last_orphan) {
5130 				ext4_msg(sb, KERN_WARNING, "Couldn't "
5131 				       "remount RDWR because of unprocessed "
5132 				       "orphan inode list.  Please "
5133 				       "umount/remount instead");
5134 				err = -EINVAL;
5135 				goto restore_opts;
5136 			}
5137 
5138 			/*
5139 			 * Mounting a RDONLY partition read-write, so reread
5140 			 * and store the current valid flag.  (It may have
5141 			 * been changed by e2fsck since we originally mounted
5142 			 * the partition.)
5143 			 */
5144 			if (sbi->s_journal)
5145 				ext4_clear_journal_err(sb, es);
5146 			sbi->s_mount_state = le16_to_cpu(es->s_state);
5147 			if (!ext4_setup_super(sb, es, 0))
5148 				sb->s_flags &= ~SB_RDONLY;
5149 			if (ext4_has_feature_mmp(sb))
5150 				if (ext4_multi_mount_protect(sb,
5151 						le64_to_cpu(es->s_mmp_block))) {
5152 					err = -EROFS;
5153 					goto restore_opts;
5154 				}
5155 			enable_quota = 1;
5156 		}
5157 	}
5158 
5159 	/*
5160 	 * Reinitialize lazy itable initialization thread based on
5161 	 * current settings
5162 	 */
5163 	if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5164 		ext4_unregister_li_request(sb);
5165 	else {
5166 		ext4_group_t first_not_zeroed;
5167 		first_not_zeroed = ext4_has_uninit_itable(sb);
5168 		ext4_register_li_request(sb, first_not_zeroed);
5169 	}
5170 
5171 	ext4_setup_system_zone(sb);
5172 	if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY))
5173 		ext4_commit_super(sb, 1);
5174 
5175 #ifdef CONFIG_QUOTA
5176 	/* Release old quota file names */
5177 	for (i = 0; i < EXT4_MAXQUOTAS; i++)
5178 		kfree(old_opts.s_qf_names[i]);
5179 	if (enable_quota) {
5180 		if (sb_any_quota_suspended(sb))
5181 			dquot_resume(sb, -1);
5182 		else if (ext4_has_feature_quota(sb)) {
5183 			err = ext4_enable_quotas(sb);
5184 			if (err)
5185 				goto restore_opts;
5186 		}
5187 	}
5188 #endif
5189 
5190 	*flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
5191 	ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5192 	kfree(orig_data);
5193 	return 0;
5194 
5195 restore_opts:
5196 	sb->s_flags = old_sb_flags;
5197 	sbi->s_mount_opt = old_opts.s_mount_opt;
5198 	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5199 	sbi->s_resuid = old_opts.s_resuid;
5200 	sbi->s_resgid = old_opts.s_resgid;
5201 	sbi->s_commit_interval = old_opts.s_commit_interval;
5202 	sbi->s_min_batch_time = old_opts.s_min_batch_time;
5203 	sbi->s_max_batch_time = old_opts.s_max_batch_time;
5204 #ifdef CONFIG_QUOTA
5205 	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5206 	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5207 		kfree(sbi->s_qf_names[i]);
5208 		sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5209 	}
5210 #endif
5211 	kfree(orig_data);
5212 	return err;
5213 }
5214 
5215 #ifdef CONFIG_QUOTA
5216 static int ext4_statfs_project(struct super_block *sb,
5217 			       kprojid_t projid, struct kstatfs *buf)
5218 {
5219 	struct kqid qid;
5220 	struct dquot *dquot;
5221 	u64 limit;
5222 	u64 curblock;
5223 
5224 	qid = make_kqid_projid(projid);
5225 	dquot = dqget(sb, qid);
5226 	if (IS_ERR(dquot))
5227 		return PTR_ERR(dquot);
5228 	spin_lock(&dquot->dq_dqb_lock);
5229 
5230 	limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5231 		 dquot->dq_dqb.dqb_bsoftlimit :
5232 		 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5233 	if (limit && buf->f_blocks > limit) {
5234 		curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5235 		buf->f_blocks = limit;
5236 		buf->f_bfree = buf->f_bavail =
5237 			(buf->f_blocks > curblock) ?
5238 			 (buf->f_blocks - curblock) : 0;
5239 	}
5240 
5241 	limit = dquot->dq_dqb.dqb_isoftlimit ?
5242 		dquot->dq_dqb.dqb_isoftlimit :
5243 		dquot->dq_dqb.dqb_ihardlimit;
5244 	if (limit && buf->f_files > limit) {
5245 		buf->f_files = limit;
5246 		buf->f_ffree =
5247 			(buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5248 			 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5249 	}
5250 
5251 	spin_unlock(&dquot->dq_dqb_lock);
5252 	dqput(dquot);
5253 	return 0;
5254 }
5255 #endif
5256 
5257 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5258 {
5259 	struct super_block *sb = dentry->d_sb;
5260 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5261 	struct ext4_super_block *es = sbi->s_es;
5262 	ext4_fsblk_t overhead = 0, resv_blocks;
5263 	u64 fsid;
5264 	s64 bfree;
5265 	resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5266 
5267 	if (!test_opt(sb, MINIX_DF))
5268 		overhead = sbi->s_overhead;
5269 
5270 	buf->f_type = EXT4_SUPER_MAGIC;
5271 	buf->f_bsize = sb->s_blocksize;
5272 	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5273 	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5274 		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5275 	/* prevent underflow in case that few free space is available */
5276 	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5277 	buf->f_bavail = buf->f_bfree -
5278 			(ext4_r_blocks_count(es) + resv_blocks);
5279 	if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5280 		buf->f_bavail = 0;
5281 	buf->f_files = le32_to_cpu(es->s_inodes_count);
5282 	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5283 	buf->f_namelen = EXT4_NAME_LEN;
5284 	fsid = le64_to_cpup((void *)es->s_uuid) ^
5285 	       le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5286 	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5287 	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5288 
5289 #ifdef CONFIG_QUOTA
5290 	if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5291 	    sb_has_quota_limits_enabled(sb, PRJQUOTA))
5292 		ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5293 #endif
5294 	return 0;
5295 }
5296 
5297 
5298 #ifdef CONFIG_QUOTA
5299 
5300 /*
5301  * Helper functions so that transaction is started before we acquire dqio_sem
5302  * to keep correct lock ordering of transaction > dqio_sem
5303  */
5304 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5305 {
5306 	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5307 }
5308 
5309 static int ext4_write_dquot(struct dquot *dquot)
5310 {
5311 	int ret, err;
5312 	handle_t *handle;
5313 	struct inode *inode;
5314 
5315 	inode = dquot_to_inode(dquot);
5316 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5317 				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5318 	if (IS_ERR(handle))
5319 		return PTR_ERR(handle);
5320 	ret = dquot_commit(dquot);
5321 	err = ext4_journal_stop(handle);
5322 	if (!ret)
5323 		ret = err;
5324 	return ret;
5325 }
5326 
5327 static int ext4_acquire_dquot(struct dquot *dquot)
5328 {
5329 	int ret, err;
5330 	handle_t *handle;
5331 
5332 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5333 				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5334 	if (IS_ERR(handle))
5335 		return PTR_ERR(handle);
5336 	ret = dquot_acquire(dquot);
5337 	err = ext4_journal_stop(handle);
5338 	if (!ret)
5339 		ret = err;
5340 	return ret;
5341 }
5342 
5343 static int ext4_release_dquot(struct dquot *dquot)
5344 {
5345 	int ret, err;
5346 	handle_t *handle;
5347 
5348 	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5349 				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5350 	if (IS_ERR(handle)) {
5351 		/* Release dquot anyway to avoid endless cycle in dqput() */
5352 		dquot_release(dquot);
5353 		return PTR_ERR(handle);
5354 	}
5355 	ret = dquot_release(dquot);
5356 	err = ext4_journal_stop(handle);
5357 	if (!ret)
5358 		ret = err;
5359 	return ret;
5360 }
5361 
5362 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5363 {
5364 	struct super_block *sb = dquot->dq_sb;
5365 	struct ext4_sb_info *sbi = EXT4_SB(sb);
5366 
5367 	/* Are we journaling quotas? */
5368 	if (ext4_has_feature_quota(sb) ||
5369 	    sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5370 		dquot_mark_dquot_dirty(dquot);
5371 		return ext4_write_dquot(dquot);
5372 	} else {
5373 		return dquot_mark_dquot_dirty(dquot);
5374 	}
5375 }
5376 
5377 static int ext4_write_info(struct super_block *sb, int type)
5378 {
5379 	int ret, err;
5380 	handle_t *handle;
5381 
5382 	/* Data block + inode block */
5383 	handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5384 	if (IS_ERR(handle))
5385 		return PTR_ERR(handle);
5386 	ret = dquot_commit_info(sb, type);
5387 	err = ext4_journal_stop(handle);
5388 	if (!ret)
5389 		ret = err;
5390 	return ret;
5391 }
5392 
5393 /*
5394  * Turn on quotas during mount time - we need to find
5395  * the quota file and such...
5396  */
5397 static int ext4_quota_on_mount(struct super_block *sb, int type)
5398 {
5399 	return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5400 					EXT4_SB(sb)->s_jquota_fmt, type);
5401 }
5402 
5403 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5404 {
5405 	struct ext4_inode_info *ei = EXT4_I(inode);
5406 
5407 	/* The first argument of lockdep_set_subclass has to be
5408 	 * *exactly* the same as the argument to init_rwsem() --- in
5409 	 * this case, in init_once() --- or lockdep gets unhappy
5410 	 * because the name of the lock is set using the
5411 	 * stringification of the argument to init_rwsem().
5412 	 */
5413 	(void) ei;	/* shut up clang warning if !CONFIG_LOCKDEP */
5414 	lockdep_set_subclass(&ei->i_data_sem, subclass);
5415 }
5416 
5417 /*
5418  * Standard function to be called on quota_on
5419  */
5420 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5421 			 const struct path *path)
5422 {
5423 	int err;
5424 
5425 	if (!test_opt(sb, QUOTA))
5426 		return -EINVAL;
5427 
5428 	/* Quotafile not on the same filesystem? */
5429 	if (path->dentry->d_sb != sb)
5430 		return -EXDEV;
5431 	/* Journaling quota? */
5432 	if (EXT4_SB(sb)->s_qf_names[type]) {
5433 		/* Quotafile not in fs root? */
5434 		if (path->dentry->d_parent != sb->s_root)
5435 			ext4_msg(sb, KERN_WARNING,
5436 				"Quota file not on filesystem root. "
5437 				"Journaled quota will not work");
5438 		sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5439 	} else {
5440 		/*
5441 		 * Clear the flag just in case mount options changed since
5442 		 * last time.
5443 		 */
5444 		sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5445 	}
5446 
5447 	/*
5448 	 * When we journal data on quota file, we have to flush journal to see
5449 	 * all updates to the file when we bypass pagecache...
5450 	 */
5451 	if (EXT4_SB(sb)->s_journal &&
5452 	    ext4_should_journal_data(d_inode(path->dentry))) {
5453 		/*
5454 		 * We don't need to lock updates but journal_flush() could
5455 		 * otherwise be livelocked...
5456 		 */
5457 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5458 		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5459 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5460 		if (err)
5461 			return err;
5462 	}
5463 
5464 	lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5465 	err = dquot_quota_on(sb, type, format_id, path);
5466 	if (err) {
5467 		lockdep_set_quota_inode(path->dentry->d_inode,
5468 					     I_DATA_SEM_NORMAL);
5469 	} else {
5470 		struct inode *inode = d_inode(path->dentry);
5471 		handle_t *handle;
5472 
5473 		/*
5474 		 * Set inode flags to prevent userspace from messing with quota
5475 		 * files. If this fails, we return success anyway since quotas
5476 		 * are already enabled and this is not a hard failure.
5477 		 */
5478 		inode_lock(inode);
5479 		handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5480 		if (IS_ERR(handle))
5481 			goto unlock_inode;
5482 		EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5483 		inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5484 				S_NOATIME | S_IMMUTABLE);
5485 		ext4_mark_inode_dirty(handle, inode);
5486 		ext4_journal_stop(handle);
5487 	unlock_inode:
5488 		inode_unlock(inode);
5489 	}
5490 	return err;
5491 }
5492 
5493 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5494 			     unsigned int flags)
5495 {
5496 	int err;
5497 	struct inode *qf_inode;
5498 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5499 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5500 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5501 		le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5502 	};
5503 
5504 	BUG_ON(!ext4_has_feature_quota(sb));
5505 
5506 	if (!qf_inums[type])
5507 		return -EPERM;
5508 
5509 	qf_inode = ext4_iget(sb, qf_inums[type]);
5510 	if (IS_ERR(qf_inode)) {
5511 		ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5512 		return PTR_ERR(qf_inode);
5513 	}
5514 
5515 	/* Don't account quota for quota files to avoid recursion */
5516 	qf_inode->i_flags |= S_NOQUOTA;
5517 	lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5518 	err = dquot_enable(qf_inode, type, format_id, flags);
5519 	iput(qf_inode);
5520 	if (err)
5521 		lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5522 
5523 	return err;
5524 }
5525 
5526 /* Enable usage tracking for all quota types. */
5527 static int ext4_enable_quotas(struct super_block *sb)
5528 {
5529 	int type, err = 0;
5530 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5531 		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5532 		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5533 		le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5534 	};
5535 	bool quota_mopt[EXT4_MAXQUOTAS] = {
5536 		test_opt(sb, USRQUOTA),
5537 		test_opt(sb, GRPQUOTA),
5538 		test_opt(sb, PRJQUOTA),
5539 	};
5540 
5541 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5542 	for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5543 		if (qf_inums[type]) {
5544 			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5545 				DQUOT_USAGE_ENABLED |
5546 				(quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5547 			if (err) {
5548 				for (type--; type >= 0; type--)
5549 					dquot_quota_off(sb, type);
5550 
5551 				ext4_warning(sb,
5552 					"Failed to enable quota tracking "
5553 					"(type=%d, err=%d). Please run "
5554 					"e2fsck to fix.", type, err);
5555 				return err;
5556 			}
5557 		}
5558 	}
5559 	return 0;
5560 }
5561 
5562 static int ext4_quota_off(struct super_block *sb, int type)
5563 {
5564 	struct inode *inode = sb_dqopt(sb)->files[type];
5565 	handle_t *handle;
5566 	int err;
5567 
5568 	/* Force all delayed allocation blocks to be allocated.
5569 	 * Caller already holds s_umount sem */
5570 	if (test_opt(sb, DELALLOC))
5571 		sync_filesystem(sb);
5572 
5573 	if (!inode || !igrab(inode))
5574 		goto out;
5575 
5576 	err = dquot_quota_off(sb, type);
5577 	if (err || ext4_has_feature_quota(sb))
5578 		goto out_put;
5579 
5580 	inode_lock(inode);
5581 	/*
5582 	 * Update modification times of quota files when userspace can
5583 	 * start looking at them. If we fail, we return success anyway since
5584 	 * this is not a hard failure and quotas are already disabled.
5585 	 */
5586 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5587 	if (IS_ERR(handle))
5588 		goto out_unlock;
5589 	EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5590 	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5591 	inode->i_mtime = inode->i_ctime = current_time(inode);
5592 	ext4_mark_inode_dirty(handle, inode);
5593 	ext4_journal_stop(handle);
5594 out_unlock:
5595 	inode_unlock(inode);
5596 out_put:
5597 	lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5598 	iput(inode);
5599 	return err;
5600 out:
5601 	return dquot_quota_off(sb, type);
5602 }
5603 
5604 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5605  * acquiring the locks... As quota files are never truncated and quota code
5606  * itself serializes the operations (and no one else should touch the files)
5607  * we don't have to be afraid of races */
5608 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5609 			       size_t len, loff_t off)
5610 {
5611 	struct inode *inode = sb_dqopt(sb)->files[type];
5612 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5613 	int offset = off & (sb->s_blocksize - 1);
5614 	int tocopy;
5615 	size_t toread;
5616 	struct buffer_head *bh;
5617 	loff_t i_size = i_size_read(inode);
5618 
5619 	if (off > i_size)
5620 		return 0;
5621 	if (off+len > i_size)
5622 		len = i_size-off;
5623 	toread = len;
5624 	while (toread > 0) {
5625 		tocopy = sb->s_blocksize - offset < toread ?
5626 				sb->s_blocksize - offset : toread;
5627 		bh = ext4_bread(NULL, inode, blk, 0);
5628 		if (IS_ERR(bh))
5629 			return PTR_ERR(bh);
5630 		if (!bh)	/* A hole? */
5631 			memset(data, 0, tocopy);
5632 		else
5633 			memcpy(data, bh->b_data+offset, tocopy);
5634 		brelse(bh);
5635 		offset = 0;
5636 		toread -= tocopy;
5637 		data += tocopy;
5638 		blk++;
5639 	}
5640 	return len;
5641 }
5642 
5643 /* Write to quotafile (we know the transaction is already started and has
5644  * enough credits) */
5645 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5646 				const char *data, size_t len, loff_t off)
5647 {
5648 	struct inode *inode = sb_dqopt(sb)->files[type];
5649 	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5650 	int err, offset = off & (sb->s_blocksize - 1);
5651 	int retries = 0;
5652 	struct buffer_head *bh;
5653 	handle_t *handle = journal_current_handle();
5654 
5655 	if (EXT4_SB(sb)->s_journal && !handle) {
5656 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5657 			" cancelled because transaction is not started",
5658 			(unsigned long long)off, (unsigned long long)len);
5659 		return -EIO;
5660 	}
5661 	/*
5662 	 * Since we account only one data block in transaction credits,
5663 	 * then it is impossible to cross a block boundary.
5664 	 */
5665 	if (sb->s_blocksize - offset < len) {
5666 		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5667 			" cancelled because not block aligned",
5668 			(unsigned long long)off, (unsigned long long)len);
5669 		return -EIO;
5670 	}
5671 
5672 	do {
5673 		bh = ext4_bread(handle, inode, blk,
5674 				EXT4_GET_BLOCKS_CREATE |
5675 				EXT4_GET_BLOCKS_METADATA_NOFAIL);
5676 	} while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5677 		 ext4_should_retry_alloc(inode->i_sb, &retries));
5678 	if (IS_ERR(bh))
5679 		return PTR_ERR(bh);
5680 	if (!bh)
5681 		goto out;
5682 	BUFFER_TRACE(bh, "get write access");
5683 	err = ext4_journal_get_write_access(handle, bh);
5684 	if (err) {
5685 		brelse(bh);
5686 		return err;
5687 	}
5688 	lock_buffer(bh);
5689 	memcpy(bh->b_data+offset, data, len);
5690 	flush_dcache_page(bh->b_page);
5691 	unlock_buffer(bh);
5692 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
5693 	brelse(bh);
5694 out:
5695 	if (inode->i_size < off + len) {
5696 		i_size_write(inode, off + len);
5697 		EXT4_I(inode)->i_disksize = inode->i_size;
5698 		ext4_mark_inode_dirty(handle, inode);
5699 	}
5700 	return len;
5701 }
5702 
5703 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5704 {
5705 	const struct quota_format_ops	*ops;
5706 
5707 	if (!sb_has_quota_loaded(sb, qid->type))
5708 		return -ESRCH;
5709 	ops = sb_dqopt(sb)->ops[qid->type];
5710 	if (!ops || !ops->get_next_id)
5711 		return -ENOSYS;
5712 	return dquot_get_next_id(sb, qid);
5713 }
5714 #endif
5715 
5716 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5717 		       const char *dev_name, void *data)
5718 {
5719 	return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5720 }
5721 
5722 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5723 static inline void register_as_ext2(void)
5724 {
5725 	int err = register_filesystem(&ext2_fs_type);
5726 	if (err)
5727 		printk(KERN_WARNING
5728 		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5729 }
5730 
5731 static inline void unregister_as_ext2(void)
5732 {
5733 	unregister_filesystem(&ext2_fs_type);
5734 }
5735 
5736 static inline int ext2_feature_set_ok(struct super_block *sb)
5737 {
5738 	if (ext4_has_unknown_ext2_incompat_features(sb))
5739 		return 0;
5740 	if (sb_rdonly(sb))
5741 		return 1;
5742 	if (ext4_has_unknown_ext2_ro_compat_features(sb))
5743 		return 0;
5744 	return 1;
5745 }
5746 #else
5747 static inline void register_as_ext2(void) { }
5748 static inline void unregister_as_ext2(void) { }
5749 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5750 #endif
5751 
5752 static inline void register_as_ext3(void)
5753 {
5754 	int err = register_filesystem(&ext3_fs_type);
5755 	if (err)
5756 		printk(KERN_WARNING
5757 		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5758 }
5759 
5760 static inline void unregister_as_ext3(void)
5761 {
5762 	unregister_filesystem(&ext3_fs_type);
5763 }
5764 
5765 static inline int ext3_feature_set_ok(struct super_block *sb)
5766 {
5767 	if (ext4_has_unknown_ext3_incompat_features(sb))
5768 		return 0;
5769 	if (!ext4_has_feature_journal(sb))
5770 		return 0;
5771 	if (sb_rdonly(sb))
5772 		return 1;
5773 	if (ext4_has_unknown_ext3_ro_compat_features(sb))
5774 		return 0;
5775 	return 1;
5776 }
5777 
5778 static struct file_system_type ext4_fs_type = {
5779 	.owner		= THIS_MODULE,
5780 	.name		= "ext4",
5781 	.mount		= ext4_mount,
5782 	.kill_sb	= kill_block_super,
5783 	.fs_flags	= FS_REQUIRES_DEV,
5784 };
5785 MODULE_ALIAS_FS("ext4");
5786 
5787 /* Shared across all ext4 file systems */
5788 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5789 
5790 static int __init ext4_init_fs(void)
5791 {
5792 	int i, err;
5793 
5794 	ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5795 	ext4_li_info = NULL;
5796 	mutex_init(&ext4_li_mtx);
5797 
5798 	/* Build-time check for flags consistency */
5799 	ext4_check_flag_values();
5800 
5801 	for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5802 		init_waitqueue_head(&ext4__ioend_wq[i]);
5803 
5804 	err = ext4_init_es();
5805 	if (err)
5806 		return err;
5807 
5808 	err = ext4_init_pageio();
5809 	if (err)
5810 		goto out5;
5811 
5812 	err = ext4_init_system_zone();
5813 	if (err)
5814 		goto out4;
5815 
5816 	err = ext4_init_sysfs();
5817 	if (err)
5818 		goto out3;
5819 
5820 	err = ext4_init_mballoc();
5821 	if (err)
5822 		goto out2;
5823 	err = init_inodecache();
5824 	if (err)
5825 		goto out1;
5826 	register_as_ext3();
5827 	register_as_ext2();
5828 	err = register_filesystem(&ext4_fs_type);
5829 	if (err)
5830 		goto out;
5831 
5832 	return 0;
5833 out:
5834 	unregister_as_ext2();
5835 	unregister_as_ext3();
5836 	destroy_inodecache();
5837 out1:
5838 	ext4_exit_mballoc();
5839 out2:
5840 	ext4_exit_sysfs();
5841 out3:
5842 	ext4_exit_system_zone();
5843 out4:
5844 	ext4_exit_pageio();
5845 out5:
5846 	ext4_exit_es();
5847 
5848 	return err;
5849 }
5850 
5851 static void __exit ext4_exit_fs(void)
5852 {
5853 	ext4_destroy_lazyinit_thread();
5854 	unregister_as_ext2();
5855 	unregister_as_ext3();
5856 	unregister_filesystem(&ext4_fs_type);
5857 	destroy_inodecache();
5858 	ext4_exit_mballoc();
5859 	ext4_exit_sysfs();
5860 	ext4_exit_system_zone();
5861 	ext4_exit_pageio();
5862 	ext4_exit_es();
5863 }
5864 
5865 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5866 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5867 MODULE_LICENSE("GPL");
5868 module_init(ext4_init_fs)
5869 module_exit(ext4_exit_fs)
5870